diff options
Diffstat (limited to 'src/gallium')
343 files changed, 31057 insertions, 7297 deletions
diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 98487d43bd6..63983c52201 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -31,8 +31,8 @@ INCLUDES = \ default: depend lib$(LIBNAME).a -lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template - $(MKLIB) -o $(LIBNAME) -static $(OBJECTS) +lib$(LIBNAME).a: $(OBJECTS) $(EXTRA_OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template + $(MKLIB) -o $(LIBNAME) -static $(OBJECTS) $(EXTRA_OBJECTS) depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) rm -f depend @@ -41,7 +41,7 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) # Emacs tags tags: - etags `find . -name \*.[ch]` `find ../include` + etags `find . -name \*.[ch]` `find $(TOP)/src/gallium/include -name \*.h` # Remove .o and backup files clean: diff --git a/src/gallium/SConscript b/src/gallium/SConscript index 32b7c2980ab..89c69d7205e 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -31,6 +31,7 @@ for driver in env['drivers']: SConscript('state_trackers/python/SConscript') SConscript('state_trackers/glx/xlib/SConscript') SConscript('state_trackers/dri/SConscript') +SConscript('state_trackers/xorg/SConscript') if platform == 'windows': SConscript('state_trackers/wgl/SConscript') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 0bc77a57287..e6dce3f0e5b 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -361,6 +361,10 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, void cso_cache_delete(struct cso_cache *sc) { assert(sc); + + if (!sc) + return; + /* delete driver data */ cso_for_each_state(sc, CSO_BLEND, delete_blend_state, 0); cso_for_each_state(sc, CSO_DEPTH_STENCIL_ALPHA, delete_depth_stencil_state, 0); diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index f388bf5d951..36c882acb75 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -55,14 +55,14 @@ struct cso_context { void *samplers[PIPE_MAX_SAMPLERS]; unsigned nr_samplers; - void *samplers_saved[PIPE_MAX_SAMPLERS]; unsigned nr_samplers_saved; + void *samplers_saved[PIPE_MAX_SAMPLERS]; struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; uint nr_textures; - struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS]; uint nr_textures_saved; + struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS]; /** Current and saved state. * The saved state is used as a 1-deep stack. diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 3cde9d36d3a..1c6d657297c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -158,6 +158,60 @@ static void do_triangle( struct draw_context *draw, +#define QUAD(i0,i1,i2,i3) \ + do_triangle( draw, \ + ( DRAW_PIPE_RESET_STIPPLE | \ + DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_2 ), \ + verts + stride * elts[i0], \ + verts + stride * elts[i1], \ + verts + stride * elts[i3]); \ + do_triangle( draw, \ + ( DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_1 ), \ + verts + stride * elts[i1], \ + verts + stride * elts[i2], \ + verts + stride * elts[i3]) + +#define TRIANGLE(flags,i0,i1,i2) \ + do_triangle( draw, \ + elts[i0], /* flags */ \ + verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * elts[i1], \ + verts + stride * elts[i2]) + +#define LINE(flags,i0,i1) \ + do_line( draw, \ + elts[i0], \ + verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * elts[i1]) + +#define POINT(i0) \ + do_point( draw, \ + verts + stride * elts[i0] ) + +#define FUNC pipe_run +#define ARGS \ + struct draw_context *draw, \ + unsigned prim, \ + struct vertex_header *vertices, \ + unsigned stride, \ + const ushort *elts + +#define LOCAL_VARS \ + char *verts = (char *)vertices; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned i; \ + ushort flags + +#define FLUSH + +#include "draw_pt_decompose.h" +#undef ARGS +#undef LOCAL_VARS + + /* Code to run the pipeline on a fairly arbitary collection of vertices. * @@ -178,34 +232,12 @@ void draw_pipeline_run( struct draw_context *draw, unsigned count ) { char *verts = (char *)vertices; - unsigned i; draw->pipeline.verts = verts; draw->pipeline.vertex_stride = stride; draw->pipeline.vertex_count = vertex_count; - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i++) - do_point( draw, - verts + stride * elts[i] ); - break; - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) - do_line( draw, - elts[i+0], /* flags */ - verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), - verts + stride * elts[i+1]); - break; - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) - do_triangle( draw, - elts[i+0], /* flags */ - verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), - verts + stride * elts[i+1], - verts + stride * elts[i+2]); - break; - } + pipe_run(draw, prim, vertices, stride, elts, count); draw->pipeline.verts = NULL; draw->pipeline.vertex_count = 0; @@ -216,22 +248,22 @@ void draw_pipeline_run( struct draw_context *draw, ( DRAW_PIPE_RESET_STIPPLE | \ DRAW_PIPE_EDGE_FLAG_0 | \ DRAW_PIPE_EDGE_FLAG_2 ), \ - verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ - verts + stride * (i1), \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ verts + stride * (i3)); \ - do_triangle( draw, \ - ( DRAW_PIPE_EDGE_FLAG_0 | \ - DRAW_PIPE_EDGE_FLAG_1 ), \ - verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \ - verts + stride * (i2), \ - verts + stride * (i3)) + do_triangle( draw, \ + ( DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_1 ), \ + verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i2), \ + verts + stride * (i3)) #define TRIANGLE(flags,i0,i1,i2) \ do_triangle( draw, \ flags, /* flags */ \ - verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ - verts + stride * (i1), \ - verts + stride * (i2)) + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ + verts + stride * (i2)) #define LINE(flags,i0,i1) \ do_line( draw, \ diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 30a6d2919d9..283502cdf3e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -256,7 +256,10 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, uint size = 4; immed = tgsi_default_full_immediate(); immed.Immediate.NrTokens = 1 + size; /* one for the token itself */ - immed.u.Pointer = (void *) value; + immed.u[0].Float = value[0]; + immed.u[1].Float = value[1]; + immed.u[2].Float = value[2]; + immed.u[3].Float = value[3]; ctx->emit_immediate(ctx, &immed); } diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index a5d840b96ea..1a5269c0de9 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -159,8 +159,19 @@ vbuf_tri( struct draw_stage *stage, check_space( vbuf, 3 ); - for (i = 0; i < 3; i++) { - vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); + if (vbuf->stage.draw->rasterizer->flatshade_first) { + /* Put provoking vertex in position expected by the driver. + * Emit last provoking vertex in first pos. + * Swap verts 0 & 1 to preserve polygon winding. + */ + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[2] ); + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] ); + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[1] ); + } + else { + for (i = 0; i < 3; i++) { + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); + } } } diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index d84bab9eaae..7d76a7dbf39 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -28,6 +28,30 @@ /* Authors: Keith Whitwell <keith@tungstengraphics.com> */ +/** + * Notes on wide points and sprite mode: + * + * In wide point/sprite mode we effectively need to convert each incoming + * vertex into four outgoing vertices specifying the corners of a quad. + * Since we don't (yet) have geometry shaders, we have to handle this here + * in the draw module. + * + * For sprites, it also means that this is where we have to handle texcoords + * for the vertices of the quad. OpenGL's GL_COORD_REPLACE state specifies + * if/how enabled texcoords are automatically generated for sprites. We pass + * that info through gallium in the pipe_rasterizer_state::sprite_coord_mode + * array. + * + * Additionally, GLSL's gl_PointCoord fragment attribute has to be handled + * here as well. This is basically an additional texture/generic attribute + * that varies .x from 0 to 1 horizontally across the point and varies .y + * vertically from 0 to 1 down the sprite. + * + * With geometry shaders, the state tracker could create a GS to do + * most/all of this. + */ + + #include "util/u_math.h" #include "util/u_memory.h" #include "pipe/p_defines.h" @@ -52,7 +76,7 @@ struct widepoint_stage { int psize_slot; - int point_coord_fs_input; /**< input for pointcoord (and fog) */ + int point_coord_fs_input; /**< input for pointcoord */ }; @@ -64,8 +88,6 @@ widepoint_stage( struct draw_stage *stage ) } - - /** * Set the vertex texcoords for sprite mode. * Coords may be left untouched or set to a right-side-up or upside-down @@ -89,10 +111,12 @@ static void set_texcoords(const struct widepoint_stage *wide, } if (wide->point_coord_fs_input >= 0) { - /* put gl_PointCoord into extra vertex output's zw components */ - uint k = wide->stage.draw->extra_vp_outputs.slot; - v->data[k][2] = tc[0]; - v->data[k][3] = tc[1]; + /* put gl_PointCoord into the extra vertex slot */ + uint slot = wide->stage.draw->extra_vp_outputs.slot; + v->data[slot][0] = tc[0]; + v->data[slot][1] = tc[1]; + v->data[slot][2] = 0.0F; + v->data[slot][3] = 1.0F; } } @@ -181,6 +205,16 @@ static void widepoint_point( struct draw_stage *stage, } +static int +find_pntc_input_attrib(struct draw_context *draw) +{ + /* Scan the fragment program's input decls to find the pointcoord + * attribute. The xy components will store the point coord. + */ + return 0; /* XXX fix this */ +} + + static void widepoint_first_point( struct draw_stage *stage, struct prim_header *header ) { @@ -219,8 +253,8 @@ static void widepoint_first_point( struct draw_stage *stage, } wide->num_texcoords = j; - /* find fragment shader PointCoord/Fog input */ - wide->point_coord_fs_input = 0; /* XXX fix this! */ + /* find fragment shader PointCoord input */ + wide->point_coord_fs_input = find_pntc_input_attrib(draw); /* setup extra vp output (point coord implemented as a texcoord) */ draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 81e4eae401c..41fcb16a0a5 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -44,7 +44,6 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_scan.h" @@ -55,6 +54,8 @@ struct draw_vertex_shader; struct draw_context; struct draw_stage; struct vbuf_render; +struct tgsi_exec_machine; +struct tgsi_sampler; /** @@ -185,7 +186,7 @@ struct draw_context uint position_output; /** TGSI program interpreter runtime state */ - struct tgsi_exec_machine machine; + struct tgsi_exec_machine *machine; uint num_samplers; struct tgsi_sampler **samplers; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 9ea0cbe5990..dbb5ac71821 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -36,6 +36,7 @@ #include "draw/draw_vs.h" #include "tgsi/tgsi_dump.h" #include "util/u_math.h" +#include "util/u_prim.h" static unsigned trim( unsigned count, unsigned first, unsigned incr ) { @@ -278,7 +279,7 @@ void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - unsigned reduced_prim = draw_pt_reduced_prim(prim); + unsigned reduced_prim = u_reduced_prim(prim); if (reduced_prim != draw->reduced_prim) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->reduced_prim = reduced_prim; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 8ef0ea8011f..7a17a9fb6b2 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -228,7 +228,6 @@ void draw_pt_post_vs_destroy( struct pt_post_vs *pvs ); * Utils: */ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr); -unsigned draw_pt_reduced_prim(unsigned prim); #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h index 3fb06956878..4ca5b520204 100644 --- a/src/gallium/auxiliary/draw/draw_pt_decompose.h +++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h @@ -47,10 +47,19 @@ static void FUNC( ARGS, case PIPE_PRIM_TRIANGLES: for (i = 0; i+2 < count; i += 3) { - TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - (i + 0), - (i + 1), - (i + 2 )); + if (flatfirst) { + /* put provoking vertex in last pos for clipper */ + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 1), + (i + 2), + (i + 0 )); + } + else { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0), + (i + 1), + (i + 2 )); + } } break; @@ -58,9 +67,9 @@ static void FUNC( ARGS, if (flatfirst) { for (i = 0; i+2 < count; i++) { TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - (i + 0), (i + 1 + (i&1)), - (i + 2 - (i&1))); + (i + 2 - (i&1)), + (i + 0) ); } } else { @@ -78,9 +87,9 @@ static void FUNC( ARGS, if (flatfirst) { for (i = 0; i+2 < count; i++) { TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - (i + 1), (i + 2), - (0 )); + 0, + (i + 1) ); } } else { @@ -96,20 +105,40 @@ static void FUNC( ARGS, case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - QUAD( (i + 0), - (i + 1), - (i + 2), - (i + 3)); + if (flatfirst) { + for (i = 0; i+3 < count; i += 4) { + QUAD( (i + 1), + (i + 2), + (i + 3), + (i + 0) ); + } + } + else { + for (i = 0; i+3 < count; i += 4) { + QUAD( (i + 0), + (i + 1), + (i + 2), + (i + 3)); + } } break; case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - QUAD( (i + 2), - (i + 0), - (i + 1), - (i + 3)); + if (flatfirst) { + for (i = 0; i+3 < count; i += 2) { + QUAD( (i + 1), + (i + 3), + (i + 2), + (i + 0) ); + } + } + else { + for (i = 0; i+3 < count; i += 2) { + QUAD( (i + 2), + (i + 0), + (i + 1), + (i + 3)); + } } break; diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index b7780fb5073..88f4d9f495a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -54,7 +54,8 @@ static unsigned elt_ubyte( const void *elts, unsigned idx ) static unsigned elt_vert( const void *elts, unsigned idx ) { - return (const ubyte *)elts - (const ubyte *)NULL + idx; + /* unsigned index is packed in the pointer */ + return (unsigned)(uintptr_t)elts + idx; } pt_elt_func draw_pt_elt_func( struct draw_context *draw ) diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c index 3bc7939c556..b61fa291436 100644 --- a/src/gallium/auxiliary/draw/draw_pt_util.c +++ b/src/gallium/auxiliary/draw/draw_pt_util.c @@ -75,28 +75,3 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr) break; } } - - -unsigned draw_pt_reduced_prim(unsigned prim) -{ - switch (prim) { - case PIPE_PRIM_POINTS: - return PIPE_PRIM_POINTS; - case PIPE_PRIM_LINES: - case PIPE_PRIM_LINE_STRIP: - case PIPE_PRIM_LINE_LOOP: - return PIPE_PRIM_LINES; - case PIPE_PRIM_TRIANGLES: - case PIPE_PRIM_TRIANGLE_STRIP: - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - case PIPE_PRIM_QUADS: - case PIPE_PRIM_QUAD_STRIP: - return PIPE_PRIM_TRIANGLES; - default: - assert(0); - return PIPE_PRIM_POINTS; - } -} - - diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 5d268a22264..1a0527be63a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -31,6 +31,7 @@ */ #include "util/u_memory.h" +#include "util/u_prim.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_pt.h" @@ -193,16 +194,30 @@ vcache_ef_quad( struct vcache_frontend *vcache, unsigned i2, unsigned i3 ) { - vcache_triangle_flags( vcache, - ( DRAW_PIPE_RESET_STIPPLE | - DRAW_PIPE_EDGE_FLAG_0 | - DRAW_PIPE_EDGE_FLAG_2 ), - i0, i1, i3 ); - - vcache_triangle_flags( vcache, - ( DRAW_PIPE_EDGE_FLAG_0 | - DRAW_PIPE_EDGE_FLAG_1 ), - i1, i2, i3 ); + if (vcache->draw->rasterizer->flatshade_first) { + vcache_triangle_flags( vcache, + ( DRAW_PIPE_RESET_STIPPLE | + DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_1 ), + i0, i1, i2 ); + + vcache_triangle_flags( vcache, + ( DRAW_PIPE_EDGE_FLAG_2 | + DRAW_PIPE_EDGE_FLAG_1 ), + i0, i2, i3 ); + } + else { + vcache_triangle_flags( vcache, + ( DRAW_PIPE_RESET_STIPPLE | + DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_2 ), + i0, i1, i3 ); + + vcache_triangle_flags( vcache, + ( DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_1 ), + i1, i2, i3 ); + } } /* At least for now, we're back to using a template include file for @@ -453,7 +468,7 @@ vcache_prepare( struct draw_pt_front_end *frontend, } vcache->input_prim = prim; - vcache->output_prim = draw_pt_reduced_prim(prim); + vcache->output_prim = u_reduced_prim(prim); vcache->middle = middle; vcache->opt = opt; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h index ec05bbeab40..62822a3d562 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -118,21 +118,39 @@ static void FUNC( struct draw_pt_front_end *frontend, case PIPE_PRIM_QUADS: for (i = 0; i+3 < count; i += 4) { - QUAD( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, i + 3)); + if (flatfirst) { + QUAD( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3) ); + } + else { + QUAD( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3) ); + } } break; case PIPE_PRIM_QUAD_STRIP: for (i = 0; i+3 < count; i += 2) { - QUAD( vcache, - get_elt(elts, i + 2), - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 3)); + if (flatfirst) { + QUAD( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3), + get_elt(elts, i + 2) ); + } + else { + QUAD( vcache, + get_elt(elts, i + 2), + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3) ); + } } break; @@ -144,19 +162,38 @@ static void FUNC( struct draw_pt_front_end *frontend, const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; + ushort edge_next, edge_finish; - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + if (flatfirst) { + flags = DRAW_PIPE_RESET_STIPPLE | edge_middle | edge_last; + edge_next = edge_last; + edge_finish = edge_first; + } + else { + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + edge_next = edge_middle; + edge_finish = edge_last; + } - for (i = 0; i+2 < count; i++, flags = edge_middle) { + for (i = 0; i+2 < count; i++, flags = edge_next) { if (i + 3 == count) - flags |= edge_last; + flags |= edge_finish; - TRIANGLE( vcache, - flags, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); + if (flatfirst) { + TRIANGLE( vcache, + flags, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + else { + TRIANGLE( vcache, + flags, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } } } break; diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index c057cd67fde..790e89ed820 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -43,6 +43,8 @@ #include "translate/translate.h" #include "translate/translate_cache.h" +#include "tgsi/tgsi_exec.h" + @@ -146,16 +148,8 @@ draw_delete_vertex_shader(struct draw_context *draw, boolean draw_vs_init( struct draw_context *draw ) { - tgsi_exec_machine_init(&draw->vs.machine); - - /* FIXME: give this machine thing a proper constructor: - */ - draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); - if (!draw->vs.machine.Inputs) - return FALSE; - - draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); - if (!draw->vs.machine.Outputs) + draw->vs.machine = tgsi_exec_machine_create(); + if (!draw->vs.machine) return FALSE; draw->vs.emit_cache = translate_cache_create(); @@ -178,12 +172,6 @@ draw_vs_init( struct draw_context *draw ) void draw_vs_destroy( struct draw_context *draw ) { - if (draw->vs.machine.Inputs) - align_free(draw->vs.machine.Inputs); - - if (draw->vs.machine.Outputs) - align_free(draw->vs.machine.Outputs); - if (draw->vs.fetch_cache) translate_cache_destroy(draw->vs.fetch_cache); @@ -196,8 +184,7 @@ draw_vs_destroy( struct draw_context *draw ) if (draw->vs.aligned_constant_storage) align_free((void*)draw->vs.aligned_constant_storage); - tgsi_exec_machine_free_data(&draw->vs.machine); - + tgsi_exec_machine_destroy(draw->vs.machine); } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 9e37a26c1e2..62e04a65f30 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1758,24 +1758,24 @@ emit_instruction( struct aos_compilation *cp, case TGSI_OPCODE_SUB: return emit_SUB(cp, inst); - case TGSI_OPCODE_LERP: + case TGSI_OPCODE_LRP: // return emit_LERP(cp, inst); return FALSE; - case TGSI_OPCODE_FRAC: + case TGSI_OPCODE_FRC: return emit_FRC(cp, inst); case TGSI_OPCODE_CLAMP: // return emit_CLAMP(cp, inst); return FALSE; - case TGSI_OPCODE_FLOOR: + case TGSI_OPCODE_FLR: return emit_FLR(cp, inst); case TGSI_OPCODE_ROUND: return emit_RND(cp, inst); - case TGSI_OPCODE_EXPBASE2: + case TGSI_OPCODE_EX2: #if FAST_MATH return emit_EXPBASE2(cp, inst); #elif 0 @@ -1787,13 +1787,13 @@ emit_instruction( struct aos_compilation *cp, return FALSE; #endif - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_LG2: return emit_LG2(cp, inst); - case TGSI_OPCODE_POWER: + case TGSI_OPCODE_POW: return emit_POW(cp, inst); - case TGSI_OPCODE_CROSSPRODUCT: + case TGSI_OPCODE_XPD: return emit_XPD(cp, inst); case TGSI_OPCODE_ABS: @@ -1891,8 +1891,9 @@ static boolean note_immediate( struct aos_compilation *cp, unsigned pos = cp->num_immediates++; unsigned j; + assert( imm->Immediate.NrTokens <= 4 + 1 ); for (j = 0; j < imm->Immediate.NrTokens - 1; j++) { - cp->vaos->machine->immediate[pos][j] = imm->u.ImmediateFloat32[j].Float; + cp->vaos->machine->immediate[pos][j] = imm->u[j].Float; } return TRUE; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index f2368dde5c2..41cc8026131 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -41,6 +41,7 @@ #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_scan.h" +#include "tgsi/tgsi_exec.h" struct exec_vertex_shader { @@ -201,7 +202,7 @@ draw_create_vs_exec(struct draw_context *draw, vs->base.run_linear = vs_exec_run_linear; vs->base.delete = vs_exec_delete; vs->base.create_varient = draw_vs_varient_generic; - vs->machine = &draw->vs.machine; + vs->machine = draw->vs.machine; return &vs->base; } diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 727977bc3af..b3535c0e48e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -119,7 +119,7 @@ draw_create_vs_llvm(struct draw_context *draw, vs->base.create_varient = draw_vs_varient_generic; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; - vs->machine = &draw->vs.machine; + vs->machine = draw->vs.machine; { struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); diff --git a/src/gallium/auxiliary/draw/draw_vs_ppc.c b/src/gallium/auxiliary/draw/draw_vs_ppc.c index d35db57d571..ad184bd696d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_ppc.c +++ b/src/gallium/auxiliary/draw/draw_vs_ppc.c @@ -48,6 +48,7 @@ #include "rtasm/rtasm_ppc.h" #include "tgsi/tgsi_ppc.h" #include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_exec.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 77ba5152f9f..702051387ac 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -48,27 +48,16 @@ #include "rtasm/rtasm_x86sse.h" #include "tgsi/tgsi_sse2.h" #include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_exec.h" #define SSE_MAX_VERTICES 4 -typedef void (PIPE_CDECL *codegen_function) ( - const struct tgsi_exec_vector *input, /* 1 */ - struct tgsi_exec_vector *output, /* 2 */ - float (*constant)[4], /* 3 */ - struct tgsi_exec_vector *temporary, /* 4 */ - float (*immediates)[4], /* 5 */ - const float (*aos_input)[4], /* 6 */ - uint num_inputs, /* 7 */ - uint input_stride, /* 8 */ - float (*aos_output)[4], /* 9 */ - uint num_outputs, /* 10 */ - uint output_stride ); /* 11 */ struct draw_sse_vertex_shader { struct draw_vertex_shader base; struct x86_function sse2_program; - codegen_function func; + tgsi_sse2_vs_func func; struct tgsi_exec_machine *machine; }; @@ -78,6 +67,10 @@ static void vs_sse_prepare( struct draw_vertex_shader *base, struct draw_context *draw ) { + struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; + struct tgsi_exec_machine *machine = shader->machine; + + machine->Samplers = draw->vs.samplers; } @@ -118,11 +111,9 @@ vs_sse_run_linear( struct draw_vertex_shader *base, /* run compiled shader */ - shader->func(machine->Inputs, - machine->Outputs, - (float (*)[4])constants, - machine->Temps, - (float (*)[4])shader->base.immediates, + shader->func(machine, + constants, + shader->base.immediates, input, base->info.num_inputs, input_stride, @@ -184,7 +175,7 @@ draw_create_vs_sse(struct draw_context *draw, vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 * sizeof(float), 16); - vs->machine = &draw->vs.machine; + vs->machine = draw->vs.machine; x86_init_func( &vs->sse2_program ); @@ -194,7 +185,7 @@ draw_create_vs_sse(struct draw_context *draw, TRUE )) goto fail; - vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); + vs->func = (tgsi_sse2_vs_func) x86_get_func( &vs->sse2_program ); if (!vs->func) { goto fail; } diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 2d2af3085e6..721b7d2d833 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -128,7 +128,7 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; m_functionsMap[TGSI_OPCODE_MIN] = "min"; m_functionsMap[TGSI_OPCODE_MAX] = "max"; - m_functionsMap[TGSI_OPCODE_POWER] = "pow"; + m_functionsMap[TGSI_OPCODE_POW] = "pow"; m_functionsMap[TGSI_OPCODE_LIT] = "lit"; m_functionsMap[TGSI_OPCODE_RSQ] = "rsq"; m_functionsMap[TGSI_OPCODE_SLT] = "slt"; @@ -311,7 +311,7 @@ std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> i std::vector<llvm::Value*> InstructionsSoa::pow(const std::vector<llvm::Value*> in1, const std::vector<llvm::Value*> in2) { - llvm::Function *func = function(TGSI_OPCODE_POWER); + llvm::Function *func = function(TGSI_OPCODE_POW); return callBuiltin(func, in1, in2); } diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 5b08200d142..bf84401e112 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -160,10 +160,11 @@ translate_immediate(Storage *storage, { float vec[4]; int i; + assert( imm->Immediate.NrTokens <= 4 + 1 ); for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) { switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; + vec[i] = imm->u[i].Float; break; default: assert(0); @@ -179,10 +180,11 @@ translate_immediateir(StorageSoa *storage, { float vec[4]; int i; + assert( imm->Immediate.NrTokens <= 4 + 1 ); for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) { switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; + vec[i] = imm->u[i].Float; break; default: assert(0); @@ -336,7 +338,7 @@ translate_instruction(llvm::Module *module, out = instr->sub(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_LERP: { + case TGSI_OPCODE_LRP: { out = instr->lerp(inputs[0], inputs[1], inputs[2]); } break; @@ -348,17 +350,11 @@ translate_instruction(llvm::Module *module, out = instr->cnd0(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_DOT2ADD: { + case TGSI_OPCODE_DP2A: { out = instr->dot2add(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: { - out = instr->neg(inputs[0]); - } - break; - case TGSI_OPCODE_FRAC: { + case TGSI_OPCODE_FRC: { out = instr->frc(inputs[0]); } break; @@ -366,30 +362,28 @@ translate_instruction(llvm::Module *module, out = instr->clamp(inputs[0]); } break; - case TGSI_OPCODE_FLOOR: { + case TGSI_OPCODE_FLR: { out = instr->floor(inputs[0]); } break; case TGSI_OPCODE_ROUND: break; - case TGSI_OPCODE_EXPBASE2: { + case TGSI_OPCODE_EX2: { out = instr->ex2(inputs[0]); } break; - case TGSI_OPCODE_LOGBASE2: { + case TGSI_OPCODE_LG2: { out = instr->lg2(inputs[0]); } break; - case TGSI_OPCODE_POWER: { + case TGSI_OPCODE_POW: { out = instr->pow(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_CROSSPRODUCT: { + case TGSI_OPCODE_XPD: { out = instr->cross(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; case TGSI_OPCODE_ABS: { out = instr->abs(inputs[0]); } @@ -522,7 +516,7 @@ translate_instruction(llvm::Module *module, return; //just update the state } break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: break; case TGSI_OPCODE_REP: break; @@ -538,7 +532,7 @@ translate_instruction(llvm::Module *module, return; //just update the state } break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: break; case TGSI_OPCODE_ENDREP: break; @@ -580,7 +574,7 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_ENDPRIM: break; - case TGSI_OPCODE_BGNLOOP2: { + case TGSI_OPCODE_BGNLOOP: { instr->beginLoop(); storage->setCurrentBlock(instr->currentBlock()); return; @@ -593,7 +587,7 @@ translate_instruction(llvm::Module *module, return; } break; - case TGSI_OPCODE_ENDLOOP2: { + case TGSI_OPCODE_ENDLOOP: { instr->endLoop(); storage->setCurrentBlock(instr->currentBlock()); return; @@ -617,14 +611,6 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; case TGSI_OPCODE_CALLNZ: break; case TGSI_OPCODE_IFC: @@ -778,44 +764,38 @@ translate_instructionir(llvm::Module *module, out = instr->sub(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_LERP: { + case TGSI_OPCODE_LRP: { } break; case TGSI_OPCODE_CND: break; case TGSI_OPCODE_CND0: break; - case TGSI_OPCODE_DOT2ADD: + case TGSI_OPCODE_DP2A: break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { + case TGSI_OPCODE_FRC: { } break; case TGSI_OPCODE_CLAMP: break; - case TGSI_OPCODE_FLOOR: { + case TGSI_OPCODE_FLR: { } break; case TGSI_OPCODE_ROUND: break; - case TGSI_OPCODE_EXPBASE2: { + case TGSI_OPCODE_EX2: { } break; - case TGSI_OPCODE_LOGBASE2: { + case TGSI_OPCODE_LG2: { } break; - case TGSI_OPCODE_POWER: { + case TGSI_OPCODE_POW: { out = instr->pow(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_CROSSPRODUCT: { + case TGSI_OPCODE_XPD: { } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; case TGSI_OPCODE_ABS: { out = instr->abs(inputs[0]); } @@ -910,7 +890,7 @@ translate_instructionir(llvm::Module *module, case TGSI_OPCODE_IF: { } break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: break; case TGSI_OPCODE_REP: break; @@ -920,7 +900,7 @@ translate_instructionir(llvm::Module *module, case TGSI_OPCODE_ENDIF: { } break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: break; case TGSI_OPCODE_ENDREP: break; @@ -961,13 +941,13 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_ENDPRIM: break; - case TGSI_OPCODE_BGNLOOP2: { + case TGSI_OPCODE_BGNLOOP: { } break; case TGSI_OPCODE_BGNSUB: { } break; - case TGSI_OPCODE_ENDLOOP2: { + case TGSI_OPCODE_ENDLOOP: { } break; case TGSI_OPCODE_ENDSUB: { @@ -983,14 +963,6 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; case TGSI_OPCODE_NRM4: break; case TGSI_OPCODE_CALLNZ: diff --git a/src/gallium/auxiliary/indices/u_indices.c b/src/gallium/auxiliary/indices/u_indices.c index 0cf7d88653c..03d7453f303 100644 --- a/src/gallium/auxiliary/indices/u_indices.c +++ b/src/gallium/auxiliary/indices/u_indices.c @@ -244,7 +244,7 @@ int u_index_generator( unsigned hw_mask, default: assert(0); - *out_generate = generate[out_idx][in_pv][out_pv][prim]; + *out_generate = generate[out_idx][in_pv][out_pv][PIPE_PRIM_POINTS]; *out_prim = PIPE_PRIM_POINTS; *out_nr = nr; return U_TRANSLATE_ERROR; diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 92b6fd00564..2590546cb4a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -72,6 +72,12 @@ struct pb_desc /** + * Size. Regular (32bit) unsigned for now. + */ +typedef unsigned pb_size; + + +/** * Base class for all pb_* buffers. */ struct pb_buffer @@ -126,7 +132,7 @@ struct pb_vtbl */ void (*get_base_buffer)( struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset ); + pb_size *offset ); }; @@ -177,7 +183,7 @@ pb_unmap(struct pb_buffer *buf) static INLINE void pb_get_base_buffer( struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset ) + pb_size *offset ) { assert(buf); if(!buf) { @@ -241,7 +247,7 @@ pb_reference(struct pb_buffer **dst, * the requested or not. */ static INLINE boolean -pb_check_alignment(size_t requested, size_t provided) +pb_check_alignment(pb_size requested, pb_size provided) { if(!requested) return TRUE; @@ -269,7 +275,7 @@ pb_check_usage(unsigned requested, unsigned provided) * hardware. */ struct pb_buffer * -pb_malloc_buffer_create(size_t size, +pb_malloc_buffer_create(pb_size size, const struct pb_desc *desc); diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 4698efa69cf..109ac7c9d63 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -65,11 +65,11 @@ struct fenced_buffer_list struct pb_fence_ops *ops; - size_t numDelayed; + pb_size numDelayed; struct list_head delayed; #ifdef DEBUG - size_t numUnfenced; + pb_size numUnfenced; struct list_head unfenced; #endif }; @@ -433,7 +433,7 @@ fenced_buffer_fence(struct pb_buffer *buf, static void fenced_buffer_get_base_buffer(struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c index 689fd74771f..6bdce5fcb06 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -53,6 +53,8 @@ static INLINE struct malloc_buffer * malloc_buffer(struct pb_buffer *buf) { assert(buf); + if (!buf) + return NULL; assert(buf->vtbl == &malloc_buffer_vtbl); return (struct malloc_buffer *)buf; } @@ -102,7 +104,7 @@ malloc_buffer_fence(struct pb_buffer *buf, static void malloc_buffer_get_base_buffer(struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { *base_buf = buf; *offset = 0; @@ -121,7 +123,7 @@ malloc_buffer_vtbl = { struct pb_buffer * -pb_malloc_buffer_create(size_t size, +pb_malloc_buffer_create(pb_size size, const struct pb_desc *desc) { struct malloc_buffer *buf; @@ -150,7 +152,7 @@ pb_malloc_buffer_create(size_t size, static struct pb_buffer * pb_malloc_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { return pb_malloc_buffer_create(size, desc); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 74077f82774..39ab8e722c1 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -73,7 +73,7 @@ struct pb_manager struct pb_buffer * (*create_buffer)( struct pb_manager *mgr, - size_t size, + pb_size size, const struct pb_desc *desc); /** @@ -106,7 +106,7 @@ pb_malloc_bufmgr_create(void); */ struct pb_manager * pool_bufmgr_create(struct pb_manager *provider, - size_t n, size_t size, + pb_size n, pb_size size, const struct pb_desc *desc); @@ -119,7 +119,7 @@ pool_bufmgr_create(struct pb_manager *provider, */ struct pb_manager * mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2); + pb_size size, pb_size align2); /** * Same as mm_bufmgr_create. @@ -128,7 +128,7 @@ mm_bufmgr_create(struct pb_manager *provider, */ struct pb_manager * mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2); + pb_size size, pb_size align2); /** @@ -136,8 +136,8 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, */ struct pb_manager * pb_slab_manager_create(struct pb_manager *provider, - size_t bufSize, - size_t slabSize, + pb_size bufSize, + pb_size slabSize, const struct pb_desc *desc); /** @@ -146,9 +146,9 @@ pb_slab_manager_create(struct pb_manager *provider, */ struct pb_manager * pb_slab_range_manager_create(struct pb_manager *provider, - size_t minBufSize, - size_t maxBufSize, - size_t slabSize, + pb_size minBufSize, + pb_size maxBufSize, + pb_size slabSize, const struct pb_desc *desc); @@ -204,7 +204,7 @@ pb_ondemand_manager_create(struct pb_manager *provider); */ struct pb_manager * pb_debug_manager_create(struct pb_manager *provider, - size_t underflow_size, size_t overflow_size); + pb_size underflow_size, pb_size overflow_size); #ifdef __cplusplus diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c index db67d46c561..f60c836f18b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c @@ -60,7 +60,7 @@ pb_alt_manager(struct pb_manager *mgr) static struct pb_buffer * pb_alt_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_alt_manager *mgr = pb_alt_manager(_mgr); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index 35358430b43..57d1ede45a4 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -81,7 +81,7 @@ struct pb_cache_manager pipe_mutex mutex; struct list_head delayed; - size_t numDelayed; + pb_size numDelayed; }; @@ -204,7 +204,7 @@ pb_cache_buffer_fence(struct pb_buffer *_buf, static void pb_cache_buffer_get_base_buffer(struct pb_buffer *_buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct pb_cache_buffer *buf = pb_cache_buffer(_buf); pb_get_base_buffer(buf->buffer, base_buf, offset); @@ -224,7 +224,7 @@ pb_cache_buffer_vtbl = { static INLINE boolean pb_cache_is_buffer_compat(struct pb_cache_buffer *buf, - size_t size, + pb_size size, const struct pb_desc *desc) { if(buf->base.base.size < size) @@ -246,7 +246,7 @@ pb_cache_is_buffer_compat(struct pb_cache_buffer *buf, static struct pb_buffer * pb_cache_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_cache_manager *mgr = pb_cache_manager(_mgr); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index cedf745bdac..1b4df28c707 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -72,8 +72,8 @@ struct pb_debug_buffer struct pb_buffer *buffer; struct pb_debug_manager *mgr; - size_t underflow_size; - size_t overflow_size; + pb_size underflow_size; + pb_size overflow_size; struct debug_stack_frame create_backtrace[PB_DEBUG_CREATE_BACKTRACE]; @@ -91,8 +91,8 @@ struct pb_debug_manager struct pb_manager *provider; - size_t underflow_size; - size_t overflow_size; + pb_size underflow_size; + pb_size overflow_size; pipe_mutex mutex; struct list_head list; @@ -124,9 +124,9 @@ static const uint8_t random_pattern[32] = { static INLINE void -fill_random_pattern(uint8_t *dst, size_t size) +fill_random_pattern(uint8_t *dst, pb_size size) { - size_t i = 0; + pb_size i = 0; while(size--) { *dst++ = random_pattern[i++]; i &= sizeof(random_pattern) - 1; @@ -135,11 +135,11 @@ fill_random_pattern(uint8_t *dst, size_t size) static INLINE boolean -check_random_pattern(const uint8_t *dst, size_t size, - size_t *min_ofs, size_t *max_ofs) +check_random_pattern(const uint8_t *dst, pb_size size, + pb_size *min_ofs, pb_size *max_ofs) { boolean result = TRUE; - size_t i; + pb_size i; *min_ofs = size; *max_ofs = 0; for(i = 0; i < size; ++i) { @@ -183,7 +183,7 @@ pb_debug_buffer_check(struct pb_debug_buffer *buf) assert(map); if(map) { boolean underflow, overflow; - size_t min_ofs, max_ofs; + pb_size min_ofs, max_ofs; underflow = !check_random_pattern(map, buf->underflow_size, &min_ofs, &max_ofs); @@ -287,7 +287,7 @@ pb_debug_buffer_unmap(struct pb_buffer *_buf) static void pb_debug_buffer_get_base_buffer(struct pb_buffer *_buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct pb_debug_buffer *buf = pb_debug_buffer(_buf); pb_get_base_buffer(buf->buffer, base_buf, offset); @@ -363,13 +363,13 @@ pb_debug_manager_dump(struct pb_debug_manager *mgr) static struct pb_buffer * pb_debug_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_debug_manager *mgr = pb_debug_manager(_mgr); struct pb_debug_buffer *buf; struct pb_desc real_desc; - size_t real_size; + pb_size real_size; buf = CALLOC_STRUCT(pb_debug_buffer); if(!buf) @@ -455,7 +455,7 @@ pb_debug_manager_destroy(struct pb_manager *_mgr) struct pb_manager * pb_debug_manager_create(struct pb_manager *provider, - size_t underflow_size, size_t overflow_size) + pb_size underflow_size, pb_size overflow_size) { struct pb_debug_manager *mgr; @@ -485,7 +485,7 @@ pb_debug_manager_create(struct pb_manager *provider, struct pb_manager * pb_debug_manager_create(struct pb_manager *provider, - size_t underflow_size, size_t overflow_size) + pb_size underflow_size, pb_size overflow_size) { return provider; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index 144db5669b6..97dd1427fda 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -62,7 +62,7 @@ fenced_pb_manager(struct pb_manager *mgr) static struct pb_buffer * fenced_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index d8decba49df..6400fc5b0a3 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -55,10 +55,10 @@ struct mm_pb_manager pipe_mutex mutex; - size_t size; + pb_size size; struct mem_block *heap; - size_t align2; + pb_size align2; struct pb_buffer *buffer; void *map; @@ -148,7 +148,7 @@ mm_buffer_fence(struct pb_buffer *buf, static void mm_buffer_get_base_buffer(struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct mm_buffer *mm_buf = mm_buffer(buf); struct mm_pb_manager *mm = mm_buf->mgr; @@ -170,15 +170,15 @@ mm_buffer_vtbl = { static struct pb_buffer * mm_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct mm_pb_manager *mm = mm_pb_manager(mgr); struct mm_buffer *mm_buf; /* We don't handle alignments larger then the one initially setup */ - assert(pb_check_alignment(desc->alignment, 1 << mm->align2)); - if(!pb_check_alignment(desc->alignment, 1 << mm->align2)) + assert(pb_check_alignment(desc->alignment, (pb_size)1 << mm->align2)); + if(!pb_check_alignment(desc->alignment, (pb_size)1 << mm->align2)) return NULL; pipe_mutex_lock(mm->mutex); @@ -198,7 +198,7 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, mm_buf->mgr = mm; - mm_buf->block = u_mmAllocMem(mm->heap, size, mm->align2, 0); + mm_buf->block = u_mmAllocMem(mm->heap, (int)size, (int)mm->align2, 0); if(!mm_buf->block) { #if 0 debug_printf("warning: heap full\n"); @@ -210,8 +210,8 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, } /* Some sanity checks */ - assert(0 <= (unsigned)mm_buf->block->ofs && (unsigned)mm_buf->block->ofs < mm->size); - assert(size <= (unsigned)mm_buf->block->size && (unsigned)mm_buf->block->ofs + (unsigned)mm_buf->block->size <= mm->size); + assert(0 <= (pb_size)mm_buf->block->ofs && (pb_size)mm_buf->block->ofs < mm->size); + assert(size <= (pb_size)mm_buf->block->size && (pb_size)mm_buf->block->ofs + (pb_size)mm_buf->block->size <= mm->size); pipe_mutex_unlock(mm->mutex); return SUPER(mm_buf); @@ -245,7 +245,7 @@ mm_bufmgr_destroy(struct pb_manager *mgr) struct pb_manager * mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2) + pb_size size, pb_size align2) { struct mm_pb_manager *mm; @@ -273,7 +273,7 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, if(!mm->map) goto failure; - mm->heap = u_mmInit(0, size); + mm->heap = u_mmInit(0, (int)size); if (!mm->heap) goto failure; @@ -292,7 +292,7 @@ if(mm->heap) struct pb_manager * mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2) + pb_size size, pb_size align2) { struct pb_buffer *buffer; struct pb_manager *mgr; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c index 4f7e6b1c4df..cb32d251367 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c @@ -55,7 +55,7 @@ struct pb_ondemand_buffer /** Real buffer */ struct pb_buffer *buffer; - size_t size; + pb_size size; struct pb_desc desc; }; @@ -74,6 +74,8 @@ static INLINE struct pb_ondemand_buffer * pb_ondemand_buffer(struct pb_buffer *buf) { assert(buf); + if (!buf) + return NULL; assert(buf->vtbl == &pb_ondemand_buffer_vtbl); return (struct pb_ondemand_buffer *)buf; } @@ -204,7 +206,7 @@ pb_ondemand_buffer_fence(struct pb_buffer *_buf, static void pb_ondemand_buffer_get_base_buffer(struct pb_buffer *_buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf); @@ -232,7 +234,7 @@ pb_ondemand_buffer_vtbl = { static struct pb_buffer * pb_ondemand_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_ondemand_manager *mgr = pb_ondemand_manager(_mgr); diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index 07fd1a22d93..7fd65ed2261 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -58,11 +58,11 @@ struct pool_pb_manager pipe_mutex mutex; - size_t bufSize; - size_t bufAlign; + pb_size bufSize; + pb_size bufAlign; - size_t numFree; - size_t numTot; + pb_size numFree; + pb_size numTot; struct list_head free; @@ -89,7 +89,7 @@ struct pool_buffer struct list_head head; - size_t start; + pb_size start; }; @@ -162,7 +162,7 @@ pool_buffer_fence(struct pb_buffer *buf, static void pool_buffer_get_base_buffer(struct pb_buffer *buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct pool_buffer *pool_buf = pool_buffer(buf); struct pool_pb_manager *pool = pool_buf->mgr; @@ -184,7 +184,7 @@ pool_buffer_vtbl = { static struct pb_buffer * pool_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pool_pb_manager *pool = pool_pb_manager(mgr); @@ -251,13 +251,13 @@ pool_bufmgr_destroy(struct pb_manager *mgr) struct pb_manager * pool_bufmgr_create(struct pb_manager *provider, - size_t numBufs, - size_t bufSize, + pb_size numBufs, + pb_size bufSize, const struct pb_desc *desc) { struct pool_pb_manager *pool; struct pool_buffer *pool_buf; - size_t i; + pb_size i; if(!provider) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 724aaadb436..e7352e90db9 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -68,7 +68,7 @@ struct pb_slab_buffer unsigned mapCount; /** Offset relative to the start of the slab buffer. */ - size_t start; + pb_size start; /** Use when validating, to signal that all mappings are finished */ /* TODO: Actually validation does not reach this stage yet */ @@ -83,8 +83,8 @@ struct pb_slab { struct list_head head; struct list_head freeBuffers; - size_t numBuffers; - size_t numFree; + pb_size numBuffers; + pb_size numFree; struct pb_slab_buffer *buffers; struct pb_slab_manager *mgr; @@ -108,10 +108,10 @@ struct pb_slab_manager struct pb_manager *provider; /** Size of the buffers we hand on downstream */ - size_t bufSize; + pb_size bufSize; /** Size of the buffers we request upstream */ - size_t slabSize; + pb_size slabSize; /** * Alignment, usage to be used to allocate the slab buffers. @@ -150,14 +150,14 @@ struct pb_slab_range_manager struct pb_manager *provider; - size_t minBufSize; - size_t maxBufSize; + pb_size minBufSize; + pb_size maxBufSize; /** @sa pb_slab_manager::desc */ struct pb_desc desc; unsigned numBuckets; - size_t *bucketSizes; + pb_size *bucketSizes; /** Array of pb_slab_manager, one for each bucket size */ struct pb_manager **buckets; @@ -270,7 +270,7 @@ pb_slab_buffer_fence(struct pb_buffer *_buf, static void pb_slab_buffer_get_base_buffer(struct pb_buffer *_buf, struct pb_buffer **base_buf, - unsigned *offset) + pb_size *offset) { struct pb_slab_buffer *buf = pb_slab_buffer(_buf); pb_get_base_buffer(buf->slab->bo, base_buf, offset); @@ -369,7 +369,7 @@ out_err0: static struct pb_buffer * pb_slab_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_slab_manager *mgr = pb_slab_manager(_mgr); @@ -450,8 +450,8 @@ pb_slab_manager_destroy(struct pb_manager *_mgr) struct pb_manager * pb_slab_manager_create(struct pb_manager *provider, - size_t bufSize, - size_t slabSize, + pb_size bufSize, + pb_size slabSize, const struct pb_desc *desc) { struct pb_slab_manager *mgr; @@ -479,11 +479,11 @@ pb_slab_manager_create(struct pb_manager *provider, static struct pb_buffer * pb_slab_range_manager_create_buffer(struct pb_manager *_mgr, - size_t size, + pb_size size, const struct pb_desc *desc) { struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr); - size_t bufSize; + pb_size bufSize; unsigned i; bufSize = mgr->minBufSize; @@ -527,13 +527,13 @@ pb_slab_range_manager_destroy(struct pb_manager *_mgr) struct pb_manager * pb_slab_range_manager_create(struct pb_manager *provider, - size_t minBufSize, - size_t maxBufSize, - size_t slabSize, + pb_size minBufSize, + pb_size maxBufSize, + pb_size slabSize, const struct pb_desc *desc) { struct pb_slab_range_manager *mgr; - size_t bufSize; + pb_size bufSize; unsigned i; if(!provider) diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc.c b/src/gallium/auxiliary/rtasm/rtasm_ppc.c index e3586482db4..ef4b306cb67 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc.c @@ -168,7 +168,7 @@ ppc_allocate_register(struct ppc_function *p) { unsigned i; for (i = 0; i < PPC_NUM_REGS; i++) { - const uint64_t mask = 1 << i; + const uint32_t mask = 1 << i; if ((p->reg_used & mask) == 0) { p->reg_used |= mask; return i; @@ -200,7 +200,7 @@ ppc_allocate_fp_register(struct ppc_function *p) { unsigned i; for (i = 0; i < PPC_NUM_FP_REGS; i++) { - const uint64_t mask = 1 << i; + const uint32_t mask = 1 << i; if ((p->fp_used & mask) == 0) { p->fp_used |= mask; return i; @@ -232,7 +232,7 @@ ppc_allocate_vec_register(struct ppc_function *p) { unsigned i; for (i = 0; i < PPC_NUM_VEC_REGS; i++) { - const uint64_t mask = 1 << i; + const uint32_t mask = 1 << i; if ((p->vec_used & mask) == 0) { p->vec_used |= mask; return i; diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 57fcf6de2ab..1acf3c373eb 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -993,6 +993,15 @@ void sse_pmovmskb( struct x86_function *p, emit_modrm(p, dst, src); } +void sse_movmskps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src) +{ + DUMP_RR( dst, src ); + emit_2ub(p, X86_TWOB, 0x50); + emit_modrm(p, dst, src); +} + /*********************************************************************** * SSE2 instructions */ diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 1b5eaaca850..731a6517968 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -223,6 +223,7 @@ void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movmskps( struct x86_function *p, struct x86_reg dst, struct x86_reg src); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index b4900e8dbaa..5f0a580b096 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -16,6 +16,7 @@ C_SOURCES = \ tgsi_sse2.c \ tgsi_text.c \ tgsi_transform.c \ + tgsi_ureg.c \ tgsi_util.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 8200cce42f5..b6bc2924f06 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -16,6 +16,7 @@ tgsi = env.ConvenienceLibrary( 'tgsi_sse2.c', 'tgsi_text.c', 'tgsi_transform.c', + 'tgsi_ureg.c', 'tgsi_util.c', ]) diff --git a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt index a3f4947c734..e63813d6936 100644 --- a/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt +++ b/src/gallium/auxiliary/tgsi/tgsi-instruction-set.txt @@ -665,9 +665,18 @@ TGSI Instruction Specification TBD -1.9.8 LOOP - Loop +1.9.8 BGNFOR - Begin a For-Loop - TBD + dst.x = floor(src.x) + dst.y = floor(src.y) + dst.z = floor(src.z) + + if (dst.y <= 0) + pc = [matching ENDFOR] + 1 + endif + + Note: The destination must be a loop register. + The source must be a constant register. 1.9.9 REP - Repeat @@ -685,9 +694,16 @@ TGSI Instruction Specification TBD -1.9.12 ENDLOOP - End Loop +1.9.12 ENDFOR - End a For-Loop - TBD + dst.x = dst.x + dst.z + dst.y = dst.y - 1.0 + + if (dst.y > 0) + pc = [matching BGNFOR instruction] + 1 + endif + + Note: The destination must be a loop register. 1.9.13 ENDREP - End Repeat @@ -840,7 +856,7 @@ TGSI Instruction Specification ---------- -1.13.1 BGNLOOP2 - Begin Loop +1.13.1 BGNLOOP - Begin a Loop TBD @@ -850,7 +866,7 @@ TGSI Instruction Specification TBD -1.13.3 ENDLOOP2 - End Loop +1.13.3 ENDLOOP - End a Loop TBD @@ -1015,12 +1031,12 @@ TGSI Instruction Specification 1.18.1 EXPP - Approximate Exponential Base 2 - Alias for EXP. + Use EXP. See also 1.19.3. 1.18.2 LOGP - Logarithm Base 2 - Alias for LG2. + Use LOG. See also 1.19.4. 1.19 vs_2_0 @@ -1037,6 +1053,16 @@ TGSI Instruction Specification Alias for ARR. +1.19.3 EXPP - Approximate Exponential Base 2 + + Use EX2. + + +1.19.4 LOGP - Logarithm Base 2 + + Use LG2. + + 2 Explanation of symbols used ============================== diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index a1891a140ac..e0cfc54420e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -139,8 +139,8 @@ tgsi_build_declaration( { struct tgsi_declaration declaration; - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + assert( file < TGSI_FILE_COUNT ); + assert( interpolate < TGSI_INTERPOLATE_COUNT ); declaration = tgsi_default_declaration(); declaration.File = file; @@ -335,7 +335,10 @@ tgsi_default_full_immediate( void ) struct tgsi_full_immediate fullimm; fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; + fullimm.u[0].Float = 0.0f; + fullimm.u[1].Float = 0.0f; + fullimm.u[2].Float = 0.0f; + fullimm.u[3].Float = 0.0f; return fullimm; } @@ -352,19 +355,19 @@ immediate_grow( header_bodysize_grow( header ); } -struct tgsi_immediate_float32 +union tgsi_immediate_data tgsi_build_immediate_float32( float value, struct tgsi_immediate *immediate, struct tgsi_header *header ) { - struct tgsi_immediate_float32 immediate_float32; + union tgsi_immediate_data immediate_data; - immediate_float32.Float = value; + immediate_data.Float = value; immediate_grow( immediate, header ); - return immediate_float32; + return immediate_data; } unsigned @@ -384,16 +387,18 @@ tgsi_build_full_immediate( *immediate = tgsi_build_immediate( header ); + assert( full_imm->Immediate.NrTokens <= 4 + 1 ); + for( i = 0; i < full_imm->Immediate.NrTokens - 1; i++ ) { - struct tgsi_immediate_float32 *if32; + union tgsi_immediate_data *data; if( maxsize <= size ) return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + data = (union tgsi_immediate_data *) &tokens[size]; size++; - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, + *data = tgsi_build_immediate_float32( + full_imm->u[i].Float, immediate, header ); } @@ -477,6 +482,8 @@ tgsi_default_full_instruction( void ) full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); } + full_instruction.Flags = 0x0; + return full_instruction; } @@ -584,6 +591,7 @@ tgsi_build_full_instruction( *dst_register = tgsi_build_dst_register( reg->DstRegister.File, reg->DstRegister.WriteMask, + reg->DstRegister.Indirect, reg->DstRegister.Index, instruction, header ); @@ -631,6 +639,28 @@ tgsi_build_full_instruction( header ); prev_token = (struct tgsi_token *) dst_register_ext_modulate; } + + if( reg->DstRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->DstRegisterInd.File, + reg->DstRegisterInd.SwizzleX, + reg->DstRegisterInd.SwizzleY, + reg->DstRegisterInd.SwizzleZ, + reg->DstRegisterInd.SwizzleW, + reg->DstRegisterInd.Negate, + reg->DstRegisterInd.Indirect, + reg->DstRegisterInd.Dimension, + reg->DstRegisterInd.Index, + instruction, + header ); + } } for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { @@ -973,7 +1003,7 @@ tgsi_build_src_register( { struct tgsi_src_register src_register; - assert( file <= TGSI_FILE_IMMEDIATE ); + assert( file < TGSI_FILE_COUNT ); assert( swizzle_x <= TGSI_SWIZZLE_W ); assert( swizzle_y <= TGSI_SWIZZLE_W ); assert( swizzle_z <= TGSI_SWIZZLE_W ); @@ -1194,13 +1224,14 @@ struct tgsi_dst_register tgsi_build_dst_register( unsigned file, unsigned mask, + unsigned indirect, int index, struct tgsi_instruction *instruction, struct tgsi_header *header ) { struct tgsi_dst_register dst_register; - assert( file <= TGSI_FILE_IMMEDIATE ); + assert( file < TGSI_FILE_COUNT ); assert( mask <= TGSI_WRITEMASK_XYZW ); assert( index >= -32768 && index <= 32767 ); @@ -1208,6 +1239,7 @@ tgsi_build_dst_register( dst_register.File = file; dst_register.WriteMask = mask; dst_register.Index = index; + dst_register.Indirect = indirect; instruction_grow( instruction, header ); @@ -1220,6 +1252,7 @@ tgsi_default_full_dst_register( void ) struct tgsi_full_dst_register full_dst_register; full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterInd = tgsi_default_src_register(); full_dst_register.DstRegisterExtConcode = tgsi_default_dst_register_ext_concode(); full_dst_register.DstRegisterExtModulate = diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h index 0fd6fabd83d..17d977b0597 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -119,7 +119,7 @@ tgsi_build_immediate( struct tgsi_full_immediate tgsi_default_full_immediate( void ); -struct tgsi_immediate_float32 +union tgsi_immediate_data tgsi_build_immediate_float32( float value, struct tgsi_immediate *immediate, @@ -289,6 +289,7 @@ struct tgsi_dst_register tgsi_build_dst_register( unsigned file, unsigned mask, + unsigned indirect, int index, struct tgsi_instruction *instruction, struct tgsi_header *header ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index a784b7cc3c1..05b07a3a73e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -27,16 +27,25 @@ #include "util/u_debug.h" #include "util/u_string.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "tgsi_dump.h" #include "tgsi_info.h" #include "tgsi_iterate.h" + +/** Number of spaces to indent for IF/LOOP/etc */ +static const int indent_spaces = 3; + + struct dump_ctx { struct tgsi_iterate_context iter; uint instno; + uint indentation; + void (*printf)(struct dump_ctx *ctx, const char *format, ...); }; @@ -80,7 +89,7 @@ static const char *processor_type_names[] = "GEOM" }; -static const char *file_names[] = +static const char *file_names[TGSI_FILE_COUNT] = { "NULL", "CONST", @@ -89,7 +98,8 @@ static const char *file_names[] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static const char *interpolate_names[] = @@ -107,7 +117,8 @@ static const char *semantic_names[] = "FOG", "PSIZE", "GENERIC", - "NORMAL" + "NORMAL", + "FACE" }; static const char *immediate_type_names[] = @@ -223,6 +234,9 @@ iter_declaration( { struct dump_ctx *ctx = (struct dump_ctx *)iter; + assert(Elements(semantic_names) == TGSI_SEMANTIC_COUNT); + assert(Elements(interpolate_names) == TGSI_INTERPOLATE_COUNT); + TXT( "DCL " ); _dump_register( @@ -289,10 +303,12 @@ iter_immediate( ENM( imm->Immediate.DataType, immediate_type_names ); TXT( " { " ); + + assert( imm->Immediate.NrTokens <= 4 + 1 ); for (i = 0; i < imm->Immediate.NrTokens - 1; i++) { switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); + FLT( imm->u[i].Float ); break; default: assert( 0 ); @@ -319,6 +335,14 @@ tgsi_dump_immediate( iter_immediate( &ctx.iter, (struct tgsi_full_immediate *)imm ); } +static void +indent(struct dump_ctx *ctx) +{ + uint i; + for (i = 0; i < ctx->indentation; i++) + TXT(" "); +} + static boolean iter_instruction( struct tgsi_iterate_context *iter, @@ -332,6 +356,15 @@ iter_instruction( INSTID( instno ); TXT( ": " ); + + /* update indentation */ + if (inst->Instruction.Opcode == TGSI_OPCODE_ENDIF || + inst->Instruction.Opcode == TGSI_OPCODE_ENDFOR || + inst->Instruction.Opcode == TGSI_OPCODE_ENDLOOP) { + ctx->indentation -= indent_spaces; + } + indent(ctx); + TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic ); switch (inst->Instruction.Saturate) { @@ -354,11 +387,22 @@ iter_instruction( CHR( ',' ); CHR( ' ' ); - _dump_register( - ctx, - dst->DstRegister.File, - dst->DstRegister.Index, - dst->DstRegister.Index ); + if (dst->DstRegister.Indirect) { + _dump_register_ind( + ctx, + dst->DstRegister.File, + dst->DstRegister.Index, + dst->DstRegisterInd.File, + dst->DstRegisterInd.Index, + dst->DstRegisterInd.SwizzleX ); + } + else { + _dump_register( + ctx, + dst->DstRegister.File, + dst->DstRegister.Index, + dst->DstRegister.Index ); + } ENM( dst->DstRegisterExtModulate.Modulate, modulate_names ); _dump_writemask( ctx, dst->DstRegister.WriteMask ); @@ -453,14 +497,22 @@ iter_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_BGNLOOP: + case TGSI_OPCODE_ENDLOOP: case TGSI_OPCODE_CAL: TXT( " :" ); UID( inst->InstructionExtLabel.Label ); break; } + /* update indentation */ + if (inst->Instruction.Opcode == TGSI_OPCODE_IF || + inst->Instruction.Opcode == TGSI_OPCODE_ELSE || + inst->Instruction.Opcode == TGSI_OPCODE_BGNFOR || + inst->Instruction.Opcode == TGSI_OPCODE_BGNLOOP) { + ctx->indentation += indent_spaces; + } + EOL(); return TRUE; @@ -475,6 +527,7 @@ tgsi_dump_instruction( ctx.instno = instno; ctx.printf = dump_ctx_printf; + ctx.indentation = 0; iter_instruction( &ctx.iter, (struct tgsi_full_instruction *)inst ); } @@ -507,6 +560,7 @@ tgsi_dump( ctx.instno = 0; ctx.printf = dump_ctx_printf; + ctx.indentation = 0; tgsi_iterate_shader( tokens, &ctx.iter ); } @@ -516,7 +570,7 @@ struct str_dump_ctx struct dump_ctx base; char *str; char *ptr; - size_t left; + int left; }; static void @@ -525,13 +579,20 @@ str_dump_ctx_printf(struct dump_ctx *ctx, const char *format, ...) struct str_dump_ctx *sctx = (struct str_dump_ctx *)ctx; if(sctx->left > 1) { - size_t written; + int written; va_list ap; va_start(ap, format); written = util_vsnprintf(sctx->ptr, sctx->left, format, ap); va_end(ap); - sctx->ptr += written; - sctx->left -= written; + + /* Some complicated logic needed to handle the return value of + * vsnprintf: + */ + if (written > 0) { + written = MIN2(sctx->left, written); + sctx->ptr += written; + sctx->left -= written; + } } } @@ -552,11 +613,12 @@ tgsi_dump_str( ctx.base.instno = 0; ctx.base.printf = &str_dump_ctx_printf; + ctx.base.indentation = 0; ctx.str = str; ctx.str[0] = 0; ctx.ptr = str; - ctx.left = size; + ctx.left = (int)size; tgsi_iterate_shader( tokens, &ctx.base.iter ); } diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c index 3dc61c48ca3..4a9c02b1413 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -69,7 +69,7 @@ static const char *TGSI_TOKEN_TYPES[] = "TOKEN_TYPE_INSTRUCTION" }; -static const char *TGSI_FILES[] = +static const char *TGSI_FILES[TGSI_FILE_COUNT] = { "FILE_NULL", "FILE_CONSTANT", @@ -78,7 +78,8 @@ static const char *TGSI_FILES[] = "FILE_TEMPORARY", "FILE_SAMPLER", "FILE_ADDRESS", - "FILE_IMMEDIATE" + "FILE_IMMEDIATE", + "FILE_LOOP" }; static const char *TGSI_INTERPOLATES[] = @@ -283,12 +284,13 @@ dump_immediate_verbose( UIX( imm->Immediate.Padding ); } + assert( imm->Immediate.NrTokens <= 4 + 1 ); for( i = 0; i < imm->Immediate.NrTokens - 1; i++ ) { EOL(); switch( imm->Immediate.DataType ) { case TGSI_IMM_FLOAT32: TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); + FLT( imm->u[i].Float ); break; default: diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index aba7a3f9374..711e86d6edf 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -62,6 +62,9 @@ #define FAST_MATH 1 +/** for tgsi_full_instruction::Flags */ +#define SOA_DEPENDENCY_FLAG 0x1 + #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 #define TILE_BOTTOM_LEFT 2 @@ -182,7 +185,7 @@ print_temp(const struct tgsi_exec_machine *mach, uint index) * MOV t3, t2; * The second instruction will have the wrong value for t0 if executed as-is. */ -static boolean +boolean tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst) { uint i, chan; @@ -301,14 +304,14 @@ tgsi_exec_machine_bind_shader( case TGSI_TOKEN_TYPE_IMMEDIATE: { uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1; - assert( size % 4 == 0 ); - assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); + assert( size <= 4 ); + assert( mach->ImmLimit + 1 <= TGSI_EXEC_NUM_IMMEDIATES ); for( i = 0; i < size; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + mach->Imms[mach->ImmLimit][i] = + parse.FullToken.FullImmediate.u[i].Float; } - mach->ImmLimit += size / 4; + mach->ImmLimit += 1; } break; @@ -328,19 +331,24 @@ tgsi_exec_machine_bind_shader( * sizeof(struct tgsi_full_instruction)); maxInstructions += 10; } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); -#if 0 if (tgsi_check_soa_dependencies(&parse.FullToken.FullInstruction)) { - debug_printf("SOA dependency in instruction:\n"); - tgsi_dump_instruction(&parse.FullToken.FullInstruction, - numInstructions); + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + parse.FullToken.FullInstruction.Flags = SOA_DEPENDENCY_FLAG; + /* XXX we only handle SOA dependencies properly for MOV/SWZ + * at this time! + */ + if (opcode != TGSI_OPCODE_MOV && opcode != TGSI_OPCODE_SWZ) { + debug_printf("Warning: SOA dependency in instruction" + " is not handled:\n"); + tgsi_dump_instruction(&parse.FullToken.FullInstruction, + numInstructions); + } } -#else - (void) tgsi_check_soa_dependencies; -#endif + + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); numInstructions++; break; @@ -365,13 +373,18 @@ tgsi_exec_machine_bind_shader( } -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach ) +struct tgsi_exec_machine * +tgsi_exec_machine_create( void ) { + struct tgsi_exec_machine *mach; uint i; - mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach = align_malloc( sizeof *mach, 16 ); + if (!mach) + goto fail; + + memset(mach, 0, sizeof(*mach)); + mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; /* Setup constants. */ @@ -393,22 +406,24 @@ tgsi_exec_machine_init( (void) print_chan; (void) print_temp; #endif + + return mach; + +fail: + align_free(mach); + return NULL; } void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) +tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach) { - if (mach->Instructions) { + if (mach) { FREE(mach->Instructions); - mach->Instructions = NULL; - mach->NumInstructions = 0; - } - if (mach->Declarations) { FREE(mach->Declarations); - mach->Declarations = NULL; - mach->NumDeclarations = 0; } + + align_free(mach); } @@ -1395,28 +1410,69 @@ store_dest( union tgsi_exec_channel null; union tgsi_exec_channel *dst; uint execmask = mach->ExecMask; + int offset = 0; /* indirection offset */ + int index; #ifdef DEBUG check_inf_or_nan(chan); #endif + /* There is an extra source register that indirectly subscripts + * a register file. The direct index now becomes an offset + * that is being added to the indirect register. + * + * file[ind[2].x+1], + * where: + * ind = DstRegisterInd.File + * [2] = DstRegisterInd.Index + * .x = DstRegisterInd.SwizzleX + */ + if (reg->DstRegister.Indirect) { + union tgsi_exec_channel index; + union tgsi_exec_channel indir_index; + uint swizzle; + + /* which address register (always zero for now) */ + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->DstRegisterInd.Index; + + /* get current value of address register[swizzle] */ + swizzle = tgsi_util_get_src_register_swizzle( ®->DstRegisterInd, CHAN_X ); + + /* fetch values from the address/indirection register */ + fetch_src_file_channel( + mach, + reg->DstRegisterInd.File, + swizzle, + &index, + &indir_index ); + + /* save indirection offset */ + offset = (int) indir_index.f[0]; + } + switch (reg->DstRegister.File) { case TGSI_FILE_NULL: dst = &null; break; case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] - + reg->DstRegister.Index].xyzw[chan_index]; + index = mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index; + dst = &mach->Outputs[offset + index].xyzw[chan_index]; break; case TGSI_FILE_TEMPORARY: - assert( reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS ); - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + index = reg->DstRegister.Index; + assert( index < TGSI_EXEC_NUM_TEMPS ); + dst = &mach->Temps[offset + index].xyzw[chan_index]; break; case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + index = reg->DstRegister.Index; + dst = &mach->Addrs[index].xyzw[chan_index]; break; default: @@ -1966,8 +2022,7 @@ exec_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ + case TGSI_OPCODE_FLR: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); micro_flr( &r[0], &r[0] ); @@ -1977,9 +2032,23 @@ exec_instruction( case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); + if (inst->Flags & SOA_DEPENDENCY_FLAG) { + /* Do all fetches into temp regs, then do all stores to avoid + * intermediate/accidental clobbering. This could be done all the + * time for MOV but for other instructions we'll need more temps... + */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[chan_index], 0, chan_index ); + } + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } + } + else { + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } } break; @@ -2236,8 +2305,7 @@ exec_instruction( } break; - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ + case TGSI_OPCODE_LRP: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); @@ -2271,8 +2339,7 @@ exec_instruction( } break; - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ + case TGSI_OPCODE_DP2A: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); micro_mul( &r[0], &r[0], &r[1] ); @@ -2290,18 +2357,7 @@ exec_instruction( } break; - case TGSI_OPCODE_INDEX: - /* XXX: considered for removal */ - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - /* XXX: considered for removal */ - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ + case TGSI_OPCODE_FRC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); micro_frc( &r[0], &r[0] ); @@ -2329,8 +2385,7 @@ exec_instruction( } break; - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ + case TGSI_OPCODE_EX2: FETCH(&r[0], 0, CHAN_X); #if FAST_MATH @@ -2344,8 +2399,7 @@ exec_instruction( } break; - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ + case TGSI_OPCODE_LG2: FETCH( &r[0], 0, CHAN_X ); micro_lg2( &r[0], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -2353,8 +2407,7 @@ exec_instruction( } break; - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ + case TGSI_OPCODE_POW: FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); @@ -2365,8 +2418,7 @@ exec_instruction( } break; - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ + case TGSI_OPCODE_XPD: FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); @@ -2408,11 +2460,6 @@ exec_instruction( } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - /* XXX: considered for removal */ - assert (0); - break; - case TGSI_OPCODE_ABS: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); @@ -3056,9 +3103,9 @@ exec_instruction( mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; @@ -3066,9 +3113,9 @@ exec_instruction( mach->ContStack[mach->ContStackTop++] = mach->ContMask; break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_ENDLOOP: /* Restore ContMask, but don't pop */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index da22baad3ef..fd9ef6f35df 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -29,6 +29,7 @@ #define TGSI_EXEC_H #include "pipe/p_compiler.h" +#include "pipe/p_state.h" #if defined __cplusplus extern "C" { @@ -94,7 +95,6 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_TEMPS 128 -#define TGSI_EXEC_NUM_TEMP_EXTRAS 6 #define TGSI_EXEC_NUM_IMMEDIATES 256 /* @@ -162,9 +162,14 @@ struct tgsi_exec_labels #define TGSI_EXEC_MASK_I (TGSI_EXEC_NUM_TEMPS + 3) #define TGSI_EXEC_MASK_C 2 +/* 4 register buffer for various purposes */ #define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4) +#define TGSI_EXEC_NUM_TEMP_R 4 + +#define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 8) +#define TGSI_EXEC_NUM_ADDRS 1 +#define TGSI_EXEC_NUM_TEMP_EXTRAS 9 -#define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 5) #define TGSI_EXEC_MAX_COND_NESTING 20 @@ -187,24 +192,21 @@ struct tgsi_exec_labels struct tgsi_exec_machine { /* Total = program temporaries + internal temporaries - * + 1 padding to align to 16 bytes */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + - TGSI_EXEC_NUM_TEMP_EXTRAS + 1]; + struct tgsi_exec_vector Temps[TGSI_EXEC_NUM_TEMPS + + TGSI_EXEC_NUM_TEMP_EXTRAS]; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + + struct tgsi_exec_vector Inputs[PIPE_MAX_ATTRIBS]; + struct tgsi_exec_vector Outputs[PIPE_MAX_ATTRIBS]; - /* - * This will point to _Temps after aligning to 16B boundary. - */ - struct tgsi_exec_vector *Temps; struct tgsi_exec_vector *Addrs; struct tgsi_sampler **Samplers; - float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; unsigned ImmLimit; const float (*Consts)[4]; - struct tgsi_exec_vector *Inputs; - struct tgsi_exec_vector *Outputs; const struct tgsi_token *Tokens; /**< Declarations, instructions */ unsigned Processor; /**< TGSI_PROCESSOR_x */ @@ -251,9 +253,11 @@ struct tgsi_exec_machine struct tgsi_exec_labels Labels; }; +struct tgsi_exec_machine * +tgsi_exec_machine_create( void ); + void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach ); +tgsi_exec_machine_destroy(struct tgsi_exec_machine *mach); void @@ -272,6 +276,10 @@ void tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); +boolean +tgsi_check_soa_dependencies(const struct tgsi_full_instruction *inst); + + static INLINE void tgsi_set_kill_mask(struct tgsi_exec_machine *mach, unsigned mask) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c index 37f2b66d1f6..ccf4b205ffb 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.c +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -26,136 +26,156 @@ **************************************************************************/ #include "util/u_debug.h" +#include "util/u_memory.h" #include "tgsi_info.h" static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = { - { 1, 1, 0, 0, "ARL", NULL, NULL }, - { 1, 1, 0, 0, "MOV", NULL, NULL }, - { 1, 1, 0, 0, "LIT", NULL, NULL }, - { 1, 1, 0, 0, "RCP", "RECIP", NULL }, - { 1, 1, 0, 0, "RSQ", "RECIPSQRT", NULL }, - { 1, 1, 0, 0, "EXP", "EXPP", NULL }, - { 1, 1, 0, 0, "LOG", NULL, NULL }, - { 1, 2, 0, 0, "MUL", NULL, NULL }, - { 1, 2, 0, 0, "ADD", NULL, NULL }, - { 1, 2, 0, 0, "DP3", "DOT3", NULL }, - { 1, 2, 0, 0, "DP4", "DOT4", NULL }, - { 1, 2, 0, 0, "DST", NULL, NULL }, - { 1, 2, 0, 0, "MIN", NULL, NULL }, - { 1, 2, 0, 0, "MAX", NULL, NULL }, - { 1, 2, 0, 0, "SLT", "SETLT", NULL }, - { 1, 2, 0, 0, "SGE", "SETGE", NULL }, - { 1, 3, 0, 0, "MAD", "MADD", NULL }, - { 1, 2, 0, 0, "SUB", NULL, NULL }, - { 1, 3, 0, 0, "LRP", "LERP", NULL }, - { 1, 3, 0, 0, "CND", NULL, NULL }, - { 1, 3, 0, 0, "CND0", NULL, NULL }, - { 1, 3, 0, 0, "DP2A", "DP2ADD", "DOT2ADD" }, - { 1, 2, 0, 0, "INDEX", NULL, NULL }, - { 1, 1, 0, 0, "NEGATE", NULL, NULL }, - { 1, 1, 0, 0, "FRC", "FRAC", NULL }, - { 1, 3, 0, 0, "CLAMP", NULL, NULL }, - { 1, 1, 0, 0, "FLR", "FLOOR", NULL }, - { 1, 1, 0, 0, "ROUND", NULL, NULL }, - { 1, 1, 0, 0, "EX2", "EXPBASE2", NULL }, - { 1, 1, 0, 0, "LG2", "LOGBASE2", "LOGP" }, - { 1, 2, 0, 0, "POW", "POWER", NULL }, - { 1, 2, 0, 0, "XPD", "CRS", "CROSSPRODUCT" }, - { 1, 2, 0, 0, "M4X4", "MULTIPLYMATRIX", NULL }, - { 1, 1, 0, 0, "ABS", NULL, NULL }, - { 1, 1, 0, 0, "RCC", NULL, NULL }, - { 1, 2, 0, 0, "DPH", NULL, NULL }, - { 1, 1, 0, 0, "COS", NULL, NULL }, - { 1, 1, 0, 0, "DDX", "DSX", NULL }, - { 1, 1, 0, 0, "DDY", "DSY", NULL }, - { 0, 0, 0, 0, "KILP", NULL, NULL }, - { 1, 1, 0, 0, "PK2H", NULL, NULL }, - { 1, 1, 0, 0, "PK2US", NULL, NULL }, - { 1, 1, 0, 0, "PK4B", NULL, NULL }, - { 1, 1, 0, 0, "PK4UB", NULL, NULL }, - { 1, 2, 0, 0, "RFL", NULL, NULL }, - { 1, 2, 0, 0, "SEQ", NULL, NULL }, - { 1, 2, 0, 0, "SFL", NULL, NULL }, - { 1, 2, 0, 0, "SGT", NULL, NULL }, - { 1, 1, 0, 0, "SIN", NULL, NULL }, - { 1, 2, 0, 0, "SLE", NULL, NULL }, - { 1, 2, 0, 0, "SNE", NULL, NULL }, - { 1, 2, 0, 0, "STR", NULL, NULL }, - { 1, 2, 1, 0, "TEX", "TEXLD", NULL }, - { 1, 4, 1, 0, "TXD", "TEXLDD", NULL }, - { 1, 2, 1, 0, "TXP", NULL, NULL }, - { 1, 1, 0, 0, "UP2H", NULL, NULL }, - { 1, 1, 0, 0, "UP2US", NULL, NULL }, - { 1, 1, 0, 0, "UP4B", NULL, NULL }, - { 1, 1, 0, 0, "UP4UB", NULL, NULL }, - { 1, 3, 0, 0, "X2D", NULL, NULL }, - { 1, 1, 0, 0, "ARA", NULL, NULL }, - { 1, 1, 0, 0, "ARR", "MOVA", NULL }, - { 0, 1, 0, 0, "BRA", NULL, NULL }, - { 0, 0, 0, 1, "CAL", "CALL", NULL }, - { 0, 0, 0, 0, "RET", NULL, NULL }, - { 1, 1, 0, 0, "SGN", "SSG", NULL }, - { 1, 3, 0, 0, "CMP", NULL, NULL }, - { 1, 1, 0, 0, "SCS", "SINCOS", NULL }, - { 1, 2, 1, 0, "TXB", "TEXLDB", NULL }, - { 1, 1, 0, 0, "NRM", NULL, NULL }, - { 1, 2, 0, 0, "DIV", NULL, NULL }, - { 1, 2, 0, 0, "DP2", NULL, NULL }, - { 1, 2, 1, 0, "TXL", NULL, NULL }, - { 0, 0, 0, 0, "BRK", "BREAK", NULL }, - { 0, 1, 0, 1, "IF", NULL, NULL }, - { 0, 0, 0, 0, "LOOP", NULL, NULL }, - { 0, 1, 0, 0, "REP", NULL, NULL }, - { 0, 0, 0, 1, "ELSE", NULL, NULL }, - { 0, 0, 0, 0, "ENDIF", NULL, NULL }, - { 0, 0, 0, 0, "ENDLOOP", NULL, NULL }, - { 0, 0, 0, 0, "ENDREP", NULL, NULL }, - { 0, 1, 0, 0, "PUSHA", NULL, NULL }, - { 1, 0, 0, 0, "POPA", NULL, NULL }, - { 1, 1, 0, 0, "CEIL", NULL, NULL }, - { 1, 1, 0, 0, "I2F", NULL, NULL }, - { 1, 1, 0, 0, "NOT", NULL, NULL }, - { 1, 1, 0, 0, "INT", "TRUNC", NULL }, - { 1, 2, 0, 0, "SHL", NULL, NULL }, - { 1, 2, 0, 0, "SHR", NULL, NULL }, - { 1, 2, 0, 0, "AND", NULL, NULL }, - { 1, 2, 0, 0, "OR", NULL, NULL }, - { 1, 2, 0, 0, "MOD", NULL, NULL }, - { 1, 2, 0, 0, "XOR", NULL, NULL }, - { 1, 3, 0, 0, "SAD", NULL, NULL }, - { 1, 2, 1, 0, "TXF", NULL, NULL }, - { 1, 2, 1, 0, "TXQ", NULL, NULL }, - { 0, 0, 0, 0, "CONT", NULL, NULL }, - { 0, 0, 0, 0, "EMIT", NULL, NULL }, - { 0, 0, 0, 0, "ENDPRIM", NULL, NULL }, - { 0, 0, 0, 1, "BGNLOOP2", NULL, NULL }, - { 0, 0, 0, 0, "BGNSUB", NULL, NULL }, - { 0, 0, 0, 1, "ENDLOOP2", NULL, NULL }, - { 0, 0, 0, 0, "ENDSUB", NULL, NULL }, - { 1, 1, 0, 0, "NOISE1", NULL, NULL }, - { 1, 1, 0, 0, "NOISE2", NULL, NULL }, - { 1, 1, 0, 0, "NOISE3", NULL, NULL }, - { 1, 1, 0, 0, "NOISE4", NULL, NULL }, - { 0, 0, 0, 0, "NOP", NULL, NULL }, - { 1, 2, 0, 0, "M4X3", NULL, NULL }, - { 1, 2, 0, 0, "M3X4", NULL, NULL }, - { 1, 2, 0, 0, "M3X3", NULL, NULL }, - { 1, 2, 0, 0, "M3X2", NULL, NULL }, - { 1, 1, 0, 0, "NRM4", NULL, NULL }, - { 0, 1, 0, 0, "CALLNZ", NULL, NULL }, - { 0, 1, 0, 0, "IFC", NULL, NULL }, - { 0, 1, 0, 0, "BREAKC", NULL, NULL }, - { 0, 1, 0, 0, "KIL", "TEXKILL", NULL }, - { 0, 0, 0, 0, "END", NULL, NULL }, - { 1, 1, 0, 0, "SWZ", NULL, NULL } + { 1, 1, 0, 0, "ARL", TGSI_OPCODE_ARL }, + { 1, 1, 0, 0, "MOV", TGSI_OPCODE_MOV }, + { 1, 1, 0, 0, "LIT", TGSI_OPCODE_LIT }, + { 1, 1, 0, 0, "RCP", TGSI_OPCODE_RCP }, + { 1, 1, 0, 0, "RSQ", TGSI_OPCODE_RSQ }, + { 1, 1, 0, 0, "EXP", TGSI_OPCODE_EXP }, + { 1, 1, 0, 0, "LOG", TGSI_OPCODE_LOG }, + { 1, 2, 0, 0, "MUL", TGSI_OPCODE_MUL }, + { 1, 2, 0, 0, "ADD", TGSI_OPCODE_ADD }, + { 1, 2, 0, 0, "DP3", TGSI_OPCODE_DP3 }, + { 1, 2, 0, 0, "DP4", TGSI_OPCODE_DP4 }, + { 1, 2, 0, 0, "DST", TGSI_OPCODE_DST }, + { 1, 2, 0, 0, "MIN", TGSI_OPCODE_MIN }, + { 1, 2, 0, 0, "MAX", TGSI_OPCODE_MAX }, + { 1, 2, 0, 0, "SLT", TGSI_OPCODE_SLT }, + { 1, 2, 0, 0, "SGE", TGSI_OPCODE_SGE }, + { 1, 3, 0, 0, "MAD", TGSI_OPCODE_MAD }, + { 1, 2, 0, 0, "SUB", TGSI_OPCODE_SUB }, + { 1, 3, 0, 0, "LRP", TGSI_OPCODE_LRP }, + { 1, 3, 0, 0, "CND", TGSI_OPCODE_CND }, + { 1, 3, 0, 0, "CND0", TGSI_OPCODE_CND0 }, + { 1, 3, 0, 0, "DP2A", TGSI_OPCODE_DP2A }, + { 0, 0, 0, 0, "", 22 }, /* removed */ + { 0, 0, 0, 0, "", 23 }, /* removed */ + { 1, 1, 0, 0, "FRC", TGSI_OPCODE_FRC }, + { 1, 3, 0, 0, "CLAMP", TGSI_OPCODE_CLAMP }, + { 1, 1, 0, 0, "FLR", TGSI_OPCODE_FLR }, + { 1, 1, 0, 0, "ROUND", TGSI_OPCODE_ROUND }, + { 1, 1, 0, 0, "EX2", TGSI_OPCODE_EX2 }, + { 1, 1, 0, 0, "LG2", TGSI_OPCODE_LG2 }, + { 1, 2, 0, 0, "POW", TGSI_OPCODE_POW }, + { 1, 2, 0, 0, "XPD", TGSI_OPCODE_XPD }, + { 0, 0, 0, 0, "", 32 }, /* removed */ + { 1, 1, 0, 0, "ABS", TGSI_OPCODE_ABS }, + { 1, 1, 0, 0, "RCC", TGSI_OPCODE_RCC }, + { 1, 2, 0, 0, "DPH", TGSI_OPCODE_DPH }, + { 1, 1, 0, 0, "COS", TGSI_OPCODE_COS }, + { 1, 1, 0, 0, "DDX", TGSI_OPCODE_DDX }, + { 1, 1, 0, 0, "DDY", TGSI_OPCODE_DDY }, + { 0, 0, 0, 0, "KILP", TGSI_OPCODE_KILP }, + { 1, 1, 0, 0, "PK2H", TGSI_OPCODE_PK2H }, + { 1, 1, 0, 0, "PK2US", TGSI_OPCODE_PK2US }, + { 1, 1, 0, 0, "PK4B", TGSI_OPCODE_PK4B }, + { 1, 1, 0, 0, "PK4UB", TGSI_OPCODE_PK4UB }, + { 1, 2, 0, 0, "RFL", TGSI_OPCODE_RFL }, + { 1, 2, 0, 0, "SEQ", TGSI_OPCODE_SEQ }, + { 1, 2, 0, 0, "SFL", TGSI_OPCODE_SFL }, + { 1, 2, 0, 0, "SGT", TGSI_OPCODE_SGT }, + { 1, 1, 0, 0, "SIN", TGSI_OPCODE_SIN }, + { 1, 2, 0, 0, "SLE", TGSI_OPCODE_SLE }, + { 1, 2, 0, 0, "SNE", TGSI_OPCODE_SNE }, + { 1, 2, 0, 0, "STR", TGSI_OPCODE_STR }, + { 1, 2, 1, 0, "TEX", TGSI_OPCODE_TEX }, + { 1, 4, 1, 0, "TXD", TGSI_OPCODE_TXD }, + { 1, 2, 1, 0, "TXP", TGSI_OPCODE_TXP }, + { 1, 1, 0, 0, "UP2H", TGSI_OPCODE_UP2H }, + { 1, 1, 0, 0, "UP2US", TGSI_OPCODE_UP2US }, + { 1, 1, 0, 0, "UP4B", TGSI_OPCODE_UP4B }, + { 1, 1, 0, 0, "UP4UB", TGSI_OPCODE_UP4UB }, + { 1, 3, 0, 0, "X2D", TGSI_OPCODE_X2D }, + { 1, 1, 0, 0, "ARA", TGSI_OPCODE_ARA }, + { 1, 1, 0, 0, "ARR", TGSI_OPCODE_ARR }, + { 0, 1, 0, 0, "BRA", TGSI_OPCODE_BRA }, + { 0, 0, 0, 1, "CAL", TGSI_OPCODE_CAL }, + { 0, 0, 0, 0, "RET", TGSI_OPCODE_RET }, + { 1, 1, 0, 0, "SSG", TGSI_OPCODE_SSG }, + { 1, 3, 0, 0, "CMP", TGSI_OPCODE_CMP }, + { 1, 1, 0, 0, "SCS", TGSI_OPCODE_SCS }, + { 1, 2, 1, 0, "TXB", TGSI_OPCODE_TXB }, + { 1, 1, 0, 0, "NRM", TGSI_OPCODE_NRM }, + { 1, 2, 0, 0, "DIV", TGSI_OPCODE_DIV }, + { 1, 2, 0, 0, "DP2", TGSI_OPCODE_DP2 }, + { 1, 2, 1, 0, "TXL", TGSI_OPCODE_TXL }, + { 0, 0, 0, 0, "BRK", TGSI_OPCODE_BRK }, + { 0, 1, 0, 1, "IF", TGSI_OPCODE_IF }, + { 1, 1, 0, 0, "BGNFOR", TGSI_OPCODE_BGNFOR }, + { 0, 1, 0, 0, "REP", TGSI_OPCODE_REP }, + { 0, 0, 0, 1, "ELSE", TGSI_OPCODE_ELSE }, + { 0, 0, 0, 0, "ENDIF", TGSI_OPCODE_ENDIF }, + { 1, 0, 0, 0, "ENDFOR", TGSI_OPCODE_ENDFOR }, + { 0, 0, 0, 0, "ENDREP", TGSI_OPCODE_ENDREP }, + { 0, 1, 0, 0, "PUSHA", TGSI_OPCODE_PUSHA }, + { 1, 0, 0, 0, "POPA", TGSI_OPCODE_POPA }, + { 1, 1, 0, 0, "CEIL", TGSI_OPCODE_CEIL }, + { 1, 1, 0, 0, "I2F", TGSI_OPCODE_I2F }, + { 1, 1, 0, 0, "NOT", TGSI_OPCODE_NOT }, + { 1, 1, 0, 0, "TRUNC", TGSI_OPCODE_TRUNC }, + { 1, 2, 0, 0, "SHL", TGSI_OPCODE_SHL }, + { 1, 2, 0, 0, "SHR", TGSI_OPCODE_SHR }, + { 1, 2, 0, 0, "AND", TGSI_OPCODE_AND }, + { 1, 2, 0, 0, "OR", TGSI_OPCODE_OR }, + { 1, 2, 0, 0, "MOD", TGSI_OPCODE_MOD }, + { 1, 2, 0, 0, "XOR", TGSI_OPCODE_XOR }, + { 1, 3, 0, 0, "SAD", TGSI_OPCODE_SAD }, + { 1, 2, 1, 0, "TXF", TGSI_OPCODE_TXF }, + { 1, 2, 1, 0, "TXQ", TGSI_OPCODE_TXQ }, + { 0, 0, 0, 0, "CONT", TGSI_OPCODE_CONT }, + { 0, 0, 0, 0, "EMIT", TGSI_OPCODE_EMIT }, + { 0, 0, 0, 0, "ENDPRIM", TGSI_OPCODE_ENDPRIM }, + { 0, 0, 0, 1, "BGNLOOP", TGSI_OPCODE_BGNLOOP }, + { 0, 0, 0, 0, "BGNSUB", TGSI_OPCODE_BGNSUB }, + { 0, 0, 0, 1, "ENDLOOP", TGSI_OPCODE_ENDLOOP }, + { 0, 0, 0, 0, "ENDSUB", TGSI_OPCODE_ENDSUB }, + { 1, 1, 0, 0, "NOISE1", TGSI_OPCODE_NOISE1 }, + { 1, 1, 0, 0, "NOISE2", TGSI_OPCODE_NOISE2 }, + { 1, 1, 0, 0, "NOISE3", TGSI_OPCODE_NOISE3 }, + { 1, 1, 0, 0, "NOISE4", TGSI_OPCODE_NOISE4 }, + { 0, 0, 0, 0, "NOP", TGSI_OPCODE_NOP }, + { 0, 0, 0, 0, "", 108 }, /* removed */ + { 0, 0, 0, 0, "", 109 }, /* removed */ + { 0, 0, 0, 0, "", 110 }, /* removed */ + { 0, 0, 0, 0, "", 111 }, /* removed */ + { 1, 1, 0, 0, "NRM4", TGSI_OPCODE_NRM4 }, + { 0, 1, 0, 0, "CALLNZ", TGSI_OPCODE_CALLNZ }, + { 0, 1, 0, 0, "IFC", TGSI_OPCODE_IFC }, + { 0, 1, 0, 0, "BREAKC", TGSI_OPCODE_BREAKC }, + { 0, 1, 0, 0, "KIL", TGSI_OPCODE_KIL }, + { 0, 0, 0, 0, "END", TGSI_OPCODE_END }, + { 1, 1, 0, 0, "SWZ", TGSI_OPCODE_SWZ } }; const struct tgsi_opcode_info * tgsi_get_opcode_info( uint opcode ) { + static boolean firsttime = 1; + + if (firsttime) { + unsigned i; + firsttime = 0; + for (i = 0; i < Elements(opcode_info); i++) + assert(opcode_info[i].opcode == i); + } + if (opcode < TGSI_OPCODE_LAST) return &opcode_info[opcode]; + assert( 0 ); return NULL; } + + +const char * +tgsi_get_opcode_name( uint opcode ) +{ + const struct tgsi_opcode_info *info = tgsi_get_opcode_info(opcode); + return info->mnemonic; +} + diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h index 077e25acd7f..b2375c69710 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_info.h +++ b/src/gallium/auxiliary/tgsi/tgsi_info.h @@ -41,13 +41,16 @@ struct tgsi_opcode_info boolean is_tex; boolean is_branch; const char *mnemonic; - const char *alt_mnemonic1; - const char *alt_mnemonic2; + uint opcode; }; const struct tgsi_opcode_info * tgsi_get_opcode_info( uint opcode ); +const char * +tgsi_get_opcode_name( uint opcode ); + + #if defined __cplusplus } #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h new file mode 100644 index 00000000000..ed594a3e2c7 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_opcode_tmp.h @@ -0,0 +1,173 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +#ifndef OP12_TEX +#define OP12_TEX(a) OP12(a) +#endif + +#ifndef OP14_TEX +#define OP14_TEX(a) OP14(a) +#endif + +#ifndef OP00_LBL +#define OP00_LBL(a) OP00(a) +#endif + +#ifndef OP01_LBL +#define OP01_LBL(a) OP01(a) +#endif + +OP11(ARL) +OP11(MOV) +OP11(LIT) +OP11(RCP) +OP11(RSQ) +OP11(EXP) +OP11(LOG) +OP12(MUL) +OP12(ADD) +OP12(DP3) +OP12(DP4) +OP12(DST) +OP12(MIN) +OP12(MAX) +OP12(SLT) +OP12(SGE) +OP13(MAD) +OP12(SUB) +OP13(LRP) +OP13(CND) +OP13(CND0) +OP13(DP2A) +OP11(FRC) +OP13(CLAMP) +OP11(FLR) +OP11(ROUND) +OP11(EX2) +OP11(LG2) +OP12(POW) +OP12(XPD) +OP11(ABS) +OP11(RCC) +OP12(DPH) +OP11(COS) +OP11(DDX) +OP11(DDY) +OP00(KILP) +OP11(PK2H) +OP11(PK2US) +OP11(PK4B) +OP11(PK4UB) +OP12(RFL) +OP12(SEQ) +OP12(SFL) +OP12(SGT) +OP11(SIN) +OP12(SLE) +OP12(SNE) +OP12(STR) +OP12_TEX(TEX) +OP14_TEX(TXD) +OP12_TEX(TXP) +OP11(UP2H) +OP11(UP2US) +OP11(UP4B) +OP11(UP4UB) +OP13(X2D) +OP11(ARA) +OP11(ARR) +OP01(BRA) +OP00_LBL(CAL) +OP00(RET) +OP11(SSG) +OP13(CMP) +OP11(SCS) +OP12_TEX(TXB) +OP11(NRM) +OP12(DIV) +OP12(DP2) +OP12_TEX(TXL) +OP00(BRK) +OP01_LBL(IF) +OP11(BGNFOR) +OP01(REP) +OP00_LBL(ELSE) +OP00(ENDIF) +OP10(ENDFOR) +OP00(ENDREP) +OP01(PUSHA) +OP10(POPA) +OP11(CEIL) +OP11(I2F) +OP11(NOT) +OP11(TRUNC) +OP12(SHL) +OP12(SHR) +OP12(AND) +OP12(OR) +OP12(MOD) +OP12(XOR) +OP13(SAD) +OP12_TEX(TXF) +OP12_TEX(TXQ) +OP00(CONT) +OP00(EMIT) +OP00(ENDPRIM) +OP00_LBL(BGNLOOP) +OP00(BGNSUB) +OP00_LBL(ENDLOOP) +OP00(ENDSUB) +OP11(NOISE1) +OP11(NOISE2) +OP11(NOISE3) +OP11(NOISE4) +OP00(NOP) +OP11(NRM4) +OP01(CALLNZ) +OP01(IFC) +OP01(BREAKC) +OP01(KIL) +OP00(END) +OP11(SWZ) + + +#undef OP00 +#undef OP01 +#undef OP10 +#undef OP11 +#undef OP12 +#undef OP13 + +#ifdef OP14 +#undef OP14 +#endif + +#undef OP00_LBL +#undef OP01_LBL + +#undef OP12_TEX +#undef OP14_TEX + diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index 0081f74ffc4..4870f82b6bd 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -42,9 +42,6 @@ void tgsi_full_token_free( union tgsi_full_token *full_token ) { - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( (void *) full_token->FullImmediate.u.Pointer ); - } } unsigned @@ -156,14 +153,8 @@ tgsi_parse_token( case TGSI_IMM_FLOAT32: { uint imm_count = imm->Immediate.NrTokens - 1; - struct tgsi_immediate_float32 *data; - - data = (struct tgsi_immediate_float32 *) MALLOC(sizeof(struct tgsi_immediate_float32) * imm_count); - if (data) { - for (i = 0; i < imm_count; i++) { - next_token(ctx, &data[i]); - } - imm->u.ImmediateFloat32 = data; + for (i = 0; i < imm_count; i++) { + next_token(ctx, &imm->u[i]); } } break; @@ -219,7 +210,6 @@ tgsi_parse_token( /* * No support for indirect or multi-dimensional addressing. */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); extended = inst->FullDstRegisters[i].DstRegister.Extended; @@ -246,6 +236,17 @@ tgsi_parse_token( extended = token.Extended; } + + if( inst->FullDstRegisters[i].DstRegister.Indirect ) { + next_token( ctx, &inst->FullDstRegisters[i].DstRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegisterInd.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegisterInd.Dimension ); + assert( !inst->FullDstRegisters[i].DstRegisterInd.Extended ); + } } assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h index 054350712d8..a26ee5ba862 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h @@ -48,6 +48,7 @@ struct tgsi_full_header struct tgsi_full_dst_register { struct tgsi_dst_register DstRegister; + struct tgsi_src_register DstRegisterInd; struct tgsi_dst_register_ext_concode DstRegisterExtConcode; struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; }; @@ -72,11 +73,7 @@ struct tgsi_full_declaration struct tgsi_full_immediate { struct tgsi_immediate Immediate; - union - { - const void *Pointer; - const struct tgsi_immediate_float32 *ImmediateFloat32; - } u; + union tgsi_immediate_data u[4]; }; #define TGSI_FULL_MAX_DST_REGISTERS 2 @@ -90,6 +87,7 @@ struct tgsi_full_instruction struct tgsi_instruction_ext_texture InstructionExtTexture; struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; + uint Flags; /**< user-defined usage */ }; union tgsi_full_token diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c index 0c64ae57131..4b1c7d4e01b 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_ppc.c +++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c @@ -38,6 +38,7 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_sse.h" +#include "tgsi/tgsi_info.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_dump.h" @@ -619,17 +620,17 @@ emit_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst) ppc_vandc(gen->f, v1, v0, bit31_vec); /* v1 = v0 & ~bit31 */ } break; - case TGSI_OPCODE_FLOOR: + case TGSI_OPCODE_FLR: ppc_vrfim(gen->f, v1, v0); /* v1 = floor(v0) */ break; - case TGSI_OPCODE_FRAC: + case TGSI_OPCODE_FRC: ppc_vrfim(gen->f, v1, v0); /* tmp = floor(v0) */ ppc_vsubfp(gen->f, v1, v0, v1); /* v1 = v0 - v1 */ break; - case TGSI_OPCODE_EXPBASE2: + case TGSI_OPCODE_EX2: ppc_vexptefp(gen->f, v1, v0); /* v1 = 2^v0 */ break; - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_LG2: /* XXX this may be broken! */ ppc_vlogefp(gen->f, v1, v0); /* v1 = log2(v0) */ break; @@ -1107,14 +1108,23 @@ static int emit_instruction(struct gen_context *gen, struct tgsi_full_instruction *inst) { + + /* we don't handle saturation/clamping yet */ + if (inst->Instruction.Saturate != TGSI_SAT_NONE) + return 0; + + /* need to use extra temps to fix SOA dependencies : */ + if (tgsi_check_soa_dependencies(inst)) + return FALSE; + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_MOV: case TGSI_OPCODE_SWZ: case TGSI_OPCODE_ABS: - case TGSI_OPCODE_FLOOR: - case TGSI_OPCODE_FRAC: - case TGSI_OPCODE_EXPBASE2: - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_FLR: + case TGSI_OPCODE_FRC: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: emit_unaryop(gen, inst); break; case TGSI_OPCODE_RSQ: @@ -1317,8 +1327,10 @@ tgsi_emit_ppc(const struct tgsi_token *tokens, ok = emit_instruction(&gen, &parse.FullToken.FullInstruction); if (!ok) { - debug_printf("failed to translate tgsi opcode %d to PPC (%s)\n", - parse.FullToken.FullInstruction.Instruction.Opcode, + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + debug_printf("failed to translate tgsi opcode %d (%s) to PPC (%s)\n", + opcode, + tgsi_get_opcode_name(opcode), parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? "vertex shader" : "fragment shader"); } @@ -1333,7 +1345,7 @@ tgsi_emit_ppc(const struct tgsi_token *tokens, assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); for (i = 0; i < size; i++) { immediates[num_immediates][i] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + parse.FullToken.FullImmediate.u[i].Float; } num_immediates++; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 6f1f5c2b4b0..4fe8553c423 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -131,7 +131,7 @@ is_register_used( return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } -static const char *file_names[] = +static const char *file_names[TGSI_FILE_COUNT] = { "NULL", "CONST", @@ -140,7 +140,8 @@ static const char *file_names[] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static boolean @@ -234,9 +235,29 @@ iter_instruction( index, "indirect", FALSE ); - if (file != TGSI_FILE_ADDRESS || index != 0) - report_warning( ctx, "Indirect register not ADDR[0]" ); + if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) { + report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]"); + } + } + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_BGNFOR: + case TGSI_OPCODE_ENDFOR: + if (inst->FullDstRegisters[0].DstRegister.File != TGSI_FILE_LOOP || + inst->FullDstRegisters[0].DstRegister.Index != 0) { + report_error(ctx, "Destination register must be LOOP[0]"); + } + break; + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_BGNFOR: + if (inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_CONSTANT && + inst->FullSrcRegisters[0].SrcRegister.File != TGSI_FILE_IMMEDIATE) { + report_error(ctx, "Source register file must be either CONST or IMM"); } + break; } ctx->num_instructions++; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index ba2bfdef062..46f2387c158 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -32,9 +32,11 @@ #include "util/u_debug.h" #include "pipe/p_shader_tokens.h" #include "util/u_math.h" +#include "util/u_memory.h" #if defined(PIPE_ARCH_SSE) #include "util/u_sse.h" #endif +#include "tgsi/tgsi_info.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_exec.h" @@ -100,37 +102,55 @@ get_const_base( void ) { return x86_make_reg( file_REG32, - reg_CX ); + reg_AX ); } static struct x86_reg -get_input_base( void ) +get_machine_base( void ) { return x86_make_reg( file_REG32, - reg_AX ); + reg_CX ); +} + +static struct x86_reg +get_input_base( void ) +{ + return x86_make_disp( + get_machine_base(), + Offset(struct tgsi_exec_machine, Inputs) ); } static struct x86_reg get_output_base( void ) { - return x86_make_reg( - file_REG32, - reg_DX ); + return x86_make_disp( + get_machine_base(), + Offset(struct tgsi_exec_machine, Outputs) ); } static struct x86_reg get_temp_base( void ) { + return x86_make_disp( + get_machine_base(), + Offset(struct tgsi_exec_machine, Temps) ); +} + +static struct x86_reg +get_coef_base( void ) +{ return x86_make_reg( file_REG32, reg_BX ); } static struct x86_reg -get_coef_base( void ) +get_sampler_base( void ) { - return get_output_base(); + return x86_make_reg( + file_REG32, + reg_DI ); } static struct x86_reg @@ -138,7 +158,7 @@ get_immediate_base( void ) { return x86_make_reg( file_REG32, - reg_DI ); + reg_DX ); } @@ -168,6 +188,15 @@ get_const( } static struct x86_reg +get_sampler_ptr( + unsigned unit ) +{ + return x86_make_disp( + get_sampler_base(), + unit * sizeof( struct tgsi_sampler * ) ); +} + +static struct x86_reg get_input( unsigned vec, unsigned chan ) @@ -241,12 +270,14 @@ emit_const( /* 'vec' is the offset from the address register's value. * We're loading CONST[ADDR+vec] into an xmm register. */ - struct x86_reg r0 = get_input_base(); - struct x86_reg r1 = get_output_base(); + struct x86_reg r0 = get_immediate_base(); + struct x86_reg r1 = get_coef_base(); uint i; assert( indirectFile == TGSI_FILE_ADDRESS ); assert( indirectIndex == 0 ); + assert( r0.mod == mod_REG ); + assert( r1.mod == mod_REG ); x86_push( func, r0 ); x86_push( func, r1 ); @@ -520,24 +551,15 @@ emit_coef_dady( * that the stack pointer is 16 byte aligned, as expected. */ static void -emit_func_call_dst( +emit_func_call( struct x86_function *func, - unsigned xmm_save, - unsigned xmm_dst, + unsigned xmm_save_mask, + const struct x86_reg *arg, + unsigned nr_args, void (PIPE_CDECL *code)() ) { struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); unsigned i, n; - unsigned xmm_mask; - - /* Bitmask of the xmm registers to save */ - xmm_mask = (1 << xmm_save) - 1; - xmm_mask &= ~(1 << xmm_dst); - - sse_movaps( - func, - get_temp( TEMP_R0, 0 ), - make_xmm( xmm_dst ) ); x86_push( func, @@ -549,8 +571,10 @@ emit_func_call_dst( func, x86_make_reg( file_REG32, reg_DX) ); + /* Store XMM regs to the stack + */ for(i = 0, n = 0; i < 8; ++i) - if(xmm_mask & (1 << i)) + if(xmm_save_mask & (1 << i)) ++n; x86_sub_imm( @@ -559,26 +583,42 @@ emit_func_call_dst( n*16); for(i = 0, n = 0; i < 8; ++i) - if(xmm_mask & (1 << i)) { + if(xmm_save_mask & (1 << i)) { sse_movups( func, x86_make_disp( x86_make_reg( file_REG32, reg_SP ), n*16 ), make_xmm( i ) ); ++n; } + + for (i = 0; i < nr_args; i++) { + /* Load the address of the buffer we use for passing arguments and + * receiving results: + */ + x86_lea( + func, + ecx, + arg[i] ); - x86_lea( - func, - ecx, - get_temp( TEMP_R0, 0 ) ); - - x86_push( func, ecx ); + /* Push actual function arguments (currently just the pointer to + * the buffer above), and call the function: + */ + x86_push( func, ecx ); + } + x86_mov_reg_imm( func, ecx, (unsigned long) code ); x86_call( func, ecx ); - x86_pop(func, ecx ); - + + /* Pop the arguments (or just add an immediate to esp) + */ + for (i = 0; i < nr_args; i++) { + x86_pop(func, ecx ); + } + + /* Pop the saved XMM regs: + */ for(i = 0, n = 0; i < 8; ++i) - if(xmm_mask & (1 << i)) { + if(xmm_save_mask & (1 << i)) { sse_movups( func, make_xmm( i ), @@ -602,34 +642,86 @@ emit_func_call_dst( x86_pop( func, x86_make_reg( file_REG32, reg_AX) ); +} + +static void +emit_func_call_dst_src1( + struct x86_function *func, + unsigned xmm_save, + unsigned xmm_dst, + unsigned xmm_src0, + void (PIPE_CDECL *code)() ) +{ + struct x86_reg store = get_temp( TEMP_R0, 0 ); + unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst); + + /* Store our input parameters (in xmm regs) to the buffer we use + * for passing arguments. We will pass a pointer to this buffer as + * the actual function argument. + */ + sse_movaps( + func, + store, + make_xmm( xmm_src0 ) ); + + emit_func_call( func, + xmm_mask, + &store, + 1, + code ); sse_movaps( func, make_xmm( xmm_dst ), - get_temp( TEMP_R0, 0 ) ); + store ); } + static void -emit_func_call_dst_src( +emit_func_call_dst_src2( struct x86_function *func, unsigned xmm_save, unsigned xmm_dst, - unsigned xmm_src, + unsigned xmm_src0, + unsigned xmm_src1, void (PIPE_CDECL *code)() ) { + struct x86_reg store = get_temp( TEMP_R0, 0 ); + unsigned xmm_mask = ((1 << xmm_save) - 1) & ~(1 << xmm_dst); + + /* Store two inputs to parameter buffer. + */ sse_movaps( func, - get_temp( TEMP_R0, 1 ), - make_xmm( xmm_src ) ); + store, + make_xmm( xmm_src0 ) ); - emit_func_call_dst( + sse_movaps( func, - xmm_save, - xmm_dst, - code ); + x86_make_disp( store, 4 * sizeof(float) ), + make_xmm( xmm_src1 ) ); + + + /* Emit the call + */ + emit_func_call( func, + xmm_mask, + &store, + 1, + code ); + + /* Retrieve the results: + */ + sse_movaps( + func, + make_xmm( xmm_dst ), + store ); } + + + #if defined(PIPE_ARCH_SSE) /* @@ -782,10 +874,11 @@ emit_cos( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, cos4f ); } @@ -812,10 +905,11 @@ emit_ex2( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, ex24f ); } @@ -857,10 +951,11 @@ emit_flr( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, flr4f ); } @@ -880,10 +975,11 @@ emit_frc( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, frc4f ); } @@ -910,10 +1006,11 @@ emit_lg2( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, lg24f ); } @@ -975,13 +1072,15 @@ emit_pow( struct x86_function *func, unsigned xmm_save, unsigned xmm_dst, - unsigned xmm_src ) + unsigned xmm_src0, + unsigned xmm_src1 ) { - emit_func_call_dst_src( + emit_func_call_dst_src2( func, xmm_save, xmm_dst, - xmm_src, + xmm_src0, + xmm_src1, pow4f ); } @@ -1017,10 +1116,11 @@ emit_rnd( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, rnd4f ); } @@ -1099,10 +1199,11 @@ emit_sgn( unsigned xmm_save, unsigned xmm_dst ) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, sgn4f ); } @@ -1121,10 +1222,11 @@ emit_sin (struct x86_function *func, unsigned xmm_save, unsigned xmm_dst) { - emit_func_call_dst( + emit_func_call_dst_src1( func, xmm_save, xmm_dst, + xmm_dst, sin4f ); } @@ -1140,6 +1242,12 @@ emit_sub( make_xmm( xmm_src ) ); } + + + + + + /** * Register fetch. */ @@ -1298,20 +1406,164 @@ emit_store( #define STORE( FUNC, INST, XMM, INDEX, CHAN )\ emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) + +static void PIPE_CDECL +fetch_texel( struct tgsi_sampler **sampler, + float *store ) +{ +#if 0 + uint j; + + debug_printf("%s sampler: %p (%p) store: %p\n", + __FUNCTION__, + sampler, *sampler, + store ); + + debug_printf("lodbias %f\n", store[12]); + + for (j = 0; j < 4; j++) + debug_printf("sample %d texcoord %f %f\n", + j, + store[0+j], + store[4+j]); +#endif + + { + float rgba[NUM_CHANNELS][QUAD_SIZE]; + (*sampler)->get_samples(*sampler, + &store[0], + &store[4], + &store[8], + 0.0f, /*store[12], lodbias */ + rgba); + + memcpy( store, rgba, 16 * sizeof(float)); + } + +#if 0 + for (j = 0; j < 4; j++) + debug_printf("sample %d result %f %f %f %f\n", + j, + store[0+j], + store[4+j], + store[8+j], + store[12+j]); +#endif +} + /** * High-level instruction translators. */ static void +emit_tex( struct x86_function *func, + const struct tgsi_full_instruction *inst, + boolean lodbias, + boolean projected) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + struct x86_reg args[2]; + unsigned count; + unsigned i; + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + count = 1; + break; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + count = 2; + break; + case TGSI_TEXTURE_SHADOW1D: + case TGSI_TEXTURE_SHADOW2D: + case TGSI_TEXTURE_SHADOWRECT: + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + count = 3; + break; + default: + assert(0); + return; + } + + if (lodbias) { + FETCH( func, *inst, 3, 0, 3 ); + } + else { + emit_tempf( + func, + 3, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + + } + + /* store lodbias whether enabled or not -- fetch_texel currently + * respects it always. + */ + sse_movaps( func, + get_temp( TEMP_R0, 3 ), + make_xmm( 3 ) ); + + + if (projected) { + FETCH( func, *inst, 3, 0, 3 ); + + emit_rcp( func, 3, 3 ); + } + + for (i = 0; i < count; i++) { + FETCH( func, *inst, i, 0, i ); + + if (projected) { + sse_mulps( + func, + make_xmm( i ), + make_xmm( 3 ) ); + } + + /* Store in the argument buffer: + */ + sse_movaps( + func, + get_temp( TEMP_R0, i ), + make_xmm( i ) ); + } + + args[0] = get_temp( TEMP_R0, 0 ); + args[1] = get_sampler_ptr( unit ); + + + emit_func_call( func, + 0, + args, + Elements(args), + fetch_texel ); + + /* If all four channels are enabled, could use a pointer to + * dst[0].x instead of TEMP_R0 for store? + */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, i ) { + + sse_movaps( + func, + make_xmm( 0 ), + get_temp( TEMP_R0, i ) ); + + STORE( func, *inst, 0, 0, i ); + } +} + + +static void emit_kil( struct x86_function *func, const struct tgsi_full_src_register *reg ) { unsigned uniquemask; - unsigned registers[4]; - unsigned nextregister = 0; - unsigned firstchan = ~0; + unsigned unique_count = 0; unsigned chan_index; + unsigned i; /* This mask stores component bits that were already tested. Note that * we test if the value is less than zero, so 1.0 and 0.0 need not to be @@ -1331,18 +1583,11 @@ emit_kil( uniquemask |= 1 << swizzle; /* allocate register */ - registers[chan_index] = nextregister; emit_fetch( func, - nextregister, + unique_count++, reg, chan_index ); - nextregister++; - - /* mark the first channel used */ - if( firstchan == ~0 ) { - firstchan = chan_index; - } } } @@ -1353,32 +1598,32 @@ emit_kil( func, x86_make_reg( file_REG32, reg_DX ) ); - FOR_EACH_CHANNEL( chan_index ) { - if( uniquemask & (1 << chan_index) ) { - sse_cmpps( + for (i = 0 ; i < unique_count; i++ ) { + struct x86_reg dataXMM = make_xmm(i); + + sse_cmpps( + func, + dataXMM, + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + + if( i == 0 ) { + sse_movmskps( func, - make_xmm( registers[chan_index] ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - - if( chan_index == firstchan ) { - sse_pmovmskb( - func, - x86_make_reg( file_REG32, reg_AX ), - make_xmm( registers[chan_index] ) ); - } - else { - sse_pmovmskb( - func, - x86_make_reg( file_REG32, reg_DX ), - make_xmm( registers[chan_index] ) ); - x86_or( - func, - x86_make_reg( file_REG32, reg_AX ), - x86_make_reg( file_REG32, reg_DX ) ); - } + x86_make_reg( file_REG32, reg_AX ), + dataXMM ); + } + else { + sse_movmskps( + func, + x86_make_reg( file_REG32, reg_DX ), + dataXMM ); + x86_or( + func, + x86_make_reg( file_REG32, reg_AX ), + x86_make_reg( file_REG32, reg_DX ) ); } } @@ -1502,6 +1747,14 @@ emit_instruction( if (indirect_temp_reference(inst)) return FALSE; + /* we don't handle saturation/clamping yet */ + if (inst->Instruction.Saturate != TGSI_SAT_NONE) + return FALSE; + + /* need to use extra temps to fix SOA dependencies : */ + if (tgsi_check_soa_dependencies(inst)) + return FALSE; + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1573,7 +1826,7 @@ emit_instruction( get_temp( TGSI_EXEC_TEMP_MINUS_128_I, TGSI_EXEC_TEMP_MINUS_128_C ) ); - emit_pow( func, 3, 1, 2 ); + emit_pow( func, 3, 1, 1, 2 ); FETCH( func, *inst, 0, 0, CHAN_X ); sse_xorps( func, @@ -1820,8 +2073,7 @@ emit_instruction( } break; - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ + case TGSI_OPCODE_LRP: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); @@ -1841,8 +2093,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ + case TGSI_OPCODE_DP2A: FETCH( func, *inst, 0, 0, CHAN_X ); /* xmm0 = src[0].x */ FETCH( func, *inst, 1, 1, CHAN_X ); /* xmm1 = src[1].x */ emit_mul( func, 0, 1 ); /* xmm0 = xmm0 * xmm1 */ @@ -1857,16 +2108,7 @@ emit_instruction( } break; - case TGSI_OPCODE_INDEX: - return 0; - break; - - case TGSI_OPCODE_NEGATE: - return 0; - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ + case TGSI_OPCODE_FRC: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); emit_frc( func, 0, 0 ); @@ -1878,8 +2120,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ + case TGSI_OPCODE_FLR: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); emit_flr( func, 0, 0 ); @@ -1895,8 +2136,7 @@ emit_instruction( } break; - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ + case TGSI_OPCODE_EX2: FETCH( func, *inst, 0, 0, CHAN_X ); emit_ex2( func, 0, 0 ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1904,8 +2144,7 @@ emit_instruction( } break; - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ + case TGSI_OPCODE_LG2: FETCH( func, *inst, 0, 0, CHAN_X ); emit_lg2( func, 0, 0 ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1913,18 +2152,16 @@ emit_instruction( } break; - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ + case TGSI_OPCODE_POW: FETCH( func, *inst, 0, 0, CHAN_X ); FETCH( func, *inst, 1, 1, CHAN_X ); - emit_pow( func, 0, 0, 1 ); + emit_pow( func, 0, 0, 0, 1 ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { STORE( func, *inst, 0, 0, chan_index ); } break; - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ + case TGSI_OPCODE_XPD: if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { FETCH( func, *inst, 1, 1, CHAN_Z ); @@ -1970,10 +2207,6 @@ emit_instruction( } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - return 0; - break; - case TGSI_OPCODE_ABS: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); @@ -2086,21 +2319,7 @@ emit_instruction( break; case TGSI_OPCODE_TEX: - if (0) { - /* Disable dummy texture code: - */ - emit_tempf( - func, - 0, - TEMP_ONE_I, - TEMP_ONE_C ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - } - else { - return 0; - } + emit_tex( func, inst, FALSE, FALSE ); break; case TGSI_OPCODE_TXD: @@ -2198,7 +2417,7 @@ emit_instruction( break; case TGSI_OPCODE_TXB: - return 0; + emit_tex( func, inst, TRUE, FALSE ); break; case TGSI_OPCODE_NRM: @@ -2306,9 +2525,13 @@ emit_instruction( break; case TGSI_OPCODE_TXL: - return 0; + emit_tex( func, inst, TRUE, FALSE ); break; + case TGSI_OPCODE_TXP: + emit_tex( func, inst, FALSE, TRUE ); + break; + case TGSI_OPCODE_BRK: return 0; break; @@ -2317,7 +2540,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: return 0; break; @@ -2333,7 +2556,7 @@ emit_instruction( return 0; break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: return 0; break; @@ -2488,7 +2711,7 @@ emit_declaration( static void aos_to_soa( struct x86_function *func, uint arg_aos, - uint arg_soa, + uint arg_machine, uint arg_num, uint arg_stride ) { @@ -2503,7 +2726,10 @@ static void aos_to_soa( struct x86_function *func, x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); x86_mov( func, aos_input, x86_fn_arg( func, arg_aos ) ); - x86_mov( func, soa_input, x86_fn_arg( func, arg_soa ) ); + x86_mov( func, soa_input, x86_fn_arg( func, arg_machine ) ); + x86_lea( func, soa_input, + x86_make_disp( soa_input, + Offset(struct tgsi_exec_machine, Inputs) ) ); x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) ); x86_mov( func, stride, x86_fn_arg( func, arg_stride ) ); @@ -2545,28 +2771,30 @@ static void aos_to_soa( struct x86_function *func, x86_jcc( func, cc_NE, inner_loop ); /* Restore EBX */ - x86_pop( func, aos_input ); + x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); } -static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) +static void soa_to_aos( struct x86_function *func, + uint arg_aos, + uint arg_machine, + uint arg_num, + uint arg_stride ) { - struct x86_reg soa_output; - struct x86_reg aos_output; - struct x86_reg num_outputs; - struct x86_reg temp; + struct x86_reg soa_output = x86_make_reg( file_REG32, reg_AX ); + struct x86_reg aos_output = x86_make_reg( file_REG32, reg_BX ); + struct x86_reg num_outputs = x86_make_reg( file_REG32, reg_CX ); + struct x86_reg temp = x86_make_reg( file_REG32, reg_DX ); int inner_loop; - soa_output = x86_make_reg( file_REG32, reg_AX ); - aos_output = x86_make_reg( file_REG32, reg_BX ); - num_outputs = x86_make_reg( file_REG32, reg_CX ); - temp = x86_make_reg( file_REG32, reg_DX ); - /* Save EBX */ - x86_push( func, aos_output ); + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); - x86_mov( func, soa_output, x86_fn_arg( func, soa ) ); - x86_mov( func, aos_output, x86_fn_arg( func, aos ) ); - x86_mov( func, num_outputs, x86_fn_arg( func, num ) ); + x86_mov( func, aos_output, x86_fn_arg( func, arg_aos ) ); + x86_mov( func, soa_output, x86_fn_arg( func, arg_machine ) ); + x86_lea( func, soa_output, + x86_make_disp( soa_output, + Offset(struct tgsi_exec_machine, Outputs) ) ); + x86_mov( func, num_outputs, x86_fn_arg( func, arg_num ) ); /* do */ inner_loop = x86_get_label( func ); @@ -2583,7 +2811,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); - x86_mov( func, temp, x86_fn_arg( func, stride ) ); + x86_mov( func, temp, x86_fn_arg( func, arg_stride ) ); x86_push( func, aos_output ); sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); @@ -2607,20 +2835,13 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, x86_jcc( func, cc_NE, inner_loop ); /* Restore EBX */ - x86_pop( func, aos_output ); + x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); } /** * Translate a TGSI vertex/fragment shader to SSE2 code. * Slightly different things are done for vertex vs. fragment shaders. * - * Note that fragment shaders are responsible for interpolating shader - * inputs. Because on x86 we have only 4 GP registers, and here we - * have 5 shader arguments (input, output, const, temp and coef), the - * code is split into two phases -- DECLARATION and INSTRUCTION phase. - * GP register holding the output argument is aliased with the coeff - * argument, as outputs are not needed in the DECLARATION phase. - * * \param tokens the TGSI input shader * \param func the output SSE code/function * \param immediates buffer to place immediates, later passed to SSE func @@ -2634,7 +2855,6 @@ tgsi_emit_sse2( boolean do_swizzles ) { struct tgsi_parse_context parse; - boolean instruction_phase = FALSE; unsigned ok = 1; uint num_immediates = 0; @@ -2646,74 +2866,48 @@ tgsi_emit_sse2( /* Can't just use EDI, EBX without save/restoring them: */ - x86_push( - func, - get_immediate_base() ); - - x86_push( - func, - get_temp_base() ); - + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + x86_push( func, x86_make_reg( file_REG32, reg_DI ) ); /* * Different function args for vertex/fragment shaders: */ - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { - /* DECLARATION phase, do not load output argument. */ - x86_mov( - func, - get_input_base(), - x86_fn_arg( func, 1 ) ); - /* skipping outputs argument here */ - x86_mov( - func, - get_const_base(), - x86_fn_arg( func, 3 ) ); - x86_mov( - func, - get_temp_base(), - x86_fn_arg( func, 4 ) ); - x86_mov( - func, - get_coef_base(), - x86_fn_arg( func, 5 ) ); - x86_mov( - func, - get_immediate_base(), - x86_fn_arg( func, 6 ) ); - } - else { - assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); - + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { if (do_swizzles) aos_to_soa( func, - 6, /* aos_input */ - 1, /* machine->input */ - 7, /* num_inputs */ - 8 ); /* input_stride */ + 4, /* aos_input */ + 1, /* machine */ + 5, /* num_inputs */ + 6 ); /* input_stride */ + } + x86_mov( + func, + get_machine_base(), + x86_fn_arg( func, 1 ) ); + x86_mov( + func, + get_const_base(), + x86_fn_arg( func, 2 ) ); + x86_mov( + func, + get_immediate_base(), + x86_fn_arg( func, 3 ) ); + + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { x86_mov( - func, - get_input_base(), - x86_fn_arg( func, 1 ) ); - x86_mov( - func, - get_output_base(), - x86_fn_arg( func, 2 ) ); - x86_mov( - func, - get_const_base(), - x86_fn_arg( func, 3 ) ); - x86_mov( - func, - get_temp_base(), - x86_fn_arg( func, 4 ) ); - x86_mov( - func, - get_immediate_base(), - x86_fn_arg( func, 5 ) ); + func, + get_coef_base(), + x86_fn_arg( func, 4 ) ); } + x86_mov( + func, + get_sampler_base(), + x86_make_disp( get_machine_base(), + Offset( struct tgsi_exec_machine, Samplers ) ) ); + + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { tgsi_parse_token( &parse ); @@ -2727,24 +2921,15 @@ tgsi_emit_sse2( break; case TGSI_TOKEN_TYPE_INSTRUCTION: - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { - if( !instruction_phase ) { - /* INSTRUCTION phase, overwrite coeff with output. */ - instruction_phase = TRUE; - x86_mov( - func, - get_output_base(), - x86_fn_arg( func, 2 ) ); - } - } - ok = emit_instruction( func, &parse.FullToken.FullInstruction ); if (!ok) { - debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", - parse.FullToken.FullInstruction.Instruction.Opcode, + uint opcode = parse.FullToken.FullInstruction.Instruction.Opcode; + debug_printf("failed to translate tgsi opcode %d (%s) to SSE (%s)\n", + opcode, + tgsi_get_opcode_name(opcode), parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? "vertex shader" : "fragment shader"); } @@ -2759,7 +2944,7 @@ tgsi_emit_sse2( assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); for( i = 0; i < size; i++ ) { immediates[num_immediates][i] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + parse.FullToken.FullImmediate.u[i].Float; } #if 0 debug_printf("SSE FS immediate[%d] = %f %f %f %f\n", @@ -2781,18 +2966,17 @@ tgsi_emit_sse2( if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { if (do_swizzles) - soa_to_aos( func, 9, 2, 10, 11 ); + soa_to_aos( func, + 7, /* aos_output */ + 1, /* machine */ + 8, /* num_outputs */ + 9 ); /* output_stride */ } /* Can't just use EBX, EDI without save/restoring them: */ - x86_pop( - func, - get_temp_base() ); - - x86_pop( - func, - get_immediate_base() ); + x86_pop( func, x86_make_reg( file_REG32, reg_DI ) ); + x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); emit_ret( func ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/tgsi_sse2.h index af838b2a25b..d81ee3d00ec 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.h @@ -34,6 +34,7 @@ extern "C" { struct tgsi_token; struct x86_function; +struct tgsi_interp_coef; unsigned tgsi_emit_sse2( @@ -42,6 +43,33 @@ tgsi_emit_sse2( float (*immediates)[4], boolean do_swizzles ); + +/* This is the function prototype generated when do_swizzles is false + * -- effectively for fragment shaders. + */ +typedef void (PIPE_CDECL *tgsi_sse2_fs_function) ( + struct tgsi_exec_machine *machine, /* 1 */ + const float (*constant)[4], /* 2 */ + const float (*immediate)[4], /* 3 */ + const struct tgsi_interp_coef *coef /* 4 */ + ); + + +/* This is the function prototype generated when do_swizzles is true + * -- effectively for vertex shaders. + */ +typedef void (PIPE_CDECL *tgsi_sse2_vs_func) ( + struct tgsi_exec_machine *machine, /* 1 */ + const float (*constant)[4], /* 2 */ + const float (*immediate)[4], /* 3 */ + const float (*aos_input)[4], /* 4 */ + uint num_inputs, /* 5 */ + uint input_stride, /* 6 */ + float (*aos_output)[4], /* 7 */ + uint num_outputs, /* 8 */ + uint output_stride ); /* 9 */ + + #if defined __cplusplus } #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index a40fcab2126..d438450b1e4 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "util/u_debug.h" +#include "util/u_memory.h" #include "tgsi_text.h" #include "tgsi_build.h" #include "tgsi_info.h" @@ -230,7 +231,8 @@ static const char *file_names[TGSI_FILE_COUNT] = "TEMP", "SAMP", "ADDR", - "IMM" + "IMM", + "LOOP" }; static boolean @@ -788,16 +790,6 @@ match_inst_mnemonic(const char **pcur, if (str_match_no_case(pcur, info->mnemonic)) { return TRUE; } - if (info->alt_mnemonic1) { - if (str_match_no_case(pcur, info->alt_mnemonic1)) { - return TRUE; - } - if (info->alt_mnemonic2) { - if (str_match_no_case(pcur, info->alt_mnemonic2)) { - return TRUE; - } - } - } return FALSE; } @@ -927,7 +919,8 @@ static const char *semantic_names[TGSI_SEMANTIC_COUNT] = "FOG", "PSIZE", "GENERIC", - "NORMAL" + "NORMAL", + "FACE" }; static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] = @@ -947,6 +940,9 @@ static boolean parse_declaration( struct translate_ctx *ctx ) const char *cur; uint advance; + assert(Elements(semantic_names) == TGSI_SEMANTIC_COUNT); + assert(Elements(interpolate_names) == TGSI_INTERPOLATE_COUNT); + if (!eat_white( &ctx->cur )) { report_error( ctx, "Syntax error" ); return FALSE; @@ -1086,7 +1082,10 @@ static boolean parse_immediate( struct translate_ctx *ctx ) imm = tgsi_default_full_immediate(); imm.Immediate.NrTokens += 4; imm.Immediate.DataType = TGSI_IMM_FLOAT32; - imm.u.Pointer = values; + imm.u[0].Float = values[0]; + imm.u[1].Float = values[1]; + imm.u[2].Float = values[2]; + imm.u[3].Float = values[3]; advance = tgsi_build_full_immediate( &imm, diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c new file mode 100644 index 00000000000..c0a0627e0b2 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c @@ -0,0 +1,860 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "tgsi/tgsi_ureg.h" +#include "tgsi/tgsi_dump.h" +#include "util/u_memory.h" +#include "util/u_math.h" + +union tgsi_any_token { + struct tgsi_version version; + struct tgsi_header header; + struct tgsi_processor processor; + struct tgsi_token token; + struct tgsi_declaration decl; + struct tgsi_declaration_range decl_range; + struct tgsi_declaration_semantic decl_semantic; + struct tgsi_immediate imm; + union tgsi_immediate_data imm_data; + struct tgsi_instruction insn; + struct tgsi_instruction_ext_nv insn_ext_nv; + struct tgsi_instruction_ext_label insn_ext_label; + struct tgsi_instruction_ext_texture insn_ext_texture; + struct tgsi_instruction_ext_predicate insn_ext_predicate; + struct tgsi_src_register src; + struct tgsi_src_register_ext_swz src_ext_swz; + struct tgsi_src_register_ext_mod src_ext_mod; + struct tgsi_dimension dim; + struct tgsi_dst_register dst; + struct tgsi_dst_register_ext_concode dst_ext_code; + struct tgsi_dst_register_ext_modulate dst_ext_mod; + struct tgsi_dst_register_ext_predicate dst_ext_pred; + unsigned value; +}; + + +struct ureg_tokens { + union tgsi_any_token *tokens; + unsigned size; + unsigned order; + unsigned count; +}; + +#define UREG_MAX_INPUT PIPE_MAX_ATTRIBS +#define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS +#define UREG_MAX_IMMEDIATE 32 +#define UREG_MAX_TEMP 256 + +#define DOMAIN_DECL 0 +#define DOMAIN_INSN 1 + +struct ureg_program +{ + unsigned processor; + struct pipe_context *pipe; + + struct { + unsigned semantic_name; + unsigned semantic_index; + unsigned interp; + } input[UREG_MAX_INPUT]; + unsigned nr_inputs; + + struct { + unsigned semantic_name; + unsigned semantic_index; + } output[UREG_MAX_OUTPUT]; + unsigned nr_outputs; + + struct { + float v[4]; + unsigned nr; + } immediate[UREG_MAX_IMMEDIATE]; + unsigned nr_immediates; + + unsigned temps_active[UREG_MAX_TEMP / 32]; + unsigned nr_temps; + + unsigned nr_constants; + unsigned nr_samplers; + unsigned nr_instructions; + + struct ureg_tokens domain[2]; +}; + +static union tgsi_any_token error_tokens[32]; + +static void tokens_error( struct ureg_tokens *tokens ) +{ + tokens->tokens = error_tokens; + tokens->size = Elements(error_tokens); + tokens->count = 0; +} + + +static void tokens_expand( struct ureg_tokens *tokens, + unsigned count ) +{ + unsigned old_size = tokens->size * sizeof(unsigned); + + if (tokens->tokens == error_tokens) + goto fail; + + while (tokens->count + count > tokens->size) { + tokens->size = (1 << ++tokens->order); + } + + tokens->tokens = REALLOC(tokens->tokens, + old_size, + tokens->size * sizeof(unsigned)); + if (tokens->tokens == NULL) + goto fail; + + return; + +fail: + tokens_error(tokens); +} + +static void set_bad( struct ureg_program *ureg ) +{ + tokens_error(&ureg->domain[0]); +} + + + +static union tgsi_any_token *get_tokens( struct ureg_program *ureg, + unsigned domain, + unsigned count ) +{ + struct ureg_tokens *tokens = &ureg->domain[domain]; + union tgsi_any_token *result; + + if (tokens->count + count > tokens->size) + tokens_expand(tokens, count); + + result = &tokens->tokens[tokens->count]; + tokens->count += count; + return result; +} + + +static union tgsi_any_token *retrieve_token( struct ureg_program *ureg, + unsigned domain, + unsigned nr ) +{ + if (ureg->domain[domain].tokens == error_tokens) + return &error_tokens[0]; + + return &ureg->domain[domain].tokens[nr]; +} + + + +static INLINE struct ureg_dst +ureg_dst_register( unsigned file, + unsigned index ) +{ + struct ureg_dst dst; + + dst.File = file; + dst.WriteMask = TGSI_WRITEMASK_XYZW; + dst.Indirect = 0; + dst.Saturate = 0; + dst.Index = index; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src_register( unsigned file, + unsigned index ) +{ + struct ureg_src src; + + src.File = file; + src.SwizzleX = TGSI_SWIZZLE_X; + src.SwizzleY = TGSI_SWIZZLE_Y; + src.SwizzleZ = TGSI_SWIZZLE_Z; + src.SwizzleW = TGSI_SWIZZLE_W; + src.Pad = 0; + src.Indirect = 0; + src.Absolute = 0; + src.Index = index; + src.Negate = 0; + + return src; +} + + + + +static struct ureg_src +ureg_DECL_input( struct ureg_program *ureg, + unsigned name, + unsigned index, + unsigned interp_mode ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_inputs; i++) { + if (ureg->input[i].semantic_name == name && + ureg->input[i].semantic_index == index) + goto out; + } + + if (ureg->nr_inputs < UREG_MAX_INPUT) { + ureg->input[i].semantic_name = name; + ureg->input[i].semantic_index = index; + ureg->input[i].interp = interp_mode; + ureg->nr_inputs++; + } + else { + set_bad( ureg ); + } + +out: + return ureg_src_register( TGSI_FILE_INPUT, i ); +} + + + +struct ureg_src +ureg_DECL_fs_input( struct ureg_program *ureg, + unsigned name, + unsigned index, + unsigned interp ) +{ + return ureg_DECL_input( ureg, name, index, interp ); +} + + +struct ureg_src +ureg_DECL_vs_input( struct ureg_program *ureg, + unsigned name, + unsigned index ) +{ + return ureg_DECL_input( ureg, name, index, TGSI_INTERPOLATE_CONSTANT ); +} + + +struct ureg_dst +ureg_DECL_output( struct ureg_program *ureg, + unsigned name, + unsigned index ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_outputs; i++) { + if (ureg->output[i].semantic_name == name && + ureg->output[i].semantic_index == index) + goto out; + } + + if (ureg->nr_outputs < UREG_MAX_OUTPUT) { + ureg->output[i].semantic_name = name; + ureg->output[i].semantic_index = index; + ureg->nr_outputs++; + } + else { + set_bad( ureg ); + } + +out: + return ureg_dst_register( TGSI_FILE_OUTPUT, i ); +} + + +/* Returns a new constant register. Keep track of which have been + * referred to so that we can emit decls later. + * + * There is nothing in this code to bind this constant to any tracked + * value or manage any constant_buffer contents -- that's the + * resposibility of the calling code. + */ +struct ureg_src ureg_DECL_constant(struct ureg_program *ureg ) +{ + return ureg_src_register( TGSI_FILE_CONSTANT, ureg->nr_constants++ ); +} + + +/* Allocate a new temporary. Temporaries greater than UREG_MAX_TEMP + * are legal, but will not be released. + */ +struct ureg_dst ureg_DECL_temporary( struct ureg_program *ureg ) +{ + unsigned i; + + for (i = 0; i < UREG_MAX_TEMP; i += 32) { + int bit = ffs(~ureg->temps_active[i/32]); + if (bit != 0) { + i += bit - 1; + goto out; + } + } + + /* No reusable temps, so allocate a new one: + */ + i = ureg->nr_temps++; + +out: + if (i < UREG_MAX_TEMP) + ureg->temps_active[i/32] |= 1 << (i % 32); + + if (i >= ureg->nr_temps) + ureg->nr_temps = i + 1; + + return ureg_dst_register( TGSI_FILE_TEMPORARY, i ); +} + + +void ureg_release_temporary( struct ureg_program *ureg, + struct ureg_dst tmp ) +{ + if(tmp.File == TGSI_FILE_TEMPORARY) + if (tmp.Index < UREG_MAX_TEMP) + ureg->temps_active[tmp.Index/32] &= ~(1 << (tmp.Index % 32)); +} + + +/* Allocate a new sampler. + */ +struct ureg_src ureg_DECL_sampler( struct ureg_program *ureg ) +{ + return ureg_src_register( TGSI_FILE_SAMPLER, ureg->nr_samplers++ ); +} + + + + +static int match_or_expand_immediate( const float *v, + unsigned nr, + float *v2, + unsigned *nr2, + unsigned *swizzle ) +{ + unsigned i, j; + + for (i = 0; i < nr; i++) { + boolean found = FALSE; + + for (j = 0; j < *nr2 && !found; j++) { + if (v[i] == v2[j]) { + *swizzle |= j << (i * 2); + found = TRUE; + } + } + + if (!found) { + if (*nr2 >= 4) + return FALSE; + + v2[*nr2] = v[i]; + *swizzle |= *nr2 << (i * 2); + (*nr2)++; + } + } + + return TRUE; +} + + + + +struct ureg_src ureg_DECL_immediate( struct ureg_program *ureg, + const float *v, + unsigned nr ) +{ + unsigned i; + unsigned swizzle = 0; + + /* Could do a first pass where we examine all existing immediates + * without expanding. + */ + + for (i = 0; i < ureg->nr_immediates; i++) { + if (match_or_expand_immediate( v, + nr, + ureg->immediate[i].v, + &ureg->immediate[i].nr, + &swizzle )) + goto out; + } + + if (ureg->nr_immediates < UREG_MAX_IMMEDIATE) { + i = ureg->nr_immediates++; + if (match_or_expand_immediate( v, + nr, + ureg->immediate[i].v, + &ureg->immediate[i].nr, + &swizzle )) + goto out; + } + + set_bad( ureg ); + +out: + return ureg_swizzle( ureg_src_register( TGSI_FILE_IMMEDIATE, i ), + (swizzle >> 0) & 0x3, + (swizzle >> 2) & 0x3, + (swizzle >> 4) & 0x3, + (swizzle >> 6) & 0x3); +} + + +void +ureg_emit_src( struct ureg_program *ureg, + struct ureg_src src ) +{ + unsigned size = (1 + + (src.Absolute ? 1 : 0) + + (src.Indirect ? 1 : 0)); + + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); + unsigned n = 0; + + out[n].value = 0; + out[n].src.File = src.File; + out[n].src.SwizzleX = src.SwizzleX; + out[n].src.SwizzleY = src.SwizzleY; + out[n].src.SwizzleZ = src.SwizzleZ; + out[n].src.SwizzleW = src.SwizzleW; + out[n].src.Indirect = src.Indirect; + out[n].src.Index = src.Index; + n++; + + if (src.Absolute) { + out[n].value = 0; + out[n].src_ext_mod.Absolute = 1; + n++; + } + + if (src.Indirect) { + out[n].value = 0; + out[n].src.File = TGSI_FILE_ADDRESS; + out[n].src.SwizzleX = TGSI_SWIZZLE_X; + out[n].src.SwizzleY = TGSI_SWIZZLE_X; + out[n].src.SwizzleZ = TGSI_SWIZZLE_X; + out[n].src.SwizzleW = TGSI_SWIZZLE_X; + out[n].src.Indirect = 0; + out[n].src.Index = 0; + n++; + } + + assert(n == size); +} + + +void +ureg_emit_dst( struct ureg_program *ureg, + struct ureg_dst dst ) +{ + unsigned size = (1 + + (dst.Indirect ? 1 : 0)); + + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size ); + unsigned n = 0; + + out[n].value = 0; + out[n].dst.File = dst.File; + out[n].dst.WriteMask = dst.WriteMask; + out[n].dst.Indirect = dst.Indirect; + out[n].dst.Index = dst.Index; + n++; + + if (dst.Indirect) { + out[n].value = 0; + out[n].src.File = TGSI_FILE_ADDRESS; + out[n].src.SwizzleX = TGSI_SWIZZLE_X; + out[n].src.SwizzleY = TGSI_SWIZZLE_X; + out[n].src.SwizzleZ = TGSI_SWIZZLE_X; + out[n].src.SwizzleW = TGSI_SWIZZLE_X; + out[n].src.Indirect = 0; + out[n].src.Index = 0; + n++; + } + + assert(n == size); +} + + + +unsigned +ureg_emit_insn(struct ureg_program *ureg, + unsigned opcode, + boolean saturate, + unsigned num_dst, + unsigned num_src ) +{ + union tgsi_any_token *out; + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + out[0].value = 0; + out[0].insn.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + out[0].insn.NrTokens = 0; + out[0].insn.Opcode = opcode; + out[0].insn.Saturate = saturate; + out[0].insn.NrTokens = 0; + out[0].insn.NumDstRegs = num_dst; + out[0].insn.NumSrcRegs = num_src; + out[0].insn.Padding = 0; + out[0].insn.Extended = 0; + + ureg->nr_instructions++; + + return ureg->domain[DOMAIN_INSN].count - 1; +} + + +void +ureg_emit_label(struct ureg_program *ureg, + unsigned insn_token, + unsigned *label_token ) +{ + union tgsi_any_token *out, *insn; + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + insn = retrieve_token( ureg, DOMAIN_INSN, insn_token ); + + insn->insn.Extended = 1; + + out[0].value = 0; + out[0].insn_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + + *label_token = ureg->domain[DOMAIN_INSN].count - 1; +} + +/* Will return a number which can be used in a label to point to the + * next instruction to be emitted. + */ +unsigned +ureg_get_instruction_number( struct ureg_program *ureg ) +{ + return ureg->nr_instructions; +} + +/* Patch a given label (expressed as a token number) to point to a + * given instruction (expressed as an instruction number). + */ +void +ureg_fixup_label(struct ureg_program *ureg, + unsigned label_token, + unsigned instruction_number ) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, label_token ); + + assert(out->insn_ext_label.Type == TGSI_INSTRUCTION_EXT_TYPE_LABEL); + out->insn_ext_label.Label = instruction_number; +} + + +void +ureg_emit_texture(struct ureg_program *ureg, + unsigned insn_token, + unsigned target ) +{ + union tgsi_any_token *out, *insn; + + out = get_tokens( ureg, DOMAIN_INSN, 1 ); + insn = retrieve_token( ureg, DOMAIN_INSN, insn_token ); + + insn->insn.Extended = 1; + + out[0].value = 0; + out[0].insn_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + out[0].insn_ext_texture.Texture = target; +} + + +void +ureg_fixup_insn_size(struct ureg_program *ureg, + unsigned insn ) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_INSN, insn ); + + assert(out->insn.Type == TGSI_TOKEN_TYPE_INSTRUCTION); + out->insn.NrTokens = ureg->domain[DOMAIN_INSN].count - insn - 1; +} + + +void +ureg_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + const struct ureg_src *src, + unsigned nr_src ) +{ + unsigned insn, i; + boolean saturate; + + saturate = nr_dst ? dst[0].Saturate : FALSE; + + insn = ureg_emit_insn( ureg, opcode, saturate, nr_dst, nr_src ); + + for (i = 0; i < nr_dst; i++) + ureg_emit_dst( ureg, dst[i] ); + + for (i = 0; i < nr_src; i++) + ureg_emit_src( ureg, src[i] ); + + ureg_fixup_insn_size( ureg, insn ); +} + + + +static void emit_decl( struct ureg_program *ureg, + unsigned file, + unsigned index, + unsigned semantic_name, + unsigned semantic_index, + unsigned interp ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 ); + + out[0].value = 0; + out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION; + out[0].decl.NrTokens = 3; + out[0].decl.File = file; + out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW; /* FIXME! */ + out[0].decl.Interpolate = interp; + out[0].decl.Semantic = 1; + + out[1].value = 0; + out[1].decl_range.First = + out[1].decl_range.Last = index; + + out[2].value = 0; + out[2].decl_semantic.SemanticName = semantic_name; + out[2].decl_semantic.SemanticIndex = semantic_index; + +} + + +static void emit_decl_range( struct ureg_program *ureg, + unsigned file, + unsigned first, + unsigned count ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 2 ); + + out[0].value = 0; + out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION; + out[0].decl.NrTokens = 2; + out[0].decl.File = file; + out[0].decl.UsageMask = 0xf; + out[0].decl.Interpolate = TGSI_INTERPOLATE_CONSTANT; + out[0].decl.Semantic = 0; + + out[1].value = 0; + out[1].decl_range.First = first; + out[1].decl_range.Last = first + count - 1; +} + +static void emit_immediate( struct ureg_program *ureg, + const float *v ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 5 ); + + out[0].value = 0; + out[0].imm.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + out[0].imm.NrTokens = 5; + out[0].imm.DataType = TGSI_IMM_FLOAT32; + out[0].imm.Padding = 0; + out[0].imm.Extended = 0; + + out[1].imm_data.Float = v[0]; + out[2].imm_data.Float = v[1]; + out[3].imm_data.Float = v[2]; + out[4].imm_data.Float = v[3]; +} + + + + +static void emit_decls( struct ureg_program *ureg ) +{ + unsigned i; + + for (i = 0; i < ureg->nr_inputs; i++) { + emit_decl( ureg, + TGSI_FILE_INPUT, + i, + ureg->input[i].semantic_name, + ureg->input[i].semantic_index, + ureg->input[i].interp ); + } + + for (i = 0; i < ureg->nr_outputs; i++) { + emit_decl( ureg, + TGSI_FILE_OUTPUT, + i, + ureg->output[i].semantic_name, + ureg->output[i].semantic_index, + TGSI_INTERPOLATE_CONSTANT ); + } + + if (ureg->nr_samplers) { + emit_decl_range( ureg, + TGSI_FILE_SAMPLER, + 0, ureg->nr_samplers ); + } + + if (ureg->nr_constants) { + emit_decl_range( ureg, + TGSI_FILE_CONSTANT, + 0, ureg->nr_constants ); + } + + if (ureg->nr_temps) { + emit_decl_range( ureg, + TGSI_FILE_TEMPORARY, + 0, ureg->nr_temps ); + } + + for (i = 0; i < ureg->nr_immediates; i++) { + emit_immediate( ureg, + ureg->immediate[i].v ); + } +} + +/* Append the instruction tokens onto the declarations to build a + * contiguous stream suitable to send to the driver. + */ +static void copy_instructions( struct ureg_program *ureg ) +{ + unsigned nr_tokens = ureg->domain[DOMAIN_INSN].count; + union tgsi_any_token *out = get_tokens( ureg, + DOMAIN_DECL, + nr_tokens ); + + memcpy(out, + ureg->domain[DOMAIN_INSN].tokens, + nr_tokens * sizeof out[0] ); +} + + +static void +fixup_header_size(struct ureg_program *ureg ) +{ + union tgsi_any_token *out = retrieve_token( ureg, DOMAIN_DECL, 1 ); + + out->header.BodySize = ureg->domain[DOMAIN_DECL].count - 3; +} + + +static void +emit_header( struct ureg_program *ureg ) +{ + union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL, 3 ); + + out[0].version.MajorVersion = 1; + out[0].version.MinorVersion = 1; + out[0].version.Padding = 0; + + out[1].header.HeaderSize = 2; + out[1].header.BodySize = 0; + + out[2].processor.Processor = ureg->processor; + out[2].processor.Padding = 0; +} + + +const struct tgsi_token *ureg_finalize( struct ureg_program *ureg ) +{ + const struct tgsi_token *tokens; + + emit_header( ureg ); + emit_decls( ureg ); + copy_instructions( ureg ); + fixup_header_size( ureg ); + + if (ureg->domain[0].tokens == error_tokens || + ureg->domain[1].tokens == error_tokens) { + debug_printf("%s: error in generated shader\n", __FUNCTION__); + assert(0); + return NULL; + } + + tokens = &ureg->domain[DOMAIN_DECL].tokens[0].token; + + if (0) { + debug_printf("%s: emitted shader %d tokens:\n", __FUNCTION__, + ureg->domain[DOMAIN_DECL].count); + tgsi_dump( tokens, 0 ); + } + + return tokens; +} + + +void *ureg_create_shader( struct ureg_program *ureg, + struct pipe_context *pipe ) +{ + struct pipe_shader_state state; + + state.tokens = ureg_finalize(ureg); + if(!state.tokens) + return NULL; + + if (ureg->processor == TGSI_PROCESSOR_VERTEX) + return pipe->create_vs_state( pipe, &state ); + else + return pipe->create_fs_state( pipe, &state ); +} + + + + +struct ureg_program *ureg_create( unsigned processor ) +{ + struct ureg_program *ureg = CALLOC_STRUCT( ureg_program ); + if (ureg == NULL) + return NULL; + + ureg->processor = processor; + return ureg; +} + + +void ureg_destroy( struct ureg_program *ureg ) +{ + unsigned i; + + for (i = 0; i < Elements(ureg->domain); i++) { + if (ureg->domain[i].tokens && + ureg->domain[i].tokens != error_tokens) + FREE(ureg->domain[i].tokens); + } + + FREE(ureg); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h new file mode 100644 index 00000000000..8836a1ea0eb --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h @@ -0,0 +1,481 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE, INC AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_UREG_H +#define TGSI_UREG_H + +#include "pipe/p_compiler.h" +#include "pipe/p_shader_tokens.h" + +struct ureg_program; + +/* Almost a tgsi_src_register, but we need to pull in the Absolute + * flag from the _ext token. Indirect flag always implies ADDR[0]. + */ +struct ureg_src +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Pad : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Negate : 1; /* BOOL */ +}; + +/* Very similar to a tgsi_dst_register, removing unsupported fields + * and adding a Saturate flag. It's easier to push saturate into the + * destination register than to try and create a _SAT varient of each + * instruction function. + */ +struct ureg_dst +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Saturate : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Pad1 : 5; + unsigned Pad2 : 1; /* BOOL */ +}; + +struct pipe_context; + +struct ureg_program * +ureg_create( unsigned processor ); + +const struct tgsi_token * +ureg_finalize( struct ureg_program * ); + +void * +ureg_create_shader( struct ureg_program *, + struct pipe_context *pipe ); + +void +ureg_destroy( struct ureg_program * ); + + +/*********************************************************************** + * Convenience routine: + */ +static INLINE void * +ureg_create_shader_and_destroy( struct ureg_program *p, + struct pipe_context *pipe ) +{ + void *result = ureg_create_shader( p, pipe ); + ureg_destroy( p ); + return result; +} + + + +/*********************************************************************** + * Build shader declarations: + */ + +struct ureg_src +ureg_DECL_fs_input( struct ureg_program *, + unsigned semantic_name, + unsigned semantic_index, + unsigned interp_mode ); + +struct ureg_src +ureg_DECL_vs_input( struct ureg_program *, + unsigned semantic_name, + unsigned semantic_index ); + +struct ureg_dst +ureg_DECL_output( struct ureg_program *, + unsigned semantic_name, + unsigned semantic_index ); + +struct ureg_src +ureg_DECL_immediate( struct ureg_program *, + const float *v, + unsigned nr ); + +struct ureg_src +ureg_DECL_constant( struct ureg_program * ); + +struct ureg_dst +ureg_DECL_temporary( struct ureg_program * ); + +void +ureg_release_temporary( struct ureg_program *ureg, + struct ureg_dst tmp ); + +struct ureg_src +ureg_DECL_sampler( struct ureg_program * ); + + +static INLINE struct ureg_src +ureg_DECL_immediate4f( struct ureg_program *ureg, + float a, float b, + float c, float d) +{ + float v[4]; + v[0] = a; + v[1] = b; + v[2] = c; + v[3] = d; + return ureg_DECL_immediate( ureg, v, 4 ); +} + +static INLINE struct ureg_src +ureg_DECL_immediate3f( struct ureg_program *ureg, + float a, float b, + float c) +{ + float v[3]; + v[0] = a; + v[1] = b; + v[2] = c; + return ureg_DECL_immediate( ureg, v, 3 ); +} + +static INLINE struct ureg_src +ureg_DECL_immediate2f( struct ureg_program *ureg, + float a, float b) +{ + float v[2]; + v[0] = a; + v[1] = b; + return ureg_DECL_immediate( ureg, v, 2 ); +} + +static INLINE struct ureg_src +ureg_DECL_immediate1f( struct ureg_program *ureg, + float a) +{ + float v[1]; + v[0] = a; + return ureg_DECL_immediate( ureg, v, 1 ); +} + +/*********************************************************************** + * Functions for patching up labels + */ + + +/* Will return a number which can be used in a label to point to the + * next instruction to be emitted. + */ +unsigned +ureg_get_instruction_number( struct ureg_program *ureg ); + + +/* Patch a given label (expressed as a token number) to point to a + * given instruction (expressed as an instruction number). + * + * Labels are obtained from instruction emitters, eg ureg_CAL(). + * Instruction numbers are obtained from ureg_get_instruction_number(), + * above. + */ +void +ureg_fixup_label(struct ureg_program *ureg, + unsigned label_token, + unsigned instruction_number ); + + +/* Generic instruction emitter. Use if you need to pass the opcode as + * a parameter, rather than using the emit_OP() varients below. + */ +void +ureg_insn(struct ureg_program *ureg, + unsigned opcode, + const struct ureg_dst *dst, + unsigned nr_dst, + const struct ureg_src *src, + unsigned nr_src ); + + +/*********************************************************************** + * Internal instruction helpers, don't call these directly: + */ + +unsigned +ureg_emit_insn(struct ureg_program *ureg, + unsigned opcode, + boolean saturate, + unsigned num_dst, + unsigned num_src ); + +void +ureg_emit_label(struct ureg_program *ureg, + unsigned insn_token, + unsigned *label_token ); + +void +ureg_emit_texture(struct ureg_program *ureg, + unsigned insn_token, + unsigned target ); + +void +ureg_emit_dst( struct ureg_program *ureg, + struct ureg_dst dst ); + +void +ureg_emit_src( struct ureg_program *ureg, + struct ureg_src src ); + +void +ureg_fixup_insn_size(struct ureg_program *ureg, + unsigned insn ); + + +#define OP00( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP01( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_src src ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP00_LBL( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + unsigned *label_token ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 0 ); \ + ureg_emit_label( ureg, insn, label_token ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP01_LBL( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_src src, \ + unsigned *label_token ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, FALSE, 0, 1 ); \ + ureg_emit_label( ureg, insn, label_token ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP10( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 0 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + + +#define OP11( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 1 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP12( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src0, \ + struct ureg_src src1 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP12_TEX( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + unsigned target, \ + struct ureg_src src0, \ + struct ureg_src src1 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 2 ); \ + ureg_emit_texture( ureg, insn, target ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP13( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + struct ureg_src src0, \ + struct ureg_src src1, \ + struct ureg_src src2 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 3 ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_emit_src( ureg, src2 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + +#define OP14_TEX( op ) \ +static INLINE void ureg_##op( struct ureg_program *ureg, \ + struct ureg_dst dst, \ + unsigned target, \ + struct ureg_src src0, \ + struct ureg_src src1, \ + struct ureg_src src2, \ + struct ureg_src src3 ) \ +{ \ + unsigned opcode = TGSI_OPCODE_##op; \ + unsigned insn = ureg_emit_insn( ureg, opcode, dst.Saturate, 1, 4 ); \ + ureg_emit_texture( ureg, insn, target ); \ + ureg_emit_dst( ureg, dst ); \ + ureg_emit_src( ureg, src0 ); \ + ureg_emit_src( ureg, src1 ); \ + ureg_emit_src( ureg, src2 ); \ + ureg_emit_src( ureg, src3 ); \ + ureg_fixup_insn_size( ureg, insn ); \ +} + + +/* Use a template include to generate a correctly-typed ureg_OP() + * function for each TGSI opcode: + */ +#include "tgsi_opcode_tmp.h" + + +/*********************************************************************** + * Inline helpers for manipulating register structs: + */ +static INLINE struct ureg_src +ureg_negate( struct ureg_src reg ) +{ + reg.Negate ^= 1; + return reg; +} + +static INLINE struct ureg_src +ureg_abs( struct ureg_src reg ) +{ + reg.Absolute = 1; + reg.Negate = 0; + return reg; +} + +static INLINE struct ureg_src +ureg_swizzle( struct ureg_src reg, + int x, int y, int z, int w ) +{ + unsigned swz = ( (reg.SwizzleX << 0) | + (reg.SwizzleY << 2) | + (reg.SwizzleZ << 4) | + (reg.SwizzleW << 6)); + + reg.SwizzleX = (swz >> (x*2)) & 0x3; + reg.SwizzleY = (swz >> (y*2)) & 0x3; + reg.SwizzleZ = (swz >> (z*2)) & 0x3; + reg.SwizzleW = (swz >> (w*2)) & 0x3; + return reg; +} + +static INLINE struct ureg_src +ureg_scalar( struct ureg_src reg, int x ) +{ + return ureg_swizzle(reg, x, x, x, x); +} + +static INLINE struct ureg_dst +ureg_writemask( struct ureg_dst reg, + unsigned writemask ) +{ + reg.WriteMask &= writemask; + return reg; +} + +static INLINE struct ureg_dst +ureg_saturate( struct ureg_dst reg ) +{ + reg.Saturate = 1; + return reg; +} + +static INLINE struct ureg_dst +ureg_dst( struct ureg_src src ) +{ + struct ureg_dst dst; + + dst.File = src.File; + dst.WriteMask = TGSI_WRITEMASK_XYZW; + dst.Indirect = src.Indirect; + dst.Saturate = 0; + dst.Index = src.Index; + dst.Pad1 = 0; + dst.Pad2 = 0; + + return dst; +} + +static INLINE struct ureg_src +ureg_src( struct ureg_dst dst ) +{ + struct ureg_src src; + + src.File = dst.File; + src.SwizzleX = TGSI_SWIZZLE_X; + src.SwizzleY = TGSI_SWIZZLE_Y; + src.SwizzleZ = TGSI_SWIZZLE_Z; + src.SwizzleW = TGSI_SWIZZLE_W; + src.Pad = 0; + src.Indirect = dst.Indirect; + src.Absolute = 0; + src.Index = dst.Index; + src.Negate = 0; + + return src; +} + + + +#endif diff --git a/src/gallium/auxiliary/trace/trace_drm.h b/src/gallium/auxiliary/trace/trace_drm.h deleted file mode 100644 index 892bd9860c6..00000000000 --- a/src/gallium/auxiliary/trace/trace_drm.h +++ /dev/null @@ -1,165 +0,0 @@ -/* - * Copyright 2009 Jakob Bornecrantz <jakob@vmware.com> - * Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#ifndef TRACE_DRM_H -#define TRACE_DRM_H - -#include "state_tracker/drm_api.h" - -#include "trace/tr_buffer.h" -#include "trace/tr_context.h" -#include "trace/tr_screen.h" -#include "trace/tr_texture.h" - -struct drm_api hooks; - -static struct pipe_screen * -trace_drm_create_screen(int fd, struct drm_create_screen_arg *arg) -{ - struct pipe_screen *screen; - - if (arg && arg->mode != DRM_CREATE_NORMAL) - return NULL; - - screen = hooks.create_screen(fd, arg); - - return trace_screen_create(screen); -}; - -static struct pipe_context * -trace_drm_create_context(struct pipe_screen *_screen) -{ - struct pipe_screen *screen; - struct pipe_context *pipe; - - if (trace_enabled()) - screen = trace_screen(_screen)->screen; - else - screen = _screen; - - pipe = hooks.create_context(screen); - - if (trace_enabled()) - pipe = trace_context_create(_screen, pipe); - - return pipe; -}; - -static boolean -trace_drm_buffer_from_texture(struct pipe_texture *_texture, - struct pipe_buffer **_buffer, - unsigned *stride) -{ - struct pipe_texture *texture; - struct pipe_buffer *buffer = NULL; - boolean result; - - if (trace_enabled()) - texture = trace_texture(_texture)->texture; - else - texture = _texture; - - result = hooks.buffer_from_texture(texture, &buffer, stride); - - if (result && _buffer) - buffer = trace_buffer_create(trace_screen(texture->screen), buffer); - - if (_buffer) - *_buffer = buffer; - else - pipe_buffer_reference(&buffer, NULL); - - return result; -} - -static struct pipe_buffer * -trace_drm_buffer_from_handle(struct pipe_screen *_screen, - const char *name, - unsigned handle) -{ - struct pipe_screen *screen; - struct pipe_buffer *result; - - if (trace_enabled()) - screen = trace_screen(_screen)->screen; - else - screen = _screen; - - result = hooks.buffer_from_handle(screen, name, handle); - - if (trace_enabled()) - result = trace_buffer_create(trace_screen(_screen), result); - - return result; -} - -static boolean -trace_drm_handle_from_buffer(struct pipe_screen *_screen, - struct pipe_buffer *_buffer, - unsigned *handle) -{ - struct pipe_screen *screen; - struct pipe_buffer *buffer; - - if (trace_enabled()) { - screen = trace_screen(_screen)->screen; - buffer = trace_buffer(_buffer)->buffer; - } else { - screen = _screen; - buffer = _buffer; - } - - return hooks.handle_from_buffer(screen, buffer, handle); -} - -static boolean -trace_drm_global_handle_from_buffer(struct pipe_screen *_screen, - struct pipe_buffer *_buffer, - unsigned *handle) -{ - struct pipe_screen *screen; - struct pipe_buffer *buffer; - - if (trace_enabled()) { - screen = trace_screen(_screen)->screen; - buffer = trace_buffer(_buffer)->buffer; - } else { - screen = _screen; - buffer = _buffer; - } - - return hooks.global_handle_from_buffer(screen, buffer, handle); -} - -struct drm_api drm_api_hooks = -{ - .create_screen = trace_drm_create_screen, - .create_context = trace_drm_create_context, - - .buffer_from_texture = trace_drm_buffer_from_texture, - .buffer_from_handle = trace_drm_buffer_from_handle, - .handle_from_buffer = trace_drm_handle_from_buffer, - .global_handle_from_buffer = trace_drm_global_handle_from_buffer, -}; - -#endif /* TRACE_DRM_H */ diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 414cf910254..cda6dbd46d7 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -45,6 +45,7 @@ #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_simple_shaders.h" +#include "util/u_surface.h" #include "cso_cache/cso_context.h" @@ -155,7 +156,11 @@ util_destroy_blit(struct blit_state *ctx) } -static unsigned get_next_slot( struct blit_state *ctx ) +/** + * Get offset of next free slot in vertex buffer for quad vertices. + */ +static unsigned +get_next_slot( struct blit_state *ctx ) { const unsigned max_slots = 4096 / sizeof ctx->vertices; @@ -173,7 +178,6 @@ static unsigned get_next_slot( struct blit_state *ctx ) } - /** * Setup vertex data for the textured quad we'll draw. * Note: y=0=top @@ -260,9 +264,38 @@ setup_vertex_data_tex(struct blit_state *ctx, return offset; } + + +/** + * \return TRUE if two regions overlap, FALSE otherwise + */ +static boolean +regions_overlap(int srcX0, int srcY0, + int srcX1, int srcY1, + int dstX0, int dstY0, + int dstX1, int dstY1) +{ + if (MAX2(srcX0, srcX1) < MIN2(dstX0, dstX1)) + return FALSE; /* src completely left of dst */ + + if (MAX2(dstX0, dstX1) < MIN2(srcX0, srcX1)) + return FALSE; /* dst completely left of src */ + + if (MAX2(srcY0, srcY1) < MIN2(dstY0, dstY1)) + return FALSE; /* src completely above dst */ + + if (MAX2(dstY0, dstY1) < MIN2(srcY0, srcY1)) + return FALSE; /* dst completely above src */ + + return TRUE; /* some overlap */ +} + + /** * Copy pixel block from src surface to dst surface. * Overlapping regions are acceptable. + * Flipping and stretching are supported. + * XXX what about clipping??? * XXX need some control over blitting Z and/or stencil. */ void @@ -285,10 +318,41 @@ util_blit_pixels(struct blit_state *ctx, const int srcLeft = MIN2(srcX0, srcX1); const int srcTop = MIN2(srcY0, srcY1); unsigned offset; + boolean overlap; assert(filter == PIPE_TEX_MIPFILTER_NEAREST || filter == PIPE_TEX_MIPFILTER_LINEAR); + assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); + + /* do the regions overlap? */ + overlap = util_same_surface(src, dst) && + regions_overlap(srcX0, srcY0, srcX1, srcY1, + dstX0, dstY0, dstX1, dstY1); + + /* + * Check for simple case: no format conversion, no flipping, no stretching, + * no overlapping. + * Filter mode should not matter since there's no stretching. + */ + if (dst->format == src->format && + srcX0 < srcX1 && + dstX0 < dstX1 && + srcY0 < srcY1 && + dstY0 < dstY1 && + (dstX1 - dstX0) == (srcX1 - srcX0) && + (dstY1 - dstY0) == (srcY1 - srcY0) && + !overlap) { + pipe->surface_copy(pipe, + dst, dstX0, dstY0, /* dest */ + src, srcX0, srcY0, /* src */ + srcW, srcH); /* size */ + return; + } + if (srcLeft != srcX0) { /* left-right flip */ int tmp = dstX0; @@ -303,20 +367,6 @@ util_blit_pixels(struct blit_state *ctx, dstY1 = tmp; } - assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_SAMPLER, 0)); - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_SAMPLER, 0)); - - if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { - /* FIXME: this will most surely fail for overlapping rectangles */ - pipe->surface_copy(pipe, - dst, dstX0, dstY0, /* dest */ - src, srcX0, srcY0, /* src */ - srcW, srcH); /* size */ - return; - } - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); diff --git a/src/gallium/auxiliary/util/u_cache.c b/src/gallium/auxiliary/util/u_cache.c index 41cd38171fa..47c16b1c927 100644 --- a/src/gallium/auxiliary/util/u_cache.c +++ b/src/gallium/auxiliary/util/u_cache.c @@ -137,6 +137,8 @@ util_cache_set(struct util_cache *cache, struct util_cache_entry *entry; assert(cache); + if (!cache) + return; entry = util_cache_entry_get(cache, key); util_cache_entry_destroy(cache, entry); @@ -158,6 +160,8 @@ util_cache_get(struct util_cache *cache, struct util_cache_entry *entry; assert(cache); + if (!cache) + return NULL; entry = util_cache_entry_get(cache, key); if(!entry->key && !entry->value) @@ -176,7 +180,9 @@ util_cache_clear(struct util_cache *cache) uint32_t i; assert(cache); - + if (!cache) + return; + for(i = 0; i < cache->size; ++i) util_cache_entry_destroy(cache, &cache->entries[i]); } @@ -186,6 +192,8 @@ void util_cache_destroy(struct util_cache *cache) { assert(cache); + if (!cache) + return; #ifdef DEBUG if(cache->count >= 20*cache->size) { diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c index 18597ef8395..96d400c839b 100644 --- a/src/gallium/auxiliary/util/u_debug.c +++ b/src/gallium/auxiliary/util/u_debug.c @@ -97,10 +97,8 @@ void _debug_vprintf(const char *format, va_list ap) buf[0] = '\0'; } #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) - /* EngDebugPrint does not handle float point arguments, so we need to use - * our own vsnprintf implementation. It is also very slow, so buffer until - * we find a newline. */ - static char buf[512 + 1] = {'\0'}; + /* OutputDebugStringA can be very slow, so buffer until we find a newline. */ + static char buf[4096] = {'\0'}; size_t len = strlen(buf); int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap); if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) { @@ -145,11 +143,9 @@ void _debug_vprintf(const char *format, va_list ap) #elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) /* TODO */ #else /* !PIPE_SUBSYSTEM_WINDOWS */ -#ifdef DEBUG fflush(stdout); vfprintf(stderr, format, ap); #endif -#endif } diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 6fa13a8ce11..edc37561ab1 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -46,10 +46,6 @@ #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" -#include "tgsi/tgsi_build.h" -#include "tgsi/tgsi_dump.h" -#include "tgsi/tgsi_parse.h" - #include "cso_cache/cso_context.h" @@ -1385,7 +1381,7 @@ set_vertex_data(struct gen_mipmap_state *ctx, * Not +/-1 to avoid cube face selection ambiguity near the edges, * though that can still sometimes happen with this scale factor... */ - const float scale = 0.9999; + const float scale = 0.9999f; const float sc = (2.0f * st[i][0] - 1.0f) * scale; const float tc = (2.0f * st[i][1] - 1.0f) * scale; diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 6da7353e259..3703718a621 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -87,6 +87,8 @@ handle_table_set_destroy(struct handle_table *ht, void (*destroy)(void *object)) { assert(ht); + if (!ht) + return; ht->destroy = destroy; } @@ -155,7 +157,7 @@ handle_table_add(struct handle_table *ht, assert(ht); assert(object); - if(!object) + if(!object || !ht) return 0; /* linear search for an empty handle */ @@ -193,7 +195,7 @@ handle_table_set(struct handle_table *ht, assert(ht); assert(handle); - if(!handle) + if(!handle || !ht) return 0; assert(object); @@ -222,7 +224,7 @@ handle_table_get(struct handle_table *ht, assert(ht); assert(handle); - if(!handle || handle > ht->size) + if(!handle || !ht || handle > ht->size) return NULL; object = ht->objects[handle - 1]; @@ -240,7 +242,7 @@ handle_table_remove(struct handle_table *ht, assert(ht); assert(handle); - if(!handle || handle > ht->size) + if(!handle || !ht || handle > ht->size) return; index = handle - 1; @@ -283,6 +285,9 @@ handle_table_destroy(struct handle_table *ht) unsigned index; assert(ht); + if (!ht) + return; + if(ht->destroy) for(index = 0; index < ht->size; ++index) handle_table_clear(ht, index); diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c index 2f83e318e44..8c2a8f454cc 100644 --- a/src/gallium/auxiliary/util/u_hash_table.c +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -148,6 +148,8 @@ hash_table_set(struct hash_table *ht, struct cso_hash_iter iter; assert(ht); + if (!ht) + return PIPE_ERROR_BAD_INPUT; key_hash = ht->hash(key); @@ -183,6 +185,8 @@ hash_table_get(struct hash_table *ht, struct hash_table_item *item; assert(ht); + if (!ht) + return NULL; key_hash = ht->hash(key); @@ -203,6 +207,8 @@ hash_table_remove(struct hash_table *ht, struct hash_table_item *item; assert(ht); + if (!ht) + return; key_hash = ht->hash(key); @@ -225,7 +231,9 @@ hash_table_clear(struct hash_table *ht) struct hash_table_item *item; assert(ht); - + if (!ht) + return; + iter = cso_hash_first_node(ht->cso); while (!cso_hash_iter_is_null(iter)) { item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter)); @@ -243,9 +251,11 @@ hash_table_foreach(struct hash_table *ht, struct cso_hash_iter iter; struct hash_table_item *item; enum pipe_error result; - + assert(ht); - + if (!ht) + return PIPE_ERROR_BAD_INPUT; + iter = cso_hash_first_node(ht->cso); while (!cso_hash_iter_is_null(iter)) { item = (struct hash_table_item *)cso_hash_iter_data(iter); @@ -264,9 +274,11 @@ hash_table_destroy(struct hash_table *ht) { struct cso_hash_iter iter; struct hash_table_item *item; - + assert(ht); - + if (!ht) + return; + iter = cso_hash_first_node(ht->cso); while (!cso_hash_iter_is_null(iter)) { item = (struct hash_table_item *)cso_hash_iter_data(iter); diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c index 3f70809efdc..508a2ee0634 100644 --- a/src/gallium/auxiliary/util/u_keymap.c +++ b/src/gallium/auxiliary/util/u_keymap.c @@ -194,6 +194,8 @@ util_keymap_insert(struct keymap *map, const void *key, struct cso_hash_iter iter; assert(map); + if (!map) + return FALSE; key_hash = hash(key, map->key_size); @@ -234,6 +236,8 @@ util_keymap_lookup(const struct keymap *map, const void *key) struct keymap_item *item; assert(map); + if (!map) + return NULL; key_hash = hash(key, map->key_size); @@ -258,6 +262,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user) struct keymap_item *item; assert(map); + if (!map) + return; key_hash = hash(key, map->key_size); @@ -267,6 +273,8 @@ util_keymap_remove(struct keymap *map, const void *key, void *user) item = hash_table_item(iter); assert(item); + if (!item) + return; map->delete_func(map, item->key, item->value, user); FREE(item->key); FREE(item); @@ -288,7 +296,9 @@ util_keymap_remove_all(struct keymap *map, void *user) struct keymap_item *item; assert(map); - + if (!map) + return; + iter = cso_hash_first_node(map->cso); while (!cso_hash_iter_is_null(iter)) { item = (struct keymap_item *) diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index e5003af01d8..57410e78b02 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -53,11 +53,11 @@ __inline double ceil(double val) { double ceil_val; - if((val - (long) val) == 0) { + if ((val - (long) val) == 0) { ceil_val = val; } else { - if(val > 0) { + if (val > 0) { ceil_val = (long) val + 1; } else { @@ -73,11 +73,11 @@ __inline double floor(double val) { double floor_val; - if((val - (long) val) == 0) { + if ((val - (long) val) == 0) { floor_val = val; } else { - if(val > 0) { + if (val > 0) { floor_val = (long) val; } else { @@ -189,7 +189,10 @@ static INLINE double log2( double x ) extern float pow2_table[POW2_TABLE_SIZE]; - +/** + * Initialize math module. This should be called before using any + * other functions in this module. + */ extern void util_init_math(void); @@ -216,23 +219,24 @@ util_fast_exp2(float x) int32_t ipart; float fpart, mpart; union fi epart; - + if(x > 129.00000f) return 3.402823466e+38f; - - if(x < -126.99999f) + + if (x < -126.99999f) return 0.0f; ipart = (int32_t) x; fpart = x - (float) ipart; - + /* same as * epart.f = (float) (1 << ipart) - * but faster and without integer overflow for ipart > 31 */ + * but faster and without integer overflow for ipart > 31 + */ epart.i = (ipart + 127 ) << 23; - + mpart = pow2_table[POW2_TABLE_OFFSET + (int)(fpart * POW2_TABLE_SCALE)]; - + return epart.f * mpart; } @@ -254,6 +258,9 @@ util_fast_exp(float x) extern float log2_table[LOG2_TABLE_SIZE]; +/** + * Fast approximation to log2(x). + */ static INLINE float util_fast_log2(float x) { @@ -267,6 +274,9 @@ util_fast_log2(float x) } +/** + * Fast approximation to x^y. + */ static INLINE float util_fast_pow(float x, float y) { @@ -274,7 +284,6 @@ util_fast_pow(float x, float y) } - /** * Floor(x), returned as int. */ @@ -284,8 +293,8 @@ util_ifloor(float f) int ai, bi; double af, bf; union fi u; - af = (3 << 22) + 0.5 + (double)f; - bf = (3 << 22) + 0.5 - (double)f; + af = (3 << 22) + 0.5 + (double) f; + bf = (3 << 22) + 0.5 - (double) f; u.f = (float) af; ai = u.i; u.f = (float) bf; bi = u.i; return (ai - bi) >> 1; @@ -305,9 +314,9 @@ util_iround(float f) #elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) int r; _asm { - fld f - fistp r - } + fld f + fistp r + } return r; #else if (f >= 0.0f) @@ -340,7 +349,7 @@ static INLINE unsigned long ffs( unsigned long u ) { unsigned long i; - if(_BitScanForward(&i, u)) + if (_BitScanForward(&i, u)) return i + 1; else return 0; @@ -351,7 +360,7 @@ unsigned ffs( unsigned u ) { unsigned i; - if( u == 0 ) { + if (u == 0) { return 0; } @@ -378,7 +387,10 @@ fui( float f ) } - +/** + * Convert ubyte to float in [0, 1]. + * XXX a 256-entry lookup table would be slightly faster. + */ static INLINE float ubyte_to_float(ubyte ub) { @@ -409,7 +421,23 @@ float_to_ubyte(float f) } +/** + * Calc log base 2 + */ +static INLINE unsigned +util_logbase2(unsigned n) +{ + unsigned log2 = 0; + while (n >>= 1) + ++log2; + return log2; +} + +/** + * Clamp X to [MIN, MAX]. + * This is a macro to allow float, int, uint, etc. types. + */ #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) @@ -422,6 +450,11 @@ align(int value, int alignment) return (value + alignment - 1) & ~(alignment - 1); } +static INLINE unsigned +minify(unsigned value) +{ + return MAX2(1, value >> 1); +} #ifndef COPY_4V #define COPY_4V( DST, SRC ) \ diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h index 0b18d043adb..c3f8c918338 100644 --- a/src/gallium/auxiliary/util/u_memory.h +++ b/src/gallium/auxiliary/util/u_memory.h @@ -100,8 +100,14 @@ ExFreePool(void *P); #define MALLOC( SIZE ) malloc( SIZE ) #define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) #define FREE( PTR ) free( PTR ) -#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) +static INLINE void * +_REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) +{ + (void) old_size; + return realloc(old_ptr, new_size); +} +#define REALLOC( a, b, c ) _REALLOC( a, b, c ) #endif diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c index 151a480d34d..4b75d4ba1d0 100644 --- a/src/gallium/auxiliary/util/u_mm.c +++ b/src/gallium/auxiliary/util/u_mm.c @@ -33,30 +33,32 @@ void u_mmDumpMemInfo(const struct mem_block *heap) { - debug_printf("Memory heap %p:\n", (void *)heap); + debug_printf("Memory heap %p:\n", (void *) heap); if (heap == 0) { debug_printf(" heap == 0\n"); - } else { + } + else { const struct mem_block *p; - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); + for (p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); } debug_printf("\nFree list:\n"); - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); + for (p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); } } debug_printf("End of memory blocks\n"); } + struct mem_block * u_mmInit(int ofs, int size) { diff --git a/src/gallium/auxiliary/util/u_mm.h b/src/gallium/auxiliary/util/u_mm.h index ce20e487635..6b158aae6e4 100644 --- a/src/gallium/auxiliary/util/u_mm.h +++ b/src/gallium/auxiliary/util/u_mm.h @@ -84,7 +84,7 @@ extern struct mem_block *u_mmFindBlock(struct mem_block *heap, int start); extern void u_mmDestroy(struct mem_block *mmInit); /** - * For debuging purpose. + * For debugging purposes. */ extern void u_mmDumpMemInfo(const struct mem_block *mmInit); diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h index d7c3995dbf0..a9b533eea70 100644 --- a/src/gallium/auxiliary/util/u_prim.h +++ b/src/gallium/auxiliary/util/u_prim.h @@ -119,7 +119,7 @@ static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr ) } -static INLINE boolean u_reduced_prim( unsigned pipe_prim ) +static INLINE unsigned u_reduced_prim( unsigned pipe_prim ) { switch (pipe_prim) { case PIPE_PRIM_POINTS: diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 74259d453b1..9866b6fc8a0 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -43,7 +43,7 @@ * src_pitch may be negative to do vertical flip of pixels from source. */ void -pipe_copy_rect(ubyte * dst, +util_copy_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, @@ -91,7 +91,7 @@ pipe_copy_rect(ubyte * dst, } void -pipe_fill_rect(ubyte * dst, +util_fill_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, @@ -204,7 +204,7 @@ util_surface_copy(struct pipe_context *pipe, if (src_map && dst_map) { /* If do_flip, invert src_y position and pass negative src stride */ - pipe_copy_rect(dst_map, + util_copy_rect(dst_map, &dst_trans->block, dst_trans->stride, 0, 0, @@ -263,7 +263,7 @@ util_surface_fill(struct pipe_context *pipe, case 1: case 2: case 4: - pipe_fill_rect(dst_map, &dst_trans->block, dst_trans->stride, + util_fill_rect(dst_map, &dst_trans->block, dst_trans->stride, 0, 0, width, height, value); break; case 8: diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h index 59e842e16d1..daa50834d36 100644 --- a/src/gallium/auxiliary/util/u_rect.h +++ b/src/gallium/auxiliary/util/u_rect.h @@ -42,13 +42,13 @@ struct pipe_surface; extern void -pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, +util_copy_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, const ubyte * src, int src_stride, unsigned src_x, int src_y); extern void -pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block, +util_fill_rect(ubyte * dst, const struct pipe_format_block *block, unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, uint32_t value); diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c index 8114b53cd0d..f01296b40fc 100644 --- a/src/gallium/auxiliary/util/u_simple_screen.c +++ b/src/gallium/auxiliary/util/u_simple_screen.c @@ -65,12 +65,13 @@ pass_surface_buffer_create(struct pipe_screen *screen, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { struct pipe_buffer *buffer = screen->winsys->surface_buffer_create(screen->winsys, width, height, - format, usage, stride); + format, usage, tex_usage, stride); buffer->screen = screen; diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index e519c354d25..d54a1d8c746 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -42,9 +42,7 @@ #include "util/u_memory.h" #include "util/u_simple_shaders.h" -#include "tgsi/tgsi_build.h" -#include "tgsi/tgsi_dump.h" -#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_ureg.h" @@ -58,93 +56,31 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, const uint *semantic_indexes) { - struct pipe_shader_state shader; - struct tgsi_token tokens[100]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_VERTEX; - uint ti, i; + struct ureg_program *ureg; + uint i; - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + ureg = ureg_create( TGSI_PROCESSOR_VERTEX ); + if (ureg == NULL) + return NULL; - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare inputs */ - for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = semantic_names[i]; - decl.Semantic.SemanticIndex = semantic_indexes[i]; - - decl.DeclarationRange.First = - decl.DeclarationRange.Last = i; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - } - - /* declare outputs */ for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = semantic_names[i]; - decl.Semantic.SemanticIndex = semantic_indexes[i]; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = i; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); + struct ureg_src src; + struct ureg_dst dst; + + src = ureg_DECL_vs_input( ureg, + semantic_names[i], + semantic_indexes[i]); + + dst = ureg_DECL_output( ureg, + semantic_names[i], + semantic_indexes[i]); + + ureg_MOV( ureg, dst, src ); } - /* emit MOV instructions */ - for (i = 0; i < num_attribs; i++) { - /* MOVE out[i], in[i]; */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = i; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = i; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - } - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; + ureg_END( ureg ); - return pipe->create_vs_state(pipe, &shader); + return ureg_create_shader_and_destroy( ureg, pipe ); } @@ -158,99 +94,29 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, void * util_make_fragment_tex_shader(struct pipe_context *pipe) { - struct pipe_shader_state shader; - struct tgsi_token tokens[100]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti; - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare TEX[0] input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - /* XXX this could be linear... */ - decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* declare color[0] output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* declare sampler */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* TEX instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_TEX; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 2; - inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - inst.FullSrcRegisters[1].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - - return pipe->create_fs_state(pipe, &shader); + struct ureg_program *ureg; + struct ureg_src sampler; + struct ureg_src tex; + struct ureg_dst out; + + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; + + sampler = ureg_DECL_sampler( ureg ); + + tex = ureg_DECL_fs_input( ureg, + TGSI_SEMANTIC_GENERIC, 0, + TGSI_INTERPOLATE_PERSPECTIVE ); + + out = ureg_DECL_output( ureg, + TGSI_SEMANTIC_COLOR, + 0 ); + + ureg_TEX( ureg, out, TGSI_TEXTURE_2D, tex, sampler ); + ureg_END( ureg ); + + return ureg_create_shader_and_destroy( ureg, pipe ); } @@ -263,87 +129,23 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) void * util_make_fragment_passthrough_shader(struct pipe_context *pipe) { - struct pipe_shader_state shader; - struct tgsi_token tokens[40]; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti; - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - /* declare output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.DeclarationRange.First = - decl.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - Elements(tokens) - ti); - - - /* MOVE out[0], in[0]; */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - Elements(tokens) - ti ); - - assert(ti < Elements(tokens)); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - - return pipe->create_fs_state(pipe, &shader); + struct ureg_program *ureg; + struct ureg_src src; + struct ureg_dst dst; + + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; + + src = ureg_DECL_fs_input( ureg, TGSI_SEMANTIC_COLOR, 0, + TGSI_INTERPOLATE_PERSPECTIVE ); + + dst = ureg_DECL_output( ureg, TGSI_SEMANTIC_COLOR, 0 ); + + ureg_MOV( ureg, dst, src ); + ureg_END( ureg ); + + return ureg_create_shader_and_destroy( ureg, pipe ); } diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index 0d54299b28a..65bdd0df2b7 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -829,7 +829,7 @@ util_vsnprintf(char *str, size_t size, const char *format, va_list args) break; default: intptr = va_arg(args, int *); - *intptr = len; + *intptr = (int)len; break; } break; diff --git a/src/gallium/auxiliary/util/u_surface.h b/src/gallium/auxiliary/util/u_surface.h index a5b73cfc20a..ce84ed7ad06 100644 --- a/src/gallium/auxiliary/util/u_surface.h +++ b/src/gallium/auxiliary/util/u_surface.h @@ -37,6 +37,23 @@ struct pipe_texture; struct pipe_surface; +/** + * Are s1 and s2 the same surface? + * Surfaces are basically views into textures so check if the two surfaces + * name the same part of the same texture. + */ +static INLINE boolean +util_same_surface(const struct pipe_surface *s1, const struct pipe_surface *s2) +{ + return (s1->texture == s2->texture && + s1->face == s2->face && + s1->level == s2->level && + s1->zslice == s2->zslice); +} + + + + extern boolean util_create_rgba_surface(struct pipe_screen *screen, uint width, uint height, diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 9747a55cbfa..9e76cfbb053 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -62,7 +62,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt, if(!src) return; - pipe_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y); + util_copy_rect(dst, &pt->block, dst_stride, 0, 0, w, h, src, pt->stride, x, y); screen->transfer_unmap(screen, pt); } @@ -90,7 +90,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt, if(!dst) return; - pipe_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0); + util_copy_rect(dst, &pt->block, pt->stride, x, y, w, h, src, src_stride, 0, 0); screen->transfer_unmap(screen, pt); } @@ -153,7 +153,7 @@ a8r8g8b8_put_tile_rgba(unsigned *dst, } -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ +/*** PIPE_FORMAT_X8R8G8B8_UNORM ***/ static void x8r8g8b8_get_tile_rgba(const unsigned *src, @@ -1202,6 +1202,19 @@ pipe_put_tile_z(struct pipe_transfer *pt, } break; case PIPE_FORMAT_S8Z24_UNORM: + { + uint *pDest = (uint *) (map + y * pt->stride + x*4); + assert(pt->usage == PIPE_TRANSFER_READ_WRITE); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 24-bit Z, preserve stencil */ + pDest[j] = (pDest[j] & 0xff000000) | ptrc[j] >> 8; + } + pDest += pt->stride/4; + ptrc += srcStride; + } + } + break; case PIPE_FORMAT_X8Z24_UNORM: { uint *pDest = (uint *) (map + y * pt->stride + x*4); @@ -1216,13 +1229,26 @@ pipe_put_tile_z(struct pipe_transfer *pt, } break; case PIPE_FORMAT_Z24S8_UNORM: + { + uint *pDest = (uint *) (map + y * pt->stride + x*4); + assert(pt->usage == PIPE_TRANSFER_READ_WRITE); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 24-bit Z, preserve stencil */ + pDest[j] = (pDest[j] & 0xff) | (ptrc[j] & 0xffffff00); + } + pDest += pt->stride/4; + ptrc += srcStride; + } + } + break; case PIPE_FORMAT_Z24X8_UNORM: { uint *pDest = (uint *) (map + y * pt->stride + x*4); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 24-bit Z (0 stencil) */ - pDest[j] = ptrc[j] << 8; + pDest[j] = ptrc[j] & 0xffffff00; } pDest += pt->stride/4; ptrc += srcStride; diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 5268cbf79ce..c16cdd0b226 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -35,7 +35,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) #include <sys/time.h> #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include <windows.h> @@ -77,7 +77,7 @@ util_time_get_frequency(void) void util_time_get(struct util_time *t) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) gettimeofday(&t->tv, NULL); #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) LONGLONG temp; @@ -102,7 +102,7 @@ util_time_add(const struct util_time *t1, int64_t usecs, struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -124,7 +124,7 @@ int64_t util_time_diff(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) @@ -144,7 +144,7 @@ util_time_micros( void ) util_time_get(&t1); -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) return t1.tv.tv_usec + t1.tv.tv_sec*1000000LL; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) util_time_get_frequency(); @@ -166,7 +166,7 @@ static INLINE int util_time_compare(const struct util_time *t1, const struct util_time *t2) { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) if (t1->tv.tv_sec < t2->tv.tv_sec) return -1; else if(t1->tv.tv_sec > t2->tv.tv_sec) diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 6bca6077a2a..7a5c54d9b23 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -38,7 +38,7 @@ #include "pipe/p_config.h" -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) #include <time.h> /* timeval */ #include <unistd.h> /* usleep */ #endif @@ -58,7 +58,7 @@ extern "C" { */ struct util_time { -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) struct timeval tv; #else int64_t counter; @@ -89,7 +89,7 @@ util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr); -#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) +#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) #define util_time_sleep usleep #else void diff --git a/src/gallium/auxiliary/util/u_timed_winsys.c b/src/gallium/auxiliary/util/u_timed_winsys.c index 77b2a3a1c87..178acdca4df 100644 --- a/src/gallium/auxiliary/util/u_timed_winsys.c +++ b/src/gallium/auxiliary/util/u_timed_winsys.c @@ -212,13 +212,14 @@ timed_surface_buffer_create(struct pipe_winsys *winsys, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { struct pipe_winsys *backend = timed_winsys(winsys)->backend; uint64_t start = time_start(); struct pipe_buffer *ret = backend->surface_buffer_create( backend, width, height, - format, usage, stride ); + format, usage, tex_usage, stride ); time_finish(winsys, start, 7, __FUNCTION__); diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c index 2eb98068c86..c90425f3e54 100644 --- a/src/gallium/auxiliary/util/u_upload_mgr.c +++ b/src/gallium/auxiliary/util/u_upload_mgr.c @@ -83,7 +83,9 @@ my_buffer_write(struct pipe_screen *screen, assert(dirty_size >= size); assert(size); - map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe_buffer_map_range(screen, buf, offset, size, + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_FLUSH_EXPLICIT); if (map == NULL) return PIPE_ERROR_OUT_OF_MEMORY; diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fp.c b/src/gallium/drivers/cell/ppu/cell_gen_fp.c index 5a889a6119d..58a8b5d0b0f 100644 --- a/src/gallium/drivers/cell/ppu/cell_gen_fp.c +++ b/src/gallium/drivers/cell/ppu/cell_gen_fp.c @@ -1834,9 +1834,9 @@ emit_instruction(struct codegen *gen, case TGSI_OPCODE_ENDIF: return emit_ENDIF(gen, inst); - case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_BGNLOOP: return emit_BGNLOOP(gen, inst); - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_ENDLOOP: return emit_ENDLOOP(gen, inst); case TGSI_OPCODE_BRK: return emit_BRK(gen, inst); @@ -1875,9 +1875,9 @@ emit_immediate(struct codegen *gen, const struct tgsi_full_immediate *immed) assert(gen->num_imm < MAX_TEMPS); for (ch = 0; ch < 4; ch++) { - float val = immed->u.ImmediateFloat32[ch].Float; + float val = immed->u[ch].Float; - if (ch > 0 && val == immed->u.ImmediateFloat32[ch - 1].Float) { + if (ch > 0 && val == immed->u[ch - 1].Float) { /* re-use previous register */ gen->imm_regs[gen->num_imm][ch] = gen->imm_regs[gen->num_imm][ch - 1]; } diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 512d85d3525..bd48ce70050 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -86,6 +86,8 @@ cell_get_param(struct pipe_screen *screen, int param) return 1; /* XXX not really true */ case PIPE_CAP_TEXTURE_MIRROR_CLAMP: return 0; /* XXX to do */ + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 1; default: return 0; } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e26594448f0..6a63a0e6ced 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -44,13 +44,6 @@ -static unsigned -minify(unsigned d) -{ - return MAX2(1, d>>1); -} - - static void cell_texture_layout(struct cell_texture *ct) { @@ -424,7 +417,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) if (!ctrans->map) return NULL; /* out of memory */ - if (transfer->usage & PIPE_TRANSFER_READ) { + if (transfer->usage == PIPE_TRANSFER_READ || + transfer->usage == PIPE_TRANSFER_READ_WRITE) { /* need to untwiddle the texture to make a linear version */ const uint bpp = pf_get_size(ct->base.format); if (bpp == 4) { @@ -465,7 +459,8 @@ cell_transfer_unmap(struct pipe_screen *screen, PIPE_BUFFER_USAGE_CPU_READ); } - if (transfer->usage & PIPE_TRANSFER_WRITE) { + if (transfer->usage == PIPE_TRANSFER_WRITE || + transfer->usage == PIPE_TRANSFER_READ_WRITE) { /* The user wrote new texture data into the mapped buffer. * We need to convert the new linear data into the twiddled/tiled format. */ diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index e27df2dfb38..6db9501128c 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -952,7 +952,6 @@ exec_instruction( break; case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ FETCH( &r[0], 0, CHAN_X ); r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -961,7 +960,6 @@ exec_instruction( break; case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ FETCH( &r[0], 0, CHAN_X ); r[0].q = micro_sqrt(r[0].q); r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); @@ -1115,7 +1113,6 @@ exec_instruction( break; case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); @@ -1136,8 +1133,7 @@ exec_instruction( } break; - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ + case TGSI_OPCODE_LRP: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); @@ -1158,21 +1154,11 @@ exec_instruction( ASSERT (0); break; - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ + case TGSI_OPCODE_DP2A: ASSERT (0); break; - case TGSI_OPCODE_INDEX: - ASSERT (0); - break; - - case TGSI_OPCODE_NEGATE: - ASSERT (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ + case TGSI_OPCODE_FRC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); r[0].q = micro_frc(r[0].q); @@ -1184,8 +1170,7 @@ exec_instruction( ASSERT (0); break; - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ + case TGSI_OPCODE_FLR: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); r[0].q = micro_flr(r[0].q); @@ -1201,8 +1186,7 @@ exec_instruction( } break; - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ + case TGSI_OPCODE_EX2: FETCH(&r[0], 0, CHAN_X); r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q); @@ -1212,8 +1196,7 @@ exec_instruction( } break; - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ + case TGSI_OPCODE_LG2: FETCH( &r[0], 0, CHAN_X ); r[0].q = micro_lg2(r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1221,8 +1204,7 @@ exec_instruction( } break; - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ + case TGSI_OPCODE_POW: FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); @@ -1233,7 +1215,7 @@ exec_instruction( } break; - case TGSI_OPCODE_CROSSPRODUCT: + case TGSI_OPCODE_XPD: /* TGSI_OPCODE_XPD */ FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); @@ -1275,10 +1257,6 @@ exec_instruction( } break; - case TGSI_OPCODE_MULTIPLYMATRIX: - ASSERT (0); - break; - case TGSI_OPCODE_ABS: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); @@ -1780,9 +1758,9 @@ exec_instruction( mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_BGNLOOP: /* push LoopMask and ContMasks */ ASSERT(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; @@ -1790,9 +1768,9 @@ exec_instruction( mach->ContStack[mach->ContStackTop++] = mach->ContMask; break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_ENDLOOP: /* Restore ContMask, but don't pop */ ASSERT(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index a433cf054de..c6e68ea38a2 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -50,8 +50,8 @@ i915_batchbuffer_check( struct i915_batchbuffer *batch, size_t dwords, size_t relocs ) { - /** TODO JB: Check relocs */ - return dwords * 4 <= batch->size - (batch->ptr - batch->map); + return dwords * 4 <= batch->size - (batch->ptr - batch->map) && + relocs <= (batch->max_relocs - batch->relocs); } static INLINE size_t diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index ccf9bb31fb0..bf69c8e9f53 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -142,10 +142,14 @@ i915_is_texture_referenced( struct pipe_context *pipe, unsigned face, unsigned level) { /** - * FIXME: Optimize. + * FIXME: Return the corrent result. We can't alays return referenced + * since it causes a double flush within the vbo module. */ - +#if 0 return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +#else + return 0; +#endif } static unsigned int @@ -153,10 +157,14 @@ i915_is_buffer_referenced( struct pipe_context *pipe, struct pipe_buffer *buf) { /** - * FIXME: Optimize. + * FIXME: Return the corrent result. We can't alays return referenced + * since it causes a double flush within the vbo module. */ - +#if 0 return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; +#else + return 0; +#endif } diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 961c1bf2134..89504ced276 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -975,8 +975,9 @@ i915_translate_instructions(struct i915_fp_compile *p, = &parse.FullToken.FullImmediate; const uint pos = p->num_immediates++; uint j; + assert( imm->Immediate.NrTokens <= 4 + 1 ); for (j = 0; j < imm->Immediate.NrTokens - 1; j++) { - p->immediates[pos][j] = imm->u.ImmediateFloat32[j].Float; + p->immediates[pos][j] = imm->u[j].Float; } } break; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index f4aa8e60d81..a3de38d5860 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -232,6 +232,8 @@ i915_get_tex_transfer(struct pipe_screen *screen, if (trans) { pipe_texture_reference(&trans->base.texture, texture); trans->base.format = trans->base.format; + trans->base.x = x; + trans->base.y = y; trans->base.width = w; trans->base.height = h; trans->base.block = texture->block; diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index ca8e87af8d1..03f0e14e7c8 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -72,11 +72,6 @@ static const int step_offsets[6][2] = { {-1, 1} }; -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - static unsigned power_of_two(unsigned x) { @@ -160,10 +155,10 @@ i915_miptree_set_image_offset(struct i915_texture *tex, /** - * Special case to deal with display targets. + * Special case to deal with scanout textures. */ static boolean -i915_displaytarget_layout(struct i915_texture *tex) +i915_scanout_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; @@ -177,9 +172,13 @@ i915_displaytarget_layout(struct i915_texture *tex) i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); if (tex->base.width[0] >= 128) { +#if 0 tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); +#else + tex->stride = 2048 * 4; /* TODO fix when backend is smarter */ +#endif tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); -#if 0 /* used for tiled display targets */ +#if 0 /* used for tiled textures */ tex->tiled = 1; #endif } else { @@ -209,9 +208,9 @@ i945_miptree_layout_2d( struct i915_texture *tex ) unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; - /* used for tiled display targets */ - if (0) - if (i915_displaytarget_layout(tex)) + /* used for scanouts that need special layouts */ + if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) + if (i915_scanout_layout(tex)) return; tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); @@ -584,6 +583,7 @@ i915_texture_create(struct pipe_screen *screen, struct i915_screen *i915screen = i915_screen(screen); struct i915_texture *tex = CALLOC_STRUCT(i915_texture); size_t tex_size; + unsigned buf_usage = 0; if (!tex) return NULL; @@ -605,9 +605,11 @@ i915_texture_create(struct pipe_screen *screen, tex_size = tex->stride * tex->total_nblocksy; - tex->buffer = screen->buffer_create(screen, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex_size); + buf_usage = PIPE_BUFFER_USAGE_PIXEL; + if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) + buf_usage |= I915_BUFFER_USAGE_SCANOUT; + + tex->buffer = screen->buffer_create(screen, 64, buf_usage, tex_size); if (!tex->buffer) goto fail; diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index ff5b34f193a..711db91c367 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -109,6 +109,7 @@ struct i915_winsys { #define I915_BUFFER_ACCESS_READ 0x2 #define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) +#define I915_BUFFER_USAGE_SCANOUT (PIPE_BUFFER_USAGE_CUSTOM << 1) /** diff --git a/src/gallium/drivers/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c index 648aaa0da52..49d80cb41c5 100644 --- a/src/gallium/drivers/i965simple/brw_draw.c +++ b/src/gallium/drivers/i965simple/brw_draw.c @@ -35,6 +35,7 @@ #include "pipe/p_context.h" #include "pipe/internal/p_winsys_screen.h" +#include "util/u_prim.h" static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = { _3DPRIM_POINTLIST, @@ -50,20 +51,6 @@ static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = { }; -static const int reduced_prim[PIPE_PRIM_POLYGON+1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - - /* When the primitive changes, set a state bit and re-validate. Not * the nicest and would rather deal with this by having all the * programs be immune to the active primitive (ie. cope with all @@ -85,8 +72,8 @@ static void brw_set_prim(struct brw_context *brw, int prim) brw->primitive = prim; brw->state.dirty.brw |= BRW_NEW_PRIMITIVE; - if (reduced_prim[prim] != brw->reduced_primitive) { - brw->reduced_primitive = reduced_prim[prim]; + if (u_reduced_prim(prim) != brw->reduced_primitive) { + brw->reduced_primitive = u_reduced_prim(prim); brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; } diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 511779dbfa0..724a69b2eee 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -60,7 +60,7 @@ brw_surface_copy(struct pipe_context *pipe, src, PIPE_BUFFER_USAGE_CPU_READ ); - pipe_copy_rect(dst_map, + util_copy_rect(dst_map, &dst->block, dst->stride, dstx, dsty, @@ -99,7 +99,7 @@ brw_surface_fill(struct pipe_context *pipe, dst, PIPE_BUFFER_USAGE_CPU_WRITE ); - pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); + util_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); pipe->screen->surface_unmap(pipe->screen, dst); } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 8aea8c05581..998ffaeac4a 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -65,11 +65,6 @@ unsigned intel_compressed_alignment(unsigned internalFormat) } #endif -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - static void intel_miptree_set_image_offset(struct brw_texture *tex, unsigned level, diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index e03d6534821..3ee82d95b3a 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -1294,10 +1294,10 @@ void brw_vs_emit(struct brw_vs_compile *c) case TGSI_TOKEN_TYPE_IMMEDIATE: { struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate; assert(imm->Immediate.NrTokens == 4 + 1); - c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u[0].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u[1].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u[2].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u[3].Float; c->prog_data.num_imm++; } break; diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c index ab6410aa607..db759639328 100644 --- a/src/gallium/drivers/i965simple/brw_wm_glsl.c +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -947,7 +947,7 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c, #endif break; - case TGSI_OPCODE_LOOP: + case TGSI_OPCODE_BGNFOR: c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8); break; case TGSI_OPCODE_BRK: @@ -958,11 +958,11 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c, brw_CONT(p); brw_set_predicate_control(p, BRW_PREDICATE_NONE); break; - case TGSI_OPCODE_ENDLOOP: + case TGSI_OPCODE_ENDFOR: c->loop_insn--; c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]); /* patch all the BREAK instructions from - last BEGINLOOP */ + last BGNFOR */ while (c->inst0 > c->loop_inst[c->loop_insn]) { c->inst0--; if (c->inst0->header.opcode == BRW_OPCODE_BREAK) { diff --git a/src/gallium/drivers/identity/Makefile b/src/gallium/drivers/identity/Makefile new file mode 100644 index 00000000000..e32b9102e59 --- /dev/null +++ b/src/gallium/drivers/identity/Makefile @@ -0,0 +1,12 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = identity + +C_SOURCES = \ + id_objects.c \ + id_context.c \ + id_screen.c \ + id_drm.c + +include ../../Makefile.template diff --git a/src/gallium/drivers/identity/SConscript b/src/gallium/drivers/identity/SConscript new file mode 100644 index 00000000000..7f079dd0a8b --- /dev/null +++ b/src/gallium/drivers/identity/SConscript @@ -0,0 +1,13 @@ +Import('*') + +env = env.Clone() + +identity = env.ConvenienceLibrary( + target = 'identity', + source = [ + 'id_screen.c', + 'id_context.c', + 'id_objects.c', + ]) + +Export('identity') diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c new file mode 100644 index 00000000000..4e700089e33 --- /dev/null +++ b/src/gallium/drivers/identity/id_context.c @@ -0,0 +1,719 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_context.h" +#include "util/u_memory.h" + +#include "id_public.h" +#include "id_context.h" +#include "id_objects.h" + + +static void +identity_destroy(struct pipe_context *_pipe) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->destroy(pipe); + + free(id_pipe); +} + +static void +identity_set_edgeflags(struct pipe_context *_pipe, + const unsigned *bitfield) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_edgeflags(pipe, + bitfield); +} + +static boolean +identity_draw_arrays(struct pipe_context *_pipe, + unsigned prim, + unsigned start, + unsigned count) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->draw_arrays(pipe, + prim, + start, + count); +} + +static boolean +identity_draw_elements(struct pipe_context *_pipe, + struct pipe_buffer *_indexBuffer, + unsigned indexSize, + unsigned prim, + unsigned start, + unsigned count) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_buffer *id_buffer = identity_buffer(_indexBuffer); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_buffer *indexBuffer = id_buffer->buffer; + + return pipe->draw_elements(pipe, + indexBuffer, + indexSize, + prim, + start, + count); +} + +static boolean +identity_draw_range_elements(struct pipe_context *_pipe, + struct pipe_buffer *_indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_buffer *id_buffer = identity_buffer(_indexBuffer); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_buffer *indexBuffer = id_buffer->buffer; + + return pipe->draw_range_elements(pipe, + indexBuffer, + indexSize, + minIndex, + maxIndex, + mode, + start, + count); +} + +static struct pipe_query * +identity_create_query(struct pipe_context *_pipe, + unsigned query_type) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_query(pipe, + query_type); +} + +static void +identity_destroy_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->destroy_query(pipe, + query); +} + +static void +identity_begin_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->begin_query(pipe, + query); +} + +static void +identity_end_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->end_query(pipe, + query); +} + +static boolean +identity_get_query_result(struct pipe_context *_pipe, + struct pipe_query *query, + boolean wait, + uint64_t *result) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->get_query_result(pipe, + query, + wait, + result); +} + +static void * +identity_create_blend_state(struct pipe_context *_pipe, + const struct pipe_blend_state *blend) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_blend_state(pipe, + blend); +} + +static void +identity_bind_blend_state(struct pipe_context *_pipe, + void *blend) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_blend_state(pipe, + blend); +} + +static void +identity_delete_blend_state(struct pipe_context *_pipe, + void *blend) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_blend_state(pipe, + blend); +} + +static void * +identity_create_sampler_state(struct pipe_context *_pipe, + const struct pipe_sampler_state *sampler) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_sampler_state(pipe, + sampler); +} + +static void +identity_bind_sampler_states(struct pipe_context *_pipe, + unsigned num, + void **samplers) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_sampler_states(pipe, + num, + samplers); +} + +static void +identity_delete_sampler_state(struct pipe_context *_pipe, + void *sampler) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_sampler_state(pipe, + sampler); +} + +static void * +identity_create_rasterizer_state(struct pipe_context *_pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_rasterizer_state(pipe, + rasterizer); +} + +static void +identity_bind_rasterizer_state(struct pipe_context *_pipe, + void *rasterizer) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_rasterizer_state(pipe, + rasterizer); +} + +static void +identity_delete_rasterizer_state(struct pipe_context *_pipe, + void *rasterizer) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_rasterizer_state(pipe, + rasterizer); +} + +static void * +identity_create_depth_stencil_alpha_state(struct pipe_context *_pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil_alpha) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_depth_stencil_alpha_state(pipe, + depth_stencil_alpha); +} + +static void +identity_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, + void *depth_stencil_alpha) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_depth_stencil_alpha_state(pipe, + depth_stencil_alpha); +} + +static void +identity_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, + void *depth_stencil_alpha) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_depth_stencil_alpha_state(pipe, + depth_stencil_alpha); +} + +static void * +identity_create_fs_state(struct pipe_context *_pipe, + const struct pipe_shader_state *fs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_fs_state(pipe, + fs); +} + +static void +identity_bind_fs_state(struct pipe_context *_pipe, + void *fs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_fs_state(pipe, + fs); +} + +static void +identity_delete_fs_state(struct pipe_context *_pipe, + void *fs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_fs_state(pipe, + fs); +} + +static void * +identity_create_vs_state(struct pipe_context *_pipe, + const struct pipe_shader_state *vs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + return pipe->create_vs_state(pipe, + vs); +} + +static void +identity_bind_vs_state(struct pipe_context *_pipe, + void *vs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->bind_vs_state(pipe, + vs); +} + +static void +identity_delete_vs_state(struct pipe_context *_pipe, + void *vs) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->delete_vs_state(pipe, + vs); +} + +static void +identity_set_blend_color(struct pipe_context *_pipe, + const struct pipe_blend_color *blend_color) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_blend_color(pipe, + blend_color); +} + +static void +identity_set_clip_state(struct pipe_context *_pipe, + const struct pipe_clip_state *clip) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_clip_state(pipe, + clip); +} + +static void +identity_set_constant_buffer(struct pipe_context *_pipe, + uint shader, + uint index, + const struct pipe_constant_buffer *_buffer) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_constant_buffer unwrapped_buffer; + struct pipe_constant_buffer *buffer = NULL; + + /* unwrap the input state */ + if (_buffer) { + unwrapped_buffer.buffer = identity_buffer_unwrap(_buffer->buffer); + buffer = &unwrapped_buffer; + } + + pipe->set_constant_buffer(pipe, + shader, + index, + buffer); +} + +static void +identity_set_framebuffer_state(struct pipe_context *_pipe, + const struct pipe_framebuffer_state *_state) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_framebuffer_state unwrapped_state; + struct pipe_framebuffer_state *state = NULL; + unsigned i; + + /* unwrap the input state */ + if (_state) { + memcpy(&unwrapped_state, _state, sizeof(unwrapped_state)); + for(i = 0; i < _state->nr_cbufs; i++) + unwrapped_state.cbufs[i] = identity_surface_unwrap(_state->cbufs[i]); + for (; i < PIPE_MAX_COLOR_BUFS; i++) + unwrapped_state.cbufs[i] = NULL; + unwrapped_state.zsbuf = identity_surface_unwrap(_state->zsbuf); + state = &unwrapped_state; + } + + pipe->set_framebuffer_state(pipe, + state); +} + +static void +identity_set_polygon_stipple(struct pipe_context *_pipe, + const struct pipe_poly_stipple *poly_stipple) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_polygon_stipple(pipe, + poly_stipple); +} + +static void +identity_set_scissor_state(struct pipe_context *_pipe, + const struct pipe_scissor_state *scissor) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_scissor_state(pipe, + scissor); +} + +static void +identity_set_viewport_state(struct pipe_context *_pipe, + const struct pipe_viewport_state *viewport) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_viewport_state(pipe, + viewport); +} + +static void +identity_set_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **_textures) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS]; + struct pipe_texture **textures = NULL; + unsigned i; + + if (_textures) { + for (i = 0; i < num_textures; i++) + unwrapped_textures[i] = identity_texture_unwrap(_textures[i]); + for (; i < PIPE_MAX_SAMPLERS; i++) + unwrapped_textures[i] = NULL; + + textures = unwrapped_textures; + } + + pipe->set_sampler_textures(pipe, + num_textures, + textures); +} + +static void +identity_set_vertex_buffers(struct pipe_context *_pipe, + unsigned num_buffers, + const struct pipe_vertex_buffer *_buffers) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_vertex_buffer unwrapped_buffers[PIPE_MAX_SHADER_INPUTS]; + struct pipe_vertex_buffer *buffers = NULL; + unsigned i; + + if (num_buffers) { + memcpy(unwrapped_buffers, _buffers, num_buffers * sizeof(*_buffers)); + for (i = 0; i < num_buffers; i++) + unwrapped_buffers[i].buffer = identity_buffer_unwrap(_buffers[i].buffer); + buffers = unwrapped_buffers; + } + + pipe->set_vertex_buffers(pipe, + num_buffers, + buffers); +} + +static void +identity_set_vertex_elements(struct pipe_context *_pipe, + unsigned num_elements, + const struct pipe_vertex_element *vertex_elements) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->set_vertex_elements(pipe, + num_elements, + vertex_elements); +} + +static void +identity_surface_copy(struct pipe_context *_pipe, + struct pipe_surface *_dst, + unsigned dstx, + unsigned dsty, + struct pipe_surface *_src, + unsigned srcx, + unsigned srcy, + unsigned width, + unsigned height) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_surface *id_surface_dst = identity_surface(_dst); + struct identity_surface *id_surface_src = identity_surface(_src); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_surface *dst = id_surface_dst->surface; + struct pipe_surface *src = id_surface_src->surface; + + pipe->surface_copy(pipe, + dst, + dstx, + dsty, + src, + srcx, + srcy, + width, + height); +} + +static void +identity_surface_fill(struct pipe_context *_pipe, + struct pipe_surface *_dst, + unsigned dstx, + unsigned dsty, + unsigned width, + unsigned height, + unsigned value) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_surface *id_surface_dst = identity_surface(_dst); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_surface *dst = id_surface_dst->surface; + + pipe->surface_fill(pipe, + dst, + dstx, + dsty, + width, + height, + value); +} + +static void +identity_clear(struct pipe_context *_pipe, + unsigned buffers, + const float *rgba, + double depth, + unsigned stencil) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->clear(pipe, + buffers, + rgba, + depth, + stencil); +} + +static void +identity_flush(struct pipe_context *_pipe, + unsigned flags, + struct pipe_fence_handle **fence) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct pipe_context *pipe = id_pipe->pipe; + + pipe->flush(pipe, + flags, + fence); +} + +static unsigned int +identity_is_texture_referenced(struct pipe_context *_pipe, + struct pipe_texture *_texture, + unsigned face, + unsigned level) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_texture *id_texture = identity_texture(_texture); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_texture *texture = id_texture->texture; + + return pipe->is_texture_referenced(pipe, + texture, + face, + level); +} + +static unsigned int +identity_is_buffer_referenced(struct pipe_context *_pipe, + struct pipe_buffer *_buffer) +{ + struct identity_context *id_pipe = identity_context(_pipe); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_context *pipe = id_pipe->pipe; + struct pipe_buffer *buffer = id_buffer->buffer; + + return pipe->is_buffer_referenced(pipe, + buffer); +} + +struct pipe_context * +identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe) +{ + struct identity_context *id_pipe; + (void)identity_screen(_screen); + + id_pipe = CALLOC_STRUCT(identity_context); + if (!id_pipe) { + return NULL; + } + + id_pipe->base.winsys = NULL; + id_pipe->base.screen = _screen; + id_pipe->base.priv = pipe->priv; + id_pipe->base.draw = NULL; + + id_pipe->base.destroy = identity_destroy; + id_pipe->base.set_edgeflags = identity_set_edgeflags; + id_pipe->base.draw_arrays = identity_draw_arrays; + id_pipe->base.draw_elements = identity_draw_elements; + id_pipe->base.draw_range_elements = identity_draw_range_elements; + id_pipe->base.create_query = identity_create_query; + id_pipe->base.destroy_query = identity_destroy_query; + id_pipe->base.begin_query = identity_begin_query; + id_pipe->base.end_query = identity_end_query; + id_pipe->base.get_query_result = identity_get_query_result; + id_pipe->base.create_blend_state = identity_create_blend_state; + id_pipe->base.bind_blend_state = identity_bind_blend_state; + id_pipe->base.delete_blend_state = identity_delete_blend_state; + id_pipe->base.create_sampler_state = identity_create_sampler_state; + id_pipe->base.bind_sampler_states = identity_bind_sampler_states; + id_pipe->base.delete_sampler_state = identity_delete_sampler_state; + id_pipe->base.create_rasterizer_state = identity_create_rasterizer_state; + id_pipe->base.bind_rasterizer_state = identity_bind_rasterizer_state; + id_pipe->base.delete_rasterizer_state = identity_delete_rasterizer_state; + id_pipe->base.create_depth_stencil_alpha_state = identity_create_depth_stencil_alpha_state; + id_pipe->base.bind_depth_stencil_alpha_state = identity_bind_depth_stencil_alpha_state; + id_pipe->base.delete_depth_stencil_alpha_state = identity_delete_depth_stencil_alpha_state; + id_pipe->base.create_fs_state = identity_create_fs_state; + id_pipe->base.bind_fs_state = identity_bind_fs_state; + id_pipe->base.delete_fs_state = identity_delete_fs_state; + id_pipe->base.create_vs_state = identity_create_vs_state; + id_pipe->base.bind_vs_state = identity_bind_vs_state; + id_pipe->base.delete_vs_state = identity_delete_vs_state; + id_pipe->base.set_blend_color = identity_set_blend_color; + id_pipe->base.set_clip_state = identity_set_clip_state; + id_pipe->base.set_constant_buffer = identity_set_constant_buffer; + id_pipe->base.set_framebuffer_state = identity_set_framebuffer_state; + id_pipe->base.set_polygon_stipple = identity_set_polygon_stipple; + id_pipe->base.set_scissor_state = identity_set_scissor_state; + id_pipe->base.set_viewport_state = identity_set_viewport_state; + id_pipe->base.set_sampler_textures = identity_set_sampler_textures; + id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers; + id_pipe->base.set_vertex_elements = identity_set_vertex_elements; + id_pipe->base.surface_copy = identity_surface_copy; + id_pipe->base.surface_fill = identity_surface_fill; + id_pipe->base.clear = identity_clear; + id_pipe->base.flush = identity_flush; + id_pipe->base.is_texture_referenced = identity_is_texture_referenced; + id_pipe->base.is_buffer_referenced = identity_is_buffer_referenced; + + id_pipe->pipe = pipe; + + return &id_pipe->base; +} diff --git a/src/gallium/drivers/identity/id_context.h b/src/gallium/drivers/identity/id_context.h new file mode 100644 index 00000000000..75b73fc7df6 --- /dev/null +++ b/src/gallium/drivers/identity/id_context.h @@ -0,0 +1,48 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ID_CONTEXT_H +#define ID_CONTEXT_H + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + +struct identity_context { + struct pipe_context base; /**< base class */ + + struct pipe_context *pipe; +}; + + +static INLINE struct identity_context * +identity_context(struct pipe_context *pipe) +{ + return (struct identity_context *)pipe; +} + +#endif /* ID_CONTEXT_H */ diff --git a/src/gallium/drivers/identity/id_drm.c b/src/gallium/drivers/identity/id_drm.c new file mode 100644 index 00000000000..e5342ac06e3 --- /dev/null +++ b/src/gallium/drivers/identity/id_drm.c @@ -0,0 +1,187 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "state_tracker/drm_api.h" + +#include "util/u_memory.h" +#include "identity/id_drm.h" +#include "identity/id_public.h" +#include "identity/id_screen.h" +#include "identity/id_objects.h" + +struct identity_drm_api +{ + struct drm_api base; + + struct drm_api *api; +}; + +static INLINE struct identity_drm_api * +identity_drm_api(struct drm_api *_api) +{ + return (struct identity_drm_api *)_api; +} + +static struct pipe_screen * +identity_drm_create_screen(struct drm_api *_api, int fd, + struct drm_create_screen_arg *arg) +{ + struct identity_drm_api *id_api = identity_drm_api(_api); + struct drm_api *api = id_api->api; + struct pipe_screen *screen; + + if (arg && arg->mode != DRM_CREATE_NORMAL) + return NULL; + + screen = api->create_screen(api, fd, arg); + + return identity_screen_create(screen); +} + +static struct pipe_context * +identity_drm_create_context(struct drm_api *_api, + struct pipe_screen *_screen) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_drm_api *id_api = identity_drm_api(_api); + struct pipe_screen *screen = id_screen->screen; + struct drm_api *api = id_api->api; + struct pipe_context *pipe; + + pipe = api->create_context(api, screen); + + pipe = identity_context_create(_screen, pipe); + + return pipe; +} + +static boolean +identity_drm_buffer_from_texture(struct drm_api *_api, + struct pipe_texture *_texture, + struct pipe_buffer **_buffer, + unsigned *stride) +{ + struct identity_texture *id_texture = identity_texture(_texture); + struct identity_drm_api *id_api = identity_drm_api(_api); + struct pipe_texture *texture = id_texture->texture; + struct drm_api *api = id_api->api; + struct pipe_buffer *buffer = NULL; + boolean result; + + result = api->buffer_from_texture(api, texture, &buffer, stride); + + if (result && _buffer) + buffer = identity_buffer_create(identity_screen(texture->screen), buffer); + + if (_buffer) + *_buffer = buffer; + else + pipe_buffer_reference(&buffer, NULL); + + return result; +} + +static struct pipe_buffer * +identity_drm_buffer_from_handle(struct drm_api *_api, + struct pipe_screen *_screen, + const char *name, + unsigned handle) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_drm_api *id_api = identity_drm_api(_api); + struct pipe_screen *screen = id_screen->screen; + struct drm_api *api = id_api->api; + struct pipe_buffer *result; + + result = api->buffer_from_handle(api, screen, name, handle); + + result = identity_buffer_create(identity_screen(_screen), result); + + return result; +} + +static boolean +identity_drm_handle_from_buffer(struct drm_api *_api, + struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct identity_drm_api *id_api = identity_drm_api(_api); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + struct drm_api *api = id_api->api; + + return api->handle_from_buffer(api, screen, buffer, handle); +} + +static boolean +identity_drm_global_handle_from_buffer(struct drm_api *_api, + struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct identity_drm_api *id_api = identity_drm_api(_api); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + struct drm_api *api = id_api->api; + + return api->global_handle_from_buffer(api, screen, buffer, handle); +} + +static void +identity_drm_destroy(struct drm_api *_api) +{ + struct identity_drm_api *id_api = identity_drm_api(_api); + struct drm_api *api = id_api->api; + api->destroy(api); + + free(id_api); +} + +struct drm_api * +identity_drm_create(struct drm_api *api) +{ + struct identity_drm_api *id_api = CALLOC_STRUCT(identity_drm_api); + + if (!id_api) + return NULL; + + id_api->base.create_screen = identity_drm_create_screen; + id_api->base.create_context = identity_drm_create_context; + id_api->base.buffer_from_texture = identity_drm_buffer_from_texture; + id_api->base.buffer_from_handle = identity_drm_buffer_from_handle; + id_api->base.handle_from_buffer = identity_drm_handle_from_buffer; + id_api->base.global_handle_from_buffer = identity_drm_global_handle_from_buffer; + id_api->base.destroy = identity_drm_destroy; + id_api->api = api; + + return &id_api->base; +} diff --git a/src/gallium/drivers/identity/id_drm.h b/src/gallium/drivers/identity/id_drm.h new file mode 100644 index 00000000000..cf2ad2ce07f --- /dev/null +++ b/src/gallium/drivers/identity/id_drm.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ID_DRM_H +#define ID_DRM_H + +struct drm_api; + +struct drm_api* identity_drm_create(struct drm_api *api); + +#endif /* ID_DRM_H */ diff --git a/src/gallium/drivers/identity/id_objects.c b/src/gallium/drivers/identity/id_objects.c new file mode 100644 index 00000000000..e893e599408 --- /dev/null +++ b/src/gallium/drivers/identity/id_objects.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "util/u_memory.h" + +#include "id_public.h" +#include "id_screen.h" +#include "id_objects.h" + +struct pipe_buffer * +identity_buffer_create(struct identity_screen *id_screen, + struct pipe_buffer *buffer) +{ + struct identity_buffer *id_buffer; + + if(!buffer) + goto error; + + assert(buffer->screen == id_screen->screen); + + id_buffer = CALLOC_STRUCT(identity_buffer); + if(!id_buffer) + goto error; + + memcpy(&id_buffer->base, buffer, sizeof(struct pipe_buffer)); + + pipe_reference_init(&id_buffer->base.reference, 1); + id_buffer->base.screen = &id_screen->base; + id_buffer->buffer = buffer; + + return &id_buffer->base; + +error: + pipe_buffer_reference(&buffer, NULL); + return NULL; +} + +void +identity_buffer_destroy(struct identity_buffer *id_buffer) +{ + pipe_buffer_reference(&id_buffer->buffer, NULL); + FREE(id_buffer); +} + + +struct pipe_texture * +identity_texture_create(struct identity_screen *id_screen, + struct pipe_texture *texture) +{ + struct identity_texture *id_texture; + + if(!texture) + goto error; + + assert(texture->screen == id_screen->screen); + + id_texture = CALLOC_STRUCT(identity_texture); + if(!id_texture) + goto error; + + memcpy(&id_texture->base, texture, sizeof(struct pipe_texture)); + + pipe_reference_init(&id_texture->base.reference, 1); + id_texture->base.screen = &id_screen->base; + id_texture->texture = texture; + + return &id_texture->base; + +error: + pipe_texture_reference(&texture, NULL); + return NULL; +} + +void +identity_texture_destroy(struct identity_texture *id_texture) +{ + pipe_texture_reference(&id_texture->texture, NULL); + FREE(id_texture); +} + + +struct pipe_surface * +identity_surface_create(struct identity_texture *id_texture, + struct pipe_surface *surface) +{ + struct identity_surface *id_surface; + + if(!surface) + goto error; + + assert(surface->texture == id_texture->texture); + + id_surface = CALLOC_STRUCT(identity_surface); + if(!id_surface) + goto error; + + memcpy(&id_surface->base, surface, sizeof(struct pipe_surface)); + + pipe_reference_init(&id_surface->base.reference, 1); + id_surface->base.texture = NULL; + pipe_texture_reference(&id_surface->base.texture, &id_texture->base); + id_surface->surface = surface; + + return &id_surface->base; + +error: + pipe_surface_reference(&surface, NULL); + return NULL; +} + +void +identity_surface_destroy(struct identity_surface *id_surface) +{ + pipe_texture_reference(&id_surface->base.texture, NULL); + pipe_surface_reference(&id_surface->surface, NULL); + FREE(id_surface); +} + + +struct pipe_transfer * +identity_transfer_create(struct identity_texture *id_texture, + struct pipe_transfer *transfer) +{ + struct identity_transfer *id_transfer; + + if(!transfer) + goto error; + + assert(transfer->texture == id_texture->texture); + + id_transfer = CALLOC_STRUCT(identity_transfer); + if(!id_transfer) + goto error; + + memcpy(&id_transfer->base, transfer, sizeof(struct pipe_transfer)); + + id_transfer->base.texture = NULL; + pipe_texture_reference(&id_transfer->base.texture, &id_texture->base); + id_transfer->transfer = transfer; + assert(id_transfer->base.texture == &id_texture->base); + + return &id_transfer->base; + +error: + transfer->texture->screen->tex_transfer_destroy(transfer); + return NULL; +} + +void +identity_transfer_destroy(struct identity_transfer *id_transfer) +{ + struct identity_screen *id_screen = identity_screen(id_transfer->base.texture->screen); + struct pipe_screen *screen = id_screen->screen; + + pipe_texture_reference(&id_transfer->base.texture, NULL); + screen->tex_transfer_destroy(id_transfer->transfer); + FREE(id_transfer); +} diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h new file mode 100644 index 00000000000..ce58faa3c7c --- /dev/null +++ b/src/gallium/drivers/identity/id_objects.h @@ -0,0 +1,169 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ID_OBJECTS_H +#define ID_OBJECTS_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + +#include "id_screen.h" + + +struct identity_buffer +{ + struct pipe_buffer base; + + struct pipe_buffer *buffer; +}; + + +struct identity_texture +{ + struct pipe_texture base; + + struct pipe_texture *texture; +}; + + +struct identity_surface +{ + struct pipe_surface base; + + struct pipe_surface *surface; +}; + + +struct identity_transfer +{ + struct pipe_transfer base; + + struct pipe_transfer *transfer; +}; + + +static INLINE struct identity_buffer * +identity_buffer(struct pipe_buffer *_buffer) +{ + if(!_buffer) + return NULL; + (void)identity_screen(_buffer->screen); + return (struct identity_buffer *)_buffer; +} + +static INLINE struct identity_texture * +identity_texture(struct pipe_texture *_texture) +{ + if(!_texture) + return NULL; + (void)identity_screen(_texture->screen); + return (struct identity_texture *)_texture; +} + +static INLINE struct identity_surface * +identity_surface(struct pipe_surface *_surface) +{ + if(!_surface) + return NULL; + (void)identity_texture(_surface->texture); + return (struct identity_surface *)_surface; +} + +static INLINE struct identity_transfer * +identity_transfer(struct pipe_transfer *_transfer) +{ + if(!_transfer) + return NULL; + (void)identity_texture(_transfer->texture); + return (struct identity_transfer *)_transfer; +} + + +static INLINE struct pipe_buffer * +identity_buffer_unwrap(struct pipe_buffer *_buffer) +{ + if(!_buffer) + return NULL; + return identity_buffer(_buffer)->buffer; +} + +static INLINE struct pipe_texture * +identity_texture_unwrap(struct pipe_texture *_texture) +{ + if(!_texture) + return NULL; + return identity_texture(_texture)->texture; +} + +static INLINE struct pipe_surface * +identity_surface_unwrap(struct pipe_surface *_surface) +{ + if(!_surface) + return NULL; + return identity_surface(_surface)->surface; +} + +static INLINE struct pipe_transfer * +identity_transfer_unwrap(struct pipe_transfer *_transfer) +{ + if(!_transfer) + return NULL; + return identity_transfer(_transfer)->transfer; +} + + +struct pipe_buffer * +identity_buffer_create(struct identity_screen *id_screen, + struct pipe_buffer *buffer); + +void +identity_buffer_destroy(struct identity_buffer *id_buffer); + +struct pipe_texture * +identity_texture_create(struct identity_screen *id_screen, + struct pipe_texture *texture); + +void +identity_texture_destroy(struct identity_texture *id_texture); + +struct pipe_surface * +identity_surface_create(struct identity_texture *id_texture, + struct pipe_surface *surface); + +void +identity_surface_destroy(struct identity_surface *id_surface); + +struct pipe_transfer * +identity_transfer_create(struct identity_texture *id_texture, + struct pipe_transfer *transfer); + +void +identity_transfer_destroy(struct identity_transfer *id_transfer); + + +#endif /* ID_OBJECTS_H */ diff --git a/src/gallium/drivers/identity/id_public.h b/src/gallium/drivers/identity/id_public.h new file mode 100644 index 00000000000..cac14cfd604 --- /dev/null +++ b/src/gallium/drivers/identity/id_public.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ID_PUBLIC_H +#define ID_PUBLIC_H + +struct pipe_screen; +struct pipe_context; + +struct pipe_screen * +identity_screen_create(struct pipe_screen *screen); + +struct pipe_context * +identity_context_create(struct pipe_screen *screen, struct pipe_context *pipe); + +#endif /* PT_PUBLIC_H */ diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c new file mode 100644 index 00000000000..26439637d08 --- /dev/null +++ b/src/gallium/drivers/identity/id_screen.c @@ -0,0 +1,483 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_screen.h" +#include "pipe/p_state.h" +#include "util/u_memory.h" + +#include "id_public.h" +#include "id_screen.h" +#include "id_objects.h" + + +static void +identity_screen_destroy(struct pipe_screen *_screen) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + screen->destroy(screen); + + FREE(id_screen); +} + +static const char * +identity_screen_get_name(struct pipe_screen *_screen) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->get_name(screen); +} + +static const char * +identity_screen_get_vendor(struct pipe_screen *_screen) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->get_vendor(screen); +} + +static int +identity_screen_get_param(struct pipe_screen *_screen, + int param) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->get_param(screen, + param); +} + +static float +identity_screen_get_paramf(struct pipe_screen *_screen, + int param) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->get_paramf(screen, + param); +} + +static boolean +identity_screen_is_format_supported(struct pipe_screen *_screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->is_format_supported(screen, + format, + target, + tex_usage, + geom_flags); +} + +static struct pipe_texture * +identity_screen_texture_create(struct pipe_screen *_screen, + const struct pipe_texture *templat) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + struct pipe_texture *result; + + result = screen->texture_create(screen, + templat); + + if (result) + return identity_texture_create(id_screen, result); + return NULL; +} + +static struct pipe_texture * +identity_screen_texture_blanket(struct pipe_screen *_screen, + const struct pipe_texture *templat, + const unsigned *stride, + struct pipe_buffer *_buffer) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + struct pipe_texture *result; + + result = screen->texture_blanket(screen, + templat, + stride, + buffer); + + if (result) + return identity_texture_create(id_screen, result); + return NULL; +} + +static void +identity_screen_texture_destroy(struct pipe_texture *_texture) +{ + identity_texture_destroy(identity_texture(_texture)); +} + +static struct pipe_surface * +identity_screen_get_tex_surface(struct pipe_screen *_screen, + struct pipe_texture *_texture, + unsigned face, + unsigned level, + unsigned zslice, + unsigned usage) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_texture *id_texture = identity_texture(_texture); + struct pipe_screen *screen = id_screen->screen; + struct pipe_texture *texture = id_texture->texture; + struct pipe_surface *result; + + result = screen->get_tex_surface(screen, + texture, + face, + level, + zslice, + usage); + + if (result) + return identity_surface_create(id_texture, result); + return NULL; +} + +static void +identity_screen_tex_surface_destroy(struct pipe_surface *_surface) +{ + identity_surface_destroy(identity_surface(_surface)); +} + +static struct pipe_transfer * +identity_screen_get_tex_transfer(struct pipe_screen *_screen, + struct pipe_texture *_texture, + unsigned face, + unsigned level, + unsigned zslice, + enum pipe_transfer_usage usage, + unsigned x, + unsigned y, + unsigned w, + unsigned h) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_texture *id_texture = identity_texture(_texture); + struct pipe_screen *screen = id_screen->screen; + struct pipe_texture *texture = id_texture->texture; + struct pipe_transfer *result; + + result = screen->get_tex_transfer(screen, + texture, + face, + level, + zslice, + usage, + x, + y, + w, + h); + + if (result) + return identity_transfer_create(id_texture, result); + return NULL; +} + +static void +identity_screen_tex_transfer_destroy(struct pipe_transfer *_transfer) +{ + identity_transfer_destroy(identity_transfer(_transfer)); +} + +static void * +identity_screen_transfer_map(struct pipe_screen *_screen, + struct pipe_transfer *_transfer) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_transfer *id_transfer = identity_transfer(_transfer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_transfer *transfer = id_transfer->transfer; + + return screen->transfer_map(screen, + transfer); +} + +static void +identity_screen_transfer_unmap(struct pipe_screen *_screen, + struct pipe_transfer *_transfer) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_transfer *id_transfer = identity_transfer(_transfer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_transfer *transfer = id_transfer->transfer; + + screen->transfer_unmap(screen, + transfer); +} + +static struct pipe_buffer * +identity_screen_buffer_create(struct pipe_screen *_screen, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *result; + + result = screen->buffer_create(screen, + alignment, + usage, + size); + + if (result) + return identity_buffer_create(id_screen, result); + return NULL; +} + +static struct pipe_buffer * +identity_screen_user_buffer_create(struct pipe_screen *_screen, + void *ptr, + unsigned bytes) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *result; + + result = screen->user_buffer_create(screen, + ptr, + bytes); + + if (result) + return identity_buffer_create(id_screen, result); + return NULL; +} + +static struct pipe_buffer * +identity_screen_surface_buffer_create(struct pipe_screen *_screen, + unsigned width, + unsigned height, + enum pipe_format format, + unsigned usage, + unsigned tex_usage, + unsigned *stride) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *result; + + result = screen->surface_buffer_create(screen, + width, + height, + format, + usage, + tex_usage, + stride); + + if (result) + return identity_buffer_create(id_screen, result); + return NULL; +} + +static void * +identity_screen_buffer_map(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned usage) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + + return screen->buffer_map(screen, + buffer, + usage); +} + +static void * +identity_screen_buffer_map_range(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned offset, + unsigned length, + unsigned usage) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + + return screen->buffer_map_range(screen, + buffer, + offset, + length, + usage); +} + +static void +identity_screen_buffer_flush_mapped_range(struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned offset, + unsigned length) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + + screen->buffer_flush_mapped_range(screen, + buffer, + offset, + length); +} + +static void +identity_screen_buffer_unmap(struct pipe_screen *_screen, + struct pipe_buffer *_buffer) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_buffer *id_buffer = identity_buffer(_buffer); + struct pipe_screen *screen = id_screen->screen; + struct pipe_buffer *buffer = id_buffer->buffer; + + screen->buffer_unmap(screen, + buffer); +} + +static void +identity_screen_buffer_destroy(struct pipe_buffer *_buffer) +{ + identity_buffer_destroy(identity_buffer(_buffer)); +} + +static void +identity_screen_flush_frontbuffer(struct pipe_screen *_screen, + struct pipe_surface *_surface, + void *context_private) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct identity_surface *id_surface = identity_surface(_surface); + struct pipe_screen *screen = id_screen->screen; + struct pipe_surface *surface = id_surface->surface; + + screen->flush_frontbuffer(screen, + surface, + context_private); +} + +static void +identity_screen_fence_reference(struct pipe_screen *_screen, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + screen->fence_reference(screen, + ptr, + fence); +} + +static int +identity_screen_fence_signalled(struct pipe_screen *_screen, + struct pipe_fence_handle *fence, + unsigned flags) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->fence_signalled(screen, + fence, + flags); +} + +static int +identity_screen_fence_finish(struct pipe_screen *_screen, + struct pipe_fence_handle *fence, + unsigned flags) +{ + struct identity_screen *id_screen = identity_screen(_screen); + struct pipe_screen *screen = id_screen->screen; + + return screen->fence_finish(screen, + fence, + flags); +} + +struct pipe_screen * +identity_screen_create(struct pipe_screen *screen) +{ + struct identity_screen *id_screen; + + id_screen = CALLOC_STRUCT(identity_screen); + if (!id_screen) { + return NULL; + } + + id_screen->base.winsys = NULL; + + id_screen->base.destroy = identity_screen_destroy; + id_screen->base.get_name = identity_screen_get_name; + id_screen->base.get_vendor = identity_screen_get_vendor; + id_screen->base.get_param = identity_screen_get_param; + id_screen->base.get_paramf = identity_screen_get_paramf; + id_screen->base.is_format_supported = identity_screen_is_format_supported; + id_screen->base.texture_create = identity_screen_texture_create; + id_screen->base.texture_blanket = identity_screen_texture_blanket; + id_screen->base.texture_destroy = identity_screen_texture_destroy; + id_screen->base.get_tex_surface = identity_screen_get_tex_surface; + id_screen->base.tex_surface_destroy = identity_screen_tex_surface_destroy; + id_screen->base.get_tex_transfer = identity_screen_get_tex_transfer; + id_screen->base.tex_transfer_destroy = identity_screen_tex_transfer_destroy; + id_screen->base.transfer_map = identity_screen_transfer_map; + id_screen->base.transfer_unmap = identity_screen_transfer_unmap; + id_screen->base.buffer_create = identity_screen_buffer_create; + id_screen->base.user_buffer_create = identity_screen_user_buffer_create; + id_screen->base.surface_buffer_create = identity_screen_surface_buffer_create; + if (screen->buffer_map) + id_screen->base.buffer_map = identity_screen_buffer_map; + if (screen->buffer_map_range) + id_screen->base.buffer_map_range = identity_screen_buffer_map_range; + if (screen->buffer_flush_mapped_range) + id_screen->base.buffer_flush_mapped_range = identity_screen_buffer_flush_mapped_range; + if (screen->buffer_unmap) + id_screen->base.buffer_unmap = identity_screen_buffer_unmap; + id_screen->base.buffer_destroy = identity_screen_buffer_destroy; + id_screen->base.flush_frontbuffer = identity_screen_flush_frontbuffer; + id_screen->base.fence_reference = identity_screen_fence_reference; + id_screen->base.fence_signalled = identity_screen_fence_signalled; + id_screen->base.fence_finish = identity_screen_fence_finish; + + id_screen->screen = screen; + + return &id_screen->base; +} diff --git a/src/gallium/drivers/identity/id_screen.h b/src/gallium/drivers/identity/id_screen.h new file mode 100644 index 00000000000..2c4f1290894 --- /dev/null +++ b/src/gallium/drivers/identity/id_screen.h @@ -0,0 +1,48 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ID_SCREEN_H +#define ID_SCREEN_H + +#include "pipe/p_screen.h" +#include "pipe/p_defines.h" + + +struct identity_screen { + struct pipe_screen base; + + struct pipe_screen *screen; +}; + + +static INLINE struct identity_screen * +identity_screen(struct pipe_screen *screen) +{ + return (struct identity_screen *)screen; +} + +#endif /* ID_SCREEN_H */ diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c index 832366e6462..e4cf91c005c 100644 --- a/src/gallium/drivers/nouveau/nouveau_screen.c +++ b/src/gallium/drivers/nouveau/nouveau_screen.c @@ -4,6 +4,8 @@ #include <util/u_memory.h> +#include <errno.h> + #include "nouveau/nouveau_bo.h" #include "nouveau_winsys.h" #include "nouveau_screen.h" @@ -141,12 +143,13 @@ nouveau_screen_bo_map_range(struct pipe_screen *pscreen, struct pipe_buffer *pb, unsigned offset, unsigned length, unsigned usage) { struct nouveau_bo *bo = nouveau_bo(pb); + uint32_t flags = nouveau_screen_map_flags(usage); int ret; - ret = nouveau_bo_map_range(bo, offset, length, - nouveau_screen_map_flags(usage)); + ret = nouveau_bo_map_range(bo, offset, length, flags); if (ret) { - debug_printf("map_range failed: %d\n", ret); + if (!(flags & NOUVEAU_BO_NOWAIT) || ret != -EBUSY) + debug_printf("map_range failed: %d\n", ret); return NULL; } diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c index 4bbedfb4d69..ff2febb668e 100644 --- a/src/gallium/drivers/nv04/nv04_screen.c +++ b/src/gallium/drivers/nv04/nv04_screen.c @@ -40,6 +40,10 @@ nv04_screen_get_param(struct pipe_screen *screen, int param) return 0; case PIPE_CAP_TEXTURE_MIRROR_REPEAT: return 1; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 0; case NOUVEAU_CAP_HW_VTXBUF: case NOUVEAU_CAP_HW_IDXBUF: return 0; @@ -79,6 +83,13 @@ nv04_screen_is_format_supported(struct pipe_screen *screen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { case PIPE_FORMAT_Z16_UNORM: return TRUE; default: diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c index 5afd028ddd2..f88e138c79d 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -15,11 +15,13 @@ nv04_surface_format(enum pipe_format format) return NV04_CONTEXT_SURFACES_2D_FORMAT_Y8; case PIPE_FORMAT_R16_SNORM: case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z16_UNORM: return NV04_CONTEXT_SURFACES_2D_FORMAT_R5G6B5; case PIPE_FORMAT_X8R8G8B8_UNORM: case PIPE_FORMAT_A8R8G8B8_UNORM: return NV04_CONTEXT_SURFACES_2D_FORMAT_A8R8G8B8; case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: return NV04_CONTEXT_SURFACES_2D_FORMAT_Y32; default: return -1; @@ -33,9 +35,11 @@ nv04_rect_format(enum pipe_format format) case PIPE_FORMAT_A8_UNORM: return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8; case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_Z16_UNORM: return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A16R5G6B5; case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8; default: return -1; @@ -110,10 +114,10 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, unsigned cx; unsigned cy; - /* POT or GTFO */ - assert(!(w & (w - 1)) && !(h & (h - 1))); +#if 0 /* That's the way she likes it */ assert(src_pitch == ((struct nv04_surface *)dst)->pitch); +#endif BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_DMA_IMAGE, 1); OUT_RELOCo(chan, dst_bo, @@ -133,7 +137,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, for (cy = 0; cy < h; cy += sub_h) { for (cx = 0; cx < w; cx += sub_w) { BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1); - OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx, cy) * + OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx+dx, cy+dy) * dst->texture->block.size, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); @@ -153,8 +157,8 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, OUT_RING (chan, src_pitch | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE); - OUT_RELOCl(chan, src_bo, src->offset + cy * src_pitch + - cx * src->texture->block.size, NOUVEAU_BO_GART | + OUT_RELOCl(chan, src_bo, src->offset + (cy+sy) * src_pitch + + (cx+sx) * src->texture->block.size, NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RING (chan, 0); } @@ -210,6 +214,43 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, } static int +nv04_surface_copy_m2mf_swizzle(struct nv04_surface_2d *ctx, + struct pipe_surface *dst, int dx, int dy, + struct pipe_surface *src, int sx, int sy) +{ + struct nouveau_channel *chan = ctx->m2mf->channel; + struct nouveau_grobj *m2mf = ctx->m2mf; + struct nouveau_bo *src_bo = nouveau_bo(ctx->buf(src)); + struct nouveau_bo *dst_bo = nouveau_bo(ctx->buf(dst)); + unsigned src_pitch = ((struct nv04_surface *)src)->pitch; + unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; + unsigned dst_offset = dst->offset + nv04_swizzle_bits(dx, dy) * + dst->texture->block.size; + unsigned src_offset = src->offset + sy * src_pitch + + sx * src->texture->block.size; + + BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2); + OUT_RELOCo(chan, src_bo, + NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); + OUT_RELOCo(chan, dst_bo, + NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); + + BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8); + OUT_RELOCl(chan, src_bo, src_offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD); + OUT_RELOCl(chan, dst_bo, dst_offset, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR); + OUT_RING (chan, src_pitch); + OUT_RING (chan, dst_pitch); + OUT_RING (chan, 1 * src->texture->block.size); + OUT_RING (chan, 1); + OUT_RING (chan, 0x0101); + OUT_RING (chan, 0); + + return 0; +} + +static int nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst, int dx, int dy, struct pipe_surface *src, int sx, int sy, int w, int h) @@ -258,8 +299,59 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst, assert(src->format == dst->format); /* Setup transfer to swizzle the texture to vram if needed */ - if (src_linear && !dst_linear && w > 1 && h > 1) { - nv04_surface_copy_swizzle(ctx, dst, dx, dy, src, sx, sy, w, h); + if (src_linear && !dst_linear) { + int x,y; + + if ((w>1) && (h>1)) { + int potWidth = 1<<log2i(w); + int potHeight = 1<<log2i(h); + int remainWidth = w-potWidth; + int remainHeight = h-potHeight; + int squareDim = (potWidth>potHeight ? potHeight : potWidth); + + /* top left is always POT, but we can only swizzle squares */ + for (y=0; y<potHeight; y+=squareDim) { + for (x=0; x<potWidth; x+= squareDim) { + nv04_surface_copy_swizzle(ctx, dst, dx+x, dy+y, + src, sx+x, sy+y, + squareDim, squareDim); + } + } + + /* top right */ + if (remainWidth>0) { + nv04_surface_copy(ctx, dst, dx+potWidth, dy, + src, sx+potWidth, sy, + remainWidth, potHeight); + } + + /* bottom left */ + if (remainHeight>0) { + nv04_surface_copy(ctx, dst, dx, dy+potHeight, + src, sx, sy+potHeight, + potWidth, remainHeight); + } + + /* bottom right */ + if ((remainWidth>0) && (remainHeight>0)) { + nv04_surface_copy(ctx, dst, dx+potWidth, dy+potHeight, + src, sx+potWidth, sy+potHeight, + remainWidth, remainHeight); + } + } else if (w==1) { + /* We have a column to copy to a swizzled texture */ + for (y=0; y<h; y++) { + nv04_surface_copy_m2mf_swizzle(ctx, dst, dx, dy+y, + src, sx, sy+y); + } + } else if (h==1) { + /* We have a row to copy to a swizzled texture */ + for (x=0; x<w; x++) { + nv04_surface_copy_m2mf_swizzle(ctx, dst, dx+x, dy, + src, sx+x, sy); + } + } + return; } @@ -267,8 +359,7 @@ nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst, * to NV_MEMORY_TO_MEMORY_FORMAT in this case. */ if ((src->offset & 63) || (dst->offset & 63) || - (src_pitch & 63) || (dst_pitch & 63) || - debug_get_bool_option("NOUVEAU_NO_COPYBLIT", FALSE)) { + (src_pitch & 63) || (dst_pitch & 63)) { nv04_surface_copy_m2mf(ctx, dst, dx, dy, src, sx, sy, w, h); return; } diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c index b03c291f9d9..4469b22d91a 100644 --- a/src/gallium/drivers/nv10/nv10_screen.c +++ b/src/gallium/drivers/nv10/nv10_screen.c @@ -35,6 +35,10 @@ nv10_screen_get_param(struct pipe_screen *screen, int param) return 12; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 0; case NOUVEAU_CAP_HW_VTXBUF: case NOUVEAU_CAP_HW_IDXBUF: return 0; @@ -74,7 +78,15 @@ nv10_screen_is_format_supported(struct pipe_screen *screen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z16_UNORM: return TRUE; default: diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c index 024356ca74c..e6924ad71eb 100644 --- a/src/gallium/drivers/nv20/nv20_screen.c +++ b/src/gallium/drivers/nv20/nv20_screen.c @@ -35,6 +35,10 @@ nv20_screen_get_param(struct pipe_screen *screen, int param) return 12; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 0; case NOUVEAU_CAP_HW_VTXBUF: case NOUVEAU_CAP_HW_IDXBUF: return 0; @@ -74,7 +78,15 @@ nv20_screen_is_format_supported(struct pipe_screen *screen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z16_UNORM: return TRUE; default: diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c index c1e588902b2..388245ecb04 100644 --- a/src/gallium/drivers/nv20/nv20_vertprog.c +++ b/src/gallium/drivers/nv20/nv20_vertprog.c @@ -617,10 +617,10 @@ nv20_vertprog_translate(struct nv20_context *nv20, assert(imm->Immediate.NrTokens == 4 + 1); vpc->imm[vpc->nr_imm++] = constant(vpc, -1, - imm->u.ImmediateFloat32[0].Float, - imm->u.ImmediateFloat32[1].Float, - imm->u.ImmediateFloat32[2].Float, - imm->u.ImmediateFloat32[3].Float); + imm->u[0].Float, + imm->u[1].Float, + imm->u[2].Float, + imm->u[3].Float); } break; case TGSI_TOKEN_TYPE_INSTRUCTION: diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c index 1d1c556fb11..a48ba9782b3 100644 --- a/src/gallium/drivers/nv30/nv30_fragprog.c +++ b/src/gallium/drivers/nv30/nv30_fragprog.c @@ -704,10 +704,10 @@ nv30_fragprog_prepare(struct nv30_fpc *fpc) assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); assert(fpc->nr_imm < MAX_IMM); - vals[0] = imm->u.ImmediateFloat32[0].Float; - vals[1] = imm->u.ImmediateFloat32[1].Float; - vals[2] = imm->u.ImmediateFloat32[2].Float; - vals[3] = imm->u.ImmediateFloat32[3].Float; + vals[0] = imm->u[0].Float; + vals[1] = imm->u[1].Float; + vals[2] = imm->u[2].Float; + vals[3] = imm->u[3].Float; fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals); } break; diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c index 31bc1f31734..f8285e4455f 100644 --- a/src/gallium/drivers/nv30/nv30_screen.c +++ b/src/gallium/drivers/nv30/nv30_screen.c @@ -46,6 +46,10 @@ nv30_screen_get_param(struct pipe_screen *pscreen, int param) return 1; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 0; case NOUVEAU_CAP_HW_VTXBUF: case NOUVEAU_CAP_HW_IDXBUF: return 1; @@ -85,7 +89,15 @@ nv30_screen_surface_format_supported(struct pipe_screen *pscreen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z16_UNORM: return TRUE; default: diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c index cb1a260eaad..44b6a74715a 100644 --- a/src/gallium/drivers/nv30/nv30_state_fb.c +++ b/src/gallium/drivers/nv30/nv30_state_fb.c @@ -65,6 +65,7 @@ nv30_state_framebuffer_validate(struct nv30_context *nv30) rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16; break; case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case 0: rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8; break; diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c index c7514efcfea..14a5c0260d0 100644 --- a/src/gallium/drivers/nv30/nv30_vertprog.c +++ b/src/gallium/drivers/nv30/nv30_vertprog.c @@ -617,10 +617,10 @@ nv30_vertprog_translate(struct nv30_context *nv30, assert(imm->Immediate.NrTokens == 4 + 1); vpc->imm[vpc->nr_imm++] = constant(vpc, -1, - imm->u.ImmediateFloat32[0].Float, - imm->u.ImmediateFloat32[1].Float, - imm->u.ImmediateFloat32[2].Float, - imm->u.ImmediateFloat32[3].Float); + imm->u[0].Float, + imm->u[1].Float, + imm->u[2].Float, + imm->u[3].Float); } break; case TGSI_TOKEN_TYPE_INSTRUCTION: diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c index 680976da56b..32d9ed1a7f8 100644 --- a/src/gallium/drivers/nv40/nv40_fragprog.c +++ b/src/gallium/drivers/nv40/nv40_fragprog.c @@ -790,10 +790,10 @@ nv40_fragprog_prepare(struct nv40_fpc *fpc) assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32); assert(fpc->nr_imm < MAX_IMM); - vals[0] = imm->u.ImmediateFloat32[0].Float; - vals[1] = imm->u.ImmediateFloat32[1].Float; - vals[2] = imm->u.ImmediateFloat32[2].Float; - vals[3] = imm->u.ImmediateFloat32[3].Float; + vals[0] = imm->u[0].Float; + vals[1] = imm->u[1].Float; + vals[2] = imm->u[2].Float; + vals[3] = imm->u[3].Float; fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals); } break; diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c index b8b2af482a9..5d2a4216c5a 100644 --- a/src/gallium/drivers/nv40/nv40_screen.c +++ b/src/gallium/drivers/nv40/nv40_screen.c @@ -44,6 +44,10 @@ nv40_screen_get_param(struct pipe_screen *pscreen, int param) return 1; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; /* We have 4 - but unsupported currently */ + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 1; case NOUVEAU_CAP_HW_VTXBUF: return 1; case NOUVEAU_CAP_HW_IDXBUF: @@ -86,7 +90,15 @@ nv40_screen_surface_format_supported(struct pipe_screen *pscreen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z16_UNORM: return TRUE; default: diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c index 273142f9e0f..c2f739157ad 100644 --- a/src/gallium/drivers/nv40/nv40_state_fb.c +++ b/src/gallium/drivers/nv40/nv40_state_fb.c @@ -73,6 +73,7 @@ nv40_state_framebuffer_validate(struct nv40_context *nv40) rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16; break; case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: case 0: rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8; break; diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c index ce45055fe82..6d92ac3db9c 100644 --- a/src/gallium/drivers/nv40/nv40_transfer.c +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -136,7 +136,7 @@ nv40_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, } static void -nv40_transfer_del(struct pipe_screen *pscreen, struct pipe_transfer *ptx) +nv40_transfer_del(struct pipe_transfer *ptx) { struct nv40_transfer *tx = (struct nv40_transfer *)ptx; diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c index e75e8d3f424..0382dbba8f6 100644 --- a/src/gallium/drivers/nv40/nv40_vertprog.c +++ b/src/gallium/drivers/nv40/nv40_vertprog.c @@ -788,10 +788,10 @@ nv40_vertprog_translate(struct nv40_context *nv40, assert(imm->Immediate.NrTokens == 4 + 1); vpc->imm[vpc->nr_imm++] = constant(vpc, -1, - imm->u.ImmediateFloat32[0].Float, - imm->u.ImmediateFloat32[1].Float, - imm->u.ImmediateFloat32[2].Float, - imm->u.ImmediateFloat32[3].Float); + imm->u[0].Float, + imm->u[1].Float, + imm->u[2].Float, + imm->u[3].Float); } break; case TGSI_TOKEN_TYPE_INSTRUCTION: diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c index e02afc4be99..6e8f4f9750d 100644 --- a/src/gallium/drivers/nv50/nv50_context.c +++ b/src/gallium/drivers/nv50/nv50_context.c @@ -31,15 +31,23 @@ static void nv50_flush(struct pipe_context *pipe, unsigned flags, struct pipe_fence_handle **fence) { - struct nv50_context *nv50 = (struct nv50_context *)pipe; - - FIRE_RING(nv50->screen->base.channel); + struct nv50_context *nv50 = nv50_context(pipe); + struct nouveau_channel *chan = nv50->screen->base.channel; + struct nouveau_grobj *eng2d = nv50->screen->eng2d; + + /* We need this in the ddx for reliable composite, not sure what we're + * actually flushing. We generate all our own flushes with flags = 0. */ + WAIT_RING(chan, 3); + BEGIN_RING(chan, eng2d, 0x0110, 1); + OUT_RING (chan, 0); + + FIRE_RING(chan); } static void nv50_destroy(struct pipe_context *pipe) { - struct nv50_context *nv50 = (struct nv50_context *)pipe; + struct nv50_context *nv50 = nv50_context(pipe); draw_destroy(nv50->draw); FREE(nv50); @@ -112,5 +120,3 @@ nv50_create(struct pipe_screen *pscreen, unsigned pctx_id) return &nv50->pipe; } - - diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h index 9b8cc4d37d0..4de6e8cfa29 100644 --- a/src/gallium/drivers/nv50/nv50_context.h +++ b/src/gallium/drivers/nv50/nv50_context.h @@ -71,6 +71,7 @@ struct nv50_sampler_stateobj { struct nv50_miptree_level { int *image_offset; unsigned pitch; + unsigned tile_mode; }; struct nv50_miptree { @@ -119,6 +120,7 @@ struct nv50_state { struct nouveau_stateobj *fragprog; struct nouveau_stateobj *vtxfmt; struct nouveau_stateobj *vtxbuf; + struct nouveau_stateobj *vtxattr; }; struct nv50_context { diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 6b605ba416c..dd1b0303bd5 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -42,8 +42,14 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) mt->base.screen = pscreen; switch (pt->format) { + case PIPE_FORMAT_Z32_FLOAT: + tile_flags = 0x4800; + break; case PIPE_FORMAT_Z24S8_UNORM: - case PIPE_FORMAT_Z16_UNORM: + tile_flags = 0x1800; + break; + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: tile_flags = 0x2800; break; default: @@ -81,20 +87,27 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) lvl->image_offset = CALLOC(mt->image_nr, sizeof(int)); lvl->pitch = align(pt->width[l] * pt->block.size, 64); + lvl->tile_mode = tile_mode; width = MAX2(1, width >> 1); height = MAX2(1, height >> 1); depth = MAX2(1, depth >> 1); + + if (tile_mode && height <= (tile_h >> 1)) { + tile_mode--; + tile_h >>= 1; + } } for (i = 0; i < mt->image_nr; i++) { for (l = 0; l <= pt->last_level; l++) { struct nv50_miptree_level *lvl = &mt->level[l]; int size; + tile_h = 1 << (lvl->tile_mode + 2); size = align(pt->width[l], 8) * pt->block.size; size = align(size, 64); - size *= align(pt->height[l], tile_h) * pt->block.size; + size *= align(pt->height[l], tile_h); lvl->image_offset[i] = mt->total_size; @@ -103,12 +116,12 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) } ret = nouveau_bo_new_tile(dev, NOUVEAU_BO_VRAM, 256, mt->total_size, - tile_mode, tile_flags, &mt->bo); + mt->level[0].tile_mode, tile_flags, &mt->bo); if (ret) { FREE(mt); return NULL; } - + return &mt->base; } @@ -145,7 +158,7 @@ nv50_miptree_destroy(struct pipe_texture *pt) struct nv50_miptree *mt = nv50_miptree(pt); nouveau_bo_ref(NULL, &mt->bo); - FREE(mt); + FREE(mt); } static struct pipe_surface * @@ -188,8 +201,8 @@ nv50_miptree_surface_del(struct pipe_surface *ps) { struct nv50_surface *s = nv50_surface(ps); - pipe_texture_reference(&ps->texture, NULL); - FREE(s); + pipe_texture_reference(&ps->texture, NULL); + FREE(s); } void diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c index 5f7d06dbecb..289c3485e08 100644 --- a/src/gallium/drivers/nv50/nv50_program.c +++ b/src/gallium/drivers/nv50/nv50_program.c @@ -251,7 +251,7 @@ alloc_temp4(struct nv50_pc *pc, struct nv50_reg *dst[4], int idx) if (pc->r_temp[idx] || pc->r_temp[idx + 1] || pc->r_temp[idx + 2] || pc->r_temp[idx + 3]) - return alloc_temp4(pc, dst, idx + 1); + return alloc_temp4(pc, dst, idx + 4); for (i = 0; i < 4; i++) { dst[i] = CALLOC_STRUCT(nv50_reg); @@ -296,7 +296,7 @@ kill_temp_temp(struct nv50_pc *pc) static int ctor_immd(struct nv50_pc *pc, float x, float y, float z, float w) { - pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * r * sizeof(float)), + pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * 4 * sizeof(float)), (pc->immd_nr + 1) * 4 * sizeof(float)); pc->immd_buf[(pc->immd_nr * 4) + 0] = x; pc->immd_buf[(pc->immd_nr * 4) + 1] = y; @@ -1014,6 +1014,7 @@ emit_tex(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask, break; } + /* some cards need t[0]'s hw index to be a multiple of 4 */ alloc_temp4(pc, t, 0); if (proj) { @@ -1809,10 +1810,10 @@ nv50_program_tx_prep(struct nv50_pc *pc) const struct tgsi_full_immediate *imm = &p.FullToken.FullImmediate; - ctor_immd(pc, imm->u.ImmediateFloat32[0].Float, - imm->u.ImmediateFloat32[1].Float, - imm->u.ImmediateFloat32[2].Float, - imm->u.ImmediateFloat32[3].Float); + ctor_immd(pc, imm->u[0].Float, + imm->u[1].Float, + imm->u[2].Float, + imm->u[3].Float); } break; case TGSI_TOKEN_TYPE_DECLARATION: @@ -2221,9 +2222,9 @@ nv50_program_upload_data(struct nv50_context *nv50, float *map, while (count) { unsigned nr = count > 2047 ? 2047 : count; - BEGIN_RING(chan, tesla, 0x00000f00, 1); + BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1); OUT_RING (chan, (cbuf << 0) | (start << 8)); - BEGIN_RING(chan, tesla, 0x40000f04, nr); + BEGIN_RING(chan, tesla, NV50TCL_CB_DATA(0) | 0x40000000, nr); OUT_RINGp (chan, map, nr); map += nr; @@ -2345,7 +2346,7 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) } so = so_new(4,2); - so_method(so, nv50->screen->tesla, 0x1280, 3); + so_method(so, nv50->screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, p->bo, 0, flags | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4)); @@ -2364,9 +2365,9 @@ nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p) continue; } - BEGIN_RING(chan, tesla, 0x0f00, 1); + BEGIN_RING(chan, tesla, NV50TCL_CB_ADDR, 1); OUT_RING (chan, (start << 8) | NV50_CB_PUPLOAD); - BEGIN_RING(chan, tesla, 0x40000f04, nr); + BEGIN_RING(chan, tesla, NV50TCL_CB_DATA(0) | 0x40000000, nr); OUT_RINGp (chan, up + start, nr); start += nr; @@ -2399,15 +2400,15 @@ nv50_vertprog_validate(struct nv50_context *nv50) NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); - so_method(so, tesla, 0x1650, 2); + so_method(so, tesla, NV50TCL_VP_ATTR_EN_0, 2); so_data (so, p->cfg.vp.attr[0]); so_data (so, p->cfg.vp.attr[1]); - so_method(so, tesla, 0x16b8, 1); + so_method(so, tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1); so_data (so, p->cfg.high_result); - so_method(so, tesla, 0x16ac, 2); + so_method(so, tesla, NV50TCL_VP_RESULT_MAP_SIZE, 2); so_data (so, p->cfg.high_result); //8); so_data (so, p->cfg.high_temp); - so_method(so, tesla, 0x140c, 1); + so_method(so, tesla, NV50TCL_VP_START_ID, 1); so_data (so, 0); /* program start offset */ so_ref(so, &nv50->state.vertprog); so_ref(NULL, &so); @@ -2436,24 +2437,24 @@ nv50_fragprog_validate(struct nv50_context *nv50) NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, p->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); - so_method(so, tesla, 0x1904, 4); + so_method(so, tesla, NV50TCL_MAP_SEMANTIC_0, 4); so_data (so, p->cfg.fp.regs[0]); /* 0x01000404 / 0x00040404 */ so_data (so, 0x00000004); so_data (so, 0x00000000); so_data (so, 0x00000000); - so_method(so, tesla, 0x16bc, p->cfg.fp.high_map); + so_method(so, tesla, NV50TCL_VP_RESULT_MAP(0), p->cfg.fp.high_map); for (i = 0; i < p->cfg.fp.high_map; i++) so_data(so, p->cfg.fp.map[i]); - so_method(so, tesla, 0x1988, 2); + so_method(so, tesla, NV50TCL_FP_INTERPOLANT_CTRL, 2); so_data (so, p->cfg.fp.regs[1]); /* 0x08040404 / 0x0f000401 */ so_data (so, p->cfg.high_temp); - so_method(so, tesla, 0x1298, 1); + so_method(so, tesla, NV50TCL_FP_RESULT_COUNT, 1); so_data (so, p->cfg.high_result); - so_method(so, tesla, 0x19a8, 1); + so_method(so, tesla, NV50TCL_FP_CTRL_UNK19A8, 1); so_data (so, p->cfg.fp.regs[2]); - so_method(so, tesla, 0x196c, 1); + so_method(so, tesla, NV50TCL_FP_CTRL_UNK196C, 1); so_data (so, p->cfg.fp.regs[3]); - so_method(so, tesla, 0x1414, 1); + so_method(so, tesla, NV50TCL_FP_START_ID, 1); so_data (so, 0); /* program start offset */ so_ref(so, &nv50->state.fragprog); so_ref(NULL, &so); @@ -2478,4 +2479,3 @@ nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p) p->translated = 0; } - diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c index 940e04365f2..5305c93d59a 100644 --- a/src/gallium/drivers/nv50/nv50_query.c +++ b/src/gallium/drivers/nv50/nv50_query.c @@ -94,7 +94,7 @@ nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq) struct nv50_query *q = nv50_query(pq); WAIT_RING (chan, 5); - BEGIN_RING(chan, tesla, 0x1b00, 4); + BEGIN_RING(chan, tesla, NV50TCL_QUERY_ADDRESS_HIGH, 4); OUT_RELOCh(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RELOCl(chan, q->bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); OUT_RING (chan, 0x00000000); @@ -107,13 +107,13 @@ nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq, boolean wait, uint64_t *result) { struct nv50_query *q = nv50_query(pq); - - /*XXX: Want to be able to return FALSE here instead of blocking - * until the result is available.. - */ + int ret; if (!q->ready) { - nouveau_bo_map(q->bo, NOUVEAU_BO_RD); + ret = nouveau_bo_map(q->bo, NOUVEAU_BO_RD | + wait ? 0 : NOUVEAU_BO_NOWAIT); + if (ret) + return false; q->result = ((uint32_t *)q->bo->map)[1]; q->ready = TRUE; nouveau_bo_unmap(q->bo); diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c index fd39fa738b7..c7f80a22037 100644 --- a/src/gallium/drivers/nv50/nv50_screen.c +++ b/src/gallium/drivers/nv50/nv50_screen.c @@ -37,8 +37,17 @@ nv50_screen_is_format_supported(struct pipe_screen *pscreen, switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: + return TRUE; + default: + break; + } + } else + if (tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL) { + switch (format) { + case PIPE_FORMAT_Z32_FLOAT: case PIPE_FORMAT_Z24S8_UNORM: - case PIPE_FORMAT_Z16_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: return TRUE; default: break; @@ -101,6 +110,10 @@ nv50_screen_get_param(struct pipe_screen *pscreen, int param) return 1; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: return 0; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 1; case NOUVEAU_CAP_HW_VTXBUF: return 1; case NOUVEAU_CAP_HW_IDXBUF: @@ -176,7 +189,8 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) nv50_transfer_init_screen_functions(pscreen); /* DMA engine object */ - ret = nouveau_grobj_alloc(chan, 0xbeef5039, 0x5039, &screen->m2mf); + ret = nouveau_grobj_alloc(chan, 0xbeef5039, + NV50_MEMORY_TO_MEMORY_FORMAT, &screen->m2mf); if (ret) { NOUVEAU_ERR("Error creating M2MF object: %d\n", ret); nv50_screen_destroy(pscreen); @@ -185,7 +199,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) BIND_RING(chan, screen->m2mf, 1); /* 2D object */ - ret = nouveau_grobj_alloc(chan, 0xbeef502d, 0x502d, &screen->eng2d); + ret = nouveau_grobj_alloc(chan, 0xbeef502d, NV50_2D, &screen->eng2d); if (ret) { NOUVEAU_ERR("Error creating 2D object: %d\n", ret); nv50_screen_destroy(pscreen); @@ -196,14 +210,15 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) /* 3D object */ switch (chipset & 0xf0) { case 0x50: - tesla_class = 0x5097; + tesla_class = NV50TCL; break; case 0x80: case 0x90: - tesla_class = 0x8297; + /* this stupid name should be corrected. */ + tesla_class = NV54TCL; break; case 0xa0: - tesla_class = 0x8397; + tesla_class = NVA0TCL; break; default: NOUVEAU_ERR("Not a known NV50 chipset: NV%02x\n", chipset); @@ -217,7 +232,8 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) return NULL; } - ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class, &screen->tesla); + ret = nouveau_grobj_alloc(chan, 0xbeef5097, tesla_class, + &screen->tesla); if (ret) { NOUVEAU_ERR("Error creating 3D object: %d\n", ret); nv50_screen_destroy(pscreen); @@ -235,7 +251,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) /* Static M2MF init */ so = so_new(32, 0); - so_method(so, screen->m2mf, 0x0180, 3); + so_method(so, screen->m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 3); so_data (so, screen->sync->handle); so_data (so, chan->vram->handle); so_data (so, chan->vram->handle); @@ -278,9 +294,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_method(so, screen->tesla, 0x13bc, 1); so_data (so, 0x54); + /* origin is top left (set to 1 for bottom left) */ so_method(so, screen->tesla, 0x13ac, 1); - so_data (so, 1); - so_method(so, screen->tesla, 0x16b8, 1); + so_data (so, 0); + so_method(so, screen->tesla, NV50TCL_VP_REG_ALLOC_RESULT, 1); so_data (so, 8); /* constant buffers for immediates and VP/FP parameters */ @@ -318,33 +335,33 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_data (so, 0x000BBNP1); */ - so_method(so, screen->tesla, 0x1280, 3); + so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->constbuf_misc[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_PMISC << 16) | 0x00000800); - so_method(so, screen->tesla, 0x1694, 1); + so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x00000001 | (NV50_CB_PMISC << 12)); - so_method(so, screen->tesla, 0x1694, 1); + so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x00000031 | (NV50_CB_PMISC << 12)); - so_method(so, screen->tesla, 0x1280, 3); + so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->constbuf_parm[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_PVP << 16) | 0x00000800); - so_method(so, screen->tesla, 0x1694, 1); + so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x00000101 | (NV50_CB_PVP << 12)); - so_method(so, screen->tesla, 0x1280, 3); + so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->constbuf_parm[1], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_PFP << 16) | 0x00000800); - so_method(so, screen->tesla, 0x1694, 1); + so_method(so, screen->tesla, NV50TCL_SET_PROGRAM_CB, 1); so_data (so, 0x00000131 | (NV50_CB_PFP << 12)); /* Texture sampler/image unit setup - we abuse the constant buffer @@ -358,13 +375,13 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) return NULL; } - so_method(so, screen->tesla, 0x1280, 3); + so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_TIC << 16) | 0x0800); - so_method(so, screen->tesla, 0x1574, 3); + so_method(so, screen->tesla, NV50TCL_TIC_ADDRESS_HIGH, 3); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM | @@ -377,13 +394,13 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) return NULL; } - so_method(so, screen->tesla, 0x1280, 3); + so_method(so, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); so_data (so, (NV50_CB_TSC << 16) | 0x0800); - so_method(so, screen->tesla, 0x155c, 3); + so_method(so, screen->tesla, NV50TCL_TSC_ADDRESS_HIGH, 3); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0); so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM | @@ -393,7 +410,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) /* Vertex array limits - max them out */ for (i = 0; i < 16; i++) { - so_method(so, screen->tesla, 0x1080 + (i * 8), 2); + so_method(so, screen->tesla, NV50TCL_UNK1080_OFFSET_HIGH(i), 2); so_data (so, 0x000000ff); so_data (so, 0xffffffff); } @@ -405,6 +422,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev) so_method(so, screen->tesla, 0x1234, 1); so_data (so, 1); + /* activate first scissor rectangle */ + so_method(so, screen->tesla, NV50TCL_SCISSOR_ENABLE, 1); + so_data (so, 1); + so_emit(chan, so); so_ref (so, &screen->static_init); so_ref (NULL, &so); diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c index 116866a8e78..ef4154d3036 100644 --- a/src/gallium/drivers/nv50/nv50_state.c +++ b/src/gallium/drivers/nv50/nv50_state.c @@ -205,11 +205,16 @@ nv50_sampler_state_create(struct pipe_context *pipe, } limit = CLAMP(cso->lod_bias, -16.0, 15.0); - tsc[1] |= ((int)(limit * 256.0) & 0x1fff) << 11; + tsc[1] |= ((int)(limit * 256.0) & 0x1fff) << 12; tsc[2] |= ((int)CLAMP(cso->max_lod, 0.0, 15.0) << 20) | ((int)CLAMP(cso->min_lod, 0.0, 15.0) << 8); + tsc[4] = fui(cso->border_color[0]); + tsc[5] = fui(cso->border_color[1]); + tsc[6] = fui(cso->border_color[2]); + tsc[7] = fui(cso->border_color[3]); + sso->normalized = cso->normalized_coords; return (void *)sso; } @@ -404,35 +409,35 @@ nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe, so_data (so, 0); } - /*XXX: yes, I know they're backwards.. header needs fixing */ + /* XXX: keep hex values until header is updated (names reversed) */ if (cso->stencil[0].enabled) { - so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5); + so_method(so, tesla, 0x1380, 8); so_data (so, 1); so_data (so, nvgl_stencil_op(cso->stencil[0].fail_op)); so_data (so, nvgl_stencil_op(cso->stencil[0].zfail_op)); so_data (so, nvgl_stencil_op(cso->stencil[0].zpass_op)); so_data (so, nvgl_comparison_op(cso->stencil[0].func)); - so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 3); so_data (so, cso->stencil[0].ref_value); so_data (so, cso->stencil[0].writemask); so_data (so, cso->stencil[0].valuemask); } else { - so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1); + so_method(so, tesla, 0x1380, 1); so_data (so, 0); } if (cso->stencil[1].enabled) { - so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 8); + so_method(so, tesla, 0x1594, 5); so_data (so, 1); so_data (so, nvgl_stencil_op(cso->stencil[1].fail_op)); so_data (so, nvgl_stencil_op(cso->stencil[1].zfail_op)); so_data (so, nvgl_stencil_op(cso->stencil[1].zpass_op)); so_data (so, nvgl_comparison_op(cso->stencil[1].func)); + so_method(so, tesla, 0x0f54, 3); so_data (so, cso->stencil[1].ref_value); so_data (so, cso->stencil[1].writemask); so_data (so, cso->stencil[1].valuemask); } else { - so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1); + so_method(so, tesla, 0x1594, 1); so_data (so, 0); } diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c index 0caf4b4e914..99d5b96e456 100644 --- a/src/gallium/drivers/nv50/nv50_state_validate.c +++ b/src/gallium/drivers/nv50/nv50_state_validate.c @@ -55,15 +55,15 @@ nv50_state_validate_fb(struct nv50_context *nv50) NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); switch (fb->cbufs[i]->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - so_data(so, 0xcf); + so_data(so, NV50TCL_RT_FORMAT_A8R8G8B8_UNORM); break; case PIPE_FORMAT_R5G6B5_UNORM: - so_data(so, 0xe8); + so_data(so, NV50TCL_RT_FORMAT_R5G6B5_UNORM); break; default: NOUVEAU_ERR("AIIII unknown format %s\n", pf_name(fb->cbufs[i]->format)); - so_data(so, 0xe6); + so_data(so, NV50TCL_RT_FORMAT_X8R8G8B8_UNORM); break; } so_data(so, bo->tile_mode << 4); @@ -92,16 +92,22 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_reloc (so, bo, fb->zsbuf->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW | NOUVEAU_BO_RDWR, 0, 0); switch (fb->zsbuf->format) { + case PIPE_FORMAT_Z32_FLOAT: + so_data(so, NV50TCL_ZETA_FORMAT_Z32_FLOAT); + break; case PIPE_FORMAT_Z24S8_UNORM: - so_data(so, 0x16); + so_data(so, NV50TCL_ZETA_FORMAT_Z24S8_UNORM); + break; + case PIPE_FORMAT_X8Z24_UNORM: + so_data(so, NV50TCL_ZETA_FORMAT_X8Z24_UNORM); break; - case PIPE_FORMAT_Z16_UNORM: - so_data(so, 0x15); + case PIPE_FORMAT_S8Z24_UNORM: + so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM); break; default: NOUVEAU_ERR("AIIII unknown format %s\n", pf_name(fb->zsbuf->format)); - so_data(so, 0x16); + so_data(so, NV50TCL_ZETA_FORMAT_S8Z24_UNORM); break; } so_data(so, bo->tile_mode << 4); @@ -109,7 +115,7 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_method(so, tesla, 0x1538, 1); so_data (so, 1); - so_method(so, tesla, 0x1228, 3); + so_method(so, tesla, NV50TCL_ZETA_HORIZ, 3); so_data (so, fb->zsbuf->width); so_data (so, fb->zsbuf->height); so_data (so, 0x00010001); @@ -118,12 +124,18 @@ nv50_state_validate_fb(struct nv50_context *nv50) so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ, 2); so_data (so, w << 16); so_data (so, h << 16); - so_method(so, tesla, 0x0e04, 2); + /* set window lower left corner */ + so_method(so, tesla, NV50TCL_WINDOW_LEFT, 2); + so_data (so, 0); + so_data (so, 0); + /* set screen scissor rectangle */ + so_method(so, tesla, NV50TCL_SCREEN_SCISSOR_HORIZ, 2); so_data (so, w << 16); so_data (so, h << 16); - so_method(so, tesla, 0xdf8, 2); - so_data (so, 0); - so_data (so, h); + + /* we set scissors to framebuffer size when they're 'turned off' */ + nv50->dirty |= NV50_NEW_SCISSOR; + so_ref(NULL, &nv50->state.scissor); so_ref(so, &nv50->state.fb); so_ref(NULL, &so); @@ -136,7 +148,32 @@ nv50_state_emit(struct nv50_context *nv50) struct nouveau_channel *chan = screen->base.channel; if (nv50->pctx_id != screen->cur_pctx) { - nv50->state.dirty |= 0xffffffff; + if (nv50->state.fb) + nv50->state.dirty |= NV50_NEW_FRAMEBUFFER; + if (nv50->state.blend) + nv50->state.dirty |= NV50_NEW_BLEND; + if (nv50->state.zsa) + nv50->state.dirty |= NV50_NEW_ZSA; + if (nv50->state.vertprog) + nv50->state.dirty |= NV50_NEW_VERTPROG; + if (nv50->state.fragprog) + nv50->state.dirty |= NV50_NEW_FRAGPROG; + if (nv50->state.rast) + nv50->state.dirty |= NV50_NEW_RASTERIZER; + if (nv50->state.blend_colour) + nv50->state.dirty |= NV50_NEW_BLEND_COLOUR; + if (nv50->state.stipple) + nv50->state.dirty |= NV50_NEW_STIPPLE; + if (nv50->state.scissor) + nv50->state.dirty |= NV50_NEW_SCISSOR; + if (nv50->state.viewport) + nv50->state.dirty |= NV50_NEW_VIEWPORT; + if (nv50->state.tsc_upload) + nv50->state.dirty |= NV50_NEW_SAMPLER; + if (nv50->state.tic_upload) + nv50->state.dirty |= NV50_NEW_TEXTURE; + if (nv50->state.vtxfmt && nv50->state.vtxbuf) + nv50->state.dirty |= NV50_NEW_ARRAYS; screen->cur_pctx = nv50->pctx_id; } @@ -167,6 +204,8 @@ nv50_state_emit(struct nv50_context *nv50) if (nv50->state.dirty & NV50_NEW_ARRAYS) { so_emit(chan, nv50->state.vtxfmt); so_emit(chan, nv50->state.vtxbuf); + if (nv50->state.vtxattr) + so_emit(chan, nv50->state.vtxattr); } nv50->state.dirty = 0; @@ -232,13 +271,13 @@ nv50_state_validate(struct nv50_context *nv50) nv50->state.scissor_enabled = rast->scissor; so = so_new(3, 0); - so_method(so, tesla, 0x0ff4, 2); + so_method(so, tesla, NV50TCL_SCISSOR_HORIZ, 2); if (nv50->state.scissor_enabled) { - so_data(so, ((s->maxx - s->minx) << 16) | s->minx); - so_data(so, ((s->maxy - s->miny) << 16) | s->miny); + so_data(so, (s->maxx << 16) | s->minx); + so_data(so, (s->maxy << 16) | s->miny); } else { - so_data(so, (8192 << 16)); - so_data(so, (8192 << 16)); + so_data(so, (nv50->framebuffer.width << 16)); + so_data(so, (nv50->framebuffer.height << 16)); } so_ref(so, &nv50->state.scissor); so_ref(NULL, &so); @@ -262,20 +301,22 @@ scissor_uptodate: so = so_new(12, 0); if (!bypass) { - so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3); + so_method(so, tesla, NV50TCL_VIEWPORT_TRANSLATE(0), 3); so_data (so, fui(nv50->viewport.translate[0])); so_data (so, fui(nv50->viewport.translate[1])); so_data (so, fui(nv50->viewport.translate[2])); - so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3); + so_method(so, tesla, NV50TCL_VIEWPORT_SCALE(0), 3); so_data (so, fui(nv50->viewport.scale[0])); - so_data (so, fui(-nv50->viewport.scale[1])); + so_data (so, fui(nv50->viewport.scale[1])); so_data (so, fui(nv50->viewport.scale[2])); - so_method(so, tesla, 0x192c, 1); + + so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1); so_data (so, 1); + /* no idea what 0f90 does */ so_method(so, tesla, 0x0f90, 1); so_data (so, 0); } else { - so_method(so, tesla, 0x192c, 1); + so_method(so, tesla, NV50TCL_VIEWPORT_TRANSFORM_EN, 1); so_data (so, 0); so_method(so, tesla, 0x0f90, 1); so_data (so, 1); @@ -291,9 +332,10 @@ viewport_uptodate: int i; so = so_new(nv50->sampler_nr * 8 + 3, 0); - so_method(so, tesla, 0x0f00, 1); + so_method(so, tesla, NV50TCL_CB_ADDR, 1); so_data (so, NV50_CB_TSC); - so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8); + so_method(so, tesla, NV50TCL_CB_DATA(0) | 0x40000000, + nv50->sampler_nr * 8); for (i = 0; i < nv50->sampler_nr; i++) so_datap (so, nv50->sampler[i]->tsc, 8); so_ref(so, &nv50->state.tsc_upload); diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c index 8db3b6d344d..edaf4b055a1 100644 --- a/src/gallium/drivers/nv50/nv50_surface.c +++ b/src/gallium/drivers/nv50/nv50_surface.c @@ -35,14 +35,13 @@ nv50_format(enum pipe_format format) { switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_Z24S8_UNORM: - return NV50_2D_DST_FORMAT_32BPP; + return NV50_2D_DST_FORMAT_A8R8G8B8_UNORM; case PIPE_FORMAT_X8R8G8B8_UNORM: - return NV50_2D_DST_FORMAT_24BPP; + return NV50_2D_DST_FORMAT_X8R8G8B8_UNORM; case PIPE_FORMAT_R5G6B5_UNORM: - return NV50_2D_DST_FORMAT_16BPP; + return NV50_2D_DST_FORMAT_R5G6B5_UNORM; case PIPE_FORMAT_A8_UNORM: - return NV50_2D_DST_FORMAT_8BPP; + return NV50_2D_DST_FORMAT_R8_UNORM; default: return -1; } @@ -145,7 +144,7 @@ nv50_surface_copy(struct pipe_context *pipe, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - struct nv50_context *nv50 = (struct nv50_context *)pipe; + struct nv50_context *nv50 = nv50_context(pipe); struct nv50_screen *screen = nv50->screen; assert(src->format == dest->format); @@ -159,7 +158,7 @@ nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest, unsigned destx, unsigned desty, unsigned width, unsigned height, unsigned value) { - struct nv50_context *nv50 = (struct nv50_context *)pipe; + struct nv50_context *nv50 = nv50_context(pipe); struct nv50_screen *screen = nv50->screen; struct nouveau_channel *chan = screen->eng2d->channel; struct nouveau_grobj *eng2d = screen->eng2d; diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c index ff40c2ad81b..14c68b96e14 100644 --- a/src/gallium/drivers/nv50/nv50_tex.c +++ b/src/gallium/drivers/nv50/nv50_tex.c @@ -32,30 +32,30 @@ nv50_tex_construct(struct nv50_context *nv50, struct nouveau_stateobj *so, switch (mt->base.format) { case PIPE_FORMAT_A8R8G8B8_UNORM: so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | - NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM | NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | - NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | NV50TIC_0_0_FMT_8_8_8_8); break; case PIPE_FORMAT_A1R5G5B5_UNORM: so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | - NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM | NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | - NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | NV50TIC_0_0_FMT_1_5_5_5); break; case PIPE_FORMAT_A4R4G4B4_UNORM: so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM | - NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM | NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | - NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | NV50TIC_0_0_FMT_4_4_4_4); break; case PIPE_FORMAT_R5G6B5_UNORM: so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM | - NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM | + NV50TIC_0_0_MAPR_C2 | NV50TIC_0_0_TYPER_UNORM | NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM | - NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM | + NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM | NV50TIC_0_0_FMT_5_6_5); break; case PIPE_FORMAT_L8_UNORM: @@ -145,25 +145,28 @@ nv50_tex_validate(struct nv50_context *nv50) push += MAX2(nv50->miptree_nr, nv50->state.miptree_nr) * 2; so = so_new(push, nv50->miptree_nr * 2); - so_method(so, tesla, 0x0f00, 1); + so_method(so, tesla, NV50TCL_CB_ADDR, 1); so_data (so, NV50_CB_TIC); for (unit = 0; unit < nv50->miptree_nr; unit++) { struct nv50_miptree *mt = nv50->miptree[unit]; - so_method(so, tesla, 0x40000f04, 8); + so_method(so, tesla, NV50TCL_CB_DATA(0) | 0x40000000, 8); if (nv50_tex_construct(nv50, so, mt, unit)) { NOUVEAU_ERR("failed tex validate\n"); so_ref(NULL, &so); return; } - so_method(so, tesla, 0x1458, 1); - so_data (so, (unit << 9) | (unit << 1) | 1); + so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1); + so_data (so, (unit << NV50TCL_SET_SAMPLER_TEX_TIC_SHIFT) | + (unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) | + NV50TCL_SET_SAMPLER_TEX_VALID); } for (; unit < nv50->state.miptree_nr; unit++) { - so_method(so, tesla, 0x1458, 1); - so_data (so, (unit << 1) | 0); + so_method(so, tesla, NV50TCL_SET_SAMPLER_TEX, 1); + so_data (so, + (unit << NV50TCL_SET_SAMPLER_TEX_SAMPLER_SHIFT) | 0); } so_ref(so, &nv50->state.tic_upload); diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h index aca622c73b1..207fb039f70 100644 --- a/src/gallium/drivers/nv50/nv50_texture.h +++ b/src/gallium/drivers/nv50/nv50_texture.h @@ -14,13 +14,13 @@ #define NV50TIC_0_0_MAPA_C2 0x20000000 #define NV50TIC_0_0_MAPA_C3 0x28000000 #define NV50TIC_0_0_MAPA_ONE 0x38000000 -#define NV50TIC_0_0_MAPR_MASK 0x07000000 -#define NV50TIC_0_0_MAPR_ZERO 0x00000000 -#define NV50TIC_0_0_MAPR_C0 0x02000000 -#define NV50TIC_0_0_MAPR_C1 0x03000000 -#define NV50TIC_0_0_MAPR_C2 0x04000000 -#define NV50TIC_0_0_MAPR_C3 0x05000000 -#define NV50TIC_0_0_MAPR_ONE 0x07000000 +#define NV50TIC_0_0_MAPB_MASK 0x07000000 +#define NV50TIC_0_0_MAPB_ZERO 0x00000000 +#define NV50TIC_0_0_MAPB_C0 0x02000000 +#define NV50TIC_0_0_MAPB_C1 0x03000000 +#define NV50TIC_0_0_MAPB_C2 0x04000000 +#define NV50TIC_0_0_MAPB_C3 0x05000000 +#define NV50TIC_0_0_MAPB_ONE 0x07000000 #define NV50TIC_0_0_MAPG_MASK 0x00e00000 #define NV50TIC_0_0_MAPG_ZERO 0x00000000 #define NV50TIC_0_0_MAPG_C0 0x00400000 @@ -28,31 +28,49 @@ #define NV50TIC_0_0_MAPG_C2 0x00800000 #define NV50TIC_0_0_MAPG_C3 0x00a00000 #define NV50TIC_0_0_MAPG_ONE 0x00e00000 -#define NV50TIC_0_0_MAPB_MASK 0x001c0000 -#define NV50TIC_0_0_MAPB_ZERO 0x00000000 -#define NV50TIC_0_0_MAPB_C0 0x00080000 -#define NV50TIC_0_0_MAPB_C1 0x000c0000 -#define NV50TIC_0_0_MAPB_C2 0x00100000 -#define NV50TIC_0_0_MAPB_C3 0x00140000 -#define NV50TIC_0_0_MAPB_ONE 0x001c0000 +#define NV50TIC_0_0_MAPR_MASK 0x001c0000 +#define NV50TIC_0_0_MAPR_ZERO 0x00000000 +#define NV50TIC_0_0_MAPR_C0 0x00080000 +#define NV50TIC_0_0_MAPR_C1 0x000c0000 +#define NV50TIC_0_0_MAPR_C2 0x00100000 +#define NV50TIC_0_0_MAPR_C3 0x00140000 +#define NV50TIC_0_0_MAPR_ONE 0x001c0000 #define NV50TIC_0_0_TYPEA_MASK 0x00038000 #define NV50TIC_0_0_TYPEA_UNORM 0x00010000 -#define NV50TIC_0_0_TYPER_MASK 0x00007000 -#define NV50TIC_0_0_TYPER_UNORM 0x00002000 +#define NV50TIC_0_0_TYPEA_SNORM 0x00008000 +#define NV50TIC_0_0_TYPEA_FLOAT 0x00038000 +#define NV50TIC_0_0_TYPEB_MASK 0x00007000 +#define NV50TIC_0_0_TYPEB_UNORM 0x00002000 +#define NV50TIC_0_0_TYPEB_SNORM 0x00001000 +#define NV50TIC_0_0_TYPEB_FLOAT 0x00007000 #define NV50TIC_0_0_TYPEG_MASK 0x00000e00 #define NV50TIC_0_0_TYPEG_UNORM 0x00000400 -#define NV50TIC_0_0_TYPEB_MASK 0x000001c0 -#define NV50TIC_0_0_TYPEB_UNORM 0x00000080 -#define NV50TIC_0_0_FMT_MASK 0x0000003c +#define NV50TIC_0_0_TYPEG_SNORM 0x00000200 +#define NV50TIC_0_0_TYPEG_FLOAT 0x00000e00 +#define NV50TIC_0_0_TYPER_MASK 0x000001c0 +#define NV50TIC_0_0_TYPER_UNORM 0x00000080 +#define NV50TIC_0_0_TYPER_SNORM 0x00000040 +#define NV50TIC_0_0_TYPER_FLOAT 0x000001c0 +#define NV50TIC_0_0_FMT_MASK 0x0000003f +#define NV50TIC_0_0_FMT_32_32_32_32 0x00000001 +#define NV50TIC_0_0_FMT_16_16_16_16 0x00000003 +#define NV50TIC_0_0_FMT_32_32 0x00000004 #define NV50TIC_0_0_FMT_8_8_8_8 0x00000008 +#define NV50TIC_0_0_FMT_2_10_10_10 0x00000009 +#define NV50TIC_0_0_FMT_32 0x0000000f #define NV50TIC_0_0_FMT_4_4_4_4 0x00000012 -#define NV50TIC_0_0_FMT_1_5_5_5 0x00000013 +/* #define NV50TIC_0_0_FMT_1_5_5_5 0x00000013 */ +#define NV50TIC_0_0_FMT_1_5_5_5 0x00000014 #define NV50TIC_0_0_FMT_5_6_5 0x00000015 #define NV50TIC_0_0_FMT_8_8 0x00000018 +#define NV50TIC_0_0_FMT_16 0x0000001b #define NV50TIC_0_0_FMT_8 0x0000001d +#define NV50TIC_0_0_FMT_10_11_11 0x00000021 #define NV50TIC_0_0_FMT_DXT1 0x00000024 #define NV50TIC_0_0_FMT_DXT3 0x00000025 #define NV50TIC_0_0_FMT_DXT5 0x00000026 +#define NV50TIC_0_0_FMT_RGTC1 0x00000027 +#define NV50TIC_0_0_FMT_RGTC2 0x00000028 #define NV50TIC_0_1_OFFSET_LOW_MASK 0xffffffff #define NV50TIC_0_1_OFFSET_LOW_SHIFT 0 @@ -102,6 +120,7 @@ #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_EDGE 0x00000140 #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_BORDER 0x00000180 #define NV50TSC_1_0_WRAPR_MIRROR_CLAMP 0x000001c0 +#define NV50TSC_1_0_MAX_ANISOTROPY_MASK 0x00700000 #define NV50TSC_1_1_MAGF_MASK 0x00000003 #define NV50TSC_1_1_MAGF_NEAREST 0x00000001 @@ -113,17 +132,19 @@ #define NV50TSC_1_1_MIPF_NONE 0x00000040 #define NV50TSC_1_1_MIPF_NEAREST 0x00000080 #define NV50TSC_1_1_MIPF_LINEAR 0x000000c0 +#define NV50TSC_1_1_LOD_BIAS_MASK 0x01fff000 -#define NV50TSC_1_2_UNKNOWN_MASK 0xffffffff +#define NV50TSC_1_2_MIN_LOD_MASK 0x00000f00 +#define NV50TSC_1_2_MAX_LOD_MASK 0x00f00000 #define NV50TSC_1_3_UNKNOWN_MASK 0xffffffff -#define NV50TSC_1_4_UNKNOWN_MASK 0xffffffff +#define NV50TSC_1_4_BORDER_COLOR_RED_MASK 0xffffffff -#define NV50TSC_1_5_UNKNOWN_MASK 0xffffffff +#define NV50TSC_1_5_BORDER_COLOR_GREEN_MASK 0xffffffff -#define NV50TSC_1_6_UNKNOWN_MASK 0xffffffff +#define NV50TSC_1_6_BORDER_COLOR_BLUE_MASK 0xffffffff -#define NV50TSC_1_7_UNKNOWN_MASK 0xffffffff +#define NV50TSC_1_7_BORDER_COLOR_ALPHA_MASK 0xffffffff #endif diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index d0b7f0bef43..d2b5e4d75d4 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -8,6 +8,7 @@ struct nv50_transfer { struct pipe_transfer base; struct nouveau_bo *bo; unsigned level_offset; + unsigned level_tiling; int level_pitch; int level_width; int level_height; @@ -16,11 +17,14 @@ struct nv50_transfer { }; static void -nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, - unsigned src_offset, int src_pitch, int sx, int sy, - int sw, int sh, struct nouveau_bo *dst_bo, - unsigned dst_offset, int dst_pitch, int dx, int dy, - int dw, int dh, int cpp, int width, int height, +nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, + struct nouveau_bo *src_bo, unsigned src_offset, + int src_pitch, unsigned src_tile_mode, + int sx, int sy, int sw, int sh, + struct nouveau_bo *dst_bo, unsigned dst_offset, + int dst_pitch, unsigned dst_tile_mode, + int dx, int dy, int dw, int dh, + int cpp, int width, int height, unsigned src_reloc, unsigned dst_reloc) { struct nv50_screen *screen = nv50_screen(pscreen); @@ -33,15 +37,18 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, WAIT_RING (chan, 14); if (!src_bo->tile_flags) { - BEGIN_RING(chan, m2mf, 0x0200, 1); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 1); OUT_RING (chan, 1); - BEGIN_RING(chan, m2mf, 0x0314, 1); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_IN, 1); OUT_RING (chan, src_pitch); src_offset += (sy * src_pitch) + (sx * cpp); } else { - BEGIN_RING(chan, m2mf, 0x0200, 6); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_IN, 6); OUT_RING (chan, 0); - OUT_RING (chan, src_bo->tile_mode << 4); + OUT_RING (chan, src_tile_mode << 4); OUT_RING (chan, sw * cpp); OUT_RING (chan, sh); OUT_RING (chan, 1); @@ -49,15 +56,18 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, } if (!dst_bo->tile_flags) { - BEGIN_RING(chan, m2mf, 0x021c, 1); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 1); OUT_RING (chan, 1); - BEGIN_RING(chan, m2mf, 0x0318, 1); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT, 1); OUT_RING (chan, dst_pitch); dst_offset += (dy * dst_pitch) + (dx * cpp); } else { - BEGIN_RING(chan, m2mf, 0x021c, 6); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_LINEAR_OUT, 6); OUT_RING (chan, 0); - OUT_RING (chan, dst_bo->tile_mode << 4); + OUT_RING (chan, dst_tile_mode << 4); OUT_RING (chan, dw * cpp); OUT_RING (chan, dh); OUT_RING (chan, 1); @@ -68,25 +78,30 @@ nv50_transfer_rect_m2mf(struct pipe_screen *pscreen, struct nouveau_bo *src_bo, int line_count = height > 2047 ? 2047 : height; WAIT_RING (chan, 15); - BEGIN_RING(chan, m2mf, 0x0238, 2); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN_HIGH, 2); OUT_RELOCh(chan, src_bo, src_offset, src_reloc); OUT_RELOCh(chan, dst_bo, dst_offset, dst_reloc); - BEGIN_RING(chan, m2mf, 0x030c, 2); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 2); OUT_RELOCl(chan, src_bo, src_offset, src_reloc); OUT_RELOCl(chan, dst_bo, dst_offset, dst_reloc); if (src_bo->tile_flags) { - BEGIN_RING(chan, m2mf, 0x0218, 1); - OUT_RING (chan, (dy << 16) | sx); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_IN, 1); + OUT_RING (chan, (sy << 16) | sx); } else { src_offset += (line_count * src_pitch); } if (dst_bo->tile_flags) { - BEGIN_RING(chan, m2mf, 0x0234, 1); - OUT_RING (chan, (sy << 16) | dx); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_TILING_POSITION_OUT, 1); + OUT_RING (chan, (dy << 16) | dx); } else { dst_offset += (line_count * dst_pitch); } - BEGIN_RING(chan, m2mf, 0x031c, 4); + BEGIN_RING(chan, m2mf, + NV50_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN, 4); OUT_RING (chan, width * cpp); OUT_RING (chan, line_count); OUT_RING (chan, 0x00000101); @@ -136,6 +151,7 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->level_width = mt->base.width[level]; tx->level_height = mt->base.height[level]; tx->level_offset = lvl->image_offset[image]; + tx->level_tiling = lvl->tile_mode; tx->level_x = x; tx->level_y = y; ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, @@ -147,9 +163,11 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, if (usage != PIPE_TRANSFER_WRITE) { nv50_transfer_rect_m2mf(pscreen, mt->bo, tx->level_offset, - tx->level_pitch, x, y, tx->level_width, - tx->level_height, tx->bo, 0, - tx->base.stride, 0, 0, + tx->level_pitch, tx->level_tiling, + x, y, + tx->level_width, tx->level_height, + tx->bo, 0, tx->base.stride, + tx->bo->tile_mode, 0, 0, tx->base.width, tx->base.height, tx->base.block.size, w, h, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, @@ -168,12 +186,14 @@ nv50_transfer_del(struct pipe_transfer *ptx) if (ptx->usage != PIPE_TRANSFER_READ) { struct pipe_screen *pscreen = ptx->texture->screen; nv50_transfer_rect_m2mf(pscreen, tx->bo, 0, tx->base.stride, - 0, 0, tx->base.width, tx->base.height, - mt->bo, tx->level_offset, - tx->level_pitch, tx->level_x, - tx->level_y, tx->level_width, - tx->level_height, tx->base.block.size, + tx->bo->tile_mode, 0, 0, tx->base.width, tx->base.height, + mt->bo, tx->level_offset, + tx->level_pitch, tx->level_tiling, + tx->level_x, tx->level_y, + tx->level_width, tx->level_height, + tx->base.block.size, tx->base.width, + tx->base.height, NOUVEAU_BO_GART, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART); } diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c index f81929f2387..36f1b24b2f0 100644 --- a/src/gallium/drivers/nv50/nv50_vbo.c +++ b/src/gallium/drivers/nv50/nv50_vbo.c @@ -49,6 +49,82 @@ nv50_prim(unsigned mode) return NV50TCL_VERTEX_BEGIN_POINTS; } +static INLINE uint32_t +nv50_vbo_type_to_hw(unsigned type) +{ + switch (type) { + case PIPE_FORMAT_TYPE_FLOAT: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_FLOAT; + case PIPE_FORMAT_TYPE_UNORM: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UNORM; + case PIPE_FORMAT_TYPE_SNORM: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SNORM; + case PIPE_FORMAT_TYPE_USCALED: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_USCALED; + case PIPE_FORMAT_TYPE_SSCALED: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SSCALED; + /* + case PIPE_FORMAT_TYPE_UINT: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_UINT; + case PIPE_FORMAT_TYPE_SINT: + return NV50TCL_VERTEX_ARRAY_ATTRIB_TYPE_SINT; */ + default: + return 0; + } +} + +static INLINE uint32_t +nv50_vbo_size_to_hw(unsigned size, unsigned nr_c) +{ + static const uint32_t hw_values[] = { + 0, 0, 0, 0, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_8_8_8_8, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_16_16_16_16, + 0, 0, 0, 0, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32, + NV50TCL_VERTEX_ARRAY_ATTRIB_SIZE_32_32_32_32 }; + + /* we'd also have R11G11B10 and R10G10B10A2 */ + + assert(nr_c > 0 && nr_c <= 4); + + if (size > 32) + return 0; + size >>= (3 - 2); + + return hw_values[size + (nr_c - 1)]; +} + +static INLINE uint32_t +nv50_vbo_vtxelt_to_hw(struct pipe_vertex_element *ve) +{ + uint32_t hw_type, hw_size; + enum pipe_format pf = ve->src_format; + unsigned size = pf_size_x(pf) << pf_exp2(pf); + + hw_type = nv50_vbo_type_to_hw(pf_type(pf)); + hw_size = nv50_vbo_size_to_hw(size, ve->nr_components); + + if (!hw_type || !hw_size) { + NOUVEAU_ERR("unsupported vbo format: %s\n", pf_name(pf)); + abort(); + return 0x24e80000; + } + + if (pf_swizzle_x(pf) == 2) /* BGRA */ + hw_size |= (1 << 31); /* no real swizzle bits :-( */ + + return (hw_type | hw_size); +} + boolean nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) @@ -139,7 +215,7 @@ nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map, } static INLINE void -nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map, +nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint32_t *map, unsigned start, unsigned count) { struct nouveau_channel *chan = nv50->screen->tesla->channel; @@ -201,49 +277,101 @@ nv50_draw_elements(struct pipe_context *pipe, return TRUE; } +static INLINE boolean +nv50_vbo_static_attrib(struct nv50_context *nv50, unsigned attrib, + struct nouveau_stateobj **pso, + struct pipe_vertex_element *ve, + struct pipe_vertex_buffer *vb) + +{ + struct nouveau_stateobj *so; + struct nouveau_grobj *tesla = nv50->screen->tesla; + struct nouveau_bo *bo = nouveau_bo(vb->buffer); + float *v; + int ret; + enum pipe_format pf = ve->src_format; + + if ((pf_type(pf) != PIPE_FORMAT_TYPE_FLOAT) || + (pf_size_x(pf) << pf_exp2(pf)) != 32) + return FALSE; + + ret = nouveau_bo_map(bo, NOUVEAU_BO_RD); + if (ret) + return FALSE; + v = (float *)(bo->map + (vb->buffer_offset + ve->src_offset)); + + so = *pso; + if (!so) + *pso = so = so_new(nv50->vtxelt_nr * 5, 0); + + switch (ve->nr_components) { + case 4: + so_method(so, tesla, NV50TCL_VTX_ATTR_4F_X(attrib), 4); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + so_data (so, fui(v[3])); + break; + case 3: + so_method(so, tesla, NV50TCL_VTX_ATTR_3F_X(attrib), 4); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + so_data (so, fui(v[2])); + break; + case 2: + so_method(so, tesla, NV50TCL_VTX_ATTR_2F_X(attrib), 4); + so_data (so, fui(v[0])); + so_data (so, fui(v[1])); + break; + case 1: + so_method(so, tesla, NV50TCL_VTX_ATTR_1F(attrib), 4); + so_data (so, fui(v[0])); + break; + default: + nouveau_bo_unmap(bo); + return FALSE; + } + + nouveau_bo_unmap(bo); + return TRUE; +} + void nv50_vbo_validate(struct nv50_context *nv50) { struct nouveau_grobj *tesla = nv50->screen->tesla; - struct nouveau_stateobj *vtxbuf, *vtxfmt; - int i; + struct nouveau_stateobj *vtxbuf, *vtxfmt, *vtxattr; + unsigned i; + + /* don't validate if Gallium took away our buffers */ + if (nv50->vtxbuf_nr == 0) + return; - vtxbuf = so_new(nv50->vtxelt_nr * 4, nv50->vtxelt_nr * 2); + vtxattr = NULL; + vtxbuf = so_new(nv50->vtxelt_nr * 7, nv50->vtxelt_nr * 4); vtxfmt = so_new(nv50->vtxelt_nr + 1, 0); - so_method(vtxfmt, tesla, 0x1ac0, nv50->vtxelt_nr); + so_method(vtxfmt, tesla, NV50TCL_VERTEX_ARRAY_ATTRIB(0), + nv50->vtxelt_nr); for (i = 0; i < nv50->vtxelt_nr; i++) { struct pipe_vertex_element *ve = &nv50->vtxelt[i]; struct pipe_vertex_buffer *vb = &nv50->vtxbuf[ve->vertex_buffer_index]; struct nouveau_bo *bo = nouveau_bo(vb->buffer); + uint32_t hw = nv50_vbo_vtxelt_to_hw(ve); - switch (ve->src_format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - so_data(vtxfmt, 0x7e080000 | i); - break; - case PIPE_FORMAT_R32G32B32_FLOAT: - so_data(vtxfmt, 0x7e100000 | i); - break; - case PIPE_FORMAT_R32G32_FLOAT: - so_data(vtxfmt, 0x7e200000 | i); - break; - case PIPE_FORMAT_R32_FLOAT: - so_data(vtxfmt, 0x7e900000 | i); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - so_data(vtxfmt, 0x24500000 | i); - break; - default: - { - NOUVEAU_ERR("invalid vbo format %s\n", - pf_name(ve->src_format)); - assert(0); - return; - } + if (!vb->stride && + nv50_vbo_static_attrib(nv50, i, &vtxattr, ve, vb)) { + so_data(vtxfmt, hw | (1 << 4)); + + so_method(vtxbuf, tesla, + NV50TCL_VERTEX_ARRAY_FORMAT(i), 1); + so_data (vtxbuf, 0); + continue; } + so_data(vtxfmt, hw | i); - so_method(vtxbuf, tesla, 0x900 + (i * 16), 3); + so_method(vtxbuf, tesla, NV50TCL_VERTEX_ARRAY_FORMAT(i), 3); so_data (vtxbuf, 0x20000000 | vb->stride); so_reloc (vtxbuf, bo, vb->buffer_offset + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | @@ -251,11 +379,22 @@ nv50_vbo_validate(struct nv50_context *nv50) so_reloc (vtxbuf, bo, vb->buffer_offset + ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0); + + /* vertex array limits */ + so_method(vtxbuf, tesla, 0x1080 + (i * 8), 2); + so_reloc (vtxbuf, bo, vb->buffer->size - 1, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | + NOUVEAU_BO_HIGH, 0, 0); + so_reloc (vtxbuf, bo, vb->buffer->size - 1, + NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD | + NOUVEAU_BO_LOW, 0, 0); } so_ref (vtxfmt, &nv50->state.vtxfmt); so_ref (vtxbuf, &nv50->state.vtxbuf); + so_ref (vtxattr, &nv50->state.vtxattr); so_ref (NULL, &vtxbuf); so_ref (NULL, &vtxfmt); + so_ref (NULL, &vtxattr); } diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile index e44f9b9dfc5..d7a2c8c462c 100644 --- a/src/gallium/drivers/r300/Makefile +++ b/src/gallium/drivers/r300/Makefile @@ -4,21 +4,38 @@ include $(TOP)/configs/current LIBNAME = r300 C_SOURCES = \ + r3xx_fs.c \ + r5xx_fs.c \ r300_chipset.c \ r300_clear.c \ r300_context.c \ - r300_debug.c \ r300_emit.c \ r300_flush.c \ + r300_fs.c \ r300_query.c \ r300_render.c \ r300_screen.c \ r300_state.c \ r300_state_derived.c \ r300_state_invariant.c \ - r300_state_shader.c \ - r300_state_tcl.c \ + r300_vs.c \ r300_surface.c \ - r300_texture.c + r300_texture.c \ + r300_tgsi_to_rc.c + +LIBRARY_INCLUDES = \ + -I$(TOP)/src/mesa/drivers/dri/r300/compiler \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include + +COMPILER_ARCHIVE = $(TOP)/src/mesa/drivers/dri/r300/compiler/libr300compiler.a + +EXTRA_OBJECTS = \ + $(COMPILER_ARCHIVE) include ../../Makefile.template + +.PHONY : $(COMPILER_ARCHIVE) + +$(COMPILER_ARCHIVE): + cd $(TOP)/src/mesa/drivers/dri/r300/compiler; make diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript index 182ed2d459a..493d7b28bc3 100644 --- a/src/gallium/drivers/r300/SConscript +++ b/src/gallium/drivers/r300/SConscript @@ -5,20 +5,22 @@ env = env.Clone() r300 = env.ConvenienceLibrary( target = 'r300', source = [ + 'r3xx_fs.c', + 'r5xx_fs.c', 'r300_chipset.c', 'r300_clear.c', 'r300_context.c', 'r300_debug.c', 'r300_emit.c', 'r300_flush.c', + 'r300_fs.c', 'r300_query.c', 'r300_render.c', 'r300_screen.c', 'r300_state.c', 'r300_state_derived.c', 'r300_state_invariant.c', - 'r300_state_shader.c', - 'r300_state_tcl.c', + 'r300_vs.c', 'r300_surface.c', 'r300_texture.c', ]) diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c index 00fae8d26f3..d138866d33c 100644 --- a/src/gallium/drivers/r300/r300_chipset.c +++ b/src/gallium/drivers/r300/r300_chipset.c @@ -30,9 +30,11 @@ void r300_parse_chipset(struct r300_capabilities* caps) { /* Reasonable defaults */ + caps->num_vert_fpus = 4; caps->has_tcl = getenv("RADEON_NO_TCL") ? FALSE : TRUE; caps->is_r500 = FALSE; - caps->num_vert_fpus = 4; + caps->high_second_pipe = FALSE; + /* Note: These are not ordered by PCI ID. I leave that task to GCC, * which will perform the ordering while collating jump tables. Instead, @@ -40,6 +42,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) switch (caps->pci_id) { case 0x4144: caps->family = CHIP_FAMILY_R300; + caps->high_second_pipe = TRUE; break; case 0x4145: @@ -50,6 +53,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) case 0x4E46: case 0x4E47: caps->family = CHIP_FAMILY_R300; + caps->high_second_pipe = TRUE; break; case 0x4150: @@ -66,6 +70,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) case 0x4E54: case 0x4E56: caps->family = CHIP_FAMILY_RV350; + caps->high_second_pipe = TRUE; break; case 0x4148: @@ -76,10 +81,12 @@ void r300_parse_chipset(struct r300_capabilities* caps) case 0x4E49: case 0x4E4B: caps->family = CHIP_FAMILY_R350; + caps->high_second_pipe = TRUE; break; case 0x4E4A: caps->family = CHIP_FAMILY_R360; + caps->high_second_pipe = TRUE; break; case 0x5460: @@ -91,6 +98,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) case 0x5B64: case 0x5B65: caps->family = CHIP_FAMILY_RV370; + caps->high_second_pipe = TRUE; break; case 0x3150: @@ -99,6 +107,7 @@ void r300_parse_chipset(struct r300_capabilities* caps) case 0x3E50: case 0x3E54: caps->family = CHIP_FAMILY_RV380; + caps->high_second_pipe = TRUE; break; case 0x4A48: diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h index 5b2e1f05688..322d4a57e41 100644 --- a/src/gallium/drivers/r300/r300_chipset.h +++ b/src/gallium/drivers/r300/r300_chipset.h @@ -34,6 +34,8 @@ struct r300_capabilities { int family; /* The number of vertex floating-point units */ int num_vert_fpus; + /* The number of fragment pipes */ + int num_frag_pipes; /* Whether or not TCL is physically present */ boolean has_tcl; /* Whether or not this is an RV515 or newer; R500s have many differences @@ -42,6 +44,8 @@ struct r300_capabilities { * - Blend color is split across two registers * - Universal Shader (US) block used for fragment shaders */ boolean is_r500; + /* Whether or not the second pixel pipe is accessed with the high bit */ + boolean high_second_pipe; }; /* Enumerations for legibility and telling which card we're running on. */ diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 233a32b53c5..da67bc29b89 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -52,7 +52,7 @@ static boolean r300_draw_range_elements(struct pipe_context* pipe, draw_set_mapped_constant_buffer(r300->draw, r300->shader_constants[PIPE_SHADER_VERTEX].constants, - r300->shader_constants[PIPE_SHADER_VERTEX].user_count * + r300->shader_constants[PIPE_SHADER_VERTEX].count * (sizeof(float) * 4)); draw_arrays(r300->draw, mode, start, count); @@ -88,9 +88,21 @@ static boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, static void r300_destroy_context(struct pipe_context* context) { struct r300_context* r300 = r300_context(context); + struct r300_query* query, * temp; draw_destroy(r300->draw); + /* Free the OQ BO. */ + context->screen->buffer_destroy(r300->oqbo); + + /* If there are any queries pending or not destroyed, remove them now. */ + if (r300->query_list) { + foreach_s(query, temp, r300->query_list) { + remove_from_list(query); + FREE(query); + } + } + FREE(r300->blend_color_state); FREE(r300->rs_block); FREE(r300->scissor_state); @@ -145,6 +157,11 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.is_texture_referenced = r300_is_texture_referenced; r300->context.is_buffer_referenced = r300_is_buffer_referenced; + r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); + r300->rs_block = CALLOC_STRUCT(r300_rs_block); + r300->scissor_state = CALLOC_STRUCT(r300_scissor_state); + r300->viewport_state = CALLOC_STRUCT(r300_viewport_state); + /* Create a Draw. This is used for vert collation and SW TCL. */ r300->draw = draw_create(); /* Enable our renderer. */ @@ -155,10 +172,9 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, * transform in hardware, always. */ draw_set_viewport_state(r300->draw, &r300_viewport_identity); - r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); - r300->rs_block = CALLOC_STRUCT(r300_rs_block); - r300->scissor_state = CALLOC_STRUCT(r300_scissor_state); - r300->viewport_state = CALLOC_STRUCT(r300_viewport_state); + /* Open up the OQ BO. */ + r300->oqbo = screen->buffer_create(screen, 4096, + PIPE_BUFFER_USAGE_VERTEX, 4096); r300_init_flush_functions(r300); diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 27bc7fd1a93..f78492d4aa9 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -25,15 +25,22 @@ #include "draw/draw_context.h" #include "draw/draw_vertex.h" + #include "pipe/p_context.h" + #include "tgsi/tgsi_scan.h" + #include "util/u_memory.h" +#include "util/u_simple_list.h" #include "r300_clear.h" #include "r300_query.h" #include "r300_screen.h" #include "r300_winsys.h" +struct r300_fragment_shader; +struct r300_vertex_shader; + struct r300_blend_state { uint32_t blend_control; /* R300_RB3D_CBLEND: 0x4e04 */ uint32_t alpha_blend_control; /* R300_RB3D_ABLEND: 0x4e08 */ @@ -143,69 +150,31 @@ struct r300_constant_buffer { /* Buffer of constants */ /* XXX first number should be raised */ float constants[32][4]; - /* Number of user-defined constants */ - unsigned user_count; /* Total number of constants */ unsigned count; }; -struct r3xx_fragment_shader { - /* Parent class */ - struct pipe_shader_state state; - struct tgsi_shader_info info; - - /* Has this shader been translated yet? */ - boolean translated; - - /* Pixel stack size */ - int stack_size; - - /* Are there immediates in this shader? - * If not, we can heavily optimize recompilation. */ - boolean uses_imms; -}; - -struct r300_fragment_shader { - /* Parent class */ - struct r3xx_fragment_shader shader; - - /* Number of ALU instructions */ - int alu_instruction_count; - - /* Number of texture instructions */ - int tex_instruction_count; - - /* Number of texture indirections */ - int indirections; - - /* Indirection node offsets */ - int alu_offset[4]; - - /* Machine instructions */ - struct { - uint32_t alu_rgb_inst; - uint32_t alu_rgb_addr; - uint32_t alu_alpha_inst; - uint32_t alu_alpha_addr; - } instructions[64]; /* XXX magic num */ -}; - -struct r500_fragment_shader { - /* Parent class */ - struct r3xx_fragment_shader shader; - - /* Number of used instructions */ - int instruction_count; - - /* Machine instructions */ - struct { - uint32_t inst0; - uint32_t inst1; - uint32_t inst2; - uint32_t inst3; - uint32_t inst4; - uint32_t inst5; - } instructions[256]; /*< XXX magic number */ +/* Query object. + * + * This is not a subclass of pipe_query because pipe_query is never + * actually fully defined. So, rather than have it as a member, and do + * subclass-style casting, we treat pipe_query as an opaque, and just + * trust that our state tracker does not ever mess up query objects. + */ +struct r300_query { + /* The kind of query. Currently only OQ is supported. */ + unsigned type; + /* Whether this query is currently active. Only active queries will + * get emitted into the command stream, and only active queries get + * tallied. */ + boolean active; + /* The current count of this query. Required to be at least 32 bits. */ + unsigned int count; + /* The offset of this query into the query buffer, in bytes. */ + unsigned offset; + /* Linked list members. */ + struct r300_query* prev; + struct r300_query* next; }; struct r300_texture { @@ -217,7 +186,7 @@ struct r300_texture { /* Stride (pitch?) of this texture in bytes */ unsigned stride; - + /* Total size of this texture, in bytes. */ unsigned size; @@ -242,33 +211,6 @@ struct r300_vertex_format { int fs_tab[16]; }; -struct r300_vertex_shader { - /* Parent class */ - struct pipe_shader_state state; - struct tgsi_shader_info info; - - /* Fallback shader, because Draw has issues */ - struct draw_vertex_shader* draw; - - /* Has this shader been translated yet? */ - boolean translated; - - /* Are there immediates in this shader? - * If not, we can heavily optimize recompilation. */ - boolean uses_imms; - - /* Number of used instructions */ - int instruction_count; - - /* Machine instructions */ - struct { - uint32_t inst0; - uint32_t inst1; - uint32_t inst2; - uint32_t inst3; - } instructions[128]; /*< XXX magic number */ -}; - static struct pipe_viewport_state r300_viewport_identity = { .scale = {1.0, 1.0, 1.0, 1.0}, .translate = {0.0, 0.0, 0.0, 0.0}, @@ -288,6 +230,11 @@ struct r300_context { /* Offset into the VBO. */ size_t vbo_offset; + /* Occlusion query buffer. */ + struct pipe_buffer* oqbo; + /* Query list. */ + struct r300_query* query_list; + /* Various CSO state objects. */ /* Blend state. */ struct r300_blend_state* blend_state; @@ -300,7 +247,7 @@ struct r300_context { /* Depth, stencil, and alpha state. */ struct r300_dsa_state* dsa_state; /* Fragment shader. */ - struct r3xx_fragment_shader* fs; + struct r300_fragment_shader* fs; /* Framebuffer state. We currently don't need our own version of this. */ struct pipe_framebuffer_state framebuffer_state; /* Rasterizer state. */ @@ -331,7 +278,8 @@ struct r300_context { }; /* Convenience cast wrapper. */ -static struct r300_context* r300_context(struct pipe_context* context) { +static INLINE struct r300_context* r300_context(struct pipe_context* context) +{ return (struct r300_context*)context; } diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c deleted file mode 100644 index 678cd2b8121..00000000000 --- a/src/gallium/drivers/r300/r300_debug.c +++ /dev/null @@ -1,198 +0,0 @@ -/* - * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#include "r300_debug.h" - -static void r300_dump_fs(struct r300_fragment_shader* fs) -{ - int i; - - for (i = 0; i < fs->alu_instruction_count; i++) { - } -} - -void r500_fs_dump(struct r500_fragment_shader* fs) -{ - int i; - uint32_t inst; - - for (i = 0; i < fs->instruction_count; i++) { - inst = fs->instructions[i].inst0; - debug_printf("%d: 0: CMN_INST 0x%08x:", i, inst); - switch (inst & 0x3) { - case R500_INST_TYPE_ALU: - debug_printf("ALU "); - break; - case R500_INST_TYPE_OUT: - debug_printf("OUT "); - break; - case R500_INST_TYPE_FC: - debug_printf("FC "); - break; - case R500_INST_TYPE_TEX: - debug_printf("TEX "); - break; - } - debug_printf("%s %s %s %s ", - inst & R500_INST_TEX_SEM_WAIT ? "TEX_WAIT" : "", - inst & R500_INST_LAST ? "LAST" : "", - inst & R500_INST_NOP ? "NOP" : "", - inst & R500_INST_ALU_WAIT ? "ALU_WAIT" : ""); - debug_printf("wmask: %s omask: %s\n", - r500_fs_mask[(inst >> 11) & 0xf], - r500_fs_mask[(inst >> 15) & 0xf]); - switch (inst & 0x3) { - case R500_INST_TYPE_ALU: - case R500_INST_TYPE_OUT: - inst = fs->instructions[i].inst1; - debug_printf(" 1: RGB_ADDR 0x%08x:", inst); - debug_printf("Addr0: %d%c, Addr1: %d%c, " - "Addr2: %d%c, srcp:%d\n", - inst & 0xff, (inst & (1 << 8)) ? 'c' : 't', - (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't', - (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't', - (inst >> 30)); - - inst = fs->instructions[i].inst2; - debug_printf(" 2: ALPHA_ADDR 0x%08x:", inst); - debug_printf("Addr0: %d%c, Addr1: %d%c, " - "Addr2: %d%c, srcp:%d\n", - inst & 0xff, (inst & (1 << 8)) ? 'c' : 't', - (inst >> 10) & 0xff, (inst & (1 << 18)) ? 'c' : 't', - (inst >> 20) & 0xff, (inst & (1 << 28)) ? 'c' : 't', - (inst >> 30)); - - inst = fs->instructions[i].inst3; - debug_printf(" 3: RGB_INST 0x%08x:", inst); - debug_printf("rgb_A_src:%d %s/%s/%s %d " - "rgb_B_src:%d %s/%s/%s %d\n", - inst & 0x3, r500_fs_swiz[(inst >> 2) & 0x7], - r500_fs_swiz[(inst >> 5) & 0x7], - r500_fs_swiz[(inst >> 8) & 0x7], - (inst >> 11) & 0x3, (inst >> 13) & 0x3, - r500_fs_swiz[(inst >> 15) & 0x7], - r500_fs_swiz[(inst >> 18) & 0x7], - r500_fs_swiz[(inst >> 21) & 0x7], - (inst >> 24) & 0x3); - - inst = fs->instructions[i].inst4; - debug_printf(" 4: ALPHA_INST 0x%08x:", inst); - debug_printf("%s dest:%d%s alp_A_src:%d %s %d " - "alp_B_src:%d %s %d w:%d\n", - r500_fs_op_alpha[inst & 0xf], (inst >> 4) & 0x7f, - inst & (1<<11) ? "(rel)":"", (inst >> 12) & 0x3, - r500_fs_swiz[(inst >> 14) & 0x7], (inst >> 17) & 0x3, - (inst >> 19) & 0x3, r500_fs_swiz[(inst >> 21) & 0x7], - (inst >> 24) & 0x3, (inst >> 31) & 0x1); - - inst = fs->instructions[i].inst5; - debug_printf(" 5: RGBA_INST 0x%08x:", inst); - debug_printf("%s dest:%d%s rgb_C_src:%d %s/%s/%s %d " - "alp_C_src:%d %s %d\n", - r500_fs_op_rgb[inst & 0xf], (inst >> 4) & 0x7f, - inst & (1 << 11) ? "(rel)":"", (inst >> 12) & 0x3, - r500_fs_swiz[(inst >> 14) & 0x7], - r500_fs_swiz[(inst >> 17) & 0x7], - r500_fs_swiz[(inst >> 20) & 0x7], - (inst >> 23) & 0x3, (inst >> 25) & 0x3, - r500_fs_swiz[(inst >> 27) & 0x7], (inst >> 30) & 0x3); - break; - case R500_INST_TYPE_FC: - /* XXX don't even bother yet */ - break; - case R500_INST_TYPE_TEX: - inst = fs->instructions[i].inst1; - debug_printf(" 1: TEX_INST 0x%08x: id: %d " - "op:%s, %s, %s %s\n", - inst, (inst >> 16) & 0xf, - r500_fs_tex[(inst >> 22) & 0x7], - (inst & (1 << 25)) ? "ACQ" : "", - (inst & (1 << 26)) ? "IGNUNC" : "", - (inst & (1 << 27)) ? "UNSCALED" : "SCALED"); - - inst = fs->instructions[i].inst2; - debug_printf(" 2: TEX_ADDR 0x%08x: " - "src: %d%s %s/%s/%s/%s dst: %d%s %s/%s/%s/%s\n", - inst, inst & 0x7f, inst & (1 << 7) ? "(rel)" : "", - r500_fs_swiz[(inst >> 8) & 0x3], - r500_fs_swiz[(inst >> 10) & 0x3], - r500_fs_swiz[(inst >> 12) & 0x3], - r500_fs_swiz[(inst >> 14) & 0x3], - (inst >> 16) & 0x7f, inst & (1 << 23) ? "(rel)" : "", - r500_fs_swiz[(inst >> 24) & 0x3], - r500_fs_swiz[(inst >> 26) & 0x3], - r500_fs_swiz[(inst >> 28) & 0x3], - r500_fs_swiz[(inst >> 30) & 0x3]); - - inst = fs->instructions[i].inst3; - debug_printf(" 3: TEX_DXDY 0x%08x\n", inst); - break; - } - } -} - -static void r300_vs_op_dump(uint32_t op) -{ - debug_printf(" dst: %d%s op: ", - (op >> 13) & 0x7f, r300_vs_dst_debug[(op >> 8) & 0x7]); - if (op & 0x80) { - if (op & 0x1) { - debug_printf("PVS_MACRO_OP_2CLK_M2X_ADD\n"); - } else { - debug_printf(" PVS_MACRO_OP_2CLK_MADD\n"); - } - } else if (op & 0x40) { - debug_printf("%s\n", r300_vs_me_ops[op & 0x1f]); - } else { - debug_printf("%s\n", r300_vs_ve_ops[op & 0x1f]); - } -} - -void r300_vs_src_dump(uint32_t src) -{ - debug_printf(" reg: %d%s swiz: %s%s/%s%s/%s%s/%s%s\n", - (src >> 5) & 0x7f, r300_vs_src_debug[src & 0x3], - src & (1 << 25) ? "-" : " ", - r300_vs_swiz_debug[(src >> 13) & 0x7], - src & (1 << 26) ? "-" : " ", - r300_vs_swiz_debug[(src >> 16) & 0x7], - src & (1 << 27) ? "-" : " ", - r300_vs_swiz_debug[(src >> 19) & 0x7], - src & (1 << 28) ? "-" : " ", - r300_vs_swiz_debug[(src >> 22) & 0x7]); -} - -void r300_vs_dump(struct r300_vertex_shader* vs) -{ - int i; - - for (i = 0; i < vs->instruction_count; i++) { - debug_printf("%d: op: 0x%08x", i, vs->instructions[i].inst0); - r300_vs_op_dump(vs->instructions[i].inst0); - debug_printf(" src0: 0x%08x", vs->instructions[i].inst1); - r300_vs_src_dump(vs->instructions[i].inst1); - debug_printf(" src1: 0x%08x", vs->instructions[i].inst2); - r300_vs_src_dump(vs->instructions[i].inst2); - debug_printf(" src2: 0x%08x", vs->instructions[i].inst3); - r300_vs_src_dump(vs->instructions[i].inst3); - } -} diff --git a/src/gallium/drivers/r300/r300_debug.h b/src/gallium/drivers/r300/r300_debug.h deleted file mode 100644 index c86410ec0a9..00000000000 --- a/src/gallium/drivers/r300/r300_debug.h +++ /dev/null @@ -1,210 +0,0 @@ -/* - * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#ifndef R300_DEBUG_H -#define R300_DEBUG_H - -#include "r300_reg.h" -#include "r300_state_shader.h" -#include "r300_state_tcl.h" - -static char* r500_fs_swiz[] = { - " R", - " G", - " B", - " A", - " 0", - ".5", - " 1", - " U", -}; - -static char* r500_fs_op_rgb[] = { - "MAD", - "DP3", - "DP4", - "D2A", - "MIN", - "MAX", - "---", - "CND", - "CMP", - "FRC", - "SOP", - "MDH", - "MDV", -}; - -static char* r500_fs_op_alpha[] = { - "MAD", - " DP", - "MIN", - "MAX", - "---", - "CND", - "CMP", - "FRC", - "EX2", - "LN2", - "RCP", - "RSQ", - "SIN", - "COS", - "MDH", - "MDV", -}; - -static char* r500_fs_mask[] = { - "NONE", - "R ", - " G ", - "RG ", - " B ", - "R B ", - " GB ", - "RGB ", - " A", - "R A", - " G A", - "RG A", - " BA", - "R BA", - " GBA", - "RGBA", -}; - -static char* r500_fs_tex[] = { - " NOP", - " LD", - "TEXKILL", - " PROJ", - "LODBIAS", - " LOD", - " DXDY", -}; - -static char* r300_vs_ve_ops[] = { - /* R300 vector ops */ - " VE_NO_OP", - " VE_DOT_PRODUCT", - " VE_MULTIPLY", - " VE_ADD", - " VE_MULTIPLY_ADD", - " VE_DISTANCE_FACTOR", - " VE_FRACTION", - " VE_MAXIMUM", - " VE_MINIMUM", - "VE_SET_GREATER_THAN_EQUAL", - " VE_SET_LESS_THAN", - " VE_MULTIPLYX2_ADD", - " VE_MULTIPLY_CLAMP", - " VE_FLT2FIX_DX", - " VE_FLT2FIX_DX_RND", - /* R500 vector ops */ - " VE_PRED_SET_EQ_PUSH", - " VE_PRED_SET_GT_PUSH", - " VE_PRED_SET_GTE_PUSH", - " VE_PRED_SET_NEQ_PUSH", - " VE_COND_WRITE_EQ", - " VE_COND_WRITE_GT", - " VE_COND_WRITE_GTE", - " VE_COND_WRITE_NEQ", - " VE_SET_GREATER_THAN", - " VE_SET_EQUAL", - " VE_SET_NOT_EQUAL", - " (reserved)", - " (reserved)", - " (reserved)", -}; - -static char* r300_vs_me_ops[] = { - /* R300 math ops */ - " ME_NO_OP", - " ME_EXP_BASE2_DX", - " ME_LOG_BASE2_DX", - " ME_EXP_BASEE_FF", - " ME_LIGHT_COEFF_DX", - " ME_POWER_FUNC_FF", - " ME_RECIP_DX", - " ME_RECIP_FF", - " ME_RECIP_SQRT_DX", - " ME_RECIP_SQRT_FF", - " ME_MULTIPLY", - " ME_EXP_BASE2_FULL_DX", - " ME_LOG_BASE2_FULL_DX", - " ME_POWER_FUNC_FF_CLAMP_B", - "ME_POWER_FUNC_FF_CLAMP_B1", - "ME_POWER_FUNC_FF_CLAMP_01", - " ME_SIN", - " ME_COS", - /* R500 math ops */ - " ME_LOG_BASE2_IEEE", - " ME_RECIP_IEEE", - " ME_RECIP_SQRT_IEEE", - " ME_PRED_SET_EQ", - " ME_PRED_SET_GT", - " ME_PRED_SET_GTE", - " ME_PRED_SET_NEQ", - " ME_PRED_SET_CLR", - " ME_PRED_SET_INV", - " ME_PRED_SET_POP", - " ME_PRED_SET_RESTORE", - " (reserved)", - " (reserved)", - " (reserved)", -}; - -/* XXX refactor to avoid clashing symbols */ -static char* r300_vs_src_debug[] = { - "t", - "i", - "c", - "a", -}; - -static char* r300_vs_dst_debug[] = { - "t", - "a0", - "o", - "ox", - "a", - "i", - "u", - "u", -}; - -static char* r300_vs_swiz_debug[] = { - "X", - "Y", - "Z", - "W", - "0", - "1", - "U", - "U", -}; - -void r500_fs_dump(struct r500_fragment_shader* fs); - -void r300_vs_dump(struct r300_vertex_shader* vs); - -#endif /* R300_DEBUG_H */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 93cf6909a33..bd4d59e6f1a 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -24,6 +24,9 @@ #include "r300_emit.h" +#include "r300_fs.h" +#include "r300_vs.h" + void r300_emit_blend_state(struct r300_context* r300, struct r300_blend_state* blend) { @@ -109,73 +112,158 @@ void r300_emit_dsa_state(struct r300_context* r300, END_CS; } -void r300_emit_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs) +static const float * get_shader_constant( + struct r300_context * r300, + struct rc_constant * constant, + struct r300_constant_buffer * externals) +{ + static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 }; + switch(constant->Type) { + case RC_CONSTANT_EXTERNAL: + return externals->constants[constant->u.External]; + + case RC_CONSTANT_IMMEDIATE: + return constant->u.Immediate; + + default: + debug_printf("r300: Implementation error: Unhandled constant type %i\n", + constant->Type); + return zero; + } +} + +/* Convert a normal single-precision float into the 7.16 format + * used by the R300 fragment shader. + */ +static uint32_t pack_float24(float f) +{ + union { + float fl; + uint32_t u; + } u; + float mantissa; + int exponent; + uint32_t float24 = 0; + + if (f == 0.0) + return 0; + + u.fl = f; + + mantissa = frexpf(f, &exponent); + + /* Handle -ve */ + if (mantissa < 0) { + float24 |= (1 << 23); + mantissa = mantissa * -1.0; + } + /* Handle exponent, bias of 63 */ + exponent += 62; + float24 |= (exponent << 16); + /* Kill 7 LSB of mantissa */ + float24 |= (u.u & 0x7FFFFF) >> 7; + + return float24; +} + +void r300_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals) { + struct r300_fragment_program_code * code = &generic_code->code.r300; + struct rc_constant_list * constants = &generic_code->constants; int i; CS_LOCALS(r300); - BEGIN_CS(22); - - OUT_CS_REG(R300_US_CONFIG, fs->indirections); - OUT_CS_REG(R300_US_PIXSIZE, fs->shader.stack_size); - /* XXX figure out exactly how big the sizes are on this reg */ - OUT_CS_REG(R300_US_CODE_OFFSET, 0x40); - /* XXX figure these ones out a bit better kthnx */ - OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0); - OUT_CS_REG(R300_US_CODE_ADDR_3, 0x40 | R300_RGBA_OUT); - - for (i = 0; i < fs->alu_instruction_count; i++) { - OUT_CS_REG(R300_US_ALU_RGB_INST_0 + (4 * i), - fs->instructions[i].alu_rgb_inst); - OUT_CS_REG(R300_US_ALU_RGB_ADDR_0 + (4 * i), - fs->instructions[i].alu_rgb_addr); - OUT_CS_REG(R300_US_ALU_ALPHA_INST_0 + (4 * i), - fs->instructions[i].alu_alpha_inst); - OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0 + (4 * i), - fs->instructions[i].alu_alpha_addr); + BEGIN_CS(15 + + code->alu.length * 4 + + (code->tex.length ? (1 + code->tex.length) : 0) + + (constants->Count ? (1 + constants->Count * 4) : 0)); + + OUT_CS_REG(R300_US_CONFIG, code->config); + OUT_CS_REG(R300_US_PIXSIZE, code->pixsize); + OUT_CS_REG(R300_US_CODE_OFFSET, code->code_offset); + + OUT_CS_REG_SEQ(R300_US_CODE_ADDR_0, 4); + for(i = 0; i < 4; ++i) + OUT_CS(code->code_addr[i]); + + OUT_CS_REG_SEQ(R300_US_ALU_RGB_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].rgb_inst); + + OUT_CS_REG_SEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].rgb_addr); + + OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].alpha_inst); + + OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) + OUT_CS(code->alu.inst[i].alpha_addr); + + if (code->tex.length) { + OUT_CS_REG_SEQ(R300_US_TEX_INST_0, code->tex.length); + for(i = 0; i < code->tex.length; ++i) + OUT_CS(code->tex.inst[i]); + } + + if (constants->Count) { + OUT_CS_ONE_REG(R300_PFS_PARAM_0_X, constants->Count * 4); + for(i = 0; i < constants->Count; ++i) { + const float * data = get_shader_constant(r300, &constants->Constants[i], externals); + OUT_CS(pack_float24(data[0])); + OUT_CS(pack_float24(data[1])); + OUT_CS(pack_float24(data[2])); + OUT_CS(pack_float24(data[3])); + } } END_CS; } -void r500_emit_fragment_shader(struct r300_context* r300, - struct r500_fragment_shader* fs) +void r500_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals) { + struct r500_fragment_program_code * code = &generic_code->code.r500; + struct rc_constant_list * constants = &generic_code->constants; int i; - struct r300_constant_buffer* constants = - &r300->shader_constants[PIPE_SHADER_FRAGMENT]; CS_LOCALS(r300); - BEGIN_CS(9 + (fs->instruction_count * 6) + (constants->count ? 3 : 0) + - (constants->count * 4)); - OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_CS_REG(R500_US_PIXSIZE, fs->shader.stack_size); - OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) | - R500_US_CODE_END_ADDR(fs->instruction_count)); + BEGIN_CS(13 + + ((code->inst_end + 1) * 6) + + (constants->Count ? (3 + (constants->Count * 4)) : 0)); + OUT_CS_REG(R500_US_CONFIG, 0); + OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx); + OUT_CS_REG(R500_US_CODE_RANGE, + R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end)); + OUT_CS_REG(R500_US_CODE_OFFSET, 0); + OUT_CS_REG(R500_US_CODE_ADDR, + R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(code->inst_end)); OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR); - OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, fs->instruction_count * 6); - for (i = 0; i < fs->instruction_count; i++) { - OUT_CS(fs->instructions[i].inst0); - OUT_CS(fs->instructions[i].inst1); - OUT_CS(fs->instructions[i].inst2); - OUT_CS(fs->instructions[i].inst3); - OUT_CS(fs->instructions[i].inst4); - OUT_CS(fs->instructions[i].inst5); - } - - if (constants->count) { - OUT_CS_REG(R500_GA_US_VECTOR_INDEX, - R500_GA_US_VECTOR_INDEX_TYPE_CONST); - OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->count * 4); - for (i = 0; i < constants->count; i++) { - OUT_CS_32F(constants->constants[i][0]); - OUT_CS_32F(constants->constants[i][1]); - OUT_CS_32F(constants->constants[i][2]); - OUT_CS_32F(constants->constants[i][3]); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, (code->inst_end + 1) * 6); + for (i = 0; i <= code->inst_end; i++) { + OUT_CS(code->inst[i].inst0); + OUT_CS(code->inst[i].inst1); + OUT_CS(code->inst[i].inst2); + OUT_CS(code->inst[i].inst3); + OUT_CS(code->inst[i].inst4); + OUT_CS(code->inst[i].inst5); + } + + if (constants->Count) { + OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4); + for (i = 0; i < constants->Count; i++) { + const float * data = get_shader_constant(r300, &constants->Constants[i], externals); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); } } @@ -190,7 +278,7 @@ void r300_emit_fb_state(struct r300_context* r300, int i; CS_LOCALS(r300); - BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4); + BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4); for (i = 0; i < fb->nr_cbufs; i++) { tex = (struct r300_texture*)fb->cbufs[i]->texture; assert(tex && tex->buffer && "cbuf is marked, but NULL!"); @@ -199,8 +287,10 @@ void r300_emit_fb_state(struct r300_context* r300, OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), pixpitch | - r300_translate_colorformat(tex->tex.format)); + OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1); + OUT_CS_RELOC(tex->buffer, pixpitch | + r300_translate_colorformat(tex->tex.format), 0, + RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), r300_translate_out_fmt(fb->cbufs[i]->format)); @@ -216,7 +306,8 @@ void r300_emit_fb_state(struct r300_context* r300, OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format)); - OUT_CS_REG(R300_ZB_DEPTHPITCH, pixpitch); + OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1); + OUT_CS_RELOC(tex->buffer, pixpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); } OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, @@ -228,6 +319,79 @@ void r300_emit_fb_state(struct r300_context* r300, END_CS; } +void r300_emit_query_begin(struct r300_context* r300, + struct r300_query* query) +{ + CS_LOCALS(r300); + + /* XXX This will almost certainly not return good results + * for overlapping queries. */ + BEGIN_CS(2); + OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); + END_CS; +} + +void r300_emit_query_end(struct r300_context* r300, + struct r300_query* query) +{ + struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; + CS_LOCALS(r300); + + if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, + 0, RADEON_GEM_DOMAIN_GTT)) { + debug_printf("r300: There wasn't room for the OQ buffer!?" + " Oh noes!\n"); + } + + assert(caps->num_frag_pipes); + BEGIN_CS(6 * caps->num_frag_pipes + 2); + /* I'm not so sure I like this switch, but it's hard to be elegant + * when there's so many special cases... + * + * So here's the basic idea. For each pipe, enable writes to it only, + * then put out the relocation for ZPASS_ADDR, taking into account a + * 4-byte offset for each pipe. RV380 and older are special; they have + * only two pipes, and the second pipe's enable is on bit 3, not bit 1, + * so there's a chipset cap for that. */ + switch (caps->num_frag_pipes) { + case 4: + /* pipe 3 only */ + OUT_CS_REG(R300_SU_REG_DEST, 1 << 3); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 3), + 0, RADEON_GEM_DOMAIN_GTT, 0); + case 3: + /* pipe 2 only */ + OUT_CS_REG(R300_SU_REG_DEST, 1 << 2); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 2), + 0, RADEON_GEM_DOMAIN_GTT, 0); + case 2: + /* pipe 1 only */ + /* As mentioned above, accomodate RV380 and older. */ + OUT_CS_REG(R300_SU_REG_DEST, + 1 << (caps->high_second_pipe ? 3 : 1)); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 1), + 0, RADEON_GEM_DOMAIN_GTT, 0); + case 1: + /* pipe 0 only */ + OUT_CS_REG(R300_SU_REG_DEST, 1 << 0); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0), + 0, RADEON_GEM_DOMAIN_GTT, 0); + default: + debug_printf("r300: Implementation error: Chipset reports %d" + " pixel pipes!\n", caps->num_frag_pipes); + assert(0); + } + + /* And, finally, reset it to normal... */ + OUT_CS_REG(R300_SU_REG_DEST, 0xF); + END_CS; + +} + void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) { CS_LOCALS(r300); @@ -289,18 +453,6 @@ void r300_emit_rs_block_state(struct r300_context* r300, END_CS; } -void r300_emit_sampler(struct r300_context* r300, - struct r300_sampler_state* sampler, unsigned offset) -{ - CS_LOCALS(r300); - - BEGIN_CS(6); - OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); - OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); - OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); - END_CS; -} - void r300_emit_scissor_state(struct r300_context* r300, struct r300_scissor_state* scissor) { @@ -314,11 +466,17 @@ void r300_emit_scissor_state(struct r300_context* r300, } void r300_emit_texture(struct r300_context* r300, - struct r300_texture* tex, unsigned offset) + struct r300_sampler_state* sampler, + struct r300_texture* tex, + unsigned offset) { CS_LOCALS(r300); - BEGIN_CS(10); + BEGIN_CS(16); + OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); + OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); + OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); + OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0); OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1); OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2); @@ -386,13 +544,13 @@ void r300_emit_vertex_format_state(struct r300_context* r300) END_CS; } -void r300_emit_vertex_shader(struct r300_context* r300, - struct r300_vertex_shader* vs) +void r300_emit_vertex_program_code(struct r300_context* r300, + struct r300_vertex_program_code* code, + struct r300_constant_buffer* constants) { int i; struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct r300_constant_buffer* constants = - &r300->shader_constants[PIPE_SHADER_VERTEX]; + unsigned instruction_count = code->length / 4; CS_LOCALS(r300); if (!r300screen->caps->has_tcl) { @@ -401,10 +559,10 @@ void r300_emit_vertex_shader(struct r300_context* r300, return; } - if (constants->count) { - BEGIN_CS(14 + (vs->instruction_count * 4) + (constants->count * 4)); + if (code->constants.Count) { + BEGIN_CS(14 + code->length + (code->constants.Count * 4)); } else { - BEGIN_CS(11 + (vs->instruction_count * 4)); + BEGIN_CS(11 + code->length); } /* R300_VAP_PVS_CODE_CNTL_0 @@ -414,30 +572,27 @@ void r300_emit_vertex_shader(struct r300_context* r300, * XXX these could be optimized to select better values... */ OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3); OUT_CS(R300_PVS_FIRST_INST(0) | - R300_PVS_XYZW_VALID_INST(vs->instruction_count - 1) | - R300_PVS_LAST_INST(vs->instruction_count - 1)); - OUT_CS(R300_PVS_MAX_CONST_ADDR(constants->count - 1)); - OUT_CS(vs->instruction_count - 1); + R300_PVS_XYZW_VALID_INST(instruction_count - 1) | + R300_PVS_LAST_INST(instruction_count - 1)); + OUT_CS(R300_PVS_MAX_CONST_ADDR(code->constants.Count - 1)); + OUT_CS(instruction_count - 1); OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0); - OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4); - for (i = 0; i < vs->instruction_count; i++) { - OUT_CS(vs->instructions[i].inst0); - OUT_CS(vs->instructions[i].inst1); - OUT_CS(vs->instructions[i].inst2); - OUT_CS(vs->instructions[i].inst3); - } + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->length); + for (i = 0; i < code->length; i++) + OUT_CS(code->body.d[i]); - if (constants->count) { + if (code->constants.Count) { OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, (r300screen->caps->is_r500 ? R500_PVS_CONST_START : R300_PVS_CONST_START)); - OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->count * 4); - for (i = 0; i < constants->count; i++) { - OUT_CS_32F(constants->constants[i][0]); - OUT_CS_32F(constants->constants[i][1]); - OUT_CS_32F(constants->constants[i][2]); - OUT_CS_32F(constants->constants[i][3]); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->constants.Count * 4); + for (i = 0; i < code->constants.Count; i++) { + const float * data = get_shader_constant(r300, &code->constants.Constants[i], constants); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); } } @@ -449,6 +604,12 @@ void r300_emit_vertex_shader(struct r300_context* r300, END_CS; } +void r300_emit_vertex_shader(struct r300_context* r300, + struct r300_vertex_shader* vs) +{ + r300_emit_vertex_program_code(r300, &vs->code, &r300->shader_constants[PIPE_SHADER_VERTEX]); +} + void r300_emit_viewport_state(struct r300_context* r300, struct r300_viewport_state* viewport) { @@ -527,6 +688,12 @@ validate: goto validate; } } + /* ...occlusion query buffer... */ + if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, + 0, RADEON_GEM_DOMAIN_GTT)) { + r300->context.flush(&r300->context, 0, NULL); + goto validate; + } /* ...and vertex buffer. */ if (r300->vbo) { if (!r300->winsys->add_buffer(r300->winsys, r300->vbo, @@ -537,10 +704,11 @@ validate: } else { debug_printf("No VBO while emitting dirty state!\n"); } - if (r300->winsys->validate(r300->winsys)) { + if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); if (invalid) { /* Well, hell. */ + debug_printf("r300: Stuck in validation loop, gonna quit now."); exit(1); } invalid = TRUE; @@ -569,11 +737,9 @@ validate: if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { if (r300screen->caps->is_r500) { - r500_emit_fragment_shader(r300, - (struct r500_fragment_shader*)r300->fs); + r500_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); } else { - r300_emit_fragment_shader(r300, - (struct r300_fragment_shader*)r300->fs); + r300_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; } @@ -593,29 +759,27 @@ validate: r300->dirty_state &= ~R300_NEW_RS_BLOCK; } - if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) { - for (i = 0; i < r300->sampler_count; i++) { - if (r300->dirty_state & (R300_NEW_SAMPLER << i)) { - r300_emit_sampler(r300, r300->sampler_states[i], i); - r300->dirty_state &= ~(R300_NEW_SAMPLER << i); - dirty_tex++; - } - } - } - if (r300->dirty_state & R300_NEW_SCISSOR) { r300_emit_scissor_state(r300, r300->scissor_state); r300->dirty_state &= ~R300_NEW_SCISSOR; } - if (r300->dirty_state & R300_ANY_NEW_TEXTURES) { - for (i = 0; i < r300->texture_count; i++) { - if (r300->dirty_state & (R300_NEW_TEXTURE << i)) { - r300_emit_texture(r300, r300->textures[i], i); - r300->dirty_state &= ~(R300_NEW_TEXTURE << i); + /* Samplers and textures are tracked separately but emitted together. */ + if (r300->dirty_state & + (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) { + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { + if (r300->dirty_state & + ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { + r300_emit_texture(r300, + r300->sampler_states[i], + r300->textures[i], + i); + r300->dirty_state &= + ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i)); dirty_tex++; } } + r300->dirty_state &= ~(R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES); } if (r300->dirty_state & R300_NEW_VIEWPORT) { @@ -637,6 +801,10 @@ validate: r300->dirty_state &= ~R300_NEW_VERTEX_SHADER; } + /* XXX + assert(r300->dirty_state == 0); + */ + /* Finally, emit the VBO. */ r300_emit_vertex_buffer(r300); diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 946f625bd89..350691d592d 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -30,6 +30,9 @@ #include "r300_screen.h" #include "r300_state_inlines.h" +struct rX00_fragment_program_code; +struct r300_vertex_program_code; + void r300_emit_blend_state(struct r300_context* r300, struct r300_blend_state* blend); @@ -42,11 +45,13 @@ void r300_emit_clip_state(struct r300_context* r300, void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); -void r300_emit_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs); +void r300_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals); -void r500_emit_fragment_shader(struct r300_context* r300, - struct r500_fragment_shader* fs); +void r500_emit_fragment_program_code(struct r300_context* r300, + struct rX00_fragment_program_code* generic_code, + struct r300_constant_buffer* externals); void r300_emit_fb_state(struct r300_context* r300, struct pipe_framebuffer_state* fb); @@ -56,19 +61,22 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs); void r300_emit_rs_block_state(struct r300_context* r300, struct r300_rs_block* rs); -void r300_emit_sampler(struct r300_context* r300, - struct r300_sampler_state* sampler, unsigned offset); - void r300_emit_scissor_state(struct r300_context* r300, struct r300_scissor_state* scissor); void r300_emit_texture(struct r300_context* r300, - struct r300_texture* tex, unsigned offset); + struct r300_sampler_state* sampler, + struct r300_texture* tex, + unsigned offset); void r300_emit_vertex_buffer(struct r300_context* r300); void r300_emit_vertex_format_state(struct r300_context* r300); +void r300_emit_vertex_program_code(struct r300_context* r300, + struct r300_vertex_program_code* code, + struct r300_constant_buffer* constants); + void r300_emit_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs); diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index 89a5f2b20cf..0dff1c6f4fb 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -29,7 +29,11 @@ static void r300_flush(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); CS_LOCALS(r300); - draw_flush(r300->draw); + /* We probably need to flush Draw, but we may have been called from + * within Draw. This feels kludgy, but it might be the best thing. */ + if (!r300->draw->flushing) { + draw_flush(r300->draw); + } if (r300->dirty_hw) { FLUSH_CS; diff --git a/src/gallium/drivers/r300/r300_flush.h b/src/gallium/drivers/r300/r300_flush.h index a1b224b39ce..9a83d89daab 100644 --- a/src/gallium/drivers/r300/r300_flush.h +++ b/src/gallium/drivers/r300/r300_flush.h @@ -23,6 +23,8 @@ #ifndef R300_FLUSH_H #define R300_FLUSH_H +#include "draw/draw_private.h" + #include "pipe/p_context.h" #include "r300_context.h" diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c new file mode 100644 index 00000000000..36463b9a2eb --- /dev/null +++ b/src/gallium/drivers/r300/r300_fs.c @@ -0,0 +1,137 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_fs.h" + +#include "r300_tgsi_to_rc.h" + +#include "radeon_compiler.h" + +static void find_output_registers(struct r300_fragment_program_compiler * compiler, + struct r300_fragment_shader * fs) +{ + unsigned i; + + /* Mark the outputs as not present initially */ + compiler->OutputColor = fs->info.num_outputs; + compiler->OutputDepth = fs->info.num_outputs; + + /* Now see where they really are. */ + for(i = 0; i < fs->info.num_outputs; ++i) { + switch(fs->info.output_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: + compiler->OutputColor = i; + break; + case TGSI_SEMANTIC_POSITION: + compiler->OutputDepth = i; + break; + } + } +} + +static void allocate_hardware_inputs( + struct r300_fragment_program_compiler * c, + void (*allocate)(void * data, unsigned input, unsigned hwreg), + void * mydata) +{ + struct tgsi_shader_info* info = &((struct r300_fragment_shader*)c->UserData)->info; + int total_colors = 0; + int colors = 0; + int total_generic = 0; + int generic = 0; + int i; + + for (i = 0; i < info->num_inputs; i++) { + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: + total_colors++; + break; + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + total_generic++; + break; + } + } + + for(i = 0; i < info->num_inputs; i++) { + switch (info->input_semantic_name[i]) { + case TGSI_SEMANTIC_COLOR: + allocate(mydata, i, colors); + colors++; + break; + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + allocate(mydata, i, total_colors + generic); + generic++; + break; + } + } +} + +void r300_translate_fragment_shader(struct r300_context* r300, + struct r300_fragment_shader* fs) +{ + struct r300_fragment_program_compiler compiler; + struct tgsi_to_rc ttr; + + memset(&compiler, 0, sizeof(compiler)); + rc_init(&compiler.Base); + compiler.Base.Debug = 1; + + compiler.code = &fs->code; + compiler.is_r500 = r300_screen(r300->context.screen)->caps->is_r500; + compiler.AllocateHwInputs = &allocate_hardware_inputs; + compiler.UserData = fs; + + /* TODO: Program compilation depends on texture compare modes, + * which are sampler state. Therefore, programs need to be recompiled + * depending on this state as in the classic Mesa driver. + * + * This is not yet handled correctly. + */ + + find_output_registers(&compiler, fs); + + if (compiler.Base.Debug) { + debug_printf("r300: Initial fragment program\n"); + tgsi_dump(fs->state.tokens, 0); + } + + /* Translate TGSI to our internal representation */ + ttr.compiler = &compiler.Base; + ttr.info = &fs->info; + + r300_tgsi_to_rc(&ttr, fs->state.tokens); + + /* Invoke the compiler */ + r3xx_compile_fragment_program(&compiler); + if (compiler.Base.Error) { + /* Todo: Fail gracefully */ + fprintf(stderr, "r300 FP: Compiler error\n"); + abort(); + } + + /* And, finally... */ + rc_destroy(&compiler.Base); + fs->translated = TRUE; +} diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h new file mode 100644 index 00000000000..9fab7894024 --- /dev/null +++ b/src/gallium/drivers/r300/r300_fs.h @@ -0,0 +1,51 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_FS_H +#define R300_FS_H + +#include "tgsi/tgsi_dump.h" + +#include "r300_context.h" +#include "r3xx_fs.h" +#include "r5xx_fs.h" + +#include "radeon_code.h" + +struct r300_fragment_shader { + /* Parent class */ + struct pipe_shader_state state; + struct tgsi_shader_info info; + + /* Has this shader been translated yet? */ + boolean translated; + + /* Compiled code */ + struct rX00_fragment_program_code code; +}; + + +void r300_translate_fragment_shader(struct r300_context* r300, + struct r300_fragment_shader* fs); + + #endif /* R300_FS_H */ diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c index 8fc61c2dec7..1d5185b417e 100644 --- a/src/gallium/drivers/r300/r300_query.c +++ b/src/gallium/drivers/r300/r300_query.c @@ -25,14 +25,30 @@ static struct pipe_query* r300_create_query(struct pipe_context* pipe, unsigned query_type) { - struct r300_query* q = CALLOC_STRUCT(r300_query); + struct r300_context* r300 = r300_context(pipe); + struct r300_screen* r300screen = r300_screen(r300->context.screen); + unsigned query_size = r300screen->caps->num_frag_pipes * 4; + struct r300_query* q, * qptr; + + q = CALLOC_STRUCT(r300_query); q->type = query_type; assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER); - /* XXX this is to force winsys to give us a GTT buffer */ - q->buf = pipe->screen->buffer_create(pipe->screen, 64, - PIPE_BUFFER_USAGE_VERTEX, 64); + q->active = FALSE; + + if (!r300->query_list) { + r300->query_list = q; + } else if (!is_empty_list(r300->query_list)) { + qptr = last_elem(r300->query_list); + q->offset = qptr->offset + query_size; + insert_at_tail(r300->query_list, q); + } + + /* XXX */ + if (q->offset >= 4096) { + q->offset = 0; + } return (struct pipe_query*)q; } @@ -40,6 +56,9 @@ static struct pipe_query* r300_create_query(struct pipe_context* pipe, static void r300_destroy_query(struct pipe_context* pipe, struct pipe_query* query) { + struct r300_query* q = (struct r300_query*)query; + + remove_from_list(q); FREE(query); } @@ -49,15 +68,15 @@ static void r300_begin_query(struct pipe_context* pipe, uint32_t* map; struct r300_context* r300 = r300_context(pipe); struct r300_query* q = (struct r300_query*)query; - CS_LOCALS(r300); - map = pipe_buffer_map(pipe->screen, q->buf, PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + map += q->offset / 4; *map = ~0; - pipe_buffer_unmap(pipe->screen, q->buf); + pipe->screen->buffer_unmap(pipe->screen, r300->oqbo); - BEGIN_CS(2); - OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); - END_CS; + r300_emit_dirty_state(r300); + r300_emit_query_begin(r300, q); } static void r300_end_query(struct pipe_context* pipe, @@ -65,12 +84,9 @@ static void r300_end_query(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); struct r300_query* q = (struct r300_query*)query; - CS_LOCALS(r300); - BEGIN_CS(4); - OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); - OUT_CS_RELOC(q->buf, 0, 0, RADEON_GEM_DOMAIN_GTT, 0); - END_CS; + r300_emit_dirty_state(r300); + r300_emit_query_end(r300, q); } static boolean r300_get_query_result(struct pipe_context* pipe, @@ -78,22 +94,38 @@ static boolean r300_get_query_result(struct pipe_context* pipe, boolean wait, uint64_t* result) { + struct r300_context* r300 = r300_context(pipe); + struct r300_screen* r300screen = r300_screen(r300->context.screen); struct r300_query* q = (struct r300_query*)query; + unsigned flags = PIPE_BUFFER_USAGE_CPU_READ; uint32_t* map; uint32_t temp; + unsigned i; if (wait) { - /* Well, we're expected to just sit here and spin, so let's go ahead - * and flush so we can be sure that the card's spinning... */ - /* XXX double-check these params */ pipe->flush(pipe, 0, NULL); + } else { + flags |= PIPE_BUFFER_USAGE_DONTBLOCK; } - map = pipe_buffer_map(pipe->screen, q->buf, PIPE_BUFFER_USAGE_CPU_READ); - temp = *map; - pipe_buffer_unmap(pipe->screen, q->buf); + map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, flags); + map += q->offset / 4; + for (i = 0; i < r300screen->caps->num_frag_pipes; i++) { + if (*map == ~0) { + /* Looks like our results aren't ready yet. */ + if (wait) { + debug_printf("r300: Despite waiting, OQ results haven't" + " come in yet.\n"); + } + temp = ~0; + break; + } + temp += *map; + map++; + } + pipe->screen->buffer_unmap(pipe->screen, r300->oqbo); - if (temp < 0) { + if (temp == ~0) { /* Our results haven't been written yet... */ return FALSE; } diff --git a/src/gallium/drivers/r300/r300_query.h b/src/gallium/drivers/r300/r300_query.h index 4f447ea45b8..4f50e8f8440 100644 --- a/src/gallium/drivers/r300/r300_query.h +++ b/src/gallium/drivers/r300/r300_query.h @@ -27,12 +27,7 @@ #include "r300_cs.h" #include "r300_reg.h" -struct r300_query { - /* The kind of query. Currently only OQ is supported. */ - unsigned type; - /* Buffer object where we want our results to reside. */ - struct pipe_buffer* buf; -}; +struct r300_context; static INLINE struct r300_query* r300_query(struct pipe_query* q) { diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 3bb9bc47b50..03cd219cde9 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -1064,8 +1064,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | \ R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \ R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | \ - R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT | \ - R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) + R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT ) # define R300_SHADE_MODEL_SMOOTH ( \ R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | \ @@ -1075,8 +1074,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | \ R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \ R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \ - R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST ) + R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD ) /* Specifies red & green components of fill color -- S312 format -- Backwards comp. */ #define R300_GA_SOLID_RG 0x427c @@ -1480,6 +1478,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_PITCH_EN (1 << 31) # define R300_TX_WIDTH(x) ((x) << 0) # define R300_TX_HEIGHT(x) ((x) << 11) +# define R300_TX_NUM_LEVELS(x) ((x) << 26) #define R300_TX_FORMAT1_0 0x44C0 /* The interpretation of the format word by Wladimir van der Laan */ @@ -1487,9 +1486,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. They are given meanings as R, G, B and Alpha by the swizzle specification */ # define R300_TX_FORMAT_X8 0x0 -# define R500_TX_FORMAT_X1 0x0 // bit set in format 2 # define R300_TX_FORMAT_X16 0x1 -# define R500_TX_FORMAT_X1_REV 0x0 // bit set in format 2 # define R300_TX_FORMAT_Y4X4 0x2 # define R300_TX_FORMAT_Y8X8 0x3 # define R300_TX_FORMAT_Y16X16 0x4 @@ -1506,31 +1503,29 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_FORMAT_DXT1 0xF # define R300_TX_FORMAT_DXT3 0x10 # define R300_TX_FORMAT_DXT5 0x11 -# define R300_TX_FORMAT_D3DMFT_CxV8U8 0x12 /* no swizzle */ -# define R300_TX_FORMAT_A8R8G8B8 0x13 /* no swizzle */ -# define R300_TX_FORMAT_B8G8_B8G8 0x14 /* no swizzle */ -# define R300_TX_FORMAT_G8R8_G8B8 0x15 /* no swizzle */ - - /* These two values are wrong, but they're the only values that - * produce any even vaguely correct results. Can r300 only do 16-bit - * depth textures? - */ -# define R300_TX_FORMAT_X24_Y8 0x1e -# define R300_TX_FORMAT_X32 0x1e - - /* 0x16 - some 16 bit green format.. ?? */ +# define R300_TX_FORMAT_Y8 0x12 +# define R300_TX_FORMAT_AVYU444 0x13 +# define R300_TX_FORMAT_VYUY422 0x14 +# define R300_TX_FORMAT_YVYU422 0x15 +# define R300_TX_FORMAT_16_MPEG 0x16 +# define R300_TX_FORMAT_16_16_MPEG 0x17 +# define R300_TX_FORMAT_16F 0x18 +# define R300_TX_FORMAT_16F_16F 0x19 +# define R300_TX_FORMAT_16F_16F_16F_16F 0x1A +# define R300_TX_FORMAT_32F 0x1B +# define R300_TX_FORMAT_32F_32F 0x1C +# define R300_TX_FORMAT_32F_32F_32F_32F 0x1D +# define R300_TX_FORMAT_W24_FP 0x1E + +# define R300_TX_FORMAT_SIGNED_W (1 << 5) +# define R300_TX_FORMAT_SIGNED_Z (1 << 6) +# define R300_TX_FORMAT_SIGNED_Y (1 << 7) +# define R300_TX_FORMAT_SIGNED_X (1 << 8) +# define R300_TX_FORMAT_SIGNED (0xf << 5) + # define R300_TX_FORMAT_3D (1 << 25) # define R300_TX_FORMAT_CUBIC_MAP (2 << 25) - /* gap */ - /* Floating point formats */ - /* Note - hardware supports both 16 and 32 bit floating point */ -# define R300_TX_FORMAT_FL_I16 0x18 -# define R300_TX_FORMAT_FL_I16A16 0x19 -# define R300_TX_FORMAT_FL_R16G16B16A16 0x1A -# define R300_TX_FORMAT_FL_I32 0x1B -# define R300_TX_FORMAT_FL_I32A32 0x1C -# define R300_TX_FORMAT_FL_R32G32B32A32 0x1D /* alpha modes, convenience mostly */ /* if you have alpha, pick constant appropriate to the number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */ @@ -1571,7 +1566,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_FORMAT_CONST_Z (4<<5) # define R300_TX_FORMAT_CONST_W (8<<5) -# define R300_TX_FORMAT_YUV_MODE 0x00800000 +# define R300_TX_FORMAT_GAMMA (1 << 21) +# define R300_TX_FORMAT_YUV_TO_RGB (1 << 22) #define R300_TX_FORMAT2_0 0x4500 /* obvious missing in gap */ # define R300_TX_PITCHMASK_SHIFT 0 @@ -3316,6 +3312,10 @@ enum { #define R200_3D_DRAW_IMMD_2 0xC0003500 +/* XXX Oh look, stuff not brought over from docs yet */ + +#define R300_SU_REG_DEST 0x42C8 + #endif /* _R300_REG_H */ /* *INDENT-ON* */ diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index a6f1efe3563..15740f61252 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -94,8 +94,7 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) return 0; } case PIPE_CAP_S3TC: - /* IN THEORY */ - return 0; + return 1; case PIPE_CAP_ANISOTROPIC_FILTER: return 1; case PIPE_CAP_POINT_SPRITE: @@ -145,6 +144,11 @@ static int r300_get_param(struct pipe_screen* pscreen, int param) case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: /* XXX guessing (what a terrible guess) */ return 2; + case PIPE_CAP_TGSI_CONT_SUPPORTED: + /* XXX */ + return 0; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 1; default: debug_printf("r300: Implementation error: Bad param %d\n", param); @@ -179,21 +183,58 @@ static float r300_get_paramf(struct pipe_screen* pscreen, int param) } } -static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500) +static boolean check_tex_2d_format(enum pipe_format format, uint32_t usage, + boolean is_r500) { switch (format) { + /* Supported formats. */ /* Colorbuffer */ case PIPE_FORMAT_A4R4G4B4_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: case PIPE_FORMAT_A1R5G5B5_UNORM: - case PIPE_FORMAT_A8R8G8B8_UNORM: + return usage & + (PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_PRIMARY); + + /* Texture */ + case PIPE_FORMAT_A8R8G8B8_SRGB: + case PIPE_FORMAT_R8G8B8A8_SRGB: + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + case PIPE_FORMAT_YCBCR: + return usage & PIPE_TEXTURE_USAGE_SAMPLER; + /* Colorbuffer or texture */ + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: case PIPE_FORMAT_I8_UNORM: + return usage & + (PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_PRIMARY | + PIPE_TEXTURE_USAGE_SAMPLER); + /* Z buffer */ case PIPE_FORMAT_Z16_UNORM: - /* Z buffer with stencil */ + return usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL; + + /* Z buffer with stencil or texture */ case PIPE_FORMAT_Z24S8_UNORM: - return TRUE; + return usage & + (PIPE_TEXTURE_USAGE_DEPTH_STENCIL | + PIPE_TEXTURE_USAGE_SAMPLER); + + /* Definitely unsupported formats. */ + /* Non-usable Z buffer/stencil formats. */ + case PIPE_FORMAT_Z24X8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + debug_printf("r300: Note: Got unsupported format: %s in %s\n", + pf_name(format), __FUNCTION__); + return FALSE; /* XXX These don't even exist case PIPE_FORMAT_A32R32G32B32: @@ -216,7 +257,8 @@ static boolean check_tex_2d_format(enum pipe_format format, boolean is_r500) return FALSE; */ default: - debug_printf("r300: Warning: Got unsupported format: %s in %s\n", + /* Unknown format... */ + debug_printf("r300: Warning: Got unknown format: %s in %s\n", pf_name(format), __FUNCTION__); break; } @@ -233,7 +275,7 @@ static boolean r300_is_format_supported(struct pipe_screen* pscreen, { switch (target) { case PIPE_TEXTURE_2D: - return check_tex_2d_format(format, + return check_tex_2d_format(format, tex_usage, r300_screen(pscreen)->caps->is_r500); case PIPE_TEXTURE_1D: case PIPE_TEXTURE_3D: @@ -280,13 +322,14 @@ r300_get_tex_transfer(struct pipe_screen *screen, trans = CALLOC_STRUCT(r300_transfer); if (trans) { pipe_texture_reference(&trans->transfer.texture, texture); - trans->transfer.format = trans->transfer.format; + trans->transfer.format = texture->format; trans->transfer.width = w; trans->transfer.height = h; trans->transfer.block = texture->block; trans->transfer.nblocksx = texture->nblocksx[level]; trans->transfer.nblocksy = texture->nblocksy[level]; - trans->transfer.stride = tex->stride; + trans->transfer.stride = align(pf_get_stride(&trans->transfer.block, + texture->width[level]), 32); trans->transfer.usage = usage; trans->offset = offset; } @@ -349,6 +392,7 @@ struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys) return NULL; caps->pci_id = r300_winsys->pci_id; + caps->num_frag_pipes = r300_winsys->gb_pipes; r300_parse_chipset(caps); diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h index 3f52dbc3bea..2a0e41fbc3b 100644 --- a/src/gallium/drivers/r300/r300_screen.h +++ b/src/gallium/drivers/r300/r300_screen.h @@ -49,7 +49,7 @@ struct r300_transfer { }; /* Convenience cast wrapper. */ -static struct r300_screen* r300_screen(struct pipe_screen* screen) { +static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) { return (struct r300_screen*)screen; } diff --git a/src/gallium/drivers/r300/r300_shader_inlines.h b/src/gallium/drivers/r300/r300_shader_inlines.h new file mode 100644 index 00000000000..a04f45b03e2 --- /dev/null +++ b/src/gallium/drivers/r300/r300_shader_inlines.h @@ -0,0 +1,47 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_SHADER_INLINES_H +#define R300_SHADER_INLINES_H + +/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're + * not using enough of it. */ +static const struct tgsi_full_src_register r300_constant_zero = { + .SrcRegister.Extended = TRUE, + .SrcRegister.File = TGSI_FILE_NULL, + .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO, + .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO, +}; + +static const struct tgsi_full_src_register r300_constant_one = { + .SrcRegister.Extended = TRUE, + .SrcRegister.File = TGSI_FILE_NULL, + .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE, + .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE, +}; + +#endif /* R300_SHADER_INLINES_H */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 01e2b511534..27680a38631 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -24,12 +24,15 @@ #include "util/u_pack_color.h" #include "util/u_debug.h" + +#include "pipe/p_config.h" #include "pipe/internal/p_winsys_screen.h" #include "r300_context.h" #include "r300_reg.h" #include "r300_state_inlines.h" -#include "r300_state_shader.h" +#include "r300_fs.h" +#include "r300_vs.h" /* r300_state: Functions used to intialize state context by translating * Gallium state objects into semi-native r300 state objects. */ @@ -135,7 +138,6 @@ static void const struct pipe_constant_buffer* buffer) { struct r300_context* r300 = r300_context(pipe); - int i = r300->shader_constants[shader].user_count; /* This entire chunk of code seems ever-so-slightly baked. * It's as if I've got pipe_buffer* matryoshkas... */ @@ -146,26 +148,13 @@ static void map, buffer->buffer->size); pipe->winsys->buffer_unmap(pipe->winsys, buffer->buffer); - r300->shader_constants[shader].user_count = + r300->shader_constants[shader].count = buffer->buffer->size / (sizeof(float) * 4); } else { - r300->shader_constants[shader].user_count = 0; + r300->shader_constants[shader].count = 0; } r300->dirty_state |= R300_NEW_CONSTANTS; - - /* If the number of constants have changed, invalidate the shader. */ - if (r300->shader_constants[shader].user_count != i) { - if (shader == PIPE_SHADER_FRAGMENT && r300->fs && - r300->fs->uses_imms) { - r300->fs->translated = FALSE; - r300_translate_fragment_shader(r300, r300->fs); - } else if (shader == PIPE_SHADER_VERTEX && r300->vs && - r300->vs->uses_imms) { - r300->vs->translated = FALSE; - r300_translate_vertex_shader(r300, r300->vs); - } - } } /* Create a new depth, stencil, and alpha state based on the CSO dsa state. @@ -235,7 +224,8 @@ static void* dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f, 0, 1023); } else { - dsa->z_buffer_top = R300_ZTOP_ENABLE; + /* XXX need to fix this to be dynamically set + dsa->z_buffer_top = R300_ZTOP_ENABLE; */ } return (void*)dsa; @@ -282,16 +272,9 @@ static void static void* r300_create_fs_state(struct pipe_context* pipe, const struct pipe_shader_state* shader) { - struct r300_context* r300 = r300_context(pipe); - struct r3xx_fragment_shader* fs = NULL; + struct r300_fragment_shader* fs = NULL; - if (r300_screen(r300->context.screen)->caps->is_r500) { - fs = - (struct r3xx_fragment_shader*)CALLOC_STRUCT(r500_fragment_shader); - } else { - fs = - (struct r3xx_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader); - } + fs = (struct r300_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader); /* Copy state directly into shader. */ fs->state = *shader; @@ -306,7 +289,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe, static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) { struct r300_context* r300 = r300_context(pipe); - struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader; + struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader; if (fs == NULL) { r300->fs = NULL; @@ -315,7 +298,6 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) r300_translate_fragment_shader(r300, fs); } - fs->translated = TRUE; r300->fs = fs; r300->dirty_state |= R300_NEW_FRAGMENT_SHADER; @@ -324,7 +306,8 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) /* Delete fragment shader state. */ static void r300_delete_fs_state(struct pipe_context* pipe, void* shader) { - struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader; + struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader; + rc_constants_destroy(&fs->code.constants); FREE(fs->state.tokens); FREE(shader); } @@ -352,14 +335,19 @@ static void* r300_create_rs_state(struct pipe_context* pipe, rs->enable_vte = !state->bypass_vs_clip_and_viewport; +#ifdef PIPE_ARCH_LITTLE_ENDIAN + rs->vap_control_status = R300_VC_NO_SWAP; +#else + rs->vap_control_status = R300_VC_32BIT_SWAP; +#endif + /* If bypassing TCL, or if no TCL engine is present, turn off the HW TCL. * Else, enable HW TCL and force Draw's TCL off. */ if (state->bypass_vs_clip_and_viewport || !r300_screen(pipe->screen)->caps->has_tcl) { - rs->vap_control_status = R300_VAP_TCL_BYPASS; + rs->vap_control_status |= R300_VAP_TCL_BYPASS; } else { rs->rs.bypass_vs_clip_and_viewport = TRUE; - rs->vap_control_status = 0; } rs->point_size = pack_float_16_6x(state->point_size) | @@ -423,6 +411,10 @@ static void* r300_create_rs_state(struct pipe_context* pipe, rs->color_control = R300_SHADE_MODEL_SMOOTH; } + if (!state->flatshade_first) { + rs->color_control |= R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST; + } + return (void*)rs; } @@ -679,6 +671,7 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader) if (r300_screen(pipe->screen)->caps->has_tcl) { struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader; + rc_constants_destroy(&vs->code.constants); draw_delete_vertex_shader(r300->draw, vs->draw); FREE(vs->state.tokens); FREE(shader); diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 2477b30822b..c01e61a9b19 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -22,6 +22,9 @@ #include "r300_state_derived.h" +#include "r300_fs.h" +#include "r300_vs.h" + /* r300_state_derived: Various bits of state which are dependent upon * currently bound CSO data. */ @@ -46,59 +49,63 @@ static void r300_vs_tab_routes(struct r300_context* r300, assert(info->num_inputs <= 16); - if (r300screen->caps->has_tcl) { - /* Just copy vert attribs over as-is. */ + if (!r300screen->caps->has_tcl || !r300->rs_state->enable_vte) + { for (i = 0; i < info->num_inputs; i++) { - tab[i] = i; - } - for (i = 0; i < info->num_outputs; i++) { - switch (info->output_semantic_name[i]) { + switch (info->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: pos = TRUE; + tab[i] = 0; break; case TGSI_SEMANTIC_COLOR: + tab[i] = 2 + cols; cols++; break; case TGSI_SEMANTIC_PSIZE: psize = TRUE; + tab[i] = 15; break; case TGSI_SEMANTIC_FOG: fog = TRUE; /* Fall through */ case TGSI_SEMANTIC_GENERIC: + tab[i] = 6 + texs; texs++; break; default: - debug_printf("r300: Unknown vertex output %d\n", - info->output_semantic_name[i]); + debug_printf("r300: Unknown vertex input %d\n", + info->input_semantic_name[i]); break; } } - } else { + } + else + { + /* Just copy vert attribs over as-is. */ for (i = 0; i < info->num_inputs; i++) { - switch (info->input_semantic_name[i]) { + tab[i] = i; + } + + for (i = 0; i < info->num_outputs; i++) { + switch (info->output_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: pos = TRUE; - tab[i] = 0; break; case TGSI_SEMANTIC_COLOR: - tab[i] = 2 + cols; cols++; break; case TGSI_SEMANTIC_PSIZE: psize = TRUE; - tab[i] = 15; break; case TGSI_SEMANTIC_FOG: fog = TRUE; /* Fall through */ case TGSI_SEMANTIC_GENERIC: - tab[i] = 6 + texs; texs++; break; default: - debug_printf("r300: Unknown vertex input %d\n", - info->input_semantic_name[i]); + debug_printf("r300: Unknown vertex output %d\n", + info->output_semantic_name[i]); break; } } diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index 22c8e199ae3..91b93fc367e 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -353,6 +353,25 @@ static INLINE uint32_t r300_translate_out_fmt(enum pipe_format format) /* Non-CSO state. (For now.) */ +static INLINE uint32_t r300_translate_gb_pipes(int pipe_count) +{ + switch (pipe_count) { + case 1: + return R300_GB_TILE_PIPE_COUNT_RV300; + break; + case 2: + return R300_GB_TILE_PIPE_COUNT_R300; + break; + case 3: + return R300_GB_TILE_PIPE_COUNT_R420_3P; + break; + case 4: + return R300_GB_TILE_PIPE_COUNT_R420; + break; + } + return 0; +} + static INLINE uint32_t translate_vertex_data_type(int type) { switch (type) { case EMIT_1F: diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index e438114010c..1e92374a4e9 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -34,14 +34,17 @@ void r300_emit_invariant_state(struct r300_context* r300) struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); - BEGIN_CS(26 + (caps->has_tcl ? 2: 0)); + BEGIN_CS(24 + (caps->has_tcl ? 2: 0)); /*** Graphics Backend (GB) ***/ /* Various GB enables */ OUT_CS_REG(R300_GB_ENABLE, 0x0); - /* Subpixel multisampling for AA */ - OUT_CS_REG(R300_GB_MSPOS0, 0x66666666); - OUT_CS_REG(R300_GB_MSPOS1, 0x6666666); + /* Subpixel multisampling for AA + * These are commented out because glisse's CS checker doesn't like them. + * I presume these will be re-enabled later. + * OUT_CS_REG(R300_GB_MSPOS0, 0x66666666); + * OUT_CS_REG(R300_GB_MSPOS1, 0x6666666); + */ /* Source of fog depth */ OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W); /* AA enable */ @@ -53,6 +56,7 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x0); OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x0); OUT_CS_REG(R300_FG_DEPTH_SRC, 0x0); + OUT_CS_REG(R300_US_W_FMT, 0x0); /*** VAP ***/ /* Max and min vertex index clamp. */ @@ -69,7 +73,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(77 + (caps->has_tcl ? 5 : 0)); + BEGIN_CS(64 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -77,17 +81,12 @@ void r300_emit_invariant_state(struct r300_context* r300) R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT); - /* XXX endian */ if (caps->has_tcl) { - OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP); OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4); OUT_CS_32F(1.0); OUT_CS_32F(1.0); OUT_CS_32F(1.0); OUT_CS_32F(1.0); - } else { - OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP | - R300_VAP_TCL_BYPASS); } /* XXX point tex stuffing */ OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1); @@ -117,8 +116,10 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); - OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); - OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF); + if (caps->is_r500) { + OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); + OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF); + } OUT_CS_REG(R300_ZB_FORMAT, 0x00000002); OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003); OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000); @@ -132,11 +133,5 @@ void r300_emit_invariant_state(struct r300_context* r300) /* XXX */ OUT_CS_REG(R300_SC_CLIP_RULE, 0xaaaa); - OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4); - OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS(R300_US_OUT_FMT_UNUSED); - OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0); END_CS; } diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r300_state_shader.c deleted file mode 100644 index cc7f6a7c4b0..00000000000 --- a/src/gallium/drivers/r300/r300_state_shader.c +++ /dev/null @@ -1,718 +0,0 @@ -/* - * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#include "r300_state_shader.h" - -static void r300_fs_declare(struct r300_fs_asm* assembler, - struct tgsi_full_declaration* decl) -{ - switch (decl->Declaration.File) { - case TGSI_FILE_INPUT: - switch (decl->Semantic.SemanticName) { - case TGSI_SEMANTIC_COLOR: - assembler->color_count++; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - assembler->tex_count++; - break; - default: - debug_printf("r300: fs: Bad semantic declaration %d\n", - decl->Semantic.SemanticName); - break; - } - break; - case TGSI_FILE_OUTPUT: - /* Depth write. Mark the position of the output so we can - * identify it later. */ - if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { - assembler->depth_output = decl->DeclarationRange.First; - } - break; - case TGSI_FILE_CONSTANT: - break; - case TGSI_FILE_TEMPORARY: - assembler->temp_count++; - break; - default: - debug_printf("r300: fs: Bad file %d\n", decl->Declaration.File); - break; - } - - assembler->temp_offset = assembler->color_count + assembler->tex_count; -} - -static INLINE unsigned r300_fs_src(struct r300_fs_asm* assembler, - struct tgsi_src_register* src) -{ - switch (src->File) { - case TGSI_FILE_NULL: - return 0; - case TGSI_FILE_INPUT: - /* XXX may be wrong */ - return src->Index; - break; - case TGSI_FILE_TEMPORARY: - return src->Index + assembler->temp_offset; - break; - case TGSI_FILE_IMMEDIATE: - return (src->Index + assembler->imm_offset) | (1 << 8); - break; - case TGSI_FILE_CONSTANT: - /* XXX magic */ - return src->Index | (1 << 8); - break; - default: - debug_printf("r300: fs: Unimplemented src %d\n", src->File); - break; - } - return 0; -} - -static INLINE unsigned r300_fs_dst(struct r300_fs_asm* assembler, - struct tgsi_dst_register* dst) -{ - switch (dst->File) { - case TGSI_FILE_NULL: - /* This happens during KIL instructions. */ - return 0; - break; - case TGSI_FILE_OUTPUT: - return 0; - break; - case TGSI_FILE_TEMPORARY: - return dst->Index + assembler->temp_offset; - break; - default: - debug_printf("r300: fs: Unimplemented dst %d\n", dst->File); - break; - } - return 0; -} - -static INLINE boolean r300_fs_is_depr(struct r300_fs_asm* assembler, - struct tgsi_dst_register* dst) -{ - return (assembler->writes_depth && - (dst->File == TGSI_FILE_OUTPUT) && - (dst->Index == assembler->depth_output)); -} - -static INLINE unsigned r500_fix_swiz(unsigned s) -{ - /* For historical reasons, the swizzle values x, y, z, w, and 0 are - * equivalent to the actual machine code, but 1 is not. Thus, we just - * adjust it a bit... */ - if (s == TGSI_EXTSWIZZLE_ONE) { - return R500_SWIZZLE_ONE; - } else { - return s; - } -} - -static uint32_t r500_rgba_swiz(struct tgsi_full_src_register* reg) -{ - if (reg->SrcRegister.Extended) { - return r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleX) | - (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleY) << 3) | - (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleZ) << 6) | - (r500_fix_swiz(reg->SrcRegisterExtSwz.ExtSwizzleW) << 9); - } else { - return reg->SrcRegister.SwizzleX | - (reg->SrcRegister.SwizzleY << 3) | - (reg->SrcRegister.SwizzleZ << 6) | - (reg->SrcRegister.SwizzleW << 9); - } -} - -static uint32_t r500_strq_swiz(struct tgsi_full_src_register* reg) -{ - return reg->SrcRegister.SwizzleX | - (reg->SrcRegister.SwizzleY << 2) | - (reg->SrcRegister.SwizzleZ << 4) | - (reg->SrcRegister.SwizzleW << 6); -} - -static INLINE uint32_t r500_rgb_swiz(struct tgsi_full_src_register* reg) -{ - /* Only the first 9 bits... */ - return (r500_rgba_swiz(reg) & 0x1ff) | - (reg->SrcRegister.Negate ? (1 << 9) : 0) | - (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0); -} - -static INLINE uint32_t r500_alpha_swiz(struct tgsi_full_src_register* reg) -{ - /* Only the last 3 bits... */ - return (r500_rgba_swiz(reg) >> 9) | - (reg->SrcRegister.Negate ? (1 << 9) : 0) | - (reg->SrcRegisterExtMod.Absolute ? (1 << 10) : 0); -} - -static INLINE uint32_t r300_rgb_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_MOV: - return R300_ALU_OUTC_CMP; - default: - return 0; - } -} - -static INLINE uint32_t r300_alpha_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_MOV: - return R300_ALU_OUTA_CMP; - default: - return 0; - } -} - -static INLINE uint32_t r500_rgba_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_COS: - case TGSI_OPCODE_EX2: - case TGSI_OPCODE_LG2: - case TGSI_OPCODE_RCP: - case TGSI_OPCODE_RSQ: - case TGSI_OPCODE_SIN: - return R500_ALU_RGBA_OP_SOP; - case TGSI_OPCODE_DDX: - return R500_ALU_RGBA_OP_MDH; - case TGSI_OPCODE_DDY: - return R500_ALU_RGBA_OP_MDV; - case TGSI_OPCODE_FRC: - return R500_ALU_RGBA_OP_FRC; - case TGSI_OPCODE_DP3: - return R500_ALU_RGBA_OP_DP3; - case TGSI_OPCODE_DP4: - case TGSI_OPCODE_DPH: - return R500_ALU_RGBA_OP_DP4; - case TGSI_OPCODE_ABS: - case TGSI_OPCODE_CMP: - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - return R500_ALU_RGBA_OP_CMP; - case TGSI_OPCODE_ADD: - case TGSI_OPCODE_MAD: - case TGSI_OPCODE_MUL: - case TGSI_OPCODE_SUB: - return R500_ALU_RGBA_OP_MAD; - default: - return 0; - } -} - -static INLINE uint32_t r500_alpha_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_COS: - return R500_ALPHA_OP_COS; - case TGSI_OPCODE_EX2: - return R500_ALPHA_OP_EX2; - case TGSI_OPCODE_LG2: - return R500_ALPHA_OP_LN2; - case TGSI_OPCODE_RCP: - return R500_ALPHA_OP_RCP; - case TGSI_OPCODE_RSQ: - return R500_ALPHA_OP_RSQ; - case TGSI_OPCODE_FRC: - return R500_ALPHA_OP_FRC; - case TGSI_OPCODE_SIN: - return R500_ALPHA_OP_SIN; - case TGSI_OPCODE_DDX: - return R500_ALPHA_OP_MDH; - case TGSI_OPCODE_DDY: - return R500_ALPHA_OP_MDV; - case TGSI_OPCODE_DP3: - case TGSI_OPCODE_DP4: - case TGSI_OPCODE_DPH: - return R500_ALPHA_OP_DP; - case TGSI_OPCODE_ABS: - case TGSI_OPCODE_CMP: - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - return R500_ALPHA_OP_CMP; - case TGSI_OPCODE_ADD: - case TGSI_OPCODE_MAD: - case TGSI_OPCODE_MUL: - case TGSI_OPCODE_SUB: - return R500_ALPHA_OP_MAD; - default: - return 0; - } -} - -static INLINE uint32_t r500_tex_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_KIL: - return R500_TEX_INST_TEXKILL; - case TGSI_OPCODE_TEX: - return R500_TEX_INST_LD; - case TGSI_OPCODE_TXB: - return R500_TEX_INST_LODBIAS; - case TGSI_OPCODE_TXP: - return R500_TEX_INST_PROJ; - default: - return 0; - } -} - -static INLINE void r300_emit_maths(struct r300_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_src_register* src, - struct tgsi_full_dst_register* dst, - unsigned op, - unsigned count) -{ - int i = fs->alu_instruction_count; - - fs->instructions[i].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | - r300_rgb_op(op); - fs->instructions[i].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | - R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ; - fs->instructions[i].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | - r300_alpha_op(op); - fs->instructions[i].alu_alpha_addr = R300_ALPHA_ADDR0(0) | - R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT; - - fs->alu_instruction_count++; -} - -/* Setup an ALU operation. */ -static INLINE void r500_emit_maths(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_src_register* src, - struct tgsi_full_dst_register* dst, - unsigned op, - unsigned count) -{ - int i = fs->instruction_count; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - fs->instructions[i].inst0 = R500_INST_TYPE_OUT; - if (r300_fs_is_depr(assembler, dst)) { - fs->instructions[i].inst4 = R500_W_OMASK; - } else { - fs->instructions[i].inst0 |= - R500_ALU_OMASK(dst->DstRegister.WriteMask); - } - } else { - fs->instructions[i].inst0 = R500_INST_TYPE_ALU | - R500_ALU_WMASK(dst->DstRegister.WriteMask); - } - - fs->instructions[i].inst0 |= R500_INST_TEX_SEM_WAIT; - - fs->instructions[i].inst4 |= - R500_ALPHA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); - fs->instructions[i].inst5 = - R500_ALU_RGBA_ADDRD(r300_fs_dst(assembler, &dst->DstRegister)); - - switch (count) { - case 3: - fs->instructions[i].inst1 = - R500_RGB_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister)); - fs->instructions[i].inst2 = - R500_ALPHA_ADDR2(r300_fs_src(assembler, &src[2].SrcRegister)); - fs->instructions[i].inst5 |= - R500_ALU_RGBA_SEL_C_SRC2 | - R500_SWIZ_RGBA_C(r500_rgb_swiz(&src[2])) | - R500_ALU_RGBA_ALPHA_SEL_C_SRC2 | - R500_SWIZ_ALPHA_C(r500_alpha_swiz(&src[2])); - case 2: - fs->instructions[i].inst1 |= - R500_RGB_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister)); - fs->instructions[i].inst2 |= - R500_ALPHA_ADDR1(r300_fs_src(assembler, &src[1].SrcRegister)); - fs->instructions[i].inst3 = - R500_ALU_RGB_SEL_B_SRC1 | - R500_SWIZ_RGB_B(r500_rgb_swiz(&src[1])); - fs->instructions[i].inst4 |= - R500_ALPHA_SEL_B_SRC1 | - R500_SWIZ_ALPHA_B(r500_alpha_swiz(&src[1])); - case 1: - case 0: - default: - fs->instructions[i].inst1 |= - R500_RGB_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister)); - fs->instructions[i].inst2 |= - R500_ALPHA_ADDR0(r300_fs_src(assembler, &src[0].SrcRegister)); - fs->instructions[i].inst3 |= - R500_ALU_RGB_SEL_A_SRC0 | - R500_SWIZ_RGB_A(r500_rgb_swiz(&src[0])); - fs->instructions[i].inst4 |= - R500_ALPHA_SEL_A_SRC0 | - R500_SWIZ_ALPHA_A(r500_alpha_swiz(&src[0])); - break; - } - - fs->instructions[i].inst4 |= r500_alpha_op(op); - fs->instructions[i].inst5 |= r500_rgba_op(op); - - fs->instruction_count++; -} - -static INLINE void r500_emit_tex(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_src_register* src, - struct tgsi_full_dst_register* dst, - uint32_t op) -{ - int i = fs->instruction_count; - - fs->instructions[i].inst0 = R500_INST_TYPE_TEX | - R500_TEX_WMASK(dst->DstRegister.WriteMask) | - R500_INST_TEX_SEM_WAIT; - fs->instructions[i].inst1 = R500_TEX_ID(0) | - R500_TEX_SEM_ACQUIRE | //R500_TEX_IGNORE_UNCOVERED | - r500_tex_op(op); - fs->instructions[i].inst2 = - R500_TEX_SRC_ADDR(r300_fs_src(assembler, &src->SrcRegister)) | - R500_SWIZ_TEX_STRQ(r500_strq_swiz(src)) | - R500_TEX_DST_ADDR(r300_fs_dst(assembler, &dst->DstRegister)) | - R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | - R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - fs->instructions[i].inst2 |= - R500_TEX_DST_ADDR(assembler->temp_count + - assembler->temp_offset); - - fs->instruction_count++; - - /* Setup and emit a MOV. */ - src[0].SrcRegister.Index = assembler->temp_count; - src[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - - src[1] = src[0]; - src[2] = r500_constant_zero; - r500_emit_maths(fs, assembler, src, dst, TGSI_OPCODE_MOV, 3); - } else { - fs->instruction_count++; - } -} - -static void r300_fs_instruction(struct r300_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_instruction* inst) -{ - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_MOV: - /* src0 -> src1 and src2 forced to zero */ - inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0]; - inst->FullSrcRegisters[2] = r500_constant_zero; - r300_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - case TGSI_OPCODE_END: - break; - default: - debug_printf("r300: fs: Bad opcode %d\n", - inst->Instruction.Opcode); - break; - } -} - -static void r500_fs_instruction(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler, - struct tgsi_full_instruction* inst) -{ - /* Switch between opcodes. When possible, prefer using the official - * AMD/ATI names for opcodes, please, as it facilitates using the - * documentation. */ - switch (inst->Instruction.Opcode) { - /* XXX trig needs extra prep */ - case TGSI_OPCODE_COS: - case TGSI_OPCODE_SIN: - /* The simple scalar ops. */ - case TGSI_OPCODE_EX2: - case TGSI_OPCODE_LG2: - case TGSI_OPCODE_RCP: - case TGSI_OPCODE_RSQ: - /* Copy red swizzle to alpha for src0 */ - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX; - inst->FullSrcRegisters[0].SrcRegister.SwizzleW = - inst->FullSrcRegisters[0].SrcRegister.SwizzleX; - /* Fall through */ - case TGSI_OPCODE_DDX: - case TGSI_OPCODE_DDY: - case TGSI_OPCODE_FRC: - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 1); - break; - - /* The dot products. */ - case TGSI_OPCODE_DPH: - /* Set alpha swizzle to one for src0 */ - if (!inst->FullSrcRegisters[0].SrcRegister.Extended) { - inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX = - inst->FullSrcRegisters[0].SrcRegister.SwizzleX; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY = - inst->FullSrcRegisters[0].SrcRegister.SwizzleY; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ = - inst->FullSrcRegisters[0].SrcRegister.SwizzleZ; - } - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = - TGSI_EXTSWIZZLE_ONE; - /* Fall through */ - case TGSI_OPCODE_DP3: - case TGSI_OPCODE_DP4: - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 2); - break; - - /* Simple three-source operations. */ - case TGSI_OPCODE_CMP: - /* Swap src0 and src2 */ - inst->FullSrcRegisters[3] = inst->FullSrcRegisters[2]; - inst->FullSrcRegisters[2] = inst->FullSrcRegisters[0]; - inst->FullSrcRegisters[0] = inst->FullSrcRegisters[3]; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - - /* The MAD variants. */ - case TGSI_OPCODE_SUB: - /* Just like ADD, but flip the negation on src1 first */ - inst->FullSrcRegisters[1].SrcRegister.Negate = - !inst->FullSrcRegisters[1].SrcRegister.Negate; - /* Fall through */ - case TGSI_OPCODE_ADD: - /* Force src0 to one, move all registers over */ - inst->FullSrcRegisters[2] = inst->FullSrcRegisters[1]; - inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0]; - inst->FullSrcRegisters[0] = r500_constant_one; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - case TGSI_OPCODE_MUL: - /* Force our src2 to zero */ - inst->FullSrcRegisters[2] = r500_constant_zero; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - case TGSI_OPCODE_MAD: - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - - /* The MOV variants. */ - case TGSI_OPCODE_ABS: - /* Set absolute value modifiers. */ - inst->FullSrcRegisters[0].SrcRegisterExtMod.Absolute = TRUE; - /* Fall through */ - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - /* src0 -> src1 and src2 forced to zero */ - inst->FullSrcRegisters[1] = inst->FullSrcRegisters[0]; - inst->FullSrcRegisters[2] = r500_constant_zero; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, 3); - break; - - /* The compound and hybrid insts. */ - case TGSI_OPCODE_LRP: - /* LRP DST A, B, C -> MAD TMP -A, C, C; MAD DST A, B, TMP */ - inst->FullSrcRegisters[3] = inst->FullSrcRegisters[1]; - inst->FullSrcRegisters[1] = inst->FullSrcRegisters[2]; - inst->FullSrcRegisters[0].SrcRegister.Negate = - !(inst->FullSrcRegisters[0].SrcRegister.Negate); - inst->FullDstRegisters[1] = inst->FullDstRegisters[0]; - inst->FullDstRegisters[0].DstRegister.Index = - assembler->temp_count; - inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3); - inst->FullSrcRegisters[2].SrcRegister.Index = - assembler->temp_count; - inst->FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY; - inst->FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; - inst->FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; - inst->FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; - inst->FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; - inst->FullSrcRegisters[1] = inst->FullSrcRegisters[3]; - inst->FullSrcRegisters[0].SrcRegister.Negate = - !(inst->FullSrcRegisters[0].SrcRegister.Negate); - inst->FullDstRegisters[0] = inst->FullDstRegisters[1]; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], TGSI_OPCODE_MAD, 3); - break; - case TGSI_OPCODE_POW: - /* POW DST A, B -> LG2 TMP A; MUL TMP TMP, B; EX2 DST TMP */ - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX; - inst->FullSrcRegisters[0].SrcRegister.SwizzleW = - inst->FullSrcRegisters[0].SrcRegister.SwizzleX; - inst->FullDstRegisters[1] = inst->FullDstRegisters[0]; - inst->FullDstRegisters[0].DstRegister.Index = - assembler->temp_count; - inst->FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], TGSI_OPCODE_LG2, 1); - inst->FullSrcRegisters[0].SrcRegister.Index = - assembler->temp_count; - inst->FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - inst->FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; - inst->FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; - inst->FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; - inst->FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; - inst->FullSrcRegisters[2] = r500_constant_zero; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], TGSI_OPCODE_MUL, 3); - inst->FullDstRegisters[0] = inst->FullDstRegisters[1]; - r500_emit_maths(fs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], TGSI_OPCODE_EX2, 1); - break; - - /* The texture instruction set. */ - case TGSI_OPCODE_KIL: - case TGSI_OPCODE_TEX: - case TGSI_OPCODE_TXB: - case TGSI_OPCODE_TXP: - r500_emit_tex(fs, assembler, &inst->FullSrcRegisters[0], - &inst->FullDstRegisters[0], inst->Instruction.Opcode); - break; - - /* This is the end. My only friend, the end. */ - case TGSI_OPCODE_END: - break; - default: - debug_printf("r300: fs: Bad opcode %d\n", - inst->Instruction.Opcode); - break; - } - - /* Clamp, if saturation flags are set. */ - if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) { - fs->instructions[fs->instruction_count - 1].inst0 |= - R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP; - } -} - -static void r300_fs_finalize(struct r3xx_fragment_shader* fs, - struct r300_fs_asm* assembler) -{ - fs->stack_size = assembler->temp_count + assembler->temp_offset + 1; -} - -static void r500_fs_finalize(struct r500_fragment_shader* fs, - struct r300_fs_asm* assembler) -{ - /* XXX should this just go with OPCODE_END? */ - fs->instructions[fs->instruction_count - 1].inst0 |= - R500_INST_LAST; -} - -void r300_translate_fragment_shader(struct r300_context* r300, - struct r3xx_fragment_shader* fs) -{ - struct tgsi_parse_context parser; - int i; - boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500; - struct r300_constant_buffer* consts = - &r300->shader_constants[PIPE_SHADER_FRAGMENT]; - - struct r300_fs_asm* assembler = CALLOC_STRUCT(r300_fs_asm); - if (assembler == NULL) { - return; - } - /* Setup starting offset for immediates. */ - assembler->imm_offset = consts->user_count; - /* Enable depth writes, if needed. */ - assembler->writes_depth = fs->info.writes_z; - - /* Make sure we start at the beginning of the shader. */ - if (is_r500) { - ((struct r500_fragment_shader*)fs)->instruction_count = 0; - } - - tgsi_parse_init(&parser, fs->state.tokens); - - while (!tgsi_parse_end_of_tokens(&parser)) { - tgsi_parse_token(&parser); - - /* This is seriously the lamest way to create fragment programs ever. - * I blame TGSI. */ - switch (parser.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* Allocated registers sitting at the beginning - * of the program. */ - r300_fs_declare(assembler, &parser.FullToken.FullDeclaration); - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - debug_printf("r300: Emitting immediate to constant buffer, " - "position %d\n", - assembler->imm_offset + assembler->imm_count); - /* I am not amused by the length of these. */ - for (i = 0; i < 4; i++) { - consts->constants[assembler->imm_offset + - assembler->imm_count][i] = - parser.FullToken.FullImmediate.u.ImmediateFloat32[i] - .Float; - } - assembler->imm_count++; - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - if (is_r500) { - r500_fs_instruction((struct r500_fragment_shader*)fs, - assembler, &parser.FullToken.FullInstruction); - } else { - r300_fs_instruction((struct r300_fragment_shader*)fs, - assembler, &parser.FullToken.FullInstruction); - } - break; - } - } - - debug_printf("r300: fs: %d texs and %d colors, first free reg is %d\n", - assembler->tex_count, assembler->color_count, - assembler->tex_count + assembler->color_count); - - consts->count = consts->user_count + assembler->imm_count; - fs->uses_imms = assembler->imm_count; - debug_printf("r300: fs: %d total constants, " - "%d from user and %d from immediates\n", consts->count, - consts->user_count, assembler->imm_count); - r300_fs_finalize(fs, assembler); - if (is_r500) { - r500_fs_finalize((struct r500_fragment_shader*)fs, assembler); - } - - tgsi_dump(fs->state.tokens); - /* XXX finish r300 dumper too */ - if (is_r500) { - r500_fs_dump((struct r500_fragment_shader*)fs); - } - - tgsi_parse_free(&parser); - FREE(assembler); -} diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r300_state_shader.h deleted file mode 100644 index b6087404cef..00000000000 --- a/src/gallium/drivers/r300/r300_state_shader.h +++ /dev/null @@ -1,220 +0,0 @@ -/* - * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#ifndef R300_STATE_SHADER_H -#define R300_STATE_SHADER_H - -#include "tgsi/tgsi_parse.h" - -#include "r300_context.h" -#include "r300_debug.h" -#include "r300_reg.h" -#include "r300_screen.h" - -/* XXX this all should find its way back to r300_reg */ -/* Swizzle tools */ -#define R500_SWIZZLE_ZERO 4 -#define R500_SWIZZLE_HALF 5 -#define R500_SWIZZLE_ONE 6 -#define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6)) -#define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6)) -#define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6)) -#define R500_SWIZ_MOD_NEG 1 -#define R500_SWIZ_MOD_ABS 2 -#define R500_SWIZ_MOD_NEG_ABS 3 -/* Swizzles for inst2 */ -#define R500_SWIZ_TEX_STRQ(x) ((x) << 8) -#define R500_SWIZ_TEX_RGBA(x) ((x) << 24) -/* Swizzles for inst3 */ -#define R500_SWIZ_RGB_A(x) ((x) << 2) -#define R500_SWIZ_RGB_B(x) ((x) << 15) -/* Swizzles for inst4 */ -#define R500_SWIZ_ALPHA_A(x) ((x) << 14) -#define R500_SWIZ_ALPHA_B(x) ((x) << 21) -/* Swizzle for inst5 */ -#define R500_SWIZ_RGBA_C(x) ((x) << 14) -#define R500_SWIZ_ALPHA_C(x) ((x) << 27) -/* Writemasks */ -#define R500_TEX_WMASK(x) ((x) << 11) -#define R500_ALU_WMASK(x) ((x) << 11) -#define R500_ALU_OMASK(x) ((x) << 15) -#define R500_W_OMASK (1 << 31) - -/* TGSI constants. TGSI is like XML: If it can't solve your problems, you're - * not using enough of it. */ -static const struct tgsi_full_src_register r500_constant_zero = { - .SrcRegister.Extended = TRUE, - .SrcRegister.File = TGSI_FILE_NULL, - .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO, -}; - -static const struct tgsi_full_src_register r500_constant_one = { - .SrcRegister.Extended = TRUE, - .SrcRegister.File = TGSI_FILE_NULL, - .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ONE, - .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ONE, - .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ONE, - .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ONE, -}; - -/* Temporary struct used to hold assembly state while putting together - * fragment programs. */ -struct r300_fs_asm { - /* Pipe context. */ - struct r300_context* r300; - /* Number of colors. */ - unsigned color_count; - /* Number of texcoords. */ - unsigned tex_count; - /* Offset for temporary registers. Inputs and temporaries have no - * distinguishing markings, so inputs start at 0 and the first usable - * temporary register is after all inputs. */ - unsigned temp_offset; - /* Number of requested temporary registers. */ - unsigned temp_count; - /* Offset for immediate constants. Neither R300 nor R500 can do four - * inline constants per source, so instead we copy immediates into the - * constant buffer. */ - unsigned imm_offset; - /* Number of immediate constants. */ - unsigned imm_count; - /* Are depth writes enabled? */ - boolean writes_depth; - /* Depth write offset. This is the TGSI output that corresponds to - * depth writes. */ - unsigned depth_output; -}; - -void r300_translate_fragment_shader(struct r300_context* r300, - struct r3xx_fragment_shader* fs); - -static struct r300_fragment_shader r300_passthrough_fragment_shader = { - .alu_instruction_count = 1, - .tex_instruction_count = 0, - .indirections = 0, - .shader.stack_size = 1, - - .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | - R300_ALU_OUTC_CMP, - .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | - R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ, - .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | - R300_ALU_OUTA_CMP, - .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) | - R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT, -}; - -static struct r500_fragment_shader r500_passthrough_fragment_shader = { - .shader.stack_size = 0, - .instruction_count = 1, - .instructions[0].inst0 = R500_INST_TYPE_OUT | - R500_INST_TEX_SEM_WAIT | R500_INST_LAST | - R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | - R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, - .instructions[0].inst1 = - R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST, - .instructions[0].inst2 = - R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST, - .instructions[0].inst3 = - R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B, - .instructions[0].inst4 = - R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A, - .instructions[0].inst5 = - R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0, -}; - -static struct r300_fragment_shader r300_texture_fragment_shader = { - .alu_instruction_count = 1, - .tex_instruction_count = 0, - .indirections = 0, - .shader.stack_size = 1, - - .instructions[0].alu_rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) | - R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | - R300_ALU_OUTC_CMP, - .instructions[0].alu_rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | - R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ, - .instructions[0].alu_alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) | - R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | - R300_ALU_OUTA_CMP, - .instructions[0].alu_alpha_addr = R300_ALPHA_ADDR0(0) | - R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT, -}; - -static struct r500_fragment_shader r500_texture_fragment_shader = { - .shader.stack_size = 1, - .instruction_count = 2, - .instructions[0].inst0 = R500_INST_TYPE_TEX | - R500_INST_TEX_SEM_WAIT | - R500_INST_RGB_WMASK_RGB | R500_INST_ALPHA_WMASK | - R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, - .instructions[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD | - R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED, - .instructions[0].inst2 = R500_TEX_SRC_ADDR(0) | - R500_TEX_SRC_S_SWIZ_R | R500_TEX_SRC_T_SWIZ_G | - R500_TEX_SRC_R_SWIZ_B | R500_TEX_SRC_Q_SWIZ_A | - R500_TEX_DST_ADDR(0) | - R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | - R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A, - .instructions[0].inst3 = 0x0, - .instructions[0].inst4 = 0x0, - .instructions[0].inst5 = 0x0, - .instructions[1].inst0 = R500_INST_TYPE_OUT | - R500_INST_TEX_SEM_WAIT | R500_INST_LAST | - R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | - R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, - .instructions[1].inst1 = - R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST, - .instructions[1].inst2 = - R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST, - .instructions[1].inst3 = - R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B, - .instructions[1].inst4 = - R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A, - .instructions[1].inst5 = - R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0, -}; - -#endif /* R300_STATE_SHADER_H */ diff --git a/src/gallium/drivers/r300/r300_state_tcl.c b/src/gallium/drivers/r300/r300_state_tcl.c deleted file mode 100644 index 8cf8250425e..00000000000 --- a/src/gallium/drivers/r300/r300_state_tcl.c +++ /dev/null @@ -1,412 +0,0 @@ -/* - * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. */ - -#include "r300_state_tcl.h" - -static void r300_vs_declare(struct r300_vs_asm* assembler, - struct tgsi_full_declaration* decl) -{ - switch (decl->Declaration.File) { - case TGSI_FILE_INPUT: - break; - case TGSI_FILE_OUTPUT: - switch (decl->Semantic.SemanticName) { - case TGSI_SEMANTIC_POSITION: - assembler->tab[decl->DeclarationRange.First] = 0; - break; - case TGSI_SEMANTIC_COLOR: - assembler->tab[decl->DeclarationRange.First] = - (assembler->point_size ? 1 : 0) + - assembler->out_colors; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - /* XXX multiple? */ - assembler->tab[decl->DeclarationRange.First] = - (assembler->point_size ? 1 : 0) + - assembler->out_colors + - assembler->out_texcoords; - break; - case TGSI_SEMANTIC_PSIZE: - assembler->tab[decl->DeclarationRange.First] = 1; - break; - default: - debug_printf("r300: vs: Bad semantic declaration %d\n", - decl->Semantic.SemanticName); - break; - } - break; - case TGSI_FILE_CONSTANT: - break; - case TGSI_FILE_TEMPORARY: - assembler->temp_count++; - break; - default: - debug_printf("r300: vs: Bad file %d\n", decl->Declaration.File); - break; - } -} - -static INLINE unsigned r300_vs_src_type(struct r300_vs_asm* assembler, - struct tgsi_src_register* src) -{ - switch (src->File) { - case TGSI_FILE_NULL: - case TGSI_FILE_INPUT: - /* Probably a zero or one swizzle */ - return R300_PVS_SRC_REG_INPUT; - case TGSI_FILE_TEMPORARY: - return R300_PVS_SRC_REG_TEMPORARY; - case TGSI_FILE_CONSTANT: - case TGSI_FILE_IMMEDIATE: - return R300_PVS_SRC_REG_CONSTANT; - default: - debug_printf("r300: vs: Unimplemented src type %d\n", src->File); - break; - } - return 0; -} - -static INLINE unsigned r300_vs_src(struct r300_vs_asm* assembler, - struct tgsi_src_register* src) -{ - switch (src->File) { - case TGSI_FILE_NULL: - case TGSI_FILE_INPUT: - case TGSI_FILE_TEMPORARY: - case TGSI_FILE_CONSTANT: - return src->Index; - case TGSI_FILE_IMMEDIATE: - return src->Index + assembler->imm_offset; - default: - debug_printf("r300: vs: Unimplemented src type %d\n", src->File); - break; - } - return 0; -} - -static INLINE unsigned r300_vs_dst_type(struct r300_vs_asm* assembler, - struct tgsi_dst_register* dst) -{ - switch (dst->File) { - case TGSI_FILE_TEMPORARY: - return R300_PVS_DST_REG_TEMPORARY; - case TGSI_FILE_OUTPUT: - return R300_PVS_DST_REG_OUT; - default: - debug_printf("r300: vs: Unimplemented dst type %d\n", dst->File); - break; - } - return 0; -} - -static INLINE unsigned r300_vs_dst(struct r300_vs_asm* assembler, - struct tgsi_dst_register* dst) -{ - switch (dst->File) { - case TGSI_FILE_TEMPORARY: - return dst->Index; - case TGSI_FILE_OUTPUT: - return assembler->tab[dst->Index]; - default: - debug_printf("r300: vs: Unimplemented dst %d\n", dst->File); - break; - } - return 0; -} - -static uint32_t r300_vs_op(unsigned op) -{ - switch (op) { - case TGSI_OPCODE_DP3: - case TGSI_OPCODE_DP4: - return R300_VE_DOT_PRODUCT; - case TGSI_OPCODE_MUL: - return R300_VE_MULTIPLY; - case TGSI_OPCODE_ADD: - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SUB: - case TGSI_OPCODE_SWZ: - return R300_VE_ADD; - case TGSI_OPCODE_MAX: - return R300_VE_MAXIMUM; - case TGSI_OPCODE_SLT: - return R300_VE_SET_LESS_THAN; - case TGSI_OPCODE_RSQ: - return R300_PVS_DST_MATH_INST | R300_ME_RECIP_DX; - case TGSI_OPCODE_MAD: - return R300_PVS_DST_MACRO_INST | R300_PVS_MACRO_OP_2CLK_MADD; - default: - break; - } - return 0; -} - -static uint32_t r300_vs_swiz(struct tgsi_full_src_register* reg) -{ - if (reg->SrcRegister.Extended) { - return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | - reg->SrcRegisterExtSwz.ExtSwizzleX | - (reg->SrcRegisterExtSwz.ExtSwizzleY << 3) | - (reg->SrcRegisterExtSwz.ExtSwizzleZ << 6) | - (reg->SrcRegisterExtSwz.ExtSwizzleW << 9); - } else { - return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | - reg->SrcRegister.SwizzleX | - (reg->SrcRegister.SwizzleY << 3) | - (reg->SrcRegister.SwizzleZ << 6) | - (reg->SrcRegister.SwizzleW << 9); - } -} - -/* XXX icky icky icky icky */ -static uint32_t r300_vs_scalar_swiz(struct tgsi_full_src_register* reg) -{ - if (reg->SrcRegister.Extended) { - return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | - reg->SrcRegisterExtSwz.ExtSwizzleX | - (reg->SrcRegisterExtSwz.ExtSwizzleX << 3) | - (reg->SrcRegisterExtSwz.ExtSwizzleX << 6) | - (reg->SrcRegisterExtSwz.ExtSwizzleX << 9); - } else { - return (reg->SrcRegister.Negate ? (0xf << 12) : 0) | - reg->SrcRegister.SwizzleX | - (reg->SrcRegister.SwizzleX << 3) | - (reg->SrcRegister.SwizzleX << 6) | - (reg->SrcRegister.SwizzleX << 9); - } -} - -/* XXX scalar stupidity */ -static void r300_vs_emit_inst(struct r300_vertex_shader* vs, - struct r300_vs_asm* assembler, - struct tgsi_full_src_register* src, - struct tgsi_full_dst_register* dst, - unsigned op, - unsigned count, - boolean is_scalar) -{ - int i = vs->instruction_count; - vs->instructions[i].inst0 = R300_PVS_DST_OPCODE(r300_vs_op(op)) | - R300_PVS_DST_REG_TYPE(r300_vs_dst_type(assembler, &dst->DstRegister)) | - R300_PVS_DST_OFFSET(r300_vs_dst(assembler, &dst->DstRegister)) | - R300_PVS_DST_WE(dst->DstRegister.WriteMask); - switch (count) { - case 3: - vs->instructions[i].inst3 = - R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, - &src[2].SrcRegister)) | - R300_PVS_SRC_OFFSET(r300_vs_src(assembler, - &src[2].SrcRegister)) | - R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[2])); - /* Fall through */ - case 2: - vs->instructions[i].inst2 = - R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, - &src[1].SrcRegister)) | - R300_PVS_SRC_OFFSET(r300_vs_src(assembler, - &src[1].SrcRegister)) | - R300_PVS_SRC_SWIZZLE(r300_vs_swiz(&src[1])); - /* Fall through */ - case 1: - vs->instructions[i].inst1 = - R300_PVS_SRC_REG_TYPE(r300_vs_src_type(assembler, - &src[0].SrcRegister)) | - R300_PVS_SRC_OFFSET(r300_vs_src(assembler, - &src[0].SrcRegister)) | - /* XXX the icky, it burns */ - R300_PVS_SRC_SWIZZLE(is_scalar ? r300_vs_scalar_swiz(&src[0]) - : r300_vs_swiz(&src[0])); - break; - } - vs->instruction_count++; -} - -static void r300_vs_instruction(struct r300_vertex_shader* vs, - struct r300_vs_asm* assembler, - struct tgsi_full_instruction* inst) -{ - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_RSQ: - r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 1, TRUE); - break; - case TGSI_OPCODE_SUB: - inst->FullSrcRegisters[1].SrcRegister.Negate = - !inst->FullSrcRegisters[1].SrcRegister.Negate; - /* Fall through */ - case TGSI_OPCODE_ADD: - case TGSI_OPCODE_MUL: - case TGSI_OPCODE_MAX: - case TGSI_OPCODE_SLT: - r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 2, FALSE); - break; - case TGSI_OPCODE_DP3: - /* Set alpha swizzle to zero for src0 and src1 */ - if (!inst->FullSrcRegisters[0].SrcRegister.Extended) { - inst->FullSrcRegisters[0].SrcRegister.Extended = TRUE; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleX = - inst->FullSrcRegisters[0].SrcRegister.SwizzleX; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleY = - inst->FullSrcRegisters[0].SrcRegister.SwizzleY; - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleZ = - inst->FullSrcRegisters[0].SrcRegister.SwizzleZ; - } - inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtSwizzleW = - TGSI_EXTSWIZZLE_ZERO; - if (!inst->FullSrcRegisters[1].SrcRegister.Extended) { - inst->FullSrcRegisters[1].SrcRegister.Extended = TRUE; - inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleX = - inst->FullSrcRegisters[1].SrcRegister.SwizzleX; - inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleY = - inst->FullSrcRegisters[1].SrcRegister.SwizzleY; - inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleZ = - inst->FullSrcRegisters[1].SrcRegister.SwizzleZ; - } - inst->FullSrcRegisters[1].SrcRegisterExtSwz.ExtSwizzleW = - TGSI_EXTSWIZZLE_ZERO; - /* Fall through */ - case TGSI_OPCODE_DP4: - r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 2, FALSE); - break; - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - inst->FullSrcRegisters[1] = r300_constant_zero; - r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 2, FALSE); - break; - case TGSI_OPCODE_MAD: - r300_vs_emit_inst(vs, assembler, inst->FullSrcRegisters, - &inst->FullDstRegisters[0], inst->Instruction.Opcode, - 3, FALSE); - break; - case TGSI_OPCODE_END: - break; - default: - debug_printf("r300: vs: Bad opcode %d\n", - inst->Instruction.Opcode); - break; - } -} - -static void r300_vs_init(struct r300_vertex_shader* vs, - struct r300_vs_asm* assembler) -{ - struct tgsi_shader_info* info = &vs->info; - int i; - - for (i = 0; i < info->num_outputs; i++) { - switch (info->output_semantic_name[i]) { - case TGSI_SEMANTIC_PSIZE: - assembler->point_size = TRUE; - break; - case TGSI_SEMANTIC_COLOR: - assembler->out_colors++; - break; - case TGSI_SEMANTIC_FOG: - case TGSI_SEMANTIC_GENERIC: - assembler->out_texcoords++; - break; - } - } - - vs->instruction_count = 0; -} - -void r300_translate_vertex_shader(struct r300_context* r300, - struct r300_vertex_shader* vs) -{ - struct tgsi_parse_context parser; - int i; - struct r300_constant_buffer* consts = - &r300->shader_constants[PIPE_SHADER_VERTEX]; - - struct r300_vs_asm* assembler = CALLOC_STRUCT(r300_vs_asm); - if (assembler == NULL) { - return; - } - - /* Init assembler. */ - r300_vs_init(vs, assembler); - - /* Setup starting offset for immediates. */ - assembler->imm_offset = consts->user_count; - - tgsi_parse_init(&parser, vs->state.tokens); - - while (!tgsi_parse_end_of_tokens(&parser)) { - tgsi_parse_token(&parser); - - /* This is seriously the lamest way to create fragment programs ever. - * I blame TGSI. */ - switch (parser.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* Allocated registers sitting at the beginning - * of the program. */ - r300_vs_declare(assembler, &parser.FullToken.FullDeclaration); - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - debug_printf("r300: Emitting immediate to constant buffer, " - "position %d\n", - assembler->imm_offset + assembler->imm_count); - /* I am not amused by the length of these. */ - for (i = 0; i < 4; i++) { - consts->constants[assembler->imm_offset + - assembler->imm_count][i] = - parser.FullToken.FullImmediate.u.ImmediateFloat32[i] - .Float; - } - assembler->imm_count++; - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - r300_vs_instruction(vs, assembler, - &parser.FullToken.FullInstruction); - break; - } - } - - debug_printf("r300: vs: %d texs and %d colors, first free reg is %d\n", - assembler->tex_count, assembler->color_count, - assembler->tex_count + assembler->color_count); - - consts->count = consts->user_count + assembler->imm_count; - vs->uses_imms = assembler->imm_count; - debug_printf("r300: vs: %d total constants, " - "%d from user and %d from immediates\n", consts->count, - consts->user_count, assembler->imm_count); - - debug_printf("r300: vs: tab: %d %d %d %d\n", assembler->tab[0], - assembler->tab[1], assembler->tab[2], assembler->tab[3]); - - tgsi_dump(vs->state.tokens); - /* XXX finish r300 vertex shader dumper */ - r300_vs_dump(vs); - - tgsi_parse_free(&parser); - FREE(assembler); -} diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c index c9e2dff14ed..a093f839454 100644 --- a/src/gallium/drivers/r300/r300_surface.c +++ b/src/gallium/drivers/r300/r300_surface.c @@ -37,7 +37,7 @@ static void r300_surface_setup(struct r300_context* r300, r300_emit_dsa_state(r300, &dsa_clear_state); r300_emit_rs_state(r300, &rs_clear_state); - BEGIN_CS(24); + BEGIN_CS(26); /* Viewport setup */ OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6); @@ -78,8 +78,10 @@ static void r300_surface_setup(struct r300_context* r300, /* Setup colorbuffer. */ OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1); OUT_CS_RELOC(dest->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_CS_REG(R300_RB3D_COLORPITCH0, pixpitch | - r300_translate_colorformat(dest->tex.format)); + OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0, 1); + OUT_CS_RELOC(dest->buffer, pixpitch | + r300_translate_colorformat(dest->tex.format), 0, + RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0xf); END_CS; @@ -125,9 +127,10 @@ validate: r300->context.flush(&r300->context, 0, NULL); goto validate; } - if (r300->winsys->validate(r300->winsys)) { + if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); if (invalid) { + debug_printf("r300: Stuck in validation loop, gonna fallback."); goto fallback; } invalid = TRUE; @@ -138,10 +141,14 @@ validate: /* Vertex shader setup */ if (caps->has_tcl) { - r300_emit_vertex_shader(r300, &r300_passthrough_vertex_shader); + r300_emit_vertex_program_code(r300, &r300_passthrough_vertex_shader, 0); } else { BEGIN_CS(4); - OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS); + OUT_CS_REG(R300_VAP_CNTL_STATUS, +#ifdef PIPE_ARCH_BIG_ENDIAN + R300_VC_32BIT_SWAP | +#endif + R300_VAP_TCL_BYPASS); OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(5) | R300_PVS_NUM_CNTLRS(5) | R300_PVS_NUM_FPUS(caps->num_vert_fpus) | @@ -151,11 +158,11 @@ validate: /* Fragment shader setup */ if (caps->is_r500) { - r500_emit_fragment_shader(r300, &r500_passthrough_fragment_shader); - r300_emit_rs_block_state(r300, &r500_rs_block_clear_state); + r500_emit_fragment_program_code(r300, &r5xx_passthrough_fragment_shader, 0); + r300_emit_rs_block_state(r300, &r5xx_rs_block_clear_state); } else { - r300_emit_fragment_shader(r300, &r300_passthrough_fragment_shader); - r300_emit_rs_block_state(r300, &r300_rs_block_clear_state); + r300_emit_fragment_program_code(r300, &r3xx_passthrough_fragment_shader, 0); + r300_emit_rs_block_state(r300, &r3xx_rs_block_clear_state); } BEGIN_CS(26); @@ -228,13 +235,14 @@ static void r300_surface_copy(struct pipe_context* pipe, struct r300_texture* desttex = (struct r300_texture*)dest->texture; unsigned pixpitch = srctex->stride / srctex->tex.block.size; boolean invalid = FALSE; + float fsrcx = srcx, fsrcy = srcy, fdestx = destx, fdesty = desty; CS_LOCALS(r300); debug_printf("r300: Copying surface %p at (%d,%d) to %p at (%d, %d)," " dimensions %dx%d (pixel pitch %d)\n", src, srcx, srcy, dest, destx, desty, w, h, pixpitch); - if ((srctex == desttex) && + if ((srctex->buffer == desttex->buffer) && ((destx < srcx + w) || (srcx < destx + w)) && ((desty < srcy + h) || (srcy < desty + h))) { fallback: @@ -255,9 +263,10 @@ validate: r300->context.flush(&r300->context, 0, NULL); goto validate; } - if (r300->winsys->validate(r300->winsys)) { + if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); if (invalid) { + debug_printf("r300: Stuck in validation loop, gonna fallback."); goto fallback; } invalid = TRUE; @@ -267,21 +276,21 @@ validate: r300_surface_setup(r300, desttex, destx, desty, w, h); /* Setup the texture. */ - r300_emit_sampler(r300, &r300_sampler_copy_state, 0); - r300_emit_texture(r300, srctex, 0); + r300_emit_texture(r300, &r300_sampler_copy_state, srctex, 0); /* Flush and enable. */ - BEGIN_CS(4); - OUT_CS_REG(R300_TX_INVALTAGS, 0); - OUT_CS_REG(R300_TX_ENABLE, 0x1); - END_CS; + r300_flush_textures(r300); /* Vertex shader setup */ if (caps->has_tcl) { - r300_emit_vertex_shader(r300, &r300_passthrough_vertex_shader); + r300_emit_vertex_program_code(r300, &r300_passthrough_vertex_shader, 0); } else { BEGIN_CS(4); - OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VAP_TCL_BYPASS); + OUT_CS_REG(R300_VAP_CNTL_STATUS, +#ifdef PIPE_ARCH_BIG_ENDIAN + R300_VC_32BIT_SWAP | +#endif + R300_VAP_TCL_BYPASS); OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(5) | R300_PVS_NUM_CNTLRS(5) | R300_PVS_NUM_FPUS(caps->num_vert_fpus) | @@ -291,11 +300,11 @@ validate: /* Fragment shader setup */ if (caps->is_r500) { - r500_emit_fragment_shader(r300, &r500_texture_fragment_shader); - r300_emit_rs_block_state(r300, &r500_rs_block_copy_state); + r500_emit_fragment_program_code(r300, &r5xx_texture_fragment_shader, 0); + r300_emit_rs_block_state(r300, &r5xx_rs_block_copy_state); } else { - r300_emit_fragment_shader(r300, &r300_texture_fragment_shader); - r300_emit_rs_block_state(r300, &r300_rs_block_copy_state); + r300_emit_fragment_program_code(r300, &r3xx_texture_fragment_shader, 0); + r300_emit_rs_block_state(r300, &r3xx_rs_block_copy_state); } BEGIN_CS(30); @@ -329,25 +338,25 @@ validate: OUT_CS(R300_PRIM_TYPE_QUADS | R300_PRIM_WALK_RING | (4 << R300_PRIM_NUM_VERTICES_SHIFT)); /* (x , y ) */ - OUT_CS_32F((float)(destx / dest->width)); - OUT_CS_32F((float)(desty / dest->height)); - OUT_CS_32F((float)(srcx / dest->width)); - OUT_CS_32F((float)(srcy / dest->height)); + OUT_CS_32F(fdestx / dest->width); + OUT_CS_32F(fdesty / dest->height); + OUT_CS_32F(fsrcx / src->width); + OUT_CS_32F(fsrcy / src->height); /* (x , y + h) */ - OUT_CS_32F((float)(destx / dest->width)); - OUT_CS_32F((float)((desty + h) / dest->height)); - OUT_CS_32F((float)(srcx / dest->width)); - OUT_CS_32F((float)((srcy + h) / dest->height)); + OUT_CS_32F(fdestx / dest->width); + OUT_CS_32F((fdesty + h) / dest->height); + OUT_CS_32F(fsrcx / src->width); + OUT_CS_32F((fsrcy + h) / src->height); /* (x + w, y + h) */ - OUT_CS_32F((float)((destx + w) / dest->width)); - OUT_CS_32F((float)((desty + h) / dest->height)); - OUT_CS_32F((float)((srcx + w) / dest->width)); - OUT_CS_32F((float)((srcy + h) / dest->height)); + OUT_CS_32F((fdestx + w) / dest->width); + OUT_CS_32F((fdesty + h) / dest->height); + OUT_CS_32F((fsrcx + w) / src->width); + OUT_CS_32F((fsrcy + h) / src->height); /* (x + w, y ) */ - OUT_CS_32F((float)((destx + w) / dest->width)); - OUT_CS_32F((float)(desty / dest->height)); - OUT_CS_32F((float)((srcx + w) / dest->width)); - OUT_CS_32F((float)(srcy / dest->height)); + OUT_CS_32F((fdestx + w) / dest->width); + OUT_CS_32F(fdesty / dest->height); + OUT_CS_32F((fsrcx + w) / src->width); + OUT_CS_32F(fsrcy / src->height); OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA); diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h index 9a4c39f58bd..d01f0b143f5 100644 --- a/src/gallium/drivers/r300/r300_surface.h +++ b/src/gallium/drivers/r300/r300_surface.h @@ -31,8 +31,8 @@ #include "r300_context.h" #include "r300_cs.h" #include "r300_emit.h" -#include "r300_state_shader.h" -#include "r300_state_tcl.h" +#include "r300_fs.h" +#include "r300_vs.h" #include "r300_state_inlines.h" static struct r300_blend_state blend_clear_state = { @@ -72,7 +72,7 @@ static struct r300_rs_state rs_clear_state = { .color_control = R300_SHADE_MODEL_FLAT, }; -static struct r300_rs_block r300_rs_block_clear_state = { +static struct r300_rs_block r3xx_rs_block_clear_state = { .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) | R500_RS_SEL_T(R300_RS_SEL_K0) | R500_RS_SEL_R(R300_RS_SEL_K0) | @@ -82,7 +82,7 @@ static struct r300_rs_block r300_rs_block_clear_state = { .inst_count = 0, }; -static struct r300_rs_block r500_rs_block_clear_state = { +static struct r300_rs_block r5xx_rs_block_clear_state = { .ip[0] = R500_RS_SEL_S(R500_RS_IP_PTR_K0) | R500_RS_SEL_T(R500_RS_IP_PTR_K0) | R500_RS_SEL_R(R500_RS_IP_PTR_K0) | @@ -94,7 +94,7 @@ static struct r300_rs_block r500_rs_block_clear_state = { /* The following state is used for surface_copy only. */ -static struct r300_rs_block r300_rs_block_copy_state = { +static struct r300_rs_block r3xx_rs_block_copy_state = { .ip[0] = R500_RS_SEL_S(R300_RS_SEL_K0) | R500_RS_SEL_T(R300_RS_SEL_K0) | R500_RS_SEL_R(R300_RS_SEL_K0) | @@ -104,7 +104,7 @@ static struct r300_rs_block r300_rs_block_copy_state = { .inst_count = R300_RS_TX_OFFSET(0), }; -static struct r300_rs_block r500_rs_block_copy_state = { +static struct r300_rs_block r5xx_rs_block_copy_state = { .ip[0] = R500_RS_SEL_S(0) | R500_RS_SEL_T(1) | R500_RS_SEL_R(R500_RS_IP_PTR_K0) | diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 5ea9f56247b..590052509cc 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -22,33 +22,35 @@ #include "r300_texture.h" -static int minify(int i) -{ - return MAX2(1, i >> 1); -} - static void r300_setup_texture_state(struct r300_texture* tex, unsigned width, unsigned height, - unsigned pitch) + unsigned pitch, + unsigned levels) { struct r300_texture_state* state = &tex->state; state->format0 = R300_TX_WIDTH((width - 1) & 0x7ff) | - R300_TX_HEIGHT((height - 1) & 0x7ff) | R300_TX_PITCH_EN; + R300_TX_HEIGHT((height - 1) & 0x7ff) | + R300_TX_NUM_LEVELS(levels) | + R300_TX_PITCH_EN; /* XXX */ - state->format1 = R300_TX_FORMAT_A8R8G8B8; + state->format1 = r300_translate_texformat(tex->tex.format); state->format2 = pitch - 1; - /* XXX + /* Assume (somewhat foolishly) that oversized textures will + * not be permitted by the state tracker. */ if (width > 2048) { - state->pitch |= R300_TXWIDTH_11; + state->format2 |= R500_TXWIDTH_BIT11; } if (height > 2048) { - state->pitch |= R300_TXHEIGHT_11; - } */ + state->format2 |= R500_TXHEIGHT_BIT11; + } + + debug_printf("r300: Set texture state (%dx%d, pitch %d, %d levels)\n", + width, height, pitch, levels); } static void r300_setup_miptree(struct r300_texture* tex) @@ -65,16 +67,24 @@ static void r300_setup_miptree(struct r300_texture* tex) } base->nblocksx[i] = pf_get_nblocksx(&base->block, base->width[i]); - base->nblocksy[i] = pf_get_nblocksy(&base->block, base->width[i]); - - /* Radeons enjoy things in multiples of 32. */ - /* XXX this can be 32 when POT */ - stride = (base->nblocksx[i] * base->block.size + 63) & ~63; + base->nblocksy[i] = pf_get_nblocksy(&base->block, base->height[i]); + + /* Radeons enjoy things in multiples of 64. + * + * XXX + * POT, uncompressed, unmippmapped textures can be aligned to 32, + * instead of 64. */ + stride = align(pf_get_stride(&base->block, base->width[i]), 32); size = stride * base->nblocksy[i] * base->depth[i]; - tex->offset[i] = (tex->size + 63) & ~63; + tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; + debug_printf("r300: Texture miptree: Level %d " + "(%dx%dx%d px, pitch %d bytes)\n", + i, base->width[i], base->height[i], base->depth[i], + stride); + /* Save stride of first level to the texture. */ if (i == 0) { tex->stride = stride; } @@ -98,11 +108,10 @@ static struct pipe_texture* r300_setup_miptree(tex); - /* XXX */ - r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0], - tex->tex.width[0]); + r300_setup_texture_state(tex, template->width[0], template->height[0], + template->width[0], template->last_level); - tex->buffer = screen->buffer_create(screen, 64, + tex->buffer = screen->buffer_create(screen, 1024, PIPE_BUFFER_USAGE_PIXEL, tex->size); @@ -164,6 +173,7 @@ static struct pipe_texture* { struct r300_texture* tex; + /* XXX we should start doing mips now... */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || base->depth[0] != 1) { @@ -181,8 +191,9 @@ static struct pipe_texture* tex->stride = *stride; + /* XXX */ r300_setup_texture_state(tex, tex->tex.width[0], tex->tex.height[0], - tex->stride); + tex->stride, 0); pipe_buffer_reference(&tex->buffer, buffer); diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h index 98fb5c9a08f..3b56f0307c0 100644 --- a/src/gallium/drivers/r300/r300_texture.h +++ b/src/gallium/drivers/r300/r300_texture.h @@ -32,6 +32,52 @@ void r300_init_screen_texture_functions(struct pipe_screen* screen); +/* Note the signature of R300_EASY_TX_FORMAT(A, R, G, B, FORMAT)... */ +static INLINE uint32_t r300_translate_texformat(enum pipe_format format) +{ + switch (format) { + /* X8 */ + case PIPE_FORMAT_I8_UNORM: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + /* W8Z8Y8X8 */ + case PIPE_FORMAT_A8R8G8B8_UNORM: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); + case PIPE_FORMAT_R8G8B8A8_UNORM: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case PIPE_FORMAT_A8R8G8B8_SRGB: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8) | + R300_TX_FORMAT_GAMMA; + case PIPE_FORMAT_R8G8B8A8_SRGB: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | + R300_TX_FORMAT_GAMMA; + /* DXT1 */ + case PIPE_FORMAT_DXT1_RGB: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1); + case PIPE_FORMAT_DXT1_RGBA: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1); + /* DXT3 */ + case PIPE_FORMAT_DXT3_RGBA: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3); + /* DXT5 */ + case PIPE_FORMAT_DXT5_RGBA: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5); + /* YVYU422 */ + case PIPE_FORMAT_YCBCR: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, YVYU422) | + R300_TX_FORMAT_YUV_TO_RGB; + /* W24_FP */ + case PIPE_FORMAT_Z24S8_UNORM: + return R300_EASY_TX_FORMAT(X, X, X, X, W24_FP); + default: + debug_printf("r300: Implementation error: " + "Got unsupported texture format %s in %s\n", + pf_name(format), __FUNCTION__); + assert(0); + break; + } + return 0; +} + #ifndef R300_WINSYS_H boolean r300_get_texture_buffer(struct pipe_texture* texture, diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c new file mode 100644 index 00000000000..3adbb715f37 --- /dev/null +++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c @@ -0,0 +1,337 @@ +/* + * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r300_tgsi_to_rc.h" + +#include "radeon_compiler.h" +#include "radeon_program.h" + +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_scan.h" +#include "tgsi/tgsi_util.h" + + +static unsigned translate_opcode(unsigned opcode) +{ + switch(opcode) { + case TGSI_OPCODE_ARL: return OPCODE_ARL; + case TGSI_OPCODE_MOV: return OPCODE_MOV; + case TGSI_OPCODE_LIT: return OPCODE_LIT; + case TGSI_OPCODE_RCP: return OPCODE_RCP; + case TGSI_OPCODE_RSQ: return OPCODE_RSQ; + case TGSI_OPCODE_EXP: return OPCODE_EXP; + case TGSI_OPCODE_LOG: return OPCODE_LOG; + case TGSI_OPCODE_MUL: return OPCODE_MUL; + case TGSI_OPCODE_ADD: return OPCODE_ADD; + case TGSI_OPCODE_DP3: return OPCODE_DP3; + case TGSI_OPCODE_DP4: return OPCODE_DP4; + case TGSI_OPCODE_DST: return OPCODE_DST; + case TGSI_OPCODE_MIN: return OPCODE_MIN; + case TGSI_OPCODE_MAX: return OPCODE_MAX; + case TGSI_OPCODE_SLT: return OPCODE_SLT; + case TGSI_OPCODE_SGE: return OPCODE_SGE; + case TGSI_OPCODE_MAD: return OPCODE_MAD; + case TGSI_OPCODE_SUB: return OPCODE_SUB; + case TGSI_OPCODE_LRP: return OPCODE_LRP; + /* case TGSI_OPCODE_CND: return OPCODE_CND; */ + /* case TGSI_OPCODE_CND0: return OPCODE_CND0; */ + case TGSI_OPCODE_DP2A: return OPCODE_DP2A; + /* gap */ + case TGSI_OPCODE_FRC: return OPCODE_FRC; + /* case TGSI_OPCODE_CLAMP: return OPCODE_CLAMP; */ + case TGSI_OPCODE_FLR: return OPCODE_FLR; + /* case TGSI_OPCODE_ROUND: return OPCODE_ROUND; */ + case TGSI_OPCODE_EX2: return OPCODE_EX2; + case TGSI_OPCODE_LG2: return OPCODE_LG2; + case TGSI_OPCODE_POW: return OPCODE_POW; + case TGSI_OPCODE_XPD: return OPCODE_XPD; + /* gap */ + case TGSI_OPCODE_ABS: return OPCODE_ABS; + case TGSI_OPCODE_RCC: return OPCODE_RCC; + case TGSI_OPCODE_DPH: return OPCODE_DPH; + case TGSI_OPCODE_COS: return OPCODE_COS; + case TGSI_OPCODE_DDX: return OPCODE_DDX; + case TGSI_OPCODE_DDY: return OPCODE_DDY; + /* case TGSI_OPCODE_KILP: return OPCODE_KILP; */ + case TGSI_OPCODE_PK2H: return OPCODE_PK2H; + case TGSI_OPCODE_PK2US: return OPCODE_PK2US; + case TGSI_OPCODE_PK4B: return OPCODE_PK4B; + case TGSI_OPCODE_PK4UB: return OPCODE_PK4UB; + case TGSI_OPCODE_RFL: return OPCODE_RFL; + case TGSI_OPCODE_SEQ: return OPCODE_SEQ; + case TGSI_OPCODE_SFL: return OPCODE_SFL; + case TGSI_OPCODE_SGT: return OPCODE_SGT; + case TGSI_OPCODE_SIN: return OPCODE_SIN; + case TGSI_OPCODE_SLE: return OPCODE_SLE; + case TGSI_OPCODE_SNE: return OPCODE_SNE; + case TGSI_OPCODE_STR: return OPCODE_STR; + case TGSI_OPCODE_TEX: return OPCODE_TEX; + case TGSI_OPCODE_TXD: return OPCODE_TXD; + case TGSI_OPCODE_TXP: return OPCODE_TXP; + case TGSI_OPCODE_UP2H: return OPCODE_UP2H; + case TGSI_OPCODE_UP2US: return OPCODE_UP2US; + case TGSI_OPCODE_UP4B: return OPCODE_UP4B; + case TGSI_OPCODE_UP4UB: return OPCODE_UP4UB; + case TGSI_OPCODE_X2D: return OPCODE_X2D; + case TGSI_OPCODE_ARA: return OPCODE_ARA; + case TGSI_OPCODE_ARR: return OPCODE_ARR; + case TGSI_OPCODE_BRA: return OPCODE_BRA; + case TGSI_OPCODE_CAL: return OPCODE_CAL; + case TGSI_OPCODE_RET: return OPCODE_RET; + case TGSI_OPCODE_SSG: return OPCODE_SSG; + case TGSI_OPCODE_CMP: return OPCODE_CMP; + case TGSI_OPCODE_SCS: return OPCODE_SCS; + case TGSI_OPCODE_TXB: return OPCODE_TXB; + /* case TGSI_OPCODE_NRM: return OPCODE_NRM; */ + /* case TGSI_OPCODE_DIV: return OPCODE_DIV; */ + case TGSI_OPCODE_DP2: return OPCODE_DP2; + case TGSI_OPCODE_TXL: return OPCODE_TXL; + case TGSI_OPCODE_BRK: return OPCODE_BRK; + case TGSI_OPCODE_IF: return OPCODE_IF; + /* case TGSI_OPCODE_LOOP: return OPCODE_LOOP; */ + /* case TGSI_OPCODE_REP: return OPCODE_REP; */ + case TGSI_OPCODE_ELSE: return OPCODE_ELSE; + case TGSI_OPCODE_ENDIF: return OPCODE_ENDIF; + case TGSI_OPCODE_ENDLOOP: return OPCODE_ENDLOOP; + /* case TGSI_OPCODE_ENDREP: return OPCODE_ENDREP; */ + case TGSI_OPCODE_PUSHA: return OPCODE_PUSHA; + case TGSI_OPCODE_POPA: return OPCODE_POPA; + /* case TGSI_OPCODE_CEIL: return OPCODE_CEIL; */ + /* case TGSI_OPCODE_I2F: return OPCODE_I2F; */ + case TGSI_OPCODE_NOT: return OPCODE_NOT; + case TGSI_OPCODE_TRUNC: return OPCODE_TRUNC; + /* case TGSI_OPCODE_SHL: return OPCODE_SHL; */ + /* case TGSI_OPCODE_SHR: return OPCODE_SHR; */ + case TGSI_OPCODE_AND: return OPCODE_AND; + case TGSI_OPCODE_OR: return OPCODE_OR; + /* case TGSI_OPCODE_MOD: return OPCODE_MOD; */ + case TGSI_OPCODE_XOR: return OPCODE_XOR; + /* case TGSI_OPCODE_SAD: return OPCODE_SAD; */ + /* case TGSI_OPCODE_TXF: return OPCODE_TXF; */ + /* case TGSI_OPCODE_TXQ: return OPCODE_TXQ; */ + case TGSI_OPCODE_CONT: return OPCODE_CONT; + /* case TGSI_OPCODE_EMIT: return OPCODE_EMIT; */ + /* case TGSI_OPCODE_ENDPRIM: return OPCODE_ENDPRIM; */ + /* case TGSI_OPCODE_BGNLOOP2: return OPCODE_BGNLOOP2; */ + case TGSI_OPCODE_BGNSUB: return OPCODE_BGNSUB; + /* case TGSI_OPCODE_ENDLOOP2: return OPCODE_ENDLOOP2; */ + case TGSI_OPCODE_ENDSUB: return OPCODE_ENDSUB; + case TGSI_OPCODE_NOISE1: return OPCODE_NOISE1; + case TGSI_OPCODE_NOISE2: return OPCODE_NOISE2; + case TGSI_OPCODE_NOISE3: return OPCODE_NOISE3; + case TGSI_OPCODE_NOISE4: return OPCODE_NOISE4; + case TGSI_OPCODE_NOP: return OPCODE_NOP; + /* gap */ + case TGSI_OPCODE_NRM4: return OPCODE_NRM4; + /* case TGSI_OPCODE_CALLNZ: return OPCODE_CALLNZ; */ + /* case TGSI_OPCODE_IFC: return OPCODE_IFC; */ + /* case TGSI_OPCODE_BREAKC: return OPCODE_BREAKC; */ + case TGSI_OPCODE_KIL: return OPCODE_KIL; + case TGSI_OPCODE_END: return OPCODE_END; + case TGSI_OPCODE_SWZ: return OPCODE_SWZ; + } + + fprintf(stderr, "Unknown opcode: %i\n", opcode); + abort(); +} + +static unsigned translate_saturate(unsigned saturate) +{ + switch(saturate) { + case TGSI_SAT_NONE: return SATURATE_OFF; + case TGSI_SAT_ZERO_ONE: return SATURATE_ZERO_ONE; + case TGSI_SAT_MINUS_PLUS_ONE: return SATURATE_PLUS_MINUS_ONE; + } + + fprintf(stderr, "Unknown saturate mode: %i\n", saturate); + abort(); +} + +static unsigned translate_register_file(unsigned file) +{ + switch(file) { + case TGSI_FILE_CONSTANT: return PROGRAM_CONSTANT; + case TGSI_FILE_IMMEDIATE: return PROGRAM_CONSTANT; + case TGSI_FILE_INPUT: return PROGRAM_INPUT; + case TGSI_FILE_OUTPUT: return PROGRAM_OUTPUT; + case TGSI_FILE_TEMPORARY: return PROGRAM_TEMPORARY; + case TGSI_FILE_ADDRESS: return PROGRAM_ADDRESS; + } + + fprintf(stderr, "Unhandled register file: %i\n", file); + abort(); +} + +static int translate_register_index( + struct tgsi_to_rc * ttr, + unsigned file, + int index) +{ + if (file == TGSI_FILE_IMMEDIATE) + return ttr->immediate_offset + index; + + return index; +} + +static void transform_dstreg( + struct tgsi_to_rc * ttr, + struct prog_dst_register * dst, + struct tgsi_full_dst_register * src) +{ + dst->File = translate_register_file(src->DstRegister.File); + dst->Index = translate_register_index(ttr, src->DstRegister.File, src->DstRegister.Index); + dst->WriteMask = src->DstRegister.WriteMask; + dst->RelAddr = src->DstRegister.Indirect; +} + +static void transform_srcreg( + struct tgsi_to_rc * ttr, + struct prog_src_register * dst, + struct tgsi_full_src_register * src) +{ + dst->File = translate_register_file(src->SrcRegister.File); + dst->Index = translate_register_index(ttr, src->SrcRegister.File, src->SrcRegister.Index); + dst->RelAddr = src->SrcRegister.Indirect; + dst->Swizzle = tgsi_util_get_full_src_register_extswizzle(src, 0); + dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 1) << 3; + dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 2) << 6; + dst->Swizzle |= tgsi_util_get_full_src_register_extswizzle(src, 3) << 9; + dst->Abs = src->SrcRegisterExtMod.Absolute; + dst->Negate = + src->SrcRegisterExtSwz.NegateX | + (src->SrcRegisterExtSwz.NegateY << 1) | + (src->SrcRegisterExtSwz.NegateZ << 2) | + (src->SrcRegisterExtSwz.NegateW << 3); + dst->Negate ^= src->SrcRegister.Negate ? NEGATE_XYZW : 0; +} + +static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_ext_texture src) +{ + switch(src.Texture) { + case TGSI_TEXTURE_1D: + dst->I.TexSrcTarget = TEXTURE_1D_INDEX; + break; + case TGSI_TEXTURE_2D: + dst->I.TexSrcTarget = TEXTURE_2D_INDEX; + break; + case TGSI_TEXTURE_3D: + dst->I.TexSrcTarget = TEXTURE_3D_INDEX; + break; + case TGSI_TEXTURE_CUBE: + dst->I.TexSrcTarget = TEXTURE_CUBE_INDEX; + break; + case TGSI_TEXTURE_RECT: + dst->I.TexSrcTarget = TEXTURE_RECT_INDEX; + break; + case TGSI_TEXTURE_SHADOW1D: + dst->I.TexSrcTarget = TEXTURE_1D_INDEX; + dst->I.TexShadow = 1; + break; + case TGSI_TEXTURE_SHADOW2D: + dst->I.TexSrcTarget = TEXTURE_2D_INDEX; + dst->I.TexShadow = 1; + break; + case TGSI_TEXTURE_SHADOWRECT: + dst->I.TexSrcTarget = TEXTURE_RECT_INDEX; + dst->I.TexShadow = 1; + break; + } +} + +static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_instruction * src) +{ + if (src->Instruction.Opcode == TGSI_OPCODE_END) + return; + + struct rc_instruction * dst = rc_insert_new_instruction(ttr->compiler, ttr->compiler->Program.Instructions.Prev); + int i; + + dst->I.Opcode = translate_opcode(src->Instruction.Opcode); + dst->I.SaturateMode = translate_saturate(src->Instruction.Saturate); + + if (src->Instruction.NumDstRegs) + transform_dstreg(ttr, &dst->I.DstReg, &src->FullDstRegisters[0]); + + for(i = 0; i < src->Instruction.NumSrcRegs; ++i) { + if (src->FullSrcRegisters[i].SrcRegister.File == TGSI_FILE_SAMPLER) + dst->I.TexSrcUnit = src->FullSrcRegisters[i].SrcRegister.Index; + else + transform_srcreg(ttr, &dst->I.SrcReg[i], &src->FullSrcRegisters[i]); + } + + /* Texturing. */ + transform_texture(dst, src->InstructionExtTexture); +} + +static void handle_immediate(struct tgsi_to_rc * ttr, struct tgsi_full_immediate * imm) +{ + struct rc_constant constant; + int i; + + constant.Type = RC_CONSTANT_IMMEDIATE; + constant.Size = 4; + for(i = 0; i < 4; ++i) + constant.u.Immediate[i] = imm->u[i].Float; + rc_constants_add(&ttr->compiler->Program.Constants, &constant); +} + +void r300_tgsi_to_rc(struct tgsi_to_rc * ttr, const struct tgsi_token * tokens) +{ + struct tgsi_parse_context parser; + int i; + + /* Allocate constants placeholders. + * + * Note: What if declared constants are not contiguous? */ + for(i = 0; i <= ttr->info->file_max[TGSI_FILE_CONSTANT]; ++i) { + struct rc_constant constant; + memset(&constant, 0, sizeof(constant)); + constant.Type = RC_CONSTANT_EXTERNAL; + constant.Size = 4; + constant.u.External = i; + rc_constants_add(&ttr->compiler->Program.Constants, &constant); + } + + ttr->immediate_offset = ttr->compiler->Program.Constants.Count; + + tgsi_parse_init(&parser, tokens); + + while (!tgsi_parse_end_of_tokens(&parser)) { + tgsi_parse_token(&parser); + + switch (parser.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + handle_immediate(ttr, &parser.FullToken.FullImmediate); + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + transform_instruction(ttr, &parser.FullToken.FullInstruction); + break; + } + } + + tgsi_parse_free(&parser); + + rc_calculate_inputs_outputs(ttr->compiler); +} + diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.h b/src/gallium/drivers/r300/r300_tgsi_to_rc.h new file mode 100644 index 00000000000..93e90ec6d2c --- /dev/null +++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.h @@ -0,0 +1,41 @@ +/* + * Copyright 2009 Nicolai Hähnle <nhaehnle@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_TGSI_TO_RC_H +#define R300_TGSI_TO_RC_H + +struct radeon_compiler; + +struct tgsi_full_declaration; +struct tgsi_shader_info; +struct tgsi_token; + +struct tgsi_to_rc { + struct radeon_compiler * compiler; + const struct tgsi_shader_info * info; + + int immediate_offset; +}; + +void r300_tgsi_to_rc(struct tgsi_to_rc * ttr, const struct tgsi_token * tokens); + +#endif /* R300_TGSI_TO_RC_H */ diff --git a/src/gallium/drivers/r300/r300_state_tcl.h b/src/gallium/drivers/r300/r300_vs.c index 2c8b586c2f5..2cb903bba2f 100644 --- a/src/gallium/drivers/r300/r300_state_tcl.h +++ b/src/gallium/drivers/r300/r300_vs.c @@ -20,15 +20,133 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#ifndef R300_STATE_TCL_H -#define R300_STATE_TCL_H +#include "r300_vs.h" +#include "r300_context.h" +#include "r300_tgsi_to_rc.h" + +#include "tgsi/tgsi_dump.h" #include "tgsi/tgsi_parse.h" -#include "r300_context.h" -#include "r300_debug.h" -#include "r300_reg.h" -#include "r300_screen.h" +#include "radeon_compiler.h" + + +static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) +{ + struct r300_vertex_shader * vs = c->UserData; + struct tgsi_shader_info* info = &vs->info; + boolean pointsize = false; + int out_colors = 0; + int colors = 0; + int out_generic = 0; + int generic = 0; + int i; + + /* Fill in the input mapping */ + for (i = 0; i < info->num_inputs; i++) + c->code->inputs[i] = i; + + /* Fill in the output mapping */ + for (i = 0; i < info->num_outputs; i++) { + switch (info->output_semantic_name[i]) { + case TGSI_SEMANTIC_PSIZE: + pointsize = true; + break; + case TGSI_SEMANTIC_COLOR: + out_colors++; + break; + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + out_generic++; + break; + } + } + + struct tgsi_parse_context parser; + + tgsi_parse_init(&parser, vs->state.tokens); + + while (!tgsi_parse_end_of_tokens(&parser)) { + tgsi_parse_token(&parser); + + if (parser.FullToken.Token.Type != TGSI_TOKEN_TYPE_DECLARATION) + continue; + + struct tgsi_full_declaration * decl = &parser.FullToken.FullDeclaration; + + if (decl->Declaration.File != TGSI_FILE_OUTPUT) + continue; + + switch (decl->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: + c->code->outputs[decl->DeclarationRange.First] = 0; + break; + case TGSI_SEMANTIC_PSIZE: + c->code->outputs[decl->DeclarationRange.First] = 1; + break; + case TGSI_SEMANTIC_COLOR: + c->code->outputs[decl->DeclarationRange.First] = 1 + + (pointsize ? 1 : 0) + + colors++; + break; + case TGSI_SEMANTIC_FOG: + case TGSI_SEMANTIC_GENERIC: + c->code->outputs[decl->DeclarationRange.First] = 1 + + (pointsize ? 1 : 0) + + out_colors + + generic++; + break; + default: + debug_printf("r300: vs: Bad semantic declaration %d\n", + decl->Semantic.SemanticName); + break; + } + } + + tgsi_parse_free(&parser); +} + + +void r300_translate_vertex_shader(struct r300_context* r300, + struct r300_vertex_shader* vs) +{ + struct r300_vertex_program_compiler compiler; + struct tgsi_to_rc ttr; + + /* Setup the compiler */ + rc_init(&compiler.Base); + + compiler.Base.Debug = 1; + compiler.code = &vs->code; + compiler.UserData = vs; + + if (compiler.Base.Debug) { + debug_printf("r300: Initial vertex program\n"); + tgsi_dump(vs->state.tokens, 0); + } + + /* Translate TGSI to our internal representation */ + ttr.compiler = &compiler.Base; + ttr.info = &vs->info; + + r300_tgsi_to_rc(&ttr, vs->state.tokens); + + compiler.RequiredOutputs = ~(~0 << vs->info.num_outputs); + compiler.SetHwInputOutput = &set_vertex_inputs_outputs; + + /* Invoke the compiler */ + r3xx_compile_vertex_program(&compiler); + if (compiler.Base.Error) { + /* Todo: Fail gracefully */ + fprintf(stderr, "r300 VP: Compiler error\n"); + abort(); + } + + /* And, finally... */ + rc_destroy(&compiler.Base); + vs->translated = TRUE; +} + /* XXX get these to r300_reg */ #define R300_PVS_DST_OPCODE(x) ((x) << 0) @@ -84,81 +202,33 @@ (R300_PVS_MODIFIER_X | R300_PVS_MODIFIER_Y | \ R300_PVS_MODIFIER_Z | R300_PVS_MODIFIER_W) -static const struct tgsi_full_src_register r300_constant_zero = { - .SrcRegister.Extended = TRUE, - .SrcRegister.File = TGSI_FILE_NULL, - .SrcRegisterExtSwz.ExtSwizzleX = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleY = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleZ = TGSI_EXTSWIZZLE_ZERO, - .SrcRegisterExtSwz.ExtSwizzleW = TGSI_EXTSWIZZLE_ZERO, -}; +struct r300_vertex_program_code r300_passthrough_vertex_shader = { + .length = 8, /* two instructions */ -/* Temporary struct used to hold assembly state while putting together - * fragment programs. */ -struct r300_vs_asm { - /* Pipe context. */ - struct r300_context* r300; - /* Number of colors. */ - unsigned color_count; - /* Number of texcoords. */ - unsigned tex_count; - /* Number of requested temporary registers. */ - unsigned temp_count; - /* Offset for immediate constants. Neither R300 nor R500 can do four - * inline constants per source, so instead we copy immediates into the - * constant buffer. */ - unsigned imm_offset; - /* Number of immediate constants. */ - unsigned imm_count; - /* Number of colors to write. */ - unsigned out_colors; - /* Number of texcoords to write. */ - unsigned out_texcoords; - /* Whether to emit point size. */ - boolean point_size; - /* Tab of declared outputs to OVM outputs. */ - unsigned tab[16]; -}; - -static struct r300_vertex_shader r300_passthrough_vertex_shader = { - /* XXX translate these back into normal instructions */ - .instruction_count = 2, - .instructions[0].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | + /* MOV out[0], in[0] */ + .body.d[0] = R300_PVS_DST_OPCODE(R300_VE_ADD) | R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | R300_PVS_DST_OFFSET(0) | R300_PVS_DST_WE_XYZW, - .instructions[0].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | + .body.d[1] = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | R300_PVS_SRC_OFFSET(0) | R300_PVS_SRC_SWIZZLE_XYZW, - .instructions[0].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, - .instructions[0].inst3 = 0x0, - .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | - R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | - R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW, - .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | - R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW, - .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, - .instructions[1].inst3 = 0x0, -}; + .body.d[2] = R300_PVS_SRC_SWIZZLE_ZERO, + .body.d[3] = 0x0, -static struct r300_vertex_shader r300_texture_vertex_shader = { - /* XXX translate these back into normal instructions */ - .instruction_count = 2, - .instructions[0].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | - R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | - R300_PVS_DST_OFFSET(0) | R300_PVS_DST_WE_XYZW, - .instructions[0].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | - R300_PVS_SRC_OFFSET(0) | R300_PVS_SRC_SWIZZLE_XYZW, - .instructions[0].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, - .instructions[0].inst3 = 0x0, - .instructions[1].inst0 = R300_PVS_DST_OPCODE(R300_VE_ADD) | + /* MOV out[1], in[1] */ + .body.d[4] = R300_PVS_DST_OPCODE(R300_VE_ADD) | R300_PVS_DST_REG_TYPE(R300_PVS_DST_REG_OUT) | R300_PVS_DST_OFFSET(1) | R300_PVS_DST_WE_XYZW, - .instructions[1].inst1 = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | + .body.d[5] = R300_PVS_SRC_REG_TYPE(R300_PVS_SRC_REG_INPUT) | R300_PVS_SRC_OFFSET(1) | R300_PVS_SRC_SWIZZLE_XYZW, - .instructions[1].inst2 = R300_PVS_SRC_SWIZZLE_ZERO, - .instructions[1].inst3 = 0x0, -}; + .body.d[6] = R300_PVS_SRC_SWIZZLE_ZERO, + .body.d[7] = 0x0, -void r300_translate_vertex_shader(struct r300_context* r300, - struct r300_vertex_shader* vs); + .inputs[0] = 0, + .inputs[1] = 1, + .outputs[0] = 0, + .outputs[1] = 1, + + .InputsRead = 3, + .OutputsWritten = 3 +}; -#endif /* R300_STATE_TCL_H */ diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h new file mode 100644 index 00000000000..2a4ce315e32 --- /dev/null +++ b/src/gallium/drivers/r300/r300_vs.h @@ -0,0 +1,54 @@ +/* + * Copyright 2009 Corbin Simpson <MostAwesomeDude@gmail.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R300_VS_H +#define R300_VS_H + +#include "pipe/p_state.h" +#include "tgsi/tgsi_scan.h" + +#include "radeon_code.h" + +struct r300_context; + +struct r300_vertex_shader { + /* Parent class */ + struct pipe_shader_state state; + struct tgsi_shader_info info; + + /* Fallback shader, because Draw has issues */ + struct draw_vertex_shader* draw; + + /* Has this shader been translated yet? */ + boolean translated; + + /* Machine code (if translated) */ + struct r300_vertex_program_code code; +}; + + +extern struct r300_vertex_program_code r300_passthrough_vertex_shader; + +void r300_translate_vertex_shader(struct r300_context* r300, + struct r300_vertex_shader* vs); + +#endif /* R300_VS_H */ diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h index d2893c3b9d7..f18ad75a47d 100644 --- a/src/gallium/drivers/r300/r300_winsys.h +++ b/src/gallium/drivers/r300/r300_winsys.h @@ -45,6 +45,9 @@ struct r300_winsys { /* PCI ID */ uint32_t pci_id; + /* GB pipe count */ + uint32_t gb_pipes; + /* GART size. */ uint32_t gart_size; diff --git a/src/gallium/drivers/r300/r3xx_fs.c b/src/gallium/drivers/r300/r3xx_fs.c new file mode 100644 index 00000000000..c1c1194d58e --- /dev/null +++ b/src/gallium/drivers/r300/r3xx_fs.c @@ -0,0 +1,74 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r3xx_fs.h" + +#include "r300_reg.h" + +struct rX00_fragment_program_code r3xx_passthrough_fragment_shader = { + .code.r300.alu.length = 1, + .code.r300.tex.length = 0, + + .code.r300.config = 0, + .code.r300.pixsize = 0, + .code.r300.code_offset = 0, + .code.r300.code_addr[3] = R300_RGBA_OUT, + + .code.r300.alu.inst[0].rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | + R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) | + R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | + R300_ALU_OUTC_CMP, + .code.r300.alu.inst[0].rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | + R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ, + .code.r300.alu.inst[0].alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | + R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) | + R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | + R300_ALU_OUTA_CMP, + .code.r300.alu.inst[0].alpha_addr = R300_ALPHA_ADDR0(0) | + R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT, +}; + +struct rX00_fragment_program_code r3xx_texture_fragment_shader = { + .code.r300.alu.length = 1, + .code.r300.tex.length = 1, + + .code.r300.config = R300_PFS_CNTL_FIRST_NODE_HAS_TEX, + .code.r300.pixsize = 0, + .code.r300.code_offset = 0, + .code.r300.code_addr[3] = R300_RGBA_OUT, + + .code.r300.tex.inst[0] = R300_TEX_OP_LD << R300_TEX_INST_SHIFT, + + .code.r300.alu.inst[0].rgb_inst = R300_RGB_SWIZA(R300_ALU_ARGC_SRC0C_XYZ) | + R300_RGB_SWIZB(R300_ALU_ARGC_SRC0C_XYZ) | + R300_RGB_SWIZC(R300_ALU_ARGC_ZERO) | + R300_ALU_OUTC_CMP, + .code.r300.alu.inst[0].rgb_addr = R300_RGB_ADDR0(0) | R300_RGB_ADDR1(0) | + R300_RGB_ADDR2(0) | R300_ALU_DSTC_OUTPUT_XYZ, + .code.r300.alu.inst[0].alpha_inst = R300_ALPHA_SWIZA(R300_ALU_ARGA_SRC0A) | + R300_ALPHA_SWIZB(R300_ALU_ARGA_SRC0A) | + R300_ALPHA_SWIZC(R300_ALU_ARGA_ZERO) | + R300_ALU_OUTA_CMP, + .code.r300.alu.inst[0].alpha_addr = R300_ALPHA_ADDR0(0) | + R300_ALPHA_ADDR1(0) | R300_ALPHA_ADDR2(0) | R300_ALU_DSTA_OUTPUT, +}; diff --git a/src/gallium/drivers/r300/r3xx_fs.h b/src/gallium/drivers/r300/r3xx_fs.h new file mode 100644 index 00000000000..51cd245724d --- /dev/null +++ b/src/gallium/drivers/r300/r3xx_fs.h @@ -0,0 +1,32 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R3XX_FS_H +#define R3XX_FS_H + +#include "radeon_code.h" + +struct rX00_fragment_program_code r3xx_passthrough_fragment_shader; +struct rX00_fragment_program_code r3xx_texture_fragment_shader; + +#endif /* R3XX_FS_H */ diff --git a/src/gallium/drivers/r300/r5xx_fs.c b/src/gallium/drivers/r300/r5xx_fs.c new file mode 100644 index 00000000000..f072deab0d9 --- /dev/null +++ b/src/gallium/drivers/r300/r5xx_fs.c @@ -0,0 +1,125 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "r5xx_fs.h" + +#include "r300_reg.h" + +/* XXX this all should find its way back to r300_reg */ +/* Swizzle tools */ +#define R500_SWIZZLE_ZERO 4 +#define R500_SWIZZLE_HALF 5 +#define R500_SWIZZLE_ONE 6 +#define R500_SWIZ_RGB_ZERO ((4 << 0) | (4 << 3) | (4 << 6)) +#define R500_SWIZ_RGB_ONE ((6 << 0) | (6 << 3) | (6 << 6)) +#define R500_SWIZ_RGB_RGB ((0 << 0) | (1 << 3) | (2 << 6)) +#define R500_SWIZ_MOD_NEG 1 +#define R500_SWIZ_MOD_ABS 2 +#define R500_SWIZ_MOD_NEG_ABS 3 +/* Swizzles for inst2 */ +#define R500_SWIZ_TEX_STRQ(x) ((x) << 8) +#define R500_SWIZ_TEX_RGBA(x) ((x) << 24) +/* Swizzles for inst3 */ +#define R500_SWIZ_RGB_A(x) ((x) << 2) +#define R500_SWIZ_RGB_B(x) ((x) << 15) +/* Swizzles for inst4 */ +#define R500_SWIZ_ALPHA_A(x) ((x) << 14) +#define R500_SWIZ_ALPHA_B(x) ((x) << 21) +/* Swizzle for inst5 */ +#define R500_SWIZ_RGBA_C(x) ((x) << 14) +#define R500_SWIZ_ALPHA_C(x) ((x) << 27) +/* Writemasks */ +#define R500_TEX_WMASK(x) ((x) << 11) +#define R500_ALU_WMASK(x) ((x) << 11) +#define R500_ALU_OMASK(x) ((x) << 15) +#define R500_W_OMASK (1 << 31) + +struct rX00_fragment_program_code r5xx_passthrough_fragment_shader = { + .code.r500.max_temp_idx = 0, + .code.r500.inst_end = 0, + + .code.r500.inst[0].inst0 = R500_INST_TYPE_OUT | + R500_INST_TEX_SEM_WAIT | R500_INST_LAST | + R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | + R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, + .code.r500.inst[0].inst1 = + R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | + R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST, + .code.r500.inst[0].inst2 = + R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | + R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST, + .code.r500.inst[0].inst3 = + R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | + R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | + R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | + R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B, + .code.r500.inst[0].inst4 = + R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A, + .code.r500.inst[0].inst5 = + R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | + R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | + R500_ALU_RGBA_A_SWIZ_0, +}; + +struct rX00_fragment_program_code r5xx_texture_fragment_shader = { + .code.r500.max_temp_idx = 0, + .code.r500.inst_end = 1, + + .code.r500.inst[0].inst0 = R500_INST_TYPE_TEX | + R500_INST_TEX_SEM_WAIT | + R500_INST_RGB_WMASK_RGB | R500_INST_ALPHA_WMASK | + R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, + .code.r500.inst[0].inst1 = R500_TEX_ID(0) | R500_TEX_INST_LD | + R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED, + .code.r500.inst[0].inst2 = R500_TEX_SRC_ADDR(0) | + R500_TEX_SRC_S_SWIZ_R | R500_TEX_SRC_T_SWIZ_G | + R500_TEX_SRC_R_SWIZ_B | R500_TEX_SRC_Q_SWIZ_A | + R500_TEX_DST_ADDR(0) | + R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | + R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A, + .code.r500.inst[0].inst3 = 0x0, + .code.r500.inst[0].inst4 = 0x0, + .code.r500.inst[0].inst5 = 0x0, + + .code.r500.inst[1].inst0 = R500_INST_TYPE_OUT | + R500_INST_TEX_SEM_WAIT | R500_INST_LAST | + R500_INST_RGB_OMASK_RGB | R500_INST_ALPHA_OMASK | + R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP, + .code.r500.inst[1].inst1 = + R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST | + R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST, + .code.r500.inst[1].inst2 = + R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST | + R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST, + .code.r500.inst[1].inst3 = + R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R | + R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B | + R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R | + R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B, + .code.r500.inst[1].inst4 = + R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A, + .code.r500.inst[1].inst5 = + R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 | + R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 | + R500_ALU_RGBA_A_SWIZ_0, +}; diff --git a/src/gallium/drivers/r300/r5xx_fs.h b/src/gallium/drivers/r300/r5xx_fs.h new file mode 100644 index 00000000000..a4addde32b2 --- /dev/null +++ b/src/gallium/drivers/r300/r5xx_fs.h @@ -0,0 +1,32 @@ +/* + * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com> + * Joakim Sindholt <opensource@zhasha.com> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#ifndef R5XX_FS_H +#define R5XX_FS_H + +#include "radeon_code.h" + +struct rX00_fragment_program_code r5xx_passthrough_fragment_shader; +struct rX00_fragment_program_code r5xx_texture_fragment_shader; + +#endif /* R5XX_FS_H */ diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 62e8d99cfd0..86df320ea8a 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -126,6 +126,22 @@ softpipe_is_texture_referenced( struct pipe_context *pipe, struct pipe_texture *texture, unsigned face, unsigned level) { + struct softpipe_context *softpipe = softpipe_context( pipe ); + unsigned i; + + if(softpipe->dirty_render_cache) { + for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) { + if(softpipe->framebuffer.cbufs[i] && + softpipe->framebuffer.cbufs[i]->texture == texture) + return PIPE_REFERENCED_FOR_WRITE; + } + if(softpipe->framebuffer.zsbuf && + softpipe->framebuffer.zsbuf->texture == texture) + return PIPE_REFERENCED_FOR_WRITE; + } + + /* FIXME: we also need to do the same for the texture cache */ + return PIPE_UNREFERENCED; } diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index b89a7292e5b..7888c2f644b 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -145,6 +145,8 @@ struct softpipe_context { struct draw_stage *vbuf; struct softpipe_vbuf_render *vbuf_render; + boolean dirty_render_cache; + struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; struct softpipe_tile_cache *zsbuf_cache; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index f117096bf73..d4045816d03 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -35,6 +35,7 @@ #include "pipe/p_context.h" #include "pipe/internal/p_winsys_screen.h" #include "pipe/p_inlines.h" +#include "util/u_prim.h" #include "sp_context.h" #include "sp_state.h" @@ -65,6 +66,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) size); } + static void softpipe_unmap_constant_buffers(struct softpipe_context *sp) { @@ -86,20 +88,6 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) } -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - - boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) @@ -108,15 +96,11 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, } - /** * Draw vertex arrays, with optional indexing. * Basically, map the vertex buffers (and drawing surfaces), then hand off * the drawing to the 'draw' module. - * - * XXX should the element buffer be specified/bound with a separate function? */ - boolean softpipe_draw_range_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, @@ -129,7 +113,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe, struct draw_context *draw = sp->draw; unsigned i; - sp->reduced_api_prim = reduced_prim[mode]; + sp->reduced_api_prim = u_reduced_prim(mode); if (sp->dirty) softpipe_update_derived( sp ); @@ -147,6 +131,7 @@ softpipe_draw_range_elements(struct pipe_context *pipe, PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } + /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes @@ -159,10 +144,10 @@ softpipe_draw_range_elements(struct pipe_context *pipe, } else { /* no index/element buffer */ - draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL); + draw_set_mapped_element_buffer_range(draw, 0, start, + start + count - 1, NULL); } - /* draw! */ draw_arrays(draw, mode, start, count); @@ -182,9 +167,12 @@ softpipe_draw_range_elements(struct pipe_context *pipe, /* Note: leave drawing surfaces mapped */ softpipe_unmap_constant_buffers(sp); + sp->dirty_render_cache = TRUE; + return TRUE; } + boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, @@ -198,11 +186,9 @@ softpipe_draw_elements(struct pipe_context *pipe, } - void softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) { struct softpipe_context *sp = softpipe_context(pipe); draw_set_edgeflags(sp->draw, edgeflags); } - diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index 035f4b963eb..4a14d49686e 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -71,6 +71,8 @@ softpipe_flush( struct pipe_context *pipe, * to unmap surfaces when flushing. */ softpipe_unmap_transfers(softpipe); + + softpipe->dirty_render_cache = FALSE; } /* Enable to dump BMPs of the color/depth buffers each frame */ diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 31c3ca21c51..f4fa0905d74 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -45,17 +45,6 @@ #include "rtasm/rtasm_x86sse.h" -/* Surely this should be defined somewhere in a tgsi header: - */ -typedef void (PIPE_CDECL *codegen_function)( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - const float (*constant)[4], - struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef, - float (*immediates)[4] - //, const struct tgsi_exec_vector *quadPos - ); /** @@ -65,7 +54,7 @@ struct sp_sse_fragment_shader { struct sp_fragment_shader base; struct x86_function sse2_program; - codegen_function func; + tgsi_sse2_fs_function func; float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; @@ -83,6 +72,7 @@ fs_sse_prepare( const struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, struct tgsi_sampler **samplers ) { + machine->Samplers = samplers; } @@ -107,12 +97,10 @@ fs_sse_run( const struct sp_fragment_shader *base, tgsi_set_kill_mask(machine, 0x0); tgsi_set_exec_mask(machine, 1, 1, 1, 1); - shader->func( machine->Inputs, - machine->Outputs, + shader->func( machine, machine->Consts, - machine->Temps, - machine->InterpCoefs, - shader->immediates + (const float (*)[4])shader->immediates, + machine->InterpCoefs // , &machine->QuadPos ); @@ -151,7 +139,7 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, return NULL; } - shader->func = (codegen_function) x86_get_func( &shader->sse2_program ); + shader->func = (tgsi_sse2_fs_function) x86_get_func( &shader->sse2_program ); if (!shader->func) { x86_release_func( &shader->sse2_program ); FREE(shader); diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 06725fd09b4..42021789ea8 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -238,65 +238,117 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) case PIPE_PRIM_TRIANGLES: for (i = 2; i < nr; i += 3) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-1], stride), - get_vert(vertex_buffer, indices[i-0], stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[i-2], stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } } break; case PIPE_PRIM_TRIANGLE_STRIP: for (i = 2; i < nr; i += 1) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i+(i&1)-2], stride), - get_vert(vertex_buffer, indices[i-(i&1)-1], stride), - get_vert(vertex_buffer, indices[i-0], stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i+(i&1)-1], stride), + get_vert(vertex_buffer, indices[i-(i&1)], stride), + get_vert(vertex_buffer, indices[i-2], stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i+(i&1)-2], stride), + get_vert(vertex_buffer, indices[i-(i&1)-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } } break; case PIPE_PRIM_TRIANGLE_FAN: for (i = 2; i < nr; i += 1) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[0], stride), - get_vert(vertex_buffer, indices[i-1], stride), - get_vert(vertex_buffer, indices[i-0], stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[0], stride), + get_vert(vertex_buffer, indices[i-1], stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[0], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } } break; case PIPE_PRIM_QUADS: for (i = 3; i < nr; i += 4) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-3], stride), - get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-0], stride)); - - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-1], stride), - get_vert(vertex_buffer, indices[i-0], stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-3], stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[i-3], stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } } break; case PIPE_PRIM_QUAD_STRIP: for (i = 3; i < nr; i += 2) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-3], stride), - get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-0], stride)); - - setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-1], stride), - get_vert(vertex_buffer, indices[i-3], stride), - get_vert(vertex_buffer, indices[i-0], stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-3], stride)); + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[i-3], stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } } break; case PIPE_PRIM_POLYGON: + /* Almost same as tri fan but the _first_ vertex specifies the flat + * shading color. Note that the first polygon vertex is passed as + * the last triangle vertex here. + * flatshade_first state makes no difference. + */ for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[0-1], stride), get_vert(vertex_buffer, indices[i-0], stride), - get_vert(vertex_buffer, indices[0], stride)); + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[0], stride) ); } break; @@ -368,58 +420,104 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) } break; - case PIPE_PRIM_TRIANGLES: for (i = 2; i < nr; i += 3) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-1, stride), - get_vert(vertex_buffer, i-0, stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, i-2, stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } } break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 2; i < nr; i += 1) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, i+(i&1)-2, stride), - get_vert(vertex_buffer, i-(i&1)-1, stride), - get_vert(vertex_buffer, i-0, stride)); + for (i = 2; i < nr; i++) { + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i+(i&1)-1, stride), + get_vert(vertex_buffer, i-(i&1), stride), + get_vert(vertex_buffer, i-2, stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i+(i&1)-2, stride), + get_vert(vertex_buffer, i-(i&1)-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } } break; case PIPE_PRIM_TRIANGLE_FAN: for (i = 2; i < nr; i += 1) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, 0, stride), - get_vert(vertex_buffer, i-1, stride), - get_vert(vertex_buffer, i-0, stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, 0, stride), + get_vert(vertex_buffer, i-1, stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, 0, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } } break; case PIPE_PRIM_QUADS: for (i = 3; i < nr; i += 4) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, i-3, stride), - get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-0, stride)); - - setup_tri( setup_ctx, - get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-1, stride), - get_vert(vertex_buffer, i-0, stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-3, stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, i-3, stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-0, stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } } break; + case PIPE_PRIM_QUAD_STRIP: for (i = 3; i < nr; i += 2) { - setup_tri( setup_ctx, - get_vert(vertex_buffer, i-3, stride), - get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-0, stride)); - - setup_tri( setup_ctx, - get_vert(vertex_buffer, i-1, stride), - get_vert(vertex_buffer, i-3, stride), - get_vert(vertex_buffer, i-0, stride)); + if (softpipe->rasterizer->flatshade_first) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-3, stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, i-3, stride) ); + } + else { + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-0, stride) ); + setup_tri( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-0, stride) ); + } } break; @@ -427,12 +525,13 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) /* Almost same as tri fan but the _first_ vertex specifies the flat * shading color. Note that the first polygon vertex is passed as * the last triangle vertex here. + * flatshade_first state makes no difference. */ for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, i-1, stride), get_vert(vertex_buffer, i-0, stride), - get_vert(vertex_buffer, 0, stride)); + get_vert(vertex_buffer, 0, stride) ); } break; diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index e134e443374..b1e18805c70 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -72,6 +72,24 @@ do { \ R[3] = A[3] - B[3]; \ } while (0) +/** Add and limit result to ceiling of 1.0 */ +#define VEC4_ADD_SAT(R, A, B) \ +do { \ + R[0] = A[0] + B[0]; if (R[0] > 1.0f) R[0] = 1.0f; \ + R[1] = A[1] + B[1]; if (R[1] > 1.0f) R[1] = 1.0f; \ + R[2] = A[2] + B[2]; if (R[2] > 1.0f) R[2] = 1.0f; \ + R[3] = A[3] + B[3]; if (R[3] > 1.0f) R[3] = 1.0f; \ +} while (0) + +/** Subtract and limit result to floor of 0.0 */ +#define VEC4_SUB_SAT(R, A, B) \ +do { \ + R[0] = A[0] - B[0]; if (R[0] < 0.0f) R[0] = 0.0f; \ + R[1] = A[1] - B[1]; if (R[1] < 0.0f) R[1] = 0.0f; \ + R[2] = A[2] - B[2]; if (R[2] < 0.0f) R[2] = 0.0f; \ + R[3] = A[3] - B[3]; if (R[3] < 0.0f) R[3] = 0.0f; \ +} while (0) + #define VEC4_MUL(R, A, B) \ do { \ R[0] = A[0] * B[0]; \ @@ -676,19 +694,19 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->rgb_func) { case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ - VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ - VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ + VEC4_ADD_SAT(quadColor[0], source[0], dest[0]); /* R */ + VEC4_ADD_SAT(quadColor[1], source[1], dest[1]); /* G */ + VEC4_ADD_SAT(quadColor[2], source[2], dest[2]); /* B */ break; case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ - VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ - VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ + VEC4_SUB_SAT(quadColor[0], source[0], dest[0]); /* R */ + VEC4_SUB_SAT(quadColor[1], source[1], dest[1]); /* G */ + VEC4_SUB_SAT(quadColor[2], source[2], dest[2]); /* B */ break; case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ - VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ - VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ + VEC4_SUB_SAT(quadColor[0], dest[0], source[0]); /* R */ + VEC4_SUB_SAT(quadColor[1], dest[1], source[1]); /* G */ + VEC4_SUB_SAT(quadColor[2], dest[2], source[2]); /* B */ break; case PIPE_BLEND_MIN: VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ @@ -709,13 +727,13 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->alpha_func) { case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ + VEC4_ADD_SAT(quadColor[3], source[3], dest[3]); /* A */ break; case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ + VEC4_SUB_SAT(quadColor[3], source[3], dest[3]); /* A */ break; case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ + VEC4_SUB_SAT(quadColor[3], dest[3], source[3]); /* A */ break; case PIPE_BLEND_MIN: VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index ca637a1d6a4..28f8d1a60ea 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -52,7 +52,7 @@ struct quad_shade_stage { struct quad_stage stage; /**< base class */ - struct tgsi_exec_machine machine; + struct tgsi_exec_machine *machine; struct tgsi_exec_vector *inputs, *outputs; }; @@ -73,7 +73,7 @@ shade_quad(struct quad_stage *qs, struct quad_header *quad) { struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; - struct tgsi_exec_machine *machine = &qss->machine; + struct tgsi_exec_machine *machine = qss->machine; boolean z_written; /* Consts do not require 16 byte alignment. */ @@ -146,7 +146,7 @@ shade_begin(struct quad_stage *qs) struct softpipe_context *softpipe = qs->softpipe; softpipe->fs->prepare( softpipe->fs, - &qss->machine, + qss->machine, (struct tgsi_sampler **) softpipe->tgsi.frag_samplers_list ); @@ -159,9 +159,8 @@ shade_destroy(struct quad_stage *qs) { struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; - tgsi_exec_machine_free_data(&qss->machine); - FREE( qss->inputs ); - FREE( qss->outputs ); + tgsi_exec_machine_destroy(qss->machine); + FREE( qs ); } @@ -170,19 +169,24 @@ struct quad_stage * sp_quad_shade_stage( struct softpipe_context *softpipe ) { struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); - - /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16); - qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16); - qss->machine.Inputs = align16(qss->inputs); - qss->machine.Outputs = align16(qss->outputs); + if (!qss) + goto fail; qss->stage.softpipe = softpipe; qss->stage.begin = shade_begin; qss->stage.run = shade_quad; qss->stage.destroy = shade_destroy; - tgsi_exec_machine_init( &qss->machine ); + qss->machine = tgsi_exec_machine_create(); + if (!qss->machine) + goto fail; return &qss->stage; + +fail: + if (qss && qss->machine) + tgsi_exec_machine_destroy(qss->machine); + + FREE(qss); + return NULL; } diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 692deeb8fd3..6178c4ac7e3 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -87,6 +87,10 @@ softpipe_get_param(struct pipe_screen *screen, int param) return 8; /* max 128x128x128 */ case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: return 13; /* max 4Kx4K */ + case PIPE_CAP_TGSI_CONT_SUPPORTED: + return 1; + case PIPE_CAP_BLEND_EQUATION_SEPARATE: + return 1; default: return 0; } diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index c6844a26498..de3ae3c3696 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -444,7 +444,8 @@ static void flush_spans( struct setup_context *setup ) mask |= MASK_TOP_RIGHT; if (x+1 >= xleft1 && x+1 < xright1) mask |= MASK_BOTTOM_RIGHT; - EMIT_QUAD( setup, x, setup->span.y, mask ); + if (mask) + EMIT_QUAD( setup, x, setup->span.y, mask ); } break; @@ -458,7 +459,8 @@ static void flush_spans( struct setup_context *setup ) mask |= MASK_TOP_LEFT; if (x+1 >= xleft0 && x+1 < xright0) mask |= MASK_TOP_RIGHT; - EMIT_QUAD( setup, x, setup->span.y, mask ); + if (mask) + EMIT_QUAD( setup, x, setup->span.y, mask ); } break; @@ -472,7 +474,8 @@ static void flush_spans( struct setup_context *setup ) mask |= MASK_BOTTOM_LEFT; if (x+1 >= xleft1 && x+1 < xright1) mask |= MASK_BOTTOM_RIGHT; - EMIT_QUAD( setup, x, setup->span.y, mask ); + if (mask) + EMIT_QUAD( setup, x, setup->span.y, mask ); } break; @@ -784,11 +787,10 @@ static void setup_tri_coefficients( struct setup_context *setup ) assert(0); } - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) { + setup->coef[fragSlot].a0[0] = 1.0f - setup->quad.input.facing; + setup->coef[fragSlot].dadx[0] = 0.0; + setup->coef[fragSlot].dady[0] = 0.0; } } } @@ -1051,7 +1053,10 @@ setup_line_coefficients(struct setup_context *setup, float area; /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = v1; + if (softpipe->rasterizer->flatshade_first) + setup->vprovoke = v0; + else + setup->vprovoke = v1; setup->vmin = v0; setup->vmax = v1; @@ -1095,11 +1100,10 @@ setup_line_coefficients(struct setup_context *setup, assert(0); } - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) { + setup->coef[fragSlot].a0[0] = 1.0f - setup->quad.input.facing; + setup->coef[fragSlot].dadx[0] = 0.0; + setup->coef[fragSlot].dady[0] = 0.0; } } return TRUE; @@ -1341,11 +1345,10 @@ setup_point( struct setup_context *setup, assert(0); } - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) { + setup->coef[fragSlot].a0[0] = 1.0f - setup->quad.input.facing; + setup->coef[fragSlot].dadx[0] = 0.0; + setup->coef[fragSlot].dady[0] = 0.0; } } diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 6b6a4c3ff34..75551000c9b 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -110,6 +110,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) break; case TGSI_SEMANTIC_GENERIC: + case TGSI_SEMANTIC_FACE: /* this includes texcoords and varying vars */ src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_GENERIC, spfs->info.input_semantic_index[i]); diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index adbd0cb7f0c..a1d3bade27a 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -705,15 +705,18 @@ get_texel(const struct tgsi_sampler *tgsi_sampler, * Compare texcoord 'p' (aka R) against texture value 'rgba[0]' * When we sampled the depth texture, the depth value was put into all * RGBA channels. We look at the red channel here. + * \param rgba quad of (depth) texel values + * \param p texture 'P' components for four pixels in quad + * \param j which pixel in the quad to test [0..3] */ static INLINE void -shadow_compare(uint compare_func, +shadow_compare(const struct pipe_sampler_state *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE], const float p[QUAD_SIZE], uint j) { int k; - switch (compare_func) { + switch (sampler->compare_func) { case PIPE_FUNC_LESS: k = p[j] < rgba[0][j]; break; @@ -744,7 +747,81 @@ shadow_compare(uint compare_func, break; } + /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */ rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; + rgba[3][j] = 1.0F; +} + + +/** + * As above, but do four z/texture comparisons. + */ +static INLINE void +shadow_compare4(const struct pipe_sampler_state *sampler, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const float p[QUAD_SIZE]) +{ + int j, k0, k1, k2, k3; + float val; + + /* compare four texcoords vs. four texture samples */ + switch (sampler->compare_func) { + case PIPE_FUNC_LESS: + k0 = p[0] < rgba[0][0]; + k1 = p[1] < rgba[0][1]; + k2 = p[2] < rgba[0][2]; + k3 = p[3] < rgba[0][3]; + break; + case PIPE_FUNC_LEQUAL: + k0 = p[0] <= rgba[0][0]; + k1 = p[1] <= rgba[0][1]; + k2 = p[2] <= rgba[0][2]; + k3 = p[3] <= rgba[0][3]; + break; + case PIPE_FUNC_GREATER: + k0 = p[0] > rgba[0][0]; + k1 = p[1] > rgba[0][1]; + k2 = p[2] > rgba[0][2]; + k3 = p[3] > rgba[0][3]; + break; + case PIPE_FUNC_GEQUAL: + k0 = p[0] >= rgba[0][0]; + k1 = p[1] >= rgba[0][1]; + k2 = p[2] >= rgba[0][2]; + k3 = p[3] >= rgba[0][3]; + break; + case PIPE_FUNC_EQUAL: + k0 = p[0] == rgba[0][0]; + k1 = p[1] == rgba[0][1]; + k2 = p[2] == rgba[0][2]; + k3 = p[3] == rgba[0][3]; + break; + case PIPE_FUNC_NOTEQUAL: + k0 = p[0] != rgba[0][0]; + k1 = p[1] != rgba[0][1]; + k2 = p[2] != rgba[0][2]; + k3 = p[3] != rgba[0][3]; + break; + case PIPE_FUNC_ALWAYS: + k0 = k1 = k2 = k3 = 1; + break; + case PIPE_FUNC_NEVER: + k0 = k1 = k2 = k3 = 0; + break; + default: + k0 = k1 = k2 = k3 = 0; + assert(0); + break; + } + + /* convert four pass/fail values to an intensity in [0,1] */ + val = 0.25F * (k0 + k1 + k2 + k3); + + /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */ + for (j = 0; j < 4; j++) { + rgba[0][j] = rgba[1][j] = rgba[2][j] = val; + rgba[3][j] = 1.0F; + } } @@ -767,7 +844,6 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, const uint unit = samp->unit; const struct pipe_texture *texture = sp->texture[unit]; const struct pipe_sampler_state *sampler = sp->sampler[unit]; - const uint compare_func = sampler->compare_func; unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; @@ -792,7 +868,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, for (j = 0; j < QUAD_SIZE; j++) { get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, rgba, p, j); + shadow_compare(sampler, rgba, p, j); } if (level0 != level1) { @@ -804,7 +880,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0, rgba2, j); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ - shadow_compare(compare_func, rgba2, p, j); + shadow_compare(sampler, rgba2, p, j); } for (c = 0; c < NUM_CHANNELS; c++) { @@ -831,10 +907,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2); get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); + shadow_compare4(sampler, tx, p); } /* interpolate R, G, B, A */ @@ -856,10 +929,7 @@ sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler, get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2); get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); + shadow_compare4(sampler, tx, p); } /* interpolate R, G, B, A */ @@ -1074,7 +1144,6 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler, const struct pipe_texture *texture = sp->texture[unit]; const struct pipe_sampler_state *sampler = sp->sampler[unit]; const uint face = 0; - const uint compare_func = sampler->compare_func; unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; @@ -1099,7 +1168,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler, for (j = 0; j < QUAD_SIZE; j++) { get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, rgba, p, j); + shadow_compare(sampler, rgba, p, j); } } } @@ -1119,10 +1188,7 @@ sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler, get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2); get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3); if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); + shadow_compare4(sampler, tx, p); } for (c = 0; c < 4; c++) { rgba[c][j] = lerp_2d(xw[j], yw[j], diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 7a533dad9f0..70f09324311 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -48,11 +48,6 @@ /* Simple, maximally packed layout. */ -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - /* Conventional allocation path for non-display textures: */ @@ -100,6 +95,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, { unsigned usage = (PIPE_BUFFER_USAGE_CPU_READ_WRITE | PIPE_BUFFER_USAGE_GPU_READ_WRITE); + unsigned tex_usage = spt->base.tex_usage; spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); @@ -109,6 +105,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, spt->base.height[0], spt->base.format, usage, + tex_usage, &spt->stride[0]); return spt->buffer != NULL; @@ -130,7 +127,8 @@ softpipe_texture_create(struct pipe_screen *screen, pipe_reference_init(&spt->base.reference, 1); spt->base.screen = screen; - if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + if (spt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_PRIMARY)) { if (!softpipe_displaytarget_layout(screen, spt)) goto fail; } @@ -224,12 +222,6 @@ softpipe_get_tex_surface(struct pipe_screen *screen, if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ) ps->usage |= PIPE_BUFFER_USAGE_CPU_READ; - if (ps->usage & (PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_GPU_WRITE)) { - /* Mark the surface as dirty. The tile cache will look for this. */ - spt->modified = TRUE; - } - ps->face = face; ps->level = level; ps->zslice = zslice; @@ -376,6 +368,11 @@ softpipe_transfer_unmap(struct pipe_screen *screen, spt = softpipe_texture(transfer->texture); pipe_buffer_unmap( screen, spt->buffer ); + + if (transfer->usage != PIPE_TRANSFER_READ) { + /* Mark the texture as dirty to expire the tile caches. */ + spt->modified = TRUE; + } } diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile index 4aeb8e3d7ed..dd6831c70ab 100644 --- a/src/gallium/drivers/trace/Makefile +++ b/src/gallium/drivers/trace/Makefile @@ -11,6 +11,7 @@ C_SOURCES = \ tr_screen.c \ tr_state.c \ tr_rbug.c \ + tr_drm.c \ tr_texture.c include ../../Makefile.template diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index e635fed77d8..c1675d1c165 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -7,6 +7,7 @@ trace = env.ConvenienceLibrary( source = [ 'tr_buffer.c', 'tr_context.c', + 'tr_drm.c', 'tr_dump.c', 'tr_dump_state.c', 'tr_screen.c', diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c new file mode 100644 index 00000000000..93c569c73a8 --- /dev/null +++ b/src/gallium/drivers/trace/tr_drm.c @@ -0,0 +1,211 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "state_tracker/drm_api.h" + +#include "util/u_memory.h" +#include "trace/tr_drm.h" +#include "trace/tr_screen.h" +#include "trace/tr_context.h" +#include "trace/tr_buffer.h" +#include "trace/tr_texture.h" + +struct trace_drm_api +{ + struct drm_api base; + + struct drm_api *api; +}; + +static INLINE struct trace_drm_api * +trace_drm_api(struct drm_api *_api) +{ + return (struct trace_drm_api *)_api; +} + +static struct pipe_screen * +trace_drm_create_screen(struct drm_api *_api, int fd, + struct drm_create_screen_arg *arg) +{ + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct drm_api *api = tr_api->api; + struct pipe_screen *screen; + + /* TODO trace call */ + + if (arg && arg->mode != DRM_CREATE_NORMAL) + return NULL; + + screen = api->create_screen(api, fd, arg); + + return trace_screen_create(screen); +} + +static struct pipe_context * +trace_drm_create_context(struct drm_api *_api, + struct pipe_screen *_screen) +{ + struct trace_screen *tr_screen = trace_screen(_screen); + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct pipe_screen *screen = tr_screen->screen; + struct drm_api *api = tr_api->api; + struct pipe_context *pipe; + + /* TODO trace call */ + + pipe = api->create_context(api, screen); + + pipe = trace_context_create(_screen, pipe); + + return pipe; +} + +static boolean +trace_drm_buffer_from_texture(struct drm_api *_api, + struct pipe_texture *_texture, + struct pipe_buffer **_buffer, + unsigned *stride) +{ + struct trace_texture *tr_texture = trace_texture(_texture); + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct pipe_texture *texture = tr_texture->texture; + struct drm_api *api = tr_api->api; + struct pipe_buffer *buffer = NULL; + boolean result; + + /* TODO trace call */ + + result = api->buffer_from_texture(api, texture, &buffer, stride); + + if (result && _buffer) + buffer = trace_buffer_create(trace_screen(_texture->screen), buffer); + + if (_buffer) + *_buffer = buffer; + else + pipe_buffer_reference(&buffer, NULL); + + return result; +} + +static struct pipe_buffer * +trace_drm_buffer_from_handle(struct drm_api *_api, + struct pipe_screen *_screen, + const char *name, + unsigned handle) +{ + struct trace_screen *tr_screen = trace_screen(_screen); + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct pipe_screen *screen = tr_screen->screen; + struct drm_api *api = tr_api->api; + struct pipe_buffer *result; + + /* TODO trace call */ + + result = api->buffer_from_handle(api, screen, name, handle); + + result = trace_buffer_create(trace_screen(_screen), result); + + return result; +} + +static boolean +trace_drm_handle_from_buffer(struct drm_api *_api, + struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct trace_screen *tr_screen = trace_screen(_screen); + struct trace_buffer *tr_buffer = trace_buffer(_buffer); + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct pipe_screen *screen = tr_screen->screen; + struct pipe_buffer *buffer = tr_buffer->buffer; + struct drm_api *api = tr_api->api; + + /* TODO trace call */ + + return api->handle_from_buffer(api, screen, buffer, handle); +} + +static boolean +trace_drm_global_handle_from_buffer(struct drm_api *_api, + struct pipe_screen *_screen, + struct pipe_buffer *_buffer, + unsigned *handle) +{ + struct trace_screen *tr_screen = trace_screen(_screen); + struct trace_buffer *tr_buffer = trace_buffer(_buffer); + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct pipe_screen *screen = tr_screen->screen; + struct pipe_buffer *buffer = tr_buffer->buffer; + struct drm_api *api = tr_api->api; + + /* TODO trace call */ + + return api->global_handle_from_buffer(api, screen, buffer, handle); +} + +static void +trace_drm_destroy(struct drm_api *_api) +{ + struct trace_drm_api *tr_api = trace_drm_api(_api); + struct drm_api *api = tr_api->api; + api->destroy(api); + + free(tr_api); +} + +struct drm_api * +trace_drm_create(struct drm_api *api) +{ + struct trace_drm_api *tr_api; + + if (!api) + goto error; + + if (!trace_enabled()) + goto error; + + tr_api = CALLOC_STRUCT(trace_drm_api); + + if (!tr_api) + goto error; + + tr_api->base.create_screen = trace_drm_create_screen; + tr_api->base.create_context = trace_drm_create_context; + tr_api->base.buffer_from_texture = trace_drm_buffer_from_texture; + tr_api->base.buffer_from_handle = trace_drm_buffer_from_handle; + tr_api->base.handle_from_buffer = trace_drm_handle_from_buffer; + tr_api->base.global_handle_from_buffer = trace_drm_global_handle_from_buffer; + tr_api->base.destroy = trace_drm_destroy; + tr_api->api = api; + + return &tr_api->base; + +error: + return api; +} diff --git a/src/gallium/drivers/trace/tr_drm.h b/src/gallium/drivers/trace/tr_drm.h new file mode 100644 index 00000000000..845c66a32aa --- /dev/null +++ b/src/gallium/drivers/trace/tr_drm.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_DRM_H +#define TR_DRM_H + +struct drm_api; + +struct drm_api* trace_drm_create(struct drm_api *api); + +#endif /* ID_DRM_H */ diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 3a1409e95a7..7e2ccbcfdc5 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -351,7 +351,7 @@ void trace_dump_call_begin_locked(const char *klass, const char *method) trace_dump_indent(1); trace_dump_writes("<call no=\'"); trace_dump_writef("%lu", call_no); - trace_dump_writes("\' class =\'"); + trace_dump_writes("\' class=\'"); trace_dump_escape(klass); trace_dump_writes("\' method=\'"); trace_dump_escape(method); @@ -451,11 +451,11 @@ void trace_dump_float(double value) } void trace_dump_bytes(const void *data, - long unsigned size) + size_t size) { static const char hex_table[16] = "0123456789ABCDEF"; const uint8_t *p = data; - long unsigned i; + size_t i; if (!dumping) return; diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 31ac70802f0..32592bab12f 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -91,7 +91,7 @@ void trace_dump_bool(int value); void trace_dump_int(long long int value); void trace_dump_uint(long long unsigned value); void trace_dump_float(double value); -void trace_dump_bytes(const void *data, long unsigned size); +void trace_dump_bytes(const void *data, size_t size); void trace_dump_string(const char *str); void trace_dump_enum(const char *value); void trace_dump_array_begin(void); @@ -130,7 +130,7 @@ void trace_dump_transfer_ptr(struct pipe_transfer *_transfer); #define trace_dump_array(_type, _obj, _size) \ do { \ - unsigned long idx; \ + size_t idx; \ trace_dump_array_begin(); \ for(idx = 0; idx < (_size); ++idx) { \ trace_dump_elem_begin(); \ @@ -142,7 +142,7 @@ void trace_dump_transfer_ptr(struct pipe_transfer *_transfer); #define trace_dump_struct_array(_type, _obj, _size) \ do { \ - unsigned long idx; \ + size_t idx; \ trace_dump_array_begin(); \ for(idx = 0; idx < (_size); ++idx) { \ trace_dump_elem_begin(); \ diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c index 23a2473b574..bcf6751af4f 100644 --- a/src/gallium/drivers/trace/tr_dump_state.c +++ b/src/gallium/drivers/trace/tr_dump_state.c @@ -36,12 +36,18 @@ void trace_dump_format(enum pipe_format format) { + if (!trace_dumping_enabled_locked()) + return; + trace_dump_enum(pf_name(format) ); } void trace_dump_block(const struct pipe_format_block *block) { + if (!trace_dumping_enabled_locked()) + return; + trace_dump_struct_begin("pipe_format_block"); trace_dump_member(uint, block, size); trace_dump_member(uint, block, width); @@ -52,6 +58,9 @@ void trace_dump_block(const struct pipe_format_block *block) static void trace_dump_reference(const struct pipe_reference *reference) { + if (!trace_dumping_enabled_locked()) + return; + trace_dump_struct_begin("pipe_reference"); trace_dump_member(int, &reference->count, count); trace_dump_struct_end(); @@ -60,6 +69,9 @@ static void trace_dump_reference(const struct pipe_reference *reference) void trace_dump_template(const struct pipe_texture *templat) { + if (!trace_dumping_enabled_locked()) + return; + if(!templat) { trace_dump_null(); return; @@ -95,6 +107,9 @@ void trace_dump_template(const struct pipe_texture *templat) void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -141,6 +156,9 @@ void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -160,6 +178,9 @@ void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) void trace_dump_viewport_state(const struct pipe_viewport_state *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -176,6 +197,9 @@ void trace_dump_viewport_state(const struct pipe_viewport_state *state) void trace_dump_scissor_state(const struct pipe_scissor_state *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -196,6 +220,9 @@ void trace_dump_clip_state(const struct pipe_clip_state *state) { unsigned i; + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -221,6 +248,9 @@ void trace_dump_clip_state(const struct pipe_clip_state *state) void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -238,6 +268,9 @@ void trace_dump_shader_state(const struct pipe_shader_state *state) { static char str[8192]; + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -259,6 +292,9 @@ void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_ { unsigned i; + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -307,6 +343,9 @@ void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_ void trace_dump_blend_state(const struct pipe_blend_state *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -336,6 +375,9 @@ void trace_dump_blend_state(const struct pipe_blend_state *state) void trace_dump_blend_color(const struct pipe_blend_color *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -351,6 +393,9 @@ void trace_dump_blend_color(const struct pipe_blend_color *state) void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) { + if (!trace_dumping_enabled_locked()) + return; + trace_dump_struct_begin("pipe_framebuffer_state"); trace_dump_member(uint, state, width); @@ -365,6 +410,9 @@ void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) void trace_dump_sampler_state(const struct pipe_sampler_state *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -382,7 +430,6 @@ void trace_dump_sampler_state(const struct pipe_sampler_state *state) trace_dump_member(uint, state, compare_func); trace_dump_member(bool, state, normalized_coords); trace_dump_member(uint, state, prefilter); - trace_dump_member(float, state, shadow_ambient); trace_dump_member(float, state, lod_bias); trace_dump_member(float, state, min_lod); trace_dump_member(float, state, max_lod); @@ -395,6 +442,9 @@ void trace_dump_sampler_state(const struct pipe_sampler_state *state) void trace_dump_surface(const struct pipe_surface *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -423,6 +473,9 @@ void trace_dump_surface(const struct pipe_surface *state) void trace_dump_transfer(const struct pipe_transfer *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -454,6 +507,9 @@ void trace_dump_transfer(const struct pipe_transfer *state) void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; @@ -472,6 +528,9 @@ void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) void trace_dump_vertex_element(const struct pipe_vertex_element *state) { + if (!trace_dumping_enabled_locked()) + return; + if(!state) { trace_dump_null(); return; diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 920f418ebf0..26f1c04594f 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -38,6 +38,7 @@ static boolean trace = FALSE; +static boolean rbug = FALSE; static const char * trace_screen_get_name(struct pipe_screen *_screen) @@ -461,6 +462,7 @@ trace_screen_surface_buffer_create(struct pipe_screen *_screen, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *pstride) { struct trace_screen *tr_scr = trace_screen(_screen); @@ -475,11 +477,13 @@ trace_screen_surface_buffer_create(struct pipe_screen *_screen, trace_dump_arg(uint, height); trace_dump_arg(format, format); trace_dump_arg(uint, usage); + trace_dump_arg(uint, tex_usage); result = screen->surface_buffer_create(screen, width, height, format, usage, + tex_usage, pstride); stride = *pstride; @@ -837,18 +841,11 @@ trace_screen_destroy(struct pipe_screen *_screen) boolean trace_enabled(void) { - return trace; -} - -struct pipe_screen * -trace_screen_create(struct pipe_screen *screen) -{ - struct trace_screen *tr_scr; - struct pipe_winsys *winsys; - boolean rbug = FALSE; + static boolean firstrun = TRUE; - if(!screen) - goto error1; + if (!firstrun) + return trace; + firstrun = FALSE; trace_dump_init(); @@ -862,7 +859,19 @@ trace_screen_create(struct pipe_screen *screen) rbug = TRUE; } - if (!trace) + return trace; +} + +struct pipe_screen * +trace_screen_create(struct pipe_screen *screen) +{ + struct trace_screen *tr_scr; + struct pipe_winsys *winsys; + + if(!screen) + goto error1; + + if (!trace_enabled()) goto error1; trace_dump_call_begin("", "pipe_screen_create"); diff --git a/src/gallium/include/pipe/internal/p_winsys_screen.h b/src/gallium/include/pipe/internal/p_winsys_screen.h index f4a29e63c7e..a1542dada70 100644 --- a/src/gallium/include/pipe/internal/p_winsys_screen.h +++ b/src/gallium/include/pipe/internal/p_winsys_screen.h @@ -140,6 +140,7 @@ struct pipe_winsys unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride); diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index 63238ea46e9..4152d6ac36a 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -102,6 +102,19 @@ /* + * Endian detection. + */ + +#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64) +#define PIPE_ARCH_LITTLE_ENDIAN +#elif defined(PIPE_ARCH_PPC) || defined(PIPE_ARCH_PPC_64) +#define PIPE_ARCH_BIG_ENDIAN +#else +#define PIPE_ARCH_UNKNOWN_ENDIAN +#endif + + +/* * Operating system family. * * See subsystem below for a more fine-grained distinction. @@ -119,6 +132,10 @@ #define PIPE_OS_SOLARIS #endif +#if defined(__APPLE__) +#define PIPE_OS_APPLE +#endif + #if defined(_WIN32) || defined(WIN32) #define PIPE_OS_WINDOWS #endif diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 57e966ac3b0..39620a71980 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -98,7 +98,7 @@ struct pipe_context { */ /*@{*/ struct pipe_query *(*create_query)( struct pipe_context *pipe, - unsigned query_type ); + unsigned query_type ); void (*destroy_query)(struct pipe_context *pipe, struct pipe_query *q); @@ -215,10 +215,12 @@ struct pipe_context { /** * Clear the specified set of currently bound buffers to specified values. + * The entire buffers are cleared (no scissor, no colormask, etc). * - * buffers is a bitfield of PIPE_CLEAR_* values. - * - * rgba is a pointer to an array of one float for each of r, g, b, a. + * \param buffers bitfield of PIPE_CLEAR_* values. + * \param rgba pointer to an array of one float for each of r, g, b, a. + * \param depth depth clear value in [0,1]. + * \param stencil stencil clear value */ void (*clear)(struct pipe_context *pipe, unsigned buffers, @@ -226,7 +228,9 @@ struct pipe_context { double depth, unsigned stencil); - /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */ + /** Flush rendering + * \param flags bitmask of PIPE_FLUSH_x tokens) + */ void (*flush)( struct pipe_context *pipe, unsigned flags, struct pipe_fence_handle **fence ); @@ -242,10 +246,10 @@ struct pipe_context { * \param texture texture to check. * \param face cubemap face. Use 0 for non-cubemap texture. */ - unsigned int (*is_texture_referenced) (struct pipe_context *pipe, struct pipe_texture *texture, unsigned face, unsigned level); + /** * Check whether a buffer is referenced by an unflushed hw command. * The state-tracker uses this function to optimize away unnecessary @@ -255,7 +259,6 @@ struct pipe_context { * checked. * \param buf Buffer to check. */ - unsigned int (*is_buffer_referenced) (struct pipe_context *pipe, struct pipe_buffer *buf); }; diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 9924046cb2f..b01ab6d137c 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -191,9 +191,9 @@ enum pipe_texture_target { * Transfer object usage flags */ enum pipe_transfer_usage { - PIPE_TRANSFER_READ, - PIPE_TRANSFER_WRITE, - PIPE_TRANSFER_READ_WRITE /**< Read/modify/write */ + PIPE_TRANSFER_READ = (1 << 0), + PIPE_TRANSFER_WRITE = (1 << 1), + PIPE_TRANSFER_READ_WRITE = PIPE_TRANSFER_READ | PIPE_TRANSFER_WRITE /**< Read/modify/write */ }; @@ -210,6 +210,7 @@ enum pipe_transfer_usage { #define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) #define PIPE_BUFFER_USAGE_DISCARD (1 << 8) #define PIPE_BUFFER_USAGE_DONTBLOCK (1 << 9) +#define PIPE_BUFFER_USAGE_FLUSH_EXPLICIT (1 << 10) /**< See pipe_screen::buffer_flush_mapped_range */ /** Pipe driver custom usage flags should be greater or equal to this value */ #define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) @@ -302,6 +303,8 @@ enum pipe_transfer_usage { #define PIPE_CAP_TEXTURE_MIRROR_CLAMP 24 #define PIPE_CAP_TEXTURE_MIRROR_REPEAT 25 #define PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS 26 +#define PIPE_CAP_TGSI_CONT_SUPPORTED 27 +#define PIPE_CAP_BLEND_EQUATION_SEPARATE 28 /** diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index a279eefef9e..c4469d4a9e9 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -536,14 +536,40 @@ pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned h return pf_get_nblocksx(block, width)*pf_get_nblocksy(block, height); } +static INLINE size_t +pf_get_stride(const struct pipe_format_block *block, unsigned width) +{ + return pf_get_nblocksx(block, width)*block->size; +} + +static INLINE size_t +pf_get_2d_size(const struct pipe_format_block *block, size_t stride, unsigned height) +{ + return pf_get_nblocksy(block, height)*stride; +} + static INLINE boolean -pf_is_depth_stencil( enum pipe_format format ) +pf_is_depth_or_stencil( enum pipe_format format ) { return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; } static INLINE boolean +pf_is_depth_and_stencil( enum pipe_format format ) +{ + return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) != 0 && + pf_get_component_bits( format, PIPE_FORMAT_COMP_S ) != 0); +} + +/** DEPRECATED: For backwards compatibility */ +static INLINE boolean +pf_is_depth_stencil( enum pipe_format format ) +{ + return pf_is_depth_or_stencil( format ); +} + +static INLINE boolean pf_is_compressed( enum pipe_format format ) { return pf_layout(format) == PIPE_FORMAT_LAYOUT_DXT ? TRUE : FALSE; diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 1232c879682..a5c1e8270a3 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -63,6 +63,13 @@ pipe_buffer_map(struct pipe_screen *screen, if(screen->buffer_map_range) { unsigned offset = 0; unsigned length = buf->size; + + /* XXX: Actually we should be using/detecting DISCARD + * instead of assuming that WRITE implies discard */ + if((usage & PIPE_BUFFER_USAGE_CPU_WRITE) && + !(usage & PIPE_BUFFER_USAGE_DISCARD)) + usage |= PIPE_BUFFER_USAGE_CPU_READ; + return screen->buffer_map_range(screen, buf, offset, length, usage); } else @@ -117,7 +124,9 @@ pipe_buffer_write(struct pipe_screen *screen, assert(offset + size <= buf->size); assert(size); - map = pipe_buffer_map_range(screen, buf, offset, size, PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe_buffer_map_range(screen, buf, offset, size, + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_FLUSH_EXPLICIT); assert(map); if(map) { memcpy(map + offset, data, size); diff --git a/src/gallium/include/pipe/p_refcnt.h b/src/gallium/include/pipe/p_refcnt.h index 1f89453e09a..1f9088b3e9c 100644 --- a/src/gallium/include/pipe/p_refcnt.h +++ b/src/gallium/include/pipe/p_refcnt.h @@ -62,29 +62,29 @@ pipe_is_referenced(struct pipe_reference *reference) * Set 'ptr' to point to 'reference' and update reference counting. * The old thing pointed to, if any, will be unreferenced first. * 'reference' may be NULL. - * - * XXX: thread safety issues! */ static INLINE bool pipe_reference(struct pipe_reference **ptr, struct pipe_reference *reference) { bool destroy = FALSE; - /* bump the reference.count first */ - if (reference) { - assert(pipe_is_referenced(reference)); - p_atomic_inc(&reference->count); - } - - if (*ptr) { - assert(pipe_is_referenced(*ptr)); - if (p_atomic_dec_zero(&(*ptr)->count)) { - destroy = TRUE; + if(*ptr != reference) { + /* bump the reference.count first */ + if (reference) { + assert(pipe_is_referenced(reference)); + p_atomic_inc(&reference->count); } + + if (*ptr) { + assert(pipe_is_referenced(*ptr)); + if (p_atomic_dec_zero(&(*ptr)->count)) { + destroy = TRUE; + } + } + + *ptr = reference; } - *ptr = reference; - return destroy; } diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index b449522fac3..3f30c52a169 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -194,6 +194,7 @@ struct pipe_screen { unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride); @@ -221,23 +222,31 @@ struct pipe_screen { /** * Notify a range that was actually written into. * + * Can only be used if the buffer was mapped with the + * PIPE_BUFFER_USAGE_CPU_WRITE and PIPE_BUFFER_USAGE_FLUSH_EXPLICIT flags + * set. + * * The range is relative to the buffer start, regardless of the range * specified to buffer_map_range. This is different from the * ARB_map_buffer_range semantics because we don't forbid multiple mappings * of the same buffer (yet). * - * If the buffer was mapped for writing and no buffer_flush_mapped_range - * call was done until the buffer_unmap is called then the pipe driver will - * assumed that the whole buffer was written. This is for backward - * compatibility purposes and may affect performance -- the state tracker - * should always specify exactly what got written while the buffer was - * mapped. */ void (*buffer_flush_mapped_range)( struct pipe_screen *screen, struct pipe_buffer *buf, unsigned offset, unsigned length); + /** + * Unmap buffer. + * + * If the buffer was mapped with PIPE_BUFFER_USAGE_CPU_WRITE flag but not + * PIPE_BUFFER_USAGE_FLUSH_EXPLICIT then the pipe driver will + * assume that the whole buffer was written. This is mostly for backward + * compatibility purposes and may affect performance -- the state tracker + * should always specify exactly what got written while the buffer was + * mapped. + */ void (*buffer_unmap)( struct pipe_screen *screen, struct pipe_buffer *buf ); diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 4dafdd6f0a9..f0ba4fb308c 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -78,6 +78,7 @@ enum tgsi_file_type { TGSI_FILE_SAMPLER =5, TGSI_FILE_ADDRESS =6, TGSI_FILE_IMMEDIATE =7, + TGSI_FILE_LOOP =8, TGSI_FILE_COUNT /**< how many TGSI_FILE_ types */ }; @@ -131,7 +132,8 @@ struct tgsi_declaration_range #define TGSI_SEMANTIC_PSIZE 4 #define TGSI_SEMANTIC_GENERIC 5 #define TGSI_SEMANTIC_NORMAL 6 -#define TGSI_SEMANTIC_COUNT 7 /**< number of semantic values */ +#define TGSI_SEMANTIC_FACE 7 +#define TGSI_SEMANTIC_COUNT 8 /**< number of semantic values */ struct tgsi_declaration_semantic { @@ -151,13 +153,16 @@ struct tgsi_immediate unsigned Extended : 1; /**< BOOL */ }; -struct tgsi_immediate_float32 +union tgsi_immediate_data { float Float; }; -/* - * GL_NV_vertex_program +/* TGSI opcodes. + * + * For more information on semantics of opcodes and + * which APIs are known to use which opcodes, see + * auxiliary/tgsi/tgsi-instruction-set.txt */ #define TGSI_OPCODE_ARL 0 #define TGSI_OPCODE_MOV 1 @@ -176,62 +181,32 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_SLT 14 #define TGSI_OPCODE_SGE 15 #define TGSI_OPCODE_MAD 16 - -/* - * GL_ATI_fragment_shader - */ #define TGSI_OPCODE_SUB 17 -#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 -#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 -#define TGSI_OPCODE_LERP 18 +#define TGSI_OPCODE_LRP 18 #define TGSI_OPCODE_CND 19 #define TGSI_OPCODE_CND0 20 -#define TGSI_OPCODE_DOT2ADD 21 - -/* - * GL_EXT_vertex_shader - */ -#define TGSI_OPCODE_INDEX 22 /* considered for removal */ -#define TGSI_OPCODE_NEGATE 23 /* considered for removal */ -#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD -#define TGSI_OPCODE_FRAC 24 -#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE -#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT +#define TGSI_OPCODE_DP2A 21 + /* gap */ +#define TGSI_OPCODE_FRC 24 #define TGSI_OPCODE_CLAMP 25 -#define TGSI_OPCODE_FLOOR 26 +#define TGSI_OPCODE_FLR 26 #define TGSI_OPCODE_ROUND 27 -#define TGSI_OPCODE_EXPBASE2 28 -#define TGSI_OPCODE_LOGBASE2 29 -#define TGSI_OPCODE_POWER 30 -#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP -#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ -#define TGSI_OPCODE_CROSSPRODUCT 31 -#define TGSI_OPCODE_MULTIPLYMATRIX 32 /* considered for removal */ - -/* - * GL_NV_vertex_program1_1 - */ +#define TGSI_OPCODE_EX2 28 +#define TGSI_OPCODE_LG2 29 +#define TGSI_OPCODE_POW 30 +#define TGSI_OPCODE_XPD 31 + /* gap */ #define TGSI_OPCODE_ABS 33 #define TGSI_OPCODE_RCC 34 #define TGSI_OPCODE_DPH 35 - -/* - * GL_NV_fragment_program - */ #define TGSI_OPCODE_COS 36 #define TGSI_OPCODE_DDX 37 #define TGSI_OPCODE_DDY 38 -#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 -#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR -#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC #define TGSI_OPCODE_KILP 39 /* predicated kill */ -#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 -#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP #define TGSI_OPCODE_PK2H 40 #define TGSI_OPCODE_PK2US 41 #define TGSI_OPCODE_PK4B 42 #define TGSI_OPCODE_PK4UB 43 -#define TGSI_OPCODE_POW TGSI_OPCODE_POWER #define TGSI_OPCODE_RFL 44 #define TGSI_OPCODE_SEQ 45 #define TGSI_OPCODE_SFL 46 @@ -248,66 +223,29 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_UP4B 57 #define TGSI_OPCODE_UP4UB 58 #define TGSI_OPCODE_X2D 59 - -/* - * GL_NV_vertex_program2 - */ #define TGSI_OPCODE_ARA 60 #define TGSI_OPCODE_ARR 61 #define TGSI_OPCODE_BRA 62 #define TGSI_OPCODE_CAL 63 #define TGSI_OPCODE_RET 64 -#define TGSI_OPCODE_SSG 65 - -/* - * GL_ARB_vertex_program - */ -#define TGSI_OPCODE_SWZ 118 -#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT - -/* - * GL_ARB_fragment_program - */ +#define TGSI_OPCODE_SSG 65 /* SGN */ #define TGSI_OPCODE_CMP 66 -#define TGSI_OPCODE_KIL 116 /* conditional kill */ #define TGSI_OPCODE_SCS 67 #define TGSI_OPCODE_TXB 68 - -/* - * GL_NV_fragment_program_option - */ -/* No new opcode */ - -/* - * GL_NV_fragment_program2 - */ #define TGSI_OPCODE_NRM 69 #define TGSI_OPCODE_DIV 70 #define TGSI_OPCODE_DP2 71 -#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD #define TGSI_OPCODE_TXL 72 #define TGSI_OPCODE_BRK 73 #define TGSI_OPCODE_IF 74 -#define TGSI_OPCODE_LOOP 75 +#define TGSI_OPCODE_BGNFOR 75 #define TGSI_OPCODE_REP 76 #define TGSI_OPCODE_ELSE 77 #define TGSI_OPCODE_ENDIF 78 -#define TGSI_OPCODE_ENDLOOP 79 +#define TGSI_OPCODE_ENDFOR 79 #define TGSI_OPCODE_ENDREP 80 - -/* - * GL_NV_vertex_program2_option - */ - -/* - * GL_NV_vertex_program3 - */ #define TGSI_OPCODE_PUSHA 81 #define TGSI_OPCODE_POPA 82 - -/* - * GL_NV_gpu_program4 - */ #define TGSI_OPCODE_CEIL 83 #define TGSI_OPCODE_I2F 84 #define TGSI_OPCODE_NOT 85 @@ -322,103 +260,25 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_TXF 94 #define TGSI_OPCODE_TXQ 95 #define TGSI_OPCODE_CONT 96 - -/* - * GL_NV_vertex_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_fragment_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_geometry_program4 - */ -/* Same as GL_NV_gpu_program4 */ #define TGSI_OPCODE_EMIT 97 #define TGSI_OPCODE_ENDPRIM 98 - -/* - * GLSL - */ -#define TGSI_OPCODE_BGNLOOP2 99 +#define TGSI_OPCODE_BGNLOOP 99 #define TGSI_OPCODE_BGNSUB 100 -#define TGSI_OPCODE_ENDLOOP2 101 +#define TGSI_OPCODE_ENDLOOP 101 #define TGSI_OPCODE_ENDSUB 102 -#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC #define TGSI_OPCODE_NOISE1 103 #define TGSI_OPCODE_NOISE2 104 #define TGSI_OPCODE_NOISE3 105 #define TGSI_OPCODE_NOISE4 106 #define TGSI_OPCODE_NOP 107 - -/* - * ps_1_1 - */ -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL - -/* - * ps_1_2 - */ -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_3 - */ -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_4 - */ -#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX - -/* - * ps_2_0 - */ -#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX -#define TGSI_OPCODE_M4X3 108 -#define TGSI_OPCODE_M3X4 109 -#define TGSI_OPCODE_M3X3 110 -#define TGSI_OPCODE_M3X2 111 -#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD + /* gap */ #define TGSI_OPCODE_NRM4 112 -#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS -#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A - -/* - * ps_2_x - */ -#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL #define TGSI_OPCODE_CALLNZ 113 #define TGSI_OPCODE_IFC 114 -#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK #define TGSI_OPCODE_BREAKC 115 -#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX -#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY -#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD - -/* - * vs_1_1 - */ -#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP -#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 - -/* - * vs_2_0 - */ -#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG -#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR -/* EXPP - use TGSI_OPCODE_EX2 */ - -/* - * vs_2_x - */ - +#define TGSI_OPCODE_KIL 116 /* conditional kill */ #define TGSI_OPCODE_END 117 /* aka HALT */ - +#define TGSI_OPCODE_SWZ 118 #define TGSI_OPCODE_LAST 119 #define TGSI_SAT_NONE 0 /* do not saturate */ diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 4b590bdc906..626bedb35a8 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -75,11 +75,11 @@ struct pipe_surface; */ struct pipe_buffer { - struct pipe_reference reference; - struct pipe_screen *screen; - unsigned alignment; - unsigned usage; - unsigned size; + struct pipe_reference reference; + unsigned size; + struct pipe_screen *screen; + unsigned alignment; + unsigned usage; }; @@ -265,7 +265,6 @@ struct pipe_sampler_state unsigned compare_func:3; /**< PIPE_FUNC_x */ unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ unsigned prefilter:4; /**< Wierd sampling state exposed by some api's */ - float shadow_ambient; /**< shadow test fail color/intensity */ float lod_bias; /**< LOD/lambda bias */ float min_lod, max_lod; /**< LOD clamp range, after bias */ float border_color[4]; @@ -287,10 +286,10 @@ struct pipe_surface unsigned offset; /**< offset from start of buffer, in bytes */ unsigned usage; /**< PIPE_BUFFER_USAGE_* */ + unsigned zslice; struct pipe_texture *texture; /**< texture into which this is a view */ unsigned face; unsigned level; - unsigned zslice; }; diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index df6d38904a6..96e8e087447 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -36,6 +36,7 @@ #include "pipe/p_compiler.h" +#include "util/u_debug.h" /* for assert */ #if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h index 5790b2f6c7e..7a38b508fbe 100644 --- a/src/gallium/include/state_tracker/drm_api.h +++ b/src/gallium/include/state_tracker/drm_api.h @@ -32,33 +32,37 @@ struct drm_api * Special buffer functions */ /*@{*/ - struct pipe_screen* (*create_screen)(int drm_fd, - struct drm_create_screen_arg *arg); - struct pipe_context* (*create_context)(struct pipe_screen *screen); + struct pipe_screen* (*create_screen)(struct drm_api *api, int drm_fd, + struct drm_create_screen_arg *arg); + struct pipe_context* (*create_context)(struct drm_api *api, + struct pipe_screen *screen); /*@}*/ /** * Special buffer functions */ /*@{*/ - boolean (*buffer_from_texture)(struct pipe_texture *texture, - struct pipe_buffer **buffer, - unsigned *stride); - struct pipe_buffer* (*buffer_from_handle)(struct pipe_screen *screen, + boolean (*buffer_from_texture)(struct drm_api *api, + struct pipe_texture *texture, + struct pipe_buffer **buffer, + unsigned *stride); + struct pipe_buffer* (*buffer_from_handle)(struct drm_api *api, + struct pipe_screen *screen, const char *name, unsigned handle); - boolean (*handle_from_buffer)(struct pipe_screen *screen, - struct pipe_buffer *buffer, - unsigned *handle); - boolean (*global_handle_from_buffer)(struct pipe_screen *screen, - struct pipe_buffer *buffer, - unsigned *handle); + boolean (*handle_from_buffer)(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); + boolean (*global_handle_from_buffer)(struct drm_api *api, + struct pipe_screen *screen, + struct pipe_buffer *buffer, + unsigned *handle); /*@}*/ + + void (*destroy)(struct drm_api *api); }; -/** - * A driver needs to export this symbol - */ -extern struct drm_api drm_api_hooks; +extern struct drm_api * drm_api_create(void); #endif diff --git a/src/gallium/state_trackers/dri/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c index 45eaec4ed39..830e511fefd 100644 --- a/src/gallium/state_trackers/dri/dri_context.c +++ b/src/gallium/state_trackers/dri/dri_context.c @@ -69,7 +69,7 @@ dri_create_context(const __GLcontextModes * visual, driParseConfigFiles(&ctx->optionCache, &screen->optionCache, sPriv->myNum, "dri"); - ctx->pipe = drm_api_hooks.create_context(screen->pipe_screen); + ctx->pipe = screen->api->create_context(screen->api, screen->pipe_screen); if (ctx->pipe == NULL) goto fail; @@ -109,9 +109,6 @@ dri_destroy_context(__DRIcontextPrivate * cPriv) */ st_flush(ctx->st, 0, NULL); - if (screen->dummyContext == ctx) - screen->dummyContext = NULL; - /* Also frees ctx->pipe? */ st_destroy_context(ctx->st); @@ -153,11 +150,6 @@ dri_make_current(__DRIcontextPrivate * cPriv, ++ctx->bind_count; - /* This is for situations in which we need a rendering context but - * there may not be any currently bound. - */ - screen->dummyContext = ctx; - if (ctx->dPriv != driDrawPriv) { ctx->dPriv = driDrawPriv; ctx->d_stamp = driDrawPriv->lastStamp - 1; diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c index 865cc8d0b6c..c7ca7f78a3d 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.c +++ b/src/gallium/state_trackers/dri/dri_drawable.c @@ -44,16 +44,9 @@ #include "util/u_memory.h" -static void -dri_copy_to_front(__DRIdrawablePrivate * dPriv, - struct pipe_surface *from, - int x, int y, unsigned w, unsigned h) -{ - /* TODO send a message to the Xserver to copy to the real front buffer */ -} - static struct pipe_surface * -dri_surface_from_handle(struct pipe_screen *screen, +dri_surface_from_handle(struct drm_api *api, + struct pipe_screen *screen, unsigned handle, enum pipe_format format, unsigned width, unsigned height, unsigned pitch) @@ -63,9 +56,11 @@ dri_surface_from_handle(struct pipe_screen *screen, struct pipe_texture templat; struct pipe_buffer *buf = NULL; - buf = drm_api_hooks.buffer_from_handle(screen, "dri2 buffer", handle); - if (!buf) + buf = api->buffer_from_handle(api, screen, "dri2 buffer", handle); + if (!buf) { + debug_printf("%s: Failed to get buffer from handle\n", __func__); return NULL; + } memset(&templat, 0, sizeof(templat)); templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; @@ -82,8 +77,10 @@ dri_surface_from_handle(struct pipe_screen *screen, /* we don't need the buffer from this point on */ pipe_buffer_reference(&buf, NULL); - if (!texture) + if (!texture) { + debug_printf("%s: Failed to blanket the buffer with a texture\n", __func__); return NULL; + } surface = screen->get_tex_surface(screen, texture, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | @@ -95,17 +92,48 @@ dri_surface_from_handle(struct pipe_screen *screen, } /** + * Pixmaps have will have the same name of fake front and front. + */ +static boolean +dri2_check_if_pixmap(__DRIbuffer *buffers, int count) +{ + boolean found = FALSE; + boolean is_pixmap = FALSE; + unsigned name; + int i; + + for (i = 0; i < count; i++) { + switch (buffers[i].attachment) { + case __DRI_BUFFER_FRONT_LEFT: + case __DRI_BUFFER_FAKE_FRONT_LEFT: + if (found) { + is_pixmap = buffers[i].name == name; + } else { + name = buffers[i].name; + found = TRUE; + } + default: + continue; + } + } + + return is_pixmap; +} + +/** * This will be called a drawable is known to have been resized. */ void dri_get_buffers(__DRIdrawablePrivate * dPriv) { + struct dri_drawable *drawable = dri_drawable(dPriv); struct pipe_surface *surface = NULL; struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen; __DRIbuffer *buffers = NULL; __DRIscreen *dri_screen = drawable->sPriv; __DRIdrawable *dri_drawable = drawable->dPriv; + struct drm_api *api = ((struct dri_screen*)(dri_screen->private))->api; boolean have_depth = FALSE; int i, count; @@ -150,36 +178,35 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) memcpy(drawable->old, buffers, sizeof(__DRIbuffer) * count); } + drawable->is_pixmap = dri2_check_if_pixmap(buffers, count); + for (i = 0; i < count; i++) { enum pipe_format format = 0; int index = 0; switch (buffers[i].attachment) { case __DRI_BUFFER_FRONT_LEFT: - index = ST_SURFACE_FRONT_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; - break; + continue; case __DRI_BUFFER_FAKE_FRONT_LEFT: index = ST_SURFACE_FRONT_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; + format = drawable->color_format; break; case __DRI_BUFFER_BACK_LEFT: index = ST_SURFACE_BACK_LEFT; - format = PIPE_FORMAT_A8R8G8B8_UNORM; + format = drawable->color_format; break; case __DRI_BUFFER_DEPTH: index = ST_SURFACE_DEPTH; - format = PIPE_FORMAT_Z24S8_UNORM; + format = drawable->depth_format; break; case __DRI_BUFFER_STENCIL: index = ST_SURFACE_DEPTH; - format = PIPE_FORMAT_Z24S8_UNORM; + format = drawable->stencil_format; break; case __DRI_BUFFER_ACCUM: default: assert(0); } - assert(buffers[i].cpp == 4); if (index == ST_SURFACE_DEPTH) { if (have_depth) @@ -188,7 +215,8 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) have_depth = TRUE; } - surface = dri_surface_from_handle(screen, + surface = dri_surface_from_handle(api, + screen, buffers[i].name, format, dri_drawable->w, @@ -201,13 +229,54 @@ dri_get_buffers(__DRIdrawablePrivate * dPriv) st_resize_framebuffer(drawable->stfb, dri_drawable->w, dri_drawable->h); } +/** + * These are used for GLX_EXT_texture_from_pixmap + */ +void dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target, + GLint format, __DRIdrawable *dPriv) +{ + struct dri_drawable *drawable = dri_drawable(dPriv); + struct pipe_surface *ps; + + dri_get_buffers(drawable->dPriv); + st_get_framebuffer_surface(drawable->stfb, ST_SURFACE_FRONT_LEFT, &ps); + + st_bind_texture_surface(ps, target == GL_TEXTURE_2D ? ST_TEXTURE_2D : + ST_TEXTURE_RECT, 0, + format == GLX_TEXTURE_FORMAT_RGBA_EXT ? + PIPE_FORMAT_R8G8B8A8_UNORM : PIPE_FORMAT_R8G8B8X8_UNORM); +} + +void dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target, + __DRIdrawable *dPriv) +{ + dri2_set_tex_buffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv); +} + void dri_flush_frontbuffer(struct pipe_screen *screen, struct pipe_surface *surf, void *context_private) { struct dri_context *ctx = (struct dri_context *)context_private; + struct dri_drawable *drawable = dri_drawable(ctx->dPriv); + __DRIdrawable *dri_drawable = ctx->dPriv; + __DRIscreen *dri_screen = ctx->sPriv; + + /* XXX Does this function get called with DRI1? */ + + if (ctx->dPriv == NULL) { + debug_printf("%s: no drawable bound to context\n", __func__); + return; + } + +#if 0 + /* TODO if rendering to pixmaps is slow enable this code. */ + if (drawable->is_pixmap) + return; +#endif - dri_copy_to_front(ctx->dPriv, surf, 0, 0, surf->width, surf->height); + (*dri_screen->dri2.loader->flushFrontBuffer)(dri_drawable, + dri_drawable->loaderPrivate); } /** @@ -218,10 +287,8 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, __DRIdrawablePrivate * dPriv, const __GLcontextModes * visual, boolean isPixmap) { - enum pipe_format colorFormat, depthFormat, stencilFormat; struct dri_screen *screen = sPriv->private; struct dri_drawable *drawable = NULL; - struct pipe_screen *pscreen = screen->pipe_screen; int i; if (isPixmap) @@ -231,41 +298,55 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, if (drawable == NULL) goto fail; - /* XXX: todo: use the pipe_screen queries to figure out which - * render targets are supportable. - */ - assert(visual->redBits == 8); - assert(visual->depthBits == 24 || visual->depthBits == 0); - assert(visual->stencilBits == 8 || visual->stencilBits == 0); - - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits) { - if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, - PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET | - PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) - depthFormat = PIPE_FORMAT_Z24S8_UNORM; - else - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - } else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits) { - if (pscreen->is_format_supported(pscreen, PIPE_FORMAT_Z24S8_UNORM, - PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET | - PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0)) - stencilFormat = PIPE_FORMAT_Z24S8_UNORM; + if (visual->redBits == 8) { + if (visual->alphaBits == 8) + drawable->color_format = PIPE_FORMAT_A8R8G8B8_UNORM; else - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - } else - stencilFormat = PIPE_FORMAT_NONE; + drawable->color_format = PIPE_FORMAT_X8R8G8B8_UNORM; + } else { + drawable->color_format = PIPE_FORMAT_R5G6B5_UNORM; + } + + switch(visual->depthBits) { + default: + case 0: + drawable->depth_format = PIPE_FORMAT_NONE; + break; + case 16: + drawable->depth_format = PIPE_FORMAT_Z16_UNORM; + break; + case 24: + if (visual->stencilBits == 0) { + drawable->depth_format = (screen->d_depth_bits_last) ? + PIPE_FORMAT_X8Z24_UNORM: + PIPE_FORMAT_Z24X8_UNORM; + } else { + drawable->depth_format = (screen->sd_depth_bits_last) ? + PIPE_FORMAT_S8Z24_UNORM: + PIPE_FORMAT_Z24S8_UNORM; + } + break; + case 32: + drawable->depth_format = PIPE_FORMAT_Z32_UNORM; + break; + } + + switch(visual->stencilBits) { + default: + case 0: + drawable->stencil_format = PIPE_FORMAT_NONE; + break; + case 8: + drawable->stencil_format = (screen->sd_depth_bits_last) ? + PIPE_FORMAT_S8Z24_UNORM: + PIPE_FORMAT_Z24S8_UNORM; + break; + } drawable->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, + drawable->color_format, + drawable->depth_format, + drawable->stencil_format, dPriv->w, dPriv->h, (void *)drawable); if (drawable->stfb == NULL) @@ -276,12 +357,11 @@ dri_create_buffer(__DRIscreenPrivate * sPriv, dPriv->driverPrivate = (void *)drawable; /* setup dri2 buffers information */ + /* TODO incase of double buffer visual, delay fake creation */ i = 0; drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT; -#if 0 - /* TODO incase of double buffer visual, delay fake creation */ drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT; -#endif + if (visual->doubleBufferMode) drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT; if (visual->depthBits) diff --git a/src/gallium/state_trackers/dri/dri_drawable.h b/src/gallium/state_trackers/dri/dri_drawable.h index 0f654d804a7..eaf0b954bd4 100644 --- a/src/gallium/state_trackers/dri/dri_drawable.h +++ b/src/gallium/state_trackers/dri/dri_drawable.h @@ -46,6 +46,8 @@ struct dri_drawable unsigned attachments[8]; unsigned num_attachments; + boolean is_pixmap; + __DRIbuffer old[8]; unsigned old_num; unsigned old_w; @@ -58,6 +60,10 @@ struct dri_drawable unsigned int tail; unsigned int desired_fences; unsigned int cur_fences; + + enum pipe_format color_format; + enum pipe_format depth_format; + enum pipe_format stencil_format; }; static INLINE struct dri_drawable * @@ -87,6 +93,12 @@ void dri_get_buffers(__DRIdrawablePrivate * dPriv); void dri_destroy_buffer(__DRIdrawablePrivate * dPriv); +void dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target, + GLint glx_texture_format, __DRIdrawable *dPriv); + +void dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target, + __DRIdrawable *dPriv); + void dri1_update_drawables(struct dri_context *ctx, struct dri_drawable *draw, struct dri_drawable *read); diff --git a/src/gallium/state_trackers/dri/dri_extensions.c b/src/gallium/state_trackers/dri/dri_extensions.c index 2f48162526b..4349a4d1d2f 100644 --- a/src/gallium/state_trackers/dri/dri_extensions.c +++ b/src/gallium/state_trackers/dri/dri_extensions.c @@ -33,6 +33,7 @@ #include "dri_context.h" #include "state_tracker/st_context.h" +#define need_GL_ARB_map_buffer_range #define need_GL_ARB_multisample #define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters @@ -61,6 +62,7 @@ */ const struct dri_extension card_extensions[] = { {"GL_ARB_fragment_shader", NULL}, + {"GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions}, {"GL_ARB_multisample", GL_ARB_multisample_functions}, {"GL_ARB_multitexture", NULL}, {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, @@ -82,8 +84,7 @@ const struct dri_extension card_extensions[] = { {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", - GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, {"GL_EXT_blend_subtract", NULL}, diff --git a/src/gallium/state_trackers/dri/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c index d3392ee690b..884b6d50111 100644 --- a/src/gallium/state_trackers/dri/dri_screen.c +++ b/src/gallium/state_trackers/dri/dri_screen.c @@ -57,82 +57,144 @@ PUBLIC const char __driConfigOptions[] = const uint __driNConfigOptions = 3; +static const __DRItexBufferExtension dri2TexBufferExtension = { + { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION }, + dri2_set_tex_buffer, + dri2_set_tex_buffer2, +}; + static const __DRIextension *dri_screen_extensions[] = { &driReadDrawableExtension, &driCopySubBufferExtension.base, &driSwapControlExtension.base, &driFrameTrackingExtension.base, &driMediaStreamCounterExtension.base, + &dri2TexBufferExtension.base, NULL }; struct dri1_api *__dri1_api_hooks = NULL; static const __DRIconfig ** -dri_fill_in_modes(__DRIscreenPrivate * psp, - unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) +dri_fill_in_modes(struct dri_screen *screen, + unsigned pixel_bits) { - __DRIconfig **configs; - __GLcontextModes *m; + __DRIconfig **configs = NULL; unsigned num_modes; - uint8_t depth_bits_array[3]; - uint8_t stencil_bits_array[3]; + uint8_t depth_bits_array[5]; + uint8_t stencil_bits_array[5]; uint8_t msaa_samples_array[1]; unsigned depth_buffer_factor; unsigned back_buffer_factor; unsigned msaa_samples_factor; - GLenum fb_format; - GLenum fb_type; - int i; + struct pipe_screen *p_screen = screen->pipe_screen; + boolean pf_r5g6b5, pf_a8r8g8b8, pf_x8r8g8b8; + boolean pf_z16, pf_x8z24, pf_z24x8, pf_s8z24, pf_z24s8, pf_z32; static const GLenum back_buffer_modes[] = { GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - /* TODO probe the hardware of what is supports */ depth_bits_array[0] = 0; - depth_bits_array[1] = 24; - depth_bits_array[2] = 24; - - stencil_bits_array[0] = 0; /* no depth or stencil */ - stencil_bits_array[1] = 0; /* z24x8 */ - stencil_bits_array[2] = 8; /* z24s8 */ + stencil_bits_array[0] = 0; + depth_buffer_factor = 1; + + pf_z16 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z16_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_z32 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z32_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_x8z24 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_X8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_z24x8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24X8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_s8z24 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_S8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_z24s8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + pf_r5g6b5 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_R5G6B5_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0); + pf_a8r8g8b8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0); + pf_x8r8g8b8 = p_screen->is_format_supported(p_screen, PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0); + + if (pf_z16) { + depth_bits_array[depth_buffer_factor] = 16; + stencil_bits_array[depth_buffer_factor++] = 0; + } + if (pf_x8z24 || pf_z24x8) { + depth_bits_array[depth_buffer_factor] = 24; + stencil_bits_array[depth_buffer_factor++] = 0; + screen->d_depth_bits_last = pf_x8z24; + } + if (pf_s8z24 || pf_z24s8) { + depth_bits_array[depth_buffer_factor] = 24; + stencil_bits_array[depth_buffer_factor++] = 8; + screen->sd_depth_bits_last = pf_s8z24; + } + if (pf_z32) { + depth_bits_array[depth_buffer_factor] = 32; + stencil_bits_array[depth_buffer_factor++] = 0; + } msaa_samples_array[0] = 0; - - depth_buffer_factor = 3; back_buffer_factor = 3; msaa_samples_factor = 1; num_modes = depth_buffer_factor * back_buffer_factor * msaa_samples_factor * 4; - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; + if (pixel_bits == 16 && pf_r5g6b5) { + configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, + msaa_samples_array, 1); } else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + __DRIconfig **configs_a8r8g8b8 = NULL; + __DRIconfig **configs_x8r8g8b8 = NULL; + + if (pf_a8r8g8b8) + configs_a8r8g8b8 = driCreateConfigs(GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, + depth_bits_array, + stencil_bits_array, + depth_buffer_factor, + back_buffer_modes, + back_buffer_factor, + msaa_samples_array, 1); + if (pf_x8r8g8b8) + configs_x8r8g8b8 = driCreateConfigs(GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV, + depth_bits_array, + stencil_bits_array, + depth_buffer_factor, + back_buffer_modes, + back_buffer_factor, + msaa_samples_array, 1); + + if (configs_a8r8g8b8 && configs_x8r8g8b8) + configs = driConcatConfigs(configs_x8r8g8b8, configs_a8r8g8b8); + else if (configs_a8r8g8b8) + configs = configs_a8r8g8b8; + else if (configs_x8r8g8b8) + configs = configs_x8r8g8b8; + else + configs = NULL; } - configs = driCreateConfigs(fb_format, fb_type, - depth_bits_array, - stencil_bits_array, depth_buffer_factor, - back_buffer_modes, back_buffer_factor, - msaa_samples_array, msaa_samples_factor); if (configs == NULL) { debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__); return NULL; } - for (i = 0; configs[i]; i++) { - m = &configs[i]->modes; - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - return (const const __DRIconfig **)configs; } @@ -170,6 +232,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv) if (!screen) return NULL; + screen->api = drm_api_create(); screen->sPriv = sPriv; screen->fd = sPriv->fd; screen->drmLock = (drmLock *) & sPriv->pSAREA->lock; @@ -187,7 +250,7 @@ dri_init_screen(__DRIscreenPrivate * sPriv) dri_copy_version(&arg.drm_version, &sPriv->drm_version); arg.api = NULL; - screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg.base); + screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg.base); if (!screen->pipe_screen || !arg.api) { debug_printf("%s: failed to create dri1 screen\n", __FUNCTION__); @@ -200,7 +263,13 @@ dri_init_screen(__DRIscreenPrivate * sPriv) driParseOptionInfo(&screen->optionCache, __driConfigOptions, __driNConfigOptions); - configs = dri_fill_in_modes(sPriv, sPriv->fbBPP, 24, 8, 1); + /** + * FIXME: If the driver supports format conversion swapbuffer blits, we might + * want to support other color bit depths than the server is currently + * using. + */ + + configs = dri_fill_in_modes(screen, sPriv->fbBPP); if (!configs) goto out_no_configs; @@ -230,13 +299,14 @@ dri_init_screen2(__DRIscreenPrivate * sPriv) if (!screen) goto fail; + screen->api = drm_api_create(); screen->sPriv = sPriv; screen->fd = sPriv->fd; sPriv->private = (void *)screen; sPriv->extensions = dri_screen_extensions; arg.mode = DRM_CREATE_NORMAL; - screen->pipe_screen = drm_api_hooks.create_screen(screen->fd, &arg); + screen->pipe_screen = screen->api->create_screen(screen->api, screen->fd, &arg); if (!screen->pipe_screen) { debug_printf("%s: failed to create pipe_screen\n", __FUNCTION__); goto fail; @@ -248,7 +318,7 @@ dri_init_screen2(__DRIscreenPrivate * sPriv) driParseOptionInfo(&screen->optionCache, __driConfigOptions, __driNConfigOptions); - return dri_fill_in_modes(sPriv, 4 * 8, 24, 8, 1); + return dri_fill_in_modes(screen, 32); fail: return NULL; } diff --git a/src/gallium/state_trackers/dri/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h index 100d9e50e07..f6c56d0f0c5 100644 --- a/src/gallium/state_trackers/dri/dri_screen.h +++ b/src/gallium/state_trackers/dri/dri_screen.h @@ -49,19 +49,16 @@ struct dri_screen */ driOptionCache optionCache; - /** - * Temporary(?) context to use for SwapBuffers or other situations in - * which we need a rendering context, but none is currently bound. - */ - struct dri_context *dummyContext; - /* drm */ int fd; drmLock *drmLock; /* gallium */ + struct drm_api *api; struct pipe_winsys *pipe_winsys; struct pipe_screen *pipe_screen; + boolean d_depth_bits_last; + boolean sd_depth_bits_last; }; /** cast wrapper */ diff --git a/src/gallium/state_trackers/egl/egl_context.c b/src/gallium/state_trackers/egl/egl_context.c index 36548fae263..c4f7361ca0b 100644 --- a/src/gallium/state_trackers/egl/egl_context.c +++ b/src/gallium/state_trackers/egl/egl_context.c @@ -83,23 +83,16 @@ const struct dri_extension card_extensions[] = { {NULL, NULL} }; -EGLContext -drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) +_EGLContext * +drm_create_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list) { - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev = lookup_drm_device(dpy); struct drm_context *ctx; struct drm_context *share = NULL; struct st_context *st_share = NULL; - _EGLConfig *conf; int i; __GLcontextModes *visual; - conf = _eglLookupConfig(drv, dpy, config); - if (!conf) { - _eglError(EGL_BAD_CONFIG, "eglCreateContext"); - return EGL_NO_CONTEXT; - } - for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { switch (attrib_list[i]) { /* no attribs defined for now */ @@ -113,9 +106,9 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext if (!ctx) goto err_c; - _eglInitContext(drv, dpy, &ctx->base, config, attrib_list); + _eglInitContext(drv, &ctx->base, conf, attrib_list); - ctx->pipe = drm_api_hooks.create_context(dev->screen); + ctx->pipe = dev->api->create_context(dev->api, dev->screen); if (!ctx->pipe) goto err_pipe; @@ -129,28 +122,21 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext if (!ctx->st) goto err_gl; - /* generate handle and insert into hash table */ - _eglSaveContext(&ctx->base); - assert(_eglGetContextHandle(&ctx->base)); - - return _eglGetContextHandle(&ctx->base); + return &ctx->base; err_gl: ctx->pipe->destroy(ctx->pipe); err_pipe: free(ctx); err_c: - return EGL_NO_CONTEXT; + return NULL; } EGLBoolean -drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) +drm_destroy_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *context) { struct drm_context *c = lookup_drm_context(context); - _eglRemoveContext(&c->base); - if (c->base.IsBound) { - c->base.DeletePending = EGL_TRUE; - } else { + if (!_eglIsContextBound(&c->base)) { st_destroy_context(c->st); c->pipe->destroy(c->pipe); free(c); @@ -159,7 +145,7 @@ drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) } EGLBoolean -drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context) +drm_make_current(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *context) { struct drm_surface *readSurf = lookup_drm_surface(read); struct drm_surface *drawSurf = lookup_drm_surface(draw); diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c index 489aa8d9af5..3ef1945b93c 100644 --- a/src/gallium/state_trackers/egl/egl_surface.c +++ b/src/gallium/state_trackers/egl/egl_surface.c @@ -67,39 +67,21 @@ drm_create_framebuffer(const __GLcontextModes *visual, } static void -drm_create_texture(_EGLDriver *drv, +drm_create_texture(_EGLDisplay *dpy, struct drm_screen *scrn, unsigned w, unsigned h) { - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev = lookup_drm_device(dpy); struct pipe_screen *screen = dev->screen; struct pipe_surface *surface; struct pipe_texture *texture; struct pipe_texture templat; - struct pipe_buffer *buf; - unsigned stride = 1024; + struct pipe_buffer *buf = NULL; unsigned pitch = 0; - unsigned size = 0; - - /* ugly */ - if (stride < w) - stride = 2048; - - pitch = stride * 4; - size = h * 2 * pitch; - - buf = pipe_buffer_create(screen, - 0, /* alignment */ - PIPE_BUFFER_USAGE_GPU_READ_WRITE | - PIPE_BUFFER_USAGE_CPU_READ_WRITE, - size); - - if (!buf) - goto err_buf; memset(&templat, 0, sizeof(templat)); - templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; - templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET; + templat.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY; templat.target = PIPE_TEXTURE_2D; templat.last_level = 0; templat.depth[0] = 1; @@ -108,13 +90,16 @@ drm_create_texture(_EGLDriver *drv, templat.height[0] = h; pf_get_block(templat.format, &templat.block); - texture = screen->texture_blanket(dev->screen, - &templat, - &pitch, - buf); + texture = screen->texture_create(dev->screen, + &templat); + if (!texture) goto err_tex; + dev->api->buffer_from_texture(dev->api, texture, &buf, &pitch); + if (!buf) + goto err_buf; + surface = screen->get_tex_surface(screen, texture, 0, @@ -125,14 +110,13 @@ drm_create_texture(_EGLDriver *drv, if (!surface) goto err_surf; - scrn->tex = texture; scrn->surface = surface; scrn->buffer = buf; scrn->front.width = w; scrn->front.height = h; scrn->front.pitch = pitch; - drm_api_hooks.handle_from_buffer(screen, scrn->buffer, &scrn->front.handle); + dev->api->handle_from_buffer(dev->api, screen, scrn->buffer, &scrn->front.handle); if (0) goto err_handle; @@ -142,9 +126,9 @@ err_handle: pipe_surface_reference(&surface, NULL); err_surf: pipe_texture_reference(&texture, NULL); +err_buf: err_tex: pipe_buffer_reference(&buf, NULL); -err_buf: return; } @@ -153,9 +137,9 @@ err_buf: */ void -drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) +drm_takedown_shown_screen(_EGLDisplay *dpy, struct drm_screen *screen) { - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev = lookup_drm_device(dpy); screen->surf = NULL; @@ -178,22 +162,22 @@ drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) screen->shown = 0; } -EGLSurface -drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list) +_EGLSurface * +drm_create_window_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list) { - return EGL_NO_SURFACE; + return NULL; } -EGLSurface -drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list) +_EGLSurface * +drm_create_pixmap_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list) { - return EGL_NO_SURFACE; + return NULL; } -EGLSurface -drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, +_EGLSurface * +drm_create_pbuffer_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list) { int i; @@ -201,13 +185,6 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, int height = -1; struct drm_surface *surf = NULL; __GLcontextModes *visual; - _EGLConfig *conf; - - conf = _eglLookupConfig(drv, dpy, config); - if (!conf) { - _eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface"); - return EGL_NO_CONTEXT; - } for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { switch (attrib_list[i]) { @@ -225,14 +202,14 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, if (width < 1 || height < 1) { _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); - return EGL_NO_SURFACE; + return NULL; } surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface)); if (!surf) goto err; - if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list)) + if (!_eglInitSurface(drv, &surf->base, EGL_PBUFFER_BIT, conf, attrib_list)) goto err_surf; surf->w = width; @@ -245,17 +222,16 @@ drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, (void*)surf); drm_visual_modes_destroy(visual); - _eglSaveSurface(&surf->base); - return surf->base.Handle; + return &surf->base; err_surf: free(surf); err: - return EGL_NO_SURFACE; + return NULL; } -EGLSurface -drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, +_EGLSurface * +drm_create_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *cfg, const EGLint *attrib_list) { EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list); @@ -264,22 +240,21 @@ drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, } EGLBoolean -drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, - EGLScreenMESA screen, - EGLSurface surface, EGLModeMESA m) +drm_show_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, + _EGLScreen *screen, + _EGLSurface *surface, _EGLMode *mode) { - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev = lookup_drm_device(dpy); struct drm_surface *surf = lookup_drm_surface(surface); - struct drm_screen *scrn = lookup_drm_screen(dpy, screen); - _EGLMode *mode = _eglLookupMode(dpy, m); + struct drm_screen *scrn = lookup_drm_screen(screen); int ret; unsigned int i, k; if (scrn->shown) - drm_takedown_shown_screen(drv, scrn); + drm_takedown_shown_screen(dpy, scrn); - drm_create_texture(drv, scrn, mode->Width, mode->Height); + drm_create_texture(dpy, scrn, mode->Width, mode->Height); if (!scrn->buffer) return EGL_FALSE; @@ -361,15 +336,12 @@ err_bo: } EGLBoolean -drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +drm_destroy_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface) { struct drm_surface *surf = lookup_drm_surface(surface); - _eglRemoveSurface(&surf->base); - if (surf->base.IsBound) { - surf->base.DeletePending = EGL_TRUE; - } else { + if (!_eglIsSurfaceBound(&surf->base)) { if (surf->screen) - drm_takedown_shown_screen(drv, surf->screen); + drm_takedown_shown_screen(dpy, surf->screen); st_unreference_framebuffer(surf->stfb); free(surf); } @@ -377,8 +349,9 @@ drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) } EGLBoolean -drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +drm_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw) { + struct drm_device *dev = lookup_drm_device(dpy); struct drm_surface *surf = lookup_drm_surface(draw); struct pipe_surface *back_surf; @@ -396,7 +369,6 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) st_notify_swapbuffers(surf->stfb); if (surf->screen) { - surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL); surf->user->pipe->surface_copy(surf->user->pipe, surf->screen->surface, 0, 0, @@ -404,7 +376,15 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) 0, 0, surf->w, surf->h); surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL); - /* TODO stuff here */ + +#ifdef DRM_MODE_FEATURE_DIRTYFB + /* TODO query connector property to see if this is needed */ + drmModeDirtyFB(dev->drmFD, surf->screen->fbID, NULL, 0); +#else + (void)dev; +#endif + + /* TODO more stuff here */ } } diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c index 8e620084617..5140755001c 100644 --- a/src/gallium/state_trackers/egl/egl_tracker.c +++ b/src/gallium/state_trackers/egl/egl_tracker.c @@ -6,6 +6,8 @@ #include <string.h> #include "egl_tracker.h" +#include <fcntl.h> + #include "egllog.h" #include "state_tracker/drm_api.h" @@ -21,44 +23,46 @@ extern const struct dri_extension card_extensions[]; * Exported functions */ +static void +drm_unload(_EGLDriver *drv) +{ + free(drv); +} + /** * The bootstrap function. Return a new drm_driver object and * plug in API functions. */ _EGLDriver * -_eglMain(_EGLDisplay *dpy, const char *args) +_eglMain(const char *args) { - struct drm_device *drm; + _EGLDriver *drv; - drm = (struct drm_device *) calloc(1, sizeof(struct drm_device)); - if (!drm) { + drv = (_EGLDriver *) calloc(1, sizeof(_EGLDriver)); + if (!drv) { return NULL; } /* First fill in the dispatch table with defaults */ - _eglInitDriverFallbacks(&drm->base); + _eglInitDriverFallbacks(drv); /* then plug in our Drm-specific functions */ - drm->base.API.Initialize = drm_initialize; - drm->base.API.Terminate = drm_terminate; - drm->base.API.CreateContext = drm_create_context; - drm->base.API.MakeCurrent = drm_make_current; - drm->base.API.CreateWindowSurface = drm_create_window_surface; - drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface; - drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface; - drm->base.API.DestroySurface = drm_destroy_surface; - drm->base.API.DestroyContext = drm_destroy_context; - drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa; - drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; - drm->base.API.SwapBuffers = drm_swap_buffers; - - drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; - drm->base.Name = "DRM/Gallium/Win"; - - /* enable supported extensions */ - drm->base.Extensions.MESA_screen_surface = EGL_TRUE; - drm->base.Extensions.MESA_copy_context = EGL_TRUE; - - return &drm->base; + drv->API.Initialize = drm_initialize; + drv->API.Terminate = drm_terminate; + drv->API.CreateContext = drm_create_context; + drv->API.MakeCurrent = drm_make_current; + drv->API.CreateWindowSurface = drm_create_window_surface; + drv->API.CreatePixmapSurface = drm_create_pixmap_surface; + drv->API.CreatePbufferSurface = drm_create_pbuffer_surface; + drv->API.DestroySurface = drm_destroy_surface; + drv->API.DestroyContext = drm_destroy_context; + drv->API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa; + drv->API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; + drv->API.SwapBuffers = drm_swap_buffers; + + drv->Name = "DRM/Gallium/Win"; + drv->Unload = drm_unload; + + return drv; } static void @@ -126,11 +130,18 @@ drm_find_dpms(struct drm_device *dev, struct drm_screen *screen) screen->dpms = p; } +static int drm_open_minor(int minor) +{ + char buf[64]; + + sprintf(buf, DRM_DEV_NAME, DRM_DIR_NAME, minor); + return open(buf, O_RDWR, 0); +} + EGLBoolean -drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) +drm_initialize(_EGLDriver *drv, _EGLDisplay *disp, EGLint *major, EGLint *minor) { - _EGLDisplay *disp = _eglLookupDisplay(dpy); - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev; struct drm_screen *screen = NULL; drmModeConnectorPtr connector = NULL; drmModeResPtr res = NULL; @@ -139,14 +150,20 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) EGLint i; int fd; - fd = drmOpen("i915", NULL); + dev = (struct drm_device *) calloc(1, sizeof(struct drm_device)); + if (!dev) + return EGL_FALSE; + dev->api = drm_api_create(); + + /* try the first node */ + fd = drm_open_minor(0); if (fd < 0) goto err_fd; dev->drmFD = fd; drm_get_device_id(dev); - dev->screen = drm_api_hooks.create_screen(dev->drmFD, NULL); + dev->screen = dev->api->create_screen(dev->api, dev->drmFD, NULL); if (!dev->screen) goto err_screen; dev->winsys = dev->screen->winsys; @@ -184,6 +201,8 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) } dev->count_screens = num_screens; + disp->DriverData = dev; + /* for now we only have one config */ _EGLConfig *config = calloc(1, sizeof(*config)); memset(config, 1, sizeof(*config)); @@ -198,7 +217,10 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); _eglAddConfig(disp, config); - drv->Initialized = EGL_TRUE; + disp->ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + /* enable supported extensions */ + disp->Extensions.MESA_screen_surface = EGL_TRUE; + disp->Extensions.MESA_copy_context = EGL_TRUE; *major = 1; *minor = 4; @@ -208,23 +230,27 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) err_screen: drmClose(fd); err_fd: + free(dev); return EGL_FALSE; } EGLBoolean -drm_terminate(_EGLDriver *drv, EGLDisplay dpy) +drm_terminate(_EGLDriver *drv, _EGLDisplay *dpy) { - struct drm_device *dev = (struct drm_device *)drv; + struct drm_device *dev = lookup_drm_device(dpy); struct drm_screen *screen; int i = 0; + _eglReleaseDisplayResources(drv, dpy); + _eglCleanupDisplay(dpy); + drmFreeVersion(dev->version); for (i = 0; i < dev->count_screens; i++) { screen = dev->screens[i]; if (screen->shown) - drm_takedown_shown_screen(drv, screen); + drm_takedown_shown_screen(dpy, screen); drmModeFreeProperty(screen->dpms); drmModeFreeConnector(screen->connector); @@ -237,8 +263,9 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy) drmClose(dev->drmFD); - _eglCleanupDisplay(_eglLookupDisplay(dpy)); + dev->api->destroy(dev->api); free(dev); + dpy->DriverData = NULL; return EGL_TRUE; } diff --git a/src/gallium/state_trackers/egl/egl_tracker.h b/src/gallium/state_trackers/egl/egl_tracker.h index ce2717de639..dd4730f9579 100644 --- a/src/gallium/state_trackers/egl/egl_tracker.h +++ b/src/gallium/state_trackers/egl/egl_tracker.h @@ -32,12 +32,11 @@ struct drm_context; struct drm_device { - _EGLDriver base; /* base class/object */ - /* * pipe */ + struct drm_api *api; struct pipe_winsys *winsys; struct pipe_screen *screen; @@ -135,25 +134,29 @@ struct drm_screen }; +static INLINE struct drm_device * +lookup_drm_device(_EGLDisplay *d) +{ + return (struct drm_device *) d->DriverData; +} + + static INLINE struct drm_context * -lookup_drm_context(EGLContext context) +lookup_drm_context(_EGLContext *c) { - _EGLContext *c = _eglLookupContext(context); return (struct drm_context *) c; } static INLINE struct drm_surface * -lookup_drm_surface(EGLSurface surface) +lookup_drm_surface(_EGLSurface *s) { - _EGLSurface *s = _eglLookupSurface(surface); return (struct drm_surface *) s; } static INLINE struct drm_screen * -lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen) +lookup_drm_screen(_EGLScreen *s) { - _EGLScreen *s = _eglLookupScreen(dpy, screen); return (struct drm_screen *) s; } @@ -170,25 +173,25 @@ __GLcontextModes* drm_visual_from_config(_EGLConfig *conf); * egl_surface.h */ /*@{*/ -void drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen); +void drm_takedown_shown_screen(_EGLDisplay *dpy, struct drm_screen *screen); /*@}*/ /** * All function exported to the egl side. */ /*@{*/ -EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor); -EGLBoolean drm_terminate(_EGLDriver *drv, EGLDisplay dpy); -EGLContext drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list); -EGLBoolean drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context); -EGLSurface drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list); -EGLSurface drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list); -EGLSurface drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list); -EGLSurface drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, const EGLint *attrib_list); -EGLBoolean drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface surface, EGLModeMESA m); -EGLBoolean drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface); -EGLBoolean drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context); -EGLBoolean drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw); +EGLBoolean drm_initialize(_EGLDriver *drv, _EGLDisplay *dpy, EGLint *major, EGLint *minor); +EGLBoolean drm_terminate(_EGLDriver *drv, _EGLDisplay *dpy); +_EGLContext *drm_create_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, _EGLContext *share_list, const EGLint *attrib_list); +EGLBoolean drm_destroy_context(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *context); +_EGLSurface *drm_create_window_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list); +_EGLSurface *drm_create_pixmap_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, NativePixmapType pixmap, const EGLint *attrib_list); +_EGLSurface *drm_create_pbuffer_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list); +_EGLSurface *drm_create_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, const EGLint *attrib_list); +EGLBoolean drm_show_screen_surface_mesa(_EGLDriver *drv, _EGLDisplay *dpy, _EGLScreen *screen, _EGLSurface *surface, _EGLMode *mode); +EGLBoolean drm_destroy_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface); +EGLBoolean drm_make_current(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw, _EGLSurface *read, _EGLContext *context); +EGLBoolean drm_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw); /*@}*/ #endif diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c index 16d4f1e32c9..20d682de3fb 100644 --- a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c +++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c @@ -82,7 +82,6 @@ static int vlResizeFrameBuffer template.width[0] = width; template.height[0] = height; template.depth[0] = 1; - template.compressed = 0; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET; @@ -669,7 +668,6 @@ static int vlInit sampler.compare_func = PIPE_FUNC_ALWAYS; sampler.normalized_coords = 1; /*sampler.prefilter = ;*/ - /*sampler.shadow_ambient = ;*/ /*sampler.lod_bias = ;*/ /*sampler.min_lod = ;*/ /*sampler.max_lod = ;*/ diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c index eb8270ecad3..23631adb693 100644 --- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c +++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c @@ -1074,7 +1074,6 @@ static int vlInit sampler.compare_func = PIPE_FUNC_ALWAYS; sampler.normalized_coords = 1; /*sampler.prefilter = ;*/ - /*sampler.shadow_ambient = ;*/ /*sampler.lod_bias = ;*/ sampler.min_lod = 0; /*sampler.max_lod = ;*/ @@ -1090,7 +1089,6 @@ static int vlInit template.width[0] = vlRoundUpPOT(mc->picture_width); template.height[0] = vlRoundUpPOT(mc->picture_height); template.depth[0] = 1; - template.compressed = 0; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_DYNAMIC; diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c index 92388f7978d..7f60852cae8 100644 --- a/src/gallium/state_trackers/g3dvl/vl_surface.c +++ b/src/gallium/state_trackers/g3dvl/vl_surface.c @@ -45,7 +45,6 @@ int vlCreateSurface template.width[0] = vlRoundUpPOT(sfc->width); template.height[0] = vlRoundUpPOT(sfc->height); template.depth[0] = 1; - template.compressed = 0; pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET; diff --git a/src/gallium/state_trackers/glx/xlib/Makefile b/src/gallium/state_trackers/glx/xlib/Makefile index 6d10b090aa0..7b2adc62c34 100644 --- a/src/gallium/state_trackers/glx/xlib/Makefile +++ b/src/gallium/state_trackers/glx/xlib/Makefile @@ -5,13 +5,12 @@ LIBNAME = xlib LIBRARY_INCLUDES = \ -I$(TOP)/include \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main + -I$(TOP)/src/mesa C_SOURCES = \ - glxapi.c \ - fakeglx.c \ - fakeglx_fonts.c \ + glx_api.c \ + glx_getproc.c \ + glx_usefont.c \ xm_api.c include ../../../Makefile.template diff --git a/src/gallium/state_trackers/glx/xlib/SConscript b/src/gallium/state_trackers/glx/xlib/SConscript index 0dbe3413972..04a44c30671 100644 --- a/src/gallium/state_trackers/glx/xlib/SConscript +++ b/src/gallium/state_trackers/glx/xlib/SConscript @@ -18,9 +18,10 @@ if env['platform'] == 'linux' \ st_xlib = env.ConvenienceLibrary( target = 'st_xlib', - source = [ 'glxapi.c', - 'fakeglx.c', - 'fakeglx_fonts.c', + source = [ + 'glx_api.c', + 'glx_getproc.c', + 'glx_usefont.c', 'xm_api.c', ] ) diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.h b/src/gallium/state_trackers/glx/xlib/fakeglx.h deleted file mode 100644 index e5fd960072e..00000000000 --- a/src/gallium/state_trackers/glx/xlib/fakeglx.h +++ /dev/null @@ -1,41 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 3.5 - * - * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef FAKEGLX_H -#define FAKEGLX_H - - -#include <X11/Xlib.h> - -struct _glxapi_table; - -extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); - -extern void Fake_glXUseXFont( Font font, int first, int count, int listbase ); - - -#endif - diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/glx_api.c index 85e7ecfb9e8..d1a98f89917 100644 --- a/src/gallium/state_trackers/glx/xlib/fakeglx.c +++ b/src/gallium/state_trackers/glx/xlib/glx_api.c @@ -1,8 +1,9 @@ /* * Mesa 3-D graphics library - * Version: 7.1 + * Version: 7.6 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2009 VMware, Inc. All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -23,31 +24,21 @@ */ -/* - * This is an emulation of the GLX API which allows Mesa/GLX-based programs - * to run on X servers which do not have the real GLX extension. - * - * Thanks to the contributors: - * - * Initial version: Philip Brown (phil@bolthole.com) - * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu) - * Further visual-handling refinements: Wolfram Gloger - * (wmglo@Dent.MED.Uni-Muenchen.DE). - * - * Notes: - * Don't be fooled, stereo isn't supported yet. +/** + * "Fake" GLX API implemented in terms of the XMesa*() functions. */ -#include "glxapi.h" +#define GLX_GLXEXT_PROTOTYPES +#include "GL/glx.h" + #include "xm_api.h" -#include "context.h" -#include "config.h" -#include "macros.h" -#include "imports.h" -#include "version.h" -#include "fakeglx.h" +#include "main/context.h" +#include "main/config.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/version.h" #include "state_tracker/st_context.h" #include "state_tracker/st_public.h" @@ -80,24 +71,42 @@ "GLX_SGIX_fbconfig " \ "GLX_SGIX_pbuffer " -/* - * Our fake GLX context will contain a "real" GLX context and an XMesa context. - * - * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context, - * and vice versa. - * - * We really just need this structure in order to make the libGL functions - * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay() - * work correctly. +#define DEFAULT_DIRECT GL_TRUE + + + +/** + * The GLXContext typedef is defined as a pointer to this structure. */ -struct fake_glx_context { - __GLXcontext glxContext; /* this MUST be first! */ +struct __GLXcontextRec +{ + Display *currentDpy; + GLboolean isDirect; + GLXDrawable currentDrawable; + GLXDrawable currentReadable; + XID xid; + XMesaContext xmesaContext; }; -#define DEFAULT_DIRECT GL_TRUE +static pipe_tsd ContextTSD; + +/** Set current context for calling thread */ +static void +SetCurrentContext(GLXContext c) +{ + pipe_tsd_set(&ContextTSD, c); +} + +/** Get current context for calling thread */ +static GLXContext +GetCurrentContext(void) +{ + return pipe_tsd_get(&ContextTSD); +} + /**********************************************************************/ @@ -366,10 +375,6 @@ find_glx_visual( Display *dpy, XVisualInfo *vinfo ) } - - - - /** * Try to get an X visual which matches the given arguments. */ @@ -418,7 +423,6 @@ get_visual( Display *dpy, int scr, unsigned int depth, int xclass ) } - /* * Retrieve the value of the given environment variable and find * the X visual which matches it. @@ -1002,8 +1006,8 @@ choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) } -static XVisualInfo * -Fake_glXChooseVisual( Display *dpy, int screen, int *list ) +XVisualInfo * +glXChooseVisual( Display *dpy, int screen, int *list ) { XMesaVisual xmvis; @@ -1024,18 +1028,18 @@ Fake_glXChooseVisual( Display *dpy, int screen, int *list ) } -static GLXContext -Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, - GLXContext share_list, Bool direct ) +GLXContext +glXCreateContext( Display *dpy, XVisualInfo *visinfo, + GLXContext share_list, Bool direct ) { XMesaVisual xmvis; - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + GLXContext glxCtx; + GLXContext shareCtx = share_list; if (!dpy || !visinfo) return 0; - glxCtx = CALLOC_STRUCT(fake_glx_context); + glxCtx = CALLOC_STRUCT(__GLXcontextRec); if (!glxCtx) return 0; @@ -1062,13 +1066,11 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, return NULL; } - glxCtx->glxContext.isDirect = DEFAULT_DIRECT; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ - - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + glxCtx->isDirect = DEFAULT_DIRECT; + glxCtx->currentDpy = dpy; + glxCtx->xid = (XID) glxCtx; /* self pointer */ - return (GLXContext) glxCtx; + return glxCtx; } @@ -1081,11 +1083,11 @@ static XMesaBuffer MakeCurrent_PrevReadBuffer = 0; /* GLX 1.3 and later */ -static Bool -Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext ctx ) +Bool +glXMakeContextCurrent( Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext ctx ) { - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + GLXContext glxCtx = ctx; static boolean firsttime = 1, no_rast = 0; if (firsttime) { @@ -1093,7 +1095,6 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, firsttime = 0; } - if (ctx && draw && read) { XMesaBuffer drawBuffer, readBuffer; XMesaContext xmctx = glxCtx->xmesaContext; @@ -1148,9 +1149,10 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, /* Now make current! */ if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) { - ((__GLXcontext *) ctx)->currentDpy = dpy; - ((__GLXcontext *) ctx)->currentDrawable = draw; - ((__GLXcontext *) ctx)->currentReadable = read; + ctx->currentDpy = dpy; + ctx->currentDrawable = draw; + ctx->currentReadable = read; + SetCurrentContext(ctx); return True; } else { @@ -1165,6 +1167,7 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, MakeCurrent_PrevReadable = 0; MakeCurrent_PrevDrawBuffer = 0; MakeCurrent_PrevReadBuffer = 0; + SetCurrentContext(NULL); return True; } else { @@ -1176,15 +1179,61 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, } -static Bool -Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ) +Bool +glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ) +{ + return glXMakeContextCurrent( dpy, drawable, drawable, ctx ); +} + + +GLXContext +glXGetCurrentContext(void) { - return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx ); + return GetCurrentContext(); } -static GLXPixmap -Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) +Display * +glXGetCurrentDisplay(void) +{ + GLXContext glxCtx = glXGetCurrentContext(); + + return glxCtx ? glxCtx->currentDpy : NULL; +} + + +Display * +glXGetCurrentDisplayEXT(void) +{ + return glXGetCurrentDisplay(); +} + + +GLXDrawable +glXGetCurrentDrawable(void) +{ + GLXContext gc = glXGetCurrentContext(); + return gc ? gc->currentDrawable : 0; +} + + +GLXDrawable +glXGetCurrentReadDrawable(void) +{ + GLXContext gc = glXGetCurrentContext(); + return gc ? gc->currentReadable : 0; +} + + +GLXDrawable +glXGetCurrentReadDrawableSGI(void) +{ + return glXGetCurrentReadDrawable(); +} + + +GLXPixmap +glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) { XMesaVisual v; XMesaBuffer b; @@ -1208,9 +1257,9 @@ Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) /*** GLX_MESA_pixmap_colormap ***/ -static GLXPixmap -Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, - Pixmap pixmap, Colormap cmap ) +GLXPixmap +glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, + Pixmap pixmap, Colormap cmap ) { XMesaVisual v; XMesaBuffer b; @@ -1232,8 +1281,8 @@ Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, } -static void -Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) +void +glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) { XMesaBuffer b = XMesaFindBuffer(dpy, pixmap); if (b) { @@ -1245,14 +1294,12 @@ Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) } -static void -Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, - unsigned long mask ) +void +glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, + unsigned long mask ) { - struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src; - struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst; - XMesaContext xm_src = fakeSrc->xmesaContext; - XMesaContext xm_dst = fakeDst->xmesaContext; + XMesaContext xm_src = src->xmesaContext; + XMesaContext xm_dst = dst->xmesaContext; (void) dpy; if (MakeCurrent_PrevContext == src) { _mesa_Flush(); @@ -1261,8 +1308,8 @@ Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, } -static Bool -Fake_glXQueryExtension( Display *dpy, int *errorb, int *event ) +Bool +glXQueryExtension( Display *dpy, int *errorb, int *event ) { /* Mesa's GLX isn't really an X extension but we try to act like one. */ (void) dpy; @@ -1272,18 +1319,10 @@ Fake_glXQueryExtension( Display *dpy, int *errorb, int *event ) } -extern void _kw_ungrab_all( Display *dpy ); -void _kw_ungrab_all( Display *dpy ) -{ - XUngrabPointer( dpy, CurrentTime ); - XUngrabKeyboard( dpy, CurrentTime ); -} - - -static void -Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) +void +glXDestroyContext( Display *dpy, GLXContext ctx ) { - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + GLXContext glxCtx = ctx; (void) dpy; MakeCurrent_PrevContext = 0; MakeCurrent_PrevDrawable = 0; @@ -1296,18 +1335,18 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) } -static Bool -Fake_glXIsDirect( Display *dpy, GLXContext ctx ) +Bool +glXIsDirect( Display *dpy, GLXContext ctx ) { - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + GLXContext glxCtx = ctx; (void) ctx; - return glxCtx->glxContext.isDirect; + return glxCtx->isDirect; } -static void -Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) +void +glXSwapBuffers( Display *dpy, GLXDrawable drawable ) { XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); static boolean firsttime = 1, no_rast = 0; @@ -1333,8 +1372,8 @@ Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) /*** GLX_MESA_copy_sub_buffer ***/ -static void -Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, +void +glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, int x, int y, int width, int height ) { XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); @@ -1347,8 +1386,8 @@ Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, } -static Bool -Fake_glXQueryVersion( Display *dpy, int *maj, int *min ) +Bool +glXQueryVersion( Display *dpy, int *maj, int *min ) { (void) dpy; /* Return GLX version, not Mesa version */ @@ -1564,8 +1603,8 @@ get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig ) } -static int -Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo, +int +glXGetConfig( Display *dpy, XVisualInfo *visinfo, int attrib, int *value ) { XMesaVisual xmvis; @@ -1594,8 +1633,8 @@ Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo, } -static void -Fake_glXWaitGL( void ) +void +glXWaitGL( void ) { XMesaContext xmesa = XMesaGetCurrentContext(); XMesaFlush( xmesa ); @@ -1603,8 +1642,8 @@ Fake_glXWaitGL( void ) -static void -Fake_glXWaitX( void ) +void +glXWaitX( void ) { XMesaContext xmesa = XMesaGetCurrentContext(); XMesaFlush( xmesa ); @@ -1620,8 +1659,8 @@ get_extensions( void ) /* GLX 1.1 and later */ -static const char * -Fake_glXQueryExtensionsString( Display *dpy, int screen ) +const char * +glXQueryExtensionsString( Display *dpy, int screen ) { (void) dpy; (void) screen; @@ -1631,8 +1670,8 @@ Fake_glXQueryExtensionsString( Display *dpy, int screen ) /* GLX 1.1 and later */ -static const char * -Fake_glXQueryServerString( Display *dpy, int screen, int name ) +const char * +glXQueryServerString( Display *dpy, int screen, int name ) { static char version[1000]; _mesa_sprintf(version, "%d.%d %s", @@ -1656,8 +1695,8 @@ Fake_glXQueryServerString( Display *dpy, int screen, int name ) /* GLX 1.1 and later */ -static const char * -Fake_glXGetClientString( Display *dpy, int name ) +const char * +glXGetClientString( Display *dpy, int name ) { static char version[1000]; _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION, @@ -1684,8 +1723,8 @@ Fake_glXGetClientString( Display *dpy, int name ) */ -static int -Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, +int +glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, int attribute, int *value ) { XMesaVisual v = (XMesaVisual) config; @@ -1699,8 +1738,8 @@ Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, } -static GLXFBConfig * -Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements ) +GLXFBConfig * +glXGetFBConfigs( Display *dpy, int screen, int *nelements ) { XVisualInfo *visuals, visTemplate; const long visMask = VisualScreenMask; @@ -1725,15 +1764,15 @@ Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements ) } -static GLXFBConfig * -Fake_glXChooseFBConfig( Display *dpy, int screen, +GLXFBConfig * +glXChooseFBConfig( Display *dpy, int screen, const int *attribList, int *nitems ) { XMesaVisual xmvis; if (!attribList || !attribList[0]) { /* return list of all configs (per GLX_SGIX_fbconfig spec) */ - return Fake_glXGetFBConfigs(dpy, screen, nitems); + return glXGetFBConfigs(dpy, screen, nitems); } xmvis = choose_visual(dpy, screen, attribList, GL_TRUE); @@ -1754,8 +1793,8 @@ Fake_glXChooseFBConfig( Display *dpy, int screen, } -static XVisualInfo * -Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) +XVisualInfo * +glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) { if (dpy && config) { XMesaVisual xmvis = (XMesaVisual) config; @@ -1776,8 +1815,8 @@ Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) } -static GLXWindow -Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, +GLXWindow +glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, const int *attribList ) { XMesaVisual xmvis = (XMesaVisual) config; @@ -1796,8 +1835,8 @@ Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, } -static void -Fake_glXDestroyWindow( Display *dpy, GLXWindow window ) +void +glXDestroyWindow( Display *dpy, GLXWindow window ) { XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable) window); if (b) @@ -1807,8 +1846,8 @@ Fake_glXDestroyWindow( Display *dpy, GLXWindow window ) /* XXX untested */ -static GLXPixmap -Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, +GLXPixmap +glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList ) { XMesaVisual v = (XMesaVisual) config; @@ -1820,7 +1859,7 @@ Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, if (!dpy || !config || !pixmap) return 0; - for (attr = attribList; *attr; attr++) { + for (attr = attribList; attr && *attr; attr++) { switch (*attr) { case GLX_TEXTURE_FORMAT_EXT: attr++; @@ -1917,8 +1956,8 @@ Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, } -static void -Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) +void +glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) { XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable)pixmap); if (b) @@ -1927,8 +1966,8 @@ Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) } -static GLXPbuffer -Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, +GLXPbuffer +glXCreatePbuffer( Display *dpy, GLXFBConfig config, const int *attribList ) { XMesaVisual xmvis = (XMesaVisual) config; @@ -1980,8 +2019,8 @@ Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, } -static void -Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) +void +glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) { XMesaBuffer b = XMesaFindBuffer(dpy, pbuf); if (b) { @@ -1990,8 +2029,8 @@ Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) } -static void -Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, +void +glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, unsigned int *value ) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw); @@ -2032,19 +2071,19 @@ Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, } -static GLXContext -Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, +GLXContext +glXCreateNewContext( Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct ) { - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList; + GLXContext glxCtx; + GLXContext shareCtx = shareList; XMesaVisual xmvis = (XMesaVisual) config; if (!dpy || !config || (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE)) return 0; - glxCtx = CALLOC_STRUCT(fake_glx_context); + glxCtx = CALLOC_STRUCT(__GLXcontextRec); if (!glxCtx) return 0; @@ -2058,20 +2097,18 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, return NULL; } - glxCtx->glxContext.isDirect = DEFAULT_DIRECT; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + glxCtx->isDirect = DEFAULT_DIRECT; + glxCtx->currentDpy = dpy; + glxCtx->xid = (XID) glxCtx; /* self pointer */ - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); - - return (GLXContext) glxCtx; + return glxCtx; } -static int -Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) +int +glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) { - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + GLXContext glxCtx = ctx; XMesaContext xmctx = glxCtx->xmesaContext; (void) dpy; @@ -2097,8 +2134,8 @@ Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) } -static void -Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) +void +glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); if (xmbuf) @@ -2106,8 +2143,8 @@ Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) } -static void -Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, +void +glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, unsigned long *mask ) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); @@ -2121,8 +2158,8 @@ Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, /*** GLX_SGI_swap_control ***/ -static int -Fake_glXSwapIntervalSGI(int interval) +int +glXSwapIntervalSGI(int interval) { (void) interval; return 0; @@ -2134,16 +2171,16 @@ Fake_glXSwapIntervalSGI(int interval) static unsigned int FrameCounter = 0; -static int -Fake_glXGetVideoSyncSGI(unsigned int *count) +int +glXGetVideoSyncSGI(unsigned int *count) { /* this is a bogus implementation */ *count = FrameCounter++; return 0; } -static int -Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) +int +glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) { if (divisor <= 0 || remainder < 0) return GLX_BAD_VALUE; @@ -2159,15 +2196,15 @@ Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) /*** GLX_SGI_make_current_read ***/ -static Bool -Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +Bool +glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) { - return Fake_glXMakeContextCurrent( dpy, draw, read, ctx ); + return glXMakeContextCurrent( dpy, draw, read, ctx ); } /* not used static GLXDrawable -Fake_glXGetCurrentReadDrawableSGI(void) +glXGetCurrentReadDrawableSGI(void) { return 0; } @@ -2177,8 +2214,8 @@ Fake_glXGetCurrentReadDrawableSGI(void) /*** GLX_SGIX_video_source ***/ #if defined(_VL_H) -static GLXVideoSourceSGIX -Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) +GLXVideoSourceSGIX +glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) { (void) dpy; (void) screen; @@ -2189,8 +2226,8 @@ Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPa return 0; } -static void -Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) +void +glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) { (void) dpy; (void) src; @@ -2201,30 +2238,30 @@ Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) /*** GLX_EXT_import_context ***/ -static void -Fake_glXFreeContextEXT(Display *dpy, GLXContext context) +void +glXFreeContextEXT(Display *dpy, GLXContext context) { (void) dpy; (void) context; } -static GLXContextID -Fake_glXGetContextIDEXT(const GLXContext context) +GLXContextID +glXGetContextIDEXT(const GLXContext context) { (void) context; return 0; } -static GLXContext -Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID) +GLXContext +glXImportContextEXT(Display *dpy, GLXContextID contextID) { (void) dpy; (void) contextID; return 0; } -static int -Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value) +int +glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value) { (void) dpy; (void) context; @@ -2237,21 +2274,21 @@ Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int /*** GLX_SGIX_fbconfig ***/ -static int -Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) +int +glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) { - return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value); + return glXGetFBConfigAttrib(dpy, config, attribute, value); } -static GLXFBConfigSGIX * -Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) +GLXFBConfigSGIX * +glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) { - return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements); + return (GLXFBConfig *) glXChooseFBConfig(dpy, screen, attrib_list, nelements); } -static GLXPixmap -Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) +GLXPixmap +glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) { XMesaVisual xmvis = (XMesaVisual) config; XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0); @@ -2259,14 +2296,14 @@ Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixm } -static GLXContext -Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) +GLXContext +glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) { XMesaVisual xmvis = (XMesaVisual) config; - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + GLXContext glxCtx; + GLXContext shareCtx = share_list; - glxCtx = CALLOC_STRUCT(fake_glx_context); + glxCtx = CALLOC_STRUCT(__GLXcontextRec); if (!glxCtx) return 0; @@ -2280,25 +2317,23 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re return NULL; } - glxCtx->glxContext.isDirect = DEFAULT_DIRECT; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ - - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + glxCtx->isDirect = DEFAULT_DIRECT; + glxCtx->currentDpy = dpy; + glxCtx->xid = (XID) glxCtx; /* self pointer */ - return (GLXContext) glxCtx; + return glxCtx; } -static XVisualInfo * -Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) +XVisualInfo * +glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) { - return Fake_glXGetVisualFromFBConfig(dpy, config); + return glXGetVisualFromFBConfig(dpy, config); } -static GLXFBConfigSGIX -Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) +GLXFBConfigSGIX +glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) { XMesaVisual xmvis = find_glx_visual(dpy, vis); if (!xmvis) { @@ -2313,8 +2348,8 @@ Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) /*** GLX_SGIX_pbuffer ***/ -static GLXPbufferSGIX -Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, +GLXPbufferSGIX +glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attribList) { @@ -2352,8 +2387,8 @@ Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, } -static void -Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) +void +glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); if (xmbuf) { @@ -2362,8 +2397,8 @@ Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) } -static int -Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) +int +glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) { const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); @@ -2395,8 +2430,8 @@ Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, un } -static void -Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) +void +glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); if (xmbuf) { @@ -2406,8 +2441,8 @@ Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) } -static void -Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) +void +glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) { XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); if (xmbuf) { @@ -2422,8 +2457,8 @@ Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long * /*** GLX_SGI_cushion ***/ -static void -Fake_glXCushionSGI(Display *dpy, Window win, float cushion) +void +glXCushionSGI(Display *dpy, Window win, float cushion) { (void) dpy; (void) win; @@ -2434,8 +2469,8 @@ Fake_glXCushionSGI(Display *dpy, Window win, float cushion) /*** GLX_SGIX_video_resize ***/ -static int -Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) +int +glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) { (void) dpy; (void) screen; @@ -2444,8 +2479,8 @@ Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window w return 0; } -static int -Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) +int +glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) { (void) dpy; (void) screen; @@ -2457,8 +2492,8 @@ Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int return 0; } -static int -Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) +int +glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) { (void) dpy; (void) screen; @@ -2470,8 +2505,8 @@ Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int return 0; } -static int -Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) +int +glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) { (void) dpy; (void) screen; @@ -2483,8 +2518,8 @@ Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, i return 0; } -static int -Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) +int +glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) { (void) dpy; (void) screen; @@ -2498,8 +2533,8 @@ Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum syncty /*** GLX_SGIX_dmbuffer **/ #if defined(_DM_BUFFER_H_) -static Bool -Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) +Bool +glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) { (void) dpy; (void) pbuffer; @@ -2512,8 +2547,8 @@ Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *p /*** GLX_SGIX_swap_group ***/ -static void -Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) +void +glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) { (void) dpy; (void) drawable; @@ -2524,16 +2559,16 @@ Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member /*** GLX_SGIX_swap_barrier ***/ -static void -Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) +void +glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) { (void) dpy; (void) drawable; (void) barrier; } -static Bool -Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) +Bool +glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) { (void) dpy; (void) screen; @@ -2545,8 +2580,8 @@ Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) /*** GLX_SUN_get_transparent_index ***/ -static Status -Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) +Status +glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) { (void) dpy; (void) overlay; @@ -2563,8 +2598,8 @@ Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, lo * Release the depth, stencil, accum buffers attached to a GLXDrawable * (a window or pixmap) prior to destroying the GLXDrawable. */ -static Bool -Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) +Bool +glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) { XMesaBuffer b = XMesaFindBuffer(dpy, d); if (b) { @@ -2576,8 +2611,8 @@ Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) /*** GLX_EXT_texture_from_pixmap ***/ -static void -Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, +void +glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, const int *attrib_list) { XMesaBuffer b = XMesaFindBuffer(dpy, drawable); @@ -2585,162 +2620,10 @@ Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, XMesaBindTexImage(dpy, b, buffer, attrib_list); } -static void -Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) +void +glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) { XMesaBuffer b = XMesaFindBuffer(dpy, drawable); if (b) XMesaReleaseTexImage(dpy, b, buffer); } - - - -/** - * Create a new GLX API dispatch table with its function pointers - * initialized to point to Mesa's "fake" GLX API functions. - * - * Note: there used to be a similar function - * (_real_GetGLXDispatchTable) that returns a new dispatch table with - * all pointers initalized to point to "real" GLX functions (which - * understand GLX wire protocol, etc). - */ -struct _glxapi_table * -_mesa_GetGLXDispatchTable(void) -{ - static struct _glxapi_table glx; - - /* be sure our dispatch table size <= libGL's table */ - { - GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); - (void) size; - assert(_glxapi_get_dispatch_table_size() >= size); - } - - /* initialize the whole table to no-ops */ - _glxapi_set_no_op_table(&glx); - - /* now initialize the table with the functions I implement */ - glx.ChooseVisual = Fake_glXChooseVisual; - glx.CopyContext = Fake_glXCopyContext; - glx.CreateContext = Fake_glXCreateContext; - glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap; - glx.DestroyContext = Fake_glXDestroyContext; - glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap; - glx.GetConfig = Fake_glXGetConfig; - /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/ - /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/ - glx.IsDirect = Fake_glXIsDirect; - glx.MakeCurrent = Fake_glXMakeCurrent; - glx.QueryExtension = Fake_glXQueryExtension; - glx.QueryVersion = Fake_glXQueryVersion; - glx.SwapBuffers = Fake_glXSwapBuffers; - glx.UseXFont = Fake_glXUseXFont; - glx.WaitGL = Fake_glXWaitGL; - glx.WaitX = Fake_glXWaitX; - - /*** GLX_VERSION_1_1 ***/ - glx.GetClientString = Fake_glXGetClientString; - glx.QueryExtensionsString = Fake_glXQueryExtensionsString; - glx.QueryServerString = Fake_glXQueryServerString; - - /*** GLX_VERSION_1_2 ***/ - /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/ - - /*** GLX_VERSION_1_3 ***/ - glx.ChooseFBConfig = Fake_glXChooseFBConfig; - glx.CreateNewContext = Fake_glXCreateNewContext; - glx.CreatePbuffer = Fake_glXCreatePbuffer; - glx.CreatePixmap = Fake_glXCreatePixmap; - glx.CreateWindow = Fake_glXCreateWindow; - glx.DestroyPbuffer = Fake_glXDestroyPbuffer; - glx.DestroyPixmap = Fake_glXDestroyPixmap; - glx.DestroyWindow = Fake_glXDestroyWindow; - /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/ - glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib; - glx.GetFBConfigs = Fake_glXGetFBConfigs; - glx.GetSelectedEvent = Fake_glXGetSelectedEvent; - glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig; - glx.MakeContextCurrent = Fake_glXMakeContextCurrent; - glx.QueryContext = Fake_glXQueryContext; - glx.QueryDrawable = Fake_glXQueryDrawable; - glx.SelectEvent = Fake_glXSelectEvent; - - /*** GLX_SGI_swap_control ***/ - glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI; - - /*** GLX_SGI_video_sync ***/ - glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI; - glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI; - - /*** GLX_SGI_make_current_read ***/ - glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI; - /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/ - -/*** GLX_SGIX_video_source ***/ -#if defined(_VL_H) - glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX; - glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX; -#endif - - /*** GLX_EXT_import_context ***/ - glx.FreeContextEXT = Fake_glXFreeContextEXT; - glx.GetContextIDEXT = Fake_glXGetContextIDEXT; - /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/ - glx.ImportContextEXT = Fake_glXImportContextEXT; - glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT; - - /*** GLX_SGIX_fbconfig ***/ - glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX; - glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX; - glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX; - glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX; - glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX; - glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX; - - /*** GLX_SGIX_pbuffer ***/ - glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX; - glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX; - glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX; - glx.SelectEventSGIX = Fake_glXSelectEventSGIX; - glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX; - - /*** GLX_SGI_cushion ***/ - glx.CushionSGI = Fake_glXCushionSGI; - - /*** GLX_SGIX_video_resize ***/ - glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX; - glx.ChannelRectSGIX = Fake_glXChannelRectSGIX; - glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX; - glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX; - glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX; - - /*** GLX_SGIX_dmbuffer **/ -#if defined(_DM_BUFFER_H_) - glx.AssociateDMPbufferSGIX = NULL; -#endif - - /*** GLX_SGIX_swap_group ***/ - glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX; - - /*** GLX_SGIX_swap_barrier ***/ - glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX; - glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX; - - /*** GLX_SUN_get_transparent_index ***/ - glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN; - - /*** GLX_MESA_copy_sub_buffer ***/ - glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA; - - /*** GLX_MESA_release_buffers ***/ - glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA; - - /*** GLX_MESA_pixmap_colormap ***/ - glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA; - - /*** GLX_EXT_texture_from_pixmap ***/ - glx.BindTexImageEXT = Fake_glXBindTexImageEXT; - glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT; - - return &glx; -} diff --git a/src/gallium/state_trackers/glx/xlib/glx_getproc.c b/src/gallium/state_trackers/glx/xlib/glx_getproc.c new file mode 100644 index 00000000000..ca7d88c9227 --- /dev/null +++ b/src/gallium/state_trackers/glx/xlib/glx_getproc.c @@ -0,0 +1,214 @@ +/* + * Mesa 3-D graphics library + * Version: 7.6 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * Copyright (C) 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * glXGetProcAddress() + */ + + +#define GLX_GLXEXT_PROTOTYPES + +#include <string.h> +#include "GL/glx.h" +#include "glapi/glapi.h" + + +struct name_address_pair { + const char *Name; + __GLXextFuncPtr Address; +}; + + +static struct name_address_pair GLX_functions[] = { + /*** GLX_VERSION_1_0 ***/ + { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual }, + { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext }, + { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext }, + { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap }, + { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext }, + { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap }, + { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig }, + { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext }, + { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable }, + { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect }, + { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent }, + { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension }, + { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion }, + { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers }, + { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont }, + { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL }, + { "glXWaitX", (__GLXextFuncPtr) glXWaitX }, + + /*** GLX_VERSION_1_1 ***/ + { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString }, + { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString }, + { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString }, + + /*** GLX_VERSION_1_2 ***/ + { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay }, + + /*** GLX_VERSION_1_3 ***/ + { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig }, + { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext }, + { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer }, + { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap }, + { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow }, + { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer }, + { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap }, + { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow }, + { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable }, + { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib }, + { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs }, + { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent }, + { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig }, + { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent }, + { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext }, + { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable }, + { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent }, + + /*** GLX_VERSION_1_4 ***/ + { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress }, + + /*** GLX_SGI_swap_control ***/ + { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI }, + + /*** GLX_SGI_video_sync ***/ + { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI }, + { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI }, + + /*** GLX_SGI_make_current_read ***/ + { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI }, + { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI }, + + /*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX }, + { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX }, +#endif + + /*** GLX_EXT_import_context ***/ + { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT }, + { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT }, + { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT }, + { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT }, + { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT }, + + /*** GLX_SGIX_fbconfig ***/ + { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX }, + { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX }, + { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX }, + { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX }, + { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX }, + { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX }, + + /*** GLX_SGIX_pbuffer ***/ + { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX }, + { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX }, + { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX }, + { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX }, + { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX }, + + /*** GLX_SGI_cushion ***/ + { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI }, + + /*** GLX_SGIX_video_resize ***/ + { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX }, + { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX }, + { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX }, + { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX }, + { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX }, + + /*** GLX_SGIX_dmbuffer **/ +#if defined(_DM_BUFFER_H_) + { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX }, +#endif + + /*** GLX_SGIX_swap_group ***/ + { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX }, + + /*** GLX_SGIX_swap_barrier ***/ + { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX }, + { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX }, + + /*** GLX_SUN_get_transparent_index ***/ + { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN }, + + /*** GLX_MESA_copy_sub_buffer ***/ + { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA }, + + /*** GLX_MESA_pixmap_colormap ***/ + { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA }, + + /*** GLX_MESA_release_buffers ***/ + { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA }, + + /*** GLX_ARB_get_proc_address ***/ + { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB }, + + /*** GLX_EXT_texture_from_pixmap ***/ + { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT }, + { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT }, + + { NULL, NULL } /* end of list */ +}; + + + +/** + * Return address of named glX function, or NULL if not found. + */ +static __GLXextFuncPtr +_glxapi_get_proc_address(const char *funcName) +{ + GLuint i; + for (i = 0; GLX_functions[i].Name; i++) { + if (strcmp(GLX_functions[i].Name, funcName) == 0) + return GLX_functions[i].Address; + } + return NULL; +} + + +__GLXextFuncPtr +glXGetProcAddressARB(const GLubyte *procName) +{ + __GLXextFuncPtr f; + + f = _glxapi_get_proc_address((const char *) procName); + if (f) { + return f; + } + + f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName); + return f; +} + + +/* GLX 1.4 */ +void (*glXGetProcAddress(const GLubyte *procName))() +{ + return glXGetProcAddressARB(procName); +} diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c b/src/gallium/state_trackers/glx/xlib/glx_usefont.c index e3590467563..acc64df62b6 100644 --- a/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c +++ b/src/gallium/state_trackers/glx/xlib/glx_usefont.c @@ -1,9 +1,10 @@ - /* * Mesa 3-D graphics library - * Version: 3.5 + * Version: 7.6 * - * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. + * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * Copyright (C) 2009 VMware, Inc. All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -24,13 +25,13 @@ */ -/* xfonts.c -- glXUseXFont() for Mesa written by - * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de +/** + * Fake implementation of glXUseXFont(). */ -#include "context.h" -#include "imports.h" -#include "fakeglx.h" + +#include "main/context.h" +#include "main/imports.h" #include <GL/glx.h> @@ -210,7 +211,7 @@ isvalid(XFontStruct * fs, unsigned int which) void -Fake_glXUseXFont(Font font, int first, int count, int listbase) +glXUseXFont(Font font, int first, int count, int listbase) { Display *dpy; Window win; @@ -228,7 +229,8 @@ Fake_glXUseXFont(Font font, int first, int count, int listbase) dpy = glXGetCurrentDisplay(); if (!dpy) return; /* I guess glXMakeCurrent wasn't called */ - win = RootWindow(dpy, DefaultScreen(dpy)); + i = DefaultScreen(dpy); + win = RootWindow(dpy, i); fs = XQueryFont(dpy, font); if (!fs) { diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.c b/src/gallium/state_trackers/glx/xlib/glxapi.c deleted file mode 100644 index c2cb34d7cf7..00000000000 --- a/src/gallium/state_trackers/glx/xlib/glxapi.c +++ /dev/null @@ -1,1254 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * This is the GLX API dispatcher. Calls to the glX* functions are - * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions. - * See the glxapi.h file for more details. - */ - - -#include <assert.h> -#include <stdlib.h> -#include <stdio.h> -#include <string.h> -#include "glapi/glapi.h" -#include "glxapi.h" -#include "fakeglx.h" -#include "pipe/p_thread.h" - - -#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 303 -# define PUBLIC __attribute__((visibility("default"))) -# define USED __attribute__((used)) -#else -# define PUBLIC -# define USED -#endif - - -struct display_dispatch { - Display *Dpy; - struct _glxapi_table *Table; - struct display_dispatch *Next; -}; - -static struct display_dispatch *DispatchList = NULL; - - -/* Display -> Dispatch caching */ -static Display *prevDisplay = NULL; -static struct _glxapi_table *prevTable = NULL; - - -static struct _glxapi_table * -get_dispatch(Display *dpy) -{ - if (!dpy) - return NULL; - - /* search list of display/dispatch pairs for this display */ - { - const struct display_dispatch *d = DispatchList; - while (d) { - if (d->Dpy == dpy) { - prevDisplay = dpy; - prevTable = d->Table; - return d->Table; /* done! */ - } - d = d->Next; - } - } - - /* A new display, determine if we should use real GLX - * or Mesa's pseudo-GLX. - */ - { - struct _glxapi_table *t = _mesa_GetGLXDispatchTable(); - - if (t) { - struct display_dispatch *d; - d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch)); - if (d) { - d->Dpy = dpy; - d->Table = t; - /* insert at head of list */ - d->Next = DispatchList; - DispatchList = d; - /* update cache */ - prevDisplay = dpy; - prevTable = t; - return t; - } - } - } - - /* If we get here that means we can't use real GLX on this display - * and the Mesa pseudo-GLX software renderer wasn't compiled in. - * Or, we ran out of memory! - */ - return NULL; -} - - -/* Don't use the GET_DISPATCH defined in glthread.h */ -#undef GET_DISPATCH - -#define GET_DISPATCH(DPY, TABLE) \ - if (DPY == prevDisplay) { \ - TABLE = prevTable; \ - } \ - else if (!DPY) { \ - TABLE = NULL; \ - } \ - else { \ - TABLE = get_dispatch(DPY); \ - } - - - - -/** - * GLX API current context. - */ -pipe_tsd ContextTSD; - - -static void -SetCurrentContext(GLXContext c) -{ - pipe_tsd_set(&ContextTSD, c); -} - - -/* - * GLX API entrypoints - */ - -/*** GLX_VERSION_1_0 ***/ - -XVisualInfo PUBLIC * -glXChooseVisual(Display *dpy, int screen, int *list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->ChooseVisual)(dpy, screen, list); -} - - -void PUBLIC -glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CopyContext)(dpy, src, dst, mask); -} - - -GLXContext PUBLIC -glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateContext)(dpy, visinfo, shareList, direct); -} - - -GLXPixmap PUBLIC -glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmap)(dpy, visinfo, pixmap); -} - - -void PUBLIC -glXDestroyContext(Display *dpy, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - if (glXGetCurrentContext() == ctx) - SetCurrentContext(NULL); - (t->DestroyContext)(dpy, ctx); -} - - -void PUBLIC -glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyGLXPixmap)(dpy, pixmap); -} - - -int PUBLIC -glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return GLX_NO_EXTENSION; - return (t->GetConfig)(dpy, visinfo, attrib, value); -} - - -GLXContext PUBLIC -glXGetCurrentContext(void) -{ - return (GLXContext) pipe_tsd_get(&ContextTSD); -} - - -GLXDrawable PUBLIC -glXGetCurrentDrawable(void) -{ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - return gc ? gc->currentDrawable : 0; -} - - -Bool PUBLIC -glXIsDirect(Display *dpy, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->IsDirect)(dpy, ctx); -} - - -Bool PUBLIC -glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx) -{ - Bool b; - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) { - return False; - } - b = (*t->MakeCurrent)(dpy, drawable, ctx); - if (b) { - SetCurrentContext(ctx); - } - return b; -} - - -Bool PUBLIC -glXQueryExtension(Display *dpy, int *errorb, int *event) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->QueryExtension)(dpy, errorb, event); -} - - -Bool PUBLIC -glXQueryVersion(Display *dpy, int *maj, int *min) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->QueryVersion)(dpy, maj, min); -} - - -void PUBLIC -glXSwapBuffers(Display *dpy, GLXDrawable drawable) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SwapBuffers)(dpy, drawable); -} - - -void PUBLIC -glXUseXFont(Font font, int first, int count, int listBase) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->UseXFont)(font, first, count, listBase); -} - - -void PUBLIC -glXWaitGL(void) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->WaitGL)(); -} - - -void PUBLIC -glXWaitX(void) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->WaitX)(); -} - - - -/*** GLX_VERSION_1_1 ***/ - -const char PUBLIC * -glXGetClientString(Display *dpy, int name) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->GetClientString)(dpy, name); -} - - -const char PUBLIC * -glXQueryExtensionsString(Display *dpy, int screen) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->QueryExtensionsString)(dpy, screen); -} - - -const char PUBLIC * -glXQueryServerString(Display *dpy, int screen, int name) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->QueryServerString)(dpy, screen, name); -} - - -/*** GLX_VERSION_1_2 ***/ - -Display PUBLIC * -glXGetCurrentDisplay(void) -{ - /* Same code as in libGL's glxext.c */ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - if (NULL == gc) return NULL; - return gc->currentDpy; -} - - - -/*** GLX_VERSION_1_3 ***/ - -GLXFBConfig PUBLIC * -glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChooseFBConfig)(dpy, screen, attribList, nitems); -} - - -GLXContext PUBLIC -glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateNewContext)(dpy, config, renderType, shareList, direct); -} - - -GLXPbuffer PUBLIC -glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreatePbuffer)(dpy, config, attribList); -} - - -GLXPixmap PUBLIC -glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreatePixmap)(dpy, config, pixmap, attribList); -} - - -GLXWindow PUBLIC -glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateWindow)(dpy, config, win, attribList); -} - - -void PUBLIC -glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyPbuffer)(dpy, pbuf); -} - - -void PUBLIC -glXDestroyPixmap(Display *dpy, GLXPixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyPixmap)(dpy, pixmap); -} - - -void PUBLIC -glXDestroyWindow(Display *dpy, GLXWindow window) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyWindow)(dpy, window); -} - - -GLXDrawable PUBLIC -glXGetCurrentReadDrawable(void) -{ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - return gc ? gc->currentReadable : 0; -} - - -int PUBLIC -glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return GLX_NO_EXTENSION; - return (t->GetFBConfigAttrib)(dpy, config, attribute, value); -} - - -GLXFBConfig PUBLIC * -glXGetFBConfigs(Display *dpy, int screen, int *nelements) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigs)(dpy, screen, nelements); -} - -void PUBLIC -glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->GetSelectedEvent)(dpy, drawable, mask); -} - - -XVisualInfo PUBLIC * -glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->GetVisualFromFBConfig)(dpy, config); -} - - -Bool PUBLIC -glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) -{ - Bool b; - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - b = (t->MakeContextCurrent)(dpy, draw, read, ctx); - if (b) { - SetCurrentContext(ctx); - } - return b; -} - - -int PUBLIC -glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - assert(t); - if (!t) - return 0; /* XXX correct? */ - return (t->QueryContext)(dpy, ctx, attribute, value); -} - - -void PUBLIC -glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->QueryDrawable)(dpy, draw, attribute, value); -} - - -void PUBLIC -glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SelectEvent)(dpy, drawable, mask); -} - - - -/*** GLX_SGI_swap_control ***/ - -int PUBLIC -glXSwapIntervalSGI(int interval) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->SwapIntervalSGI)(interval); -} - - - -/*** GLX_SGI_video_sync ***/ - -int PUBLIC -glXGetVideoSyncSGI(unsigned int *count) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t || !glXGetCurrentContext()) - return GLX_BAD_CONTEXT; - return (t->GetVideoSyncSGI)(count); -} - -int PUBLIC -glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t || !glXGetCurrentContext()) - return GLX_BAD_CONTEXT; - return (t->WaitVideoSyncSGI)(divisor, remainder, count); -} - - - -/*** GLX_SGI_make_current_read ***/ - -Bool PUBLIC -glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx); -} - -GLXDrawable PUBLIC -glXGetCurrentReadDrawableSGI(void) -{ - return glXGetCurrentReadDrawable(); -} - - -#if defined(_VL_H) - -GLXVideoSourceSGIX PUBLIC -glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode); -} - -void PUBLIC -glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->DestroyGLXVideoSourceSGIX)(dpy, src); -} - -#endif - - -/*** GLX_EXT_import_context ***/ - -void PUBLIC -glXFreeContextEXT(Display *dpy, GLXContext context) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->FreeContextEXT)(dpy, context); -} - -GLXContextID PUBLIC -glXGetContextIDEXT(const GLXContext context) -{ - return ((__GLXcontext *) context)->xid; -} - -Display PUBLIC * -glXGetCurrentDisplayEXT(void) -{ - return glXGetCurrentDisplay(); -} - -GLXContext PUBLIC -glXImportContextEXT(Display *dpy, GLXContextID contextID) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ImportContextEXT)(dpy, contextID); -} - -int PUBLIC -glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; /* XXX ok? */ - return (t->QueryContextInfoEXT)(dpy, context, attribute, value); -} - - - -/*** GLX_SGIX_fbconfig ***/ - -int PUBLIC -glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value); -} - -GLXFBConfigSGIX PUBLIC * -glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements); -} - -GLXPixmap PUBLIC -glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap); -} - -GLXContext PUBLIC -glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct); -} - -XVisualInfo PUBLIC * -glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetVisualFromFBConfigSGIX)(dpy, config); -} - -GLXFBConfigSGIX PUBLIC -glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigFromVisualSGIX)(dpy, vis); -} - - - -/*** GLX_SGIX_pbuffer ***/ - -GLXPbufferSGIX PUBLIC -glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list); -} - -void PUBLIC -glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyGLXPbufferSGIX)(dpy, pbuf); -} - -int PUBLIC -glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value); -} - -void PUBLIC -glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SelectEventSGIX)(dpy, drawable, mask); -} - -void PUBLIC -glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->GetSelectedEventSGIX)(dpy, drawable, mask); -} - - - -/*** GLX_SGI_cushion ***/ - -void PUBLIC -glXCushionSGI(Display *dpy, Window win, float cushion) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CushionSGI)(dpy, win, cushion); -} - - - -/*** GLX_SGIX_video_resize ***/ - -int PUBLIC -glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window); -} - -int PUBLIC -glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h); -} - -int PUBLIC -glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h); -} - -int PUBLIC -glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh); -} - -int PUBLIC -glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype); -} - - - -#if defined(_DM_BUFFER_H_) - -Bool PUBLIC -glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer); -} - -#endif - - -/*** GLX_SGIX_swap_group ***/ - -void PUBLIC -glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (*t->JoinSwapGroupSGIX)(dpy, drawable, member); -} - - -/*** GLX_SGIX_swap_barrier ***/ - -void PUBLIC -glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier); -} - -Bool PUBLIC -glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max); -} - - - -/*** GLX_SUN_get_transparent_index ***/ - -Status PUBLIC -glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent); -} - - - -/*** GLX_MESA_copy_sub_buffer ***/ - -void PUBLIC -glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height); -} - - - -/*** GLX_MESA_release_buffers ***/ - -Bool PUBLIC -glXReleaseBuffersMESA(Display *dpy, Window w) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->ReleaseBuffersMESA)(dpy, w); -} - - - -/*** GLX_MESA_pixmap_colormap ***/ - -GLXPixmap PUBLIC -glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap); -} - -/*** GLX_EXT_texture_from_pixmap */ - -void -glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, - const int *attrib_list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (t) - t->BindTexImageEXT(dpy, drawable, buffer, attrib_list); -} - -void -glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (t) - t->ReleaseTexImageEXT(dpy, drawable, buffer); -} - - -/**********************************************************************/ -/* GLX API management functions */ -/**********************************************************************/ - - -const char * -_glxapi_get_version(void) -{ - return "1.3"; -} - - - -/* - * Return size of the GLX dispatch table, in entries, not bytes. - */ -GLuint -_glxapi_get_dispatch_table_size(void) -{ - return sizeof(struct _glxapi_table) / sizeof(void *); -} - - -static int -generic_no_op_func(void) -{ - return 0; -} - - -/* - * Initialize all functions in given dispatch table to be no-ops - */ -void -_glxapi_set_no_op_table(struct _glxapi_table *t) -{ - typedef int (*nop_func)(void); - nop_func *dispatch = (nop_func *) t; - GLuint n = _glxapi_get_dispatch_table_size(); - GLuint i; - for (i = 0; i < n; i++) { - dispatch[i] = generic_no_op_func; - } -} - - -struct name_address_pair { - const char *Name; - __GLXextFuncPtr Address; -}; - -static struct name_address_pair GLX_functions[] = { - /*** GLX_VERSION_1_0 ***/ - { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual }, - { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext }, - { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext }, - { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap }, - { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext }, - { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap }, - { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig }, - { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext }, - { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable }, - { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect }, - { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent }, - { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension }, - { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion }, - { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers }, - { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont }, - { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL }, - { "glXWaitX", (__GLXextFuncPtr) glXWaitX }, - - /*** GLX_VERSION_1_1 ***/ - { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString }, - { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString }, - { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString }, - - /*** GLX_VERSION_1_2 ***/ - { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay }, - - /*** GLX_VERSION_1_3 ***/ - { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig }, - { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext }, - { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer }, - { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap }, - { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow }, - { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer }, - { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap }, - { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow }, - { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable }, - { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib }, - { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs }, - { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent }, - { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig }, - { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent }, - { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext }, - { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable }, - { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent }, - - /*** GLX_VERSION_1_4 ***/ - { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress }, - - /*** GLX_SGI_swap_control ***/ - { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI }, - - /*** GLX_SGI_video_sync ***/ - { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI }, - { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI }, - - /*** GLX_SGI_make_current_read ***/ - { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI }, - { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI }, - - /*** GLX_SGIX_video_source ***/ -#if defined(_VL_H) - { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX }, - { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX }, -#endif - - /*** GLX_EXT_import_context ***/ - { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT }, - { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT }, - { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT }, - { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT }, - { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT }, - - /*** GLX_SGIX_fbconfig ***/ - { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX }, - { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX }, - { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX }, - { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX }, - { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX }, - { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX }, - - /*** GLX_SGIX_pbuffer ***/ - { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX }, - { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX }, - { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX }, - { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX }, - { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX }, - - /*** GLX_SGI_cushion ***/ - { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI }, - - /*** GLX_SGIX_video_resize ***/ - { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX }, - { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX }, - { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX }, - { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX }, - { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX }, - - /*** GLX_SGIX_dmbuffer **/ -#if defined(_DM_BUFFER_H_) - { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX }, -#endif - - /*** GLX_SGIX_swap_group ***/ - { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX }, - - /*** GLX_SGIX_swap_barrier ***/ - { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX }, - { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX }, - - /*** GLX_SUN_get_transparent_index ***/ - { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN }, - - /*** GLX_MESA_copy_sub_buffer ***/ - { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA }, - - /*** GLX_MESA_pixmap_colormap ***/ - { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA }, - - /*** GLX_MESA_release_buffers ***/ - { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA }, - - /*** GLX_ARB_get_proc_address ***/ - { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB }, - - /*** GLX_EXT_texture_from_pixmap ***/ - { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT }, - { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT }, - - { NULL, NULL } /* end of list */ -}; - - - -/* - * Return address of named glX function, or NULL if not found. - */ -__GLXextFuncPtr -_glxapi_get_proc_address(const char *funcName) -{ - GLuint i; - for (i = 0; GLX_functions[i].Name; i++) { - if (strcmp(GLX_functions[i].Name, funcName) == 0) - return GLX_functions[i].Address; - } - return NULL; -} - - - -/* - * This function does not get dispatched through the dispatch table - * since it's really a "meta" function. - */ -__GLXextFuncPtr -glXGetProcAddressARB(const GLubyte *procName) -{ - __GLXextFuncPtr f; - - f = _glxapi_get_proc_address((const char *) procName); - if (f) { - return f; - } - - f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName); - return f; -} - - -/* GLX 1.4 */ -void (*glXGetProcAddress(const GLubyte *procName))() -{ - return glXGetProcAddressARB(procName); -} diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.h b/src/gallium/state_trackers/glx/xlib/glxapi.h deleted file mode 100644 index b4e12b4162b..00000000000 --- a/src/gallium/state_trackers/glx/xlib/glxapi.h +++ /dev/null @@ -1,213 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.3 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef _glxapi_h_ -#define _glxapi_h_ - - -#define GLX_GLXEXT_PROTOTYPES -#include "GL/glx.h" - - -/* The GLX API dispatcher (i.e. this code) is being built into stand-alone - * Mesa. We don't know anything about XFree86 or real GLX so we define a - * minimal __GLXContextRec here so some of the functions in this file can - * work properly. - */ -typedef struct __GLXcontextRec { - Display *currentDpy; - GLboolean isDirect; - GLXDrawable currentDrawable; - GLXDrawable currentReadable; - XID xid; -} __GLXcontext; - - -/* - * Almost all the GLX API functions get routed through this dispatch table. - * The exceptions are the glXGetCurrentXXX() functions. - * - * This dispatch table allows multiple GLX client-side modules to coexist. - * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's - * pseudo-GLX can be present at the same time. The former being used on - * GLX-enabled X servers and the later on non-GLX X servers. - * - * Red Hat has been using this since Red Hat Linux 7.0 (I think). - * This'll be a standard feature in XFree86 4.3. It basically allows one - * libGL to do both DRI-rendering and "fake GLX" rendering to X displays - * that lack the GLX extension. - */ -struct _glxapi_table { - /*** GLX_VERSION_1_0 ***/ - XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list); - void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask); - GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct); - GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap); - void (*DestroyContext)(Display *dpy, GLXContext ctx); - void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap); - int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value); - /*GLXContext (*GetCurrentContext)(void);*/ - /*GLXDrawable (*GetCurrentDrawable)(void);*/ - Bool (*IsDirect)(Display *dpy, GLXContext ctx); - Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx); - Bool (*QueryExtension)(Display *dpy, int *errorb, int *event); - Bool (*QueryVersion)(Display *dpy, int *maj, int *min); - void (*SwapBuffers)(Display *dpy, GLXDrawable drawable); - void (*UseXFont)(Font font, int first, int count, int listBase); - void (*WaitGL)(void); - void (*WaitX)(void); - - /*** GLX_VERSION_1_1 ***/ - const char *(*GetClientString)(Display *dpy, int name); - const char *(*QueryExtensionsString)(Display *dpy, int screen); - const char *(*QueryServerString)(Display *dpy, int screen, int name); - - /*** GLX_VERSION_1_2 ***/ - /*Display *(*GetCurrentDisplay)(void);*/ - - /*** GLX_VERSION_1_3 ***/ - GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems); - GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct); - GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList); - GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList); - GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList); - void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf); - void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap); - void (*DestroyWindow)(Display *dpy, GLXWindow window); - /*GLXDrawable (*GetCurrentReadDrawable)(void);*/ - int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value); - GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements); - void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask); - XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config); - Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); - int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value); - void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value); - void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask); - - /*** GLX_SGI_swap_control ***/ - int (*SwapIntervalSGI)(int); - - /*** GLX_SGI_video_sync ***/ - int (*GetVideoSyncSGI)(unsigned int *count); - int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count); - - /*** GLX_SGI_make_current_read ***/ - Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext); - /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/ - - /*** GLX_SGIX_video_source (needs video library) ***/ -#if defined(_VL_H_) - GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode); - void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX); -#else - void *CreateGLXVideoSourceSGIX; - void *DestroyGLXVideoSourceSGIX; -#endif - - /*** GLX_EXT_import_context ***/ - void (*FreeContextEXT)(Display *dpy, GLXContext context); - GLXContextID (*GetContextIDEXT)(const GLXContext context); - /*Display *(*GetCurrentDisplayEXT)(void);*/ - GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID); - int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value); - - /*** GLX_SGIX_fbconfig ***/ - int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *); - GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *); - GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap); - GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool); - XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX); - GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *); - - /*** GLX_SGIX_pbuffer ***/ - GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *); - void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX); - int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *); - void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long); - void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *); - - /*** GLX_SGI_cushion ***/ - void (*CushionSGI)(Display *, Window, float); - - /*** GLX_SGIX_video_resize ***/ - int (*BindChannelToWindowSGIX)(Display *, int, int, Window); - int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int); - int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *); - int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *); - int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum); - - /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/ -#if defined (_DM_BUFFER_H_) - Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer); -#else - void *AssociciateDMPbufferSGIX; -#endif - - /*** GLX_SGIX_swap_group ***/ - void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable); - - /*** GLX_SGIX_swap_barrier ***/ - void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int); - Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *); - - /*** GLX_SUN_get_transparent_index ***/ - Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *); - - /*** GLX_MESA_copy_sub_buffer ***/ - void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height); - - /*** GLX_MESA_release_buffers ***/ - Bool (*ReleaseBuffersMESA)(Display *dpy, Window w); - - /*** GLX_MESA_pixmap_colormap ***/ - GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap); - - /*** GLX_EXT_texture_from_pixmap ***/ - void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer, - const int *attrib_list); - void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer); -}; - - - -extern const char * -_glxapi_get_version(void); - - - - -extern GLuint -_glxapi_get_dispatch_table_size(void); - - -extern void -_glxapi_set_no_op_table(struct _glxapi_table *t); - - -extern __GLXextFuncPtr -_glxapi_get_proc_address(const char *funcName); - - -#endif diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.h b/src/gallium/state_trackers/glx/xlib/xm_api.h index bdd434cd364..ce97a3ec768 100644 --- a/src/gallium/state_trackers/glx/xlib/xm_api.h +++ b/src/gallium/state_trackers/glx/xlib/xm_api.h @@ -57,7 +57,7 @@ and create a window, you must do the following to use the X/Mesa interface: #define XMESA_H -#include "mtypes.h" +#include "main/mtypes.h" #include "state_tracker/st_context.h" #include "state_tracker/st_public.h" #include "pipe/p_thread.h" diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index 1581182aec2..ec385e7c447 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -15,6 +15,19 @@ if 'python' in env['statetrackers']: env.Append(CPPPATH = '.') + if env['platform'] == 'windows': + env.Append(LIBS = [ + 'opengl32', + 'gdi32', + 'user32', + 'kernel32', + ]) + else: + env.Append(LIBS = [ + 'GL', + 'X11', + ]) + pyst = env.ConvenienceLibrary( target = 'pyst', source = [ diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index 5d4d04498b2..6f0bd6ae52a 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -43,19 +43,27 @@ except ImportError: return struct.unpack(fmt, buf[offset:offset + size]) -def make_image(surface): +def make_image(surface, x=None, y=None, w=None, h=None): + if x is None: + x = 0 + if y is None: + y = 0 + if w is None: + w = surface.width - x + if h is None: + h = surface.height - y data = surface.get_tile_rgba8(0, 0, surface.width, surface.height) import Image outimage = Image.fromstring('RGBA', (surface.width, surface.height), data, "raw", 'RGBA', 0, 1) return outimage -def save_image(filename, surface): - outimage = make_image(surface) +def save_image(filename, surface, x=None, y=None, w=None, h=None): + outimage = make_image(surface, x, y, w, h) outimage.save(filename, "PNG") -def show_image(surface, title): - outimage = make_image(surface) +def show_image(surface, title, x=None, y=None, w=None, h=None): + outimage = make_image(surface, x, y, w, h) import Tkinter as tk from PIL import Image, ImageTk @@ -305,7 +313,11 @@ class Screen(Object): def get_tex_transfer(self, texture, face, level, zslice, usage, x, y, w, h): if texture is None: return None - return Transfer(texture.get_surface(face, level, zslice), x, y, w, h) + transfer = Transfer(texture.get_surface(face, level, zslice), x, y, w, h) + if transfer and usage != gallium.PIPE_TRANSFER_WRITE: + if self.interpreter.options.all: + self.interpreter.present(transfer.surface, 'transf_read', x, y, w, h) + return transfer def tex_transfer_destroy(self, transfer): self.interpreter.unregister_object(transfer) @@ -314,6 +326,8 @@ class Screen(Object): if transfer is None: return transfer.surface.put_tile_raw(transfer.x, transfer.y, transfer.w, transfer.h, data, stride) + if self.interpreter.options.all: + self.interpreter.present(transfer.surface, 'transf_write', transfer.x, transfer.y, transfer.w, transfer.h) def user_buffer_create(self, data, size): # We don't really care to distinguish between user and regular buffers @@ -442,6 +456,7 @@ class Context(Object): x, y, z, w = unpack_from(format, data, offset) sys.stdout.write('\tCONST[%2u] = {%10.4f, %10.4f, %10.4f, %10.4f}\n' % (index, x, y, z, w)) index += 1 + sys.stdout.flush() def set_constant_buffer(self, shader, index, buffer): if buffer is not None: @@ -523,6 +538,7 @@ class Context(Object): sys.stdout.write('\t\t{' + ', '.join(map(str, values)) + '},\n') assert len(values) == velem.nr_components sys.stdout.write('\t},\n') + sys.stdout.flush() def dump_indices(self, ibuf, isize, start, count): if not self.interpreter.verbosity(2): @@ -550,6 +566,7 @@ class Context(Object): minindex = min(minindex, index) maxindex = max(maxindex, index) sys.stdout.write('\t},\n') + sys.stdout.flush() return minindex, maxindex @@ -577,6 +594,28 @@ class Context(Object): self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count) self._set_dirty() + def surface_copy(self, dest, destx, desty, src, srcx, srcy, width, height): + if dest is not None and src is not None: + if self.interpreter.options.all: + self.interpreter.present(src, 'surface_copy_src', srcx, srcy, width, height) + self.real.surface_copy(dest, destx, desty, src, srcx, srcy, width, height) + if dest in self.cbufs: + self._set_dirty() + flags = gallium.PIPE_FLUSH_FRAME + else: + flags = 0 + self.flush(flags) + if self.interpreter.options.all: + self.interpreter.present(dest, 'surface_copy_dest', destx, desty, width, height) + + def is_texture_referenced(self, texture, face, level): + #return self.real.is_texture_referenced(format, texture, face, level) + pass + + def is_buffer_referenced(self, buf): + #return self.real.is_buffer_referenced(format, buf) + pass + def _set_dirty(self): if self.interpreter.options.step: self._present() @@ -602,6 +641,9 @@ class Context(Object): if self.cbufs and self.cbufs[0]: self.interpreter.present(self.cbufs[0], "cbuf") + if self.zsbuf: + if self.interpreter.options.all: + self.interpreter.present(self.zsbuf, "zsbuf") class Interpreter(parser.TraceDumper): @@ -635,7 +677,7 @@ class Interpreter(parser.TraceDumper): self.interpret_call(call) def handle_call(self, call): - if self.options.stop and call.no >= self.options.stop: + if self.options.stop and call.no > self.options.stop: sys.exit(0) if (call.klass, call.method) in self.ignore_calls: @@ -645,6 +687,7 @@ class Interpreter(parser.TraceDumper): if self.verbosity(1): parser.TraceDumper.handle_call(self, call) + sys.stdout.flush() args = [(str(name), self.interpret_arg(arg)) for name, arg in call.args] @@ -671,16 +714,16 @@ class Interpreter(parser.TraceDumper): def verbosity(self, level): return self.options.verbosity >= level - def present(self, surface, description): + def present(self, surface, description, x=None, y=None, w=None, h=None): if self.call_no < self.options.start: return if self.options.images: - filename = '%s_%04u.png' % (description, self.call_no) - save_image(filename, surface) + filename = '%04u_%s.png' % (self.call_no, description) + save_image(filename, surface, x, y, w, h) else: title = '%u. %s' % (self.call_no, description) - show_image(surface, title) + show_image(surface, title, x, y, w, h) class Main(parser.Main): @@ -690,6 +733,7 @@ class Main(parser.Main): optparser.add_option("-q", "--quiet", action="store_const", const=0, dest="verbosity", help="no messages") optparser.add_option("-v", "--verbose", action="count", dest="verbosity", default=1, help="increase verbosity level") optparser.add_option("-i", "--images", action="store_true", dest="images", default=False, help="save images instead of showing them") + optparser.add_option("-a", "--all", action="store_true", dest="all", default=False, help="show depth, stencil, and transfers") optparser.add_option("-s", "--step", action="store_true", dest="step", default=False, help="step trhough every draw") optparser.add_option("-f", "--from", action="store", type="int", dest="start", default=0, help="from call no") optparser.add_option("-t", "--to", action="store", type="int", dest="stop", default=0, help="until call no") diff --git a/src/gallium/state_trackers/python/retrace/parse.py b/src/gallium/state_trackers/python/retrace/parse.py index b0f3e8a432f..b08d3686715 100755 --- a/src/gallium/state_trackers/python/retrace/parse.py +++ b/src/gallium/state_trackers/python/retrace/parse.py @@ -371,7 +371,7 @@ class Main: stream = GzipFile(arg, 'rt') elif arg.endswith('.bz2'): from bz2 import BZ2File - stream = BZ2File(arg, 'rt') + stream = BZ2File(arg, 'rU') else: stream = open(arg, 'rt') self.process_arg(stream, options) diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index 4b9659861df..b721e0b5750 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -139,7 +139,7 @@ def test(dev): tex_usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET, ).get_surface() zbuf = dev.texture_create( - PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z16_UNORM, width, height, tex_usage=PIPE_TEXTURE_USAGE_DEPTH_STENCIL, ).get_surface() diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 8246b378ce0..ea7d18738f6 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -44,8 +44,14 @@ static void st_device_really_destroy(struct st_device *st_dev) { - if(st_dev->screen) + if(st_dev->screen) { + /* FIXME: Don't really destroy until we keep track of every single + * reference or we end up causing a segmentation fault every time + * python exits. */ +#if 0 st_dev->screen->destroy(st_dev->screen); +#endif + } FREE(st_dev); } diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c index 8b33c70fd71..43aaaabf2a1 100644 --- a/src/gallium/state_trackers/python/st_hardpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c @@ -28,31 +28,206 @@ /** * @file - * Stub for hardware pipe driver support. + * Get a hardware accelerated Gallium screen/context from the OpenGL driver. */ #include "pipe/p_compiler.h" +#ifdef PIPE_OS_WINDOWS +#include <windows.h> +#include <GL/gl.h> +#else +#include <X11/Xlib.h> +#include <X11/Xutil.h> +#include <GL/gl.h> +#include <GL/glx.h> +#endif + #include "st_winsys.h" +typedef struct pipe_screen * (GLAPIENTRY *PFNGETGALLIUMSCREENMESAPROC) (void); +typedef struct pipe_context * (GLAPIENTRY* PFNCREATEGALLIUMCONTEXTMESAPROC) (void); + +static PFNGETGALLIUMSCREENMESAPROC pfnGetGalliumScreenMESA = NULL; +static PFNCREATEGALLIUMCONTEXTMESAPROC pfnCreateGalliumContextMESA = NULL; + + /* XXX: Force init_gallium symbol to be linked */ extern void init_gallium(void); void (*force_init_gallium_linkage)(void) = &init_gallium; +#ifdef PIPE_OS_WINDOWS + +static INLINE boolean +st_hardpipe_load(void) +{ + WNDCLASS wc; + HWND hwnd; + HGLRC hglrc; + HDC hdc; + PIXELFORMATDESCRIPTOR pfd; + int iPixelFormat; + + if(pfnGetGalliumScreenMESA && pfnCreateGalliumContextMESA) + return TRUE; + + memset(&wc, 0, sizeof wc); + wc.lpfnWndProc = DefWindowProc; + wc.lpszClassName = "gallium"; + wc.style = CS_OWNDC | CS_HREDRAW | CS_VREDRAW; + RegisterClass(&wc); + + hwnd = CreateWindow(wc.lpszClassName, "gallium", 0, 0, 0, 0, 0, NULL, 0, wc.hInstance, NULL); + if (!hwnd) + return FALSE; + + hdc = GetDC(hwnd); + if (!hdc) + return FALSE; + + pfd.cColorBits = 3; + pfd.cRedBits = 1; + pfd.cGreenBits = 1; + pfd.cBlueBits = 1; + pfd.dwFlags = PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL; + pfd.iLayerType = PFD_MAIN_PLANE; + pfd.iPixelType = PFD_TYPE_RGBA; + pfd.nSize = sizeof(pfd); + pfd.nVersion = 1; + + iPixelFormat = ChoosePixelFormat(hdc, &pfd); + if (!iPixelFormat) { + pfd.dwFlags |= PFD_DOUBLEBUFFER; + iPixelFormat = ChoosePixelFormat(hdc, &pfd); + } + if (!iPixelFormat) + return FALSE; + + SetPixelFormat(hdc, iPixelFormat, &pfd); + hglrc = wglCreateContext(hdc); + if (!hglrc) + return FALSE; + + if (!wglMakeCurrent(hdc, hglrc)) + return FALSE; + + pfnGetGalliumScreenMESA = (PFNGETGALLIUMSCREENMESAPROC)wglGetProcAddress("wglGetGalliumScreenMESA"); + if(!pfnGetGalliumScreenMESA) + return FALSE; + + pfnCreateGalliumContextMESA = (PFNCREATEGALLIUMCONTEXTMESAPROC)wglGetProcAddress("wglCreateGalliumContextMESA"); + if(!pfnCreateGalliumContextMESA) + return FALSE; + + DestroyWindow(hwnd); + + return TRUE; +} + +#else + +static INLINE boolean +st_hardpipe_load(void) +{ + Display *dpy; + int scrnum; + Window root; + int attribSingle[] = { + GLX_RGBA, + GLX_RED_SIZE, 1, + GLX_GREEN_SIZE, 1, + GLX_BLUE_SIZE, 1, + None }; + int attribDouble[] = { + GLX_RGBA, + GLX_RED_SIZE, 1, + GLX_GREEN_SIZE, 1, + GLX_BLUE_SIZE, 1, + GLX_DOUBLEBUFFER, + None }; + XVisualInfo *visinfo; + GLXContext ctx = NULL; + XSetWindowAttributes attr; + unsigned long mask; + int width = 100, height = 100; + Window win; + + dpy = XOpenDisplay(NULL); + if (!dpy) + return FALSE; + + scrnum = 0; + + root = RootWindow(dpy, scrnum); + + visinfo = glXChooseVisual(dpy, scrnum, attribSingle); + if (!visinfo) + visinfo = glXChooseVisual(dpy, scrnum, attribDouble); + if (!visinfo) + return FALSE; + + ctx = glXCreateContext( dpy, visinfo, NULL, True ); + + if (!ctx) + return FALSE; + + attr.background_pixel = 0; + attr.border_pixel = 0; + attr.colormap = XCreateColormap(dpy, root, visinfo->visual, AllocNone); + attr.event_mask = StructureNotifyMask | ExposureMask; + + mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask; + + win = XCreateWindow(dpy, root, 0, 0, width, height, + 0, visinfo->depth, InputOutput, + visinfo->visual, mask, &attr); + + if (!glXMakeCurrent(dpy, win, ctx)) + return FALSE; + + pfnGetGalliumScreenMESA = (PFNGETGALLIUMSCREENMESAPROC)glXGetProcAddressARB((const GLubyte *)"glXGetGalliumScreenMESA"); + if(!pfnGetGalliumScreenMESA) + return FALSE; + + pfnCreateGalliumContextMESA = (PFNCREATEGALLIUMCONTEXTMESAPROC)glXGetProcAddressARB((const GLubyte *)"glXCreateGalliumContextMESA"); + if(!pfnCreateGalliumContextMESA) + return FALSE; + + glXDestroyContext(dpy, ctx); + XFree(visinfo); + XDestroyWindow(dpy, win); + XCloseDisplay(dpy); + + return TRUE; +} + +#endif + + static struct pipe_screen * st_hardpipe_screen_create(void) { - return st_softpipe_winsys.screen_create(); + if(st_hardpipe_load()) + return pfnGetGalliumScreenMESA(); + else + return st_softpipe_winsys.screen_create(); } static struct pipe_context * st_hardpipe_context_create(struct pipe_screen *screen) { - return st_softpipe_winsys.context_create(screen); + if(st_hardpipe_load()) { + if(screen == pfnGetGalliumScreenMESA()) + return pfnCreateGalliumContextMESA(); + else + return NULL; + } + else + return st_softpipe_winsys.context_create(screen); } diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index 1fa7fe6f3b3..202ccfc3509 100755 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -46,6 +46,14 @@ for name, value in globals().items(): if name.startswith("PIPE_FORMAT_") and isinstance(value, int): formats[value] = name +def is_depth_stencil_format(format): + # FIXME: make and use binding to pf_is_depth_stencil + return format in ( + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_Z16_UNORM, + ) def make_image(width, height, rgba): import Image diff --git a/src/gallium/state_trackers/python/tests/texture_sample.py b/src/gallium/state_trackers/python/tests/texture_sample.py index a382424667c..c7b78abbbec 100755 --- a/src/gallium/state_trackers/python/tests/texture_sample.py +++ b/src/gallium/state_trackers/python/tests/texture_sample.py @@ -99,7 +99,7 @@ def is_pot(n): return n & (n - 1) == 0 -class TextureTest(TestCase): +class TextureColorSampleTest(TestCase): tags = ( 'target', @@ -286,6 +286,206 @@ class TextureTest(TestCase): self.assert_rgba(cbuf, x, y, w, h, expected_rgba, 4.0/256, 0.85) +class TextureDepthSampleTest(TestCase): + + tags = ( + 'target', + 'format', + 'width', + 'height', + 'depth', + 'last_level', + 'face', + 'level', + 'zslice', + ) + + def test(self): + dev = self.dev + + target = self.target + format = self.format + width = self.width + height = self.height + depth = self.depth + last_level = self.last_level + face = self.face + level = self.level + zslice = self.zslice + + tex_usage = PIPE_TEXTURE_USAGE_SAMPLER + geom_flags = 0 + if width != height: + geom_flags |= PIPE_TEXTURE_GEOM_NON_SQUARE + if not is_pot(width) or not is_pot(height) or not is_pot(depth): + geom_flags |= PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO + + if not dev.is_format_supported(format, target, tex_usage, geom_flags): + raise TestSkip + + ctx = self.dev.context_create() + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + depth_stencil_alpha.depth.enabled = 1 + depth_stencil_alpha.depth.writemask = 1 + depth_stencil_alpha.depth.func = PIPE_FUNC_LESS + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_vs_clip_and_viewport = 1 + ctx.set_rasterizer(rasterizer) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + sampler.min_lod = 0 + sampler.max_lod = PIPE_MAX_TEXTURE_LEVELS - 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create( + target = target, + format = format, + width = width, + height = height, + depth = depth, + last_level = last_level, + tex_usage = tex_usage, + ) + + expected_rgba = FloatArray(height*width*4) + texture.get_surface( + face = face, + level = level, + zslice = zslice, + ).sample_rgba(expected_rgba) + + ctx.set_sampler_texture(0, texture) + + # framebuffer + cbuf_tex = dev.texture_create( + PIPE_FORMAT_A8R8G8B8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) + + zsbuf_tex = dev.texture_create( + PIPE_FORMAT_Z24X8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) + + cbuf = cbuf_tex.get_surface() + zsbuf = zsbuf_tex.get_surface() + fb = Framebuffer() + fb.width = width + fb.height = height + fb.nr_cbufs = 1 + fb.set_cbuf(0, cbuf) + fb.set_zsbuf(zsbuf) + ctx.set_framebuffer(fb) + rgba = FloatArray(4); + rgba[0] = 0.5 + rgba[1] = 0.5 + rgba[2] = 0.5 + rgba[3] = 0.5 + ctx.clear(PIPE_CLEAR_DEPTHSTENCIL, rgba, 1.0, 0) + del fb + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], GENERIC, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], GENERIC, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) + + # fragment shader + op = { + PIPE_TEXTURE_1D: "1D", + PIPE_TEXTURE_2D: "2D", + PIPE_TEXTURE_3D: "3D", + PIPE_TEXTURE_CUBE: "CUBE", + }[target] + fs = Shader(''' + FRAG1.1 + DCL IN[0], GENERIC[0], LINEAR + DCL SAMP[0], CONSTANT + DCL OUT[0].z, POSITION + 0:TEX OUT[0].z, IN[0], SAMP[0], %s + 1:END + ''' % op) + #fs.dump() + ctx.set_fragment_shader(fs) + + nverts = 4 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + x = 0 + y = 0 + w, h = minify((width, height), level) + + pos = [ + [x, y], + [x+w, y], + [x+w, y+h], + [x, y+h], + ] + + tex = tex_coords(texture, face, level, zslice) + + for i in range(0, 4): + j = 8*i + verts[j + 0] = pos[i][0] # x + verts[j + 1] = pos[i][1] # y + verts[j + 2] = 0.0 # z + verts[j + 3] = 1.0 # w + verts[j + 4] = tex[i][0] # s + verts[j + 5] = tex[i][1] # r + verts[j + 6] = tex[i][2] # q + verts[j + 7] = 1.0 + + ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, + nverts, + nattrs, + verts) + + ctx.flush() + + zsbuf = zsbuf_tex.get_surface() + + self.assert_rgba(zsbuf, x, y, w, h, expected_rgba, 4.0/256, 0.85) + + + def main(): dev = Device() @@ -297,18 +497,13 @@ def main(): PIPE_TEXTURE_3D, ] - formats = [ + color_formats = [ PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_X8R8G8B8_UNORM, #PIPE_FORMAT_A8R8G8B8_SRGB, PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_A1R5G5B5_UNORM, PIPE_FORMAT_A4R4G4B4_UNORM, - #PIPE_FORMAT_Z32_UNORM, - #PIPE_FORMAT_Z24S8_UNORM, - #PIPE_FORMAT_Z24X8_UNORM, - #PIPE_FORMAT_Z16_UNORM, - #PIPE_FORMAT_S8_UNORM, PIPE_FORMAT_A8_UNORM, PIPE_FORMAT_L8_UNORM, PIPE_FORMAT_YCBCR, @@ -318,6 +513,13 @@ def main(): #PIPE_FORMAT_DXT5_RGBA, ] + depth_formats = [ + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_Z16_UNORM, + ] + sizes = [64, 32, 16, 8, 4, 2, 1] #sizes = [1020, 508, 252, 62, 30, 14, 6, 3] #sizes = [64] @@ -332,8 +534,8 @@ def main(): PIPE_TEX_FACE_NEG_Z, ] - for target in targets: - for format in formats: + for format in color_formats: + for target in targets: for size in sizes: if target == PIPE_TEXTURE_3D: depth = size @@ -347,7 +549,7 @@ def main(): for level in range(0, last_level + 1): zslice = 0 while zslice < depth >> level: - test = TextureTest( + test = TextureColorSampleTest( dev = dev, target = target, format = format, @@ -361,6 +563,27 @@ def main(): ) suite.add_test(test) zslice = (zslice + 1)*2 - 1 + for format in depth_formats: + target = PIPE_TEXTURE_2D + depth = 1 + face = 0 + last_level = 0 + level = 0 + zslice = 0 + for size in sizes: + test = TextureDepthSampleTest( + dev = dev, + target = target, + format = format, + width = size, + height = size, + depth = depth, + last_level = last_level, + face = face, + level = level, + zslice = zslice, + ) + suite.add_test(test) suite.run() diff --git a/src/gallium/state_trackers/vega/Makefile b/src/gallium/state_trackers/vega/Makefile new file mode 100644 index 00000000000..b8c805b06c4 --- /dev/null +++ b/src/gallium/state_trackers/vega/Makefile @@ -0,0 +1,128 @@ +# src/mesa/Makefile + +TOP = ../../../.. +include $(TOP)/configs/current +GALLIUM = $(TOP) + +### Lists of source files, included by Makefiles + +VG_SOURCES = \ + api_context.c \ + api_filters.c \ + api_images.c \ + api_masks.c \ + api_misc.c \ + api_paint.c \ + api_params.c \ + api_path.c \ + api_text.c \ + api_transform.c \ + vgu.c \ + vg_context.c \ + vg_state.c \ + vg_tracker.c \ + vg_translate.c \ + polygon.c \ + bezier.c \ + path.c \ + paint.c \ + arc.c \ + image.c \ + renderer.c \ + stroker.c \ + mask.c \ + shader.c \ + shaders_cache.c + + +### All the core C sources + +ALL_SOURCES = \ + $(VG_SOURCES) + + +### Object files +VG_OBJECTS = \ + $(VG_SOURCES:.c=.o) + +### Include directories + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(GALLIUM)/include \ + -I$(GALLIUM)/src/gallium/include \ + -I$(GALLIUM)/src/gallium/auxiliary + +VG_LIB = OpenVG +VG_LIB_NAME = lib$(VG_LIB).so + +VG_MAJOR = 1 +VG_MINOR = 0 +VG_TINY = 0 + +GALLIUM_LIBS = \ + $(GALLIUM)/src/gallium/auxiliary/pipebuffer/libpipebuffer.a \ + $(GALLIUM)/src/gallium/auxiliary/sct/libsct.a \ + $(GALLIUM)/src/gallium/auxiliary/draw/libdraw.a \ + $(GALLIUM)/src/gallium/auxiliary/rtasm/librtasm.a \ + $(GALLIUM)/src/gallium/auxiliary/translate/libtranslate.a \ + $(GALLIUM)/src/gallium/auxiliary/cso_cache/libcso_cache.a \ + $(GALLIUM)/src/gallium/auxiliary/util/libutil.a \ + $(GALLIUM)/src/gallium/auxiliary/tgsi/libtgsi.a + +.SUFFIXES : .cpp + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +default: depend subdirs $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME) + +# Make the OpenVG library +$(TOP)/$(LIB_DIR)/$(VG_LIB_NAME): $(VG_OBJECTS) $(GALLIUM_LIBS) + $(TOP)/bin/mklib -o $(VG_LIB) \ + -major $(VG_MAJOR) \ + -minor $(VG_MINOR) \ + -patch $(VG_TINY) \ + -install $(TOP)/$(LIB_DIR) \ + $(VG_OBJECTS) $(GALLIUM_LIBS) \ + -Wl,--whole-archive $(LIBS) -Wl,--no-whole-archive $(SYS_LIBS) + +###################################################################### +# Generic stuff + +depend: $(ALL_SOURCES) + @ echo "running $(MKDEP)" + @ rm -f depend # workaround oops on gutsy?!? + @ touch depend + @ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ + > /dev/null 2>/dev/null + + +subdirs: + +install: default + $(INSTALL) -d $(INSTALL_DIR)/include/VG + $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 $(TOP)/include/VG/*.h $(INSTALL_DIR)/include/VG + @if [ -e $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME) ]; then \ + $(INSTALL) $(TOP)/$(LIB_DIR)/libOpenVG* $(INSTALL_DIR)/$(LIB_DIR); \ + fi + +# Emacs tags +tags: + etags `find . -name \*.[ch]` $(TOP)/include/VG/*.h + +clean: + -rm -f *.o + -rm -f */*.o + -rm -f */*/*.o + -rm -f depend depend.bak + +include depend diff --git a/src/gallium/state_trackers/vega/api_consts.h b/src/gallium/state_trackers/vega/api_consts.h new file mode 100644 index 00000000000..e1b48d4a46c --- /dev/null +++ b/src/gallium/state_trackers/vega/api_consts.h @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef API_CONSTS_H +#define API_CONSTS_H + +/*must be at least 32*/ +#define VEGA_MAX_SCISSOR_RECTS 32 + +/*must be at least 16*/ +#define VEGA_MAX_DASH_COUNT 32 + +/*must be at least 7*/ +#define VEGA_MAX_KERNEL_SIZE 7 + +/*must be at least 15*/ +#define VEGA_MAX_SEPARABLE_KERNEL_SIZE 15 + +/*must be at least 32*/ +#define VEGA_MAX_COLOR_RAMP_STOPS 256 + +#define VEGA_MAX_IMAGE_WIDTH 2048 + +#define VEGA_MAX_IMAGE_HEIGHT 2048 + +#define VEGA_MAX_IMAGE_PIXELS (2048*2048) + +#define VEGA_MAX_IMAGE_BYTES (2048*2048 * 4) + +/*must be at least 128*/ +#define VEGA_MAX_GAUSSIAN_STD_DEVIATION 128 + +#endif diff --git a/src/gallium/state_trackers/vega/api_context.c b/src/gallium/state_trackers/vega/api_context.c new file mode 100644 index 00000000000..47db102dd2d --- /dev/null +++ b/src/gallium/state_trackers/vega/api_context.c @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" + +VGErrorCode vgGetError(void) +{ + struct vg_context *ctx = vg_current_context(); + VGErrorCode error = VG_NO_CONTEXT_ERROR; + + if (!ctx) + return error; + + error = ctx->_error; + ctx->_error = VG_NO_ERROR; + + return error; +} + +void vgFlush(void) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe; + + if (!ctx) + return; + + pipe = ctx->pipe; + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); +} + +void vgFinish(void) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_fence_handle *fence = NULL; + struct pipe_context *pipe; + + if (!ctx) + return; + + pipe = ctx->pipe; + + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence); + + pipe->screen->fence_finish(pipe->screen, fence, 0); + pipe->screen->fence_reference(pipe->screen, &fence, NULL); +} diff --git a/src/gallium/state_trackers/vega/api_filters.c b/src/gallium/state_trackers/vega/api_filters.c new file mode 100644 index 00000000000..862cbb03c43 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_filters.c @@ -0,0 +1,805 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" +#include "image.h" +#include "renderer.h" +#include "shaders_cache.h" +#include "st_inlines.h" + +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "pipe/p_screen.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_memory.h" + + +#include "asm_filters.h" + + +struct filter_info { + struct vg_image *dst; + struct vg_image *src; + struct vg_shader * (*setup_shader)(struct vg_context *, void *); + void *user_data; + const void *const_buffer; + VGint const_buffer_len; + VGTilingMode tiling_mode; + struct pipe_texture *extra_texture; +}; + +static INLINE struct pipe_texture *create_texture_1d(struct vg_context *ctx, + const VGuint *color_data, + const VGint color_data_len) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_texture *tex = 0; + struct pipe_texture templ; + + memset(&templ, 0, sizeof(templ)); + templ.target = PIPE_TEXTURE_1D; + templ.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templ.last_level = 0; + templ.width[0] = color_data_len; + templ.height[0] = 1; + templ.depth[0] = 1; + pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block); + templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + + tex = screen->texture_create(screen, &templ); + + { /* upload color_data */ + struct pipe_transfer *transfer = + screen->get_tex_transfer(screen, tex, + 0, 0, 0, + PIPE_TRANSFER_READ_WRITE , + 0, 0, tex->width[0], tex->height[0]); + void *map = screen->transfer_map(screen, transfer); + memcpy(map, color_data, sizeof(VGint)*color_data_len); + screen->transfer_unmap(screen, transfer); + screen->tex_transfer_destroy(transfer); + } + + return tex; +} + +static INLINE struct pipe_surface * setup_framebuffer(struct vg_image *dst) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_framebuffer_state fb; + struct pipe_surface *dst_surf = pipe->screen->get_tex_surface( + pipe->screen, dst->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* drawing dest */ + memset(&fb, 0, sizeof(fb)); + fb.width = dst->x + dst_surf->width; + fb.height = dst->y + dst_surf->height; + fb.nr_cbufs = 1; + fb.cbufs[0] = dst_surf; + { + VGint i; + for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i) + fb.cbufs[i] = 0; + } + cso_set_framebuffer(ctx->cso_context, &fb); + + return dst_surf; +} + +static void setup_viewport(struct vg_image *dst) +{ + struct vg_context *ctx = vg_current_context(); + vg_set_viewport(ctx, VEGA_Y0_TOP); +} + +static void setup_blend() +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + if (ctx->state.vg.filter_channel_mask & VG_RED) + blend.colormask |= PIPE_MASK_R; + if (ctx->state.vg.filter_channel_mask & VG_GREEN) + blend.colormask |= PIPE_MASK_G; + if (ctx->state.vg.filter_channel_mask & VG_BLUE) + blend.colormask |= PIPE_MASK_B; + if (ctx->state.vg.filter_channel_mask & VG_ALPHA) + blend.colormask |= PIPE_MASK_A; + blend.blend_enable = 1; + cso_set_blend(ctx->cso_context, &blend); +} + +static void setup_constant_buffer(struct vg_context *ctx, const void *buffer, + VGint param_bytes) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_constant_buffer *cbuf = &ctx->filter.buffer; + + /* We always need to get a new buffer, to keep the drivers simple and + * avoid gratuitous rendering synchronization. */ + pipe_buffer_reference(&cbuf->buffer, NULL); + + cbuf->buffer = pipe_buffer_create(pipe->screen, 16, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + + if (cbuf->buffer) { + st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, + 0, param_bytes, buffer); + } + + ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); +} + +static void setup_samplers(struct vg_context *ctx, struct filter_info *info) +{ + struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + struct pipe_sampler_state sampler[3]; + int num_samplers = 0; + int num_textures = 0; + + samplers[0] = NULL; + samplers[1] = NULL; + samplers[2] = NULL; + samplers[3] = NULL; + textures[0] = NULL; + textures[1] = NULL; + textures[2] = NULL; + textures[3] = NULL; + + memset(&sampler[0], 0, sizeof(struct pipe_sampler_state)); + sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler[0].wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler[0].min_img_filter = PIPE_TEX_MIPFILTER_LINEAR; + sampler[0].mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR; + sampler[0].normalized_coords = 1; + + switch(info->tiling_mode) { + case VG_TILE_FILL: + sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER; + sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER; + memcpy(sampler[0].border_color, + ctx->state.vg.tile_fill_color, + sizeof(VGfloat) * 4); + break; + case VG_TILE_PAD: + sampler[0].wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler[0].wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + break; + case VG_TILE_REPEAT: + sampler[0].wrap_s = PIPE_TEX_WRAP_REPEAT; + sampler[0].wrap_t = PIPE_TEX_WRAP_REPEAT; + break; + case VG_TILE_REFLECT: + sampler[0].wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT; + sampler[0].wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT; + break; + default: + debug_assert(!"Unknown tiling mode"); + } + + samplers[0] = &sampler[0]; + textures[0] = info->src->texture; + ++num_samplers; + ++num_textures; + + if (info->extra_texture) { + memcpy(&sampler[1], &sampler[0], sizeof(struct pipe_sampler_state)); + samplers[1] = &sampler[1]; + textures[1] = info->extra_texture; + ++num_samplers; + ++num_textures; + } + + + cso_set_samplers(ctx->cso_context, num_samplers, (const struct pipe_sampler_state **)samplers); + cso_set_sampler_textures(ctx->cso_context, num_textures, textures); +} + +static struct vg_shader * setup_color_matrix(struct vg_context *ctx, void *user_data) +{ + struct vg_shader *shader = + shader_create_from_text(ctx->pipe, color_matrix_asm, 200, + PIPE_SHADER_FRAGMENT); + cso_set_fragment_shader_handle(ctx->cso_context, shader->driver); + return shader; +} + +static struct vg_shader * setup_convolution(struct vg_context *ctx, void *user_data) +{ + char buffer[1024]; + VGint num_consts = (VGint)(long)(user_data); + struct vg_shader *shader; + + snprintf(buffer, 1023, convolution_asm, num_consts, num_consts / 2 + 1); + + shader = shader_create_from_text(ctx->pipe, buffer, 200, + PIPE_SHADER_FRAGMENT); + + cso_set_fragment_shader_handle(ctx->cso_context, shader->driver); + return shader; +} + +static struct vg_shader * setup_lookup(struct vg_context *ctx, void *user_data) +{ + struct vg_shader *shader = + shader_create_from_text(ctx->pipe, lookup_asm, + 200, PIPE_SHADER_FRAGMENT); + + cso_set_fragment_shader_handle(ctx->cso_context, shader->driver); + return shader; +} + + +static struct vg_shader * setup_lookup_single(struct vg_context *ctx, void *user_data) +{ + char buffer[1024]; + VGImageChannel channel = (VGImageChannel)(user_data); + struct vg_shader *shader; + + switch(channel) { + case VG_RED: + snprintf(buffer, 1023, lookup_single_asm, "xxxx"); + break; + case VG_GREEN: + snprintf(buffer, 1023, lookup_single_asm, "yyyy"); + break; + case VG_BLUE: + snprintf(buffer, 1023, lookup_single_asm, "zzzz"); + break; + case VG_ALPHA: + snprintf(buffer, 1023, lookup_single_asm, "wwww"); + break; + default: + debug_assert(!"Unknown color channel"); + } + + shader = shader_create_from_text(ctx->pipe, buffer, 200, + PIPE_SHADER_FRAGMENT); + + cso_set_fragment_shader_handle(ctx->cso_context, shader->driver); + return shader; +} + +static void execute_filter(struct vg_context *ctx, + struct filter_info *info) +{ + struct pipe_surface *dst_surf; + struct vg_shader *shader; + + cso_save_framebuffer(ctx->cso_context); + cso_save_fragment_shader(ctx->cso_context); + cso_save_viewport(ctx->cso_context); + cso_save_blend(ctx->cso_context); + cso_save_samplers(ctx->cso_context); + cso_save_sampler_textures(ctx->cso_context); + + dst_surf = setup_framebuffer(info->dst); + setup_viewport(info->dst); + setup_blend(); + setup_constant_buffer(ctx, info->const_buffer, info->const_buffer_len); + shader = info->setup_shader(ctx, info->user_data); + setup_samplers(ctx, info); + + renderer_draw_texture(ctx->renderer, + info->src->texture, + info->dst->x, info->dst->y, + info->dst->x + info->dst->width, + info->dst->y + info->dst->height, + info->dst->x, info->dst->y, + info->dst->x + info->dst->width, + info->dst->y + info->dst->height); + + cso_restore_framebuffer(ctx->cso_context); + cso_restore_fragment_shader(ctx->cso_context); + cso_restore_viewport(ctx->cso_context); + cso_restore_blend(ctx->cso_context); + cso_restore_samplers(ctx->cso_context); + cso_restore_sampler_textures(ctx->cso_context); + + vg_shader_destroy(ctx, shader); + + pipe_surface_reference(&dst_surf, NULL); +} + +void vgColorMatrix(VGImage dst, VGImage src, + const VGfloat * matrix) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *d, *s; + struct filter_info info; + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!matrix || !is_aligned(matrix)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + d = (struct vg_image*)dst; + s = (struct vg_image*)src; + + if (vg_image_overlaps(d, s)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + info.dst = d; + info.src = s; + info.setup_shader = &setup_color_matrix; + info.user_data = NULL; + info.const_buffer = matrix; + info.const_buffer_len = 20 * sizeof(VGfloat); + info.tiling_mode = VG_TILE_PAD; + info.extra_texture = 0; + execute_filter(ctx, &info); +} + +static VGfloat texture_offset(VGfloat width, VGint kernelSize, VGint current, VGint shift) +{ + VGfloat diff = current - shift; + + return diff / width; +} + +void vgConvolve(VGImage dst, VGImage src, + VGint kernelWidth, VGint kernelHeight, + VGint shiftX, VGint shiftY, + const VGshort * kernel, + VGfloat scale, + VGfloat bias, + VGTilingMode tilingMode) +{ + struct vg_context *ctx = vg_current_context(); + VGfloat *buffer; + VGint buffer_len; + VGint i, j; + VGint idx = 0; + struct vg_image *d, *s; + VGint kernel_size = kernelWidth * kernelHeight; + struct filter_info info; + const VGint max_kernel_size = vgGeti(VG_MAX_KERNEL_SIZE); + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (kernelWidth <= 0 || kernelHeight <= 0 || + kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!kernel || !is_aligned_to(kernel, 2)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (tilingMode < VG_TILE_FILL || + tilingMode > VG_TILE_REFLECT) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + d = (struct vg_image*)dst; + s = (struct vg_image*)src; + + if (vg_image_overlaps(d, s)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + vg_validate_state(ctx); + + buffer_len = 8 + 2 * 4 * kernel_size; + buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat)); + + buffer[0] = 0.f; + buffer[1] = 1.f; + buffer[2] = 2.f; /*unused*/ + buffer[3] = 4.f; /*unused*/ + + buffer[4] = kernelWidth * kernelHeight; + buffer[5] = scale; + buffer[6] = bias; + buffer[7] = 0.f; + + idx = 8; + for (j = 0; j < kernelHeight; ++j) { + for (i = 0; i < kernelWidth; ++i) { + VGint index = j * kernelWidth + i; + VGfloat x, y; + + x = texture_offset(s->width, kernelWidth, i, shiftX); + y = texture_offset(s->height, kernelHeight, j, shiftY); + + buffer[idx + index*4 + 0] = x; + buffer[idx + index*4 + 1] = y; + buffer[idx + index*4 + 2] = 0.f; + buffer[idx + index*4 + 3] = 0.f; + } + } + idx += kernel_size * 4; + + for (j = 0; j < kernelHeight; ++j) { + for (i = 0; i < kernelWidth; ++i) { + /* transpose the kernel */ + VGint index = j * kernelWidth + i; + VGint kindex = (kernelWidth - i - 1) * kernelHeight + (kernelHeight - j - 1); + buffer[idx + index*4 + 0] = kernel[kindex]; + buffer[idx + index*4 + 1] = kernel[kindex]; + buffer[idx + index*4 + 2] = kernel[kindex]; + buffer[idx + index*4 + 3] = kernel[kindex]; + } + } + + info.dst = d; + info.src = s; + info.setup_shader = &setup_convolution; + info.user_data = (void*)(long)(buffer_len/4); + info.const_buffer = buffer; + info.const_buffer_len = buffer_len * sizeof(VGfloat); + info.tiling_mode = tilingMode; + info.extra_texture = 0; + execute_filter(ctx, &info); + + free(buffer); +} + +void vgSeparableConvolve(VGImage dst, VGImage src, + VGint kernelWidth, + VGint kernelHeight, + VGint shiftX, VGint shiftY, + const VGshort * kernelX, + const VGshort * kernelY, + VGfloat scale, + VGfloat bias, + VGTilingMode tilingMode) +{ + struct vg_context *ctx = vg_current_context(); + VGshort *kernel; + VGint i, j, idx = 0; + const VGint max_kernel_size = vgGeti(VG_MAX_SEPARABLE_KERNEL_SIZE); + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (kernelWidth <= 0 || kernelHeight <= 0 || + kernelWidth > max_kernel_size || kernelHeight > max_kernel_size) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!kernelX || !kernelY || + !is_aligned_to(kernelX, 2) || !is_aligned_to(kernelY, 2)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (tilingMode < VG_TILE_FILL || + tilingMode > VG_TILE_REFLECT) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + kernel = malloc(sizeof(VGshort)*kernelWidth*kernelHeight); + for (i = 0; i < kernelWidth; ++i) { + for (j = 0; j < kernelHeight; ++j) { + kernel[idx] = kernelX[i] * kernelY[j]; + ++idx; + } + } + vgConvolve(dst, src, kernelWidth, kernelHeight, shiftX, shiftY, + kernel, scale, bias, tilingMode); + free(kernel); +} + +static INLINE VGfloat compute_gaussian_componenet(VGfloat x, VGfloat y, + VGfloat stdDeviationX, + VGfloat stdDeviationY) +{ + VGfloat mult = 1 / ( 2 * M_PI * stdDeviationX * stdDeviationY); + VGfloat e = exp( - ( pow(x, 2)/(2*pow(stdDeviationX, 2)) + + pow(y, 2)/(2*pow(stdDeviationY, 2)) ) ); + return mult * e; +} + +static INLINE VGint compute_kernel_size(VGfloat deviation) +{ + VGint size = ceil(2.146 * deviation); + if (size > 11) + return 11; + return size; +} + +static void compute_gaussian_kernel(VGfloat *kernel, + VGint width, VGint height, + VGfloat stdDeviationX, + VGfloat stdDeviationY) +{ + VGint i, j; + VGfloat scale = 0.0f; + + for (j = 0; j < height; ++j) { + for (i = 0; i < width; ++i) { + VGint idx = (height - j -1) * width + (width - i -1); + kernel[idx] = compute_gaussian_componenet(i-(ceil(width/2))-1, + j-ceil(height/2)-1, + stdDeviationX, stdDeviationY); + scale += kernel[idx]; + } + } + + for (j = 0; j < height; ++j) { + for (i = 0; i < width; ++i) { + VGint idx = j * width + i; + kernel[idx] /= scale; + } + } +} + +void vgGaussianBlur(VGImage dst, VGImage src, + VGfloat stdDeviationX, + VGfloat stdDeviationY, + VGTilingMode tilingMode) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *d, *s; + VGfloat *buffer, *kernel; + VGint kernel_width, kernel_height, kernel_size; + VGint buffer_len; + VGint idx, i, j; + struct filter_info info; + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (stdDeviationX <= 0 || stdDeviationY <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (tilingMode < VG_TILE_FILL || + tilingMode > VG_TILE_REFLECT) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + d = (struct vg_image*)dst; + s = (struct vg_image*)src; + + if (vg_image_overlaps(d, s)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + kernel_width = compute_kernel_size(stdDeviationX); + kernel_height = compute_kernel_size(stdDeviationY); + kernel_size = kernel_width * kernel_height; + kernel = malloc(sizeof(VGfloat)*kernel_size); + compute_gaussian_kernel(kernel, kernel_width, kernel_height, + stdDeviationX, stdDeviationY); + + buffer_len = 8 + 2 * 4 * kernel_size; + buffer = (VGfloat*)malloc(buffer_len * sizeof(VGfloat)); + + buffer[0] = 0.f; + buffer[1] = 1.f; + buffer[2] = 2.f; /*unused*/ + buffer[3] = 4.f; /*unused*/ + + buffer[4] = kernel_width * kernel_height; + buffer[5] = 1.f;/*scale*/ + buffer[6] = 0.f;/*bias*/ + buffer[7] = 0.f; + + idx = 8; + for (j = 0; j < kernel_height; ++j) { + for (i = 0; i < kernel_width; ++i) { + VGint index = j * kernel_width + i; + VGfloat x, y; + + x = texture_offset(s->width, kernel_width, i, kernel_width/2); + y = texture_offset(s->height, kernel_height, j, kernel_height/2); + + buffer[idx + index*4 + 0] = x; + buffer[idx + index*4 + 1] = y; + buffer[idx + index*4 + 2] = 0.f; + buffer[idx + index*4 + 3] = 0.f; + } + } + idx += kernel_size * 4; + + for (j = 0; j < kernel_height; ++j) { + for (i = 0; i < kernel_width; ++i) { + /* transpose the kernel */ + VGint index = j * kernel_width + i; + VGint kindex = (kernel_width - i - 1) * kernel_height + (kernel_height - j - 1); + buffer[idx + index*4 + 0] = kernel[kindex]; + buffer[idx + index*4 + 1] = kernel[kindex]; + buffer[idx + index*4 + 2] = kernel[kindex]; + buffer[idx + index*4 + 3] = kernel[kindex]; + } + } + + info.dst = d; + info.src = s; + info.setup_shader = &setup_convolution; + info.user_data = (void*)(long)(buffer_len/4); + info.const_buffer = buffer; + info.const_buffer_len = buffer_len * sizeof(VGfloat); + info.tiling_mode = tilingMode; + info.extra_texture = 0; + execute_filter(ctx, &info); + + free(buffer); + free(kernel); +} + +void vgLookup(VGImage dst, VGImage src, + const VGubyte * redLUT, + const VGubyte * greenLUT, + const VGubyte * blueLUT, + const VGubyte * alphaLUT, + VGboolean outputLinear, + VGboolean outputPremultiplied) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *d, *s; + VGuint color_data[256]; + VGint i; + struct pipe_texture *lut_texture; + VGfloat buffer[4]; + struct filter_info info; + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!redLUT || !greenLUT || !blueLUT || !alphaLUT) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + d = (struct vg_image*)dst; + s = (struct vg_image*)src; + + if (vg_image_overlaps(d, s)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + for (i = 0; i < 256; ++i) { + color_data[i] = blueLUT[i] << 24 | greenLUT[i] << 16 | + redLUT[i] << 8 | alphaLUT[i]; + } + lut_texture = create_texture_1d(ctx, color_data, 255); + + buffer[0] = 0.f; + buffer[1] = 0.f; + buffer[2] = 1.f; + buffer[3] = 1.f; + + info.dst = d; + info.src = s; + info.setup_shader = &setup_lookup; + info.user_data = NULL; + info.const_buffer = buffer; + info.const_buffer_len = 4 * sizeof(VGfloat); + info.tiling_mode = VG_TILE_PAD; + info.extra_texture = lut_texture; + + execute_filter(ctx, &info); + + pipe_texture_reference(&lut_texture, NULL); +} + +void vgLookupSingle(VGImage dst, VGImage src, + const VGuint * lookupTable, + VGImageChannel sourceChannel, + VGboolean outputLinear, + VGboolean outputPremultiplied) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *d, *s; + struct pipe_texture *lut_texture; + VGfloat buffer[4]; + struct filter_info info; + VGuint color_data[256]; + VGint i; + + if (dst == VG_INVALID_HANDLE || src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!lookupTable || !is_aligned(lookupTable)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (sourceChannel != VG_RED && sourceChannel != VG_GREEN && + sourceChannel != VG_BLUE && sourceChannel != VG_ALPHA) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + d = (struct vg_image*)dst; + s = (struct vg_image*)src; + + if (vg_image_overlaps(d, s)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + for (i = 0; i < 256; ++i) { + VGuint rgba = lookupTable[i]; + VGubyte blue, green, red, alpha; + red = (rgba & 0xff000000)>>24; + green = (rgba & 0x00ff0000)>>16; + blue = (rgba & 0x0000ff00)>> 8; + alpha = (rgba & 0x000000ff)>> 0; + color_data[i] = blue << 24 | green << 16 | + red << 8 | alpha; + } + lut_texture = create_texture_1d(ctx, color_data, 256); + + buffer[0] = 0.f; + buffer[1] = 0.f; + buffer[2] = 1.f; + buffer[3] = 1.f; + + info.dst = d; + info.src = s; + info.setup_shader = &setup_lookup_single; + info.user_data = (void*)sourceChannel; + info.const_buffer = buffer; + info.const_buffer_len = 4 * sizeof(VGfloat); + info.tiling_mode = VG_TILE_PAD; + info.extra_texture = lut_texture; + + execute_filter(ctx, &info); + + pipe_texture_reference(&lut_texture, NULL); +} diff --git a/src/gallium/state_trackers/vega/api_images.c b/src/gallium/state_trackers/vega/api_images.c new file mode 100644 index 00000000000..c437553bc23 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_images.c @@ -0,0 +1,489 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "image.h" + +#include "VG/openvg.h" + +#include "vg_context.h" +#include "vg_translate.h" +#include "api_consts.h" +#include "image.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "util/u_blit.h" +#include "util/u_tile.h" +#include "util/u_memory.h" + +static INLINE VGboolean supported_image_format(VGImageFormat format) +{ + switch(format) { + case VG_sRGBX_8888: + case VG_sRGBA_8888: + case VG_sRGBA_8888_PRE: + case VG_sRGB_565: + case VG_sRGBA_5551: + case VG_sRGBA_4444: + case VG_sL_8: + case VG_lRGBX_8888: + case VG_lRGBA_8888: + case VG_lRGBA_8888_PRE: + case VG_lL_8: + case VG_A_8: + case VG_BW_1: +#ifdef OPENVG_VERSION_1_1 + case VG_A_1: + case VG_A_4: +#endif + case VG_sXRGB_8888: + case VG_sARGB_8888: + case VG_sARGB_8888_PRE: + case VG_sARGB_1555: + case VG_sARGB_4444: + case VG_lXRGB_8888: + case VG_lARGB_8888: + case VG_lARGB_8888_PRE: + case VG_sBGRX_8888: + case VG_sBGRA_8888: + case VG_sBGRA_8888_PRE: + case VG_sBGR_565: + case VG_sBGRA_5551: + case VG_sBGRA_4444: + case VG_lBGRX_8888: + case VG_lBGRA_8888: + case VG_lBGRA_8888_PRE: + case VG_sXBGR_8888: + case VG_sABGR_8888: + case VG_sABGR_8888_PRE: + case VG_sABGR_1555: + case VG_sABGR_4444: + case VG_lXBGR_8888: + case VG_lABGR_8888: + case VG_lABGR_8888_PRE: + return VG_TRUE; + default: + return VG_FALSE; + } + return VG_FALSE; +} + +VGImage vgCreateImage(VGImageFormat format, + VGint width, VGint height, + VGbitfield allowedQuality) +{ + struct vg_context *ctx = vg_current_context(); + + if (!supported_image_format(format)) { + vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR); + return VG_INVALID_HANDLE; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + if (width > vgGeti(VG_MAX_IMAGE_WIDTH) || + height > vgGeti(VG_MAX_IMAGE_HEIGHT)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + if (width * height > vgGeti(VG_MAX_IMAGE_PIXELS)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + if (!(allowedQuality & ((VG_IMAGE_QUALITY_NONANTIALIASED | + VG_IMAGE_QUALITY_FASTER | + VG_IMAGE_QUALITY_BETTER)))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + return (VGImage)image_create(format, width, height); +} + +void vgDestroyImage(VGImage image) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img = (struct vg_image *)image; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!vg_object_is_valid((void*)image, VG_OBJECT_IMAGE)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + image_destroy(img); +} + +void vgClearImage(VGImage image, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + img = (struct vg_image*)image; + + if (x + width < 0 || y + height < 0) + return; + + image_clear(img, x, y, width, height); + +} + +void vgImageSubData(VGImage image, + const void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!supported_image_format(dataFormat)) { + vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR); + return; + } + if (width <= 0 || height <= 0 || !data || !is_aligned(data)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + img = (struct vg_image*)(image); + image_sub_data(img, data, dataStride, dataFormat, + x, y, width, height); +} + +void vgGetImageSubData(VGImage image, + void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!supported_image_format(dataFormat)) { + vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR); + return; + } + if (width <= 0 || height <= 0 || !data || !is_aligned(data)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + img = (struct vg_image*)image; + image_get_sub_data(img, data, dataStride, dataFormat, + x, y, width, height); +} + +VGImage vgChildImage(VGImage parent, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *p; + + if (parent == VG_INVALID_HANDLE || + !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void*)parent) || + !vg_object_is_valid((void*)parent, VG_OBJECT_IMAGE)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return VG_INVALID_HANDLE; + } + if (width <= 0 || height <= 0 || x < 0 || y < 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + p = (struct vg_image *)parent; + if (x > p->width || y > p->height) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + if (x + width > p->width || y + height > p->height) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + return (VGImage)image_child_image(p, x, y, width, height); +} + +VGImage vgGetParent(VGImage image) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return VG_INVALID_HANDLE; + } + + img = (struct vg_image*)image; + if (img->parent) + return (VGImage)img->parent; + else + return image; +} + +void vgCopyImage(VGImage dst, VGint dx, VGint dy, + VGImage src, VGint sx, VGint sy, + VGint width, VGint height, + VGboolean dither) +{ + struct vg_context *ctx = vg_current_context(); + + if (src == VG_INVALID_HANDLE || dst == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + vg_validate_state(ctx); + image_copy((struct vg_image*)dst, dx, dy, + (struct vg_image*)src, sx, sy, + width, height, dither); +} + +void vgDrawImage(VGImage image) +{ + struct vg_context *ctx = vg_current_context(); + + if (!ctx) + return; + + if (image == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + vg_validate_state(ctx); + image_draw((struct vg_image*)image); +} + +void vgSetPixels(VGint dx, VGint dy, + VGImage src, VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + + vg_validate_state(ctx); + + if (src == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + image_set_pixels(dx, dy, (struct vg_image*)src, sx, sy, width, + height); +} + +void vgGetPixels(VGImage dst, VGint dx, VGint dy, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *img; + + if (dst == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + img = (struct vg_image*)dst; + + image_get_pixels(img, dx, dy, + sx, sy, width, height); +} + +void vgWritePixels(const void * data, VGint dataStride, + VGImageFormat dataFormat, + VGint dx, VGint dy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + + if (!supported_image_format(dataFormat)) { + vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR); + return; + } + if (!data || !is_aligned(data)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + vg_validate_state(ctx); + { + struct vg_image *img = image_create(dataFormat, width, height); + image_sub_data(img, data, dataStride, dataFormat, 0, 0, + width, height); +#if 0 + struct matrix *matrix = &ctx->state.vg.image_user_to_surface_matrix; + matrix_translate(matrix, dx, dy); + image_draw(img); + matrix_translate(matrix, -dx, -dy); +#else + /* this looks like a better approach */ + image_set_pixels(dx, dy, img, 0, 0, width, height); +#endif + image_destroy(img); + } + /* make sure rendering has completed */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); +} + +void vgReadPixels(void * data, VGint dataStride, + VGImageFormat dataFormat, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + + struct st_framebuffer *stfb = ctx->draw_buffer; + struct st_renderbuffer *strb = stfb->strb; + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + + VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4]; + VGfloat *df = (VGfloat*)temp; + VGint y = (fb->height - sy) - 1, yStep = -1; + VGint i; + VGubyte *dst = (VGubyte *)data; + VGint xoffset = 0, yoffset = 0; + + if (!supported_image_format(dataFormat)) { + vg_set_error(ctx, VG_UNSUPPORTED_IMAGE_FORMAT_ERROR); + return; + } + if (!data || !is_aligned(data)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + /* make sure rendering has completed */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + if (sx < 0) { + xoffset = -sx; + xoffset *= _vega_size_for_format(dataFormat); + width += sx; + sx = 0; + } + if (sy < 0) { + yoffset = -sy; + height += sy; + sy = 0; + y = (fb->height - sy) - 1; + yoffset *= dataStride; + } + + { + struct pipe_transfer *transfer; + + transfer = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0, + PIPE_TRANSFER_READ, + 0, 0, width, height); + + /* Do a row at a time to flip image data vertically */ + for (i = 0; i < height; i++) { +#if 0 + debug_printf("%d-%d == %d\n", sy, height, y); +#endif + pipe_get_tile_rgba(transfer, sx, y, width, 1, df); + y += yStep; + _vega_pack_rgba_span_float(ctx, width, temp, dataFormat, + dst + yoffset + xoffset); + dst += dataStride; + } + + screen->tex_transfer_destroy(transfer); + } +} + +void vgCopyPixels(VGint dx, VGint dy, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + struct st_renderbuffer *strb = ctx->draw_buffer->strb; + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + /* do nothing if we copy from outside the fb */ + if (dx >= (VGint)fb->width || dy >= (VGint)fb->height || + sx >= (VGint)fb->width || sy >= (VGint)fb->height) + return; + + vg_validate_state(ctx); + /* make sure rendering has completed */ + vgFinish(); + + vg_copy_surface(ctx, strb->surface, dx, dy, + strb->surface, sx, sy, width, height); +} diff --git a/src/gallium/state_trackers/vega/api_masks.c b/src/gallium/state_trackers/vega/api_masks.c new file mode 100644 index 00000000000..4f9f3dae173 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_masks.c @@ -0,0 +1,373 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "mask.h" +#include "renderer.h" + +#include "vg_context.h" +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "pipe/internal/p_winsys_screen.h" /* for winsys->update_buffer */ + +#include "util/u_pack_color.h" +#include "util/u_draw_quad.h" +#include "util/u_memory.h" + + +#define DISABLE_1_1_MASKING 1 + +/** + * Draw a screen-aligned quadrilateral. + * Coords are window coords with y=0=bottom. These coords will be transformed + * by the vertex shader and viewport transform. + */ +static void +draw_clear_quad(struct vg_context *st, + float x0, float y0, float x1, float y1, float z, + const VGfloat color[4]) +{ + struct pipe_context *pipe = st->pipe; + struct pipe_buffer *buf; + VGuint i; + + /* positions */ + st->clear.vertices[0][0][0] = x0; + st->clear.vertices[0][0][1] = y0; + + st->clear.vertices[1][0][0] = x1; + st->clear.vertices[1][0][1] = y0; + + st->clear.vertices[2][0][0] = x1; + st->clear.vertices[2][0][1] = y1; + + st->clear.vertices[3][0][0] = x0; + st->clear.vertices[3][0][1] = y1; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + st->clear.vertices[i][0][2] = z; + st->clear.vertices[i][0][3] = 1.0; + st->clear.vertices[i][1][0] = color[0]; + st->clear.vertices[i][1][1] = color[1]; + st->clear.vertices[i][1][2] = color[2]; + st->clear.vertices[i][1][3] = color[3]; + } + + + /* put vertex data into vbuf */ + buf = pipe_user_buffer_create(pipe->screen, + st->clear.vertices, + sizeof(st->clear.vertices)); + + + /* draw */ + if (buf) { + util_draw_vertex_buffer(pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference(&buf, NULL); + } +} + +/** + * Do vgClear by drawing a quadrilateral. + */ +static void +clear_with_quad(struct vg_context *st, float x0, float y0, + float width, float height, const VGfloat clear_color[4]) +{ + VGfloat x1, y1; + + vg_validate_state(st); + + x1 = x0 + width; + y1 = y0 + height; + + /* + printf("%s %f,%f %f,%f\n", __FUNCTION__, + x0, y0, + x1, y1); + */ + + if (st->pipe->winsys && st->pipe->winsys->update_buffer) + st->pipe->winsys->update_buffer( st->pipe->winsys, + st->pipe->priv ); + + cso_save_blend(st->cso_context); + cso_save_rasterizer(st->cso_context); + cso_save_fragment_shader(st->cso_context); + cso_save_vertex_shader(st->cso_context); + + /* blend state: RGBA masking */ + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask |= PIPE_MASK_R; + blend.colormask |= PIPE_MASK_G; + blend.colormask |= PIPE_MASK_B; + blend.colormask |= PIPE_MASK_A; + cso_set_blend(st->cso_context, &blend); + } + + cso_set_rasterizer(st->cso_context, &st->clear.raster); + + cso_set_fragment_shader_handle(st->cso_context, st->clear.fs); + cso_set_vertex_shader_handle(st->cso_context, vg_clear_vs(st)); + + /* draw quad matching scissor rect (XXX verify coord round-off) */ + draw_clear_quad(st, x0, y0, x1, y1, 0, clear_color); + + /* Restore pipe state */ + cso_restore_blend(st->cso_context); + cso_restore_rasterizer(st->cso_context); + cso_restore_fragment_shader(st->cso_context); + cso_restore_vertex_shader(st->cso_context); +} + + +void vgMask(VGHandle mask, VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + + if (width <=0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (operation < VG_CLEAR_MASK || operation > VG_SUBTRACT_MASK) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + + vg_validate_state(ctx); + + if (operation == VG_CLEAR_MASK) { + mask_fill(x, y, width, height, 0.f); + } else if (operation == VG_FILL_MASK) { + mask_fill(x, y, width, height, 1.f); + } else if (vg_object_is_valid((void*)mask, VG_OBJECT_IMAGE)) { + struct vg_image *image = (struct vg_image *)mask; + mask_using_image(image, operation, x, y, width, height); + } else if (vg_object_is_valid((void*)mask, VG_OBJECT_MASK)) { +#if DISABLE_1_1_MASKING + return; +#else + struct vg_mask_layer *layer = (struct vg_mask_layer *)mask; + mask_using_layer(layer, operation, x, y, width, height); +#endif + } else { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + } +} + +void vgClear(VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_framebuffer_state *fb; + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + vg_validate_state(ctx); +#if 0 + debug_printf("Clear [%d, %d, %d, %d] with [%f, %f, %f, %f]\n", + x, y, width, height, + ctx->state.vg.clear_color[0], + ctx->state.vg.clear_color[1], + ctx->state.vg.clear_color[2], + ctx->state.vg.clear_color[3]); +#endif + + fb = &ctx->state.g3d.fb; + /* check for a whole surface clear */ + if (!ctx->state.vg.scissoring && + (x == 0 && y == 0 && width == fb->width && height == fb->height)) { + ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_COLOR | PIPE_CLEAR_DEPTHSTENCIL, + ctx->state.vg.clear_color, 1., 0); + } else { + clear_with_quad(ctx, x, y, width, height, ctx->state.vg.clear_color); + } +} + + +#ifdef OPENVG_VERSION_1_1 + + +void vgRenderToMask(VGPath path, + VGbitfield paintModes, + VGMaskOperation operation) +{ + struct vg_context *ctx = vg_current_context(); + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!paintModes || (paintModes&(~(VG_STROKE_PATH|VG_FILL_PATH)))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (operation < VG_CLEAR_MASK || + operation > VG_SUBTRACT_MASK) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (!vg_object_is_valid((void*)path, VG_OBJECT_PATH)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + +#if DISABLE_1_1_MASKING + return; +#endif + + vg_validate_state(ctx); + + mask_render_to((struct path *)path, paintModes, operation); +} + +VGMaskLayer vgCreateMaskLayer(VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + + if (width <= 0 || height <= 0 || + width > vgGeti(VG_MAX_IMAGE_WIDTH) || + height > vgGeti(VG_MAX_IMAGE_HEIGHT)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + return (VGMaskLayer)mask_layer_create(width, height); +} + +void vgDestroyMaskLayer(VGMaskLayer maskLayer) +{ + struct vg_mask_layer *mask = 0; + struct vg_context *ctx = vg_current_context(); + + if (maskLayer == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + mask = (struct vg_mask_layer *)maskLayer; + mask_layer_destroy(mask); +} + +void vgFillMaskLayer(VGMaskLayer maskLayer, + VGint x, VGint y, + VGint width, VGint height, + VGfloat value) +{ + struct vg_mask_layer *mask = 0; + struct vg_context *ctx = vg_current_context(); + + if (maskLayer == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (value < 0 || value > 1) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (x < 0 || y < 0 || (x + width) < 0 || (y + height) < 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + mask = (struct vg_mask_layer*)maskLayer; + + if (x + width > mask_layer_width(mask) || + y + height > mask_layer_height(mask)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + +#if DISABLE_1_1_MASKING + return; +#endif + mask_layer_fill(mask, x, y, width, height, value); +} + +void vgCopyMask(VGMaskLayer maskLayer, + VGint sx, VGint sy, + VGint dx, VGint dy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_mask_layer *mask = 0; + + if (maskLayer == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (!vg_object_is_valid((void*)maskLayer, VG_OBJECT_MASK)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + +#if DISABLE_1_1_MASKING + return; +#endif + + mask = (struct vg_mask_layer*)maskLayer; + mask_copy(mask, sx, sy, dx, dy, width, height); +} + +#endif diff --git a/src/gallium/state_trackers/vega/api_misc.c b/src/gallium/state_trackers/vega/api_misc.c new file mode 100644 index 00000000000..78ba0bc1103 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_misc.c @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" + +/* Hardware Queries */ +VGHardwareQueryResult vgHardwareQuery(VGHardwareQueryType key, + VGint setting) +{ + struct vg_context *ctx = vg_current_context(); + + if (key < VG_IMAGE_FORMAT_QUERY || + key > VG_PATH_DATATYPE_QUERY) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_HARDWARE_UNACCELERATED; + } + + if (key == VG_IMAGE_FORMAT_QUERY) { + if (setting < VG_sRGBX_8888 || + setting > VG_lABGR_8888_PRE) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_HARDWARE_UNACCELERATED; + } + } else if (key == VG_PATH_DATATYPE_QUERY) { + if (setting < VG_PATH_DATATYPE_S_8 || + setting > VG_PATH_DATATYPE_F) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_HARDWARE_UNACCELERATED; + } + } + /* we're supposed to accelerate everything */ + return VG_HARDWARE_ACCELERATED; +} + +/* Renderer and Extension Information */ +const VGubyte *vgGetString(VGStringID name) +{ + struct vg_context *ctx = vg_current_context(); + static const VGubyte *vendor = (VGubyte *)"Tungsten Graphics, Inc"; + static const VGubyte *renderer = (VGubyte *)"Vega OpenVG 1.0"; + static const VGubyte *version = (VGubyte *)"1.0"; + + if (!ctx) + return NULL; + + switch(name) { + case VG_VENDOR: + return vendor; + case VG_RENDERER: + return renderer; + case VG_VERSION: + return version; + case VG_EXTENSIONS: + return NULL; + default: + return NULL; + } +} diff --git a/src/gallium/state_trackers/vega/api_paint.c b/src/gallium/state_trackers/vega/api_paint.c new file mode 100644 index 00000000000..dd3ac5bdb09 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_paint.c @@ -0,0 +1,166 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" +#include "paint.h" +#include "image.h" + +VGPaint vgCreatePaint(void) +{ + return (VGPaint) paint_create(vg_current_context()); +} + +void vgDestroyPaint(VGPaint p) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_paint *paint; + + if (p == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + paint = (struct vg_paint *)p; + paint_destroy(paint); +} + +void vgSetPaint(VGPaint paint, VGbitfield paintModes) +{ + struct vg_context *ctx = vg_current_context(); + + if (paint == VG_INVALID_HANDLE) { + /* restore the default */ + paint = (VGPaint)ctx->default_paint; + } else if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!(paintModes & ((VG_FILL_PATH|VG_STROKE_PATH)))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (paintModes & VG_FILL_PATH) { + ctx->state.vg.fill_paint = (struct vg_paint *)paint; + } + if (paintModes & VG_STROKE_PATH) { + ctx->state.vg.stroke_paint = (struct vg_paint *)paint; + } +} + +VGPaint vgGetPaint(VGPaintMode paintMode) +{ + struct vg_context *ctx = vg_current_context(); + VGPaint paint = VG_INVALID_HANDLE; + + if (paintMode < VG_STROKE_PATH || paintMode > VG_FILL_PATH) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + if (paintMode == VG_FILL_PATH) + paint = (VGPaint)ctx->state.vg.fill_paint; + else if (paintMode == VG_STROKE_PATH) + paint = (VGPaint)ctx->state.vg.stroke_paint; + + if (paint == (VGPaint)ctx->default_paint) + paint = VG_INVALID_HANDLE; + + return paint; +} + +void vgSetColor(VGPaint paint, VGuint rgba) +{ + struct vg_context *ctx = vg_current_context(); + + if (paint == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + { + struct vg_paint *p = (struct vg_paint *)paint; + paint_set_colori(p, rgba); + } +} + +VGuint vgGetColor(VGPaint paint) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_paint *p; + VGuint rgba = 0; + + if (paint == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return rgba; + } + + if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return rgba; + } + p = (struct vg_paint *)paint; + + return paint_colori(p); +} + +void vgPaintPattern(VGPaint paint, VGImage pattern) +{ + struct vg_context *ctx = vg_current_context(); + + if (paint == VG_INVALID_HANDLE || + !vg_context_is_object_valid(ctx, VG_OBJECT_PAINT, (void *)paint)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (pattern == VG_INVALID_HANDLE) { + paint_set_type((struct vg_paint*)paint, VG_PAINT_TYPE_COLOR); + return; + } + + if (!vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)pattern)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + + if (!vg_object_is_valid((void*)paint, VG_OBJECT_PAINT) || + !vg_object_is_valid((void*)pattern, VG_OBJECT_IMAGE)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + paint_set_pattern((struct vg_paint*)paint, + (struct vg_image*)pattern); +} + diff --git a/src/gallium/state_trackers/vega/api_params.c b/src/gallium/state_trackers/vega/api_params.c new file mode 100644 index 00000000000..db77fd9cb05 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_params.c @@ -0,0 +1,1673 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" +#include "paint.h" +#include "path.h" +#include "image.h" +#include "matrix.h" +#include "api_consts.h" + +#include "pipe/p_compiler.h" +#include "util/u_pointer.h" +#include "util/u_math.h" + +#include <math.h> + +static INLINE struct vg_state *current_state() +{ + struct vg_context *ctx = vg_current_context(); + if (!ctx) + return 0; + else + return &ctx->state.vg; +} + +static INLINE VGboolean count_in_bounds(VGParamType type, VGint count) +{ + if (count < 0) + return VG_FALSE; + + if (type == VG_SCISSOR_RECTS) + return (!(count % 4) && (count >= 0 || count <= VEGA_MAX_SCISSOR_RECTS * 4)); + else if (type == VG_STROKE_DASH_PATTERN) { + return count <= VEGA_MAX_DASH_COUNT; + } else { + VGint real_count = vgGetVectorSize(type); + return count == real_count; + } +} + +void vgSetf (VGParamType type, VGfloat value) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_state *state = current_state(); + VGErrorCode error = VG_NO_ERROR; + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + case VG_MAX_FLOAT: + vgSeti(type, floor(value)); + return; + break; + case VG_STROKE_LINE_WIDTH: + state->stroke.line_width.f = value; + state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(&value))); + break; + case VG_STROKE_MITER_LIMIT: + state->stroke.miter_limit.f = value; + state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(&value))); + break; + case VG_STROKE_DASH_PHASE: + state->stroke.dash_phase.f = value; + state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(&value))); + break; + default: + error = VG_ILLEGAL_ARGUMENT_ERROR; + break; + } + vg_set_error(ctx, error); +} + +void vgSeti (VGParamType type, VGint value) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_state *state = current_state(); + VGErrorCode error = VG_NO_ERROR; + + switch(type) { + case VG_MATRIX_MODE: + if (value < VG_MATRIX_PATH_USER_TO_SURFACE || +#ifdef OPENVG_VERSION_1_1 + value > VG_MATRIX_GLYPH_USER_TO_SURFACE) +#else + value > VG_MATRIX_STROKE_PAINT_TO_USER) +#endif + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->matrix_mode = value; + break; + case VG_FILL_RULE: + if (value < VG_EVEN_ODD || + value > VG_NON_ZERO) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->fill_rule = value; + break; + case VG_IMAGE_QUALITY: + state->image_quality = value; + break; + case VG_RENDERING_QUALITY: + if (value < VG_RENDERING_QUALITY_NONANTIALIASED || + value > VG_RENDERING_QUALITY_BETTER) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->rendering_quality = value; + break; + case VG_BLEND_MODE: + if (value < VG_BLEND_SRC || + value > VG_BLEND_ADDITIVE) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else { + ctx->state.dirty |= BLEND_DIRTY; + state->blend_mode = value; + } + break; + case VG_IMAGE_MODE: + if (value < VG_DRAW_IMAGE_NORMAL || + value > VG_DRAW_IMAGE_STENCIL) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->image_mode = value; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: + state->color_transform = value; +#endif + break; + case VG_STROKE_LINE_WIDTH: + state->stroke.line_width.f = value; + state->stroke.line_width.i = value; + break; + case VG_STROKE_CAP_STYLE: + if (value < VG_CAP_BUTT || + value > VG_CAP_SQUARE) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->stroke.cap_style = value; + break; + case VG_STROKE_JOIN_STYLE: + if (value < VG_JOIN_MITER || + value > VG_JOIN_BEVEL) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->stroke.join_style = value; + break; + case VG_STROKE_MITER_LIMIT: + state->stroke.miter_limit.f = value; + state->stroke.miter_limit.i = value; + break; + case VG_STROKE_DASH_PHASE: + state->stroke.dash_phase.f = value; + state->stroke.dash_phase.i = value; + break; + case VG_STROKE_DASH_PHASE_RESET: + state->stroke.dash_phase_reset = value; + break; + case VG_MASKING: + state->masking = value; + break; + case VG_SCISSORING: + state->scissoring = value; + ctx->state.dirty |= DEPTH_STENCIL_DIRTY; + break; + case VG_PIXEL_LAYOUT: + if (value < VG_PIXEL_LAYOUT_UNKNOWN || + value > VG_PIXEL_LAYOUT_BGR_HORIZONTAL) + error = VG_ILLEGAL_ARGUMENT_ERROR; + else + state->pixel_layout = value; + break; + case VG_SCREEN_LAYOUT: + /* read only ignore */ + break; + case VG_FILTER_FORMAT_LINEAR: + state->filter_format_linear = value; + break; + case VG_FILTER_FORMAT_PREMULTIPLIED: + state->filter_format_premultiplied = value; + break; + case VG_FILTER_CHANNEL_MASK: + state->filter_channel_mask = value; + break; + + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + case VG_MAX_FLOAT: + /* read only ignore */ + break; + default: + error = VG_ILLEGAL_ARGUMENT_ERROR; + break; + } + vg_set_error(ctx, error); +} + +void vgSetfv(VGParamType type, VGint count, + const VGfloat * values) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_state *state = current_state(); + VGErrorCode error = VG_NO_ERROR; + + if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + vgSeti(type, floor(values[0])); + return; + break; + case VG_SCISSOR_RECTS: { + VGint i; + VGuint *x = (VGuint*)values; + for (i = 0; i < count; ++i) { + state->scissor_rects[i].f = values[i]; + state->scissor_rects[i].i = float_to_int_floor(x[i]); + } + state->scissor_rects_num = count / 4; + ctx->state.dirty |= DEPTH_STENCIL_DIRTY; + } + break; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM_VALUES: { + VGint i; + for (i = 0; i < count; ++i) { + state->color_transform_values[i] = values[i]; + } + } + break; +#endif + case VG_STROKE_LINE_WIDTH: + state->stroke.line_width.f = values[0]; + state->stroke.line_width.i = float_to_int_floor(*((VGuint*)(values))); + break; + case VG_STROKE_MITER_LIMIT: + state->stroke.miter_limit.f = values[0]; + state->stroke.miter_limit.i = float_to_int_floor(*((VGuint*)(values))); + break; + case VG_STROKE_DASH_PATTERN: { + int i; + for (i = 0; i < count; ++i) { + state->stroke.dash_pattern[i].f = values[i]; + state->stroke.dash_pattern[i].i = + float_to_int_floor(*((VGuint*)(values + i))); + } + state->stroke.dash_pattern_num = count; + } + break; + case VG_STROKE_DASH_PHASE: + state->stroke.dash_phase.f = values[0]; + state->stroke.dash_phase.i = float_to_int_floor(*((VGuint*)(values))); + break; + case VG_TILE_FILL_COLOR: + state->tile_fill_color[0] = values[0]; + state->tile_fill_color[1] = values[1]; + state->tile_fill_color[2] = values[2]; + state->tile_fill_color[3] = values[3]; + + state->tile_fill_colori[0] = float_to_int_floor(*((VGuint*)(values + 0))); + state->tile_fill_colori[1] = float_to_int_floor(*((VGuint*)(values + 1))); + state->tile_fill_colori[2] = float_to_int_floor(*((VGuint*)(values + 2))); + state->tile_fill_colori[3] = float_to_int_floor(*((VGuint*)(values + 3))); + break; + case VG_CLEAR_COLOR: + state->clear_color[0] = values[0]; + state->clear_color[1] = values[1]; + state->clear_color[2] = values[2]; + state->clear_color[3] = values[3]; + + state->clear_colori[0] = float_to_int_floor(*((VGuint*)(values + 0))); + state->clear_colori[1] = float_to_int_floor(*((VGuint*)(values + 1))); + state->clear_colori[2] = float_to_int_floor(*((VGuint*)(values + 2))); + state->clear_colori[3] = float_to_int_floor(*((VGuint*)(values + 3))); + break; +#ifdef OPENVG_VERSION_1_1 + case VG_GLYPH_ORIGIN: + state->glyph_origin[0].f = values[0]; + state->glyph_origin[1].f = values[1]; + + state->glyph_origin[0].i = float_to_int_floor(*((VGuint*)(values + 0))); + state->glyph_origin[1].i = float_to_int_floor(*((VGuint*)(values + 1))); + break; +#endif + + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + case VG_MAX_FLOAT: + break; + default: + error = VG_ILLEGAL_ARGUMENT_ERROR; + break; + } + vg_set_error(ctx, error); +} + +void vgSetiv(VGParamType type, VGint count, + const VGint * values) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_state *state = current_state(); + + if ((count && !values) || !count_in_bounds(type, count) || !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + vgSeti(type, values[0]); + return; + break; + case VG_SCISSOR_RECTS: { + VGint i; + for (i = 0; i < count; ++i) { + state->scissor_rects[i].i = values[i]; + state->scissor_rects[i].f = values[i]; + } + state->scissor_rects_num = count / 4; + ctx->state.dirty |= DEPTH_STENCIL_DIRTY; + } + break; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM_VALUES: { + VGint i; + for (i = 0; i < count; ++i) { + state->color_transform_values[i] = values[i]; + } + } + break; +#endif + case VG_STROKE_LINE_WIDTH: + state->stroke.line_width.f = values[0]; + state->stroke.line_width.i = values[0]; + break; + case VG_STROKE_MITER_LIMIT: + state->stroke.miter_limit.f = values[0]; + state->stroke.miter_limit.i = values[0]; + break; + case VG_STROKE_DASH_PATTERN: { + int i; + for (i = 0; i < count; ++i) { + state->stroke.dash_pattern[i].f = values[i]; + state->stroke.dash_pattern[i].i = values[i]; + } + state->stroke.dash_pattern_num = count; + } + break; + case VG_STROKE_DASH_PHASE: + state->stroke.dash_phase.f = values[0]; + state->stroke.dash_phase.i = values[0]; + break; + case VG_TILE_FILL_COLOR: + state->tile_fill_color[0] = values[0]; + state->tile_fill_color[1] = values[1]; + state->tile_fill_color[2] = values[2]; + state->tile_fill_color[3] = values[3]; + + state->tile_fill_colori[0] = values[0]; + state->tile_fill_colori[1] = values[1]; + state->tile_fill_colori[2] = values[2]; + state->tile_fill_colori[3] = values[3]; + break; + case VG_CLEAR_COLOR: + state->clear_color[0] = values[0]; + state->clear_color[1] = values[1]; + state->clear_color[2] = values[2]; + state->clear_color[3] = values[3]; + + state->clear_colori[0] = values[0]; + state->clear_colori[1] = values[1]; + state->clear_colori[2] = values[2]; + state->clear_colori[3] = values[3]; + break; +#ifdef OPENVG_VERSION_1_1 + case VG_GLYPH_ORIGIN: + state->glyph_origin[0].f = values[0]; + state->glyph_origin[1].f = values[1]; + state->glyph_origin[0].i = values[0]; + state->glyph_origin[1].i = values[1]; + break; +#endif + + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + case VG_MAX_FLOAT: + break; + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} + +VGfloat vgGetf(VGParamType type) +{ + struct vg_context *ctx = vg_current_context(); + const struct vg_state *state = current_state(); + VGErrorCode error = VG_NO_ERROR; + VGfloat value = 0.0f; + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + return vgGeti(type); + break; + case VG_STROKE_LINE_WIDTH: + value = state->stroke.line_width.f; + break; + case VG_STROKE_MITER_LIMIT: + value = state->stroke.miter_limit.f; + break; + case VG_STROKE_DASH_PHASE: + value = state->stroke.dash_phase.f; + break; + + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + return vgGeti(type); + break; + case VG_MAX_FLOAT: + value = 1e+10;/*must be at least 1e+10*/ + break; + default: + error = VG_ILLEGAL_ARGUMENT_ERROR; + break; + } + vg_set_error(ctx, error); + return value; +} + +VGint vgGeti(VGParamType type) +{ + const struct vg_state *state = current_state(); + struct vg_context *ctx = vg_current_context(); + VGErrorCode error = VG_NO_ERROR; + VGint value = 0; + + switch(type) { + case VG_MATRIX_MODE: + value = state->matrix_mode; + break; + case VG_FILL_RULE: + value = state->fill_rule; + break; + case VG_IMAGE_QUALITY: + value = state->image_quality; + break; + case VG_RENDERING_QUALITY: + value = state->rendering_quality; + break; + case VG_BLEND_MODE: + value = state->blend_mode; + break; + case VG_IMAGE_MODE: + value = state->image_mode; + break; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: + value = state->color_transform; + break; +#endif + case VG_STROKE_LINE_WIDTH: + value = state->stroke.line_width.i; + break; + case VG_STROKE_CAP_STYLE: + value = state->stroke.cap_style; + break; + case VG_STROKE_JOIN_STYLE: + value = state->stroke.join_style; + break; + case VG_STROKE_MITER_LIMIT: + value = state->stroke.miter_limit.i; + break; + case VG_STROKE_DASH_PHASE: + value = state->stroke.dash_phase.i; + break; + case VG_STROKE_DASH_PHASE_RESET: + value = state->stroke.dash_phase_reset; + break; + case VG_MASKING: + value = state->masking; + break; + case VG_SCISSORING: + value = state->scissoring; + break; + case VG_PIXEL_LAYOUT: + value = state->pixel_layout; + break; + case VG_SCREEN_LAYOUT: + value = state->screen_layout; + break; + case VG_FILTER_FORMAT_LINEAR: + value = state->filter_format_linear; + break; + case VG_FILTER_FORMAT_PREMULTIPLIED: + value = state->filter_format_premultiplied; + break; + case VG_FILTER_CHANNEL_MASK: + value = state->filter_channel_mask; + break; + + case VG_MAX_SCISSOR_RECTS: + value = 32; /*must be at least 32*/ + break; + case VG_MAX_DASH_COUNT: + value = 16; /*must be at least 16*/ + break; + case VG_MAX_KERNEL_SIZE: + value = 7; /*must be at least 7*/ + break; + case VG_MAX_SEPARABLE_KERNEL_SIZE: + value = 15; /*must be at least 15*/ + break; + case VG_MAX_COLOR_RAMP_STOPS: + value = 256; /*must be at least 32*/ + break; + case VG_MAX_IMAGE_WIDTH: + value = 2048; + break; + case VG_MAX_IMAGE_HEIGHT: + value = 2048; + break; + case VG_MAX_IMAGE_PIXELS: + value = 2048*2048; + break; + case VG_MAX_IMAGE_BYTES: + value = 2048*2048 * 4; + break; + case VG_MAX_GAUSSIAN_STD_DEVIATION: + value = 128; /*must be at least 128*/ + break; + + case VG_MAX_FLOAT: { + VGfloat val = vgGetf(type); + value = float_to_int_floor(*((VGuint*)&val)); + } + break; + default: + error = VG_ILLEGAL_ARGUMENT_ERROR; + break; + } + vg_set_error(ctx, error); + return value; +} + +VGint vgGetVectorSize(VGParamType type) +{ + struct vg_context *ctx = vg_current_context(); + const struct vg_state *state = current_state(); + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: + return 1; + case VG_SCISSOR_RECTS: + return state->scissor_rects_num * 4; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: + return 1; + case VG_COLOR_TRANSFORM_VALUES: + return 8; +#endif + case VG_STROKE_LINE_WIDTH: + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_MITER_LIMIT: + return 1; + case VG_STROKE_DASH_PATTERN: + return state->stroke.dash_pattern_num; + case VG_STROKE_DASH_PHASE: + return 1; + case VG_STROKE_DASH_PHASE_RESET: + return 1; + case VG_TILE_FILL_COLOR: + return 4; + case VG_CLEAR_COLOR: + return 4; +#ifdef OPENVG_VERSION_1_1 + case VG_GLYPH_ORIGIN: + return 2; +#endif + case VG_MASKING: + return 1; + case VG_SCISSORING: + return 1; + case VG_PIXEL_LAYOUT: + return 1; + case VG_SCREEN_LAYOUT: + return 1; + case VG_FILTER_FORMAT_LINEAR: + return 1; + case VG_FILTER_FORMAT_PREMULTIPLIED: + return 1; + case VG_FILTER_CHANNEL_MASK: + return 1; + + case VG_MAX_COLOR_RAMP_STOPS: + return 1; + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_FLOAT: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + return 1; + default: + if (ctx) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return 0; + } +} + +void vgGetfv(VGParamType type, VGint count, + VGfloat * values) +{ + const struct vg_state *state = current_state(); + struct vg_context *ctx = vg_current_context(); + VGint real_count = vgGetVectorSize(type); + + if (!values || count <= 0 || count > real_count || !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + values[0] = vgGeti(type); + break; + case VG_MAX_FLOAT: + values[0] = vgGetf(type); + break; + case VG_SCISSOR_RECTS: { + VGint i; + for (i = 0; i < count; ++i) { + values[i] = state->scissor_rects[i].f; + } + } + break; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM_VALUES: { + memcpy(values, state->color_transform_values, + sizeof(VGfloat) * count); + } + break; +#endif + case VG_STROKE_LINE_WIDTH: + values[0] = state->stroke.line_width.f; + break; + case VG_STROKE_MITER_LIMIT: + values[0] = state->stroke.miter_limit.f; + break; + case VG_STROKE_DASH_PATTERN: { + VGint i; + for (i = 0; i < count; ++i) { + values[i] = state->stroke.dash_pattern[i].f; + } + } + break; + case VG_STROKE_DASH_PHASE: + values[0] = state->stroke.dash_phase.f; + break; + case VG_TILE_FILL_COLOR: + values[0] = state->tile_fill_color[0]; + values[1] = state->tile_fill_color[1]; + values[2] = state->tile_fill_color[2]; + values[3] = state->tile_fill_color[3]; + break; + case VG_CLEAR_COLOR: + values[0] = state->clear_color[0]; + values[1] = state->clear_color[1]; + values[2] = state->clear_color[2]; + values[3] = state->clear_color[3]; + break; +#ifdef OPENVG_VERSION_1_1 + case VG_GLYPH_ORIGIN: + values[0] = state->glyph_origin[0].f; + values[1] = state->glyph_origin[1].f; + break; +#endif + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} + +void vgGetiv(VGParamType type, VGint count, + VGint * values) +{ + const struct vg_state *state = current_state(); + struct vg_context *ctx = vg_current_context(); + VGint real_count = vgGetVectorSize(type); + + if (!values || count <= 0 || count > real_count || !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(type) { + case VG_MATRIX_MODE: + case VG_FILL_RULE: + case VG_IMAGE_QUALITY: + case VG_RENDERING_QUALITY: + case VG_BLEND_MODE: + case VG_IMAGE_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM: +#endif + case VG_STROKE_CAP_STYLE: + case VG_STROKE_JOIN_STYLE: + case VG_STROKE_DASH_PHASE_RESET: + case VG_MASKING: + case VG_SCISSORING: + case VG_PIXEL_LAYOUT: + case VG_SCREEN_LAYOUT: + case VG_FILTER_FORMAT_LINEAR: + case VG_FILTER_FORMAT_PREMULTIPLIED: + case VG_FILTER_CHANNEL_MASK: + case VG_MAX_SCISSOR_RECTS: + case VG_MAX_DASH_COUNT: + case VG_MAX_KERNEL_SIZE: + case VG_MAX_SEPARABLE_KERNEL_SIZE: + case VG_MAX_COLOR_RAMP_STOPS: + case VG_MAX_IMAGE_WIDTH: + case VG_MAX_IMAGE_HEIGHT: + case VG_MAX_IMAGE_PIXELS: + case VG_MAX_IMAGE_BYTES: + case VG_MAX_GAUSSIAN_STD_DEVIATION: + values[0] = vgGeti(type); + break; + case VG_MAX_FLOAT: { + VGfloat val = vgGetf(type); + values[0] = float_to_int_floor(*((VGuint*)&val)); + } + break; + case VG_SCISSOR_RECTS: { + VGint i; + for (i = 0; i < count; ++i) { + values[i] = state->scissor_rects[i].i; + } + } + break; +#ifdef OPENVG_VERSION_1_1 + case VG_COLOR_TRANSFORM_VALUES: { + VGint i; + VGuint *x = (VGuint*)state->color_transform_values; + for (i = 0; i < count; ++i) { + values[i] = float_to_int_floor(x[i]); + } + } + break; +#endif + case VG_STROKE_LINE_WIDTH: + values[0] = state->stroke.line_width.i; + break; + case VG_STROKE_MITER_LIMIT: + values[0] = state->stroke.miter_limit.i; + break; + case VG_STROKE_DASH_PATTERN: { + VGint i; + for (i = 0; i < count; ++i) { + values[i] = state->stroke.dash_pattern[i].i; + } + } + break; + case VG_STROKE_DASH_PHASE: + values[0] = state->stroke.dash_phase.i; + break; + case VG_TILE_FILL_COLOR: + values[0] = state->tile_fill_colori[0]; + values[1] = state->tile_fill_colori[1]; + values[2] = state->tile_fill_colori[2]; + values[3] = state->tile_fill_colori[3]; + break; + case VG_CLEAR_COLOR: + values[0] = state->clear_colori[0]; + values[1] = state->clear_colori[1]; + values[2] = state->clear_colori[2]; + values[3] = state->clear_colori[3]; + break; +#ifdef OPENVG_VERSION_1_1 + case VG_GLYPH_ORIGIN: + values[0] = state->glyph_origin[0].i; + values[1] = state->glyph_origin[1].i; + break; +#endif + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} + +void vgSetParameterf(VGHandle object, + VGint paramType, + VGfloat value) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + + if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_PATTERN_TILING_MODE: + vgSetParameteri(object, paramType, floor(value)); + return; + break; + case VG_PAINT_COLOR: + case VG_PAINT_COLOR_RAMP_STOPS: + case VG_PAINT_LINEAR_GRADIENT: + case VG_PAINT_RADIAL_GRADIENT: + /* it's an error if paramType refers to a vector parameter */ + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: { + struct vg_paint *p = (struct vg_paint *)object; + paint_set_color_ramp_premultiplied(p, value); + } + break; + + case VG_PATH_DATATYPE: + case VG_PATH_FORMAT: + case VG_PATH_SCALE: + case VG_PATH_BIAS: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: + +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + /* read only don't produce an error */ + break; +#endif + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} + +void vgSetParameteri(VGHandle object, + VGint paramType, + VGint value) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + + if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: + if (value < VG_PAINT_TYPE_COLOR || + value > VG_PAINT_TYPE_PATTERN) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)ptr; + paint_set_type(paint, value); + } + break; + case VG_PAINT_COLOR: + case VG_PAINT_COLOR_RAMP_STOPS: + case VG_PAINT_LINEAR_GRADIENT: + case VG_PAINT_RADIAL_GRADIENT: + /* it's an error if paramType refers to a vector parameter */ + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + if (value < VG_COLOR_RAMP_SPREAD_PAD || + value > VG_COLOR_RAMP_SPREAD_REFLECT) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)ptr; + paint_set_spread_mode(paint, value); + } + break; + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: { + struct vg_paint *p = (struct vg_paint *)object; + paint_set_color_ramp_premultiplied(p, value); + } + break; + case VG_PAINT_PATTERN_TILING_MODE: + if (value < VG_TILE_FILL || + value > VG_TILE_REFLECT) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)ptr; + paint_set_pattern_tiling(paint, value); + } + break; + + case VG_PATH_DATATYPE: + case VG_PATH_FORMAT: + case VG_PATH_SCALE: + case VG_PATH_BIAS: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: + +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + /* read only don't produce an error */ + break; +#endif + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } +} + +void vgSetParameterfv(VGHandle object, + VGint paramType, + VGint count, + const VGfloat * values) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + VGint real_count = vgGetParameterVectorSize(object, paramType); + + if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (count < 0 || count < real_count || + (values == NULL && count != 0) || + !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: + case VG_PAINT_PATTERN_TILING_MODE: + if (count != 1) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else + vgSetParameterf(object, paramType, values[0]); + return; + break; + case VG_PAINT_COLOR: { + if (count != 4) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)object; + paint_set_color(paint, values); + } + } + break; + case VG_PAINT_COLOR_RAMP_STOPS: { + if (count && count < 4) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)object; + count = MIN2(count, VEGA_MAX_COLOR_RAMP_STOPS); + paint_set_ramp_stops(paint, values, count); + { + VGint stopsi[VEGA_MAX_COLOR_RAMP_STOPS]; + int i = 0; + for (i = 0; i < count; ++i) { + stopsi[i] = float_to_int_floor(*((VGuint*)(values + i))); + } + paint_set_ramp_stopsi(paint, stopsi, count); + } + } + } + break; + case VG_PAINT_LINEAR_GRADIENT: { + if (count != 4) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)object; + paint_set_linear_gradient(paint, values); + { + VGint vals[4]; + vals[0] = FLT_TO_INT(values[0]); + vals[1] = FLT_TO_INT(values[1]); + vals[2] = FLT_TO_INT(values[2]); + vals[3] = FLT_TO_INT(values[3]); + paint_set_linear_gradienti(paint, vals); + } + } + } + break; + case VG_PAINT_RADIAL_GRADIENT: { + if (count != 5) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)object; + paint_set_radial_gradient(paint, values); + { + VGint vals[5]; + vals[0] = FLT_TO_INT(values[0]); + vals[1] = FLT_TO_INT(values[1]); + vals[2] = FLT_TO_INT(values[2]); + vals[3] = FLT_TO_INT(values[3]); + vals[4] = FLT_TO_INT(values[4]); + paint_set_radial_gradienti(paint, vals); + } + } + } + break; + + case VG_PATH_DATATYPE: + case VG_PATH_FORMAT: + case VG_PATH_SCALE: + case VG_PATH_BIAS: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + /* read only don't produce an error */ + break; +#endif + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } +} + +void vgSetParameteriv(VGHandle object, + VGint paramType, + VGint count, + const VGint * values) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + VGint real_count = vgGetParameterVectorSize(object, paramType); + + if (!object || object == VG_INVALID_HANDLE || !is_aligned(ptr)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (count < 0 || count < real_count || + (values == NULL && count != 0) || + !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: + case VG_PAINT_PATTERN_TILING_MODE: + if (count != 1) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else + vgSetParameteri(object, paramType, values[0]); + return; + break; + case VG_PAINT_COLOR: { + if (count != 4) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + struct vg_paint *paint = (struct vg_paint *)object; + paint_set_coloriv(paint, values); + } + } + break; + case VG_PAINT_COLOR_RAMP_STOPS: { + if ((count % 5)) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + VGfloat *vals = 0; + int i; + struct vg_paint *paint = (struct vg_paint *)object; + if (count) { + vals = malloc(sizeof(VGfloat)*count); + for (i = 0; i < count; ++i) + vals[i] = values[i]; + } + + paint_set_ramp_stopsi(paint, values, count); + paint_set_ramp_stops(paint, vals, count); + free(vals); + } + } + break; + case VG_PAINT_LINEAR_GRADIENT: { + if (count != 4) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + VGfloat vals[4]; + struct vg_paint *paint = (struct vg_paint *)object; + vals[0] = values[0]; + vals[1] = values[1]; + vals[2] = values[2]; + vals[3] = values[3]; + paint_set_linear_gradient(paint, vals); + paint_set_linear_gradienti(paint, values); + } + } + break; + case VG_PAINT_RADIAL_GRADIENT: { + if (count != 5) + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + else { + VGfloat vals[5]; + struct vg_paint *paint = (struct vg_paint *)object; + vals[0] = values[0]; + vals[1] = values[1]; + vals[2] = values[2]; + vals[3] = values[3]; + vals[4] = values[4]; + paint_set_radial_gradient(paint, vals); + paint_set_radial_gradienti(paint, values); + } + } + break; + case VG_PATH_DATATYPE: + case VG_PATH_FORMAT: + case VG_PATH_SCALE: + case VG_PATH_BIAS: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + /* read only don't produce an error */ + break; + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } +} + +VGint vgGetParameterVectorSize(VGHandle object, + VGint paramType) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + + if (!ptr || object == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return 0; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: + case VG_PAINT_PATTERN_TILING_MODE: + return 1; + case VG_PAINT_COLOR: + return 4; + case VG_PAINT_COLOR_RAMP_STOPS: { + struct vg_paint *p = (struct vg_paint *)object; + return paint_num_ramp_stops(p); + } + break; + case VG_PAINT_LINEAR_GRADIENT: + return 4; + case VG_PAINT_RADIAL_GRADIENT: + return 5; + + + case VG_PATH_FORMAT: + case VG_PATH_DATATYPE: + case VG_PATH_SCALE: + case VG_PATH_BIAS: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + return 1; + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: + return 1; + +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + return 1; +#endif + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } + return 0; +} + + +VGfloat vgGetParameterf(VGHandle object, + VGint paramType) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + + if (!ptr || object == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return 0; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: + case VG_PAINT_PATTERN_TILING_MODE: + return vgGetParameteri(object, paramType); + break; + case VG_PAINT_COLOR: + case VG_PAINT_COLOR_RAMP_STOPS: + case VG_PAINT_LINEAR_GRADIENT: + case VG_PAINT_RADIAL_GRADIENT: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + + case VG_PATH_FORMAT: + return VG_PATH_FORMAT_STANDARD; + case VG_PATH_SCALE: { + struct path *p = (struct path*)object; + return path_scale(p); + } + case VG_PATH_BIAS: { + struct path *p = (struct path*)object; + return path_bias(p); + } + case VG_PATH_DATATYPE: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + return vgGetParameteri(object, paramType); + break; + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + return vgGetParameteri(object, paramType); + break; +#endif + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } + return 0; +} + +VGint vgGetParameteri(VGHandle object, + VGint paramType) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + + if (!ptr || object == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return 0; + } + + switch(paramType) { + case VG_PAINT_TYPE: { + struct vg_paint *paint = (struct vg_paint *)ptr; + return paint_type(paint); + } + break; + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: { + struct vg_paint *p = (struct vg_paint *)object; + return paint_spread_mode(p); + } + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: { + struct vg_paint *p = (struct vg_paint *)object; + return paint_color_ramp_premultiplied(p); + } + break; + case VG_PAINT_PATTERN_TILING_MODE: { + struct vg_paint *p = (struct vg_paint *)object; + return paint_pattern_tiling(p); + } + break; + case VG_PAINT_COLOR: + case VG_PAINT_COLOR_RAMP_STOPS: + case VG_PAINT_LINEAR_GRADIENT: + case VG_PAINT_RADIAL_GRADIENT: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + + case VG_PATH_FORMAT: + return VG_PATH_FORMAT_STANDARD; + case VG_PATH_SCALE: + case VG_PATH_BIAS: + return vgGetParameterf(object, paramType); + case VG_PATH_DATATYPE: { + struct path *p = (struct path*)object; + return path_datatype(p); + } + case VG_PATH_NUM_SEGMENTS: { + struct path *p = (struct path*)object; + return path_num_segments(p); + } + case VG_PATH_NUM_COORDS: { + struct path *p = (struct path*)object; + return path_num_coords(p); + } + break; + + case VG_IMAGE_FORMAT: { + struct vg_image *img = (struct vg_image*)object; + return img->format; + } + break; + case VG_IMAGE_WIDTH: { + struct vg_image *img = (struct vg_image*)object; + return img->width; + } + break; + case VG_IMAGE_HEIGHT: { + struct vg_image *img = (struct vg_image*)object; + return img->height; + } + break; + +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: { + return 1; + } + break; +#endif + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } + return 0; +} + +void vgGetParameterfv(VGHandle object, + VGint paramType, + VGint count, + VGfloat * values) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + VGint real_count = vgGetParameterVectorSize(object, paramType); + + if (!ptr || object == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!values || count <= 0 || count > real_count || + !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: { + struct vg_paint *p = (struct vg_paint *)object; + values[0] = paint_type(p); + } + break; + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: { + struct vg_paint *p = (struct vg_paint *)object; + values[0] = paint_spread_mode(p); + } + break; + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: { + struct vg_paint *p = (struct vg_paint *)object; + values[0] = paint_color_ramp_premultiplied(p); + } + break; + case VG_PAINT_PATTERN_TILING_MODE: { + values[0] = vgGetParameterf(object, paramType); + } + break; + case VG_PAINT_COLOR: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_get_color(paint, values); + } + break; + case VG_PAINT_COLOR_RAMP_STOPS: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_ramp_stops(paint, values, count); + } + break; + case VG_PAINT_LINEAR_GRADIENT: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_linear_gradient(paint, values); + } + break; + case VG_PAINT_RADIAL_GRADIENT: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_radial_gradient(paint, values); + } + break; + + case VG_PATH_FORMAT: + case VG_PATH_DATATYPE: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + values[0] = vgGetParameteri(object, paramType); + break; + case VG_PATH_SCALE: + case VG_PATH_BIAS: + values[0] = vgGetParameterf(object, paramType); + break; + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + values[0] = vgGetParameteri(object, paramType); + break; +#endif + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} + +void vgGetParameteriv(VGHandle object, + VGint paramType, + VGint count, + VGint * values) +{ + struct vg_context *ctx = vg_current_context(); + void *ptr = (void*)object; + VGint real_count = vgGetParameterVectorSize(object, paramType); + + if (!ptr || object == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!values || count <= 0 || count > real_count || + !is_aligned(values)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + switch(paramType) { + case VG_PAINT_TYPE: + case VG_PAINT_COLOR_RAMP_SPREAD_MODE: + case VG_PAINT_COLOR_RAMP_PREMULTIPLIED: + case VG_PAINT_PATTERN_TILING_MODE: +#ifdef OPENVG_VERSION_1_1 + case VG_FONT_NUM_GLYPHS: + values[0] = vgGetParameteri(object, paramType); + break; +#endif + case VG_PAINT_COLOR: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_get_coloriv(paint, values); + } + break; + case VG_PAINT_COLOR_RAMP_STOPS: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_ramp_stopsi(paint, values, count); + } + break; + case VG_PAINT_LINEAR_GRADIENT: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_linear_gradienti(paint, values); + } + break; + case VG_PAINT_RADIAL_GRADIENT: { + struct vg_paint *paint = (struct vg_paint *)object; + paint_radial_gradienti(paint, values); + } + break; + + case VG_PATH_SCALE: + case VG_PATH_BIAS: + values[0] = vgGetParameterf(object, paramType); + break; + case VG_PATH_FORMAT: + case VG_PATH_DATATYPE: + case VG_PATH_NUM_SEGMENTS: + case VG_PATH_NUM_COORDS: + values[0] = vgGetParameteri(object, paramType); + break; + + case VG_IMAGE_FORMAT: + case VG_IMAGE_WIDTH: + case VG_IMAGE_HEIGHT: + values[0] = vgGetParameteri(object, paramType); + break; + + default: + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + break; + } +} diff --git a/src/gallium/state_trackers/vega/api_path.c b/src/gallium/state_trackers/vega/api_path.c new file mode 100644 index 00000000000..a6b7a2bb93a --- /dev/null +++ b/src/gallium/state_trackers/vega/api_path.c @@ -0,0 +1,488 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" +#include "path.h" +#include "polygon.h" +#include "paint.h" + +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "util/u_draw_quad.h" + +VGPath vgCreatePath(VGint pathFormat, + VGPathDatatype datatype, + VGfloat scale, VGfloat bias, + VGint segmentCapacityHint, + VGint coordCapacityHint, + VGbitfield capabilities) +{ + struct vg_context *ctx = vg_current_context(); + + if (pathFormat != VG_PATH_FORMAT_STANDARD) { + vg_set_error(ctx, VG_UNSUPPORTED_PATH_FORMAT_ERROR); + return VG_INVALID_HANDLE; + } + if (datatype < VG_PATH_DATATYPE_S_8 || + datatype > VG_PATH_DATATYPE_F) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + if (!scale) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + return (VGPath)path_create(datatype, scale, bias, + segmentCapacityHint, coordCapacityHint, + capabilities); +} + +void vgClearPath(VGPath path, VGbitfield capabilities) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + p = (struct path *)path; + path_clear(p, capabilities); +} + +void vgDestroyPath(VGPath p) +{ + struct path *path = 0; + struct vg_context *ctx = vg_current_context(); + + if (p == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + path = (struct path *)p; + path_destroy(path); +} + +void vgRemovePathCapabilities(VGPath path, + VGbitfield capabilities) +{ + struct vg_context *ctx = vg_current_context(); + VGbitfield current; + struct path *p; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + p = (struct path*)path; + current = path_capabilities(p); + path_set_capabilities(p, (current & + (~(capabilities & VG_PATH_CAPABILITY_ALL)))); +} + +VGbitfield vgGetPathCapabilities(VGPath path) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return 0; + } + p = (struct path*)path; + return path_capabilities(p); +} + +void vgAppendPath(VGPath dstPath, VGPath srcPath) +{ + struct vg_context *ctx = vg_current_context(); + struct path *src, *dst; + + if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + src = (struct path *)srcPath; + dst = (struct path *)dstPath; + + if (!(path_capabilities(src) & VG_PATH_CAPABILITY_APPEND_FROM) || + !(path_capabilities(dst) & VG_PATH_CAPABILITY_APPEND_TO)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + path_append_path(dst, src); +} + +void vgAppendPathData(VGPath dstPath, + VGint numSegments, + const VGubyte * pathSegments, + const void * pathData) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + VGint i; + + if (dstPath == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!pathSegments) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (numSegments <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + for (i = 0; i < numSegments; ++i) { + if (pathSegments[i] < VG_CLOSE_PATH || + pathSegments[i] > VG_LCWARC_TO_REL) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + } + + p = (struct path*)dstPath; + + if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!(path_capabilities(p)&VG_PATH_CAPABILITY_APPEND_TO)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + + path_append_data(p, numSegments, pathSegments, pathData); +} + +void vgModifyPathCoords(VGPath dstPath, + VGint startIndex, + VGint numSegments, + const void * pathData) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + + if (dstPath == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (startIndex < 0 || numSegments <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + p = (struct path *)dstPath; + + if (!pathData || !is_aligned_to(pathData, path_datatype_size(p))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (startIndex + numSegments > path_num_segments(p)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (!(path_capabilities(p)&VG_PATH_CAPABILITY_MODIFY)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + path_modify_coords(p, startIndex, numSegments, pathData); +} + +void vgTransformPath(VGPath dstPath, VGPath srcPath) +{ + struct vg_context *ctx = vg_current_context(); + struct path *src = 0, *dst = 0; + + if (dstPath == VG_INVALID_HANDLE || srcPath == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + src = (struct path *)srcPath; + dst = (struct path *)dstPath; + + if (!(path_capabilities(src) & VG_PATH_CAPABILITY_TRANSFORM_FROM) || + !(path_capabilities(dst) & VG_PATH_CAPABILITY_TRANSFORM_TO)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + path_transform(dst, src); +} + +VGboolean vgInterpolatePath(VGPath dstPath, + VGPath startPath, + VGPath endPath, + VGfloat amount) +{ + struct vg_context *ctx = vg_current_context(); + struct path *start = 0, *dst = 0, *end = 0; + + if (dstPath == VG_INVALID_HANDLE || + startPath == VG_INVALID_HANDLE || + endPath == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return VG_FALSE; + } + dst = (struct path *)dstPath; + start = (struct path *)startPath; + end = (struct path *)endPath; + + if (!(path_capabilities(dst) & VG_PATH_CAPABILITY_INTERPOLATE_TO) || + !(path_capabilities(start) & VG_PATH_CAPABILITY_INTERPOLATE_FROM) || + !(path_capabilities(end) & VG_PATH_CAPABILITY_INTERPOLATE_FROM)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return VG_FALSE; + } + + return path_interpolate(dst, + start, end, amount); +} + +VGfloat vgPathLength(VGPath path, + VGint startSegment, + VGint numSegments) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return -1; + } + if (startSegment < 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return -1; + } + if (numSegments <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return -1; + } + p = (struct path*)path; + + if (!(path_capabilities(p) & VG_PATH_CAPABILITY_PATH_LENGTH)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return -1; + } + if (startSegment + numSegments > path_num_segments(p)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return -1; + } + + return path_length(p, startSegment, numSegments); +} + +void vgPointAlongPath(VGPath path, + VGint startSegment, + VGint numSegments, + VGfloat distance, + VGfloat * x, VGfloat * y, + VGfloat * tangentX, + VGfloat * tangentY) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (startSegment < 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (numSegments <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!is_aligned(x) || !is_aligned(y) || + !is_aligned(tangentX) || !is_aligned(tangentY)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + p = (struct path*)path; + + caps = path_capabilities(p); + if (!(caps & VG_PATH_CAPABILITY_POINT_ALONG_PATH) || + !(caps & VG_PATH_CAPABILITY_TANGENT_ALONG_PATH)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + + if (startSegment + numSegments > path_num_segments(p)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + { + VGfloat point[2], normal[2]; + path_point(p, startSegment, numSegments, distance, + point, normal); + if (x) + *x = point[0]; + if (y) + *y = point[1]; + if (tangentX) + *tangentX = -normal[1]; + if (tangentY) + *tangentY = normal[0]; + } +} + +void vgPathBounds(VGPath path, + VGfloat * minX, + VGfloat * minY, + VGfloat * width, + VGfloat * height) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!minX || !minY || !width || !height) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!is_aligned(minX) || !is_aligned(minY) || + !is_aligned(width) || !is_aligned(height)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + p = (struct path*)path; + + caps = path_capabilities(p); + if (!(caps & VG_PATH_CAPABILITY_PATH_BOUNDS)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + + path_bounding_rect(p, minX, minY, width, height); +} + +void vgPathTransformedBounds(VGPath path, + VGfloat * minX, + VGfloat * minY, + VGfloat * width, + VGfloat * height) +{ + struct vg_context *ctx = vg_current_context(); + struct path *p = 0; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!minX || !minY || !width || !height) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (!is_aligned(minX) || !is_aligned(minY) || + !is_aligned(width) || !is_aligned(height)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + p = (struct path*)path; + + caps = path_capabilities(p); + if (!(caps & VG_PATH_CAPABILITY_PATH_TRANSFORMED_BOUNDS)) { + vg_set_error(ctx, VG_PATH_CAPABILITY_ERROR); + return; + } + +#if 0 + /* faster, but seems to have precision problems... */ + path_bounding_rect(p, minX, minY, width, height); + if (*width > 0 && *height > 0) { + VGfloat pts[] = {*minX, *minY, + *minX + *width, *minY, + *minX + *width, *minY + *height, + *minX, *minY + *height}; + struct matrix *matrix = &ctx->state.vg.path_user_to_surface_matrix; + VGfloat maxX, maxY; + matrix_map_point(matrix, pts[0], pts[1], pts + 0, pts + 1); + matrix_map_point(matrix, pts[2], pts[3], pts + 2, pts + 3); + matrix_map_point(matrix, pts[4], pts[5], pts + 4, pts + 5); + matrix_map_point(matrix, pts[6], pts[7], pts + 6, pts + 7); + *minX = MIN2(pts[0], MIN2(pts[2], MIN2(pts[4], pts[6]))); + *minY = MIN2(pts[1], MIN2(pts[3], MIN2(pts[5], pts[7]))); + maxX = MAX2(pts[0], MAX2(pts[2], MAX2(pts[4], pts[6]))); + maxY = MAX2(pts[1], MAX2(pts[3], MAX2(pts[5], pts[7]))); + *width = maxX - *minX; + *height = maxY - *minY; + } +#else + { + struct path *dst = path_create(VG_PATH_DATATYPE_F, 1.0, 0, + 0, 0, VG_PATH_CAPABILITY_ALL); + path_transform(dst, p); + path_bounding_rect(dst, minX, minY, width, height); + path_destroy(dst); + } +#endif +} + + +void vgDrawPath(VGPath path, VGbitfield paintModes) +{ + struct vg_context *ctx = vg_current_context(); + + if (path == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + if (!(paintModes & (VG_STROKE_PATH | VG_FILL_PATH))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (path_is_empty((struct path*)path)) + return; + path_render((struct path*)path, paintModes); +} + diff --git a/src/gallium/state_trackers/vega/api_text.c b/src/gallium/state_trackers/vega/api_text.c new file mode 100644 index 00000000000..d8411cf3e87 --- /dev/null +++ b/src/gallium/state_trackers/vega/api_text.c @@ -0,0 +1,258 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" + +#include "util/u_memory.h" + +#ifdef OPENVG_VERSION_1_1 + +struct vg_font { + struct vg_object base; + + VGint glyph_indices[200]; + VGint num_glyphs; +}; + +VGFont vgCreateFont(VGint glyphCapacityHint) +{ + struct vg_font *font = 0; + struct vg_context *ctx = vg_current_context(); + + if (glyphCapacityHint < 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return VG_INVALID_HANDLE; + } + + font = CALLOC_STRUCT(vg_font); + vg_init_object(&font->base, ctx, VG_OBJECT_FONT); + vg_context_add_object(ctx, VG_OBJECT_FONT, font); + return (VGFont)font; +} + +void vgDestroyFont(VGFont f) +{ + struct vg_font *font = (struct vg_font *)f; + struct vg_context *ctx = vg_current_context(); + + if (f == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + vg_context_remove_object(ctx, VG_OBJECT_FONT, font); + /*free(font);*/ +} + +void vgSetGlyphToPath(VGFont font, + VGuint glyphIndex, + VGPath path, + VGboolean isHinted, + VGfloat glyphOrigin [2], + VGfloat escapement[2]) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_object *pathObj; + struct vg_font *f; + + if (font == VG_INVALID_HANDLE || + !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!glyphOrigin || !escapement || + !is_aligned(glyphOrigin) || !is_aligned(escapement)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (path != VG_INVALID_HANDLE && + !vg_context_is_object_valid(ctx, VG_OBJECT_PATH, (void *)path)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + pathObj = (struct vg_object*)path; + if (pathObj && pathObj->type != VG_OBJECT_PATH) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + + f = (struct vg_font*)font; + f->glyph_indices[f->num_glyphs] = glyphIndex; + ++f->num_glyphs; +} + +void vgSetGlyphToImage(VGFont font, + VGuint glyphIndex, + VGImage image, + VGfloat glyphOrigin [2], + VGfloat escapement[2]) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_object *img_obj; + struct vg_font *f; + + if (font == VG_INVALID_HANDLE || + !vg_context_is_object_valid(ctx, VG_OBJECT_FONT, (void *)font)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (!glyphOrigin || !escapement || + !is_aligned(glyphOrigin) || !is_aligned(escapement)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (image != VG_INVALID_HANDLE && + !vg_context_is_object_valid(ctx, VG_OBJECT_IMAGE, (void *)image)) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + img_obj = (struct vg_object*)image; + if (img_obj && img_obj->type != VG_OBJECT_IMAGE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + f = (struct vg_font*)font; + f->glyph_indices[f->num_glyphs] = glyphIndex; + ++f->num_glyphs; +} + +static INLINE VGboolean font_contains_glyph(struct vg_font *font, + VGuint glyph_index) +{ + VGint i; + for (i = 0; i < font->num_glyphs; ++i) { + if (font->glyph_indices[i] == glyph_index) { + return VG_TRUE; + } + } + return VG_FALSE; +} + +void vgClearGlyph(VGFont font, + VGuint glyphIndex) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_font *f; + VGint i; + + if (font == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (glyphIndex <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + f = (struct vg_font*)font; + if (!font_contains_glyph(f, glyphIndex)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + for (i = 0; i < f->num_glyphs; ++i) { + if (f->glyph_indices[i] == glyphIndex) { + /*FIXME*/ + f->glyph_indices[f->num_glyphs] = 0; + --f->num_glyphs; + return; + } + } +} + +void vgDrawGlyph(VGFont font, + VGuint glyphIndex, + VGbitfield paintModes, + VGboolean allowAutoHinting) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_font *f; + + if (font == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (glyphIndex <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + f = (struct vg_font*)font; + if (!font_contains_glyph(f, glyphIndex)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } +} + +void vgDrawGlyphs(VGFont font, + VGint glyphCount, + VGuint *glyphIndices, + VGfloat *adjustments_x, + VGfloat *adjustments_y, + VGbitfield paintModes, + VGboolean allowAutoHinting) +{ + struct vg_context *ctx = vg_current_context(); + VGint i; + struct vg_font *f; + + if (font == VG_INVALID_HANDLE) { + vg_set_error(ctx, VG_BAD_HANDLE_ERROR); + return; + } + if (glyphCount <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (!glyphIndices || !is_aligned(glyphIndices)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (!adjustments_x || !is_aligned(adjustments_x) || + !adjustments_y || !is_aligned(adjustments_y)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + if (paintModes & (~(VG_STROKE_PATH|VG_FILL_PATH))) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + f = (struct vg_font*)font; + for (i = 0; i < glyphCount; ++i) { + VGuint glyph_index = glyphIndices[i]; + if (!font_contains_glyph(f, glyph_index)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + } +} + +#endif diff --git a/src/gallium/state_trackers/vega/api_transform.c b/src/gallium/state_trackers/vega/api_transform.c new file mode 100644 index 00000000000..763a5ec415c --- /dev/null +++ b/src/gallium/state_trackers/vega/api_transform.c @@ -0,0 +1,128 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" + +#include "vg_context.h" + +#include "matrix.h" + +void vgLoadIdentity(void) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *mat = vg_state_matrix(&ctx->state.vg); + matrix_load_identity(mat); +} + +void vgLoadMatrix(const VGfloat * m) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *mat; + + if (!ctx) + return; + + if (!m || !is_aligned(m)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + mat = vg_state_matrix(&ctx->state.vg); + matrix_init(mat, m); + if (!matrix_is_affine(mat)) { + if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) { + matrix_make_affine(mat); + } + } +} + +void vgGetMatrix(VGfloat * m) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *mat; + + if (!ctx) + return; + + if (!m || !is_aligned(m)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + mat = vg_state_matrix(&ctx->state.vg); + memcpy(m, mat->m, sizeof(VGfloat)*9); +} + +void vgMultMatrix(const VGfloat * m) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *dst, src; + + if (!ctx) + return; + + if (!m || !is_aligned(m)) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + matrix_init(&src, m); + dst = vg_state_matrix(&ctx->state.vg); + if (!matrix_is_affine(&src)) { + if (ctx->state.vg.matrix_mode != VG_MATRIX_IMAGE_USER_TO_SURFACE) { + matrix_make_affine(&src); + } + } + matrix_mult(dst, &src); + +} + +void vgTranslate(VGfloat tx, VGfloat ty) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *dst = vg_state_matrix(&ctx->state.vg); + matrix_translate(dst, tx, ty); +} + +void vgScale(VGfloat sx, VGfloat sy) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *dst = vg_state_matrix(&ctx->state.vg); + matrix_scale(dst, sx, sy); +} + +void vgShear(VGfloat shx, VGfloat shy) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *dst = vg_state_matrix(&ctx->state.vg); + matrix_shear(dst, shx, shy); +} + +void vgRotate(VGfloat angle) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *dst = vg_state_matrix(&ctx->state.vg); + matrix_rotate(dst, angle); +} diff --git a/src/gallium/state_trackers/vega/arc.c b/src/gallium/state_trackers/vega/arc.c new file mode 100644 index 00000000000..e74c7f03345 --- /dev/null +++ b/src/gallium/state_trackers/vega/arc.c @@ -0,0 +1,708 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "arc.h" + +#include "matrix.h" +#include "bezier.h" +#include "polygon.h" +#include "stroker.h" +#include "path.h" + +#include "util/u_debug.h" + +#include <math.h> + +#ifndef M_PI +#define M_PI 3.14159265358979323846 +#endif + +#define DEBUG_ARCS 0 + +static const VGfloat two_pi = M_PI * 2; + + +static const double coeffs3Low[2][4][4] = { + { + { 3.85268, -21.229, -0.330434, 0.0127842 }, + { -1.61486, 0.706564, 0.225945, 0.263682 }, + { -0.910164, 0.388383, 0.00551445, 0.00671814 }, + { -0.630184, 0.192402, 0.0098871, 0.0102527 } + }, + { + { -0.162211, 9.94329, 0.13723, 0.0124084 }, + { -0.253135, 0.00187735, 0.0230286, 0.01264 }, + { -0.0695069, -0.0437594, 0.0120636, 0.0163087 }, + { -0.0328856, -0.00926032, -0.00173573, 0.00527385 } + } +}; + +/* coefficients for error estimation + while using cubic Bézier curves for approximation + 1/4 <= b/a <= 1 */ +static const double coeffs3High[2][4][4] = { + { + { 0.0899116, -19.2349, -4.11711, 0.183362 }, + { 0.138148, -1.45804, 1.32044, 1.38474 }, + { 0.230903, -0.450262, 0.219963, 0.414038 }, + { 0.0590565, -0.101062, 0.0430592, 0.0204699 } + }, + { + { 0.0164649, 9.89394, 0.0919496, 0.00760802 }, + { 0.0191603, -0.0322058, 0.0134667, -0.0825018 }, + { 0.0156192, -0.017535, 0.00326508, -0.228157 }, + { -0.0236752, 0.0405821, -0.0173086, 0.176187 } + } +}; + +/* safety factor to convert the "best" error approximation + into a "max bound" error */ +static const double safety3[] = { + 0.001, 4.98, 0.207, 0.0067 +}; + +/* The code below is from the OpenVG 1.1 Spec + * Section 18.4 */ + +/* Given: Points (x0, y0) and (x1, y1) + * Return: TRUE if a solution exists, FALSE otherwise + * Circle centers are written to (cx0, cy0) and (cx1, cy1) + */ +static VGboolean +find_unit_circles(double x0, double y0, double x1, double y1, + double *cx0, double *cy0, + double *cx1, double *cy1) +{ + /* Compute differences and averages */ + double dx = x0 - x1; + double dy = y0 - y1; + double xm = (x0 + x1)/2; + double ym = (y0 + y1)/2; + double dsq, disc, s, sdx, sdy; + + /* Solve for intersecting unit circles */ + dsq = dx*dx + dy*dy; + if (dsq == 0.0) return VG_FALSE; /* Points are coincident */ + disc = 1.0/dsq - 1.0/4.0; + + /* the precision we care about here is around float so if we're + * around the float defined zero then make it official to avoid + * precision problems later on */ + if (floatIsZero(disc)) + disc = 0.0; + + if (disc < 0.0) return VG_FALSE; /* Points are too far apart */ + s = sqrt(disc); + sdx = s*dx; + sdy = s*dy; + *cx0 = xm + sdy; + *cy0 = ym - sdx; + *cx1 = xm - sdy; + *cy1 = ym + sdx; + return VG_TRUE; +} + + +/* Given: Ellipse parameters rh, rv, rot (in degrees), + * endpoints (x0, y0) and (x1, y1) + * Return: TRUE if a solution exists, FALSE otherwise + * Ellipse centers are written to (cx0, cy0) and (cx1, cy1) + */ +static VGboolean +find_ellipses(double rh, double rv, double rot, + double x0, double y0, double x1, double y1, + double *cx0, double *cy0, double *cx1, double *cy1) +{ + double COS, SIN, x0p, y0p, x1p, y1p, pcx0, pcy0, pcx1, pcy1; + /* Convert rotation angle from degrees to radians */ + rot *= M_PI/180.0; + /* Pre-compute rotation matrix entries */ + COS = cos(rot); SIN = sin(rot); + /* Transform (x0, y0) and (x1, y1) into unit space */ + /* using (inverse) rotate, followed by (inverse) scale */ + x0p = (x0*COS + y0*SIN)/rh; + y0p = (-x0*SIN + y0*COS)/rv; + x1p = (x1*COS + y1*SIN)/rh; + y1p = (-x1*SIN + y1*COS)/rv; + if (!find_unit_circles(x0p, y0p, x1p, y1p, + &pcx0, &pcy0, &pcx1, &pcy1)) { + return VG_FALSE; + } + /* Transform back to original coordinate space */ + /* using (forward) scale followed by (forward) rotate */ + pcx0 *= rh; pcy0 *= rv; + pcx1 *= rh; pcy1 *= rv; + *cx0 = pcx0*COS - pcy0*SIN; + *cy0 = pcx0*SIN + pcy0*COS; + *cx1 = pcx1*COS - pcy1*SIN; + *cy1 = pcx1*SIN + pcy1*COS; + return VG_TRUE; +} + +static INLINE VGboolean +try_to_fix_radii(struct arc *arc) +{ + double COS, SIN, rot, x0p, y0p, x1p, y1p; + double dx, dy, dsq, scale; + + /* Convert rotation angle from degrees to radians */ + rot = DEGREES_TO_RADIANS(arc->theta); + + /* Pre-compute rotation matrix entries */ + COS = cos(rot); SIN = sin(rot); + + /* Transform (x0, y0) and (x1, y1) into unit space */ + /* using (inverse) rotate, followed by (inverse) scale */ + x0p = (arc->x1*COS + arc->y1*SIN)/arc->a; + y0p = (-arc->x1*SIN + arc->y1*COS)/arc->b; + x1p = (arc->x2*COS + arc->y2*SIN)/arc->a; + y1p = (-arc->x2*SIN + arc->y2*COS)/arc->b; + /* Compute differences and averages */ + dx = x0p - x1p; + dy = y0p - y1p; + + dsq = dx*dx + dy*dy; +#if 0 + if (dsq <= 0.001) { + debug_printf("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAaaaaa\n"); + } +#endif + scale = 1/(2/sqrt(dsq)); + arc->a *= scale; + arc->b *= scale; + return VG_TRUE; +} + +static INLINE double vector_normalize(double *v) +{ + double sq = v[0] * v[0] + v[1] * v[1]; + return sqrt(sq); +} +static INLINE double vector_orientation(double *v) +{ + double norm = vector_normalize(v); + double cosa = v[0] / norm; + double sina = v[1] / norm; + return (sina>=0 ? acos(cosa) : 2*M_PI - acos(cosa)); +} +static INLINE double vector_dot(double *v0, + double *v1) +{ + return v0[0] * v1[0] + v0[1] * v1[1]; +} + +static INLINE double vector_angles(double *v0, + double *v1) +{ + double dot = vector_dot(v0, v1); + double norm0 = vector_normalize(v0); + double norm1 = vector_normalize(v1); + + return acos(dot / (norm0 * norm1)); +} + +static VGboolean find_angles(struct arc *arc) +{ + double vec0[2], vec1[2]; + double lambda1, lambda2; + double angle; + struct matrix matrix; + + if (floatIsZero(arc->a) || floatIsZero(arc->b)) { + return VG_FALSE; + } + /* map the points to an identity circle */ + matrix_load_identity(&matrix); + matrix_scale(&matrix, 1.f, arc->a/arc->b); + matrix_rotate(&matrix, -arc->theta); + matrix_map_point(&matrix, + arc->x1, arc->y1, + &arc->x1, &arc->y1); + matrix_map_point(&matrix, + arc->x2, arc->y2, + &arc->x2, &arc->y2); + matrix_map_point(&matrix, + arc->cx, arc->cy, + &arc->cx, &arc->cy); + +#if DEBUG_ARCS + debug_printf("Matrix 3 [%f, %f, %f| %f, %f, %f| %f, %f, %f]\n", + matrix.m[0], matrix.m[1], matrix.m[2], + matrix.m[3], matrix.m[4], matrix.m[5], + matrix.m[6], matrix.m[7], matrix.m[8]); + debug_printf("Endpoints [%f, %f], [%f, %f]\n", + arc->x1, arc->y1, arc->x2, arc->y2); +#endif + + vec0[0] = arc->x1 - arc->cx; + vec0[1] = arc->y1 - arc->cy; + vec1[0] = arc->x2 - arc->cx; + vec1[1] = arc->y2 - arc->cy; + +#if DEBUG_ARCS + debug_printf("Vec is [%f, %f], [%f, %f], [%f, %f]\n", + vec0[0], vec0[1], vec1[0], vec1[1], arc->cx, arc->cy); +#endif + + lambda1 = vector_orientation(vec0); + + if (isnan(lambda1)) + lambda1 = 0.f; + + if (arc->type == VG_SCWARC_TO || + arc->type == VG_SCCWARC_TO) + angle = vector_angles(vec0, vec1); + else if (arc->type == VG_LCWARC_TO || + arc->type == VG_LCCWARC_TO) { + angle = 2*M_PI - vector_angles(vec0, vec1); + } else + abort(); + + if (isnan(angle)) + angle = M_PI; + + + if (arc->type == VG_SCWARC_TO || + arc->type == VG_LCWARC_TO) + lambda2 = lambda1 - angle; + else + lambda2 = lambda1 + angle; + +#if DEBUG_ARCS + debug_printf("Angle is %f and (%f, %f)\n", angle, lambda1, lambda2); +#endif + +#if 0 + arc->eta1 = atan2(sin(lambda1) / arc->b, + cos(lambda1) / arc->a); + arc->eta2 = atan2(sin(lambda2) / arc->b, + cos(lambda2) / arc->a); + + /* make sure we have eta1 <= eta2 <= eta1 + 2 PI */ + arc->eta2 -= two_pi * floor((arc->eta2 - arc->eta1) / two_pi); + + /* the preceding correction fails if we have exactly et2 - eta1 = 2 PI + it reduces the interval to zero length */ + if ((lambda2 - lambda1 > M_PI) && (arc->eta2 - arc->eta1 < M_PI)) { + arc->eta2 += 2 * M_PI; + } +#else + arc->eta1 = lambda1; + arc->eta2 = lambda2; +#endif + + return VG_TRUE; +} + +#if DEBUG_ARCS +static void check_endpoints(struct arc *arc) +{ + double x1, y1, x2, y2; + + double a_cos_eta1 = arc->a * cos(arc->eta1); + double b_sin_eta1 = arc->b * sin(arc->eta1); + x1 = arc->cx + a_cos_eta1 * arc->cos_theta - + b_sin_eta1 * arc->sin_theta; + y1 = arc->cy + a_cos_eta1 * arc->sin_theta + + b_sin_eta1 * arc->cos_theta; + + double a_cos_eta2 = arc->a * cos(arc->eta2); + double b_sin_eta2 = arc->b * sin(arc->eta2); + x2 = arc->cx + a_cos_eta2 * arc->cos_theta - + b_sin_eta2 * arc->sin_theta; + y2 = arc->cy + a_cos_eta2 * arc->sin_theta + + b_sin_eta2 * arc->cos_theta; + + debug_printf("Computed (%f, %f), (%f, %f)\n", + x1, y1, x2, y2); + debug_printf("Real (%f, %f), (%f, %f)\n", + arc->x1, arc->y1, + arc->x2, arc->y2); +} +#endif + +void arc_init(struct arc *arc, + VGPathSegment type, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat rh, VGfloat rv, + VGfloat rot) +{ + assert(type == VG_SCCWARC_TO || + type == VG_SCWARC_TO || + type == VG_LCCWARC_TO || + type == VG_LCWARC_TO); + arc->type = type; + arc->x1 = x1; + arc->y1 = y1; + arc->x2 = x2; + arc->y2 = y2; + arc->a = rh; + arc->b = rv; + arc->theta = rot; + arc->cos_theta = cos(arc->theta); + arc->sin_theta = sin(arc->theta); + { + double cx0, cy0, cx1, cy1; + double cx, cy; + arc->is_valid = find_ellipses(rh, rv, rot, x1, y1, x2, y2, + &cx0, &cy0, &cx1, &cy1); + + if (!arc->is_valid && try_to_fix_radii(arc)) { + rh = arc->a; + rv = arc->b; + arc->is_valid = + find_ellipses(rh, rv, rot, x1, y1, x2, y2, + &cx0, &cy0, &cx1, &cy1); + } + + if (type == VG_SCWARC_TO || + type == VG_LCCWARC_TO) { + cx = cx1; + cy = cy1; + } else { + cx = cx0; + cy = cy0; + } +#if DEBUG_ARCS + debug_printf("Centers are : (%f, %f) , (%f, %f). Real (%f, %f)\n", + cx0, cy0, cx1, cy1, cx, cy); +#endif + arc->cx = cx; + arc->cy = cy; + if (arc->is_valid) { + arc->is_valid = find_angles(arc); +#if DEBUG_ARCS + check_endpoints(arc); +#endif + /* remap a few points. find_angles requires + * rot in angles, the rest of the code + * will need them in radians. and find_angles + * modifies the center to match an identity + * circle so lets reset it */ + arc->theta = DEGREES_TO_RADIANS(rot); + arc->cos_theta = cos(arc->theta); + arc->sin_theta = sin(arc->theta); + arc->cx = cx; + arc->cy = cy; + } + } +} + +static INLINE double rational_function(double x, const double *c) +{ + return (x * (x * c[0] + c[1]) + c[2]) / (x + c[3]); +} + +static double estimate_error(struct arc *arc, + double etaA, double etaB) +{ + double eta = 0.5 * (etaA + etaB); + + double x = arc->b / arc->a; + double dEta = etaB - etaA; + double cos2 = cos(2 * eta); + double cos4 = cos(4 * eta); + double cos6 = cos(6 * eta); + double c0, c1; + + /* select the right coeficients set according to degree and b/a */ + const double (*coeffs)[4][4]; + const double *safety; + coeffs = (x < 0.25) ? coeffs3Low : coeffs3High; + safety = safety3; + + c0 = rational_function(x, coeffs[0][0]) + + cos2 * rational_function(x, coeffs[0][1]) + + cos4 * rational_function(x, coeffs[0][2]) + + cos6 * rational_function(x, coeffs[0][3]); + + c1 = rational_function(x, coeffs[1][0]) + + cos2 * rational_function(x, coeffs[1][1]) + + cos4 * rational_function(x, coeffs[1][2]) + + cos6 * rational_function(x, coeffs[1][3]); + + return rational_function(x, safety) * arc->a * exp(c0 + c1 * dEta); +} + +struct arc_cb { + void (*move)(struct arc_cb *cb, VGfloat x, VGfloat y); + void (*point)(struct arc_cb *cb, VGfloat x, VGfloat y); + void (*bezier)(struct arc_cb *cb, struct bezier *bezier); + + void *user_data; +}; + +static void cb_null_move(struct arc_cb *cb, VGfloat x, VGfloat y) +{ +} + +static void polygon_point(struct arc_cb *cb, VGfloat x, VGfloat y) +{ + struct polygon *poly = (struct polygon*)cb->user_data; + polygon_vertex_append(poly, x, y); +} + +static void polygon_bezier(struct arc_cb *cb, struct bezier *bezier) +{ + struct polygon *poly = (struct polygon*)cb->user_data; + bezier_add_to_polygon(bezier, poly); +} + +static void stroke_point(struct arc_cb *cb, VGfloat x, VGfloat y) +{ + struct stroker *stroker = (struct stroker*)cb->user_data; + stroker_line_to(stroker, x, y); +} + +static void stroke_curve(struct arc_cb *cb, struct bezier *bezier) +{ + struct stroker *stroker = (struct stroker*)cb->user_data; + stroker_curve_to(stroker, + bezier->x2, bezier->y2, + bezier->x3, bezier->y3, + bezier->x4, bezier->y4); +} + +static void stroke_emit_point(struct arc_cb *cb, VGfloat x, VGfloat y) +{ + struct stroker *stroker = (struct stroker*)cb->user_data; + stroker_emit_line_to(stroker, x, y); +} + +static void stroke_emit_curve(struct arc_cb *cb, struct bezier *bezier) +{ + struct stroker *stroker = (struct stroker*)cb->user_data; + stroker_emit_curve_to(stroker, + bezier->x2, bezier->y2, + bezier->x3, bezier->y3, + bezier->x4, bezier->y4); +} + +static void arc_path_move(struct arc_cb *cb, VGfloat x, VGfloat y) +{ + struct path *path = (struct path*)cb->user_data; + path_move_to(path, x, y); +} + +static void arc_path_point(struct arc_cb *cb, VGfloat x, VGfloat y) +{ + struct path *path = (struct path*)cb->user_data; + path_line_to(path, x, y); +} + +static void arc_path_bezier(struct arc_cb *cb, struct bezier *bezier) +{ + struct path *path = (struct path*)cb->user_data; + path_cubic_to(path, + bezier->x2, bezier->y2, + bezier->x3, bezier->y3, + bezier->x4, bezier->y4); +} + +static INLINE int num_beziers_needed(struct arc *arc) +{ + double threshold = 0.05; + VGboolean found = VG_FALSE; + int n = 1; + double min_eta, max_eta; + + min_eta = MIN2(arc->eta1, arc->eta2); + max_eta = MAX2(arc->eta1, arc->eta2); + + while ((! found) && (n < 1024)) { + double d_eta = (max_eta - min_eta) / n; + if (d_eta <= 0.5 * M_PI) { + double eta_b = min_eta; + found = VG_TRUE; + for (int i = 0; found && (i < n); ++i) { + double etaA = eta_b; + eta_b += d_eta; + found = (estimate_error(arc, etaA, eta_b) <= threshold); + } + } + n = n << 1; + } + + return n; +} + +static void arc_to_beziers(struct arc *arc, + struct arc_cb cb, + struct matrix *matrix) +{ + int n = 1; + double d_eta, eta_b, cos_eta_b, + sin_eta_b, a_cos_eta_b, b_sin_eta_b, a_sin_eta_b, + b_cos_eta_b, x_b, y_b, x_b_dot, y_b_dot, lx, ly; + double t, alpha; + + { /* always move to the start of the arc */ + VGfloat x = arc->x1; + VGfloat y = arc->y1; + matrix_map_point(matrix, x, y, &x, &y); + cb.move(&cb, x, y); + } + + if (!arc->is_valid) { + VGfloat x = arc->x2; + VGfloat y = arc->y2; + matrix_map_point(matrix, x, y, &x, &y); + cb.point(&cb, x, y); + return; + } + + /* find the number of Bézier curves needed */ + n = num_beziers_needed(arc); + + d_eta = (arc->eta2 - arc->eta1) / n; + eta_b = arc->eta1; + + cos_eta_b = cos(eta_b); + sin_eta_b = sin(eta_b); + a_cos_eta_b = arc->a * cos_eta_b; + b_sin_eta_b = arc->b * sin_eta_b; + a_sin_eta_b = arc->a * sin_eta_b; + b_cos_eta_b = arc->b * cos_eta_b; + x_b = arc->cx + a_cos_eta_b * arc->cos_theta - + b_sin_eta_b * arc->sin_theta; + y_b = arc->cy + a_cos_eta_b * arc->sin_theta + + b_sin_eta_b * arc->cos_theta; + x_b_dot = -a_sin_eta_b * arc->cos_theta - + b_cos_eta_b * arc->sin_theta; + y_b_dot = -a_sin_eta_b * arc->sin_theta + + b_cos_eta_b * arc->cos_theta; + + { + VGfloat x = x_b, y = y_b; + matrix_map_point(matrix, x, y, &x, &y); + cb.point(&cb, x, y); + } + lx = x_b; + ly = y_b; + + t = tan(0.5 * d_eta); + alpha = sin(d_eta) * (sqrt(4 + 3 * t * t) - 1) / 3; + + for (int i = 0; i < n; ++i) { + struct bezier bezier; + double xA = x_b; + double yA = y_b; + double xADot = x_b_dot; + double yADot = y_b_dot; + + eta_b += d_eta; + cos_eta_b = cos(eta_b); + sin_eta_b = sin(eta_b); + a_cos_eta_b = arc->a * cos_eta_b; + b_sin_eta_b = arc->b * sin_eta_b; + a_sin_eta_b = arc->a * sin_eta_b; + b_cos_eta_b = arc->b * cos_eta_b; + x_b = arc->cx + a_cos_eta_b * arc->cos_theta - + b_sin_eta_b * arc->sin_theta; + y_b = arc->cy + a_cos_eta_b * arc->sin_theta + + b_sin_eta_b * arc->cos_theta; + x_b_dot = -a_sin_eta_b * arc->cos_theta - + b_cos_eta_b * arc->sin_theta; + y_b_dot = -a_sin_eta_b * arc->sin_theta + + b_cos_eta_b * arc->cos_theta; + + bezier_init(&bezier, + lx, ly, + (float) (xA + alpha * xADot), (float) (yA + alpha * yADot), + (float) (x_b - alpha * x_b_dot), (float) (y_b - alpha * y_b_dot), + (float) x_b, (float) y_b); +#if 0 + debug_printf("%d) Bezier (%f, %f), (%f, %f), (%f, %f), (%f, %f)\n", + i, + bezier.x1, bezier.y1, + bezier.x2, bezier.y2, + bezier.x3, bezier.y3, + bezier.x4, bezier.y4); +#endif + bezier_transform(&bezier, matrix); + cb.bezier(&cb, &bezier); + lx = x_b; + ly = y_b; + } +} + + +void arc_add_to_polygon(struct arc *arc, + struct polygon *poly, + struct matrix *matrix) +{ + struct arc_cb cb; + + cb.move = cb_null_move; + cb.point = polygon_point; + cb.bezier = polygon_bezier; + cb.user_data = poly; + + arc_to_beziers(arc, cb, matrix); +} + +void arc_stroke_cb(struct arc *arc, + struct stroker *stroke, + struct matrix *matrix) +{ + struct arc_cb cb; + + cb.move = cb_null_move; + cb.point = stroke_point; + cb.bezier = stroke_curve; + cb.user_data = stroke; + + arc_to_beziers(arc, cb, matrix); +} + +void arc_stroker_emit(struct arc *arc, + struct stroker *stroker, + struct matrix *matrix) +{ + struct arc_cb cb; + + cb.move = cb_null_move; + cb.point = stroke_emit_point; + cb.bezier = stroke_emit_curve; + cb.user_data = stroker; + + arc_to_beziers(arc, cb, matrix); +} + +void arc_to_path(struct arc *arc, + struct path *path, + struct matrix *matrix) +{ + struct arc_cb cb; + + cb.move = arc_path_move; + cb.point = arc_path_point; + cb.bezier = arc_path_bezier; + cb.user_data = path; + + arc_to_beziers(arc, cb, matrix); +} diff --git a/src/gallium/state_trackers/vega/arc.h b/src/gallium/state_trackers/vega/arc.h new file mode 100644 index 00000000000..3205cd5021d --- /dev/null +++ b/src/gallium/state_trackers/vega/arc.h @@ -0,0 +1,80 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ARC_H +#define ARC_H + +#include "VG/openvg.h" + +struct polygon; +struct matrix; +struct stroker; +struct path; + +struct arc { + VGPathSegment type; + + VGfloat cx, cy; + + VGfloat a, b; + + VGfloat theta; + VGfloat cos_theta, sin_theta; + + VGfloat eta1; + VGfloat eta2; + + VGfloat x1, y1, x2, y2; + + VGboolean is_valid; +}; + +void arc_init(struct arc *arc, + VGPathSegment type, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat rh, VGfloat rv, + VGfloat rot); + +void arc_add_to_polygon(struct arc *arc, + struct polygon *poly, + struct matrix *matrix); + + +void arc_to_path(struct arc *arc, + struct path *p, + struct matrix *matrix); + +void arc_stroke_cb(struct arc *arc, + struct stroker *stroke, + struct matrix *matrix); + +void arc_stroker_emit(struct arc *arc, + struct stroker *stroke, + struct matrix *matrix); + + +#endif diff --git a/src/gallium/state_trackers/vega/asm_fill.h b/src/gallium/state_trackers/vega/asm_fill.h new file mode 100644 index 00000000000..2f394ad6c55 --- /dev/null +++ b/src/gallium/state_trackers/vega/asm_fill.h @@ -0,0 +1,246 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ASM_FILL_H +#define ASM_FILL_H + +static const char solid_fill_asm[] = + "MOV %s, CONST[0]\n"; + + +static const char linear_grad_asm[] = + "MOV TEMP[0].xy, IN[0]\n" + "MOV TEMP[0].z, CONST[1].yyyy\n" + "DP3 TEMP[1], CONST[2], TEMP[0]\n" + "DP3 TEMP[2], CONST[3], TEMP[0]\n" + "DP3 TEMP[3], CONST[4], TEMP[0]\n" + "RCP TEMP[3], TEMP[3]\n" + "MUL TEMP[1], TEMP[1], TEMP[3]\n" + "MUL TEMP[2], TEMP[2], TEMP[3]\n" + "MOV TEMP[4].x, TEMP[1]\n" + "MOV TEMP[4].y, TEMP[2]\n" + "MUL TEMP[0], CONST[0].yyyy, TEMP[4].yyyy\n" + "MAD TEMP[1], CONST[0].xxxx, TEMP[4].xxxx, TEMP[0]\n" + "MUL TEMP[2], TEMP[1], CONST[0].zzzz\n" + "TEX %s, TEMP[2], SAMP[0], 1D\n"; + +static const char radial_grad_asm[] = + "MOV TEMP[0].xy, IN[0]\n" + "MOV TEMP[0].z, CONST[1].yyyy\n" + "DP3 TEMP[1], CONST[2], TEMP[0]\n" + "DP3 TEMP[2], CONST[3], TEMP[0]\n" + "DP3 TEMP[3], CONST[4], TEMP[0]\n" + "RCP TEMP[3], TEMP[3]\n" + "MUL TEMP[1], TEMP[1], TEMP[3]\n" + "MUL TEMP[2], TEMP[2], TEMP[3]\n" + "MOV TEMP[5].x, TEMP[1]\n" + "MOV TEMP[5].y, TEMP[2]\n" + "MUL TEMP[0], CONST[0].yyyy, TEMP[5].yyyy\n" + "MAD TEMP[1], CONST[0].xxxx, TEMP[5].xxxx, TEMP[0]\n" + "ADD TEMP[1], TEMP[1], TEMP[1]\n" + "MUL TEMP[3], TEMP[5].yyyy, TEMP[5].yyyy\n" + "MAD TEMP[4], TEMP[5].xxxx, TEMP[5].xxxx, TEMP[3]\n" + "MOV TEMP[4], -TEMP[4]\n" + "MUL TEMP[2], CONST[0].zzzz, TEMP[4]\n" + "MUL TEMP[0], CONST[1].wwww, TEMP[2]\n" + "MUL TEMP[3], TEMP[1], TEMP[1]\n" + "SUB TEMP[2], TEMP[3], TEMP[0]\n" + "RSQ TEMP[2], |TEMP[2]|\n" + "RCP TEMP[2], TEMP[2]\n" + "SUB TEMP[1], TEMP[2], TEMP[1]\n" + "ADD TEMP[0], CONST[0].zzzz, CONST[0].zzzz\n" + "RCP TEMP[0], TEMP[0]\n" + "MUL TEMP[2], TEMP[1], TEMP[0]\n" + "TEX %s, TEMP[2], SAMP[0], 1D\n"; + +static const char pattern_asm[] = + "MOV TEMP[0].xy, IN[0]\n" + "MOV TEMP[0].z, CONST[1].yyyy\n" + "DP3 TEMP[1], CONST[2], TEMP[0]\n" + "DP3 TEMP[2], CONST[3], TEMP[0]\n" + "DP3 TEMP[3], CONST[4], TEMP[0]\n" + "RCP TEMP[3], TEMP[3]\n" + "MUL TEMP[1], TEMP[1], TEMP[3]\n" + "MUL TEMP[2], TEMP[2], TEMP[3]\n" + "MOV TEMP[4].x, TEMP[1]\n" + "MOV TEMP[4].y, TEMP[2]\n" + "RCP TEMP[0], CONST[1].zwzw\n" + "MOV TEMP[1], TEMP[4]\n" + "MUL TEMP[1].x, TEMP[1], TEMP[0]\n" + "MUL TEMP[1].y, TEMP[1], TEMP[0]\n" + "TEX %s, TEMP[1], SAMP[0], 2D\n"; + + +static const char mask_asm[] = + "TEX TEMP[1], IN[0], SAMP[1], 2D\n" + "MUL TEMP[0].w, TEMP[0].wwww, TEMP[1].wwww\n" + "MOV %s, TEMP[0]\n"; + + +static const char image_normal_asm[] = + "TEX %s, IN[1], SAMP[3], 2D\n"; + +static const char image_multiply_asm[] = + "TEX TEMP[1], IN[1], SAMP[3], 2D\n" + "MUL %s, TEMP[0], TEMP[1]\n"; + +static const char image_stencil_asm[] = + "TEX TEMP[1], IN[1], SAMP[3], 2D\n" + "MUL %s, TEMP[0], TEMP[1]\n"; + + +#define EXTENDED_BLEND_OVER \ + "SUB TEMP[3], CONST[1].yyyy, TEMP[1].wwww\n" \ + "SUB TEMP[4], CONST[1].yyyy, TEMP[0].wwww\n" \ + "MUL TEMP[3], TEMP[0], TEMP[3]\n" \ + "MUL TEMP[4], TEMP[1], TEMP[4]\n" \ + "ADD TEMP[3], TEMP[3], TEMP[4]\n" + +static const char blend_multiply_asm[] = + "TEX TEMP[1], IN[0], SAMP[2], 2D\n" + EXTENDED_BLEND_OVER + "MUL TEMP[4], TEMP[0], TEMP[1]\n" + "ADD TEMP[1], TEMP[4], TEMP[3]\n"/*result.rgb*/ + "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n" + "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n" + "SUB TEMP[1].w, TEMP[3], TEMP[2]\n" + "MOV %s, TEMP[1]\n"; +#if 1 +static const char blend_screen_asm[] = + "TEX TEMP[1], IN[0], SAMP[2], 2D\n" + "ADD TEMP[3], TEMP[0], TEMP[1]\n" + "MUL TEMP[2], TEMP[0], TEMP[1]\n" + "SUB %s, TEMP[3], TEMP[2]\n"; +#else +static const char blend_screen_asm[] = + "TEX TEMP[1], IN[0], SAMP[2], 2D\n" + "MOV %s, TEMP[1]\n"; +#endif + +static const char blend_darken_asm[] = + "TEX TEMP[1], IN[0], SAMP[2], 2D\n" + EXTENDED_BLEND_OVER + "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n" + "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n" + "MIN TEMP[4], TEMP[4], TEMP[5]\n" + "ADD TEMP[1], TEMP[3], TEMP[4]\n" + "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n" + "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n" + "SUB TEMP[1].w, TEMP[3], TEMP[2]\n" + "MOV %s, TEMP[1]\n"; + +static const char blend_lighten_asm[] = + "TEX TEMP[1], IN[0], SAMP[2], 2D\n" + EXTENDED_BLEND_OVER + "MUL TEMP[4], TEMP[0], TEMP[1].wwww\n" + "MUL TEMP[5], TEMP[1], TEMP[0].wwww\n" + "MAX TEMP[4], TEMP[4], TEMP[5]\n" + "ADD TEMP[1], TEMP[3], TEMP[4]\n" + "MUL TEMP[2], TEMP[0].wwww, TEMP[1].wwww\n" + "ADD TEMP[3], TEMP[0].wwww, TEMP[1].wwww\n" + "SUB TEMP[1].w, TEMP[3], TEMP[2]\n" + "MOV %s, TEMP[1]\n"; + + +static const char premultiply_asm[] = + "MUL TEMP[0].xyz, TEMP[0], TEMP[0].wwww\n"; + +static const char unpremultiply_asm[] = + "TEX TEMP[0], IN[0], SAMP[1], 2D\n"; + + +static const char color_bw_asm[] = + "ADD TEMP[1], CONST[1].yyyy, CONST[1].yyyy\n" + "RCP TEMP[2], TEMP[1]\n" + "ADD TEMP[1], CONST[1].yyyy, TEMP[2]\n" + "ADD TEMP[2].x, TEMP[0].xxxx, TEMP[0].yyyy\n" + "ADD TEMP[2].x, TEMP[0].zzzz, TEMP[0].xxxx\n" + "SGE TEMP[0].xyz, TEMP[2].xxxx, TEMP[1]\n" + "SGE TEMP[0].w, TEMP[0].wwww, TEMP[2].yyyy\n" + "MOV %s, TEMP[0]\n"; + + +struct shader_asm_info { + VGint id; + VGint num_tokens; + const char * txt; + + VGboolean needs_position; + + VGint start_const; + VGint num_consts; + + VGint start_sampler; + VGint num_samplers; + + VGint start_temp; + VGint num_temps; +}; + + +static const struct shader_asm_info shaders_asm[] = { + /* fills */ + {VEGA_SOLID_FILL_SHADER, 40, solid_fill_asm, + VG_FALSE, 0, 1, 0, 0, 0, 0}, + {VEGA_LINEAR_GRADIENT_SHADER, 200, linear_grad_asm, + VG_TRUE, 0, 5, 0, 1, 0, 5}, + {VEGA_RADIAL_GRADIENT_SHADER, 200, radial_grad_asm, + VG_TRUE, 0, 5, 0, 1, 0, 6}, + {VEGA_PATTERN_SHADER, 100, pattern_asm, + VG_TRUE, 1, 4, 0, 1, 0, 5}, + + /* image draw modes */ + {VEGA_IMAGE_NORMAL_SHADER, 200, image_normal_asm, + VG_TRUE, 0, 0, 3, 1, 0, 0}, + {VEGA_IMAGE_MULTIPLY_SHADER, 200, image_multiply_asm, + VG_TRUE, 0, 0, 3, 1, 0, 2}, + {VEGA_IMAGE_STENCIL_SHADER, 200, image_stencil_asm, + VG_TRUE, 0, 0, 3, 1, 0, 2}, + + {VEGA_MASK_SHADER, 100, mask_asm, + VG_TRUE, 0, 0, 1, 1, 0, 2}, + + /* extra blend modes */ + {VEGA_BLEND_MULTIPLY_SHADER, 200, blend_multiply_asm, + VG_TRUE, 1, 1, 2, 1, 0, 5}, + {VEGA_BLEND_SCREEN_SHADER, 200, blend_screen_asm, + VG_TRUE, 0, 0, 2, 1, 0, 4}, + {VEGA_BLEND_DARKEN_SHADER, 200, blend_darken_asm, + VG_TRUE, 1, 1, 2, 1, 0, 6}, + {VEGA_BLEND_LIGHTEN_SHADER, 200, blend_lighten_asm, + VG_TRUE, 1, 1, 2, 1, 0, 6}, + + /* premultiply */ + {VEGA_PREMULTIPLY_SHADER, 100, premultiply_asm, + VG_FALSE, 0, 0, 0, 0, 0, 1}, + {VEGA_UNPREMULTIPLY_SHADER, 100, unpremultiply_asm, + VG_FALSE, 0, 0, 0, 0, 0, 1}, + + /* color transform to black and white */ + {VEGA_BW_SHADER, 150, color_bw_asm, + VG_FALSE, 1, 1, 0, 0, 0, 3}, +}; +#endif diff --git a/src/gallium/state_trackers/vega/asm_filters.h b/src/gallium/state_trackers/vega/asm_filters.h new file mode 100644 index 00000000000..9a49f2e12d0 --- /dev/null +++ b/src/gallium/state_trackers/vega/asm_filters.h @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ASM_FILTERS_H +#define ASM_FILTERS_H + +static const char color_matrix_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "DCL CONST[0..4], CONSTANT\n" + "DCL TEMP[0..4], CONSTANT\n" + "DCL SAMP[0], CONSTANT\n" + "TEX TEMP[0], IN[0], SAMP[0], 2D\n" + "MOV TEMP[1], TEMP[0].xxxx\n" + "MOV TEMP[2], TEMP[0].yyyy\n" + "MOV TEMP[3], TEMP[0].zzzz\n" + "MOV TEMP[4], TEMP[0].wwww\n" + "MUL TEMP[1], TEMP[1], CONST[0]\n" + "MUL TEMP[2], TEMP[2], CONST[1]\n" + "MUL TEMP[3], TEMP[3], CONST[2]\n" + "MUL TEMP[4], TEMP[4], CONST[3]\n" + "ADD TEMP[0], TEMP[1], CONST[4]\n" + "ADD TEMP[0], TEMP[0], TEMP[2]\n" + "ADD TEMP[0], TEMP[0], TEMP[3]\n" + "ADD TEMP[0], TEMP[0], TEMP[4]\n" + "MOV OUT[0], TEMP[0]\n" + "END\n"; + +static const char convolution_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "DCL TEMP[0..4], CONSTANT\n" + "DCL ADDR[0], CONSTANT\n" + "DCL CONST[0..%d], CONSTANT\n" + "DCL SAMP[0], CONSTANT\n" + "0: MOV TEMP[0], CONST[0].xxxx\n" + "1: MOV TEMP[1], CONST[0].xxxx\n" + "2: BGNLOOP :14\n" + "3: SGE TEMP[0].z, TEMP[0].yyyy, CONST[1].xxxx\n" + "4: IF TEMP[0].zzzz :7\n" + "5: BRK\n" + "6: ENDIF\n" + "7: ARL ADDR[0].x, TEMP[0].yyyy\n" + "8: MOV TEMP[3], CONST[ADDR[0]+2]\n" + "9: ADD TEMP[4].xy, IN[0], TEMP[3]\n" + "10: TEX TEMP[2], TEMP[4], SAMP[0], 2D\n" + "11: MOV TEMP[3], CONST[ADDR[0]+%d]\n" + "12: MAD TEMP[1], TEMP[2], TEMP[3], TEMP[1]\n" + "13: ADD TEMP[0].y, TEMP[0].yyyy, CONST[0].yyyy\n" + "14: ENDLOOP :2\n" + "15: MAD OUT[0], TEMP[1], CONST[1].yyyy, CONST[1].zzzz\n" + "16: END\n"; + + +static const char lookup_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "DCL TEMP[0..2], CONSTANT\n" + "DCL CONST[0], CONSTANT\n" + "DCL SAMP[0..1], CONSTANT\n" + "TEX TEMP[0], IN[0], SAMP[0], 2D\n" + "MOV TEMP[1], TEMP[0]\n" + /* do red */ + "TEX TEMP[2], TEMP[1].xxxx, SAMP[1], 1D\n" + "MOV TEMP[0].x, TEMP[2].xxxx\n" + /* do blue */ + "TEX TEMP[2], TEMP[1].yyyy, SAMP[1], 1D\n" + "MOV TEMP[0].y, TEMP[2].yyyy\n" + /* do green */ + "TEX TEMP[2], TEMP[1].zzzz, SAMP[1], 1D\n" + "MOV TEMP[0].z, TEMP[2].zzzz\n" + /* do alpha */ + "TEX TEMP[2], TEMP[1].wwww, SAMP[1], 1D\n" + "MOV TEMP[0].w, TEMP[2].wwww\n" + "MOV OUT[0], TEMP[0]\n" + "END\n"; + + +static const char lookup_single_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "DCL TEMP[0..2], CONSTANT\n" + "DCL CONST[0], CONSTANT\n" + "DCL SAMP[0..1], CONSTANT\n" + "TEX TEMP[0], IN[0], SAMP[0], 2D\n" + "TEX TEMP[1], TEMP[0].%s, SAMP[1], 1D\n" + "MOV OUT[0], TEMP[1]\n" + "END\n"; + +#endif diff --git a/src/gallium/state_trackers/vega/asm_util.h b/src/gallium/state_trackers/vega/asm_util.h new file mode 100644 index 00000000000..218e1d166db --- /dev/null +++ b/src/gallium/state_trackers/vega/asm_util.h @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ASM_UTIL_H +#define ASM_UTIL_H + + +static const char pass_through_depth_asm[] = + "FRAG1.1\n" + "DCL IN[0], POSITION, LINEAR\n" + "DCL OUT[0].z, POSITION, CONSTANT\n" + "0: MOV OUT[0].z, IN[0].zzzz\n" + "1: END\n"; + + + +/* μnew = μmask */ +static const char set_mask_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL SAMP[0], CONSTANT\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "0: TEX OUT[0], IN[0], SAMP[0], 2D\n"/*umask*/ + "1: END\n"; + +/* μnew = 1 – (1 – μmask)*(1 – μprev) */ +static const char union_mask_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL IN[1], POSITION, LINEAR\n" + "DCL CONST[0], CONSTANT\n" + "DCL SAMP[0..1], CONSTANT\n" + "DCL TEMP[0..3], CONSTANT\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/ + "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/ + "2: SUB TEMP[2], CONST[0], TEMP[0]\n" + "3: SUB TEMP[3], CONST[0], TEMP[1]\n" + "4: MUL TEMP[0].w, TEMP[2].wwww, TEMP[3].wwww\n" + "5: SUB OUT[0], CONST[0], TEMP[0]\n" + "6: END\n"; + +/* μnew = μmask *μprev */ +static const char intersect_mask_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL IN[1], POSITION, LINEAR\n" + "DCL CONST[0], CONSTANT\n" + "DCL SAMP[0..1], CONSTANT\n" + "DCL TEMP[0..1], CONSTANT\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "0: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/ + "1: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/ + "2: MUL OUT[0], TEMP[0].wwww, TEMP[1].wwww\n" + "3: END\n"; + +/* μnew = μprev*(1 – μmask) */ +static const char subtract_mask_asm[] = + "FRAG1.1\n" + "DCL IN[0], GENERIC[0], PERSPECTIVE\n" + "DCL IN[1], POSITION, LINEAR\n" + "DCL CONST[0], CONSTANT\n" + "DCL SAMP[0..1], CONSTANT\n" + "DCL TEMP[0..2], CONSTANT\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "0: TEX TEMP[1], IN[0], SAMP[0], 2D\n"/*umask*/ + "1: TEX TEMP[0], IN[1], SAMP[1], 2D\n"/*uprev*/ + "2: SUB TEMP[2], CONST[0], TEMP[1]\n" + "3: MUL OUT[0], TEMP[2].wwww, TEMP[0].wwww\n" + "4: END\n"; + + +static const char vs_plain_asm[] = + "VERT1.1\n" + "DCL IN[0]\n" + "DCL OUT[0], POSITION\n" + "DCL TEMP[0]\n" + "DCL CONST[0..1]\n" + "0: MUL TEMP[0], IN[0], CONST[0]\n" + "1: ADD TEMP[0], TEMP[0], CONST[1]\n" + "2: MOV OUT[0], TEMP[0]\n" + "3: END\n"; + +static const char vs_clear_asm[] = + "VERT1.1\n" + "DCL IN[0]\n" + "DCL IN[1]\n" + "DCL OUT[0], POSITION\n" + "DCL OUT[1], COLOR\n" + "DCL TEMP[0]\n" + "DCL CONST[0..1]\n" + "0: MUL TEMP[0], IN[0], CONST[0]\n" + "1: ADD TEMP[0], TEMP[0], CONST[1]\n" + "2: MOV OUT[0], TEMP[0]\n" + "3: MOV OUT[1], IN[1]\n" + "4: END\n"; + + +static const char vs_texture_asm[] = + "VERT1.1\n" + "DCL IN[0]\n" + "DCL IN[1]\n" + "DCL OUT[0], POSITION\n" + "DCL OUT[1], GENERIC\n" + "DCL TEMP[0]\n" + "DCL CONST[0..1]\n" + "0: MUL TEMP[0], IN[0], CONST[0]\n" + "1: ADD TEMP[0], TEMP[0], CONST[1]\n" + "2: MOV OUT[0], TEMP[0]\n" + "3: MOV OUT[1], IN[1]\n" + "4: END\n"; + +#endif diff --git a/src/gallium/state_trackers/vega/bezier.c b/src/gallium/state_trackers/vega/bezier.c new file mode 100644 index 00000000000..39a7ade0161 --- /dev/null +++ b/src/gallium/state_trackers/vega/bezier.c @@ -0,0 +1,704 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "bezier.h" + +#include "matrix.h" +#include "polygon.h" + +#include "pipe/p_compiler.h" +#include "util/u_debug.h" + +#include <stdlib.h> +#include <stdio.h> +#include <assert.h> +#include <math.h> + +static const float flatness = 0.5; + + +static INLINE void split_left(struct bezier *bez, VGfloat t, struct bezier* left) +{ + left->x1 = bez->x1; + left->y1 = bez->y1; + + left->x2 = bez->x1 + t * (bez->x2 - bez->x1); + left->y2 = bez->y1 + t * (bez->y2 - bez->y1); + + left->x3 = bez->x2 + t * (bez->x3 - bez->x2); + left->y3 = bez->y2 + t * (bez->y3 - bez->y2); + + bez->x3 = bez->x3 + t * (bez->x4 - bez->x3); + bez->y3 = bez->y3 + t * (bez->y4 - bez->y3); + + bez->x2 = left->x3 + t * (bez->x3 - left->x3); + bez->y2 = left->y3 + t * (bez->y3 - left->y3); + + left->x3 = left->x2 + t * (left->x3 - left->x2); + left->y3 = left->y2 + t * (left->y3 - left->y2); + + left->x4 = bez->x1 = left->x3 + t * (bez->x2 - left->x3); + left->y4 = bez->y1 = left->y3 + t * (bez->y2 - left->y3); +} + +static INLINE void split(struct bezier *bez, + struct bezier *first_half, + struct bezier *second_half) +{ + double c = (bez->x2 + bez->x3) * 0.5; + first_half->x2 = (bez->x1 + bez->x2) * 0.5; + second_half->x3 = (bez->x3 + bez->x4) * 0.5; + first_half->x1 = bez->x1; + second_half->x4 = bez->x4; + first_half->x3 = (first_half->x2 + c) * 0.5; + second_half->x2 = (second_half->x3 + c) * 0.5; + first_half->x4 = second_half->x1 = + (first_half->x3 + second_half->x2) * 0.5; + + c = (bez->y2 + bez->y3) / 2; + first_half->y2 = (bez->y1 + bez->y2) * 0.5; + second_half->y3 = (bez->y3 + bez->y4) * 0.5; + first_half->y1 = bez->y1; + second_half->y4 = bez->y4; + first_half->y3 = (first_half->y2 + c) * 0.5; + second_half->y2 = (second_half->y3 + c) * 0.5; + first_half->y4 = second_half->y1 = + (first_half->y3 + second_half->y2) * 0.5; +} + +struct polygon * bezier_to_polygon(struct bezier *bez) +{ + struct polygon *poly = polygon_create(64); + polygon_vertex_append(poly, bez->x1, bez->y1); + bezier_add_to_polygon(bez, poly); + return poly; +} + +void bezier_add_to_polygon(const struct bezier *bez, + struct polygon *poly) +{ + struct bezier beziers[32]; + struct bezier *b; + + beziers[0] = *bez; + b = beziers; + + while (b >= beziers) { + double y4y1 = b->y4 - b->y1; + double x4x1 = b->x4 - b->x1; + double l = ABS(x4x1) + ABS(y4y1); + double d; + if (l > 1.f) { + d = ABS((x4x1)*(b->y1 - b->y2) - (y4y1)*(b->x1 - b->x2)) + + ABS((x4x1)*(b->y1 - b->y3) - (y4y1)*(b->x1 - b->x3)); + } else { + d = ABS(b->x1 - b->x2) + ABS(b->y1 - b->y2) + + ABS(b->x1 - b->x3) + ABS(b->y1 - b->y3); + l = 1.; + } + if (d < flatness*l || b == beziers + 31) { + /* good enough, we pop it off and add the endpoint */ + polygon_vertex_append(poly, b->x4, b->y4); + --b; + } else { + /* split, second half of the bezier goes lower into the stack */ + split(b, b+1, b); + ++b; + } + } +} + +static void add_if_close(struct bezier *bez, VGfloat *length, VGfloat error) +{ + struct bezier left, right; /* bez poly splits */ + VGfloat len = 0.0; /* arc length */ + VGfloat chord; /* chord length */ + + len = len + line_length(bez->x1, bez->y1, bez->x2, bez->y2); + len = len + line_length(bez->x2, bez->y2, bez->x3, bez->y3); + len = len + line_length(bez->x3, bez->y3, bez->x4, bez->y4); + + chord = line_length(bez->x1, bez->y1, bez->x4, bez->y4); + + if ((len-chord) > error) { + split(bez, &left, &right); /* split in two */ + add_if_close(&left, length, error); /* try left side */ + add_if_close(&right, length, error); /* try right side */ + return; + } + + *length = *length + len; + + return; +} + +float bezier_length(struct bezier *bez, float error) +{ + VGfloat length = 0.f; + + add_if_close(bez, &length, error); + return length; +} + +void bezier_init(struct bezier *bez, + float x1, float y1, + float x2, float y2, + float x3, float y3, + float x4, float y4) +{ + bez->x1 = x1; + bez->y1 = y1; + bez->x2 = x2; + bez->y2 = y2; + bez->x3 = x3; + bez->y3 = y3; + bez->x4 = x4; + bez->y4 = y4; +#if 0 + debug_printf("bezier in [%f, %f, %f, %f, %f, %f]\n", + x1, y1, x2, y2, x3, y3, x4, y4); +#endif +} + + +static INLINE void bezier_init2v(struct bezier *bez, + float *pt1, + float *pt2, + float *pt3, + float *pt4) +{ + bez->x1 = pt1[0]; + bez->y1 = pt1[1]; + + bez->x2 = pt2[0]; + bez->y2 = pt2[1]; + + bez->x3 = pt3[0]; + bez->y3 = pt3[1]; + + bez->x4 = pt4[0]; + bez->y4 = pt4[1]; +} + + +void bezier_transform(struct bezier *bez, + struct matrix *matrix) +{ + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, bez->x1, bez->y1, &bez->x1, &bez->y1); + matrix_map_point(matrix, bez->x2, bez->y2, &bez->x2, &bez->y2); + matrix_map_point(matrix, bez->x3, bez->y3, &bez->x3, &bez->y3); + matrix_map_point(matrix, bez->x4, bez->y4, &bez->x4, &bez->y4); +} + +static INLINE void bezier_point_at(const struct bezier *bez, float t, float *pt) +{ + float a, b, c, d; + float m_t; + m_t = 1. - t; + b = m_t * m_t; + c = t * t; + d = c * t; + a = b * m_t; + b *= 3. * t; + c *= 3. * m_t; + pt[0] = a*bez->x1 + b*bez->x2 + c*bez->x3 + d*bez->x4; + pt[1] = a*bez->y1 + b*bez->y2 + c*bez->y3 + d*bez->y4; +} + +static INLINE void bezier_normal_at(const struct bezier *bez, float t, float *norm) +{ + float m_t = 1. - t; + float a = m_t * m_t; + float b = t * m_t; + float c = t * t; + + norm[0] = (bez->y2-bez->y1) * a + (bez->y3-bez->y2) * b + (bez->y4-bez->y3) * c; + norm[1] = -(bez->x2-bez->x1) * a - (bez->x3-bez->x2) * b - (bez->x4-bez->x3) * c; +} + +enum shift_result { + Ok, + Discard, + Split, + Circle +}; + +static enum shift_result good_offset(const struct bezier *b1, + const struct bezier *b2, + float offset, float threshold) +{ + const float o2 = offset*offset; + const float max_dist_line = threshold*offset*offset; + const float max_dist_normal = threshold*offset; + const float spacing = 0.25; + for (float i = spacing; i < 0.99; i += spacing) { + float p1[2],p2[2], d, l; + float normal[2]; + bezier_point_at(b1, i, p1); + bezier_point_at(b2, i, p2); + d = (p1[0] - p2[0])*(p1[0] - p2[0]) + (p1[1] - p2[1])*(p1[1] - p2[1]); + if (ABS(d - o2) > max_dist_line) + return Split; + + bezier_normal_at(b1, i, normal); + l = ABS(normal[0]) + ABS(normal[1]); + if (l != 0.) { + d = ABS(normal[0]*(p1[1] - p2[1]) - normal[1]*(p1[0] - p2[0]) ) / l; + if (d > max_dist_normal) + return Split; + } + } + return Ok; +} + +static INLINE void shift_line_by_normal(float *l, float offset) +{ + float norm[4]; + float tx, ty; + + line_normal(l, norm); + line_normalize(norm); + + tx = (norm[2] - norm[0]) * offset; + ty = (norm[3] - norm[1]) * offset; + l[0] += tx; l[1] += ty; + l[2] += tx; l[3] += ty; +} + +static INLINE VGboolean is_bezier_line(float (*points)[2], int count) +{ + float dx13 = points[2][0] - points[0][0]; + float dy13 = points[2][1] - points[0][1]; + + float dx12 = points[1][0] - points[0][0]; + float dy12 = points[1][1] - points[0][1]; + + debug_assert(count > 2); + + if (count == 3) { + return floatsEqual(dx12 * dy13, dx13 * dy12); + } else if (count == 4) { + float dx14 = points[3][0] - points[0][0]; + float dy14 = points[3][1] - points[0][1]; + + return (floatsEqual(dx12 * dy13, dx13 * dy12) && + floatsEqual(dx12 * dy14, dx14 * dy12)); + } + + return VG_FALSE; +} + +static INLINE void compute_pt_normal(float *pt1, float *pt2, float *res) +{ + float line[4]; + float normal[4]; + line[0] = 0.f; line[1] = 0.f; + line[2] = pt2[0] - pt1[0]; + line[3] = pt2[1] - pt1[1]; + line_normal(line, normal); + line_normalize(normal); + + res[0] = normal[2]; + res[1] = normal[3]; +} + +static enum shift_result shift(const struct bezier *orig, + struct bezier *shifted, + float offset, float threshold) +{ + int map[4]; + VGboolean p1_p2_equal = (orig->x1 == orig->x2 && orig->y1 == orig->y2); + VGboolean p2_p3_equal = (orig->x2 == orig->x3 && orig->y2 == orig->y3); + VGboolean p3_p4_equal = (orig->x3 == orig->x4 && orig->y3 == orig->y4); + + float points[4][2]; + int np = 0; + float bounds[4]; + float points_shifted[4][2]; + float prev_normal[2]; + + points[np][0] = orig->x1; + points[np][1] = orig->y1; + map[0] = 0; + ++np; + if (!p1_p2_equal) { + points[np][0] = orig->x2; + points[np][1] = orig->y2; + ++np; + } + map[1] = np - 1; + if (!p2_p3_equal) { + points[np][0] = orig->x3; + points[np][1] = orig->y3; + ++np; + } + map[2] = np - 1; + if (!p3_p4_equal) { + points[np][0] = orig->x4; + points[np][1] = orig->y4; + ++np; + } + map[3] = np - 1; + if (np == 1) + return Discard; + + /* We need to specialcase lines of 3 or 4 points due to numerical + instability in intersection code below */ + if (np > 2 && is_bezier_line(points, np)) { + float l[4] = { points[0][0], points[0][1], + points[np-1][0], points[np-1][1] }; + float ctrl1[2], ctrl2[2]; + if (floatsEqual(points[0][0], points[np-1][0]) && + floatsEqual(points[0][1], points[np-1][1])) + return Discard; + + shift_line_by_normal(l, offset); + line_point_at(l, 0.33, ctrl1); + line_point_at(l, 0.66, ctrl2); + bezier_init(shifted, l[0], l[1], + ctrl1[0], ctrl1[1], ctrl2[0], ctrl2[1], + l[2], l[3]); + return Ok; + } + + bezier_bounds(orig, bounds); + if (np == 4 && bounds[2] < .1*offset && bounds[3] < .1*offset) { + float l = (orig->x1 - orig->x2)*(orig->x1 - orig->x2) + + (orig->y1 - orig->y2)*(orig->y1 - orig->y1) * + (orig->x3 - orig->x4)*(orig->x3 - orig->x4) + + (orig->y3 - orig->y4)*(orig->y3 - orig->y4); + float dot = (orig->x1 - orig->x2)*(orig->x3 - orig->x4) + + (orig->y1 - orig->y2)*(orig->y3 - orig->y4); + if (dot < 0 && dot*dot < 0.8*l) + /* the points are close and reverse dirction. Approximate the whole + thing by a semi circle */ + return Circle; + } + + compute_pt_normal(points[0], points[1], prev_normal); + + points_shifted[0][0] = points[0][0] + offset * prev_normal[0]; + points_shifted[0][1] = points[0][1] + offset * prev_normal[1]; + + for (int i = 1; i < np - 1; ++i) { + float normal_sum[2], r; + float next_normal[2]; + compute_pt_normal(points[i], points[i + 1], next_normal); + + normal_sum[0] = prev_normal[0] + next_normal[0]; + normal_sum[1] = prev_normal[1] + next_normal[1]; + + r = 1.0 + prev_normal[0] * next_normal[0] + + prev_normal[1] * next_normal[1]; + + if (floatsEqual(r + 1, 1)) { + points_shifted[i][0] = points[i][0] + offset * prev_normal[0]; + points_shifted[i][1] = points[i][1] + offset * prev_normal[1]; + } else { + float k = offset / r; + points_shifted[i][0] = points[i][0] + k * normal_sum[0]; + points_shifted[i][1] = points[i][1] + k * normal_sum[1]; + } + + prev_normal[0] = next_normal[0]; + prev_normal[1] = next_normal[1]; + } + + points_shifted[np - 1][0] = points[np - 1][0] + offset * prev_normal[0]; + points_shifted[np - 1][1] = points[np - 1][1] + offset * prev_normal[1]; + + bezier_init2v(shifted, + points_shifted[map[0]], points_shifted[map[1]], + points_shifted[map[2]], points_shifted[map[3]]); + + return good_offset(orig, shifted, offset, threshold); +} + +static VGboolean make_circle(const struct bezier *b, float offset, struct bezier *o) +{ + float normals[3][2]; + float dist; + float angles[2]; + float sign = 1.f; + int i; + float circle[3][2]; + + normals[0][0] = b->y2 - b->y1; + normals[0][1] = b->x1 - b->x2; + dist = sqrt(normals[0][0]*normals[0][0] + normals[0][1]*normals[0][1]); + if (floatsEqual(dist + 1, 1.f)) + return VG_FALSE; + normals[0][0] /= dist; + normals[0][1] /= dist; + + normals[2][0] = b->y4 - b->y3; + normals[2][1] = b->x3 - b->x4; + dist = sqrt(normals[2][0]*normals[2][0] + normals[2][1]*normals[2][1]); + if (floatsEqual(dist + 1, 1.f)) + return VG_FALSE; + normals[2][0] /= dist; + normals[2][1] /= dist; + + normals[1][0] = b->x1 - b->x2 - b->x3 + b->x4; + normals[1][1] = b->y1 - b->y2 - b->y3 + b->y4; + dist = -1*sqrt(normals[1][0]*normals[1][0] + normals[1][1]*normals[1][1]); + normals[1][0] /= dist; + normals[1][1] /= dist; + + for (i = 0; i < 2; ++i) { + float cos_a = normals[i][0]*normals[i+1][0] + normals[i][1]*normals[i+1][1]; + if (cos_a > 1.) + cos_a = 1.; + if (cos_a < -1.) + cos_a = -1; + angles[i] = acos(cos_a)/M_PI; + } + + if (angles[0] + angles[1] > 1.) { + /* more than 180 degrees */ + normals[1][0] = -normals[1][0]; + normals[1][1] = -normals[1][1]; + angles[0] = 1. - angles[0]; + angles[1] = 1. - angles[1]; + sign = -1.; + } + + circle[0][0] = b->x1 + normals[0][0]*offset; + circle[0][1] = b->y1 + normals[0][1]*offset; + + circle[1][0] = 0.5*(b->x1 + b->x4) + normals[1][0]*offset; + circle[1][1] = 0.5*(b->y1 + b->y4) + normals[1][1]*offset; + + circle[2][0] = b->x4 + normals[2][0]*offset; + circle[2][1] = b->y4 + normals[2][1]*offset; + + for (i = 0; i < 2; ++i) { + float kappa = 2.*KAPPA * sign * offset * angles[i]; + + o->x1 = circle[i][0]; + o->y1 = circle[i][1]; + o->x2 = circle[i][0] - normals[i][1]*kappa; + o->y2 = circle[i][1] + normals[i][0]*kappa; + o->x3 = circle[i+1][0] + normals[i+1][1]*kappa; + o->y3 = circle[i+1][1] - normals[i+1][0]*kappa; + o->x4 = circle[i+1][0]; + o->y4 = circle[i+1][1]; + + ++o; + } + return VG_TRUE; +} + +int bezier_translate_by_normal(struct bezier *bez, + struct bezier *curves, + int max_curves, + float normal_len, + float threshold) +{ + struct bezier beziers[10]; + struct bezier *b, *o; + + /* fixme: this should really be floatsEqual */ + if (bez->x1 == bez->x2 && bez->x1 == bez->x3 && bez->x1 == bez->x4 && + bez->y1 == bez->y2 && bez->y1 == bez->y3 && bez->y1 == bez->y4) + return 0; + + --max_curves; +redo: + beziers[0] = *bez; + b = beziers; + o = curves; + + while (b >= beziers) { + int stack_segments = b - beziers + 1; + enum shift_result res; + if ((stack_segments == 10) || (o - curves == max_curves - stack_segments)) { + threshold *= 1.5; + if (threshold > 2.) + goto give_up; + goto redo; + } + res = shift(b, o, normal_len, threshold); + if (res == Discard) { + --b; + } else if (res == Ok) { + ++o; + --b; + continue; + } else if (res == Circle && max_curves - (o - curves) >= 2) { + /* add semi circle */ + if (make_circle(b, normal_len, o)) + o += 2; + --b; + } else { + split(b, b+1, b); + ++b; + } + } + +give_up: + while (b >= beziers) { + enum shift_result res = shift(b, o, normal_len, threshold); + + /* if res isn't Ok or Split then *o is undefined */ + if (res == Ok || res == Split) + ++o; + + --b; + } + + debug_assert(o - curves <= max_curves); + return o - curves; +} + +void bezier_bounds(const struct bezier *bez, + float *bounds/*x/y/width/height*/) +{ + float xmin = bez->x1; + float xmax = bez->x1; + float ymin = bez->y1; + float ymax = bez->y1; + + if (bez->x2 < xmin) + xmin = bez->x2; + else if (bez->x2 > xmax) + xmax = bez->x2; + if (bez->x3 < xmin) + xmin = bez->x3; + else if (bez->x3 > xmax) + xmax = bez->x3; + if (bez->x4 < xmin) + xmin = bez->x4; + else if (bez->x4 > xmax) + xmax = bez->x4; + + if (bez->y2 < ymin) + ymin = bez->y2; + else if (bez->y2 > ymax) + ymax = bez->y2; + if (bez->y3 < ymin) + ymin = bez->y3; + else if (bez->y3 > ymax) + ymax = bez->y3; + if (bez->y4 < ymin) + ymin = bez->y4; + else if (bez->y4 > ymax) + ymax = bez->y4; + + bounds[0] = xmin; /* x */ + bounds[1] = ymin; /* y */ + bounds[2] = xmax - xmin; /* width */ + bounds[3] = ymax - ymin; /* height */ +} + +void bezier_start_tangent(const struct bezier *bez, + float *tangent) +{ + tangent[0] = bez->x1; + tangent[1] = bez->y1; + tangent[2] = bez->x2; + tangent[3] = bez->y2; + + if (null_line(tangent)) { + tangent[0] = bez->x1; + tangent[1] = bez->y1; + tangent[2] = bez->x3; + tangent[3] = bez->y3; + } + if (null_line(tangent)) { + tangent[0] = bez->x1; + tangent[1] = bez->y1; + tangent[2] = bez->x4; + tangent[3] = bez->y4; + } +} + + +static INLINE VGfloat bezier_t_at_length(struct bezier *bez, + VGfloat at_length, + VGfloat error) +{ + VGfloat len = bezier_length(bez, error); + VGfloat t = 1.0; + VGfloat last_bigger = 1.; + + if (at_length > len || floatsEqual(at_length, len)) + return t; + + if (floatIsZero(at_length)) + return 0.f; + + t *= 0.5; + while (1) { + struct bezier right = *bez; + struct bezier left; + VGfloat tmp_len; + split_left(&right, t, &left); + tmp_len = bezier_length(&left, error); + if (ABS(tmp_len - at_length) < error) + break; + + if (tmp_len < at_length) { + t += (last_bigger - t)*.5; + } else { + last_bigger = t; + t -= t*.5; + } + } + return t; +} + +void bezier_point_at_length(struct bezier *bez, + float length, + float *point, + float *normal) +{ + /* ~0.000001 seems to be required to pass G2080x tests */ + VGfloat t = bezier_t_at_length(bez, length, 0.000001); + bezier_point_at(bez, t, point); + bezier_normal_at(bez, t, normal); + vector_unit(normal); +} + +void bezier_point_at_t(struct bezier *bez, float t, + float *point, float *normal) +{ + bezier_point_at(bez, t, point); + bezier_normal_at(bez, t, normal); + vector_unit(normal); +} + +void bezier_exact_bounds(const struct bezier *bez, + float *bounds/*x/y/width/height*/) +{ + struct polygon *poly = polygon_create(64); + polygon_vertex_append(poly, bez->x1, bez->y1); + bezier_add_to_polygon(bez, poly); + polygon_bounding_rect(poly, bounds); + polygon_destroy(poly); +} + diff --git a/src/gallium/state_trackers/vega/bezier.h b/src/gallium/state_trackers/vega/bezier.h new file mode 100644 index 00000000000..e06666051ca --- /dev/null +++ b/src/gallium/state_trackers/vega/bezier.h @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BEZIER_H +#define BEZIER_H + +struct polygon; +struct matrix; + +struct bezier { + float x1, y1; + float x2, y2; + float x3, y3; + float x4, y4; +}; + + +#define BEZIER_DEFAULT_ERROR 0.01 + +/* kappa as being l of a circle with r = 1, we can emulate any + * circle of radius r by using the formula + * l = r . kappa + * More at: + * http://www.whizkidtech.redprince.net/bezier/circle/ */ +#define KAPPA 0.5522847498 + +void bezier_init(struct bezier *bez, + float x1, float y1, + float x2, float y2, + float x3, float y3, + float x4, float y4); + +struct polygon *bezier_to_polygon(struct bezier *bez); +void bezier_add_to_polygon(const struct bezier *bez, + struct polygon *poly); +float bezier_length(struct bezier *bez, float error); +void bezier_transform(struct bezier *bez, + struct matrix *mat); + +int bezier_translate_by_normal(struct bezier *b, + struct bezier *curves, + int max_curves, + float normal_len, + float threshold); +void bezier_bounds(const struct bezier *bez, + float *bounds/*x/y/width/height*/); +void bezier_exact_bounds(const struct bezier *bez, + float *bounds/*x/y/width/height*/); + +void bezier_start_tangent(const struct bezier *bez, + float *tangent); + +void bezier_point_at_length(struct bezier *bez, float length, + float *point, float *normal); +void bezier_point_at_t(struct bezier *bez, float t, + float *point, float *normal); + +#endif diff --git a/src/gallium/state_trackers/vega/image.c b/src/gallium/state_trackers/vega/image.c new file mode 100644 index 00000000000..9a722980d52 --- /dev/null +++ b/src/gallium/state_trackers/vega/image.c @@ -0,0 +1,654 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "image.h" + +#include "vg_translate.h" +#include "vg_context.h" +#include "matrix.h" +#include "renderer.h" +#include "util_array.h" +#include "api_consts.h" +#include "shaders_cache.h" +#include "shader.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "util/u_blit.h" +#include "util/u_tile.h" +#include "util/u_memory.h" +#include "util/u_math.h" + +static enum pipe_format vg_format_to_pipe(VGImageFormat format) +{ + switch(format) { + case VG_sRGB_565: + return PIPE_FORMAT_R5G6B5_UNORM; + case VG_sRGBA_5551: + return PIPE_FORMAT_A1R5G5B5_UNORM; + case VG_sRGBA_4444: + return PIPE_FORMAT_A4R4G4B4_UNORM; + case VG_sL_8: + case VG_lL_8: + return PIPE_FORMAT_L8_UNORM; + case VG_BW_1: + return PIPE_FORMAT_A8R8G8B8_UNORM; + case VG_A_8: + return PIPE_FORMAT_A8_UNORM; +#ifdef OPENVG_VERSION_1_1 + case VG_A_1: + case VG_A_4: + return PIPE_FORMAT_A8_UNORM; +#endif + default: + return PIPE_FORMAT_A8R8G8B8_UNORM; + } +} + +static INLINE void vg_sync_size(VGfloat *src_loc, VGfloat *dst_loc) +{ + src_loc[2] = MIN2(src_loc[2], dst_loc[2]); + src_loc[3] = MIN2(src_loc[3], dst_loc[3]); + dst_loc[2] = src_loc[2]; + dst_loc[3] = src_loc[3]; +} + + +static void vg_copy_texture(struct vg_context *ctx, + struct pipe_texture *dst, VGint dx, VGint dy, + struct pipe_texture *src, VGint sx, VGint sy, + VGint width, VGint height) +{ + VGfloat dst_loc[4], src_loc[4]; + VGfloat dst_bounds[4], src_bounds[4]; + VGfloat src_shift[4], dst_shift[4], shift[4]; + + dst_loc[0] = dx; + dst_loc[1] = dy; + dst_loc[2] = width; + dst_loc[3] = height; + dst_bounds[0] = 0.f; + dst_bounds[1] = 0.f; + dst_bounds[2] = dst->width[0]; + dst_bounds[3] = dst->height[0]; + + src_loc[0] = sx; + src_loc[1] = sy; + src_loc[2] = width; + src_loc[3] = height; + src_bounds[0] = 0.f; + src_bounds[1] = 0.f; + src_bounds[2] = src->width[0]; + src_bounds[3] = src->height[0]; + + vg_bound_rect(src_loc, src_bounds, src_shift); + vg_bound_rect(dst_loc, dst_bounds, dst_shift); + shift[0] = src_shift[0] - dst_shift[0]; + shift[1] = src_shift[1] - dst_shift[1]; + + if (shift[0] < 0) + vg_shift_rectx(src_loc, src_bounds, -shift[0]); + else + vg_shift_rectx(dst_loc, dst_bounds, shift[0]); + + if (shift[1] < 0) + vg_shift_recty(src_loc, src_bounds, -shift[1]); + else + vg_shift_recty(dst_loc, dst_bounds, shift[1]); + + vg_sync_size(src_loc, dst_loc); + + if (src_loc[2] >= 0 && src_loc[3] >= 0 && + dst_loc[2] >= 0 && dst_loc[3] >= 0) { + renderer_copy_texture(ctx->renderer, + src, + src_loc[0], + src_loc[1] + src_loc[3], + src_loc[0] + src_loc[2], + src_loc[1], + dst, + dst_loc[0], + dst_loc[1] + dst_loc[3], + dst_loc[0] + dst_loc[2], + dst_loc[1]); + } + +} + +void vg_copy_surface(struct vg_context *ctx, + struct pipe_surface *dst, VGint dx, VGint dy, + struct pipe_surface *src, VGint sx, VGint sy, + VGint width, VGint height) +{ + VGfloat dst_loc[4], src_loc[4]; + VGfloat dst_bounds[4], src_bounds[4]; + VGfloat src_shift[4], dst_shift[4], shift[4]; + + dst_loc[0] = dx; + dst_loc[1] = dy; + dst_loc[2] = width; + dst_loc[3] = height; + dst_bounds[0] = 0.f; + dst_bounds[1] = 0.f; + dst_bounds[2] = dst->width; + dst_bounds[3] = dst->height; + + src_loc[0] = sx; + src_loc[1] = sy; + src_loc[2] = width; + src_loc[3] = height; + src_bounds[0] = 0.f; + src_bounds[1] = 0.f; + src_bounds[2] = src->width; + src_bounds[3] = src->height; + + vg_bound_rect(src_loc, src_bounds, src_shift); + vg_bound_rect(dst_loc, dst_bounds, dst_shift); + shift[0] = src_shift[0] - dst_shift[0]; + shift[1] = src_shift[1] - dst_shift[1]; + + if (shift[0] < 0) + vg_shift_rectx(src_loc, src_bounds, -shift[0]); + else + vg_shift_rectx(dst_loc, dst_bounds, shift[0]); + + if (shift[1] < 0) + vg_shift_recty(src_loc, src_bounds, -shift[1]); + else + vg_shift_recty(dst_loc, dst_bounds, shift[1]); + + vg_sync_size(src_loc, dst_loc); + + if (src_loc[2] > 0 && src_loc[3] > 0 && + dst_loc[2] > 0 && dst_loc[3] > 0) { + if (src == dst) + renderer_copy_surface(ctx->renderer, + src, + src_loc[0], + src->height - (src_loc[1] + src_loc[3]), + src_loc[0] + src_loc[2], + src->height - src_loc[1], + dst, + dst_loc[0], + dst->height - (dst_loc[1] + dst_loc[3]), + dst_loc[0] + dst_loc[2], + dst->height - dst_loc[1], + 0, 0); + else + renderer_copy_surface(ctx->renderer, + src, + src_loc[0], + src->height - src_loc[1], + src_loc[0] + src_loc[2], + src->height - (src_loc[1] + src_loc[3]), + dst, + dst_loc[0], + dst->height - (dst_loc[1] + dst_loc[3]), + dst_loc[0] + dst_loc[2], + dst->height - dst_loc[1], + 0, 0); + } + +} + +static struct pipe_texture *image_texture(struct vg_image *img) +{ + struct pipe_texture *tex = img->texture; + return tex; +} + + +static void image_cleari(struct vg_image *img, VGint clear_colori, + VGint x, VGint y, VGint width, VGint height) +{ + VGint *clearbuf; + VGint i; + VGfloat dwidth, dheight; + + clearbuf = malloc(sizeof(VGint)*width*height); + for (i = 0; i < width*height; ++i) + clearbuf[i] = clear_colori; + + dwidth = MIN2(width, img->width); + dheight = MIN2(height, img->height); + + image_sub_data(img, clearbuf, width * sizeof(VGint), + VG_sRGBA_8888, + x, y, dwidth, dheight); + free(clearbuf); +} + +struct vg_image * image_create(VGImageFormat format, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *image = CALLOC_STRUCT(vg_image); + enum pipe_format pformat = vg_format_to_pipe(format); + struct pipe_texture pt, *newtex; + struct pipe_screen *screen = ctx->pipe->screen; + + vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE); + + image->format = format; + image->width = width; + image->height = height; + + image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + image->sampler.normalized_coords = 1; + + assert(screen->is_format_supported(screen, pformat, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + + memset(&pt, 0, sizeof(pt)); + pt.target = PIPE_TEXTURE_2D; + pt.format = pformat; + pf_get_block(pformat, &pt.block); + pt.last_level = 0; + pt.width[0] = width; + pt.height[0] = height; + pt.depth[0] = 1; + pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + + newtex = screen->texture_create(screen, &pt); + + debug_assert(newtex); + + image->texture = newtex; + + vg_context_add_object(ctx, VG_OBJECT_IMAGE, image); + + image_cleari(image, 0, 0, 0, image->width, image->height); + return image; +} + +void image_destroy(struct vg_image *img) +{ + struct vg_context *ctx = vg_current_context(); + vg_context_remove_object(ctx, VG_OBJECT_IMAGE, img); + + + if (img->parent) { + /* remove img from the parent child array */ + int idx; + struct vg_image **array = + (struct vg_image **)img->parent->children_array->data; + + for (idx = 0; idx < img->parent->children_array->num_elements; ++idx) { + struct vg_image *child = array[idx]; + if (child == img) { + break; + } + } + debug_assert(idx < img->parent->children_array->num_elements); + array_remove_element(img->parent->children_array, idx); + } + + if (img->children_array && img->children_array->num_elements) { + /* reparent the children */ + VGint i; + struct vg_image *parent = img->parent; + struct vg_image **children = + (struct vg_image **)img->children_array->data; + if (!parent) { + VGint min_x = children[0]->x; + parent = children[0]; + + for (i = 1; i < img->children_array->num_elements; ++i) { + struct vg_image *child = children[i]; + if (child->x < min_x) { + parent = child; + } + } + } + + for (i = 0; i < img->children_array->num_elements; ++i) { + struct vg_image *child = children[i]; + if (child != parent) { + child->parent = parent; + if (!parent->children_array) { + parent->children_array = array_create( + sizeof(struct vg_image*)); + } + array_append_data(parent->children_array, + &child, 1); + } else + child->parent = NULL; + } + array_destroy(img->children_array); + } + + pipe_texture_reference(&img->texture, NULL); + free(img); +} + +void image_clear(struct vg_image *img, + VGint x, VGint y, VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + VGfloat *clear_colorf = ctx->state.vg.clear_color; + VGubyte r, g, b ,a; + VGint clear_colori; + /* FIXME: this is very nasty */ + r = float_to_ubyte(clear_colorf[0]); + g = float_to_ubyte(clear_colorf[1]); + b = float_to_ubyte(clear_colorf[2]); + a = float_to_ubyte(clear_colorf[3]); + clear_colori = r << 24 | g << 16 | b << 8 | a; + image_cleari(img, clear_colori, x, y, width, height); +} + +void image_sub_data(struct vg_image *image, + const void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint x, VGint y, + VGint width, VGint height) +{ + const VGint yStep = 1; + VGubyte *src = (VGubyte *)data; + VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4]; + VGfloat *df = (VGfloat*)temp; + VGint i; + struct vg_context *ctx = vg_current_context(); + struct pipe_screen *screen = ctx->pipe->screen; + struct pipe_texture *texture = image_texture(image); + VGint xoffset = 0, yoffset = 0; + + if (x < 0) { + xoffset -= x; + width += x; + x = 0; + } + if (y < 0) { + yoffset -= y; + height += y; + y = 0; + } + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (x > image->width || y > image->width) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + + if (x + width > image->width) { + width = image->width - x; + } + + if (y + height > image->height) { + height = image->height - y; + } + + { /* upload color_data */ + struct pipe_transfer *transfer = screen->get_tex_transfer( + screen, texture, 0, 0, 0, + PIPE_TRANSFER_WRITE, 0, 0, texture->width[0], texture->height[0]); + src += (dataStride * yoffset); + for (i = 0; i < height; i++) { + _vega_unpack_float_span_rgba(ctx, width, xoffset, src, dataFormat, temp); + pipe_put_tile_rgba(transfer, x+image->x, y+image->y, width, 1, df); + y += yStep; + src += dataStride; + } + screen->tex_transfer_destroy(transfer); + } +} + +void image_get_sub_data(struct vg_image * image, + void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4]; + VGfloat *df = (VGfloat*)temp; + VGint y = 0, yStep = 1; + VGint i; + VGubyte *dst = (VGubyte *)data; + + { + struct pipe_transfer *transfer = + screen->get_tex_transfer(screen, + image->texture, 0, 0, 0, + PIPE_TRANSFER_READ, + 0, 0, + image->x + image->width, + image->y + image->height); + /* Do a row at a time to flip image data vertically */ + for (i = 0; i < height; i++) { +#if 0 + debug_printf("%d-%d == %d\n", sy, height, y); +#endif + pipe_get_tile_rgba(transfer, sx+image->x, y, width, 1, df); + y += yStep; + _vega_pack_rgba_span_float(ctx, width, temp, dataFormat, dst); + dst += dataStride; + } + + screen->tex_transfer_destroy(transfer); + } +} + +struct vg_image * image_child_image(struct vg_image *parent, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_image *image = CALLOC_STRUCT(vg_image); + + vg_init_object(&image->base, ctx, VG_OBJECT_IMAGE); + + image->x = parent->x + x; + image->y = parent->y + y; + image->width = width; + image->height = height; + image->parent = parent; + image->texture = 0; + pipe_texture_reference(&image->texture, + parent->texture); + + image->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + image->sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + image->sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + image->sampler.normalized_coords = 1; + + if (!parent->children_array) + parent->children_array = array_create( + sizeof(struct vg_image*)); + + array_append_data(parent->children_array, + &image, 1); + + vg_context_add_object(ctx, VG_OBJECT_IMAGE, image); + + return image; +} + +void image_copy(struct vg_image *dst, VGint dx, VGint dy, + struct vg_image *src, VGint sx, VGint sy, + VGint width, VGint height, + VGboolean dither) +{ + struct vg_context *ctx = vg_current_context(); + + if (width <= 0 || height <= 0) { + vg_set_error(ctx, VG_ILLEGAL_ARGUMENT_ERROR); + return; + } + /* make sure rendering has completed */ + ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + vg_copy_texture(ctx, dst->texture, dst->x + dx, dst->y + dy, + src->texture, src->x + sx, src->y + sy, width, height); +} + +void image_draw(struct vg_image *img) +{ + struct vg_context *ctx = vg_current_context(); + VGfloat x1, y1; + VGfloat x2, y2; + VGfloat x3, y3; + VGfloat x4, y4; + struct matrix *matrix; + + x1 = 0; + y1 = 0; + x2 = img->width; + y2 = 0; + x3 = img->width; + y3 = img->height; + x4 = 0; + y4 = img->height; + + matrix = &ctx->state.vg.image_user_to_surface_matrix; + + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + matrix_map_point(matrix, x4, y4, &x4, &y4); + + shader_set_drawing_image(ctx->shader, VG_TRUE); + shader_set_paint(ctx->shader, ctx->state.vg.fill_paint); + shader_set_image(ctx->shader, img); + shader_bind(ctx->shader); + + renderer_texture_quad(ctx->renderer, image_texture(img), + img->x, img->y, img->x + img->width, img->y + img->height, + x1, y1, x2, y2, x3, y3, x4, y4); +} + +void image_set_pixels(VGint dx, VGint dy, + struct vg_image *src, VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surf; + struct st_renderbuffer *strb = ctx->draw_buffer->strb; + + /* make sure rendering has completed */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + + surf = screen->get_tex_surface(screen, image_texture(src), 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ); + + vg_copy_surface(ctx, strb->surface, dx, dy, + surf, sx+src->x, sy+src->y, width, height); + + screen->tex_surface_destroy(surf); +} + +void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surf; + struct st_renderbuffer *strb = ctx->draw_buffer->strb; + + /* flip the y coordinates */ + /*dy = dst->height - dy - height;*/ + + /* make sure rendering has completed */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + + surf = screen->get_tex_surface(screen, image_texture(dst), 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ); + vg_copy_surface(ctx, surf, dst->x + dx, dst->y + dy, + strb->surface, sx, sy, width, height); + + pipe_surface_reference(&surf, NULL); +} + + +VGboolean vg_image_overlaps(struct vg_image *dst, + struct vg_image *src) +{ + if (dst == src || dst->parent == src || + dst == src->parent) + return VG_TRUE; + if (dst->parent && dst->parent == src->parent) { + VGfloat left1 = dst->x; + VGfloat left2 = src->x; + VGfloat right1 = dst->x + dst->width; + VGfloat right2 = src->x + src->width; + VGfloat bottom1 = dst->y; + VGfloat bottom2 = src->y; + VGfloat top1 = dst->y + dst->height; + VGfloat top2 = src->y + src->height; + + return !(left2 > right1 || right2 < left1 || + top2 > bottom1 || bottom2 < top1); + } + return VG_FALSE; +} + +VGint image_bind_samplers(struct vg_image *img, struct pipe_sampler_state **samplers, + struct pipe_texture **textures) +{ + img->sampler.min_img_filter = image_sampler_filter(img->base.ctx); + img->sampler.mag_img_filter = image_sampler_filter(img->base.ctx); + samplers[3] = &img->sampler; + textures[3] = img->texture; + return 1; +} + +VGint image_sampler_filter(struct vg_context *ctx) +{ + switch(ctx->state.vg.image_quality) { + case VG_IMAGE_QUALITY_NONANTIALIASED: + return PIPE_TEX_FILTER_NEAREST; + break; + case VG_IMAGE_QUALITY_FASTER: + return PIPE_TEX_FILTER_NEAREST; + break; + case VG_IMAGE_QUALITY_BETTER: + /*return PIPE_TEX_FILTER_ANISO;*/ + return PIPE_TEX_FILTER_LINEAR; + break; + default: + debug_printf("Unknown image quality"); + } + return PIPE_TEX_FILTER_NEAREST; +} diff --git a/src/gallium/state_trackers/vega/image.h b/src/gallium/state_trackers/vega/image.h new file mode 100644 index 00000000000..78e17cffa64 --- /dev/null +++ b/src/gallium/state_trackers/vega/image.h @@ -0,0 +1,104 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef IMAGES_H +#define IMAGES_H + +#include "vg_context.h" +#include "pipe/p_state.h" + +struct pipe_texture; +struct array; +struct vg_context; +struct pipe_surface; + +struct vg_image { + struct vg_object base; + VGImageFormat format; + VGint x, y; + VGint width, height; + + struct vg_image *parent; + + struct pipe_texture *texture; + struct pipe_sampler_state sampler; + + struct array *children_array; +}; + +struct vg_image *image_create(VGImageFormat format, + VGint width, VGint height); +void image_destroy(struct vg_image *img); + +void image_clear(struct vg_image *img, + VGint x, VGint y, VGint width, VGint height); + +void image_sub_data(struct vg_image *image, + const void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint x, VGint y, + VGint width, VGint height); + +void image_get_sub_data(struct vg_image * image, + void * data, + VGint dataStride, + VGImageFormat dataFormat, + VGint x, VGint y, + VGint width, VGint height); + +struct vg_image *image_child_image(struct vg_image *parent, + VGint x, VGint y, + VGint width, VGint height); + +void image_copy(struct vg_image *dst, VGint dx, VGint dy, + struct vg_image *src, VGint sx, VGint sy, + VGint width, VGint height, + VGboolean dither); + +void image_draw(struct vg_image *img); + +void image_set_pixels(VGint dx, VGint dy, + struct vg_image *src, VGint sx, VGint sy, + VGint width, VGint height); +void image_get_pixels(struct vg_image *dst, VGint dx, VGint dy, + VGint sx, VGint sy, + VGint width, VGint height); + +VGint image_bind_samplers(struct vg_image *dst, struct pipe_sampler_state **samplers, + struct pipe_texture **textures); + +VGboolean vg_image_overlaps(struct vg_image *dst, + struct vg_image *src); + +VGint image_sampler_filter(struct vg_context *ctx); + +void vg_copy_surface(struct vg_context *ctx, + struct pipe_surface *dst, VGint dx, VGint dy, + struct pipe_surface *src, VGint sx, VGint sy, + VGint width, VGint height); + +#endif diff --git a/src/gallium/state_trackers/vega/mask.c b/src/gallium/state_trackers/vega/mask.c new file mode 100644 index 00000000000..24650a37d50 --- /dev/null +++ b/src/gallium/state_trackers/vega/mask.c @@ -0,0 +1,690 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mask.h" + +#include "path.h" +#include "image.h" +#include "shaders_cache.h" +#include "renderer.h" +#include "asm_util.h" +#include "st_inlines.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "util/u_memory.h" + +struct vg_mask_layer { + struct vg_object base; + + VGint width; + VGint height; + + struct pipe_texture *texture; +}; + +static INLINE struct pipe_surface * +alpha_mask_surface(struct vg_context *ctx, int usage) +{ + struct pipe_screen *screen = ctx->pipe->screen; + struct st_framebuffer *stfb = ctx->draw_buffer; + return screen->get_tex_surface(screen, + stfb->alpha_mask, + 0, 0, 0, + usage); +} + +static INLINE VGboolean +intersect_rectangles(VGint dwidth, VGint dheight, + VGint swidth, VGint sheight, + VGint tx, VGint ty, + VGint twidth, VGint theight, + VGint *offsets, + VGint *location) +{ + if (tx + twidth <= 0 || tx >= dwidth) + return VG_FALSE; + if (ty + theight <= 0 || ty >= dheight) + return VG_FALSE; + + offsets[0] = 0; + offsets[1] = 0; + location[0] = tx; + location[1] = ty; + + if (tx < 0) { + offsets[0] -= tx; + location[0] = 0; + + location[2] = MIN2(tx + swidth, MIN2(dwidth, tx + twidth)); + offsets[2] = location[2]; + } else { + offsets[2] = MIN2(twidth, MIN2(dwidth - tx, swidth )); + location[2] = offsets[2]; + } + + if (ty < 0) { + offsets[1] -= ty; + location[1] = 0; + + location[3] = MIN2(ty + sheight, MIN2(dheight, ty + theight)); + offsets[3] = location[3]; + } else { + offsets[3] = MIN2(theight, MIN2(dheight - ty, sheight)); + location[3] = offsets[3]; + } + + return VG_TRUE; +} + +#if DEBUG_MASKS +static void read_alpha_mask(void * data, VGint dataStride, + VGImageFormat dataFormat, + VGint sx, VGint sy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + + struct st_framebuffer *stfb = ctx->draw_buffer; + struct st_renderbuffer *strb = stfb->alpha_mask; + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + + VGfloat temp[VEGA_MAX_IMAGE_WIDTH][4]; + VGfloat *df = (VGfloat*)temp; + VGint y = (fb->height - sy) - 1, yStep = -1; + VGint i; + VGubyte *dst = (VGubyte *)data; + VGint xoffset = 0, yoffset = 0; + + /* make sure rendering has completed */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + if (sx < 0) { + xoffset = -sx; + xoffset *= _vega_size_for_format(dataFormat); + width += sx; + sx = 0; + } + if (sy < 0) { + yoffset = -sy; + height += sy; + sy = 0; + y = (fb->height - sy) - 1; + yoffset *= dataStride; + } + + { + struct pipe_surface *surf; + + surf = screen->get_tex_surface(screen, strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + + /* Do a row at a time to flip image data vertically */ + for (i = 0; i < height; i++) { +#if 0 + debug_printf("%d-%d == %d\n", sy, height, y); +#endif + pipe_get_tile_rgba(surf, sx, y, width, 1, df); + y += yStep; + _vega_pack_rgba_span_float(ctx, width, temp, dataFormat, + dst + yoffset + xoffset); + dst += dataStride; + } + + pipe_surface_reference(&surf, NULL); + } +} + +void save_alpha_to_file(const char *filename) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + VGint *data; + int i, j; + + data = malloc(sizeof(int) * fb->width * fb->height); + read_alpha_mask(data, fb->width * sizeof(int), + VG_sRGBA_8888, + 0, 0, fb->width, fb->height); + fprintf(stderr, "/*---------- start */\n"); + fprintf(stderr, "const int image_width = %d;\n", + fb->width); + fprintf(stderr, "const int image_height = %d;\n", + fb->height); + fprintf(stderr, "const int image_data = {\n"); + for (i = 0; i < fb->height; ++i) { + for (j = 0; j < fb->width; ++j) { + int rgba = data[i * fb->height + j]; + int argb = 0; + argb = (rgba >> 8); + argb |= ((rgba & 0xff) << 24); + fprintf(stderr, "0x%x, ", argb); + } + fprintf(stderr, "\n"); + } + fprintf(stderr, "};\n"); + fprintf(stderr, "/*---------- end */\n"); +} +#endif + +static void setup_mask_framebuffer(struct pipe_surface *surf, + VGint surf_width, VGint surf_height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_framebuffer_state fb; + + memset(&fb, 0, sizeof(fb)); + fb.width = surf_width; + fb.height = surf_height; + fb.nr_cbufs = 1; + fb.cbufs[0] = surf; + { + VGint i; + for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i) + fb.cbufs[i] = 0; + } + cso_set_framebuffer(ctx->cso_context, &fb); +} + + +/* setup shader constants */ +static void setup_mask_operation(VGMaskOperation operation) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf; + const VGint param_bytes = 4 * sizeof(VGfloat); + const VGfloat ones[4] = {1.f, 1.f, 1.f, 1.f}; + void *shader = 0; + + /* We always need to get a new buffer, to keep the drivers simple and + * avoid gratuitous rendering synchronization. + */ + pipe_buffer_reference(&cbuf->buffer, NULL); + + cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + if (cbuf->buffer) { + st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, + 0, param_bytes, ones); + } + + ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); + switch (operation) { + case VG_UNION_MASK: { + if (!ctx->mask.union_fs) { + ctx->mask.union_fs = shader_create_from_text(ctx->pipe, + union_mask_asm, + 200, + PIPE_SHADER_FRAGMENT); + } + shader = ctx->mask.union_fs->driver; + } + break; + case VG_INTERSECT_MASK: { + if (!ctx->mask.intersect_fs) { + ctx->mask.intersect_fs = shader_create_from_text(ctx->pipe, + intersect_mask_asm, + 200, + PIPE_SHADER_FRAGMENT); + } + shader = ctx->mask.intersect_fs->driver; + } + break; + case VG_SUBTRACT_MASK: { + if (!ctx->mask.subtract_fs) { + ctx->mask.subtract_fs = shader_create_from_text(ctx->pipe, + subtract_mask_asm, + 200, + PIPE_SHADER_FRAGMENT); + } + shader = ctx->mask.subtract_fs->driver; + } + break; + case VG_SET_MASK: { + if (!ctx->mask.set_fs) { + ctx->mask.set_fs = shader_create_from_text(ctx->pipe, + set_mask_asm, + 200, + PIPE_SHADER_FRAGMENT); + } + shader = ctx->mask.set_fs->driver; + } + break; + default: + assert(0); + break; + } + cso_set_fragment_shader_handle(ctx->cso_context, shader); +} + +static void setup_mask_samplers(struct pipe_texture *umask) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + struct st_framebuffer *fb_buffers = ctx->draw_buffer; + struct pipe_texture *uprev = NULL; + struct pipe_sampler_state sampler; + + uprev = fb_buffers->blend_texture; + sampler = ctx->mask.sampler; + sampler.normalized_coords = 1; + + samplers[0] = NULL; + samplers[1] = NULL; + samplers[2] = NULL; + textures[0] = NULL; + textures[1] = NULL; + textures[2] = NULL; + + samplers[0] = &sampler; + samplers[1] = &ctx->mask.sampler; + + textures[0] = umask; + textures[1] = uprev; + + cso_set_samplers(ctx->cso_context, 2, + (const struct pipe_sampler_state **)samplers); + cso_set_sampler_textures(ctx->cso_context, 2, textures); +} + + +/* setup shader constants */ +static void setup_mask_fill(const VGfloat color[4]) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf; + const VGint param_bytes = 4 * sizeof(VGfloat); + + /* We always need to get a new buffer, to keep the drivers simple and + * avoid gratuitous rendering synchronization. + */ + pipe_buffer_reference(&cbuf->buffer, NULL); + + cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 1, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + if (cbuf->buffer) { + st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, 0, param_bytes, color); + } + + ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); + cso_set_fragment_shader_handle(ctx->cso_context, + shaders_cache_fill(ctx->sc, + VEGA_SOLID_FILL_SHADER)); +} + +static void setup_mask_viewport() +{ + struct vg_context *ctx = vg_current_context(); + vg_set_viewport(ctx, VEGA_Y0_TOP); +} + +static void setup_mask_blend() +{ + struct vg_context *ctx = vg_current_context(); + + struct pipe_blend_state blend; + + memset(&blend, 0, sizeof(struct pipe_blend_state)); + blend.blend_enable = 1; + blend.colormask |= PIPE_MASK_R; + blend.colormask |= PIPE_MASK_G; + blend.colormask |= PIPE_MASK_B; + blend.colormask |= PIPE_MASK_A; + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + + cso_set_blend(ctx->cso_context, &blend); +} + + +static void surface_fill(struct pipe_surface *surf, + int surf_width, int surf_height, + int x, int y, int width, int height, + const VGfloat color[4]) +{ + struct vg_context *ctx = vg_current_context(); + + if (x < 0) { + width += x; + x = 0; + } + if (y < 0) { + height += y; + y = 0; + } + + cso_save_framebuffer(ctx->cso_context); + cso_save_blend(ctx->cso_context); + cso_save_fragment_shader(ctx->cso_context); + cso_save_viewport(ctx->cso_context); + + setup_mask_blend(); + setup_mask_fill(color); + setup_mask_framebuffer(surf, surf_width, surf_height); + setup_mask_viewport(); + + renderer_draw_quad(ctx->renderer, x, y, + x + width, y + height, 0.0f/*depth should be disabled*/); + + + /* make sure rendering has completed */ + ctx->pipe->flush(ctx->pipe, + PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, + NULL); + +#if DEBUG_MASKS + save_alpha_to_file(0); +#endif + + cso_restore_blend(ctx->cso_context); + cso_restore_framebuffer(ctx->cso_context); + cso_restore_fragment_shader(ctx->cso_context); + cso_restore_viewport(ctx->cso_context); +} + + +static void mask_using_texture(struct pipe_texture *texture, + VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct pipe_surface *surface = + alpha_mask_surface(ctx, PIPE_BUFFER_USAGE_GPU_WRITE); + VGint offsets[4], loc[4]; + + if (!surface) + return; + if (!intersect_rectangles(surface->width, surface->height, + texture->width[0], texture->height[0], + x, y, width, height, + offsets, loc)) + return; +#if 0 + debug_printf("Offset = [%d, %d, %d, %d]\n", offsets[0], + offsets[1], offsets[2], offsets[3]); + debug_printf("Locati = [%d, %d, %d, %d]\n", loc[0], + loc[1], loc[2], loc[3]); +#endif + + /* prepare our blend surface */ + vg_prepare_blend_surface_from_mask(ctx); + + cso_save_samplers(ctx->cso_context); + cso_save_sampler_textures(ctx->cso_context); + cso_save_framebuffer(ctx->cso_context); + cso_save_blend(ctx->cso_context); + cso_save_fragment_shader(ctx->cso_context); + cso_save_viewport(ctx->cso_context); + + setup_mask_samplers(texture); + setup_mask_blend(); + setup_mask_operation(operation); + setup_mask_framebuffer(surface, surface->width, surface->height); + setup_mask_viewport(); + + /* render the quad to propagate the rendering from stencil */ + renderer_draw_texture(ctx->renderer, texture, + offsets[0], offsets[1], + offsets[0] + offsets[2], offsets[1] + offsets[3], + loc[0], loc[1], loc[0] + loc[2], loc[1] + loc[3]); + + /* make sure rendering has completed */ + ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + cso_restore_blend(ctx->cso_context); + cso_restore_framebuffer(ctx->cso_context); + cso_restore_fragment_shader(ctx->cso_context); + cso_restore_samplers(ctx->cso_context); + cso_restore_sampler_textures(ctx->cso_context); + cso_restore_viewport(ctx->cso_context); + + pipe_surface_reference(&surface, NULL); +} + + +#ifdef OPENVG_VERSION_1_1 + +struct vg_mask_layer * mask_layer_create(VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct vg_mask_layer *mask = 0; + + mask = CALLOC_STRUCT(vg_mask_layer); + vg_init_object(&mask->base, ctx, VG_OBJECT_MASK); + mask->width = width; + mask->height = height; + + { + struct pipe_texture pt; + struct pipe_screen *screen = ctx->pipe->screen; + + memset(&pt, 0, sizeof(pt)); + pt.target = PIPE_TEXTURE_2D; + pt.format = PIPE_FORMAT_A8R8G8B8_UNORM; + pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &pt.block); + pt.last_level = 0; + pt.width[0] = width; + pt.height[0] = height; + pt.depth[0] = 1; + pt.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + pt.compressed = 0; + + mask->texture = screen->texture_create(screen, &pt); + } + + vg_context_add_object(ctx, VG_OBJECT_MASK, mask); + + return mask; +} + +void mask_layer_destroy(struct vg_mask_layer *layer) +{ + struct vg_context *ctx = vg_current_context(); + + vg_context_remove_object(ctx, VG_OBJECT_MASK, layer); + pipe_texture_release(&layer->texture); + free(layer); +} + +void mask_layer_fill(struct vg_mask_layer *layer, + VGint x, VGint y, + VGint width, VGint height, + VGfloat value) +{ + struct vg_context *ctx = vg_current_context(); + VGfloat alpha_color[4] = {0, 0, 0, 0}; + struct pipe_surface *surface; + + alpha_color[3] = value; + + surface = ctx->pipe->screen->get_tex_surface( + ctx->pipe->screen, layer->texture, + 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + surface_fill(surface, + layer->width, layer->height, + x, y, width, height, alpha_color); + + ctx->pipe->screen->tex_surface_release(ctx->pipe->screen, &surface); +} + +void mask_copy(struct vg_mask_layer *layer, + VGint sx, VGint sy, + VGint dx, VGint dy, + VGint width, VGint height) +{ + struct vg_context *ctx = vg_current_context(); + struct st_framebuffer *fb_buffers = ctx->draw_buffer; + + renderer_copy_texture(ctx->renderer, + layer->texture, + sx, sy, + sx + width, sy + height, + fb_buffers->alpha_mask, + dx, dy, + dx + width, dy + height); +} + +static void mask_layer_render_to(struct vg_mask_layer *layer, + struct path *path, + VGbitfield paint_modes) +{ + struct vg_context *ctx = vg_current_context(); + const VGfloat fill_color[4] = {1.f, 1.f, 1.f, 1.f}; + struct pipe_screen *screen = ctx->pipe->screen; + struct pipe_surface *surface; + + surface = screen->get_tex_surface(screen, layer->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + cso_save_framebuffer(ctx->cso_context); + cso_save_fragment_shader(ctx->cso_context); + cso_save_viewport(ctx->cso_context); + + setup_mask_blend(); + setup_mask_fill(fill_color); + setup_mask_framebuffer(surface, layer->width, layer->height); + setup_mask_viewport(); + + if (paint_modes & VG_FILL_PATH) { + struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix; + path_fill(path, mat); + } + + if (paint_modes & VG_STROKE_PATH){ + path_stroke(path); + } + + + /* make sure rendering has completed */ + ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + + cso_restore_framebuffer(ctx->cso_context); + cso_restore_fragment_shader(ctx->cso_context); + cso_restore_viewport(ctx->cso_context); + ctx->state.dirty |= BLEND_DIRTY; + + screen->tex_surface_release(ctx->pipe->screen, &surface); +} + +void mask_render_to(struct path *path, + VGbitfield paint_modes, + VGMaskOperation operation) +{ + struct vg_context *ctx = vg_current_context(); + struct st_framebuffer *fb_buffers = ctx->draw_buffer; + struct vg_mask_layer *temp_layer; + VGint width, height; + + width = fb_buffers->alpha_mask->width[0]; + height = fb_buffers->alpha_mask->width[0]; + + temp_layer = mask_layer_create(width, height); + + mask_layer_render_to(temp_layer, path, paint_modes); + + mask_using_layer(temp_layer, 0, 0, width, height, + operation); + + mask_layer_destroy(temp_layer); +} + +void mask_using_layer(struct vg_mask_layer *layer, + VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height) +{ + mask_using_texture(layer->texture, operation, + x, y, width, height); +} + +VGint mask_layer_width(struct vg_mask_layer *layer) +{ + return layer->width; +} + +VGint mask_layer_height(struct vg_mask_layer *layer) +{ + return layer->height; +} + + +#endif + +void mask_using_image(struct vg_image *image, + VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height) +{ + mask_using_texture(image->texture, operation, + x, y, width, height); +} + +void mask_fill(VGint x, VGint y, VGint width, VGint height, + VGfloat value) +{ + struct vg_context *ctx = vg_current_context(); + VGfloat alpha_color[4] = {.0f, .0f, .0f, value}; + struct pipe_surface *surf = alpha_mask_surface( + ctx, PIPE_BUFFER_USAGE_GPU_WRITE); + +#if DEBUG_MASKS + debug_printf("mask_fill(%d, %d, %d, %d) with rgba(%f, %f, %f, %f)\n", + x, y, width, height, + alpha_color[0], alpha_color[1], + alpha_color[2], alpha_color[3]); + debug_printf("XXX %f === %f \n", + alpha_color[3], value); +#endif + + surface_fill(surf, surf->width, surf->height, + x, y, width, height, alpha_color); + + pipe_surface_reference(&surf, NULL); +} + +VGint mask_bind_samplers(struct pipe_sampler_state **samplers, + struct pipe_texture **textures) +{ + struct vg_context *ctx = vg_current_context(); + + if (ctx->state.vg.masking) { + struct st_framebuffer *fb_buffers = ctx->draw_buffer; + + samplers[1] = &ctx->mask.sampler; + textures[1] = fb_buffers->alpha_mask; + return 1; + } else + return 0; +} diff --git a/src/gallium/state_trackers/vega/mask.h b/src/gallium/state_trackers/vega/mask.h new file mode 100644 index 00000000000..5eaaede0e3a --- /dev/null +++ b/src/gallium/state_trackers/vega/mask.h @@ -0,0 +1,68 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef MASK_H +#define MASK_H + +#include "vg_context.h" + +struct path; +struct vg_image; +struct pipe_texture; + +struct vg_mask_layer *mask_layer_create(VGint width, VGint height); +void mask_layer_destroy(struct vg_mask_layer *layer); +void mask_layer_fill(struct vg_mask_layer *layer, + VGint x, VGint y, + VGint width, VGint height, + VGfloat value); +VGint mask_layer_width(struct vg_mask_layer *layer); +VGint mask_layer_height(struct vg_mask_layer *layer); +void mask_copy(struct vg_mask_layer *layer, + VGint sx, VGint sy, + VGint dx, VGint dy, + VGint width, VGint height); + +void mask_render_to(struct path *path, + VGbitfield paint_modes, + VGMaskOperation operation); + +void mask_using_layer(struct vg_mask_layer *layer, + VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height); +void mask_using_image(struct vg_image *image, + VGMaskOperation operation, + VGint x, VGint y, + VGint width, VGint height); +void mask_fill(VGint x, VGint y, + VGint width, VGint height, + VGfloat value); + +VGint mask_bind_samplers(struct pipe_sampler_state **samplers, + struct pipe_texture **textures); + +#endif diff --git a/src/gallium/state_trackers/vega/matrix.h b/src/gallium/state_trackers/vega/matrix.h new file mode 100644 index 00000000000..4c207f912a8 --- /dev/null +++ b/src/gallium/state_trackers/vega/matrix.h @@ -0,0 +1,462 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef MATRIX_H +#define MATRIX_H + +#include "VG/openvg.h" + +#include "pipe/p_compiler.h" +#include "util/u_math.h" + +#include <stdio.h> +#include <math.h> + +#define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y))) +#define floatIsZero(x) (floatsEqual((x) + 1, 1)) +#define ABS(x) (fabsf(x)) + +#define DEGREES_TO_RADIANS(d) (0.0174532925199 * (d)) +#define FLT_TO_INT(flt) float_to_int_floor(((VGuint*)&(flt))[0]) + +static INLINE VGint float_to_int_floor(VGuint bits) +{ + int sign = (bits >> 31) ? -1 : 1; + int exp = ((bits >> 23) & 255) - 127; + int mant = bits & 0x007fffff; + int sh = 23 - exp; + + /* abs(value) >= 2^31 -> clamp. */ + + if (exp >= 31) + return (VGint)((sign < 0) ? 0x80000000u : 0x7fffffffu); + + /* abs(value) < 1 -> return -1 or 0. */ + + if (exp < 0) + return (sign < 0 && (exp > -127 || mant != 0)) ? -1 : 0; + + /* abs(value) >= 2^23 -> shift left. */ + + mant |= 0x00800000; + if (sh <= 0) + return sign * (mant << -sh); + + /* Negative -> add a rounding term. */ + + if (sign < 0) + mant += (1 << sh) - 1; + + /* Shift right to obtain the result. */ + + return sign * (mant >> sh); +} + + +struct matrix { + VGfloat m[9]; +}; + +static INLINE void matrix_init(struct matrix *mat, + const VGfloat *val) +{ + memcpy(mat->m, val, sizeof(VGfloat) * 9); +} + +static INLINE void matrix_inits(struct matrix *mat, + VGfloat m11, VGfloat m12, VGfloat m13, + VGfloat m21, VGfloat m22, VGfloat m23, + VGfloat m31, VGfloat m32, VGfloat m33) +{ + mat->m[0] = m11; mat->m[1] = m12; mat->m[2] = m13; + mat->m[3] = m21; mat->m[4] = m22; mat->m[5] = m23; + mat->m[6] = m31; mat->m[7] = m32; mat->m[8] = m33; +} + +static INLINE void matrix_load_identity(struct matrix *matrix) +{ + static const VGfloat identity[9] = {1.f, 0.f, 0.f, + 0.f, 1.f, 0.f, + 0.f, 0.f, 1.f}; + memcpy(matrix->m, identity, sizeof(identity)); +} + +static INLINE VGboolean matrix_is_identity(struct matrix *matrix) +{ + return floatsEqual(matrix->m[0], 1) && floatIsZero(matrix->m[1]) && + floatIsZero(matrix->m[2]) && + floatIsZero(matrix->m[3]) && floatsEqual(matrix->m[4], 1) && + floatIsZero(matrix->m[5]) && + floatIsZero(matrix->m[6]) && floatIsZero(matrix->m[7]) && + floatIsZero(matrix->m[8]); +} + +static INLINE VGboolean matrix_is_affine(struct matrix *matrix) +{ + return floatIsZero(matrix->m[2]) && floatIsZero(matrix->m[5]) + && floatsEqual(matrix->m[8], 1); +} + + +static INLINE void matrix_make_affine(struct matrix *matrix) +{ + matrix->m[2] = 0.f; + matrix->m[5] = 0.f; + matrix->m[8] = 1.f; +} + +static INLINE void matrix_mult(struct matrix *dst, + struct matrix *src) +{ + VGfloat m11 = dst->m[0]*src->m[0] + dst->m[3]*src->m[1] + dst->m[6]*src->m[2]; + VGfloat m12 = dst->m[0]*src->m[3] + dst->m[3]*src->m[4] + dst->m[6]*src->m[5]; + VGfloat m13 = dst->m[0]*src->m[6] + dst->m[3]*src->m[7] + dst->m[6]*src->m[8]; + + VGfloat m21 = dst->m[1]*src->m[0] + dst->m[4]*src->m[1] + dst->m[7]*src->m[2]; + VGfloat m22 = dst->m[1]*src->m[3] + dst->m[4]*src->m[4] + dst->m[7]*src->m[5]; + VGfloat m23 = dst->m[1]*src->m[6] + dst->m[4]*src->m[7] + dst->m[7]*src->m[8]; + + VGfloat m31 = dst->m[2]*src->m[0] + dst->m[5]*src->m[1] + dst->m[8]*src->m[2]; + VGfloat m32 = dst->m[2]*src->m[3] + dst->m[5]*src->m[4] + dst->m[8]*src->m[5]; + VGfloat m33 = dst->m[2]*src->m[6] + dst->m[5]*src->m[7] + dst->m[8]*src->m[8]; + + dst->m[0] = m11; dst->m[1] = m21; dst->m[2] = m31; + dst->m[3] = m12; dst->m[4] = m22; dst->m[5] = m32; + dst->m[6] = m13; dst->m[7] = m23; dst->m[8] = m33; +} + + +static INLINE void matrix_map_point(struct matrix *mat, + VGfloat x, VGfloat y, + VGfloat *out_x, VGfloat *out_y) +{ + /* to be able to do matrix_map_point(m, x, y, &x, &y) use + * temporaries */ + VGfloat tmp_x = x, tmp_y = y; + + *out_x = mat->m[0]*tmp_x + mat->m[3]*tmp_y + mat->m[6]; + *out_y = mat->m[1]*tmp_x + mat->m[4]*tmp_y + mat->m[7]; + if (!matrix_is_affine(mat)) { + VGfloat w = 1/(mat->m[2]*tmp_x + mat->m[5]*tmp_y + mat->m[8]); + *out_x *= w; + *out_y *= w; + } +} + +static INLINE void matrix_translate(struct matrix *dst, + VGfloat tx, VGfloat ty) +{ + if (!matrix_is_affine(dst)) { + struct matrix trans_matrix; + matrix_load_identity(&trans_matrix); + trans_matrix.m[6] = tx; + trans_matrix.m[7] = ty; + matrix_mult(dst, &trans_matrix); + } else { + dst->m[6] += tx*dst->m[0] + ty*dst->m[3]; + dst->m[7] += ty*dst->m[4] + tx*dst->m[1]; + } +} + +static INLINE void matrix_scale(struct matrix *dst, + VGfloat sx, VGfloat sy) +{ + if (!matrix_is_affine(dst)) { + struct matrix scale_matrix; + matrix_load_identity(&scale_matrix); + scale_matrix.m[0] = sx; + scale_matrix.m[4] = sy; + matrix_mult(dst, &scale_matrix); + } else { + dst->m[0] *= sx; dst->m[1] *= sx; + dst->m[3] *= sy; dst->m[4] *= sy; + } +} + +static INLINE void matrix_shear(struct matrix *dst, + VGfloat shx, VGfloat shy) +{ + struct matrix shear_matrix; + matrix_load_identity(&shear_matrix); + shear_matrix.m[1] = shy; + shear_matrix.m[3] = shx; + matrix_mult(dst, &shear_matrix); +} + +static INLINE void matrix_rotate(struct matrix *dst, + VGfloat angle) +{ + struct matrix mat; + float sin_val = 0; + float cos_val = 0; + + + if (floatsEqual(angle, 90) || floatsEqual(angle, -270)) + sin_val = 1.f; + else if (floatsEqual(angle, 270) || floatsEqual(angle, -90)) + sin_val = -1.f; + else if (floatsEqual(angle, 180)) + cos_val = -1.f; + else { + float radians = DEGREES_TO_RADIANS(angle); + sin_val = sin(radians); + cos_val = cos(radians); + } + + if (!matrix_is_affine(dst)) { + matrix_load_identity(&mat); + mat.m[0] = cos_val; mat.m[1] = sin_val; + mat.m[3] = -sin_val; mat.m[4] = cos_val; + + matrix_mult(dst, &mat); + } else { + VGfloat m11 = cos_val*dst->m[0] + sin_val*dst->m[3]; + VGfloat m12 = cos_val*dst->m[1] + sin_val*dst->m[4]; + VGfloat m21 = -sin_val*dst->m[0] + cos_val*dst->m[3]; + VGfloat m22 = -sin_val*dst->m[1] + cos_val*dst->m[4]; + dst->m[0] = m11; dst->m[1] = m12; + dst->m[3] = m21; dst->m[4] = m22; + } +} + + +static INLINE VGfloat matrix_determinant(struct matrix *mat) +{ + return mat->m[0]*(mat->m[8]*mat->m[4]-mat->m[7]*mat->m[5]) - + mat->m[3]*(mat->m[8]*mat->m[1]-mat->m[7]*mat->m[2])+ + mat->m[6]*(mat->m[5]*mat->m[1]-mat->m[4]*mat->m[2]); +} + + +static INLINE void matrix_adjoint(struct matrix *mat) +{ + VGfloat h[9]; + h[0] = mat->m[4]*mat->m[8] - mat->m[5]*mat->m[7]; + h[3] = mat->m[5]*mat->m[6] - mat->m[3]*mat->m[8]; + h[6] = mat->m[3]*mat->m[7] - mat->m[4]*mat->m[6]; + h[1] = mat->m[2]*mat->m[7] - mat->m[1]*mat->m[8]; + h[4] = mat->m[0]*mat->m[8] - mat->m[2]*mat->m[6]; + h[7] = mat->m[1]*mat->m[6] - mat->m[0]*mat->m[7]; + h[2] = mat->m[1]*mat->m[5] - mat->m[2]*mat->m[4]; + h[5] = mat->m[2]*mat->m[3] - mat->m[0]*mat->m[5]; + h[8] = mat->m[0]*mat->m[4] - mat->m[1]*mat->m[3]; + + + memcpy(mat->m, h, sizeof(VGfloat) * 9); +} + +static INLINE void matrix_divs(struct matrix *mat, + VGfloat s) +{ + mat->m[0] /= s; + mat->m[1] /= s; + mat->m[2] /= s; + mat->m[3] /= s; + mat->m[4] /= s; + mat->m[5] /= s; + mat->m[6] /= s; + mat->m[7] /= s; + mat->m[8] /= s; +} + +static INLINE VGboolean matrix_invert(struct matrix *mat) +{ + VGfloat det = matrix_determinant(mat); + + if (floatIsZero(det)) + return VG_FALSE; + + matrix_adjoint(mat); + matrix_divs(mat, det); + return VG_TRUE; +} + +static INLINE VGboolean matrix_is_invertible(struct matrix *mat) +{ + return !floatIsZero(matrix_determinant(mat)); +} + + +static INLINE VGboolean matrix_square_to_quad(VGfloat dx0, VGfloat dy0, + VGfloat dx1, VGfloat dy1, + VGfloat dx3, VGfloat dy3, + VGfloat dx2, VGfloat dy2, + struct matrix *mat) +{ + VGfloat ax = dx0 - dx1 + dx2 - dx3; + VGfloat ay = dy0 - dy1 + dy2 - dy3; + + if (floatIsZero(ax) && floatIsZero(ay)) { + /* affine case */ + matrix_inits(mat, + dx1 - dx0, dy1 - dy0, 0, + dx2 - dx1, dy2 - dy1, 0, + dx0, dy0, 1); + } else { + VGfloat a, b, c, d, e, f, g, h; + VGfloat ax1 = dx1 - dx2; + VGfloat ax2 = dx3 - dx2; + VGfloat ay1 = dy1 - dy2; + VGfloat ay2 = dy3 - dy2; + + /* determinants */ + VGfloat gtop = ax * ay2 - ax2 * ay; + VGfloat htop = ax1 * ay - ax * ay1; + VGfloat bottom = ax1 * ay2 - ax2 * ay1; + + if (!bottom) + return VG_FALSE; + + g = gtop / bottom; + h = htop / bottom; + + a = dx1 - dx0 + g * dx1; + b = dx3 - dx0 + h * dx3; + c = dx0; + d = dy1 - dy0 + g * dy1; + e = dy3 - dy0 + h * dy3; + f = dy0; + + matrix_inits(mat, + a, d, g, + b, e, h, + c, f, 1.f); + } + + return VG_TRUE; +} + +static INLINE VGboolean matrix_quad_to_square(VGfloat sx0, VGfloat sy0, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + VGfloat sx3, VGfloat sy3, + struct matrix *mat) +{ + if (!matrix_square_to_quad(sx0, sy0, sx1, sy1, + sx2, sy2, sx3, sy3, + mat)) + return VG_FALSE; + + return matrix_invert(mat); +} + + +static INLINE VGboolean matrix_quad_to_quad(VGfloat dx0, VGfloat dy0, + VGfloat dx1, VGfloat dy1, + VGfloat dx2, VGfloat dy2, + VGfloat dx3, VGfloat dy3, + VGfloat sx0, VGfloat sy0, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + VGfloat sx3, VGfloat sy3, + struct matrix *mat) +{ + struct matrix sqr_to_qd; + + if (!matrix_square_to_quad(dx0, dy0, dx1, dy1, + dx2, dy2, dx3, dy3, + mat)) + return VG_FALSE; + + if (!matrix_quad_to_square(sx0, sy0, sx1, sy1, + sx2, sy2, sx3, sy3, + &sqr_to_qd)) + return VG_FALSE; + + matrix_mult(mat, &sqr_to_qd); + + return VG_TRUE; +} + + +static INLINE VGboolean null_line(const VGfloat *l) +{ + return floatsEqual(l[0], l[2]) && floatsEqual(l[1], l[3]); +} + +static INLINE void line_normal(float *l, float *norm) +{ + norm[0] = l[0]; + norm[1] = l[1]; + + norm[2] = l[0] + (l[3] - l[1]); + norm[3] = l[1] - (l[2] - l[0]); +} + +static INLINE void line_normalize(float *l) +{ + float x = l[2] - l[0]; + float y = l[3] - l[1]; + float len = sqrt(x*x + y*y); + l[2] = l[0] + x/len; + l[3] = l[1] + y/len; +} + +static INLINE VGfloat line_length(VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2) +{ + VGfloat x = x2 - x1; + VGfloat y = y2 - y1; + return sqrt(x*x + y*y); +} + +static INLINE VGfloat line_lengthv(const VGfloat *l) +{ + VGfloat x = l[2] - l[0]; + VGfloat y = l[3] - l[1]; + return sqrt(x*x + y*y); +} + + +static INLINE void line_point_at(float *l, float t, float *pt) +{ + float dx = l[2] - l[0]; + float dy = l[3] - l[1]; + + pt[0] = l[0] + dx * t; + pt[1] = l[1] + dy * t; +} + +static INLINE void vector_unit(float *vec) +{ + float len = sqrt(vec[0] * vec[0] + vec[1] * vec[1]); + vec[0] /= len; + vec[1] /= len; +} + +static INLINE void line_normal_vector(float *line, float *vec) +{ + VGfloat normal[4]; + + line_normal(line, normal); + + vec[0] = normal[2] - normal[0]; + vec[1] = normal[3] - normal[1]; + + vector_unit(vec); +} + +#endif diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c new file mode 100644 index 00000000000..04a6ba9cdcd --- /dev/null +++ b/src/gallium/state_trackers/vega/paint.c @@ -0,0 +1,699 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "paint.h" + +#include "shaders_cache.h" +#include "matrix.h" +#include "image.h" +#include "st_inlines.h" + +#include "pipe/p_compiler.h" +#include "pipe/p_inlines.h" + +#include "util/u_memory.h" +#include "util/u_math.h" + +#include "cso_cache/cso_context.h" + +struct vg_paint { + struct vg_object base; + + VGPaintType type; + + struct { + VGfloat color[4]; + VGint colori[4]; + } solid; + + struct { + VGColorRampSpreadMode spread; + VGuint color_data[1024]; + struct { + VGfloat coords[4]; + VGint coordsi[4]; + } linear; + struct { + VGfloat vals[5]; + VGint valsi[5]; + } radial; + struct pipe_texture *texture; + struct pipe_sampler_state sampler; + + VGfloat *ramp_stops; + VGint *ramp_stopsi; + VGint num_stops; + + VGboolean color_ramps_premultiplied; + } gradient; + + struct { + struct pipe_texture *texture; + VGTilingMode tiling_mode; + struct pipe_sampler_state sampler; + } pattern; + + struct pipe_constant_buffer cbuf; + struct pipe_shader_state fs_state; + void *fs; +}; + +static INLINE VGuint mix_pixels(VGuint p1, VGuint a, VGuint p2, VGuint b) +{ + VGuint t = (p1 & 0xff00ff) * a + (p2 & 0xff00ff) * b; + t >>= 8; t &= 0xff00ff; + + p1 = ((p1 >> 8) & 0xff00ff) * a + ((p2 >> 8) & 0xff00ff) * b; + p1 &= 0xff00ff00; p1 |= t; + + return p1; +} + +static INLINE VGuint float4_to_argb(const VGfloat *clr) +{ + return float_to_ubyte(clr[3]) << 24 | + float_to_ubyte(clr[0]) << 16 | + float_to_ubyte(clr[1]) << 8 | + float_to_ubyte(clr[2]) << 0; +} + +static INLINE void create_gradient_data(const VGfloat *ramp_stops, + VGint num, + VGuint *data, + VGint size) +{ + VGint i; + VGint pos = 0; + VGfloat fpos = 0, incr = 1.f / size; + VGuint last_color; + + while (fpos < ramp_stops[0]) { + data[pos] = float4_to_argb(ramp_stops + 1); + fpos += incr; + ++pos; + } + + for (i = 0; i < num - 1; ++i) { + VGint rcur = 5 * i; + VGint rnext = 5 * (i + 1); + VGfloat delta = 1.f/(ramp_stops[rnext] - ramp_stops[rcur]); + while (fpos < ramp_stops[rnext] && pos < size) { + VGint dist = 256 * ((fpos - ramp_stops[rcur]) * delta); + VGint idist = 256 - dist; + VGuint current_color = float4_to_argb(ramp_stops + rcur + 1); + VGuint next_color = float4_to_argb(ramp_stops + rnext + 1); + data[pos] = mix_pixels(current_color, idist, + next_color, dist); + fpos += incr; + ++pos; + } + } + + last_color = float4_to_argb(ramp_stops + ((num - 1) * 5 + 1)); + while (pos < size) { + data[pos] = last_color; + ++pos; + } + data[size-1] = last_color; +} + +static INLINE struct pipe_texture *create_gradient_texture(struct vg_paint *p) +{ + struct pipe_context *pipe = p->base.ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_texture *tex = 0; + struct pipe_texture templ; + + memset(&templ, 0, sizeof(templ)); + templ.target = PIPE_TEXTURE_1D; + templ.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templ.last_level = 0; + templ.width[0] = 1024; + templ.height[0] = 1; + templ.depth[0] = 1; + pf_get_block(PIPE_FORMAT_A8R8G8B8_UNORM, &templ.block); + templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER; + + tex = screen->texture_create(screen, &templ); + + { /* upload color_data */ + struct pipe_transfer *transfer = + st_no_flush_get_tex_transfer(p->base.ctx, tex, 0, 0, 0, + PIPE_TRANSFER_WRITE, 0, 0, 1024, 1); + void *map = screen->transfer_map(screen, transfer); + memcpy(map, p->gradient.color_data, sizeof(VGint)*1024); + screen->transfer_unmap(screen, transfer); + screen->tex_transfer_destroy(transfer); + } + + return tex; +} + +struct vg_paint * paint_create(struct vg_context *ctx) +{ + struct vg_paint *paint = CALLOC_STRUCT(vg_paint); + const VGfloat default_color[] = {0.0f, 0.0f, 0.0f, 1.0f}; + const VGfloat def_ling[] = {0.0f, 0.0f, 1.0f, 0.0f}; + const VGfloat def_radg[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f}; + vg_init_object(&paint->base, ctx, VG_OBJECT_PAINT); + vg_context_add_object(ctx, VG_OBJECT_PAINT, paint); + + paint->type = VG_PAINT_TYPE_COLOR; + memcpy(paint->solid.color, default_color, + 4 * sizeof(VGfloat)); + paint->gradient.spread = VG_COLOR_RAMP_SPREAD_PAD; + memcpy(paint->gradient.linear.coords, def_ling, + 4 * sizeof(VGfloat)); + memcpy(paint->gradient.radial.vals, def_radg, + 5 * sizeof(VGfloat)); + + paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + paint->gradient.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + paint->gradient.sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + paint->gradient.sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + paint->gradient.sampler.normalized_coords = 1; + + memcpy(&paint->pattern.sampler, + &paint->gradient.sampler, + sizeof(struct pipe_sampler_state)); + + return paint; +} + +void paint_destroy(struct vg_paint *paint) +{ + struct vg_context *ctx = paint->base.ctx; + if (paint->pattern.texture) + pipe_texture_reference(&paint->pattern.texture, NULL); + if (ctx) + vg_context_remove_object(ctx, VG_OBJECT_PAINT, paint); + + free(paint->gradient.ramp_stopsi); + free(paint->gradient.ramp_stops); + free(paint); +} + +void paint_set_color(struct vg_paint *paint, + const VGfloat *color) +{ + paint->solid.color[0] = color[0]; + paint->solid.color[1] = color[1]; + paint->solid.color[2] = color[2]; + paint->solid.color[3] = color[3]; + + paint->solid.colori[0] = FLT_TO_INT(color[0]); + paint->solid.colori[1] = FLT_TO_INT(color[1]); + paint->solid.colori[2] = FLT_TO_INT(color[2]); + paint->solid.colori[3] = FLT_TO_INT(color[3]); +} + +static INLINE void paint_color_buffer(struct vg_paint *paint, void *buffer) +{ + VGfloat *map = (VGfloat*)buffer; + memcpy(buffer, paint->solid.color, 4 * sizeof(VGfloat)); + map[4] = 0.f; + map[5] = 1.f; + map[6] = 2.f; + map[7] = 4.f; +} + +static INLINE void paint_linear_gradient_buffer(struct vg_paint *paint, void *buffer) +{ + struct vg_context *ctx = paint->base.ctx; + VGfloat *map = (VGfloat*)buffer; + + map[0] = paint->gradient.linear.coords[2] - paint->gradient.linear.coords[0]; + map[1] = paint->gradient.linear.coords[3] - paint->gradient.linear.coords[1]; + map[2] = 1.f / (map[0] * map[0] + map[1] * map[1]); + map[3] = 1.f; + + map[4] = 0.f; + map[5] = 1.f; + map[6] = 2.f; + map[7] = 4.f; + { + struct matrix mat; + struct matrix inv; + matrix_load_identity(&mat); + matrix_translate(&mat, -paint->gradient.linear.coords[0], -paint->gradient.linear.coords[1]); + memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix, + sizeof(struct matrix)); + matrix_invert(&inv); + matrix_mult(&inv, &mat); + memcpy(&mat, &inv, + sizeof(struct matrix)); + + map[8] = mat.m[0]; map[9] = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f; + map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f; + map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f; + } +#if 0 + debug_printf("Coords (%f, %f, %f, %f)\n", + map[0], map[1], map[2], map[3]); +#endif +} + + +static INLINE void paint_radial_gradient_buffer(struct vg_paint *paint, void *buffer) +{ + VGfloat *radialCoords = paint->gradient.radial.vals; + struct vg_context *ctx = paint->base.ctx; + + VGfloat *map = (VGfloat*)buffer; + + map[0] = radialCoords[0] - radialCoords[2]; + map[1] = radialCoords[1] - radialCoords[3]; + map[2] = -map[0] * map[0] - map[1] * map[1] + + radialCoords[4] * radialCoords[4]; + map[3] = 1.f; + + map[4] = 0.f; + map[5] = 1.f; + map[6] = 2.f; + map[7] = 4.f; + + { + struct matrix mat; + struct matrix inv; + matrix_load_identity(&mat); + matrix_translate(&mat, -radialCoords[2], -radialCoords[3]); + memcpy(&inv, &ctx->state.vg.fill_paint_to_user_matrix, + sizeof(struct matrix)); + matrix_invert(&inv); + matrix_mult(&inv, &mat); + memcpy(&mat, &inv, + sizeof(struct matrix)); + + map[8] = mat.m[0]; map[9] = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f; + map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f; + map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f; + } + +#if 0 + debug_printf("Coords (%f, %f, %f, %f)\n", + map[0], map[1], map[2], map[3]); +#endif +} + + +static INLINE void paint_pattern_buffer(struct vg_paint *paint, void *buffer) +{ + struct vg_context *ctx = paint->base.ctx; + + VGfloat *map = (VGfloat *)buffer; + memcpy(map, paint->solid.color, 4 * sizeof(VGfloat)); + + map[4] = 0.f; + map[5] = 1.f; + map[6] = paint->pattern.texture->width[0]; + map[7] = paint->pattern.texture->height[0]; + { + struct matrix mat; + memcpy(&mat, &ctx->state.vg.fill_paint_to_user_matrix, + sizeof(struct matrix)); + matrix_invert(&mat); + { + struct matrix pm; + memcpy(&pm, &ctx->state.vg.path_user_to_surface_matrix, + sizeof(struct matrix)); + matrix_invert(&pm); + matrix_mult(&pm, &mat); + memcpy(&mat, &pm, sizeof(struct matrix)); + } + + map[8] = mat.m[0]; map[9] = mat.m[3]; map[10] = mat.m[6]; map[11] = 0.f; + map[12] = mat.m[1]; map[13] = mat.m[4]; map[14] = mat.m[7]; map[15] = 0.f; + map[16] = mat.m[2]; map[17] = mat.m[5]; map[18] = mat.m[8]; map[19] = 0.f; + } +} + +void paint_set_type(struct vg_paint *paint, VGPaintType type) +{ + paint->type = type; +} + +void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops, + int num) +{ + const VGfloat default_stops[] = {0.0f, 0.0f, 0.0f, 0.0f, 1.0f, + 1.0f, 1.0f, 1.0f, 1.0f, 1.0f}; + VGint i; + const VGint num_stops = num / 5; + VGfloat last_coord; + + paint->gradient.num_stops = num; + if (num) { + free(paint->gradient.ramp_stops); + paint->gradient.ramp_stops = malloc(sizeof(VGfloat)*num); + memcpy(paint->gradient.ramp_stops, stops, sizeof(VGfloat)*num); + } else + return; + + /* stops must be in increasing order. the last stop is 1.0. if the + * first one is bigger than 1 then the whole sequence is invalid*/ + if (stops[0] > 1) { + stops = default_stops; + num = 10; + } + last_coord = stops[0]; + for (i = 1; i < num_stops; ++i) { + VGint idx = 5 * i; + VGfloat coord = stops[idx]; + if (!floatsEqual(last_coord, coord) && coord < last_coord) { + stops = default_stops; + num = 10; + break; + } + last_coord = coord; + } + + create_gradient_data(stops, num / 5, paint->gradient.color_data, + 1024); + + if (paint->gradient.texture) { + pipe_texture_reference(&paint->gradient.texture, NULL); + paint->gradient.texture = 0; + } + + paint->gradient.texture = create_gradient_texture(paint); +} + +void paint_set_colori(struct vg_paint *p, + VGuint rgba) +{ + p->solid.color[0] = ((rgba >> 24) & 0xff) / 255.f; + p->solid.color[1] = ((rgba >> 16) & 0xff) / 255.f; + p->solid.color[2] = ((rgba >> 8) & 0xff) / 255.f; + p->solid.color[3] = ((rgba >> 0) & 0xff) / 255.f; +} + +VGuint paint_colori(struct vg_paint *p) +{ +#define F2B(f) (float_to_ubyte(f)) + + return ((F2B(p->solid.color[0]) << 24) | + (F2B(p->solid.color[1]) << 16) | + (F2B(p->solid.color[2]) << 8) | + (F2B(p->solid.color[3]) << 0)); +#undef F2B +} + +void paint_set_linear_gradient(struct vg_paint *paint, + const VGfloat *coords) +{ + memcpy(paint->gradient.linear.coords, coords, sizeof(VGfloat) * 4); +} + +void paint_set_spread_mode(struct vg_paint *paint, + VGint mode) +{ + paint->gradient.spread = mode; + switch(mode) { + case VG_COLOR_RAMP_SPREAD_PAD: + paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + break; + case VG_COLOR_RAMP_SPREAD_REPEAT: + paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; + break; + case VG_COLOR_RAMP_SPREAD_REFLECT: + paint->gradient.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT; + break; + } +} + +VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint) +{ + return paint->gradient.spread; +} + +void paint_set_radial_gradient(struct vg_paint *paint, + const VGfloat *values) +{ + memcpy(paint->gradient.radial.vals, values, sizeof(VGfloat) * 5); +} + +void paint_set_pattern(struct vg_paint *paint, + struct vg_image *img) +{ + if (paint->pattern.texture) + pipe_texture_reference(&paint->pattern.texture, NULL); + + paint->pattern.texture = 0; + pipe_texture_reference(&paint->pattern.texture, + img->texture); +} + +void paint_set_pattern_tiling(struct vg_paint *paint, + VGTilingMode mode) +{ + paint->pattern.tiling_mode = mode; + + switch(mode) { + case VG_TILE_FILL: + paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_BORDER; + paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_BORDER; + break; + case VG_TILE_PAD: + paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + break; + case VG_TILE_REPEAT: + paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; + paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; + break; + case VG_TILE_REFLECT: + paint->pattern.sampler.wrap_s = PIPE_TEX_WRAP_MIRROR_REPEAT; + paint->pattern.sampler.wrap_t = PIPE_TEX_WRAP_MIRROR_REPEAT; + break; + default: + debug_assert("!Unknown tiling mode"); + } +} + +void paint_get_color(struct vg_paint *paint, + VGfloat *color) +{ + color[0] = paint->solid.color[0]; + color[1] = paint->solid.color[1]; + color[2] = paint->solid.color[2]; + color[3] = paint->solid.color[3]; +} + +void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops, + int num) +{ + memcpy(stops, paint->gradient.ramp_stops, sizeof(VGfloat)*num); +} + +void paint_linear_gradient(struct vg_paint *paint, + VGfloat *coords) +{ + memcpy(coords, paint->gradient.linear.coords, sizeof(VGfloat)*4); +} + +void paint_radial_gradient(struct vg_paint *paint, + VGfloat *coords) +{ + memcpy(coords, paint->gradient.radial.vals, sizeof(VGfloat)*5); +} + +int paint_num_ramp_stops(struct vg_paint *paint) +{ + return paint->gradient.num_stops; +} + +VGPaintType paint_type(struct vg_paint *paint) +{ + return paint->type; +} + +void paint_set_coloriv(struct vg_paint *paint, + const VGint *color) +{ + paint->solid.color[0] = color[0]; + paint->solid.color[1] = color[1]; + paint->solid.color[2] = color[2]; + paint->solid.color[3] = color[3]; + + paint->solid.colori[0] = color[0]; + paint->solid.colori[1] = color[1]; + paint->solid.colori[2] = color[2]; + paint->solid.colori[3] = color[3]; +} + +void paint_get_coloriv(struct vg_paint *paint, + VGint *color) +{ + color[0] = paint->solid.colori[0]; + color[1] = paint->solid.colori[1]; + color[2] = paint->solid.colori[2]; + color[3] = paint->solid.colori[3]; +} + +void paint_set_color_ramp_premultiplied(struct vg_paint *paint, + VGboolean set) +{ + paint->gradient.color_ramps_premultiplied = set; +} + +VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint) +{ + return paint->gradient.color_ramps_premultiplied; +} + +void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops, + int num) +{ + if (num) { + free(paint->gradient.ramp_stopsi); + paint->gradient.ramp_stopsi = malloc(sizeof(VGint)*num); + memcpy(paint->gradient.ramp_stopsi, stops, sizeof(VGint)*num); + } +} + +void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops, + int num) +{ + memcpy(stops, paint->gradient.ramp_stopsi, sizeof(VGint)*num); +} + +void paint_set_linear_gradienti(struct vg_paint *paint, + const VGint *coords) +{ + memcpy(paint->gradient.linear.coordsi, coords, sizeof(VGint) * 4); +} + +void paint_linear_gradienti(struct vg_paint *paint, + VGint *coords) +{ + memcpy(coords, paint->gradient.linear.coordsi, sizeof(VGint)*4); +} + +void paint_set_radial_gradienti(struct vg_paint *paint, + const VGint *values) +{ + memcpy(paint->gradient.radial.valsi, values, sizeof(VGint) * 5); +} + +void paint_radial_gradienti(struct vg_paint *paint, + VGint *coords) +{ + memcpy(coords, paint->gradient.radial.valsi, sizeof(VGint)*5); +} + +VGTilingMode paint_pattern_tiling(struct vg_paint *paint) +{ + return paint->pattern.tiling_mode; +} + +VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers, + struct pipe_texture **textures) +{ + struct vg_context *ctx = vg_current_context(); + + switch(paint->type) { + case VG_PAINT_TYPE_LINEAR_GRADIENT: + case VG_PAINT_TYPE_RADIAL_GRADIENT: { + if (paint->gradient.texture) { + paint->gradient.sampler.min_img_filter = image_sampler_filter(ctx); + paint->gradient.sampler.mag_img_filter = image_sampler_filter(ctx); + samplers[0] = &paint->gradient.sampler; + textures[0] = paint->gradient.texture; + return 1; + } + } + break; + case VG_PAINT_TYPE_PATTERN: { + memcpy(paint->pattern.sampler.border_color, + ctx->state.vg.tile_fill_color, + sizeof(VGfloat) * 4); + paint->pattern.sampler.min_img_filter = image_sampler_filter(ctx); + paint->pattern.sampler.mag_img_filter = image_sampler_filter(ctx); + samplers[0] = &paint->pattern.sampler; + textures[0] = paint->pattern.texture; + return 1; + } + break; + default: + samplers[0] = &paint->pattern.sampler; /* dummy */ + textures[0] = 0; + return 0; + break; + } + return 0; +} + +void paint_resolve_type(struct vg_paint *paint) +{ + if (paint->type == VG_PAINT_TYPE_PATTERN && + !paint->pattern.texture) { + paint->type = VG_PAINT_TYPE_COLOR; + } +} + +VGint paint_constant_buffer_size(struct vg_paint *paint) +{ + switch(paint->type) { + case VG_PAINT_TYPE_COLOR: + return 8 * sizeof(VGfloat);/*4 color + 4 constants (0.f,1.f,2.f,4.f)*/ + break; + case VG_PAINT_TYPE_LINEAR_GRADIENT: + return 20 * sizeof(VGfloat); + break; + case VG_PAINT_TYPE_RADIAL_GRADIENT: + return 20 * sizeof(VGfloat); + break; + case VG_PAINT_TYPE_PATTERN: + return 20 * sizeof(VGfloat); + break; + default: + debug_printf("Uknown paint type: %d\n", paint->type); + } + + return 0; +} + +void paint_fill_constant_buffer(struct vg_paint *paint, + void *buffer) +{ + switch(paint->type) { + case VG_PAINT_TYPE_COLOR: + paint_color_buffer(paint, buffer); + break; + case VG_PAINT_TYPE_LINEAR_GRADIENT: + paint_linear_gradient_buffer(paint, buffer); + break; + case VG_PAINT_TYPE_RADIAL_GRADIENT: + paint_radial_gradient_buffer(paint, buffer); + break; + case VG_PAINT_TYPE_PATTERN: + paint_pattern_buffer(paint, buffer); + break; + + default: + abort(); + } +} diff --git a/src/gallium/state_trackers/vega/paint.h b/src/gallium/state_trackers/vega/paint.h new file mode 100644 index 00000000000..999b5c167ca --- /dev/null +++ b/src/gallium/state_trackers/vega/paint.h @@ -0,0 +1,118 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PAINT_H +#define PAINT_H + +#include "vg_context.h" + +#include "VG/openvg.h" +#include "pipe/p_state.h" + +struct vg_paint; +struct vg_image; +struct pipe_sampler_state; +struct pipe_texture; + +struct vg_paint *paint_create(struct vg_context *ctx); +void paint_destroy(struct vg_paint *paint); + +void paint_set_color(struct vg_paint *paint, + const VGfloat *color); +void paint_get_color(struct vg_paint *paint, + VGfloat *color); + +void paint_set_coloriv(struct vg_paint *paint, + const VGint *color); +void paint_get_coloriv(struct vg_paint *paint, + VGint *color); + +void paint_set_colori(struct vg_paint *paint, + VGuint rgba); + +VGuint paint_colori(struct vg_paint *paint); + +void paint_set_type(struct vg_paint *paint, VGPaintType type); +VGPaintType paint_type(struct vg_paint *paint); +void paint_resolve_type(struct vg_paint *paint); + +void paint_set_linear_gradient(struct vg_paint *paint, + const VGfloat *coords); +void paint_linear_gradient(struct vg_paint *paint, + VGfloat *coords); +void paint_set_linear_gradienti(struct vg_paint *paint, + const VGint *coords); +void paint_linear_gradienti(struct vg_paint *paint, + VGint *coords); + + +void paint_set_radial_gradient(struct vg_paint *paint, + const VGfloat *values); +void paint_radial_gradient(struct vg_paint *paint, + VGfloat *coords); +void paint_set_radial_gradienti(struct vg_paint *paint, + const VGint *values); +void paint_radial_gradienti(struct vg_paint *paint, + VGint *coords); + + +void paint_set_ramp_stops(struct vg_paint *paint, const VGfloat *stops, + int num); +void paint_ramp_stops(struct vg_paint *paint, VGfloat *stops, + int num); + +void paint_set_ramp_stopsi(struct vg_paint *paint, const VGint *stops, + int num); +void paint_ramp_stopsi(struct vg_paint *paint, VGint *stops, + int num); + +int paint_num_ramp_stops(struct vg_paint *paint); + +void paint_set_spread_mode(struct vg_paint *paint, + VGint mode); +VGColorRampSpreadMode paint_spread_mode(struct vg_paint *paint); + + +void paint_set_pattern(struct vg_paint *paint, + struct vg_image *img); +void paint_set_pattern_tiling(struct vg_paint *paint, + VGTilingMode mode); +VGTilingMode paint_pattern_tiling(struct vg_paint *paint); + +void paint_set_color_ramp_premultiplied(struct vg_paint *paint, + VGboolean set); +VGboolean paint_color_ramp_premultiplied(struct vg_paint *paint); + + +VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **samplers, + struct pipe_texture **textures); + +VGint paint_constant_buffer_size(struct vg_paint *paint); +void paint_fill_constant_buffer(struct vg_paint *paint, + void *buffer); + + +#endif diff --git a/src/gallium/state_trackers/vega/path.c b/src/gallium/state_trackers/vega/path.c new file mode 100644 index 00000000000..4fc23a7a278 --- /dev/null +++ b/src/gallium/state_trackers/vega/path.c @@ -0,0 +1,2034 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "path.h" + +#include "stroker.h" +#include "polygon.h" +#include "bezier.h" +#include "matrix.h" +#include "vg_context.h" +#include "util_array.h" +#include "arc.h" +#include "path_utils.h" +#include "paint.h" +#include "shader.h" + +#include "util/u_memory.h" + +#include <assert.h> + +#define DEBUG_PATH 0 + +struct path { + struct vg_object base; + VGbitfield caps; + VGboolean dirty; + VGboolean dirty_stroke; + + VGPathDatatype datatype; + + VGfloat scale; + VGfloat bias; + + VGint num_segments; + + struct array * segments; + struct array * control_points; + + struct { + struct polygon_array polygon_array; + struct matrix matrix; + } fill_polys; + + struct { + struct path *path; + struct matrix matrix; + VGfloat stroke_width; + VGfloat miter_limit; + VGCapStyle cap_style; + VGJoinStyle join_style; + } stroked; +}; + + +static INLINE void data_at(void **data, + struct path *p, + VGint start, VGint count, + VGfloat *out) +{ + VGPathDatatype dt = p->datatype; + VGint i; + VGint end = start + count; + VGfloat *itr = out; + + switch(dt) { + case VG_PATH_DATATYPE_S_8: { + VGbyte **bdata = (VGbyte **)data; + for (i = start; i < end; ++i) { + *itr = (*bdata)[i]; + ++itr; + } + *bdata += count; + } + break; + case VG_PATH_DATATYPE_S_16: { + VGshort **bdata = (VGshort **)data; + for (i = start; i < end; ++i) { + *itr = (*bdata)[i]; + ++itr; + } + *bdata += count; + } + break; + case VG_PATH_DATATYPE_S_32: { + VGint **bdata = (VGint **)data; + for (i = start; i < end; ++i) { + *itr = (*bdata)[i]; + ++itr; + } + *bdata += count; + } + break; + case VG_PATH_DATATYPE_F: { + VGfloat **fdata = (VGfloat **)data; + for (i = start; i < end; ++i) { + *itr = (*fdata)[i]; + ++itr; + } + *fdata += count; + } + break; + default: + debug_assert(!"Unknown path datatype!"); + } +} + + +void vg_float_to_datatype(VGPathDatatype datatype, + VGubyte *common_data, + const VGfloat *data, + VGint num_coords) +{ + VGint i; + switch(datatype) { + case VG_PATH_DATATYPE_S_8: { + for (i = 0; i < num_coords; ++i) { + common_data[i] = (VGubyte)data[i]; + } + } + break; + case VG_PATH_DATATYPE_S_16: { + VGshort *buf = (VGshort*)common_data; + for (i = 0; i < num_coords; ++i) { + buf[i] = (VGshort)data[i]; + } + } + break; + case VG_PATH_DATATYPE_S_32: { + VGint *buf = (VGint*)common_data; + for (i = 0; i < num_coords; ++i) { + buf[i] = (VGint)data[i]; + } + } + break; + case VG_PATH_DATATYPE_F: { + memcpy(common_data, data, sizeof(VGfloat) * num_coords); + } + break; + default: + debug_assert(!"Unknown path datatype!"); + } +} + +static void coords_adjust_by_scale_bias(struct path *p, + void *pdata, VGint num_coords, + VGfloat scale, VGfloat bias, + VGPathDatatype datatype) +{ + VGfloat data[8]; + void *coords = (VGfloat *)pdata; + VGubyte *common_data = (VGubyte *)pdata; + VGint size_dst = size_for_datatype(datatype); + VGint i; + + for (i = 0; i < num_coords; ++i) { + data_at(&coords, p, 0, 1, data); + data[0] = data[0] * scale + bias; + vg_float_to_datatype(datatype, common_data, data, 1); + common_data += size_dst; + } +} + +struct path * path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias, + VGint segmentCapacityHint, + VGint coordCapacityHint, + VGbitfield capabilities) +{ + struct path *path = CALLOC_STRUCT(path); + + vg_init_object(&path->base, vg_current_context(), VG_OBJECT_PATH); + path->caps = capabilities & VG_PATH_CAPABILITY_ALL; + vg_context_add_object(vg_current_context(), VG_OBJECT_PATH, path); + + path->datatype = dt; + path->scale = scale; + path->bias = bias; + + path->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8)); + path->control_points = array_create(size_for_datatype(dt)); + + path->dirty = VG_TRUE; + path->dirty_stroke = VG_TRUE; + + return path; +} + +void path_destroy(struct path *p) +{ + vg_context_remove_object(vg_current_context(), VG_OBJECT_PATH, p); + + array_destroy(p->segments); + array_destroy(p->control_points); + array_destroy(p->fill_polys.polygon_array.array); + + if (p->stroked.path) + path_destroy(p->stroked.path); + + free(p); +} + +VGbitfield path_capabilities(struct path *p) +{ + return p->caps; +} + +void path_set_capabilities(struct path *p, VGbitfield bf) +{ + p->caps = (bf & VG_PATH_CAPABILITY_ALL); +} + +void path_append_data(struct path *p, + VGint numSegments, + const VGubyte * pathSegments, + const void * pathData) +{ + VGint old_segments = p->num_segments; + VGint num_new_coords = num_elements_for_segments(pathSegments, numSegments); + array_append_data(p->segments, pathSegments, numSegments); + array_append_data(p->control_points, pathData, num_new_coords); + + p->num_segments += numSegments; + if (!floatsEqual(p->scale, 1.f) || !floatsEqual(p->bias, 0.f)) { + VGubyte *coords = (VGubyte*)p->control_points->data; + coords_adjust_by_scale_bias(p, + coords + old_segments * p->control_points->datatype_size, + num_new_coords, + p->scale, p->bias, p->datatype); + } + p->dirty = VG_TRUE; + p->dirty_stroke = VG_TRUE; +} + +VGint path_num_segments(struct path *p) +{ + return p->num_segments; +} + +static INLINE void map_if_relative(VGfloat ox, VGfloat oy, + VGboolean relative, + VGfloat *x, VGfloat *y) +{ + if (relative) { + if (x) + *x += ox; + if (y) + *y += oy; + } +} + +static INLINE void close_polygon(struct polygon *current, + VGfloat sx, VGfloat sy, + VGfloat ox, VGfloat oy, + struct matrix *matrix) +{ + if (!floatsEqual(sx, ox) || + !floatsEqual(sy, oy)) { + VGfloat x0 = sx; + VGfloat y0 = sy; + matrix_map_point(matrix, x0, y0, &x0, &y0); + polygon_vertex_append(current, x0, y0); + } +} + +static void convert_path(struct path *p, + VGPathDatatype to, + void *dst, + VGint num_coords) +{ + VGfloat data[8]; + void *coords = (VGfloat *)p->control_points->data; + VGubyte *common_data = (VGubyte *)dst; + VGint size_dst = size_for_datatype(to); + VGint i; + + for (i = 0; i < num_coords; ++i) { + data_at(&coords, p, 0, 1, data); + vg_float_to_datatype(to, common_data, data, 1); + common_data += size_dst; + } +} + + +static void polygon_array_calculate_bounds( struct polygon_array *polyarray ) +{ + struct array *polys = polyarray->array; + VGfloat min_x, max_x; + VGfloat min_y, max_y; + VGfloat bounds[4]; + unsigned i; + + assert(polys); + assert(polys->num_elements); + polygon_bounding_rect((((struct polygon**)polys->data)[0]), bounds); + min_x = bounds[0]; + min_y = bounds[1]; + max_x = bounds[0] + bounds[2]; + max_y = bounds[1] + bounds[3]; + for (i = 1; i < polys->num_elements; ++i) { + struct polygon *p = (((struct polygon**)polys->data)[i]); + polygon_bounding_rect(p, bounds); + min_x = MIN2(min_x, bounds[0]); + min_y = MIN2(min_y, bounds[1]); + max_x = MAX2(max_x, bounds[0] + bounds[2]); + max_y = MAX2(max_y, bounds[1] + bounds[3]); + } + + polyarray->min_x = min_x; + polyarray->min_y = min_y; + polyarray->max_x = max_x; + polyarray->max_y = max_y; +} + + +static struct polygon_array * path_get_fill_polygons(struct path *p, struct matrix *matrix) +{ + VGint i; + struct polygon *current = 0; + VGfloat sx, sy, px, py, ox, oy; + VGfloat x0, y0, x1, y1, x2, y2, x3, y3; + VGfloat data[8]; + void *coords = (VGfloat *)p->control_points->data; + struct array *array; + + if (p->fill_polys.polygon_array.array) + { + if (memcmp( &p->fill_polys.matrix, + matrix, + sizeof *matrix ) == 0 && p->dirty == VG_FALSE) + { + return &p->fill_polys.polygon_array; + } + else { + array_destroy( p->fill_polys.polygon_array.array ); + p->fill_polys.polygon_array.array = NULL; + } + } + + array = array_create(sizeof(struct array*)); + + sx = sy = px = py = ox = oy = 0.f; + + current = polygon_create(32); + + for (i = 0; i < p->num_segments; ++i) { + VGubyte segment = ((VGubyte*)(p->segments->data))[i]; + VGint command = SEGMENT_COMMAND(segment); + VGboolean relative = SEGMENT_ABS_REL(segment); + + switch(command) { + case VG_CLOSE_PATH: + close_polygon(current, sx, sy, ox, oy, matrix); + ox = sx; + oy = sy; + break; + case VG_MOVE_TO: + if (current && polygon_vertex_count(current) > 0) { + /* add polygon */ + close_polygon(current, sx, sy, ox, oy, matrix); + array_append_data(array, ¤t, 1); + current = polygon_create(32); + } + data_at(&coords, p, 0, 2, data); + x0 = data[0]; + y0 = data[1]; + map_if_relative(ox, oy, relative, &x0, &y0); + sx = x0; + sy = y0; + ox = x0; + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + polygon_vertex_append(current, x0, y0); + break; + case VG_LINE_TO: + data_at(&coords, p, 0, 2, data); + x0 = data[0]; + y0 = data[1]; + map_if_relative(ox, oy, relative, &x0, &y0); + ox = x0; + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + polygon_vertex_append(current, x0, y0); + break; + case VG_HLINE_TO: + data_at(&coords, p, 0, 1, data); + x0 = data[0]; + y0 = oy; + map_if_relative(ox, oy, relative, &x0, 0); + ox = x0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + polygon_vertex_append(current, x0, y0); + break; + case VG_VLINE_TO: + data_at(&coords, p, 0, 1, data); + x0 = ox; + y0 = data[0]; + map_if_relative(ox, oy, relative, 0, &y0); + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + polygon_vertex_append(current, x0, y0); + break; + case VG_CUBIC_TO: { + struct bezier bezier; + data_at(&coords, p, 0, 6, data); + x0 = ox; + y0 = oy; + x1 = data[0]; + y1 = data[1]; + x2 = data[2]; + y2 = data[3]; + x3 = data[4]; + y3 = data[5]; + map_if_relative(ox, oy, relative, &x1, &y1); + map_if_relative(ox, oy, relative, &x2, &y2); + map_if_relative(ox, oy, relative, &x3, &y3); + ox = x3; + oy = y3; + px = x2; + py = y2; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + bezier_init(&bezier, x0, y0, x1, y1, + x2, y2, x3, y3); + bezier_add_to_polygon(&bezier, current); + } + break; + case VG_QUAD_TO: { + struct bezier bezier; + data_at(&coords, p, 0, 4, data); + x0 = ox; + y0 = oy; + x1 = data[0]; + y1 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(ox, oy, relative, &x1, &y1); + map_if_relative(ox, oy, relative, &x3, &y3); + px = x1; + py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + ox = x3; + oy = y3; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + bezier_init(&bezier, x0, y0, x1, y1, + x2, y2, x3, y3); + bezier_add_to_polygon(&bezier, current); + } + break; + case VG_SQUAD_TO: { + struct bezier bezier; + data_at(&coords, p, 0, 2, data); + x0 = ox; + y0 = oy; + x1 = 2*ox-px; + y1 = 2*oy-py; + x3 = data[0]; + y3 = data[1]; + map_if_relative(ox, oy, relative, &x3, &y3); + px = x1; + py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + ox = x3; + oy = y3; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + bezier_init(&bezier, x0, y0, x1, y1, + x2, y2, x3, y3); + bezier_add_to_polygon(&bezier, current); + } + break; + case VG_SCUBIC_TO: { + struct bezier bezier; + data_at(&coords, p, 0, 4, data); + x0 = ox; + y0 = oy; + x1 = 2*ox-px; + y1 = 2*oy-py; + x2 = data[0]; + y2 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(ox, oy, relative, &x2, &y2); + map_if_relative(ox, oy, relative, &x3, &y3); + ox = x3; + oy = y3; + px = x2; + py = y2; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + bezier_init(&bezier, x0, y0, x1, y1, + x2, y2, x3, y3); + bezier_add_to_polygon(&bezier, current); + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + VGfloat rh, rv, rot; + struct arc arc; + + data_at(&coords, p, 0, 5, data); + x0 = ox; + y0 = oy; + rh = data[0]; + rv = data[1]; + rot = data[2]; + x1 = data[3]; + y1 = data[4]; + map_if_relative(ox, oy, relative, &x1, &y1); +#if 0 + debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n", + x0, y0, x1, y1, rh, rv, rot); +#endif + arc_init(&arc, command, x0, y0, x1, y1, + rh, rv, rot); + arc_add_to_polygon(&arc, current, + matrix); + ox = x1; + oy = y1; + px = x1; + py = y1; + } + break; + default: + abort(); + assert(!"Unknown segment!"); + } + } + if (current) { + if (polygon_vertex_count(current) > 0) { + close_polygon(current, sx, sy, ox, oy, matrix); + array_append_data(array, ¤t, 1); + } else + polygon_destroy(current); + } + + p->fill_polys.polygon_array.array = array; + p->fill_polys.matrix = *matrix; + + polygon_array_calculate_bounds( &p->fill_polys.polygon_array ); + + p->dirty = VG_FALSE; + + return &p->fill_polys.polygon_array; +} + +VGbyte path_datatype_size(struct path *p) +{ + return size_for_datatype(p->datatype); +} + +VGPathDatatype path_datatype(struct path *p) +{ + return p->datatype; +} + +VGfloat path_scale(struct path *p) +{ + return p->scale; +} + +VGfloat path_bias(struct path *p) +{ + return p->bias; +} + +VGint path_num_coords(struct path *p) +{ + return num_elements_for_segments((VGubyte*)p->segments->data, + p->num_segments); +} + +void path_modify_coords(struct path *p, + VGint startIndex, + VGint numSegments, + const void * pathData) +{ + VGubyte *segments = (VGubyte*)(p->segments->data); + VGint count = num_elements_for_segments(&segments[startIndex], numSegments); + VGint start_cp = num_elements_for_segments(segments, startIndex); + + array_change_data(p->control_points, pathData, start_cp, count); + coords_adjust_by_scale_bias(p, + ((VGubyte*)p->control_points->data) + + (startIndex * p->control_points->datatype_size), + path_num_coords(p), + p->scale, p->bias, p->datatype); + p->dirty = VG_TRUE; + p->dirty_stroke = VG_TRUE; +} + +void path_for_each_segment(struct path *path, + path_for_each_cb cb, + void *user_data) +{ + VGint i; + struct path_for_each_data p; + VGfloat data[8]; + void *coords = (VGfloat *)path->control_points->data; + + p.coords = data; + p.sx = p.sy = p.px = p.py = p.ox = p.oy = 0.f; + p.user_data = user_data; + + for (i = 0; i < path->num_segments; ++i) { + VGint command; + VGboolean relative; + + p.segment = ((VGubyte*)(path->segments->data))[i]; + command = SEGMENT_COMMAND(p.segment); + relative = SEGMENT_ABS_REL(p.segment); + + switch(command) { + case VG_CLOSE_PATH: + cb(path, &p); + break; + case VG_MOVE_TO: + data_at(&coords, path, 0, 2, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + cb(path, &p); + p.sx = data[0]; + p.sy = data[1]; + p.ox = data[0]; + p.oy = data[1]; + p.px = data[0]; + p.py = data[1]; + break; + case VG_LINE_TO: + data_at(&coords, path, 0, 2, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + cb(path, &p); + p.ox = data[0]; + p.oy = data[1]; + p.px = data[0]; + p.py = data[1]; + break; + case VG_HLINE_TO: + data_at(&coords, path, 0, 1, data); + map_if_relative(p.ox, p.oy, relative, &data[0], 0); + p.segment = VG_LINE_TO; + data[1] = p.oy; + cb(path, &p); + p.ox = data[0]; + p.oy = data[1]; + p.px = data[0]; + p.py = data[1]; + break; + case VG_VLINE_TO: + data_at(&coords, path, 0, 1, data); + map_if_relative(p.ox, p.oy, relative, 0, &data[0]); + p.segment = VG_LINE_TO; + data[1] = data[0]; + data[0] = p.ox; + cb(path, &p); + p.ox = data[0]; + p.oy = data[1]; + p.px = data[0]; + p.py = data[1]; + break; + case VG_CUBIC_TO: { + data_at(&coords, path, 0, 6, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]); + map_if_relative(p.ox, p.oy, relative, &data[4], &data[5]); + cb(path, &p); + p.px = data[2]; + p.py = data[3]; + p.ox = data[4]; + p.oy = data[5]; + } + break; + case VG_QUAD_TO: { + data_at(&coords, path, 0, 4, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]); + cb(path, &p); + p.px = data[0]; + p.py = data[1]; + p.ox = data[2]; + p.oy = data[3]; + } + break; + case VG_SQUAD_TO: { + data_at(&coords, path, 0, 2, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + cb(path, &p); + p.px = 2*p.ox-p.px; + p.py = 2*p.oy-p.py; + p.ox = data[2]; + p.oy = data[3]; + } + break; + case VG_SCUBIC_TO: { + data_at(&coords, path, 0, 4, data); + map_if_relative(p.ox, p.oy, relative, &data[0], &data[1]); + map_if_relative(p.ox, p.oy, relative, &data[2], &data[3]); + cb(path, &p); + p.px = data[0]; + p.py = data[1]; + p.ox = data[2]; + p.oy = data[3]; + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + data_at(&coords, path, 0, 5, data); + map_if_relative(p.ox, p.oy, relative, &data[3], &data[4]); +#if 0 + debug_printf("------- ARC (%f, %f), (%f, %f) %f, %f, %f\n", + p.ox, p.oy, data[3], data[4], data[0], data[1], data[2]); +#endif + cb(path, &p); + p.ox = data[3]; + p.oy = data[4]; + p.px = data[3]; + p.py = data[4]; + } + break; + default: + abort(); + assert(!"Unknown segment!"); + } + } +} + +struct transform_data { + struct array *segments; + struct array *coords; + + struct matrix *matrix; + + VGPathDatatype datatype; +}; + +static VGboolean transform_cb(struct path *p, + struct path_for_each_data *pd) +{ + struct transform_data *td = (struct transform_data *)pd->user_data; + VGint num_coords = num_elements_for_segments(&pd->segment, 1); + VGubyte segment = SEGMENT_COMMAND(pd->segment);/* abs bit is 0 */ + VGfloat data[8]; + VGubyte common_data[sizeof(VGfloat)*8]; + + memcpy(data, pd->coords, sizeof(VGfloat) * num_coords); + + switch(segment) { + case VG_CLOSE_PATH: + break; + case VG_MOVE_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + break; + case VG_LINE_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + break; + case VG_HLINE_TO: + case VG_VLINE_TO: + assert(0); + break; + case VG_QUAD_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + matrix_map_point(td->matrix, + data[2], data[3], &data[2], &data[3]); + break; + case VG_CUBIC_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + matrix_map_point(td->matrix, + data[2], data[3], &data[2], &data[3]); + matrix_map_point(td->matrix, + data[4], data[5], &data[4], &data[5]); + break; + case VG_SQUAD_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + break; + case VG_SCUBIC_TO: + matrix_map_point(td->matrix, + data[0], data[1], &data[0], &data[1]); + matrix_map_point(td->matrix, + data[2], data[3], &data[2], &data[3]); + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + struct arc arc; + struct path *path = path_create(td->datatype, + 1, 0, 0, 0, VG_PATH_CAPABILITY_ALL); + arc_init(&arc, segment, + pd->ox, pd->oy, data[3], data[4], + data[0], data[1], data[2]); + + arc_to_path(&arc, path, td->matrix); + + num_coords = path_num_coords(path); + + array_append_data(td->segments, path->segments->data, + path->num_segments); + array_append_data(td->coords, path->control_points->data, + num_coords); + path_destroy(path); + + return VG_TRUE; + } + break; + default: + break; + } + + vg_float_to_datatype(td->datatype, common_data, data, num_coords); + + array_append_data(td->segments, &pd->segment, 1); + array_append_data(td->coords, common_data, num_coords); + return VG_TRUE; +} + +void path_transform(struct path *dst, struct path *src) +{ + struct transform_data data; + struct vg_context *ctx = dst->base.ctx; + + data.segments = dst->segments; + data.coords = dst->control_points; + data.matrix = &ctx->state.vg.path_user_to_surface_matrix; + data.datatype = dst->datatype; + + path_for_each_segment(src, transform_cb, (void*)&data); + + dst->num_segments = dst->segments->num_elements; + dst->dirty = VG_TRUE; + dst->dirty_stroke = VG_TRUE; +} + +void path_append_path(struct path *dst, + struct path *src) +{ + VGint num_coords = path_num_coords(src); + void *dst_data = malloc(size_for_datatype(dst->datatype) * num_coords); + array_append_data(dst->segments, + src->segments->data, + src->num_segments); + convert_path(src, dst->datatype, + dst_data, num_coords); + array_append_data(dst->control_points, + dst_data, + num_coords); + free(dst_data); + + dst->num_segments += src->num_segments; + dst->dirty = VG_TRUE; + dst->dirty_stroke = VG_TRUE; +} + +static INLINE VGboolean is_segment_arc(VGubyte segment) +{ + VGubyte scommand = SEGMENT_COMMAND(segment); + return (scommand == VG_SCCWARC_TO || + scommand == VG_SCWARC_TO || + scommand == VG_LCCWARC_TO || + scommand == VG_LCWARC_TO); +} + +struct path_iter_data { + struct path *path; + VGubyte segment; + void *coords; + VGfloat px, py, ox, oy, sx, sy; +}; +static INLINE VGubyte normalize_coords(struct path_iter_data *pd, + VGint *num_coords, + VGfloat *data) +{ + VGint command = SEGMENT_COMMAND(pd->segment); + VGboolean relative = SEGMENT_ABS_REL(pd->segment); + + switch(command) { + case VG_CLOSE_PATH: + *num_coords = 0; + pd->ox = pd->sx; + pd->oy = pd->sy; + return VG_CLOSE_PATH; + break; + case VG_MOVE_TO: + data_at(&pd->coords, pd->path, 0, 2, data); + map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]); + pd->sx = data[0]; + pd->sy = data[1]; + pd->ox = data[0]; + pd->oy = data[1]; + pd->px = data[0]; + pd->py = data[1]; + *num_coords = 2; + return VG_MOVE_TO_ABS; + break; + case VG_LINE_TO: + data_at(&pd->coords, pd->path, 0, 2, data); + map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]); + pd->ox = data[0]; + pd->oy = data[1]; + pd->px = data[0]; + pd->py = data[1]; + *num_coords = 2; + return VG_LINE_TO_ABS; + break; + case VG_HLINE_TO: + data_at(&pd->coords, pd->path, 0, 1, data); + map_if_relative(pd->ox, pd->oy, relative, &data[0], 0); + data[1] = pd->oy; + pd->ox = data[0]; + pd->oy = data[1]; + pd->px = data[0]; + pd->py = data[1]; + *num_coords = 2; + return VG_LINE_TO_ABS; + break; + case VG_VLINE_TO: + data_at(&pd->coords, pd->path, 0, 1, data); + map_if_relative(pd->ox, pd->oy, relative, 0, &data[0]); + data[1] = data[0]; + data[0] = pd->ox; + pd->ox = data[0]; + pd->oy = data[1]; + pd->px = data[0]; + pd->py = data[1]; + *num_coords = 2; + return VG_LINE_TO_ABS; + break; + case VG_CUBIC_TO: { + data_at(&pd->coords, pd->path, 0, 6, data); + map_if_relative(pd->ox, pd->oy, relative, &data[0], &data[1]); + map_if_relative(pd->ox, pd->oy, relative, &data[2], &data[3]); + map_if_relative(pd->ox, pd->oy, relative, &data[4], &data[5]); + pd->px = data[2]; + pd->py = data[3]; + pd->ox = data[4]; + pd->oy = data[5]; + *num_coords = 6; + return VG_CUBIC_TO_ABS; + } + break; + case VG_QUAD_TO: { + VGfloat x0, y0, x1, y1, x2, y2, x3, y3; + data_at(&pd->coords, pd->path, 0, 4, data); + x0 = pd->ox; + y0 = pd->oy; + x1 = data[0]; + y1 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(pd->ox, pd->oy, relative, &x1, &y1); + map_if_relative(pd->ox, pd->oy, relative, &x3, &y3); + pd->px = x1; + pd->py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + pd->ox = x3; + pd->oy = y3; + data[0] = x1; + data[1] = y1; + data[2] = x2; + data[3] = y2; + data[4] = x3; + data[5] = y3; + *num_coords = 6; + return VG_CUBIC_TO_ABS; + } + break; + case VG_SQUAD_TO: { + VGfloat x0, y0, x1, y1, x2, y2, x3, y3; + data_at(&pd->coords, pd->path, 0, 2, data); + x0 = pd->ox; + y0 = pd->oy; + x1 = 2 * pd->ox - pd->px; + y1 = 2 * pd->oy - pd->py; + x3 = data[0]; + y3 = data[1]; + map_if_relative(pd->ox, pd->oy, relative, &x3, &y3); + pd->px = x1; + pd->py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + pd->ox = x3; + pd->oy = y3; + data[0] = x1; + data[1] = y1; + data[2] = x2; + data[3] = y2; + data[4] = x3; + data[5] = y3; + *num_coords = 6; + return VG_CUBIC_TO_ABS; + } + break; + case VG_SCUBIC_TO: { + VGfloat x0, y0, x1, y1, x2, y2, x3, y3; + data_at(&pd->coords, pd->path, 0, 4, data); + x0 = pd->ox; + y0 = pd->oy; + x1 = 2*pd->ox-pd->px; + y1 = 2*pd->oy-pd->py; + x2 = data[0]; + y2 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(pd->ox, pd->oy, relative, &x2, &y2); + map_if_relative(pd->ox, pd->oy, relative, &x3, &y3); + pd->ox = x3; + pd->oy = y3; + pd->px = x2; + pd->py = y2; + data[0] = x1; + data[1] = y1; + data[2] = x2; + data[3] = y2; + data[4] = x3; + data[5] = y3; + *num_coords = 6; + return VG_CUBIC_TO_ABS; + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + data_at(&pd->coords, pd->path, 0, 5, data); + map_if_relative(pd->ox, pd->oy, relative, &data[3], &data[4]); + pd->ox = data[3]; + pd->oy = data[4]; + pd->px = data[3]; + pd->py = data[4]; + *num_coords = 5; + return command | VG_ABSOLUTE; + } + break; + default: + abort(); + assert(!"Unknown segment!"); + } +} + +static void linearly_interpolate(VGfloat *result, + const VGfloat *start, + const VGfloat *end, + VGfloat amount, + VGint number) +{ + VGint i; + for (i = 0; i < number; ++i) { + result[i] = start[i] + (end[i] - start[i]) * amount; + } +} + +VGboolean path_interpolate(struct path *dst, + struct path *start, struct path *end, + VGfloat amount) +{ + /* temporary path that we can discard if it will turn + * out that start is not compatible with end */ + struct path *res_path = path_create(dst->datatype, + 1.0, 0.0, + 0, 0, dst->caps); + VGint i; + VGfloat start_coords[8]; + VGfloat end_coords[8]; + VGfloat results[8]; + VGubyte common_data[sizeof(VGfloat)*8]; + struct path_iter_data start_iter, end_iter; + + memset(&start_iter, 0, sizeof(struct path_iter_data)); + memset(&end_iter, 0, sizeof(struct path_iter_data)); + + start_iter.path = start; + start_iter.coords = start->control_points->data; + end_iter.path = end; + end_iter.coords = end->control_points->data; + + for (i = 0; i < start->num_segments; ++i) { + VGubyte segment; + VGubyte ssegment, esegment; + VGint snum_coords, enum_coords; + start_iter.segment = ((VGubyte*)(start->segments->data))[i]; + end_iter.segment = ((VGubyte*)(end->segments->data))[i]; + + ssegment = normalize_coords(&start_iter, &snum_coords, + start_coords); + esegment = normalize_coords(&end_iter, &enum_coords, + end_coords); + + if (is_segment_arc(ssegment)) { + if (!is_segment_arc(esegment)) { + path_destroy(res_path); + return VG_FALSE; + } + if (amount > 0.5) + segment = esegment; + else + segment = ssegment; + } else if (is_segment_arc(esegment)) { + path_destroy(res_path); + return VG_FALSE; + } + else if (ssegment != esegment) { + path_destroy(res_path); + return VG_FALSE; + } + else + segment = ssegment; + + linearly_interpolate(results, start_coords, end_coords, + amount, snum_coords); + vg_float_to_datatype(dst->datatype, common_data, results, snum_coords); + path_append_data(res_path, 1, &segment, common_data); + } + + path_append_path(dst, res_path); + path_destroy(res_path); + + dst->dirty = VG_TRUE; + dst->dirty_stroke = VG_TRUE; + + return VG_TRUE; +} + +void path_clear(struct path *p, VGbitfield capabilities) +{ + path_set_capabilities(p, capabilities); + array_destroy(p->segments); + array_destroy(p->control_points); + p->segments = array_create(size_for_datatype(VG_PATH_DATATYPE_S_8)); + p->control_points = array_create(size_for_datatype(p->datatype)); + p->num_segments = 0; + p->dirty = VG_TRUE; + p->dirty_stroke = VG_TRUE; +} + +struct path * path_create_stroke(struct path *p, + struct matrix *matrix) +{ + VGint i; + VGfloat sx, sy, px, py, ox, oy; + VGfloat x0, y0, x1, y1, x2, y2, x3, y3; + VGfloat data[8]; + void *coords = (VGfloat *)p->control_points->data; + int dashed = (p->base.ctx->state.vg.stroke.dash_pattern_num ? 1 : 0); + struct dash_stroker stroker; + struct vg_state *vg_state = &p->base.ctx->state.vg; + + if (p->stroked.path) + { + /* ### compare the dash patterns to see if we can cache them. + * for now we simply always bail out if the path is dashed. + */ + if (memcmp( &p->stroked.matrix, + matrix, + sizeof *matrix ) == 0 && + !dashed && !p->dirty_stroke && + floatsEqual(p->stroked.stroke_width, vg_state->stroke.line_width.f) && + floatsEqual(p->stroked.miter_limit, vg_state->stroke.miter_limit.f) && + p->stroked.cap_style == vg_state->stroke.cap_style && + p->stroked.join_style == vg_state->stroke.join_style) + { + return p->stroked.path; + } + else { + path_destroy( p->stroked.path ); + p->stroked.path = NULL; + } + } + + + sx = sy = px = py = ox = oy = 0.f; + + if (dashed) + dash_stroker_init((struct stroker *)&stroker, vg_state); + else + stroker_init((struct stroker *)&stroker, vg_state); + + stroker_begin((struct stroker *)&stroker); + + for (i = 0; i < p->num_segments; ++i) { + VGubyte segment = ((VGubyte*)(p->segments->data))[i]; + VGint command = SEGMENT_COMMAND(segment); + VGboolean relative = SEGMENT_ABS_REL(segment); + + switch(command) { + case VG_CLOSE_PATH: { + VGfloat x0 = sx; + VGfloat y0 = sy; + matrix_map_point(matrix, x0, y0, &x0, &y0); + stroker_line_to((struct stroker *)&stroker, x0, y0); + } + break; + case VG_MOVE_TO: + data_at(&coords, p, 0, 2, data); + x0 = data[0]; + y0 = data[1]; + map_if_relative(ox, oy, relative, &x0, &y0); + sx = x0; + sy = y0; + ox = x0; + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + stroker_move_to((struct stroker *)&stroker, x0, y0); + break; + case VG_LINE_TO: + data_at(&coords, p, 0, 2, data); + x0 = data[0]; + y0 = data[1]; + map_if_relative(ox, oy, relative, &x0, &y0); + ox = x0; + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + stroker_line_to((struct stroker *)&stroker, x0, y0); + break; + case VG_HLINE_TO: + data_at(&coords, p, 0, 1, data); + x0 = data[0]; + y0 = oy; + map_if_relative(ox, oy, relative, &x0, 0); + ox = x0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + stroker_line_to((struct stroker *)&stroker, x0, y0); + break; + case VG_VLINE_TO: + data_at(&coords, p, 0, 1, data); + x0 = ox; + y0 = data[0]; + map_if_relative(ox, oy, relative, 0, &y0); + oy = y0; + px = x0; + py = y0; + matrix_map_point(matrix, x0, y0, &x0, &y0); + stroker_line_to((struct stroker *)&stroker, x0, y0); + break; + case VG_CUBIC_TO: { + data_at(&coords, p, 0, 6, data); + x0 = ox; + y0 = oy; + x1 = data[0]; + y1 = data[1]; + x2 = data[2]; + y2 = data[3]; + x3 = data[4]; + y3 = data[5]; + map_if_relative(ox, oy, relative, &x1, &y1); + map_if_relative(ox, oy, relative, &x2, &y2); + map_if_relative(ox, oy, relative, &x3, &y3); + if (floatsEqual(x1, ox) && floatsEqual(y1, oy) && + floatsEqual(x1, x2) && floatsEqual(y1, y2) && + floatsEqual(x2, x3) && floatsEqual(y2, y3)) { + /*ignore the empty segment */ + continue; + } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) { + /* if dup vertex, emit a line */ + ox = x3; + oy = y3; + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_line_to((struct stroker *)&stroker, x3, y3); + continue; + } + ox = x3; + oy = y3; + px = x2; + py = y2; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3); + } + break; + case VG_QUAD_TO: { + data_at(&coords, p, 0, 4, data); + x0 = ox; + y0 = oy; + x1 = data[0]; + y1 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(ox, oy, relative, &x1, &y1); + map_if_relative(ox, oy, relative, &x3, &y3); + px = x1; + py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + if (floatsEqual(x1, ox) && floatsEqual(y1, oy) && + floatsEqual(x1, x2) && floatsEqual(y1, y2) && + floatsEqual(x2, x3) && floatsEqual(y2, y3)) { + /*ignore the empty segment */ + continue; + } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) { + /* if dup vertex, emit a line */ + ox = x3; + oy = y3; + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_line_to((struct stroker *)&stroker, x3, y3); + continue; + } + ox = x3; + oy = y3; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3); + } + break; + case VG_SQUAD_TO: { + data_at(&coords, p, 0, 2, data); + x0 = ox; + y0 = oy; + x1 = 2*ox-px; + y1 = 2*oy-py; + x3 = data[0]; + y3 = data[1]; + map_if_relative(ox, oy, relative, &x3, &y3); + px = x1; + py = y1; + { /* form a cubic out of it */ + x2 = (x3 + 2*x1) / 3.f; + y2 = (y3 + 2*y1) / 3.f; + x1 = (x0 + 2*x1) / 3.f; + y1 = (y0 + 2*y1) / 3.f; + } + if (floatsEqual(x1, ox) && floatsEqual(y1, oy) && + floatsEqual(x1, x2) && floatsEqual(y1, y2) && + floatsEqual(x2, x3) && floatsEqual(y2, y3)) { + /*ignore the empty segment */ + continue; + } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) { + /* if dup vertex, emit a line */ + ox = x3; + oy = y3; + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_line_to((struct stroker *)&stroker, x3, y3); + continue; + } + ox = x3; + oy = y3; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3); + } + break; + case VG_SCUBIC_TO: { + data_at(&coords, p, 0, 4, data); + x0 = ox; + y0 = oy; + x1 = 2*ox-px; + y1 = 2*oy-py; + x2 = data[0]; + y2 = data[1]; + x3 = data[2]; + y3 = data[3]; + map_if_relative(ox, oy, relative, &x2, &y2); + map_if_relative(ox, oy, relative, &x3, &y3); + if (floatsEqual(x1, ox) && floatsEqual(y1, oy) && + floatsEqual(x1, x2) && floatsEqual(y1, y2) && + floatsEqual(x2, x3) && floatsEqual(y2, y3)) { + /*ignore the empty segment */ + continue; + } else if (floatsEqual(x3, ox) && floatsEqual(y3, oy)) { + /* if dup vertex, emit a line */ + ox = x3; + oy = y3; + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_line_to((struct stroker *)&stroker, x3, y3); + continue; + } + ox = x3; + oy = y3; + px = x2; + py = y2; + assert(matrix_is_affine(matrix)); + matrix_map_point(matrix, x0, y0, &x0, &y0); + matrix_map_point(matrix, x1, y1, &x1, &y1); + matrix_map_point(matrix, x2, y2, &x2, &y2); + matrix_map_point(matrix, x3, y3, &x3, &y3); + stroker_curve_to((struct stroker *)&stroker, x1, y1, x2, y2, x3, y3); + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + VGfloat rh, rv, rot; + struct arc arc; + + data_at(&coords, p, 0, 5, data); + x0 = ox; + y0 = oy; + rh = data[0]; + rv = data[1]; + rot = data[2]; + x1 = data[3]; + y1 = data[4]; + map_if_relative(ox, oy, relative, &x1, &y1); + if (floatsEqual(x1, ox) && floatsEqual(y1, oy)) { + /* if dup vertex, emit a line */ + ox = x1; + oy = y1; + matrix_map_point(matrix, x1, y1, &x1, &y1); + stroker_line_to((struct stroker *)&stroker, x1, y1); + continue; + } + arc_init(&arc, command, x0, y0, x1, y1, + rh, rv, rot); + arc_stroke_cb(&arc, (struct stroker *)&stroker, + matrix); + ox = x1; + oy = y1; + px = x1; + py = y1; + } + break; + default: + abort(); + assert(!"Unknown segment!"); + } + } + + stroker_end((struct stroker *)&stroker); + + if (dashed) + dash_stroker_cleanup((struct dash_stroker *)&stroker); + else + stroker_cleanup((struct stroker *)&stroker); + + p->stroked.path = stroker.base.path; + p->stroked.matrix = *matrix; + p->dirty_stroke = VG_FALSE; + p->stroked.stroke_width = vg_state->stroke.line_width.f; + p->stroked.miter_limit = vg_state->stroke.miter_limit.f; + p->stroked.cap_style = vg_state->stroke.cap_style; + p->stroked.join_style = vg_state->stroke.join_style; + + return stroker.base.path; +} + +void path_render(struct path *p, VGbitfield paintModes) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix; + + vg_validate_state(ctx); + + shader_set_drawing_image(ctx->shader, VG_FALSE); + shader_set_image(ctx->shader, 0); +#if 0 + fprintf(stderr, "Matrix(11=%f 12=%f 13=%f 21=%f 22=%f 23=%f 31=%f 32=%f 33=%f)\n", + mat->m[0], mat->m[1], mat->m[2], + mat->m[3], mat->m[4], mat->m[5], + mat->m[6], mat->m[7], mat->m[8]); +#endif + if (paintModes & VG_FILL_PATH) { + /* First the fill */ + shader_set_paint(ctx->shader, ctx->state.vg.fill_paint); + shader_bind(ctx->shader); + path_fill(p, mat); + } + + if (paintModes & VG_STROKE_PATH){ + /* 8.7.5: "line width less than or equal to 0 prevents stroking from + * taking place."*/ + if (ctx->state.vg.stroke.line_width.f <= 0) + return; + shader_set_paint(ctx->shader, ctx->state.vg.stroke_paint); + shader_bind(ctx->shader); + path_stroke(p); + } +} + +void path_fill(struct path *p, struct matrix *mat) +{ + struct vg_context *ctx = vg_current_context(); + { + struct polygon_array *polygon_array = path_get_fill_polygons(p, mat); + struct array *polys = polygon_array->array; + + if (!polygon_array || !polys || !polys->num_elements) { + return; + } + polygon_array_fill(polygon_array, ctx); + } +} + +void path_stroke(struct path *p) +{ + struct vg_context *ctx = vg_current_context(); + struct matrix *mat = &ctx->state.vg.path_user_to_surface_matrix; + VGFillRule old_fill = ctx->state.vg.fill_rule; + struct matrix identity; + struct path *stroke; + + matrix_load_identity(&identity); + stroke = path_create_stroke(p, &identity); + if (stroke && !path_is_empty(stroke)) { + ctx->state.vg.fill_rule = VG_NON_ZERO; + + path_fill(stroke, mat); + + ctx->state.vg.fill_rule = old_fill; + } +} + +void path_move_to(struct path *p, float x, float y) +{ + VGubyte segment = VG_MOVE_TO_ABS; + VGubyte common_data[sizeof(VGfloat) * 2]; + VGfloat data[2] = {x, y}; + + vg_float_to_datatype(p->datatype, common_data, data, 2); + path_append_data(p, 1, &segment, common_data); +} + +void path_line_to(struct path *p, float x, float y) +{ + VGubyte segment = VG_LINE_TO_ABS; + VGubyte common_data[sizeof(VGfloat) * 2]; + VGfloat data[2] = {x, y}; + + vg_float_to_datatype(p->datatype, common_data, data, 2); + + path_append_data(p, 1, &segment, common_data); +} + +void path_cubic_to(struct path *p, float px1, float py1, + float px2, float py2, + float x, float y) +{ + VGubyte segment = VG_CUBIC_TO_ABS; + VGubyte common_data[sizeof(VGfloat) * 6]; + VGfloat data[6]; + + data[0] = px1; data[1] = py1; + data[2] = px2; data[3] = py2; + data[4] = x; data[5] = y; + + vg_float_to_datatype(p->datatype, common_data, data, 6); + + path_append_data(p, 1, &segment, common_data); +} + +static INLINE void line_bounds(VGfloat *line /*x1,y1,x2,y2*/, + VGfloat *bounds) +{ + bounds[0] = MIN2(line[0], line[2]); + bounds[1] = MIN2(line[1], line[3]); + bounds[2] = MAX2(line[0], line[2]) - bounds[0]; + bounds[3] = MAX2(line[1], line[3]) - bounds[1]; +} + +static INLINE void unite_bounds(VGfloat *bounds, + VGfloat *el) +{ + VGfloat cx1, cy1, cx2, cy2; + VGfloat nx1, ny1, nx2, ny2; + + cx1 = bounds[0]; + cy1 = bounds[1]; + cx2 = bounds[0] + bounds[2]; + cy2 = bounds[1] + bounds[3]; + + nx1 = el[0]; + ny1 = el[1]; + nx2 = el[0] + el[2]; + ny2 = el[1] + el[3]; + + bounds[0] = MIN2(cx1, nx1); + bounds[1] = MIN2(cy1, ny1); + bounds[2] = MAX2(cx2, nx2) - bounds[0]; + bounds[3] = MAX2(cy2, ny2) - bounds[1]; +} + +static INLINE void set_bounds(VGfloat *bounds, + VGfloat *element_bounds, + VGboolean *initialized) +{ + if (!(*initialized)) { + memcpy(bounds, element_bounds, 4 * sizeof(VGfloat)); + *initialized = VG_TRUE; + } else + unite_bounds(bounds, element_bounds); +} + +void path_bounding_rect(struct path *p, float *x, float *y, + float *w, float *h) +{ + VGint i; + VGfloat coords[8]; + struct path_iter_data iter; + VGint num_coords; + VGfloat bounds[4]; + VGfloat element_bounds[4]; + VGfloat ox, oy; + VGboolean bounds_inited = VG_FALSE; + + memset(&iter, 0, sizeof(struct path_iter_data)); + memset(&bounds, 0, sizeof(bounds)); + + if (!p->num_segments) { + bounds[2] = -1; + bounds[3] = -1; + } + + + iter.path = p; + iter.coords = p->control_points->data; + + for (i = 0; i < p->num_segments; ++i) { + VGubyte segment; + iter.segment = ((VGubyte*)(p->segments->data))[i]; + + ox = iter.ox; + oy = iter.oy; + + segment = normalize_coords(&iter, &num_coords, coords); + + switch(segment) { + case VG_CLOSE_PATH: + case VG_MOVE_TO_ABS: + break; + case VG_LINE_TO_ABS: { + VGfloat line[4] = {ox, oy, coords[0], coords[1]}; + line_bounds(line, element_bounds); + set_bounds(bounds, element_bounds, &bounds_inited); + } + break; + case VG_CUBIC_TO_ABS: { + struct bezier bezier; + bezier_init(&bezier, ox, oy, + coords[0], coords[1], + coords[2], coords[3], + coords[4], coords[5]); + bezier_exact_bounds(&bezier, element_bounds); + set_bounds(bounds, element_bounds, &bounds_inited); + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + struct arc arc; + struct matrix identity; + struct path *path = path_create(VG_PATH_DATATYPE_F, + 1, 0, 0, 0, VG_PATH_CAPABILITY_ALL); + + matrix_load_identity(&identity); + arc_init(&arc, segment, + ox, oy, coords[3], coords[4], + coords[0], coords[1], coords[2]); + + arc_to_path(&arc, path, &identity); + + path_bounding_rect(path, element_bounds + 0, element_bounds + 1, + element_bounds + 2, element_bounds + 3); + set_bounds(bounds, element_bounds, &bounds_inited); + } + break; + default: + assert(0); + } + } + + *x = bounds[0]; + *y = bounds[1]; + *w = bounds[2]; + *h = bounds[3]; +} + +float path_length(struct path *p, int start_segment, int num_segments) +{ + VGint i; + VGfloat coords[8]; + struct path_iter_data iter; + VGint num_coords; + VGfloat length = 0; + VGfloat ox, oy; + VGboolean in_range = VG_FALSE; + + memset(&iter, 0, sizeof(struct path_iter_data)); + + iter.path = p; + iter.coords = p->control_points->data; + + for (i = 0; i < (start_segment + num_segments); ++i) { + VGubyte segment; + + iter.segment = ((VGubyte*)(p->segments->data))[i]; + + ox = iter.ox; + oy = iter.oy; + + segment = normalize_coords(&iter, &num_coords, coords); + + in_range = (i >= start_segment) && i <= (start_segment + num_segments); + if (!in_range) + continue; + + switch(segment) { + case VG_MOVE_TO_ABS: + break; + case VG_CLOSE_PATH: { + VGfloat line[4] = {ox, oy, iter.sx, iter.sy}; + length += line_lengthv(line); + } + break; + case VG_LINE_TO_ABS: { + VGfloat line[4] = {ox, oy, coords[0], coords[1]}; + length += line_lengthv(line); + } + break; + case VG_CUBIC_TO_ABS: { + struct bezier bezier; + bezier_init(&bezier, ox, oy, + coords[0], coords[1], + coords[2], coords[3], + coords[4], coords[5]); + length += bezier_length(&bezier, BEZIER_DEFAULT_ERROR); + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + struct arc arc; + struct matrix identity; + struct path *path = path_create(VG_PATH_DATATYPE_F, + 1, 0, 0, 0, VG_PATH_CAPABILITY_ALL); + + matrix_load_identity(&identity); + arc_init(&arc, segment, + ox, oy, coords[3], coords[4], + coords[0], coords[1], coords[2]); + + arc_to_path(&arc, path, &identity); + + length += path_length(path, 0, path_num_segments(path)); + } + break; + default: + assert(0); + } + } + + return length; +} + +static INLINE VGboolean point_on_current_segment(VGfloat distance, + VGfloat length, + VGfloat segment_length) +{ + return + (((floatIsZero(distance) || distance < 0) && floatIsZero(length)) || + ((distance > length || floatsEqual(distance, length)) && + (floatsEqual(distance, length + segment_length) || + distance < (length + segment_length)))); +} + +static VGboolean path_point_segment(struct path_iter_data iter, + struct path_iter_data prev_iter, + VGfloat coords[8], + VGfloat distance, + VGfloat length, VGfloat *current_length, + VGfloat *point, VGfloat *normal) +{ + switch (iter.segment) { + case VG_MOVE_TO_ABS: + break; + case VG_CLOSE_PATH: { + VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy}; + VGboolean on_current_segment = VG_FALSE; + *current_length = line_lengthv(line); + on_current_segment = point_on_current_segment(distance, + length, + *current_length); + if (on_current_segment) { + VGfloat at = (distance - length) / line_lengthv(line); + line_normal_vector(line, normal); + line_point_at(line, at, point); + return VG_TRUE; + } + } + break; + case VG_LINE_TO_ABS: { + VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]}; + VGboolean on_current_segment = VG_FALSE; + *current_length = line_lengthv(line); + on_current_segment = point_on_current_segment(distance, + length, + *current_length); + if (on_current_segment) { + VGfloat at = (distance - length) / line_lengthv(line); + line_normal_vector(line, normal); + line_point_at(line, at, point); + return VG_TRUE; + } + } + break; + case VG_CUBIC_TO_ABS: { + struct bezier bezier; + bezier_init(&bezier, prev_iter.ox, prev_iter.oy, + coords[0], coords[1], + coords[2], coords[3], + coords[4], coords[5]); + *current_length = bezier_length(&bezier, BEZIER_DEFAULT_ERROR); + if (point_on_current_segment(distance, length, *current_length)) { + bezier_point_at_length(&bezier, distance - length, + point, normal); + return VG_TRUE; + } + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + struct arc arc; + struct matrix identity; + struct path *path = path_create(VG_PATH_DATATYPE_F, + 1, 0, 0, 0, VG_PATH_CAPABILITY_ALL); + + matrix_load_identity(&identity); + arc_init(&arc, iter.segment, + prev_iter.ox, prev_iter.oy, coords[3], coords[4], + coords[0], coords[1], coords[2]); + + arc_to_path(&arc, path, &identity); + + *current_length = path_length(path, 0, path_num_segments(path)); + if (point_on_current_segment(distance, length, *current_length)) { + path_point(path, 0, path_num_segments(path), + distance - length, point, normal); + return VG_TRUE; + } + } + break; + default: + assert(0); + } + return VG_FALSE; +} + +void path_point(struct path *p, VGint start_segment, VGint num_segments, + VGfloat distance, VGfloat *point, VGfloat *normal) +{ + VGint i; + VGfloat coords[8]; + struct path_iter_data iter, prev_iter; + VGint num_coords; + VGfloat length = 0; + VGfloat current_length = 0; + + memset(&iter, 0, sizeof(struct path_iter_data)); + memset(&prev_iter, 0, sizeof(struct path_iter_data)); + + point[0] = 0; + point[1] = 0; + + normal[0] = 0; + normal[1] = -1; + + iter.path = p; + iter.coords = p->control_points->data; + if (distance < 0) + distance = 0; + + for (i = 0; i < (start_segment + num_segments); ++i) { + VGboolean outside_range = (i < start_segment || + i >= (start_segment + num_segments)); + + prev_iter = iter; + + iter.segment = ((VGubyte*)(p->segments->data))[i]; + iter.segment = normalize_coords(&iter, &num_coords, coords); + + if (outside_range) + continue; + + if (path_point_segment(iter, prev_iter, coords, + distance, length, ¤t_length, + point, normal)) + return; + + length += current_length; + } + + /* + *OpenVG 1.0 - 8.6.11 vgPointAlongPath + * + * If distance is greater than or equal to the path length + *(i.e., the value returned by vgPathLength when called with the same + *startSegment and numSegments parameters), the visual ending point of + *the path is used. + */ + { + switch (iter.segment) { + case VG_MOVE_TO_ABS: + break; + case VG_CLOSE_PATH: { + VGfloat line[4] = {prev_iter.ox, prev_iter.oy, iter.sx, iter.sy}; + line_normal_vector(line, normal); + line_point_at(line, 1.f, point); + } + break; + case VG_LINE_TO_ABS: { + VGfloat line[4] = {prev_iter.ox, prev_iter.oy, coords[0], coords[1]}; + line_normal_vector(line, normal); + line_point_at(line, 1.f, point); + } + break; + case VG_CUBIC_TO_ABS: { + struct bezier bezier; + bezier_init(&bezier, prev_iter.ox, prev_iter.oy, + coords[0], coords[1], + coords[2], coords[3], + coords[4], coords[5]); + bezier_point_at_t(&bezier, 1.f, point, normal); + } + break; + case VG_SCCWARC_TO: + case VG_SCWARC_TO: + case VG_LCCWARC_TO: + case VG_LCWARC_TO: { + struct arc arc; + struct matrix identity; + struct path *path = path_create(VG_PATH_DATATYPE_F, + 1, 0, 0, 0, VG_PATH_CAPABILITY_ALL); + + matrix_load_identity(&identity); + arc_init(&arc, iter.segment, + prev_iter.ox, prev_iter.oy, coords[3], coords[4], + coords[0], coords[1], coords[2]); + + arc_to_path(&arc, path, &identity); + + path_point(path, 0, path_num_segments(path), + /* to make sure we're bigger than len * 2 it */ + 2 * path_length(path, 0, path_num_segments(path)), + point, normal); + } + break; + default: + assert(0); + } + } +} + +VGboolean path_is_empty(struct path *p) +{ + return p->segments->num_elements == 0; +} diff --git a/src/gallium/state_trackers/vega/path.h b/src/gallium/state_trackers/vega/path.h new file mode 100644 index 00000000000..e34538b7368 --- /dev/null +++ b/src/gallium/state_trackers/vega/path.h @@ -0,0 +1,126 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _PATH_H +#define _PATH_H + +#include "VG/openvg.h" + +struct path; +struct polygon; +struct matrix; + +enum fill_rule { + ODD_EVEN_FILL, + WINDING_FILL +}; + + +struct path_for_each_data { + VGubyte segment; + /* all coords are absolute, even if segment is relative */ + const VGfloat *coords; + VGfloat sx, sy, ox, oy, px, py; + void *user_data; +}; + +typedef VGboolean (*path_for_each_cb)(struct path *p, + struct path_for_each_data *data); + + +struct path *path_create(VGPathDatatype dt, VGfloat scale, VGfloat bias, + VGint segmentCapacityHint, + VGint coordCapacityHint, + VGbitfield capabilities); +void path_destroy(struct path *p); + +VGbitfield path_capabilities(struct path *p); +void path_set_capabilities(struct path *p, VGbitfield bf); + +void path_append_data(struct path *p, + VGint numSegments, + const VGubyte * pathSegments, + const void * pathData); + +void path_append_path(struct path *dst, + struct path *src); + +VGint path_num_segments(struct path *p); + +void path_bounding_rect(struct path *p, float *x, float *y, + float *w, float *h); +float path_length(struct path *p, int start_segment, int num_segments); + +void path_set_fill_rule(enum fill_rule fill); +enum fill_rule path_fill_rule(enum fill_rule fill); + +VGboolean path_is_empty(struct path *p); + +VGbyte path_datatype_size(struct path *p); + +VGPathDatatype path_datatype(struct path *p); +VGfloat path_scale(struct path *p); +VGfloat path_bias(struct path *p); +VGint path_num_coords(struct path *p); + +void path_modify_coords(struct path *p, + VGint startIndex, + VGint numSegments, + const void * pathData); + +struct path *path_create_stroke(struct path *p, + struct matrix *m); + +void path_for_each_segment(struct path *path, + path_for_each_cb cb, + void *user_data); + +void path_transform(struct path *dst, struct path *src); +VGboolean path_interpolate(struct path *dst, + struct path *start, struct path *end, + VGfloat amount); + +void path_clear(struct path *p, VGbitfield capabilities); +void path_render(struct path *p, VGbitfield paintModes); +void path_fill(struct path *p, struct matrix *mat); +void path_stroke(struct path *p); + +void path_move_to(struct path *p, float x, float y); +void path_line_to(struct path *p, float x, float y); +void path_cubic_to(struct path *p, float px1, float py1, + float px2, float py2, + float x, float y); + +void path_point(struct path *p, VGint startSegment, VGint numSegments, + VGfloat distance, VGfloat *point, VGfloat *normal); + + + +void vg_float_to_datatype(VGPathDatatype datatype, + VGubyte *common_data, + const VGfloat *data, + VGint num_coords); +#endif diff --git a/src/gallium/state_trackers/vega/path_utils.h b/src/gallium/state_trackers/vega/path_utils.h new file mode 100644 index 00000000000..c2b3221dc53 --- /dev/null +++ b/src/gallium/state_trackers/vega/path_utils.h @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PATH_UTILS_H +#define PATH_UTILS_H + +#include "VG/openvg.h" + +#define SEGMENT_COMMAND(command) /* Extract segment type */ \ + ((command) & 0x1e) +#define SEGMENT_ABS_REL(command) /* Extract absolute/relative bit */ \ + ((command) & 0x1) + +static INLINE VGint size_for_datatype(VGPathDatatype datatype) +{ + switch(datatype) { + case VG_PATH_DATATYPE_S_8: + return 1; + case VG_PATH_DATATYPE_S_16: + return 2; + case VG_PATH_DATATYPE_S_32: + return 4; + case VG_PATH_DATATYPE_F: + return 4; + default: + assert(!"unknown datatype"); + } + return 0; +} + +static INLINE VGint num_elements_for_segments(const VGubyte *segments, + VGint num_segments) +{ + VGint i; + VGint count = 0; + + for (i = 0; i < num_segments; ++i) { + VGubyte segment = segments[i]; + VGint command = SEGMENT_COMMAND(segment); + switch(command) { + case VG_CLOSE_PATH: + break; + case VG_MOVE_TO: + count += 2; + break; + case VG_LINE_TO: + count += 2; + break; + case VG_HLINE_TO: + count += 1; + break; + case VG_VLINE_TO: + count += 1; + break; + case VG_QUAD_TO: + count += 4; + break; + case VG_CUBIC_TO: + count += 6; + break; + case VG_SQUAD_TO: + count += 2; + break; + case VG_SCUBIC_TO: + count += 4; + break; + case VG_SCCWARC_TO: + count += 5; + break; + case VG_SCWARC_TO: + count += 5; + break; + case VG_LCCWARC_TO: + count += 5; + break; + case VG_LCWARC_TO: + count += 5; + break; + default: + assert(!"Unknown segment!"); + } + } + return count; +} + +#endif diff --git a/src/gallium/state_trackers/vega/polygon.c b/src/gallium/state_trackers/vega/polygon.c new file mode 100644 index 00000000000..b6d282d803b --- /dev/null +++ b/src/gallium/state_trackers/vega/polygon.c @@ -0,0 +1,550 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "polygon.h" + +#include "matrix.h" /*for floatsEqual*/ +#include "vg_context.h" +#include "vg_state.h" +#include "paint.h" +#include "renderer.h" +#include "util_array.h" +#include "VG/openvg.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "pipe/p_screen.h" + +#include "util/u_draw_quad.h" +#include "util/u_math.h" + +#include <string.h> +#include <stdlib.h> + +#define DEBUG_POLYGON 0 + +#define COMPONENTS 2 + +struct polygon +{ + VGfloat *data; + VGint size; + + VGint num_verts; + + VGboolean dirty; + struct pipe_buffer *vbuf; + struct pipe_screen *screen; +}; + +static float *ptr_to_vertex(float *data, int idx) +{ + return data + (idx * COMPONENTS); +} + +#if 0 +static void polygon_print(struct polygon *poly) +{ + int i; + float *vert; + debug_printf("Polygon %p, size = %d\n", poly, poly->num_verts); + for (i = 0; i < poly->num_verts; ++i) { + vert = ptr_to_vertex(poly->data, i); + debug_printf("%f, %f, ", vert[0], vert[1]); + } + debug_printf("\nend\n"); +} +#endif + + +struct polygon * polygon_create(int size) +{ + struct polygon *poly = (struct polygon*)malloc(sizeof(struct polygon)); + + poly->data = malloc(sizeof(float) * COMPONENTS * size); + poly->size = size; + poly->num_verts = 0; + poly->dirty = VG_TRUE; + poly->vbuf = NULL; + + return poly; +} + +struct polygon * polygon_create_from_data(float *data, int size) +{ + struct polygon *poly = polygon_create(size); + + memcpy(poly->data, data, sizeof(float) * COMPONENTS * size); + poly->num_verts = size; + poly->dirty = VG_TRUE; + poly->vbuf = NULL; + + return poly; +} + +void polygon_destroy(struct polygon *poly) +{ + if (poly->vbuf) + pipe_buffer_reference(&poly->vbuf, NULL); + + free(poly->data); + free(poly); +} + +void polygon_resize(struct polygon *poly, int new_size) +{ + float *data = (float*)malloc(sizeof(float) * COMPONENTS * new_size); + int size = MIN2(sizeof(float) * COMPONENTS * new_size, + sizeof(float) * COMPONENTS * poly->size); + memcpy(data, poly->data, size); + free(poly->data); + poly->data = data; + poly->size = new_size; + poly->dirty = VG_TRUE; +} + +int polygon_size(struct polygon *poly) +{ + return poly->size; +} + +int polygon_vertex_count(struct polygon *poly) +{ + return poly->num_verts; +} + +float * polygon_data(struct polygon *poly) +{ + return poly->data; +} + +void polygon_vertex_append(struct polygon *p, + float x, float y) +{ + float *vert; +#if DEBUG_POLYGON + debug_printf("Append vertex [%f, %f]\n", x, y); +#endif + if (p->num_verts >= p->size) { + polygon_resize(p, p->size * 2); + } + + vert = ptr_to_vertex(p->data, p->num_verts); + vert[0] = x; + vert[1] = y; + ++p->num_verts; + p->dirty = VG_TRUE; +} + +void polygon_set_vertex(struct polygon *p, int idx, + float x, float y) +{ + float *vert; + if (idx >= p->num_verts) { + /*fixme: error reporting*/ + abort(); + return; + } + + vert = ptr_to_vertex(p->data, idx); + vert[0] = x; + vert[1] = y; + p->dirty = VG_TRUE; +} + +void polygon_vertex(struct polygon *p, int idx, + float *vertex) +{ + float *vert; + if (idx >= p->num_verts) { + /*fixme: error reporting*/ + abort(); + return; + } + + vert = ptr_to_vertex(p->data, idx); + vertex[0] = vert[0]; + vertex[1] = vert[1]; +} + +void polygon_bounding_rect(struct polygon *p, + float *rect) +{ + int i; + float minx, miny, maxx, maxy; + float *vert = ptr_to_vertex(p->data, 0); + minx = vert[0]; + maxx = vert[0]; + miny = vert[1]; + maxy = vert[1]; + + for (i = 1; i < p->num_verts; ++i) { + vert = ptr_to_vertex(p->data, i); + minx = MIN2(vert[0], minx); + miny = MIN2(vert[1], miny); + + maxx = MAX2(vert[0], maxx); + maxy = MAX2(vert[1], maxy); + } + + rect[0] = minx; + rect[1] = miny; + rect[2] = maxx - minx; + rect[3] = maxy - miny; +} + +int polygon_contains_point(struct polygon *p, + float x, float y) +{ + return 0; +} + +void polygon_append_polygon(struct polygon *dst, + struct polygon *src) +{ + if (dst->num_verts + src->num_verts >= dst->size) { + polygon_resize(dst, dst->num_verts + src->num_verts * 1.5); + } + memcpy(ptr_to_vertex(dst->data, dst->num_verts), + src->data, src->num_verts * COMPONENTS * sizeof(VGfloat)); + dst->num_verts += src->num_verts; +} + +VGboolean polygon_is_closed(struct polygon *p) +{ + VGfloat start[2], end[2]; + + polygon_vertex(p, 0, start); + polygon_vertex(p, p->num_verts - 1, end); + + return floatsEqual(start[0], end[0]) && floatsEqual(start[1], end[1]); +} + +static void set_blend_for_fill(struct pipe_blend_state *blend) +{ + memset(blend, 0, sizeof(struct pipe_blend_state)); + blend->colormask = 0; /*disable colorwrites*/ + + blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA; +} + +static void draw_polygon(struct vg_context *ctx, + struct polygon *poly) +{ + int vert_size; + struct pipe_context *pipe; + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + + vert_size = poly->num_verts * COMPONENTS * sizeof(float); + + /*polygon_print(poly);*/ + + pipe = ctx->pipe; + + if (poly->vbuf == NULL || poly->dirty) { + if (poly->vbuf) { + pipe_buffer_reference(&poly->vbuf, + NULL); + } + poly->screen = pipe->screen; + poly->vbuf= pipe_user_buffer_create(poly->screen, + poly->data, + vert_size); + poly->dirty = VG_FALSE; + } + + + /* tell pipe about the vertex buffer */ + memset(&vbuffer, 0, sizeof(vbuffer)); + vbuffer.buffer = poly->vbuf; + vbuffer.stride = COMPONENTS * sizeof(float); /* vertex size */ + vbuffer.buffer_offset = 0; + vbuffer.max_index = poly->num_verts - 1; + pipe->set_vertex_buffers(pipe, 1, &vbuffer); + + /* tell pipe about the vertex attributes */ + velement.src_offset = 0; + velement.vertex_buffer_index = 0; + velement.src_format = PIPE_FORMAT_R32G32_FLOAT; + velement.nr_components = COMPONENTS; + pipe->set_vertex_elements(pipe, 1, &velement); + + /* draw */ + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, + 0, poly->num_verts); +} + +void polygon_fill(struct polygon *poly, struct vg_context *ctx) +{ + struct pipe_depth_stencil_alpha_state dsa; + struct pipe_blend_state blend; + VGfloat bounds[4]; + VGfloat min_x, min_y, max_x, max_y; + assert(poly); + polygon_bounding_rect(poly, bounds); + min_x = bounds[0]; + min_y = bounds[1]; + max_x = bounds[0] + bounds[2]; + max_y = bounds[1] + bounds[3]; + +#if DEBUG_POLYGON + debug_printf("Poly bounds are [%f, %f], [%f, %f]\n", + min_x, min_y, max_x, max_y); +#endif + + set_blend_for_fill(&blend); + + memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state)); + + cso_save_blend(ctx->cso_context); + cso_save_depth_stencil_alpha(ctx->cso_context); + + dsa.stencil[0].enabled = 1; + if (ctx->state.vg.fill_rule == VG_EVEN_ODD) { + dsa.stencil[0].writemask = 1; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT; + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + draw_polygon(ctx, poly); + } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) { + struct pipe_screen *screen = ctx->pipe->screen; + + if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) { + /* front */ + dsa.stencil[0].writemask = ~0; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP; + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + /* back */ + dsa.stencil[1].enabled = 1; + dsa.stencil[1].writemask = ~0; + dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP; + dsa.stencil[1].func = PIPE_FUNC_ALWAYS; + dsa.stencil[1].ref_value = 0; + dsa.stencil[1].valuemask = ~0; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + draw_polygon(ctx, poly); + } else { + struct pipe_rasterizer_state raster; + + memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state)); + + cso_save_rasterizer(ctx->cso_context); + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + cso_set_rasterizer(ctx->cso_context, &raster); + draw_polygon(ctx, poly); + + raster.cull_mode = raster.front_winding; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP; + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + cso_set_rasterizer(ctx->cso_context, &raster); + draw_polygon(ctx, poly); + + cso_restore_rasterizer(ctx->cso_context); + } + } + + /* restore color writes */ + cso_restore_blend(ctx->cso_context); + /* setup stencil ops */ + dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = dsa.stencil[0].writemask; + dsa.stencil[1].enabled = 0; + memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth, + sizeof(struct pipe_depth_state)); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + + /* render the quad to propagate the rendering from stencil */ + renderer_draw_quad(ctx->renderer, min_x, min_y, + max_x, max_y, 0.0f/*depth should be disabled*/); + + cso_restore_depth_stencil_alpha(ctx->cso_context); +} + +void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx) +{ + struct array *polys = polyarray->array; + struct pipe_depth_stencil_alpha_state dsa; + struct pipe_blend_state blend; + VGfloat min_x = polyarray->min_x; + VGfloat min_y = polyarray->min_y; + VGfloat max_x = polyarray->max_x; + VGfloat max_y = polyarray->max_y; + VGint i; + + +#if DEBUG_POLYGON + debug_printf("%s: Poly bounds are [%f, %f], [%f, %f]\n", + __FUNCTION__, + min_x, min_y, max_x, max_y); +#endif + + set_blend_for_fill(&blend); + + memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state)); + + cso_save_blend(ctx->cso_context); + cso_save_depth_stencil_alpha(ctx->cso_context); + + dsa.stencil[0].enabled = 1; + if (ctx->state.vg.fill_rule == VG_EVEN_ODD) { + dsa.stencil[0].writemask = 1; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INVERT; + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + for (i = 0; i < polys->num_elements; ++i) { + struct polygon *poly = (((struct polygon**)polys->data)[i]); + draw_polygon(ctx, poly); + } + } else if (ctx->state.vg.fill_rule == VG_NON_ZERO) { + struct pipe_screen *screen = ctx->pipe->screen; + + if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) { + /* front */ + dsa.stencil[0].writemask = ~0; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP; + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + /* back */ + dsa.stencil[1].enabled = 1; + dsa.stencil[1].writemask = ~0; + dsa.stencil[1].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[1].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[1].zpass_op = PIPE_STENCIL_OP_DECR_WRAP; + dsa.stencil[1].func = PIPE_FUNC_ALWAYS; + dsa.stencil[1].ref_value = 0; + dsa.stencil[1].valuemask = ~0; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + for (i = 0; i < polys->num_elements; ++i) { + struct polygon *poly = (((struct polygon**)polys->data)[i]); + draw_polygon(ctx, poly); + } + } else { + struct pipe_rasterizer_state raster; + + memcpy(&raster, &ctx->state.g3d.rasterizer, sizeof(struct pipe_rasterizer_state)); + + cso_save_rasterizer(ctx->cso_context); + dsa.stencil[0].func = PIPE_FUNC_ALWAYS; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = ~0; + + raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_INCR_WRAP; + + cso_set_blend(ctx->cso_context, &blend); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + cso_set_rasterizer(ctx->cso_context, &raster); + for (i = 0; i < polys->num_elements; ++i) { + struct polygon *poly = (((struct polygon**)polys->data)[i]); + draw_polygon(ctx, poly); + } + + raster.cull_mode = raster.front_winding; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_KEEP; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_DECR_WRAP; + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + cso_set_rasterizer(ctx->cso_context, &raster); + for (i = 0; i < polys->num_elements; ++i) { + struct polygon *poly = (((struct polygon**)polys->data)[i]); + draw_polygon(ctx, poly); + } + + cso_restore_rasterizer(ctx->cso_context); + } + } + + /* restore color writes */ + cso_restore_blend(ctx->cso_context); + /* setup stencil ops */ + dsa.stencil[0].func = PIPE_FUNC_NOTEQUAL; + dsa.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE; + dsa.stencil[0].ref_value = 0; + dsa.stencil[0].valuemask = dsa.stencil[0].writemask; + dsa.stencil[1].enabled = 0; + memcpy(&dsa.depth, &ctx->state.g3d.dsa.depth, + sizeof(struct pipe_depth_state)); + cso_set_depth_stencil_alpha(ctx->cso_context, &dsa); + + /* render the quad to propagate the rendering from stencil */ + renderer_draw_quad(ctx->renderer, min_x, min_y, + max_x, max_y, 0.0f/*depth should be disabled*/); + + cso_restore_depth_stencil_alpha(ctx->cso_context); +} diff --git a/src/gallium/state_trackers/vega/polygon.h b/src/gallium/state_trackers/vega/polygon.h new file mode 100644 index 00000000000..22672b728ed --- /dev/null +++ b/src/gallium/state_trackers/vega/polygon.h @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef POLYGON_H +#define POLYGON_H + +#include "VG/openvg.h" + +struct polygon; +struct vg_context; +struct vg_paint; +struct array; + +struct polygon *polygon_create(int size); +struct polygon *polygon_create_from_data(float *data, int size); +void polygon_destroy(struct polygon *poly); + +void polygon_resize(struct polygon *poly, int new_size); +int polygon_size(struct polygon *poly); + +int polygon_vertex_count(struct polygon *poly); +float * polygon_data(struct polygon *poly); + +void polygon_vertex_append(struct polygon *p, + float x, float y); +void polygon_append_polygon(struct polygon *dst, + struct polygon *src); +void polygon_set_vertex(struct polygon *p, int idx, + float x, float y); +void polygon_vertex(struct polygon *p, int idx, + float *vertex); + +void polygon_bounding_rect(struct polygon *p, + float *rect); +int polygon_contains_point(struct polygon *p, + float x, float y); + +VGboolean polygon_is_closed(struct polygon *p); + +void polygon_fill(struct polygon *p, struct vg_context *pipe); + +/* TODO: make a file/module around this struct + */ +struct polygon_array { + struct array *array; + VGfloat min_x, max_x; + VGfloat min_y, max_y; +}; + +void polygon_array_fill(struct polygon_array *polyarray, struct vg_context *ctx); + +#endif diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c new file mode 100644 index 00000000000..f7c5f2f0cdf --- /dev/null +++ b/src/gallium/state_trackers/vega/renderer.c @@ -0,0 +1,592 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "renderer.h" + +#include "vg_context.h" + +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "pipe/p_screen.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_draw_quad.h" +#include "util/u_simple_shaders.h" +#include "util/u_memory.h" + +#include "cso_cache/cso_context.h" + +struct renderer { + struct pipe_context *pipe; + struct vg_context *owner; + + struct cso_context *cso; + + void *fs; + + VGfloat vertices[4][2][4]; +}; + +static void setup_shaders(struct renderer *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + /* fragment shader */ + ctx->fs = util_make_fragment_tex_shader(pipe); +} + +static struct pipe_buffer * +setup_vertex_data(struct renderer *ctx, + float x0, float y0, float x1, float y1, float z) +{ + ctx->vertices[0][0][0] = x0; + ctx->vertices[0][0][1] = y0; + ctx->vertices[0][0][2] = z; + ctx->vertices[0][1][0] = 0.0f; /*s*/ + ctx->vertices[0][1][1] = 0.0f; /*t*/ + + ctx->vertices[1][0][0] = x1; + ctx->vertices[1][0][1] = y0; + ctx->vertices[1][0][2] = z; + ctx->vertices[1][1][0] = 1.0f; /*s*/ + ctx->vertices[1][1][1] = 0.0f; /*t*/ + + ctx->vertices[2][0][0] = x1; + ctx->vertices[2][0][1] = y1; + ctx->vertices[2][0][2] = z; + ctx->vertices[2][1][0] = 1.0f; + ctx->vertices[2][1][1] = 1.0f; + + ctx->vertices[3][0][0] = x0; + ctx->vertices[3][0][1] = y1; + ctx->vertices[3][0][2] = z; + ctx->vertices[3][1][0] = 0.0f; + ctx->vertices[3][1][1] = 1.0f; + + return pipe_user_buffer_create( ctx->pipe->screen, + ctx->vertices, + sizeof(ctx->vertices) ); +} + +static struct pipe_buffer * +setup_vertex_data_tex(struct renderer *ctx, + float x0, float y0, float x1, float y1, + float s0, float t0, float s1, float t1, + float z) +{ + ctx->vertices[0][0][0] = x0; + ctx->vertices[0][0][1] = y0; + ctx->vertices[0][0][2] = z; + ctx->vertices[0][1][0] = s0; /*s*/ + ctx->vertices[0][1][1] = t0; /*t*/ + + ctx->vertices[1][0][0] = x1; + ctx->vertices[1][0][1] = y0; + ctx->vertices[1][0][2] = z; + ctx->vertices[1][1][0] = s1; /*s*/ + ctx->vertices[1][1][1] = t0; /*t*/ + + ctx->vertices[2][0][0] = x1; + ctx->vertices[2][0][1] = y1; + ctx->vertices[2][0][2] = z; + ctx->vertices[2][1][0] = s1; + ctx->vertices[2][1][1] = t1; + + ctx->vertices[3][0][0] = x0; + ctx->vertices[3][0][1] = y1; + ctx->vertices[3][0][2] = z; + ctx->vertices[3][1][0] = s0; + ctx->vertices[3][1][1] = t1; + + return pipe_user_buffer_create( ctx->pipe->screen, + ctx->vertices, + sizeof(ctx->vertices) ); +} + + +static struct pipe_buffer * +setup_vertex_data_qtex(struct renderer *ctx, + float x0, float y0, float x1, float y1, + float x2, float y2, float x3, float y3, + float s0, float t0, float s1, float t1, + float z) +{ + ctx->vertices[0][0][0] = x0; + ctx->vertices[0][0][1] = y0; + ctx->vertices[0][0][2] = z; + ctx->vertices[0][1][0] = s0; /*s*/ + ctx->vertices[0][1][1] = t0; /*t*/ + + ctx->vertices[1][0][0] = x1; + ctx->vertices[1][0][1] = y1; + ctx->vertices[1][0][2] = z; + ctx->vertices[1][1][0] = s1; /*s*/ + ctx->vertices[1][1][1] = t0; /*t*/ + + ctx->vertices[2][0][0] = x2; + ctx->vertices[2][0][1] = y2; + ctx->vertices[2][0][2] = z; + ctx->vertices[2][1][0] = s1; + ctx->vertices[2][1][1] = t1; + + ctx->vertices[3][0][0] = x3; + ctx->vertices[3][0][1] = y3; + ctx->vertices[3][0][2] = z; + ctx->vertices[3][1][0] = s0; + ctx->vertices[3][1][1] = t1; + + return pipe_user_buffer_create( ctx->pipe->screen, + ctx->vertices, + sizeof(ctx->vertices) ); +} + +struct renderer * renderer_create(struct vg_context *owner) +{ + VGint i; + struct renderer *renderer = CALLOC_STRUCT(renderer); + + if (!renderer) + return NULL; + + renderer->owner = owner; + renderer->pipe = owner->pipe; + renderer->cso = owner->cso_context; + + setup_shaders(renderer); + + /* init vertex data that doesn't change */ + for (i = 0; i < 4; i++) { + renderer->vertices[i][0][3] = 1.0f; /* w */ + renderer->vertices[i][1][2] = 0.0f; /* r */ + renderer->vertices[i][1][3] = 1.0f; /* q */ + } + + return renderer; +} + +void renderer_destroy(struct renderer *ctx) +{ +#if 0 + if (ctx->fs) { + cso_delete_fragment_shader(ctx->cso, ctx->fs); + ctx->fs = NULL; + } +#endif + free(ctx); +} + +void renderer_draw_quad(struct renderer *r, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat depth) +{ + struct pipe_buffer *buf; + + buf = setup_vertex_data(r, x1, y1, x2, y2, depth); + + if (buf) { + util_draw_vertex_buffer(r->pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference( &buf, + NULL ); + } +} + +void renderer_draw_texture(struct renderer *r, + struct pipe_texture *tex, + VGfloat x1offset, VGfloat y1offset, + VGfloat x2offset, VGfloat y2offset, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2) +{ + struct pipe_context *pipe = r->pipe; + struct pipe_buffer *buf; + VGfloat s0, t0, s1, t1; + + assert(tex->width[0] != 0); + assert(tex->height[0] != 0); + + s0 = x1offset / tex->width[0]; + s1 = x2offset / tex->width[0]; + t0 = y1offset / tex->height[0]; + t1 = y2offset / tex->height[0]; + + cso_save_vertex_shader(r->cso); + /* shaders */ + cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner)); + + /* draw quad */ + buf = setup_vertex_data_tex(r, x1, y1, x2, y2, + s0, t0, s1, t1, 0.0f); + + if (buf) { + util_draw_vertex_buffer(pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference( &buf, + NULL ); + } + + cso_restore_vertex_shader(r->cso); +} + +void renderer_copy_texture(struct renderer *ctx, + struct pipe_texture *src, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + struct pipe_texture *dst, + VGfloat dx1, VGfloat dy1, + VGfloat dx2, VGfloat dy2) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_buffer *buf; + struct pipe_surface *dst_surf = screen->get_tex_surface( + screen, dst, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + struct pipe_framebuffer_state fb; + float s0, t0, s1, t1; + + assert(src->width[0] != 0); + assert(src->height[0] != 0); + assert(dst->width[0] != 0); + assert(dst->height[0] != 0); + +#if 0 + debug_printf("copy texture [%f, %f, %f, %f], [%f, %f, %f, %f]\n", + sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2); +#endif + +#if 1 + s0 = sx1 / src->width[0]; + s1 = sx2 / src->width[0]; + t0 = sy1 / src->height[0]; + t1 = sy2 / src->height[0]; +#else + s0 = 0; + s1 = 1; + t0 = 0; + t1 = 1; +#endif + + assert(screen->is_format_supported(screen, dst_surf->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); + + /* save state (restored below) */ + cso_save_blend(ctx->cso); + cso_save_samplers(ctx->cso); + cso_save_sampler_textures(ctx->cso); + cso_save_framebuffer(ctx->cso); + cso_save_fragment_shader(ctx->cso); + cso_save_vertex_shader(ctx->cso); + + cso_save_viewport(ctx->cso); + + + /* set misc state we care about */ + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + cso_set_blend(ctx->cso, &blend); + } + + /* sampler */ + { + struct pipe_sampler_state sampler; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; + cso_single_sampler(ctx->cso, 0, &sampler); + cso_single_sampler_done(ctx->cso); + } + + vg_set_viewport(ctx->owner, VEGA_Y0_TOP); + + /* texture */ + cso_set_sampler_textures(ctx->cso, 1, &src); + + /* shaders */ + cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner)); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + + /* drawing dest */ + memset(&fb, 0, sizeof(fb)); + fb.width = dst_surf->width; + fb.height = dst_surf->height; + fb.nr_cbufs = 1; + fb.cbufs[0] = dst_surf; + { + VGint i; + for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i) + fb.cbufs[i] = 0; + } + cso_set_framebuffer(ctx->cso, &fb); + + /* draw quad */ + buf = setup_vertex_data_tex(ctx, + dx1, dy1, + dx2, dy2, + s0, t0, s1, t1, + 0.0f); + + if (buf) { + util_draw_vertex_buffer(ctx->pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference( &buf, + NULL ); + } + + /* restore state we changed */ + cso_restore_blend(ctx->cso); + cso_restore_samplers(ctx->cso); + cso_restore_sampler_textures(ctx->cso); + cso_restore_framebuffer(ctx->cso); + cso_restore_vertex_shader(ctx->cso); + cso_restore_fragment_shader(ctx->cso); + cso_restore_viewport(ctx->cso); + + pipe_surface_reference(&dst_surf, NULL); +} + +void renderer_copy_surface(struct renderer *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, unsigned filter) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_buffer *buf; + struct pipe_texture texTemp, *tex; + struct pipe_surface *texSurf; + struct pipe_framebuffer_state fb; + struct st_framebuffer *stfb = ctx->owner->draw_buffer; + const int srcW = abs(srcX1 - srcX0); + const int srcH = abs(srcY1 - srcY0); + const int srcLeft = MIN2(srcX0, srcX1); + const int srcTop = MIN2(srcY0, srcY1); + + assert(filter == PIPE_TEX_MIPFILTER_NEAREST || + filter == PIPE_TEX_MIPFILTER_LINEAR); + + if (srcLeft != srcX0) { + /* left-right flip */ + int tmp = dstX0; + dstX0 = dstX1; + dstX1 = tmp; + } + + if (srcTop != srcY0) { + /* up-down flip */ + int tmp = dstY0; + dstY0 = dstY1; + dstY1 = tmp; + } + + assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); + + /* + * XXX for now we're always creating a temporary texture. + * Strictly speaking that's not always needed. + */ + + /* create temp texture */ + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = src->format; + texTemp.last_level = 0; + texTemp.width[0] = srcW; + texTemp.height[0] = srcH; + texTemp.depth[0] = 1; + pf_get_block(src->format, &texTemp.block); + + tex = screen->texture_create(screen, &texTemp); + if (!tex) + return; + + texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* load temp texture */ + pipe->surface_copy(pipe, + texSurf, 0, 0, /* dest */ + src, srcLeft, srcTop, /* src */ + srcW, srcH); /* size */ + + /* free the surface, update the texture if necessary.*/ + screen->tex_surface_destroy(texSurf); + + /* save state (restored below) */ + cso_save_blend(ctx->cso); + cso_save_samplers(ctx->cso); + cso_save_sampler_textures(ctx->cso); + cso_save_framebuffer(ctx->cso); + cso_save_fragment_shader(ctx->cso); + cso_save_vertex_shader(ctx->cso); + cso_save_viewport(ctx->cso); + + /* set misc state we care about */ + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + cso_set_blend(ctx->cso, &blend); + } + + vg_set_viewport(ctx->owner, VEGA_Y0_TOP); + + /* sampler */ + { + struct pipe_sampler_state sampler; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; + cso_single_sampler(ctx->cso, 0, &sampler); + cso_single_sampler_done(ctx->cso); + } + + /* texture */ + cso_set_sampler_textures(ctx->cso, 1, &tex); + + /* shaders */ + cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_vertex_shader_handle(ctx->cso, vg_texture_vs(ctx->owner)); + + /* drawing dest */ + if (stfb->strb->surface != dst) { + memset(&fb, 0, sizeof(fb)); + fb.width = dst->width; + fb.height = dst->height; + fb.nr_cbufs = 1; + fb.cbufs[0] = dst; + fb.zsbuf = stfb->dsrb->surface; + cso_set_framebuffer(ctx->cso, &fb); + } + + /* draw quad */ + buf = setup_vertex_data(ctx, + (float) dstX0, (float) dstY0, + (float) dstX1, (float) dstY1, z); + + if (buf) { + util_draw_vertex_buffer(ctx->pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference( &buf, + NULL ); + } + + + /* restore state we changed */ + cso_restore_blend(ctx->cso); + cso_restore_samplers(ctx->cso); + cso_restore_sampler_textures(ctx->cso); + cso_restore_framebuffer(ctx->cso); + cso_restore_fragment_shader(ctx->cso); + cso_restore_vertex_shader(ctx->cso); + cso_restore_viewport(ctx->cso); + + pipe_texture_reference(&tex, NULL); +} + +void renderer_texture_quad(struct renderer *r, + struct pipe_texture *tex, + VGfloat x1offset, VGfloat y1offset, + VGfloat x2offset, VGfloat y2offset, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat x3, VGfloat y3, + VGfloat x4, VGfloat y4) +{ + struct pipe_context *pipe = r->pipe; + struct pipe_buffer *buf; + VGfloat s0, t0, s1, t1; + + assert(tex->width[0] != 0); + assert(tex->height[0] != 0); + + s0 = x1offset / tex->width[0]; + s1 = x2offset / tex->width[0]; + t0 = y1offset / tex->height[0]; + t1 = y2offset / tex->height[0]; + + cso_save_vertex_shader(r->cso); + /* shaders */ + cso_set_vertex_shader_handle(r->cso, vg_texture_vs(r->owner)); + + /* draw quad */ + buf = setup_vertex_data_qtex(r, x1, y1, x2, y2, x3, y3, x4, y4, + s0, t0, s1, t1, 0.0f); + + if (buf) { + util_draw_vertex_buffer(pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference(&buf, + NULL); + } + + cso_restore_vertex_shader(r->cso); +} diff --git a/src/gallium/state_trackers/vega/renderer.h b/src/gallium/state_trackers/vega/renderer.h new file mode 100644 index 00000000000..990cd32c313 --- /dev/null +++ b/src/gallium/state_trackers/vega/renderer.h @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef RENDERER_H +#define RENDERER_H + +#include "VG/openvg.h" + +struct renderer; + +struct vg_context; +struct pipe_texture; +struct pipe_surface; + +struct renderer *renderer_create(struct vg_context *owner); +void renderer_destroy(struct renderer *); + +void renderer_draw_quad(struct renderer *, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat depth); +void renderer_draw_texture(struct renderer *, + struct pipe_texture *texture, + VGfloat x1offset, VGfloat y1offset, + VGfloat x2offset, VGfloat y2offset, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2); +void renderer_texture_quad(struct renderer *, + struct pipe_texture *texture, + VGfloat x1offset, VGfloat y1offset, + VGfloat x2offset, VGfloat y2offset, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat x3, VGfloat y3, + VGfloat x4, VGfloat y4); +void renderer_copy_texture(struct renderer *r, + struct pipe_texture *src, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + struct pipe_texture *dst, + VGfloat dx1, VGfloat dy1, + VGfloat dx2, VGfloat dy2); +void renderer_copy_surface(struct renderer *r, + struct pipe_surface *src, + int sx1, int sy1, + int sx2, int sy2, + struct pipe_surface *dst, + int dx1, int dy1, + int dx2, int dy2, + float z, unsigned filter); + + +#endif diff --git a/src/gallium/state_trackers/vega/shader.c b/src/gallium/state_trackers/vega/shader.c new file mode 100644 index 00000000000..d9074a377b3 --- /dev/null +++ b/src/gallium/state_trackers/vega/shader.c @@ -0,0 +1,310 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "shader.h" + +#include "vg_context.h" +#include "shaders_cache.h" +#include "paint.h" +#include "mask.h" +#include "image.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "util/u_memory.h" + +#define MAX_CONSTANTS 20 + +struct shader { + struct vg_context *context; + + VGboolean masking; + struct vg_paint *paint; + struct vg_image *image; + + VGboolean drawing_image; + VGImageMode image_mode; + + float constants[MAX_CONSTANTS]; + struct pipe_constant_buffer cbuf; + struct pipe_shader_state fs_state; + void *fs; +}; + +struct shader * shader_create(struct vg_context *ctx) +{ + struct shader *shader = 0; + + shader = CALLOC_STRUCT(shader); + shader->context = ctx; + + return shader; +} + +void shader_destroy(struct shader *shader) +{ + free(shader); +} + +void shader_set_masking(struct shader *shader, VGboolean set) +{ + shader->masking = set; +} + +VGboolean shader_is_masking(struct shader *shader) +{ + return shader->masking; +} + +void shader_set_paint(struct shader *shader, struct vg_paint *paint) +{ + shader->paint = paint; +} + +struct vg_paint * shader_paint(struct shader *shader) +{ + return shader->paint; +} + + +static void setup_constant_buffer(struct shader *shader) +{ + struct vg_context *ctx = shader->context; + struct pipe_context *pipe = shader->context->pipe; + struct pipe_constant_buffer *cbuf = &shader->cbuf; + VGint param_bytes = paint_constant_buffer_size(shader->paint); + float temp_buf[MAX_CONSTANTS]; + + assert(param_bytes <= sizeof(temp_buf)); + paint_fill_constant_buffer(shader->paint, temp_buf); + + if (cbuf->buffer == NULL || + memcmp(temp_buf, shader->constants, param_bytes) != 0) + { + pipe_buffer_reference(&cbuf->buffer, NULL); + + memcpy(shader->constants, temp_buf, param_bytes); + cbuf->buffer = pipe_user_buffer_create(pipe->screen, + &shader->constants, + sizeof(shader->constants)); + } + + ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); +} + +static VGint blend_bind_samplers(struct vg_context *ctx, + struct pipe_sampler_state **samplers, + struct pipe_texture **textures) +{ + VGBlendMode bmode = ctx->state.vg.blend_mode; + + if (bmode == VG_BLEND_MULTIPLY || + bmode == VG_BLEND_SCREEN || + bmode == VG_BLEND_DARKEN || + bmode == VG_BLEND_LIGHTEN) { + struct st_framebuffer *stfb = ctx->draw_buffer; + + vg_prepare_blend_surface(ctx); + + samplers[2] = &ctx->blend_sampler; + textures[2] = stfb->blend_texture; + + if (!samplers[0] || !textures[0]) { + samplers[1] = samplers[2]; + textures[1] = textures[2]; + } + if (!samplers[1] || !textures[1]) { + samplers[1] = samplers[0]; + textures[1] = textures[0]; + } + + return 1; + } + return 0; +} + +static void setup_samplers(struct shader *shader) +{ + struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + struct vg_context *ctx = shader->context; + /* a little wonky: we use the num as a boolean that just says + * whether any sampler/textures have been set. the actual numbering + * for samplers is always the same: + * 0 - paint sampler/texture for gradient/pattern + * 1 - mask sampler/texture + * 2 - blend sampler/texture + * 3 - image sampler/texture + * */ + VGint num = 0; + + samplers[0] = NULL; + samplers[1] = NULL; + samplers[2] = NULL; + samplers[3] = NULL; + textures[0] = NULL; + textures[1] = NULL; + textures[2] = NULL; + textures[3] = NULL; + + num += paint_bind_samplers(shader->paint, samplers, textures); + num += mask_bind_samplers(samplers, textures); + num += blend_bind_samplers(ctx, samplers, textures); + if (shader->drawing_image && shader->image) + num += image_bind_samplers(shader->image, samplers, textures); + + if (num) { + cso_set_samplers(ctx->cso_context, 4, (const struct pipe_sampler_state **)samplers); + cso_set_sampler_textures(ctx->cso_context, 4, textures); + } +} + +static INLINE VGboolean is_format_bw(struct shader *shader) +{ +#if 0 + struct vg_context *ctx = shader->context; + struct st_framebuffer *stfb = ctx->draw_buffer; +#endif + + if (shader->drawing_image && shader->image) { + if (shader->image->format == VG_BW_1) + return VG_TRUE; + } + + return VG_FALSE; +} + +static void setup_shader_program(struct shader *shader) +{ + struct vg_context *ctx = shader->context; + VGint shader_id = 0; + VGBlendMode blend_mode = ctx->state.vg.blend_mode; + VGboolean black_white = is_format_bw(shader); + + /* 1st stage: fill */ + if (!shader->drawing_image || + (shader->image_mode == VG_DRAW_IMAGE_MULTIPLY || shader->image_mode == VG_DRAW_IMAGE_STENCIL)) { + switch(paint_type(shader->paint)) { + case VG_PAINT_TYPE_COLOR: + shader_id |= VEGA_SOLID_FILL_SHADER; + break; + case VG_PAINT_TYPE_LINEAR_GRADIENT: + shader_id |= VEGA_LINEAR_GRADIENT_SHADER; + break; + case VG_PAINT_TYPE_RADIAL_GRADIENT: + shader_id |= VEGA_RADIAL_GRADIENT_SHADER; + break; + case VG_PAINT_TYPE_PATTERN: + shader_id |= VEGA_PATTERN_SHADER; + break; + + default: + abort(); + } + } + + /* second stage image */ + if (shader->drawing_image) { + switch(shader->image_mode) { + case VG_DRAW_IMAGE_NORMAL: + shader_id |= VEGA_IMAGE_NORMAL_SHADER; + break; + case VG_DRAW_IMAGE_MULTIPLY: + shader_id |= VEGA_IMAGE_MULTIPLY_SHADER; + break; + case VG_DRAW_IMAGE_STENCIL: + shader_id |= VEGA_IMAGE_STENCIL_SHADER; + break; + default: + debug_printf("Unknown image mode!"); + } + } + + if (shader->masking) + shader_id |= VEGA_MASK_SHADER; + + switch(blend_mode) { + case VG_BLEND_MULTIPLY: + shader_id |= VEGA_BLEND_MULTIPLY_SHADER; + break; + case VG_BLEND_SCREEN: + shader_id |= VEGA_BLEND_SCREEN_SHADER; + break; + case VG_BLEND_DARKEN: + shader_id |= VEGA_BLEND_DARKEN_SHADER; + break; + case VG_BLEND_LIGHTEN: + shader_id |= VEGA_BLEND_LIGHTEN_SHADER; + break; + default: + /* handled by pipe_blend_state */ + break; + } + + if (black_white) + shader_id |= VEGA_BW_SHADER; + + shader->fs = shaders_cache_fill(ctx->sc, shader_id); + cso_set_fragment_shader_handle(ctx->cso_context, shader->fs); +} + + +void shader_bind(struct shader *shader) +{ + /* first resolve the real paint type */ + paint_resolve_type(shader->paint); + + setup_constant_buffer(shader); + setup_samplers(shader); + setup_shader_program(shader); +} + +void shader_set_image_mode(struct shader *shader, VGImageMode image_mode) +{ + shader->image_mode = image_mode; +} + +VGImageMode shader_image_mode(struct shader *shader) +{ + return shader->image_mode; +} + +void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image) +{ + shader->drawing_image = drawing_image; +} + +VGboolean shader_drawing_image(struct shader *shader) +{ + return shader->drawing_image; +} + +void shader_set_image(struct shader *shader, struct vg_image *img) +{ + shader->image = img; +} diff --git a/src/gallium/state_trackers/vega/shader.h b/src/gallium/state_trackers/vega/shader.h new file mode 100644 index 00000000000..847eee6a310 --- /dev/null +++ b/src/gallium/state_trackers/vega/shader.h @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SHADER_H +#define SHADER_H + +#include "VG/openvg.h" + +struct shader; +struct vg_paint; +struct vg_context; +struct vg_image; + +struct shader *shader_create(struct vg_context *context); +void shader_destroy(struct shader *shader); + +void shader_set_masking(struct shader *shader, VGboolean set); +VGboolean shader_is_masking(struct shader *shader); + +void shader_set_paint(struct shader *shader, struct vg_paint *paint); +struct vg_paint *shader_paint(struct shader *shader); + +void shader_set_image_mode(struct shader *shader, VGImageMode image_mode); +VGImageMode shader_image_mode(struct shader *shader); + +void shader_set_drawing_image(struct shader *shader, VGboolean drawing_image); +VGboolean shader_drawing_image(struct shader *shader); + +void shader_set_image(struct shader *shader, struct vg_image *img); + +void shader_bind(struct shader *shader); + +#endif diff --git a/src/gallium/state_trackers/vega/shaders_cache.c b/src/gallium/state_trackers/vega/shaders_cache.c new file mode 100644 index 00000000000..fd0831fab1f --- /dev/null +++ b/src/gallium/state_trackers/vega/shaders_cache.c @@ -0,0 +1,439 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "shaders_cache.h" + +#include "vg_context.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_screen.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi/tgsi_text.h" + +#include "util/u_memory.h" +#include "util/u_math.h" +#include "util/u_debug.h" +#include "cso_cache/cso_hash.h" +#include "cso_cache/cso_context.h" + +#include "VG/openvg.h" + +#include "asm_fill.h" + +/* Essentially we construct an ubber-shader based on the state + * of the pipeline. The stages are: + * 1) Fill (mandatory, solid color/gradient/pattern/image draw) + * 2) Image composition (image mode multiply and stencil) + * 3) Mask + * 4) Extended blend (multiply/screen/darken/lighten) + * 5) Premultiply/Unpremultiply + * 6) Color transform (to black and white) + */ +#define SHADER_STAGES 6 + +struct cached_shader { + void *driver_shader; + struct pipe_shader_state state; +}; + +struct shaders_cache { + struct vg_context *pipe; + + struct cso_hash *hash; +}; + + +static INLINE struct tgsi_token *tokens_from_assembly(const char *txt, int num_tokens) +{ + struct tgsi_token *tokens; + + tokens = (struct tgsi_token *) MALLOC(num_tokens * sizeof(tokens[0])); + + tgsi_text_translate(txt, tokens, num_tokens); + +#if DEBUG_SHADERS + tgsi_dump(tokens, 0); +#endif + + return tokens; +} + +#define ALL_FILLS (VEGA_SOLID_FILL_SHADER | \ + VEGA_LINEAR_GRADIENT_SHADER | \ + VEGA_RADIAL_GRADIENT_SHADER | \ + VEGA_PATTERN_SHADER | \ + VEGA_IMAGE_NORMAL_SHADER) + + +/* +static const char max_shader_preamble[] = + "FRAG1.1\n" + "DCL IN[0], POSITION, LINEAR\n" + "DCL IN[1], GENERIC[0], PERSPECTIVE\n" + "DCL OUT[0], COLOR, CONSTANT\n" + "DCL CONST[0..9], CONSTANT\n" + "DCL TEMP[0..9], CONSTANT\n" + "DCL SAMP[0..9], CONSTANT\n"; + + max_shader_preamble strlen == 175 +*/ +#define MAX_PREAMBLE 175 + +static INLINE VGint range_min(VGint min, VGint current) +{ + if (min < 0) + min = current; + else + min = MIN2(min, current); + return min; +} + +static INLINE VGint range_max(VGint max, VGint current) +{ + return MAX2(max, current); +} + +static void +create_preamble(char *txt, + const struct shader_asm_info *shaders[SHADER_STAGES], + int num_shaders) +{ + VGboolean declare_input = VG_FALSE; + VGint start_const = -1, end_const = 0; + VGint start_temp = -1, end_temp = 0; + VGint start_sampler = -1, end_sampler = 0; + VGint i; + VGint num_consts, num_temps, num_samplers; + + for (i = 0; i < num_shaders; ++i) { + if (shaders[i]->num_consts) + start_const = range_min(start_const, shaders[i]->start_const); + if (shaders[i]->num_temps) + start_temp = range_min(start_temp, shaders[i]->start_temp); + if (shaders[i]->num_samplers) + start_sampler = range_min(start_sampler, shaders[i]->start_sampler); + + end_const = range_max(end_const, shaders[i]->start_const + + shaders[i]->num_consts); + end_temp = range_max(end_temp, shaders[i]->start_temp + + shaders[i]->num_temps); + end_sampler = range_max(end_sampler, shaders[i]->start_sampler + + shaders[i]->num_samplers); + if (shaders[i]->needs_position) + declare_input = VG_TRUE; + } + /* if they're still unitialized, initialize them */ + if (start_const < 0) + start_const = 0; + if (start_temp < 0) + start_temp = 0; + if (start_sampler < 0) + start_sampler = 0; + + num_consts = end_const - start_const; + num_temps = end_temp - start_temp; + num_samplers = end_sampler - start_sampler; + /* end exclusive */ + --end_const; + --end_temp; + --end_sampler; + + sprintf(txt, "FRAG1.1\n"); + + if (declare_input) { + sprintf(txt + strlen(txt), "DCL IN[0], POSITION, LINEAR\n"); + sprintf(txt + strlen(txt), "DCL IN[1], GENERIC[0], PERSPECTIVE\n"); + } + + /* we always have a color output */ + sprintf(txt + strlen(txt), "DCL OUT[0], COLOR, CONSTANT\n"); + + if (num_consts > 1) + sprintf(txt + strlen(txt), "DCL CONST[%d..%d], CONSTANT\n", start_const, end_const); + else if (num_consts == 1) + sprintf(txt + strlen(txt), "DCL CONST[%d], CONSTANT\n", start_const); + + if (num_temps > 1) + sprintf(txt + strlen(txt), "DCL TEMP[%d..%d], CONSTANT\n", start_temp, end_temp); + else if (num_temps > 1) + sprintf(txt + strlen(txt), "DCL TEMP[%d], CONSTANT\n", start_temp); + + if (num_samplers > 1) + sprintf(txt + strlen(txt), "DCL SAMP[%d..%d], CONSTANT\n", start_sampler, end_sampler); + else if (num_samplers == 1) + sprintf(txt + strlen(txt), "DCL SAMP[%d], CONSTANT\n", start_sampler); +} + +static void * +combine_shaders(const struct shader_asm_info *shaders[SHADER_STAGES], int num_shaders, + struct pipe_context *pipe, + struct pipe_shader_state *shader) +{ + char *combined_txt; + int combined_len = MAX_PREAMBLE; + int combined_tokens = 0; + int i = 0; + int current_shader = 0; + int current_len; + + for (i = 0; i < num_shaders; ++i) { + combined_len += strlen(shaders[i]->txt); + combined_tokens += shaders[i]->num_tokens; + } + /* add for the %s->TEMP[0] substitutions */ + combined_len += num_shaders * 7 /*TEMP[0]*/ + 4 /*"END\n"*/; + + combined_txt = (char*)malloc(combined_len); + combined_txt[0] = '\0'; + + create_preamble(combined_txt, shaders, num_shaders); + + while (current_shader < num_shaders) { + const char temp[] = "TEMP[0]"; + const char out[] = "OUT[0]"; + const char *subst = temp; + + current_len = strlen(combined_txt); + + /* if the last shader then output */ + if (current_shader + 1 == num_shaders) + subst = out; + + snprintf(combined_txt + current_len, + combined_len - current_len, + shaders[current_shader]->txt, + subst); + ++current_shader; + } + + + current_len = strlen(combined_txt); + snprintf(combined_txt + current_len, + combined_len - current_len, + "END\n"); + + debug_printf("Combined shader is : \n%s\n", + combined_txt); + + shader->tokens = tokens_from_assembly( + combined_txt, combined_tokens); + + free(combined_txt); + + return pipe->create_fs_state(pipe, shader); +} + +static void * +create_shader(struct pipe_context *pipe, + int id, + struct pipe_shader_state *shader) +{ + int idx = 0; + const struct shader_asm_info * shaders[SHADER_STAGES]; + + /* the shader has to have a fill */ + debug_assert(id & ALL_FILLS); + + /* first stage */ + if (id & VEGA_SOLID_FILL_SHADER) { + debug_assert(idx == 0); + shaders[idx] = &shaders_asm[0]; + debug_assert(shaders_asm[0].id == VEGA_SOLID_FILL_SHADER); + ++idx; + } + if ((id & VEGA_LINEAR_GRADIENT_SHADER)) { + debug_assert(idx == 0); + shaders[idx] = &shaders_asm[1]; + debug_assert(shaders_asm[1].id == VEGA_LINEAR_GRADIENT_SHADER); + ++idx; + } + if ((id & VEGA_RADIAL_GRADIENT_SHADER)) { + debug_assert(idx == 0); + shaders[idx] = &shaders_asm[2]; + debug_assert(shaders_asm[2].id == VEGA_RADIAL_GRADIENT_SHADER); + ++idx; + } + if ((id & VEGA_PATTERN_SHADER)) { + debug_assert(idx == 0); + debug_assert(shaders_asm[3].id == VEGA_PATTERN_SHADER); + shaders[idx] = &shaders_asm[3]; + ++idx; + } + if ((id & VEGA_IMAGE_NORMAL_SHADER)) { + debug_assert(idx == 0); + debug_assert(shaders_asm[4].id == VEGA_IMAGE_NORMAL_SHADER); + shaders[idx] = &shaders_asm[4]; + ++idx; + } + + /* second stage */ + if ((id & VEGA_IMAGE_MULTIPLY_SHADER)) { + debug_assert(shaders_asm[5].id == VEGA_IMAGE_MULTIPLY_SHADER); + shaders[idx] = &shaders_asm[5]; + ++idx; + } else if ((id & VEGA_IMAGE_STENCIL_SHADER)) { + debug_assert(shaders_asm[6].id == VEGA_IMAGE_STENCIL_SHADER); + shaders[idx] = &shaders_asm[6]; + ++idx; + } + + /* third stage */ + if ((id & VEGA_MASK_SHADER)) { + debug_assert(idx == 1); + debug_assert(shaders_asm[7].id == VEGA_MASK_SHADER); + shaders[idx] = &shaders_asm[7]; + ++idx; + } + + /* fourth stage */ + if ((id & VEGA_BLEND_MULTIPLY_SHADER)) { + debug_assert(shaders_asm[8].id == VEGA_BLEND_MULTIPLY_SHADER); + shaders[idx] = &shaders_asm[8]; + ++idx; + } else if ((id & VEGA_BLEND_SCREEN_SHADER)) { + debug_assert(shaders_asm[9].id == VEGA_BLEND_SCREEN_SHADER); + shaders[idx] = &shaders_asm[9]; + ++idx; + } else if ((id & VEGA_BLEND_DARKEN_SHADER)) { + debug_assert(shaders_asm[10].id == VEGA_BLEND_DARKEN_SHADER); + shaders[idx] = &shaders_asm[10]; + ++idx; + } else if ((id & VEGA_BLEND_LIGHTEN_SHADER)) { + debug_assert(shaders_asm[11].id == VEGA_BLEND_LIGHTEN_SHADER); + shaders[idx] = &shaders_asm[11]; + ++idx; + } + + /* fifth stage */ + if ((id & VEGA_PREMULTIPLY_SHADER)) { + debug_assert(shaders_asm[12].id == VEGA_PREMULTIPLY_SHADER); + shaders[idx] = &shaders_asm[12]; + ++idx; + } else if ((id & VEGA_UNPREMULTIPLY_SHADER)) { + debug_assert(shaders_asm[13].id == VEGA_UNPREMULTIPLY_SHADER); + shaders[idx] = &shaders_asm[13]; + ++idx; + } + + /* sixth stage */ + if ((id & VEGA_BW_SHADER)) { + debug_assert(shaders_asm[14].id == VEGA_BW_SHADER); + shaders[idx] = &shaders_asm[14]; + ++idx; + } + + return combine_shaders(shaders, idx, pipe, shader); +} + +/*************************************************/ + +struct shaders_cache * shaders_cache_create(struct vg_context *vg) +{ + struct shaders_cache *sc = CALLOC_STRUCT(shaders_cache); + + sc->pipe = vg; + sc->hash = cso_hash_create(); + + return sc; +} + +void shaders_cache_destroy(struct shaders_cache *sc) +{ + struct cso_hash_iter iter = cso_hash_first_node(sc->hash); + + while (!cso_hash_iter_is_null(iter)) { + struct cached_shader *cached = + (struct cached_shader *)cso_hash_iter_data(iter); + cso_delete_fragment_shader(sc->pipe->cso_context, + cached->driver_shader); + iter = cso_hash_erase(sc->hash, iter); + } + + cso_hash_delete(sc->hash); + free(sc); +} + +void * shaders_cache_fill(struct shaders_cache *sc, + int shader_key) +{ + VGint key = shader_key; + struct cached_shader *cached; + struct cso_hash_iter iter = cso_hash_find(sc->hash, key); + + if (cso_hash_iter_is_null(iter)) { + cached = CALLOC_STRUCT(cached_shader); + cached->driver_shader = create_shader(sc->pipe->pipe, key, &cached->state); + + cso_hash_insert(sc->hash, key, cached); + + return cached->driver_shader; + } + + cached = (struct cached_shader *)cso_hash_iter_data(iter); + + assert(cached->driver_shader); + return cached->driver_shader; +} + +struct vg_shader * shader_create_from_text(struct pipe_context *pipe, + const char *txt, int num_tokens, + int type) +{ + struct vg_shader *shader = (struct vg_shader *)malloc( + sizeof(struct vg_shader)); + struct tgsi_token *tokens = tokens_from_assembly(txt, num_tokens); + struct pipe_shader_state state; + + debug_assert(type == PIPE_SHADER_VERTEX || + type == PIPE_SHADER_FRAGMENT); + + state.tokens = tokens; + shader->type = type; + shader->tokens = tokens; + + if (type == PIPE_SHADER_FRAGMENT) + shader->driver = pipe->create_fs_state(pipe, &state); + else + shader->driver = pipe->create_vs_state(pipe, &state); + return shader; +} + +void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader) +{ + if (shader->type == PIPE_SHADER_FRAGMENT) + cso_delete_fragment_shader(ctx->cso_context, shader->driver); + else + cso_delete_vertex_shader(ctx->cso_context, shader->driver); + free(shader->tokens); + free(shader); +} diff --git a/src/gallium/state_trackers/vega/shaders_cache.h b/src/gallium/state_trackers/vega/shaders_cache.h new file mode 100644 index 00000000000..feca58b61a9 --- /dev/null +++ b/src/gallium/state_trackers/vega/shaders_cache.h @@ -0,0 +1,77 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SHADERS_CACHE_H +#define SHADERS_CACHE_H + + +struct vg_context; +struct pipe_context; +struct tgsi_token; +struct shaders_cache; + +enum VegaShaderType { + VEGA_SOLID_FILL_SHADER = 1 << 0, + VEGA_LINEAR_GRADIENT_SHADER = 1 << 1, + VEGA_RADIAL_GRADIENT_SHADER = 1 << 2, + VEGA_PATTERN_SHADER = 1 << 3, + VEGA_IMAGE_NORMAL_SHADER = 1 << 4, + VEGA_IMAGE_MULTIPLY_SHADER = 1 << 5, + VEGA_IMAGE_STENCIL_SHADER = 1 << 6, + + VEGA_MASK_SHADER = 1 << 7, + + VEGA_BLEND_MULTIPLY_SHADER = 1 << 8, + VEGA_BLEND_SCREEN_SHADER = 1 << 9, + VEGA_BLEND_DARKEN_SHADER = 1 << 10, + VEGA_BLEND_LIGHTEN_SHADER = 1 << 11, + + VEGA_PREMULTIPLY_SHADER = 1 << 12, + VEGA_UNPREMULTIPLY_SHADER = 1 << 13, + + VEGA_BW_SHADER = 1 << 14 +}; + +struct vg_shader { + void *driver; + struct tgsi_token *tokens; + int type;/* PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT */ +}; + +struct shaders_cache *shaders_cache_create(struct vg_context *pipe); +void shaders_cache_destroy(struct shaders_cache *sc); +void *shaders_cache_fill(struct shaders_cache *sc, + int shader_key); + +struct vg_shader *shader_create_from_text(struct pipe_context *pipe, + const char *txt, int num_tokens, + int type); + +void vg_shader_destroy(struct vg_context *ctx, struct vg_shader *shader); + + + +#endif diff --git a/src/gallium/state_trackers/vega/st_inlines.h b/src/gallium/state_trackers/vega/st_inlines.h new file mode 100644 index 00000000000..1f331dfcdb7 --- /dev/null +++ b/src/gallium/state_trackers/vega/st_inlines.h @@ -0,0 +1,159 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Functions for checking if buffers/textures are referenced when we need + * to read/write from/to them. Flush when needed. + */ + +#ifndef ST_INLINES_H +#define ST_INLINES_H + +#include "vg_context.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_state.h" + +static INLINE struct pipe_transfer * +st_cond_flush_get_tex_transfer(struct vg_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + unsigned int zslice, + enum pipe_transfer_usage usage, + unsigned int x, unsigned int y, + unsigned int w, unsigned int h) +{ + struct pipe_screen *screen = st->pipe->screen; + struct pipe_context *pipe = st->pipe; + unsigned referenced = + pipe->is_texture_referenced(pipe, pt, face, level); + + if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) || + usage == PIPE_TRANSFER_WRITE || + usage == PIPE_TRANSFER_READ_WRITE)) + vgFlush(); + + return screen->get_tex_transfer(screen, pt, face, level, zslice, usage, + x, y, w, h); +} + +static INLINE struct pipe_transfer * +st_no_flush_get_tex_transfer(struct vg_context *st, + struct pipe_texture *pt, + unsigned int face, + unsigned int level, + unsigned int zslice, + enum pipe_transfer_usage usage, + unsigned int x, unsigned int y, + unsigned int w, unsigned int h) +{ + struct pipe_screen *screen = st->pipe->screen; + + return screen->get_tex_transfer(screen, pt, face, level, + zslice, usage, x, y, w, h); +} + +static INLINE void * +st_cond_flush_pipe_buffer_map(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int map_flags) +{ + struct pipe_context *pipe = st->pipe; + unsigned int referenced = pipe->is_buffer_referenced(pipe, buf); + + if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) || + (map_flags & PIPE_BUFFER_USAGE_CPU_WRITE))) + vgFlush(); + + return pipe_buffer_map(pipe->screen, buf, map_flags); +} + +static INLINE void * +st_no_flush_pipe_buffer_map(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int map_flags) +{ + return pipe_buffer_map(st->pipe->screen, buf, map_flags); +} + + +static INLINE void +st_cond_flush_pipe_buffer_write(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + const void * data) +{ + struct pipe_context *pipe = st->pipe; + + if (pipe->is_buffer_referenced(pipe, buf)) + vgFlush(); + + pipe_buffer_write(pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_no_flush_pipe_buffer_write(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + const void * data) +{ + pipe_buffer_write(st->pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_cond_flush_pipe_buffer_read(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + void * data) +{ + struct pipe_context *pipe = st->pipe; + + if (pipe->is_buffer_referenced(pipe, buf) & PIPE_REFERENCED_FOR_WRITE) + vgFlush(); + + pipe_buffer_read(pipe->screen, buf, offset, size, data); +} + +static INLINE void +st_no_flush_pipe_buffer_read(struct vg_context *st, + struct pipe_buffer *buf, + unsigned int offset, + unsigned int size, + void * data) +{ + pipe_buffer_read(st->pipe->screen, buf, offset, size, data); +} + +#endif + diff --git a/src/gallium/state_trackers/vega/stroker.c b/src/gallium/state_trackers/vega/stroker.c new file mode 100644 index 00000000000..1b92d2b5c62 --- /dev/null +++ b/src/gallium/state_trackers/vega/stroker.c @@ -0,0 +1,1349 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "stroker.h" + +#include "path.h" +#include "vg_state.h" +#include "util_array.h" +#include "arc.h" +#include "bezier.h" +#include "matrix.h" +#include "path_utils.h" +#include "polygon.h" + +#include "math.h" + +#ifndef M_2PI +#define M_2PI 6.28318530717958647692528676655900576 +#endif + +#define STROKE_SEGMENTS 0 +#define STROKE_DEBUG 0 +#define DEBUG_EMITS 0 + +static const VGfloat curve_threshold = 0.25f; + +static const VGfloat zero_coords[] = {0.f, 0.f}; + +enum intersection_type { + NoIntersections, + BoundedIntersection, + UnboundedIntersection, +}; + +enum line_join_mode { + FlatJoin, + SquareJoin, + MiterJoin, + RoundJoin, + RoundCap +}; + +struct stroke_iterator { + void (*next)(struct stroke_iterator *); + VGboolean (*has_next)(struct stroke_iterator *); + + VGPathCommand (*current_command)(struct stroke_iterator *it); + void (*current_coords)(struct stroke_iterator *it, VGfloat *coords); + + VGint position; + VGint coord_position; + + const VGubyte *cmds; + const VGfloat *coords; + VGint num_commands; + VGint num_coords; + + struct polygon *curve_poly; + VGint curve_index; +}; + +static VGPathCommand stroke_itr_command(struct stroke_iterator *itr) +{ + return itr->current_command(itr); +} + +static void stroke_itr_coords(struct stroke_iterator *itr, VGfloat *coords) +{ + itr->current_coords(itr, coords); +} + +static void stroke_fw_itr_coords(struct stroke_iterator *itr, VGfloat *coords) +{ + if (itr->position >= itr->num_commands) + return; + switch (stroke_itr_command(itr)) { + case VG_MOVE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_LINE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_CUBIC_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + coords[2] = itr->coords[itr->coord_position + 2]; + coords[3] = itr->coords[itr->coord_position + 3]; + coords[4] = itr->coords[itr->coord_position + 4]; + coords[5] = itr->coords[itr->coord_position + 5]; + break; + default: + debug_assert(!"invalid command!\n"); + } +} + + +static void stroke_bw_itr_coords(struct stroke_iterator *itr, VGfloat *coords) +{ + if (itr->position >= itr->num_commands) + return; + switch (stroke_itr_command(itr)) { + case VG_MOVE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_LINE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_CUBIC_TO_ABS: + coords[0] = itr->coords[itr->coord_position + 4]; + coords[1] = itr->coords[itr->coord_position + 5]; + coords[2] = itr->coords[itr->coord_position + 2]; + coords[3] = itr->coords[itr->coord_position + 3]; + coords[4] = itr->coords[itr->coord_position + 0]; + coords[5] = itr->coords[itr->coord_position + 1]; + break; + default: + debug_assert(!"invalid command!\n"); + } +} + + +static VGPathCommand stroke_fw_current_command(struct stroke_iterator *it) +{ + return it->cmds[it->position]; +} + +static VGPathCommand stroke_bw_current_command(struct stroke_iterator *it) +{ + VGPathCommand prev_cmd; + if (it->position == it->num_commands -1) + return VG_MOVE_TO_ABS; + + prev_cmd = it->cmds[it->position + 1]; + return prev_cmd; +} + +static VGboolean stroke_fw_has_next(struct stroke_iterator *itr) +{ + return itr->position < (itr->num_commands - 1); +} + +static VGboolean stroke_bw_has_next(struct stroke_iterator *itr) +{ + return itr->position > 0; +} + +static void stroke_fw_next(struct stroke_iterator *itr) +{ + VGubyte cmd; + debug_assert(stroke_fw_has_next(itr)); + + cmd = stroke_itr_command(itr); + + itr->coord_position += num_elements_for_segments(&cmd, 1); + ++itr->position; +} + +static void stroke_bw_next(struct stroke_iterator *itr) +{ + VGubyte cmd; + debug_assert(stroke_bw_has_next(itr)); + + --itr->position; + cmd = stroke_itr_command(itr); + + itr->coord_position -= num_elements_for_segments(&cmd, 1); +} + +static void stroke_itr_common_init(struct stroke_iterator *itr, + struct array *cmds, + struct array *coords) +{ + itr->cmds = (VGubyte*)cmds->data; + itr->num_commands = cmds->num_elements; + + itr->coords = (VGfloat*)coords->data; + itr->num_coords = coords->num_elements; +} + +static void stroke_forward_iterator(struct stroke_iterator *itr, + struct array *cmds, + struct array *coords) +{ + stroke_itr_common_init(itr, cmds, coords); + itr->position = 0; + itr->coord_position = 0; + + itr->next = stroke_fw_next; + itr->has_next = stroke_fw_has_next; + itr->current_command = stroke_fw_current_command; + itr->current_coords = stroke_fw_itr_coords; +} + +static void stroke_backward_iterator(struct stroke_iterator *itr, + struct array *cmds, + struct array *coords) +{ + VGubyte cmd; + stroke_itr_common_init(itr, cmds, coords); + itr->position = itr->num_commands - 1; + + cmd = stroke_bw_current_command(itr); + itr->coord_position = itr->num_coords - + num_elements_for_segments(&cmd, 1); + + itr->next = stroke_bw_next; + itr->has_next = stroke_bw_has_next; + itr->current_command = stroke_bw_current_command; + itr->current_coords = stroke_bw_itr_coords; +} + + + +static void stroke_flat_next(struct stroke_iterator *itr) +{ + VGubyte cmd; + + if (itr->curve_index >= 0) { + ++itr->curve_index; + if (itr->curve_index >= polygon_vertex_count(itr->curve_poly)) { + itr->curve_index = -1; + polygon_destroy(itr->curve_poly); + itr->curve_poly = 0; + } else + return; + } + debug_assert(stroke_fw_has_next(itr)); + + cmd = itr->cmds[itr->position]; + itr->coord_position += num_elements_for_segments(&cmd, 1); + ++itr->position; + + cmd = itr->cmds[itr->position]; + + if (cmd == VG_CUBIC_TO_ABS) { + struct bezier bezier; + VGfloat bez[8]; + + bez[0] = itr->coords[itr->coord_position - 2]; + bez[1] = itr->coords[itr->coord_position - 1]; + bez[2] = itr->coords[itr->coord_position]; + bez[3] = itr->coords[itr->coord_position + 1]; + bez[4] = itr->coords[itr->coord_position + 2]; + bez[5] = itr->coords[itr->coord_position + 3]; + bez[6] = itr->coords[itr->coord_position + 4]; + bez[7] = itr->coords[itr->coord_position + 5]; + + bezier_init(&bezier, + bez[0], bez[1], + bez[2], bez[3], + bez[4], bez[5], + bez[6], bez[7]); + /* skip the first one, it's the same as the prev point */ + itr->curve_index = 1; + if (itr->curve_poly) { + polygon_destroy(itr->curve_poly); + itr->curve_poly = 0; + } + itr->curve_poly = bezier_to_polygon(&bezier); + } +} + +static VGboolean stroke_flat_has_next(struct stroke_iterator *itr) +{ + return (itr->curve_index >= 0 && + itr->curve_index < (polygon_vertex_count(itr->curve_poly)-1)) + || itr->position < (itr->num_commands - 1); +} + +static VGPathCommand stroke_flat_current_command(struct stroke_iterator *it) +{ + if (it->cmds[it->position] == VG_CUBIC_TO_ABS) { + return VG_LINE_TO_ABS; + } + return it->cmds[it->position]; +} + +static void stroke_flat_itr_coords(struct stroke_iterator *itr, VGfloat *coords) +{ + if (itr->curve_index <= -1 && itr->position >= itr->num_commands) + return; + + if (itr->curve_index >= 0) { + polygon_vertex(itr->curve_poly, itr->curve_index, + coords); + return; + } + + switch (stroke_itr_command(itr)) { + case VG_MOVE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_LINE_TO_ABS: + coords[0] = itr->coords[itr->coord_position]; + coords[1] = itr->coords[itr->coord_position + 1]; + break; + case VG_CUBIC_TO_ABS: + default: + debug_assert(!"invalid command!\n"); + } +} + +static void stroke_flat_iterator(struct stroke_iterator *itr, + struct array *cmds, + struct array *coords) +{ + stroke_itr_common_init(itr, cmds, coords); + itr->position = 0; + itr->coord_position = 0; + + itr->next = stroke_flat_next; + itr->has_next = stroke_flat_has_next; + itr->current_command = stroke_flat_current_command; + itr->current_coords = stroke_flat_itr_coords; + itr->curve_index = -1; + itr->curve_poly = 0; +} + + +static INLINE VGboolean finite_coords4(const VGfloat *c) +{ + return + isfinite(c[0]) && isfinite(c[1]) && + isfinite(c[2]) && isfinite(c[3]); +} + +/* from Graphics Gems II */ +#define SAME_SIGNS(a, b) ((a) * (b) >= 0) +static VGboolean do_lines_intersect(VGfloat x1, VGfloat y1, VGfloat x2, VGfloat y2, + VGfloat x3, VGfloat y3, VGfloat x4, VGfloat y4) +{ + VGfloat a1, a2, b1, b2, c1, c2; /* Coefficients of line eqns */ + VGfloat r1, r2, r3, r4; /* 'sign' values */ + + a1 = y2 - y1; + b1 = x1 - x2; + c1 = x2 * y1 - x1 * y2; + + r3 = a1 * x3 + b1 * y3 + c1; + r4 = a1 * x4 + b1 * y4 + c1; + + if (r3 != 0 && r4 != 0 && SAME_SIGNS(r3, r4)) + return VG_FALSE; + + a2 = y4 - y3; + b2 = x3 - x4; + c2 = x4 * y3 - x3 * y4; + + r1 = a2 * x1 + b2 * y1 + c2; + r2 = a2 * x2 + b2 * y2 + c2; + + if (r1 != 0 && r2 != 0 && SAME_SIGNS(r1, r2)) + return VG_FALSE; + + return VG_TRUE; +} + +static INLINE VGfloat line_dx(const VGfloat *l) +{ + return l[2] - l[0]; +} + +static INLINE VGfloat line_dy(const VGfloat *l) +{ + return l[3] - l[1]; +} + +static INLINE VGfloat line_angle(const VGfloat *l) +{ + const VGfloat dx = line_dx(l); + const VGfloat dy = line_dy(l); + + const VGfloat theta = atan2(-dy, dx) * 360.0 / M_2PI; + + const VGfloat theta_normalized = theta < 0 ? theta + 360 : theta; + + if (floatsEqual(theta_normalized, 360.f)) + return 0; + else + return theta_normalized; +} + +static INLINE void line_set_length(VGfloat *l, VGfloat len) +{ + VGfloat uv[] = {l[0], l[1], l[2], l[3]}; + if (null_line(l)) + return; + line_normalize(uv); + l[2] = l[0] + line_dx(uv) * len; + l[3] = l[1] + line_dy(uv) * len; +} + +static INLINE void line_translate(VGfloat *l, VGfloat x, VGfloat y) +{ + l[0] += x; + l[1] += y; + l[2] += x; + l[3] += y; +} + +static INLINE VGfloat line_angle_to(const VGfloat *l1, + const VGfloat *l2) +{ + VGfloat a1, a2, delta, delta_normalized; + if (null_line(l1) || null_line(l1)) + return 0; + + a1 = line_angle(l1); + a2 = line_angle(l2); + + delta = a2 - a1; + delta_normalized = delta < 0 ? delta + 360 : delta; + + if (floatsEqual(delta, 360.f)) + return 0; + else + return delta_normalized; +} + +static INLINE VGfloat line_angles(const VGfloat *l1, + const VGfloat *l2) +{ + VGfloat cos_line, rad = 0; + + if (null_line(l1) || null_line(l2)) + return 0; + + cos_line = (line_dx(l1)*line_dx(l2) + line_dy(l1)*line_dy(l2)) / + (line_lengthv(l1)*line_lengthv(l2)); + rad = 0; + + if (cos_line >= -1.0 && cos_line <= 1.0) + rad = acos(cos_line); + return rad * 360 / M_2PI; +} + + +static INLINE VGfloat adapted_angle_on_x(const VGfloat *line) +{ + const VGfloat identity[] = {0, 0, 1, 0}; + VGfloat angle = line_angles(line, identity); + if (line_dy(line) > 0) + angle = 360 - angle; + return angle; +} + +static enum intersection_type line_intersect(const VGfloat *l1, + const VGfloat *l2, + float *intersection_point) +{ + VGfloat isect[2]; + enum intersection_type type; + VGboolean dx_zero, ldx_zero; + + if (null_line(l1) || null_line(l2) || + !finite_coords4(l1) || !finite_coords4(l2)) + return NoIntersections; + + type = do_lines_intersect(l1[0], l1[1], l1[2], l1[3], l2[0], l2[1], l2[2], l2[3]) + ? BoundedIntersection : UnboundedIntersection; + + dx_zero = floatsEqual(line_dx(l1) + 1, 1); + ldx_zero = floatsEqual(line_dx(l2) + 1, 1); + + /* one of the lines is vertical */ + if (dx_zero && ldx_zero) { + type = NoIntersections; + } else if (dx_zero) { + VGfloat la = line_dy(l2) / line_dx(l2); + isect[0] = l1[0]; + isect[1] = la * l1[0] + l2[1] - la * l2[0]; + } else if (ldx_zero) { + VGfloat ta = line_dy(l1) / line_dx(l1); + isect[0] = l2[0]; + isect[1] = ta * l2[0] + l1[1] - ta*l1[0]; + } else { + VGfloat x; + VGfloat ta = line_dy(l1) / line_dx(l1); + VGfloat la = line_dy(l2) / line_dx(l2); + if (ta == la) + return NoIntersections; + + x = ( - l2[1] + la * l2[0] + l1[1] - ta * l1[0] ) / (la - ta); + isect[0] = x; + isect[1] = ta*(x - l1[0]) + l1[1]; + } + if (intersection_point) { + intersection_point[0] = isect[0]; + intersection_point[1] = isect[1]; + } + return type; +} + +static INLINE enum line_join_mode stroker_join_mode(struct stroker *s) +{ + switch(s->join_style) { + case VG_JOIN_MITER: + return MiterJoin; + case VG_JOIN_ROUND: + return RoundJoin; + case VG_JOIN_BEVEL: + return FlatJoin; + default: + return FlatJoin; + } +} + +static INLINE enum line_join_mode stroker_cap_mode(struct stroker *s) +{ + switch(s->cap_style) { + case VG_CAP_BUTT: + return FlatJoin; + case VG_CAP_ROUND: + return RoundCap; + case VG_CAP_SQUARE: + return SquareJoin; + default: + return FlatJoin; + } +} + +void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y) +{ + VGubyte cmds = VG_MOVE_TO_ABS; + VGfloat coords[2] = {x, y}; +#if DEBUG_EMITS + debug_printf("emit move %f, %f\n", x, y); +#endif + stroker->back2_x = stroker->back1_x; + stroker->back2_y = stroker->back1_y; + stroker->back1_x = x; + stroker->back1_y = y; + + path_append_data(stroker->path, + 1, + &cmds, &coords); +} + +void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y) +{ + VGubyte cmds = VG_LINE_TO_ABS; + VGfloat coords[2] = {x, y}; +#if DEBUG_EMITS + debug_printf("emit line %f, %f\n", x, y); +#endif + stroker->back2_x = stroker->back1_x; + stroker->back2_y = stroker->back1_y; + stroker->back1_x = x; + stroker->back1_y = y; + path_append_data(stroker->path, + 1, + &cmds, &coords); +} + +void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1, + VGfloat px2, VGfloat py2, + VGfloat x, VGfloat y) +{ + VGubyte cmds = VG_CUBIC_TO_ABS; + VGfloat coords[6] = {px1, py1, px2, py2, x, y}; +#if DEBUG_EMITS + debug_printf("emit curve %f, %f, %f, %f, %f, %f\n", px1, py1, + px2, py2, x, y); +#endif + + if (px2 == x && py2 == y) { + if (px1 == x && py1 == y) { + stroker->back2_x = stroker->back1_x; + stroker->back2_y = stroker->back1_y; + } else { + stroker->back2_x = px1; + stroker->back2_y = py1; + } + } else { + stroker->back2_x = px2; + stroker->back2_y = py2; + } + stroker->back1_x = x; + stroker->back1_y = y; + + path_append_data(stroker->path, + 1, + &cmds, &coords); +} + +static INLINE void create_round_join(struct stroker *stroker, + VGfloat x1, VGfloat y1, + VGfloat x2, VGfloat y2, + VGfloat width, VGfloat height) +{ + struct arc arc; + struct matrix matrix; + + matrix_load_identity(&matrix); + + /*stroker_emit_line_to(stroker, nx, ny);*/ + + arc_init(&arc, VG_SCCWARC_TO_ABS, + x1, y1, x2, y2, width/2, height/2, 0); + arc_stroker_emit(&arc, stroker, &matrix); +} + + +static void create_joins(struct stroker *stroker, + VGfloat focal_x, VGfloat focal_y, + const VGfloat *next_line, enum line_join_mode join) +{ +#if DEBUG_EMITS + debug_printf("create_joins: focal=[%f, %f], next_line=[%f, %f,%f, %f]\n", + focal_x, focal_y, + next_line[0], next_line[1], next_line[2], next_line[3]); +#endif + /* if we're alredy connected do nothing */ + if (floatsEqual(stroker->back1_x, next_line[0]) && + floatsEqual(stroker->back1_y, next_line[1])) + return; + + if (join == FlatJoin) { + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + } else { + VGfloat prev_line[] = {stroker->back2_x, stroker->back2_y, + stroker->back1_x, stroker->back1_y}; + + VGfloat isect[2]; + enum intersection_type type = line_intersect(prev_line, next_line, isect); + + if (join == SquareJoin) { + VGfloat offset = stroker->stroke_width / 2; + VGfloat l1[4] = {prev_line[0], + prev_line[1], + prev_line[2], + prev_line[3]}; + VGfloat l2[4] = {next_line[2], + next_line[3], + next_line[0], + next_line[1]}; + + line_translate(l1, line_dx(l1), line_dy(l1)); + line_set_length(l1, offset); + + line_translate(l2, line_dx(l2), line_dy(l2)); + line_set_length(l2, offset); + + stroker_emit_line_to(stroker, l1[2], l1[3]); + stroker_emit_line_to(stroker, l2[2], l2[3]); + stroker_emit_line_to(stroker, l2[0], l2[1]); + } else if (join == RoundJoin) { + VGfloat offset = stroker->stroke_width / 2; + VGfloat short_cut[4] = {prev_line[2], prev_line[3], + next_line[0], next_line[1]}; + VGfloat angle = line_angles(prev_line, short_cut); + + if (type == BoundedIntersection || + (angle > 90 && !floatsEqual(angle, 90.f))) { + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + return; + } + create_round_join(stroker, prev_line[2], prev_line[3], + next_line[0], next_line[1], + offset * 2, offset * 2); + + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + } else if (join == RoundCap) { + VGfloat offset = stroker->stroke_width / 2; + VGfloat l1[4] = { prev_line[0], prev_line[1], + prev_line[2], prev_line[3] }; + VGfloat l2[4] = {focal_x, focal_y, + prev_line[2], prev_line[3]}; + + line_translate(l1, line_dx(l1), line_dy(l1)); + line_set_length(l1, KAPPA * offset); + + /* normal between prev_line and focal */ + line_translate(l2, -line_dy(l2), line_dx(l2)); + line_set_length(l2, KAPPA * offset); + + stroker_emit_curve_to(stroker, l1[2], l1[3], + l2[2], l2[3], + l2[0], l2[1]); + + l2[0] = l2[0]; + l2[1] = l2[1]; + l2[2] = l2[0] - line_dx(l2); + l2[3] = l2[1] - line_dy(l2); + + line_translate(l1, next_line[0] - l1[0], next_line[1] - l1[1]); + + stroker_emit_curve_to(stroker, + l2[2], l2[3], + l1[2], l1[3], + l1[0], l1[1]); + } else if (join == MiterJoin) { + VGfloat miter_line[4] = {stroker->back1_x, stroker->back1_y, + isect[0], isect[1]}; + VGfloat sl = (stroker->stroke_width * stroker->miter_limit); + VGfloat inside_line[4] = {prev_line[2], prev_line[3], + next_line[0], next_line[1]}; + VGfloat angle = line_angle_to(inside_line, prev_line); + + if (type == BoundedIntersection || + (angle > 90 && !floatsEqual(angle, 90.f))) { + /* + debug_printf("f = %f, nl = %f, pl = %f, is = %f\n", + focal_x, next_line[0], + prev_line[2], isect[0]);*/ + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + return; + } + + if (type == NoIntersections || line_lengthv(miter_line) > sl) { + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + } else { + stroker_emit_line_to(stroker, isect[0], isect[1]); + stroker_emit_line_to(stroker, next_line[0], next_line[1]); + } + } else { + debug_assert(!"create_joins bad join style"); + } + } +} + +static void stroker_add_segment(struct stroker *stroker, + VGPathCommand cmd, + const VGfloat *coords, + VGint num_coords) +{ + /* skip duplicated points */ + if (stroker->segments->num_elements && + stroker->last_cmd == cmd) { + VGfloat *data = stroker->control_points->data; + data += stroker->control_points->num_elements; + data -= num_coords; + switch (cmd) { + case VG_MOVE_TO_ABS: + if (floatsEqual(coords[0], data[0]) && + floatsEqual(coords[1], data[1])) + return; + break; + case VG_LINE_TO_ABS: + if (floatsEqual(coords[0], data[0]) && + floatsEqual(coords[1], data[1])) + return; + break; + case VG_CUBIC_TO_ABS: + if (floatsEqual(coords[0], data[0]) && + floatsEqual(coords[1], data[1]) && + floatsEqual(coords[2], data[2]) && + floatsEqual(coords[3], data[3]) && + floatsEqual(coords[4], data[4]) && + floatsEqual(coords[5], data[5])) + return; + break; + default: + debug_assert(!"Invalid stroke segment"); + } + } else if (stroker->last_cmd == VG_CUBIC_TO_ABS && + cmd == VG_LINE_TO_ABS) { + VGfloat *data = stroker->control_points->data; + data += stroker->control_points->num_elements; + data -= 2; + if (floatsEqual(coords[0], data[0]) && + floatsEqual(coords[1], data[1])) + return; + } + stroker->last_cmd = cmd; + array_append_data(stroker->segments, &cmd, 1); + array_append_data(stroker->control_points, coords, num_coords); +} + +void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y) +{ + VGfloat coords[2] = {x, y}; +#if STROKE_SEGMENTS + debug_printf("stroker_move_to(%f, %f)\n", x, y); +#endif + + if (stroker->segments->num_elements > 0) + stroker->process_subpath(stroker); + + array_reset(stroker->segments); + array_reset(stroker->control_points); + + stroker_add_segment(stroker, VG_MOVE_TO_ABS, coords, 2); +} + +void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y) +{ + VGfloat coords[] = {x, y}; + +#if STROKE_SEGMENTS + debug_printf("stroker_line_to(%f, %f)\n", x, y); +#endif + if (!stroker->segments->num_elements) + stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2); + + stroker_add_segment(stroker, VG_LINE_TO_ABS, coords, 2); +} + +void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1, + VGfloat px2, VGfloat py2, + VGfloat x, VGfloat y) +{ + VGfloat coords[] = {px1, py1, + px2, py2, + x, y}; +#if STROKE_SEGMENTS + debug_printf("stroker_curve_to(%f, %f, %f, %f, %f, %f)\n", + px1, py1, px2, py2, x, y); +#endif + if (!stroker->segments->num_elements) + stroker_add_segment(stroker, VG_MOVE_TO_ABS, zero_coords, 2); + + stroker_add_segment(stroker, VG_CUBIC_TO_ABS, coords, 6); +} + +static INLINE VGboolean is_segment_null(VGPathCommand cmd, + VGfloat *coords, + VGfloat *res) +{ + switch(cmd) { + case VG_MOVE_TO_ABS: + case VG_LINE_TO_ABS: + return floatsEqual(coords[0], res[0]) && + floatsEqual(coords[1], res[1]); + break; + case VG_CUBIC_TO_ABS: + return floatsEqual(coords[0], res[0]) && + floatsEqual(coords[1], res[1]) && + floatsEqual(coords[2], res[0]) && + floatsEqual(coords[3], res[1]) && + floatsEqual(coords[4], res[0]) && + floatsEqual(coords[5], res[1]); + break; + default: + assert(0); + } + return VG_FALSE; +} + +static VGboolean vg_stroke_outline(struct stroke_iterator *it, + struct stroker *stroker, + VGboolean cap_first, + VGfloat *start_tangent) +{ + const int MAX_OFFSET = 16; + struct bezier offset_curves[MAX_OFFSET]; + VGPathCommand first_element; + VGfloat start[2], prev[2]; + VGboolean first = VG_TRUE; + VGfloat offset; + + first_element = stroke_itr_command(it); + if (first_element != VG_MOVE_TO_ABS) { + stroker_emit_move_to(stroker, 0.f, 0.f); + prev[0] = 0.f; + prev[1] = 0.f; + } + stroke_itr_coords(it, start); +#if STROKE_DEBUG + debug_printf(" -> (side) [%.2f, %.2f]\n", + start[0], + start[1]); +#endif + + prev[0] = start[0]; + prev[1] = start[1]; + + offset = stroker->stroke_width / 2; + + if (!it->has_next(it)) { + /* single point */ + + return VG_TRUE; + } + + while (it->has_next(it)) { + VGPathCommand cmd; + VGfloat coords[8]; + + it->next(it); + cmd = stroke_itr_command(it); + stroke_itr_coords(it, coords); + + if (cmd == VG_LINE_TO_ABS) { + VGfloat line[4] = {prev[0], prev[1], coords[0], coords[1]}; + VGfloat normal[4]; + line_normal(line, normal); + +#if STROKE_DEBUG + debug_printf("\n ---> (side) lineto [%.2f, %.2f]\n", coords[0], coords[1]); +#endif + line_set_length(normal, offset); + line_translate(line, line_dx(normal), line_dy(normal)); + + /* if we are starting a new subpath, move to correct starting point */ + if (first) { + if (cap_first) + create_joins(stroker, prev[0], prev[1], line, + stroker_cap_mode(stroker)); + else + stroker_emit_move_to(stroker, line[0], line[1]); + memcpy(start_tangent, line, + sizeof(VGfloat) * 4); + first = VG_FALSE; + } else { + create_joins(stroker, prev[0], prev[1], line, + stroker_join_mode(stroker)); + } + + /* add the stroke for this line */ + stroker_emit_line_to(stroker, line[2], line[3]); + prev[0] = coords[0]; + prev[1] = coords[1]; + } else if (cmd == VG_CUBIC_TO_ABS) { +#if STROKE_DEBUG + debug_printf("\n ---> (side) cubicTo [%.2f, %.2f]\n", + coords[4], + coords[5]); +#endif + struct bezier bezier; + int count; + + bezier_init(&bezier, + prev[0], prev[1], coords[0], coords[1], + coords[2], coords[3], coords[4], coords[5]); + + count = bezier_translate_by_normal(&bezier, + offset_curves, + MAX_OFFSET, + offset, + curve_threshold); + + if (count) { + /* if we are starting a new subpath, move to correct starting point */ + VGfloat tangent[4]; + VGint i; + + bezier_start_tangent(&bezier, tangent); + line_translate(tangent, + offset_curves[0].x1 - bezier.x1, + offset_curves[0].y1 - bezier.y1); + if (first) { + VGfloat pt[2] = {offset_curves[0].x1, + offset_curves[0].y1}; + + if (cap_first) { + create_joins(stroker, prev[0], prev[1], tangent, + stroker_cap_mode(stroker)); + } else { + stroker_emit_move_to(stroker, pt[0], pt[1]); + } + start_tangent[0] = tangent[0]; + start_tangent[1] = tangent[1]; + start_tangent[2] = tangent[2]; + start_tangent[3] = tangent[3]; + first = VG_FALSE; + } else { + create_joins(stroker, prev[0], prev[1], tangent, + stroker_join_mode(stroker)); + } + + /* add these beziers */ + for (i = 0; i < count; ++i) { + struct bezier *bez = &offset_curves[i]; + stroker_emit_curve_to(stroker, + bez->x2, bez->y2, + bez->x3, bez->y3, + bez->x4, bez->y4); + } + } + + prev[0] = coords[4]; + prev[1] = coords[5]; + } + } + + if (floatsEqual(start[0], prev[0]) && + floatsEqual(start[1], prev[1])) { + /* closed subpath, join first and last point */ +#if STROKE_DEBUG + debug_printf("\n stroker: closed subpath\n"); +#endif + create_joins(stroker, prev[0], prev[1], start_tangent, + stroker_join_mode(stroker)); + return VG_TRUE; + } else { +#if STROKE_DEBUG + debug_printf("\n stroker: open subpath\n"); +#endif + return VG_FALSE; + } +} + +static void stroker_process_subpath(struct stroker *stroker) +{ + VGboolean fwclosed, bwclosed; + VGfloat fw_start_tangent[4], bw_start_tangent[4]; + struct stroke_iterator fwit; + struct stroke_iterator bwit; + debug_assert(stroker->segments->num_elements > 0); + + memset(fw_start_tangent, 0, + sizeof(VGfloat)*4); + memset(bw_start_tangent, 0, + sizeof(VGfloat)*4); + + stroke_forward_iterator(&fwit, stroker->segments, + stroker->control_points); + stroke_backward_iterator(&bwit, stroker->segments, + stroker->control_points); + + debug_assert(fwit.cmds[0] == VG_MOVE_TO_ABS); + + fwclosed = vg_stroke_outline(&fwit, stroker, VG_FALSE, fw_start_tangent); + bwclosed = vg_stroke_outline(&bwit, stroker, !fwclosed, bw_start_tangent); + + if (!bwclosed) + create_joins(stroker, + fwit.coords[0], fwit.coords[1], fw_start_tangent, + stroker_cap_mode(stroker)); + else { + /* hack to handle the requirement of the VG spec that says that strokes + * of len==0 that have butt cap or round cap still need + * to be rendered. (8.7.4 Stroke Generation) */ + if (stroker->segments->num_elements <= 3) { + VGPathCommand cmd; + VGfloat data[8], coords[8]; + struct stroke_iterator *it = &fwit; + + stroke_forward_iterator(it, stroker->segments, + stroker->control_points); + cmd = stroke_itr_command(it); + stroke_itr_coords(it, coords); + if (cmd != VG_MOVE_TO_ABS) { + memset(data, 0, sizeof(VGfloat) * 8); + if (!is_segment_null(cmd, coords, data)) + return; + } else { + data[0] = coords[0]; + data[1] = coords[1]; + } + while (it->has_next(it)) { + it->next(it); + cmd = stroke_itr_command(it); + stroke_itr_coords(it, coords); + if (!is_segment_null(cmd, coords, data)) + return; + } + /* generate the square/round cap */ + if (stroker->cap_style == VG_CAP_SQUARE) { + VGfloat offset = stroker->stroke_width / 2; + stroker_emit_move_to(stroker, data[0] - offset, + data[1] - offset); + stroker_emit_line_to(stroker, data[0] + offset, + data[1] - offset); + stroker_emit_line_to(stroker, data[0] + offset, + data[1] + offset); + stroker_emit_line_to(stroker, data[0] - offset, + data[1] + offset); + stroker_emit_line_to(stroker, data[0] - offset, + data[1] - offset); + } else if (stroker->cap_style == VG_CAP_ROUND) { + VGfloat offset = stroker->stroke_width / 2; + VGfloat cx = data[0], cy = data[1]; + { /*circle */ + struct arc arc; + struct matrix matrix; + matrix_load_identity(&matrix); + + stroker_emit_move_to(stroker, cx + offset, cy); + arc_init(&arc, VG_SCCWARC_TO_ABS, + cx + offset, cy, + cx - offset, cy, + offset, offset, 0); + arc_stroker_emit(&arc, stroker, &matrix); + arc_init(&arc, VG_SCCWARC_TO_ABS, + cx - offset, cy, + cx + offset, cy, + offset, offset, 0); + arc_stroker_emit(&arc, stroker, &matrix); + } + } + } + } +} + +static INLINE VGfloat dash_pattern(struct dash_stroker *stroker, + VGint idx) +{ + if (stroker->dash_pattern[idx] < 0) + return 0.f; + return stroker->dash_pattern[idx]; +} + +static void dash_stroker_process_subpath(struct stroker *str) +{ + struct dash_stroker *stroker = (struct dash_stroker *)str; + VGfloat sum_length = 0; + VGint i; + VGint idash = 0; + VGfloat pos = 0; + VGfloat elen = 0; + VGfloat doffset = stroker->dash_phase; + VGfloat estart = 0; + VGfloat estop = 0; + VGfloat cline[4]; + struct stroke_iterator it; + VGfloat prev[2]; + VGfloat move_to_pos[2]; + VGfloat line_to_pos[2]; + + VGboolean has_move_to = VG_FALSE; + + stroke_flat_iterator(&it, stroker->base.segments, + stroker->base.control_points); + + stroke_itr_coords(&it, prev); + move_to_pos[0] = prev[0]; + move_to_pos[1] = prev[1]; + + debug_assert(stroker->dash_pattern_num > 0); + + for (i = 0; i < stroker->dash_pattern_num; ++i) { + sum_length += dash_pattern(stroker, i); + } + + if (floatIsZero(sum_length)) { + return; + } + + doffset -= floorf(doffset / sum_length) * sum_length; + + while (!floatIsZero(doffset) && doffset >= dash_pattern(stroker, idash)) { + doffset -= dash_pattern(stroker, idash); + idash = (idash + 1) % stroker->dash_pattern_num; + } + + while (it.has_next(&it)) { + VGPathCommand cmd; + VGfloat coords[8]; + VGboolean done; + + it.next(&it); + cmd = stroke_itr_command(&it); + stroke_itr_coords(&it, coords); + + debug_assert(cmd == VG_LINE_TO_ABS); + cline[0] = prev[0]; + cline[1] = prev[1]; + cline[2] = coords[0]; + cline[3] = coords[1]; + + elen = line_lengthv(cline); + + estop = estart + elen; + + done = pos >= estop; + while (!done) { + VGfloat p2[2]; + + VGint idash_incr = 0; + VGboolean has_offset = doffset > 0; + VGfloat dpos = pos + dash_pattern(stroker, idash) - doffset - estart; + + debug_assert(dpos >= 0); + + if (dpos > elen) { /* dash extends this line */ + doffset = dash_pattern(stroker, idash) - (dpos - elen); + pos = estop; + done = VG_TRUE; + p2[0] = cline[2]; + p2[1] = cline[3]; + } else { /* Dash is on this line */ + line_point_at(cline, dpos/elen, p2); + pos = dpos + estart; + done = pos >= estop; + idash_incr = 1; + doffset = 0; + } + + if (idash % 2 == 0) { + line_to_pos[0] = p2[0]; + line_to_pos[1] = p2[1]; + + if (!has_offset || !has_move_to) { + stroker_move_to(&stroker->stroker, move_to_pos[0], move_to_pos[1]); + has_move_to = VG_TRUE; + } + stroker_line_to(&stroker->stroker, line_to_pos[0], line_to_pos[1]); + } else { + move_to_pos[0] = p2[0]; + move_to_pos[1] = p2[1]; + } + + idash = (idash + idash_incr) % stroker->dash_pattern_num; + } + + estart = estop; + prev[0] = coords[0]; + prev[1] = coords[1]; + } + + if (it.curve_poly) { + polygon_destroy(it.curve_poly); + it.curve_poly = 0; + } + + stroker->base.path = stroker->stroker.path; +} + +static void default_begin(struct stroker *stroker) +{ + array_reset(stroker->segments); + array_reset(stroker->control_points); +} + +static void default_end(struct stroker *stroker) +{ + if (stroker->segments->num_elements > 0) + stroker->process_subpath(stroker); +} + + +static void dash_stroker_begin(struct stroker *stroker) +{ + struct dash_stroker *dasher = + (struct dash_stroker *)stroker; + + default_begin(&dasher->stroker); + default_begin(stroker); +} + +static void dash_stroker_end(struct stroker *stroker) +{ + struct dash_stroker *dasher = + (struct dash_stroker *)stroker; + + default_end(stroker); + default_end(&dasher->stroker); +} + +void stroker_init(struct stroker *stroker, + struct vg_state *state) +{ + stroker->stroke_width = state->stroke.line_width.f; + stroker->miter_limit = state->stroke.miter_limit.f; + stroker->cap_style = state->stroke.cap_style; + stroker->join_style = state->stroke.join_style; + + stroker->begin = default_begin; + stroker->process_subpath = stroker_process_subpath; + stroker->end = default_end; + + stroker->segments = array_create(sizeof(VGubyte)); + stroker->control_points = array_create(sizeof(VGfloat)); + + stroker->back1_x = 0; + stroker->back1_y = 0; + stroker->back2_x = 0; + stroker->back2_y = 0; + + stroker->path = path_create(VG_PATH_DATATYPE_F, 1.0f, 0.0f, + 0, 0, VG_PATH_CAPABILITY_ALL); + + stroker->last_cmd = VG_CLOSE_PATH; +} + +void dash_stroker_init(struct stroker *str, + struct vg_state *state) +{ + struct dash_stroker *stroker = (struct dash_stroker *)str; + int i; + + stroker_init(str, state); + stroker_init(&stroker->stroker, state); + + { + int real_num = state->stroke.dash_pattern_num; + if (real_num % 2)/* if odd, ignore the last one */ + --real_num; + for (i = 0; i < real_num; ++i) + stroker->dash_pattern[i] = state->stroke.dash_pattern[i].f; + stroker->dash_pattern_num = real_num; + } + + stroker->dash_phase = state->stroke.dash_phase.f; + stroker->dash_phase_reset = state->stroke.dash_phase_reset; + + stroker->base.begin = dash_stroker_begin; + stroker->base.process_subpath = dash_stroker_process_subpath; + stroker->base.end = dash_stroker_end; + path_destroy(stroker->base.path); + stroker->base.path = NULL; +} + +void stroker_begin(struct stroker *stroker) +{ + stroker->begin(stroker); +} + +void stroker_end(struct stroker *stroker) +{ + stroker->end(stroker); +} + +void stroker_cleanup(struct stroker *stroker) +{ + array_destroy(stroker->segments); + array_destroy(stroker->control_points); +} + +void dash_stroker_cleanup(struct dash_stroker *stroker) +{ + /* if stroker->base.path is null means we never + * processed a valid path so delete the temp one + * we already created */ + if (!stroker->base.path) + path_destroy(stroker->stroker.path); + stroker_cleanup(&stroker->stroker); + stroker_cleanup((struct stroker*)stroker); +} diff --git a/src/gallium/state_trackers/vega/stroker.h b/src/gallium/state_trackers/vega/stroker.h new file mode 100644 index 00000000000..36543cd9237 --- /dev/null +++ b/src/gallium/state_trackers/vega/stroker.h @@ -0,0 +1,89 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STROKER_H +#define STROKER_H + +#include "VG/openvg.h" +#include "api_consts.h" + +struct path; +struct vg_state; +struct array; + +struct stroker { + void (*begin)(struct stroker *stroker); + void (*process_subpath)(struct stroker *stroker); + void (*end)(struct stroker *stroker); + + struct array *segments; + struct array *control_points; + struct path *path; + + VGfloat back1_x, back1_y; + VGfloat back2_x, back2_y; + + VGfloat stroke_width; + VGfloat miter_limit; + VGCapStyle cap_style; + VGJoinStyle join_style; + + VGPathCommand last_cmd; +}; + +struct dash_stroker { + struct stroker base; + + struct stroker stroker; + + VGfloat dash_pattern[VEGA_MAX_DASH_COUNT]; + VGint dash_pattern_num; + VGfloat dash_phase; + VGboolean dash_phase_reset; +}; + +void stroker_init(struct stroker *stroker, + struct vg_state *state); +void dash_stroker_init(struct stroker *stroker, + struct vg_state *state); +void dash_stroker_cleanup(struct dash_stroker *stroker); +void stroker_cleanup(struct stroker *stroker); + +void stroker_begin(struct stroker *stroker); +void stroker_move_to(struct stroker *stroker, VGfloat x, VGfloat y); +void stroker_line_to(struct stroker *stroker, VGfloat x, VGfloat y); +void stroker_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1, + VGfloat px2, VGfloat py2, + VGfloat x, VGfloat y); +void stroker_end(struct stroker *stroker); + +void stroker_emit_move_to(struct stroker *stroker, VGfloat x, VGfloat y); +void stroker_emit_line_to(struct stroker *stroker, VGfloat x, VGfloat y); +void stroker_emit_curve_to(struct stroker *stroker, VGfloat px1, VGfloat py1, + VGfloat px2, VGfloat py2, + VGfloat x, VGfloat y); + +#endif diff --git a/src/gallium/state_trackers/vega/util_array.h b/src/gallium/state_trackers/vega/util_array.h new file mode 100644 index 00000000000..4343dfe36e2 --- /dev/null +++ b/src/gallium/state_trackers/vega/util_array.h @@ -0,0 +1,122 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef UTIL_ARRAY_H +#define UTIL_ARRAY_H + +#include "util/u_memory.h" + +#define DEFAULT_ARRAY_SIZE 256 + +struct array { + VGint datatype_size; + void *data; + VGint size; + VGint num_elements; +}; + +static INLINE struct array *array_create(VGint datatype_size) +{ + struct array *array = CALLOC_STRUCT(array); + array->datatype_size = datatype_size; + + array->size = DEFAULT_ARRAY_SIZE; + array->data = malloc(array->size * array->datatype_size); + + return array; +} + + +static INLINE struct array *array_create_size(VGint datatype_size, VGint size) +{ + struct array *array = CALLOC_STRUCT(array); + array->datatype_size = datatype_size; + + array->size = size; + array->data = malloc(array->size * array->datatype_size); + + return array; +} + +static INLINE void array_destroy(struct array *array) +{ + if (array) + free(array->data); + free(array); +} + +static INLINE void array_resize(struct array *array, int num) +{ + VGint size = array->datatype_size * num; + void *data = malloc(size); + memcpy(data, array->data, array->size * array->datatype_size); + free(array->data); + array->data = data; + array->size = num; + array->num_elements = (array->num_elements > num) ? num : + array->num_elements; +} + +static INLINE void array_append_data(struct array *array, + const void *data, int num_elements) +{ + VGbyte *adata; + + while (array->num_elements + num_elements > array->size) { + array_resize(array, (array->num_elements + num_elements) * 1.5); + } + adata = (VGbyte *)array->data; + memcpy(adata + (array->num_elements * array->datatype_size), data, + num_elements * array->datatype_size); + array->num_elements += num_elements; +} + +static INLINE void array_change_data(struct array *array, + const void *data, + int start_idx, + int num_elements) +{ + VGbyte *adata = (VGbyte *)array->data; + memcpy(adata + (start_idx * array->datatype_size), data, + num_elements * array->datatype_size); +} + +static INLINE void array_remove_element(struct array *array, + int idx) +{ + VGbyte *adata = (VGbyte *)array->data; + memmove(adata + (idx * array->datatype_size), + adata + ((idx + 1) * array->datatype_size), + (array->num_elements - idx - 1) * array->datatype_size); + --array->num_elements; +} + +static INLINE void array_reset(struct array *array) +{ + array->num_elements = 0; +} + +#endif diff --git a/src/gallium/state_trackers/vega/vg_context.c b/src/gallium/state_trackers/vega/vg_context.c new file mode 100644 index 00000000000..e0ff02f3a99 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_context.c @@ -0,0 +1,543 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vg_context.h" + +#include "paint.h" +#include "renderer.h" +#include "shaders_cache.h" +#include "shader.h" +#include "asm_util.h" +#include "st_inlines.h" + +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" + +#include "cso_cache/cso_context.h" + +#include "util/u_simple_shaders.h" +#include "util/u_memory.h" +#include "util/u_blit.h" + +struct vg_context *_vg_context = 0; + +struct vg_context * vg_current_context(void) +{ + return _vg_context; +} + +static void init_clear(struct vg_context *st) +{ + struct pipe_context *pipe = st->pipe; + + /* rasterizer state: bypass clipping */ + memset(&st->clear.raster, 0, sizeof(st->clear.raster)); + st->clear.raster.gl_rasterization_rules = 1; + + /* fragment shader state: color pass-through program */ + st->clear.fs = + util_make_fragment_passthrough_shader(pipe); +} +void vg_set_current_context(struct vg_context *ctx) +{ + _vg_context = ctx; +} + +struct vg_context * vg_create_context(struct pipe_context *pipe, + const void *visual, + struct vg_context *share) +{ + struct vg_context *ctx; + + ctx = CALLOC_STRUCT(vg_context); + + ctx->pipe = pipe; + + vg_init_state(&ctx->state.vg); + ctx->state.dirty = ALL_DIRTY; + + ctx->cso_context = cso_create_context(pipe); + + init_clear(ctx); + + ctx->default_paint = paint_create(ctx); + ctx->state.vg.stroke_paint = ctx->default_paint; + ctx->state.vg.fill_paint = ctx->default_paint; + + + ctx->mask.sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->mask.sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->mask.sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + ctx->mask.sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + ctx->mask.sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + ctx->mask.sampler.normalized_coords = 0; + + ctx->blend_sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->blend_sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->blend_sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + ctx->blend_sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + ctx->blend_sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + ctx->blend_sampler.normalized_coords = 0; + + vg_set_error(ctx, VG_NO_ERROR); + + ctx->owned_objects[VG_OBJECT_PAINT] = cso_hash_create(); + ctx->owned_objects[VG_OBJECT_IMAGE] = cso_hash_create(); + ctx->owned_objects[VG_OBJECT_MASK] = cso_hash_create(); + ctx->owned_objects[VG_OBJECT_FONT] = cso_hash_create(); + ctx->owned_objects[VG_OBJECT_PATH] = cso_hash_create(); + + ctx->renderer = renderer_create(ctx); + ctx->sc = shaders_cache_create(ctx); + ctx->shader = shader_create(ctx); + + ctx->blit = util_create_blit(ctx->pipe, ctx->cso_context); + + return ctx; +} + +void vg_destroy_context(struct vg_context *ctx) +{ + struct pipe_constant_buffer *cbuf = &ctx->mask.cbuf; + struct pipe_constant_buffer *vsbuf = &ctx->vs_const_buffer; + + util_destroy_blit(ctx->blit); + renderer_destroy(ctx->renderer); + shaders_cache_destroy(ctx->sc); + shader_destroy(ctx->shader); + paint_destroy(ctx->default_paint); + + if (cbuf && cbuf->buffer) + pipe_buffer_reference(&cbuf->buffer, NULL); + + if (vsbuf && vsbuf->buffer) + pipe_buffer_reference(&vsbuf->buffer, NULL); + + if (ctx->clear.fs) { + cso_delete_fragment_shader(ctx->cso_context, ctx->clear.fs); + ctx->clear.fs = NULL; + } + + if (ctx->plain_vs) { + vg_shader_destroy(ctx, ctx->plain_vs); + ctx->plain_vs = NULL; + } + if (ctx->clear_vs) { + vg_shader_destroy(ctx, ctx->clear_vs); + ctx->clear_vs = NULL; + } + if (ctx->texture_vs) { + vg_shader_destroy(ctx, ctx->texture_vs); + ctx->texture_vs = NULL; + } + + if (ctx->pass_through_depth_fs) + vg_shader_destroy(ctx, ctx->pass_through_depth_fs); + if (ctx->mask.union_fs) + vg_shader_destroy(ctx, ctx->mask.union_fs); + if (ctx->mask.intersect_fs) + vg_shader_destroy(ctx, ctx->mask.intersect_fs); + if (ctx->mask.subtract_fs) + vg_shader_destroy(ctx, ctx->mask.subtract_fs); + if (ctx->mask.set_fs) + vg_shader_destroy(ctx, ctx->mask.set_fs); + + cso_release_all(ctx->cso_context); + cso_destroy_context(ctx->cso_context); + + cso_hash_delete(ctx->owned_objects[VG_OBJECT_PAINT]); + cso_hash_delete(ctx->owned_objects[VG_OBJECT_IMAGE]); + cso_hash_delete(ctx->owned_objects[VG_OBJECT_MASK]); + cso_hash_delete(ctx->owned_objects[VG_OBJECT_FONT]); + cso_hash_delete(ctx->owned_objects[VG_OBJECT_PATH]); + + free(ctx); +} + +void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type) +{ + obj->type = type; + obj->ctx = ctx; +} + +VGboolean vg_context_is_object_valid(struct vg_context *ctx, + enum vg_object_type type, + void *ptr) +{ + if (ctx) { + struct cso_hash *hash = ctx->owned_objects[type]; + if (!hash) + return VG_FALSE; + return cso_hash_contains(hash, (unsigned)(long)ptr); + } + return VG_FALSE; +} + +void vg_context_add_object(struct vg_context *ctx, + enum vg_object_type type, + void *ptr) +{ + if (ctx) { + struct cso_hash *hash = ctx->owned_objects[type]; + if (!hash) + return; + cso_hash_insert(hash, (unsigned)(long)ptr, ptr); + } +} + +void vg_context_remove_object(struct vg_context *ctx, + enum vg_object_type type, + void *ptr) +{ + if (ctx) { + struct cso_hash *hash = ctx->owned_objects[type]; + if (!hash) + return; + cso_hash_take(hash, (unsigned)(long)ptr); + } +} + +static void update_clip_state(struct vg_context *ctx) +{ + struct pipe_depth_stencil_alpha_state *dsa = &ctx->state.g3d.dsa; + struct vg_state *state = &ctx->state.vg; + + memset(dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state)); + + if (state->scissoring) { + struct pipe_blend_state *blend = &ctx->state.g3d.blend; + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + dsa->depth.writemask = 1;/*glDepthMask(TRUE);*/ + dsa->depth.func = PIPE_FUNC_ALWAYS; + dsa->depth.enabled = 1; + + cso_save_blend(ctx->cso_context); + cso_save_fragment_shader(ctx->cso_context); + /* set a passthrough shader */ + if (!ctx->pass_through_depth_fs) + ctx->pass_through_depth_fs = shader_create_from_text(ctx->pipe, + pass_through_depth_asm, + 40, + PIPE_SHADER_FRAGMENT); + cso_set_fragment_shader_handle(ctx->cso_context, + ctx->pass_through_depth_fs->driver); + cso_set_depth_stencil_alpha(ctx->cso_context, dsa); + + ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL, NULL, 1.0, 0); + + /* disable color writes */ + blend->colormask = 0; /*disable colorwrites*/ + cso_set_blend(ctx->cso_context, blend); + + /* enable scissoring */ + for (int i = 0; i < state->scissor_rects_num; ++i) { + const float x = state->scissor_rects[i * 4 + 0].f; + const float y = state->scissor_rects[i * 4 + 1].f; + const float width = state->scissor_rects[i * 4 + 2].f; + const float height = state->scissor_rects[i * 4 + 3].f; + VGfloat minx, miny, maxx, maxy; + + minx = 0; + miny = 0; + maxx = fb->width; + maxy = fb->height; + + if (x > minx) + minx = x; + if (y > miny) + miny = y; + + if (x + width < maxx) + maxx = x + width; + if (y + height < maxy) + maxy = y + height; + + /* check for null space */ + if (minx >= maxx || miny >= maxy) + minx = miny = maxx = maxy = 0; + + /*glClear(GL_DEPTH_BUFFER_BIT);*/ + renderer_draw_quad(ctx->renderer, minx, miny, maxx, maxy, 0.0f); + } + + blend->colormask = 1; /*enable colorwrites*/ + cso_restore_blend(ctx->cso_context); + cso_restore_fragment_shader(ctx->cso_context); + + dsa->depth.enabled = 1; /* glEnable(GL_DEPTH_TEST); */ + dsa->depth.writemask = 0;/*glDepthMask(FALSE);*/ + dsa->depth.func = PIPE_FUNC_GEQUAL; + } +} + +void vg_validate_state(struct vg_context *ctx) +{ + if ((ctx->state.dirty & BLEND_DIRTY)) { + struct pipe_blend_state *blend = &ctx->state.g3d.blend; + memset(blend, 0, sizeof(struct pipe_blend_state)); + blend->blend_enable = 1; + blend->colormask |= PIPE_MASK_R; + blend->colormask |= PIPE_MASK_G; + blend->colormask |= PIPE_MASK_B; + blend->colormask |= PIPE_MASK_A; + + switch (ctx->state.vg.blend_mode) { + case VG_BLEND_SRC: + blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + break; + case VG_BLEND_SRC_OVER: + blend->rgb_src_factor = PIPE_BLENDFACTOR_SRC_ALPHA; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_INV_SRC_ALPHA; + break; + case VG_BLEND_DST_OVER: + blend->rgb_src_factor = PIPE_BLENDFACTOR_INV_DST_ALPHA; + blend->alpha_src_factor = PIPE_BLENDFACTOR_INV_DST_ALPHA; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_DST_ALPHA; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_DST_ALPHA; + break; + case VG_BLEND_SRC_IN: + blend->rgb_src_factor = PIPE_BLENDFACTOR_DST_ALPHA; + blend->alpha_src_factor = PIPE_BLENDFACTOR_DST_ALPHA; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + break; + case VG_BLEND_DST_IN: + blend->rgb_src_factor = PIPE_BLENDFACTOR_ZERO; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ZERO; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_SRC_ALPHA; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_SRC_ALPHA; + break; + case VG_BLEND_MULTIPLY: + case VG_BLEND_SCREEN: + case VG_BLEND_DARKEN: + case VG_BLEND_LIGHTEN: + blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + break; + case VG_BLEND_ADDITIVE: + blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE; + break; + default: + assert(!"not implemented blend mode"); + } + cso_set_blend(ctx->cso_context, &ctx->state.g3d.blend); + } + if ((ctx->state.dirty & RASTERIZER_DIRTY)) { + struct pipe_rasterizer_state *raster = &ctx->state.g3d.rasterizer; + memset(raster, 0, sizeof(struct pipe_rasterizer_state)); + raster->gl_rasterization_rules = 1; + cso_set_rasterizer(ctx->cso_context, &ctx->state.g3d.rasterizer); + } + if ((ctx->state.dirty & VIEWPORT_DIRTY)) { + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + const VGint param_bytes = 8 * sizeof(VGfloat); + VGfloat vs_consts[8] = { + 2.f/fb->width, 2.f/fb->height, 1, 1, + -1, -1, 0, 0 + }; + struct pipe_constant_buffer *cbuf = &ctx->vs_const_buffer; + + vg_set_viewport(ctx, VEGA_Y0_BOTTOM); + + pipe_buffer_reference(&cbuf->buffer, NULL); + cbuf->buffer = pipe_buffer_create(ctx->pipe->screen, 16, + PIPE_BUFFER_USAGE_CONSTANT, + param_bytes); + + if (cbuf->buffer) { + st_no_flush_pipe_buffer_write(ctx, cbuf->buffer, + 0, param_bytes, vs_consts); + } + ctx->pipe->set_constant_buffer(ctx->pipe, PIPE_SHADER_VERTEX, 0, cbuf); + } + if ((ctx->state.dirty & VS_DIRTY)) { + cso_set_vertex_shader_handle(ctx->cso_context, + vg_plain_vs(ctx)); + } + + /* must be last because it renders to the depth buffer*/ + if ((ctx->state.dirty & DEPTH_STENCIL_DIRTY)) { + update_clip_state(ctx); + cso_set_depth_stencil_alpha(ctx->cso_context, &ctx->state.g3d.dsa); + } + + shader_set_masking(ctx->shader, ctx->state.vg.masking); + shader_set_image_mode(ctx->shader, ctx->state.vg.image_mode); + + ctx->state.dirty = NONE_DIRTY; +} + +VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type) +{ + struct vg_object *obj = ptr; + if (ptr && is_aligned(obj) && obj->type == type) + return VG_TRUE; + else + return VG_FALSE; +} + +void vg_set_error(struct vg_context *ctx, + VGErrorCode code) +{ + /*vgGetError returns the oldest error code provided by + * an API call on the current context since the previous + * call to vgGetError on that context (or since the creation + of the context).*/ + if (ctx->_error == VG_NO_ERROR) + ctx->_error = code; +} + +void vg_prepare_blend_surface(struct vg_context *ctx) +{ + struct pipe_surface *dest_surface = NULL; + struct pipe_context *pipe = ctx->pipe; + struct st_framebuffer *stfb = ctx->draw_buffer; + struct st_renderbuffer *strb = stfb->strb; + + /* first finish all pending rendering */ + vgFinish(); + + dest_surface = pipe->screen->get_tex_surface(pipe->screen, + stfb->blend_texture, + 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + /* flip it, because we want to use it as a sampler */ + util_blit_pixels_tex(ctx->blit, + strb->texture, + 0, strb->height, + strb->width, 0, + dest_surface, + 0, 0, + strb->width, strb->height, + 0.0, PIPE_TEX_MIPFILTER_NEAREST); + + if (dest_surface) + pipe_surface_reference(&dest_surface, NULL); + + /* make sure it's complete */ + vgFinish(); +} + + +void vg_prepare_blend_surface_from_mask(struct vg_context *ctx) +{ + struct pipe_surface *dest_surface = NULL; + struct pipe_context *pipe = ctx->pipe; + struct st_framebuffer *stfb = ctx->draw_buffer; + struct st_renderbuffer *strb = stfb->strb; + + vg_validate_state(ctx); + + /* first finish all pending rendering */ + vgFinish(); + + dest_surface = pipe->screen->get_tex_surface(pipe->screen, + stfb->blend_texture, + 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + /* flip it, because we want to use it as a sampler */ + util_blit_pixels_tex(ctx->blit, + stfb->alpha_mask, + 0, strb->height, + strb->width, 0, + dest_surface, + 0, 0, + strb->width, strb->height, + 0.0, PIPE_TEX_MIPFILTER_NEAREST); + + /* make sure it's complete */ + vgFinish(); + + if (dest_surface) + pipe_surface_reference(&dest_surface, NULL); +} + +void * vg_plain_vs(struct vg_context *ctx) +{ + if (!ctx->plain_vs) { + ctx->plain_vs = shader_create_from_text(ctx->pipe, + vs_plain_asm, + 200, + PIPE_SHADER_VERTEX); + } + + return ctx->plain_vs->driver; +} + + +void * vg_clear_vs(struct vg_context *ctx) +{ + if (!ctx->clear_vs) { + ctx->clear_vs = shader_create_from_text(ctx->pipe, + vs_clear_asm, + 200, + PIPE_SHADER_VERTEX); + } + + return ctx->clear_vs->driver; +} + +void * vg_texture_vs(struct vg_context *ctx) +{ + if (!ctx->texture_vs) { + ctx->texture_vs = shader_create_from_text(ctx->pipe, + vs_texture_asm, + 200, + PIPE_SHADER_VERTEX); + } + + return ctx->texture_vs->driver; +} + +void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation) +{ + struct pipe_viewport_state viewport; + struct pipe_framebuffer_state *fb = &ctx->state.g3d.fb; + VGfloat y_scale = (orientation == VEGA_Y0_BOTTOM) ? -2.f : 2.f; + + viewport.scale[0] = fb->width / 2.f; + viewport.scale[1] = fb->height / y_scale; + viewport.scale[2] = 1.0; + viewport.scale[3] = 1.0; + viewport.translate[0] = fb->width / 2.f; + viewport.translate[1] = fb->height / 2.f; + viewport.translate[2] = 0.0; + viewport.translate[3] = 0.0; + + cso_set_viewport(ctx->cso_context, &viewport); +} diff --git a/src/gallium/state_trackers/vega/vg_context.h b/src/gallium/state_trackers/vega/vg_context.h new file mode 100644 index 00000000000..ccc8889c8c5 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_context.h @@ -0,0 +1,292 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef VG_CONTEXT_H +#define VG_CONTEXT_H + +#include "vg_state.h" + +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "util/u_pointer.h" +#include "util/u_math.h" + +#include "cso_cache/cso_hash.h" +#include "cso_cache/cso_context.h" + +struct renderer; +struct shaders_cache; +struct shader; +struct vg_shader; + +struct st_renderbuffer { + enum pipe_format format; + struct pipe_surface *surface; + struct pipe_texture *texture; + VGint width, height; +}; + +struct st_framebuffer { + VGint init_width, init_height; + struct st_renderbuffer *strb; + struct st_renderbuffer *dsrb; + + struct pipe_texture *alpha_mask; + + struct pipe_texture *blend_texture; + + void *privateData; +}; + +enum vg_object_type { + VG_OBJECT_UNKNOWN = 0, + VG_OBJECT_PAINT, + VG_OBJECT_IMAGE, + VG_OBJECT_MASK, + VG_OBJECT_FONT, + VG_OBJECT_PATH, + + VG_OBJECT_LAST +}; +enum dirty_state { + NONE_DIRTY = 0<<0, + BLEND_DIRTY = 1<<1, + RASTERIZER_DIRTY = 1<<2, + VIEWPORT_DIRTY = 1<<3, + VS_DIRTY = 1<<4, + DEPTH_STENCIL_DIRTY = 1<<5, + ALL_DIRTY = BLEND_DIRTY | RASTERIZER_DIRTY | + VIEWPORT_DIRTY | VS_DIRTY | DEPTH_STENCIL_DIRTY +}; + +struct vg_context +{ + struct pipe_context *pipe; + + struct { + struct vg_state vg; + struct { + struct pipe_blend_state blend; + struct pipe_rasterizer_state rasterizer; + struct pipe_shader_state vs_state; + struct pipe_depth_stencil_alpha_state dsa; + struct pipe_framebuffer_state fb; + } g3d; + VGbitfield dirty; + } state; + + VGErrorCode _error; + + struct st_framebuffer *draw_buffer; + + struct cso_hash *owned_objects[VG_OBJECT_LAST]; + + struct { + struct pipe_shader_state vert_shader; + struct pipe_shader_state frag_shader; + struct pipe_rasterizer_state raster; + void *fs; + float vertices[4][2][4]; /**< vertex pos + color */ + } clear; + + struct { + struct pipe_constant_buffer cbuf; + struct pipe_sampler_state sampler; + + struct vg_shader *union_fs; + struct vg_shader *intersect_fs; + struct vg_shader *subtract_fs; + struct vg_shader *set_fs; + } mask; + + struct vg_shader *pass_through_depth_fs; + + struct cso_context *cso_context; + + struct pipe_buffer *stencil_quad; + VGfloat stencil_vertices[4][2][4]; + + struct renderer *renderer; + struct shaders_cache *sc; + struct shader *shader; + + struct pipe_sampler_state blend_sampler; + struct { + struct pipe_constant_buffer buffer; + void *color_matrix_fs; + } filter; + struct vg_paint *default_paint; + + struct blit_state *blit; + + struct vg_shader *plain_vs; + struct vg_shader *clear_vs; + struct vg_shader *texture_vs; + struct pipe_constant_buffer vs_const_buffer; +}; + +struct vg_object { + enum vg_object_type type; + struct vg_context *ctx; +}; +void vg_init_object(struct vg_object *obj, struct vg_context *ctx, enum vg_object_type type); +VGboolean vg_object_is_valid(void *ptr, enum vg_object_type type); + +struct vg_context *vg_create_context(struct pipe_context *pipe, + const void *visual, + struct vg_context *share); +void vg_destroy_context(struct vg_context *ctx); +struct vg_context *vg_current_context(void); +void vg_set_current_context(struct vg_context *ctx); + +VGboolean vg_context_is_object_valid(struct vg_context *ctx, + enum vg_object_type type, + void *ptr); +void vg_context_add_object(struct vg_context *ctx, + enum vg_object_type type, + void *ptr); +void vg_context_remove_object(struct vg_context *ctx, + enum vg_object_type type, + void *ptr); + +void vg_validate_state(struct vg_context *ctx); + +void vg_set_error(struct vg_context *ctx, + VGErrorCode code); + +void vg_prepare_blend_surface(struct vg_context *ctx); +void vg_prepare_blend_surface_from_mask(struct vg_context *ctx); + + +static INLINE VGboolean is_aligned_to(const void *ptr, VGbyte alignment) +{ + void *aligned = align_pointer(ptr, alignment); + return (ptr == aligned) ? VG_TRUE : VG_FALSE; +} + +static INLINE VGboolean is_aligned(const void *ptr) +{ + return is_aligned_to(ptr, 4); +} + +static INLINE void vg_shift_rectx(VGfloat coords[4], + const VGfloat *bounds, + const VGfloat shift) +{ + coords[0] += shift; + coords[2] -= shift; + if (bounds) { + coords[2] = MIN2(coords[2], bounds[2]); + /* bound x/y + width/height */ + if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) { + coords[2] = (bounds[0] + bounds[2]) - coords[0]; + } + } +} + +static INLINE void vg_shift_recty(VGfloat coords[4], + const VGfloat *bounds, + const VGfloat shift) +{ + coords[1] += shift; + coords[3] -= shift; + if (bounds) { + coords[3] = MIN2(coords[3], bounds[3]); + if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) { + coords[3] = (bounds[1] + bounds[3]) - coords[1]; + } + } +} + +static INLINE void vg_bound_rect(VGfloat coords[4], + const VGfloat bounds[4], + VGfloat shift[4]) +{ + /* if outside the bounds */ + if (coords[0] > (bounds[0] + bounds[2]) || + coords[1] > (bounds[1] + bounds[3]) || + (coords[0] + coords[2]) < bounds[0] || + (coords[1] + coords[3]) < bounds[1]) { + coords[0] = 0.f; + coords[1] = 0.f; + coords[2] = 0.f; + coords[3] = 0.f; + shift[0] = 0.f; + shift[1] = 0.f; + return; + } + + /* bound x */ + if (coords[0] < bounds[0]) { + shift[0] = bounds[0] - coords[0]; + coords[2] -= shift[0]; + coords[0] = bounds[0]; + } else + shift[0] = 0.f; + + /* bound y */ + if (coords[1] < bounds[1]) { + shift[1] = bounds[1] - coords[1]; + coords[3] -= shift[1]; + coords[1] = bounds[1]; + } else + shift[1] = 0.f; + + shift[2] = bounds[2] - coords[2]; + shift[3] = bounds[3] - coords[3]; + /* bound width/height */ + coords[2] = MIN2(coords[2], bounds[2]); + coords[3] = MIN2(coords[3], bounds[3]); + + /* bound x/y + width/height */ + if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) { + coords[2] = (bounds[0] + bounds[2]) - coords[0]; + } + if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) { + coords[3] = (bounds[1] + bounds[3]) - coords[1]; + } + + /* if outside the bounds */ + if ((coords[0] + coords[2]) < bounds[0] || + (coords[1] + coords[3]) < bounds[1]) { + coords[0] = 0.f; + coords[1] = 0.f; + coords[2] = 0.f; + coords[3] = 0.f; + return; + } +} + +void *vg_plain_vs(struct vg_context *ctx); +void *vg_clear_vs(struct vg_context *ctx); +void *vg_texture_vs(struct vg_context *ctx); +typedef enum { + VEGA_Y0_TOP, + VEGA_Y0_BOTTOM +} VegaOrientation; +void vg_set_viewport(struct vg_context *ctx, VegaOrientation orientation); + +#endif diff --git a/src/gallium/state_trackers/vega/vg_state.c b/src/gallium/state_trackers/vega/vg_state.c new file mode 100644 index 00000000000..6f6bfdaf7a2 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_state.c @@ -0,0 +1,124 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vg_state.h" + +#include <string.h> + +void vg_init_state(struct vg_state *state) +{ + state->matrix_mode = VG_MATRIX_PATH_USER_TO_SURFACE; + state->fill_rule = VG_EVEN_ODD; + state->image_quality = VG_IMAGE_QUALITY_FASTER; + state->rendering_quality = VG_RENDERING_QUALITY_BETTER; + state->blend_mode = VG_BLEND_SRC_OVER; + state->image_mode = VG_DRAW_IMAGE_NORMAL; + + memset(state->scissor_rects, 0, sizeof(state->scissor_rects)); + state->scissor_rects_num = 0; + + state->color_transform = VG_FALSE; + state->color_transform_values[0] = 1.0f; + state->color_transform_values[1] = 1.0f; + state->color_transform_values[2] = 1.0f; + state->color_transform_values[3] = 1.0f; + state->color_transform_values[4] = 0.0f; + state->color_transform_values[5] = 0.0f; + state->color_transform_values[6] = 0.0f; + state->color_transform_values[7] = 0.0f; + + /* Stroke parameters */ + state->stroke.line_width.f = 1.0f; + state->stroke.line_width.i = 1; + state->stroke.cap_style = VG_CAP_BUTT; + state->stroke.join_style = VG_JOIN_MITER; + state->stroke.miter_limit.f = 4.0f; + state->stroke.miter_limit.i = 4; + state->stroke.dash_pattern_num = 0; + state->stroke.dash_phase.f = 0.0f; + state->stroke.dash_phase.i = 0; + state->stroke.dash_phase_reset = VG_FALSE; + + /* Edge fill color for VG_TILE_FILL tiling mode */ + state->tile_fill_color[0] = 0.0f; + state->tile_fill_color[1] = 0.0f; + state->tile_fill_color[2] = 0.0f; + state->tile_fill_color[3] = 0.0f; + + /* Color for vgClear */ + state->clear_color[0] = 0.0f; + state->clear_color[1] = 0.0f; + state->clear_color[2] = 0.0f; + state->clear_color[3] = 0.0f; + + /* Glyph origin */ + state->glyph_origin[0].f = 0.0f; + state->glyph_origin[1].f = 0.0f; + state->glyph_origin[0].i = 0; + state->glyph_origin[1].i = 0; + + /* Enable/disable alpha masking and scissoring */ + state->masking = VG_FALSE; + state->scissoring = VG_FALSE; + + /* Pixel layout information */ + state->pixel_layout = VG_PIXEL_LAYOUT_UNKNOWN; + state->screen_layout = VG_PIXEL_LAYOUT_UNKNOWN; + + /* Source format selection for image filters */ + state->filter_format_linear = VG_FALSE; + state->filter_format_premultiplied = VG_FALSE; + + /* Destination write enable mask for image filters */ + state->filter_channel_mask = (VG_RED | VG_GREEN | VG_BLUE | VG_ALPHA); + + matrix_load_identity(&state->path_user_to_surface_matrix); + matrix_load_identity(&state->image_user_to_surface_matrix); + matrix_load_identity(&state->fill_paint_to_user_matrix); + matrix_load_identity(&state->stroke_paint_to_user_matrix); + matrix_load_identity(&state->glyph_user_to_surface_matrix); +} + +struct matrix *vg_state_matrix(struct vg_state *state) +{ + switch(state->matrix_mode) { + case VG_MATRIX_PATH_USER_TO_SURFACE: + return &state->path_user_to_surface_matrix; + case VG_MATRIX_IMAGE_USER_TO_SURFACE: + return &state->image_user_to_surface_matrix; + case VG_MATRIX_FILL_PAINT_TO_USER: + return &state->fill_paint_to_user_matrix; + case VG_MATRIX_STROKE_PAINT_TO_USER: + return &state->stroke_paint_to_user_matrix; +#ifdef OPENVG_VERSION_1_1 + case VG_MATRIX_GLYPH_USER_TO_SURFACE: + return &state->glyph_user_to_surface_matrix; +#endif + default: + break; + } + return NULL; +} diff --git a/src/gallium/state_trackers/vega/vg_state.h b/src/gallium/state_trackers/vega/vg_state.h new file mode 100644 index 00000000000..ed90689f917 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_state.h @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef VG_STATE_H +#define VG_STATE_H + +#include "VG/openvg.h" + +#include "api_consts.h" +#include "matrix.h" + +struct vg_value +{ + VGfloat f; + VGint i; +}; + +struct vg_state { + /* Mode settings */ + VGMatrixMode matrix_mode; + VGFillRule fill_rule; + VGImageQuality image_quality; + VGRenderingQuality rendering_quality; + VGBlendMode blend_mode; + VGImageMode image_mode; + + /* Scissoring rectangles */ + struct vg_value scissor_rects[32*4]; + VGint scissor_rects_num; + + /* Color Transformation */ + VGboolean color_transform; + VGfloat color_transform_values[8]; + + /* Stroke parameters */ + struct { + struct vg_value line_width; + VGCapStyle cap_style; + VGJoinStyle join_style; + struct vg_value miter_limit; + struct vg_value dash_pattern[VEGA_MAX_DASH_COUNT]; + VGint dash_pattern_num; + struct vg_value dash_phase; + VGboolean dash_phase_reset; + } stroke; + + /* Edge fill color for VG_TILE_FILL tiling mode */ + VGfloat tile_fill_color[4]; + VGint tile_fill_colori[4]; + + /* Color for vgClear */ + VGfloat clear_color[4]; + VGint clear_colori[4]; + + /* Glyph origin */ + struct vg_value glyph_origin[2]; + + /* Enable/disable alpha masking and scissoring */ + VGboolean masking; + VGboolean scissoring; + + /* Pixel layout information */ + VGPixelLayout pixel_layout; + VGPixelLayout screen_layout; + + /* Source format selection for image filters */ + VGboolean filter_format_linear; + VGboolean filter_format_premultiplied; + + /* Destination write enable mask for image filters */ + VGbitfield filter_channel_mask; + + struct matrix path_user_to_surface_matrix; + struct matrix image_user_to_surface_matrix; + struct matrix fill_paint_to_user_matrix; + struct matrix stroke_paint_to_user_matrix; + struct matrix glyph_user_to_surface_matrix; + + struct vg_paint *stroke_paint; + struct vg_paint *fill_paint; +}; + +void vg_init_state(struct vg_state *state); +struct matrix * vg_state_matrix(struct vg_state *state); + +#endif diff --git a/src/gallium/state_trackers/vega/vg_tracker.c b/src/gallium/state_trackers/vega/vg_tracker.c new file mode 100644 index 00000000000..56cc60aebe1 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_tracker.c @@ -0,0 +1,416 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vg_context.h" +#include "vg_tracker.h" +#include "mask.h" + +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "pipe/p_screen.h" +#include "util/u_memory.h" +#include "util/u_math.h" + +static struct pipe_texture * +create_texture(struct pipe_context *pipe, enum pipe_format format, + VGint width, VGint height) +{ + struct pipe_texture templ; + + memset(&templ, 0, sizeof(templ)); + + if (format != PIPE_FORMAT_NONE) { + templ.format = format; + } + else { + templ.format = PIPE_FORMAT_A8R8G8B8_UNORM; + } + + templ.target = PIPE_TEXTURE_2D; + pf_get_block(templ.format, &templ.block); + templ.width[0] = width; + templ.height[0] = height; + templ.depth[0] = 1; + templ.last_level = 0; + + if (pf_get_component_bits(format, PIPE_FORMAT_COMP_S)) { + templ.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; + } else { + templ.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_SAMPLER); + } + + return pipe->screen->texture_create(pipe->screen, &templ); +} + +/** + * Allocate a renderbuffer for a an on-screen window (not a user-created + * renderbuffer). The window system code determines the format. + */ +static struct st_renderbuffer * +st_new_renderbuffer_fb(enum pipe_format format) +{ + struct st_renderbuffer *strb; + + strb = CALLOC_STRUCT(st_renderbuffer); + if (!strb) { + /*_vega_error(NULL, VG_OUT_OF_MEMORY, "creating renderbuffer");*/ + return NULL; + } + + strb->format = format; + + return strb; +} + + +/** + * This is called to allocate the original drawing surface, and + * during window resize. + */ +static VGboolean +st_renderbuffer_alloc_storage(struct vg_context * ctx, + struct st_renderbuffer *strb, + VGuint width, VGuint height) +{ + struct pipe_context *pipe = ctx->pipe; + unsigned surface_usage; + + /* Free the old surface and texture + */ + pipe_surface_reference(&strb->surface, NULL); + pipe_texture_reference(&strb->texture, NULL); + + + /* Probably need dedicated flags for surface usage too: + */ + surface_usage = (PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + strb->texture = create_texture(pipe, strb->format, + width, height); + + if (!strb->texture) + return FALSE; + + strb->surface = pipe->screen->get_tex_surface(pipe->screen, + strb->texture, + 0, 0, 0, + surface_usage); + strb->width = width; + strb->height = height; + + assert(strb->surface->width == width); + assert(strb->surface->height == height); + + return strb->surface != NULL; +} + +struct vg_context * st_create_context(struct pipe_context *pipe, + const void *visual, + struct vg_context *share) +{ + struct vg_context *ctx = vg_create_context(pipe, visual, share); + /*debug_printf("--------- CREATE CONTEXT %p\n", ctx);*/ + return ctx; +} + +void st_destroy_context(struct vg_context *st) +{ + /*debug_printf("--------- DESTROY CONTEXT %p\n", st);*/ + vg_destroy_context(st); +} + +void st_copy_context_state(struct vg_context *dst, struct vg_context *src, + uint mask) +{ + fprintf(stderr, "FIXME: %s\n", __FUNCTION__); +} + +void st_get_framebuffer_dimensions(struct st_framebuffer *stfb, + uint *width, + uint *height) +{ + *width = stfb->strb->width; + *height = stfb->strb->height; +} + +struct st_framebuffer * st_create_framebuffer(const void *visual, + enum pipe_format colorFormat, + enum pipe_format depthFormat, + enum pipe_format stencilFormat, + uint width, uint height, + void *privateData) +{ + struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); + if (stfb) { + struct st_renderbuffer *rb = + st_new_renderbuffer_fb(colorFormat); + stfb->strb = rb; +#if 0 + if (doubleBuffer) { + struct st_renderbuffer *rb = + st_new_renderbuffer_fb(colorFormat); + } +#endif + + /* we want to combine the depth/stencil */ + if (stencilFormat == depthFormat) + stfb->dsrb = st_new_renderbuffer_fb(stencilFormat); + else + stfb->dsrb = st_new_renderbuffer_fb(PIPE_FORMAT_S8Z24_UNORM); + + /*### currently we always allocate it but it's possible it's + not necessary if EGL_ALPHA_MASK_SIZE was 0 + */ + stfb->alpha_mask = 0; + + stfb->init_width = width; + stfb->init_height = height; + stfb->privateData = privateData; + } + + return stfb; +} + +static void setup_new_alpha_mask(struct vg_context *ctx, + struct st_framebuffer *stfb, + uint width, uint height) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_texture *old_texture = stfb->alpha_mask; + + /* + we use PIPE_FORMAT_A8R8G8B8_UNORM because we want to render to + this texture and use it as a sampler, so while this wastes some + space it makes both of those a lot simpler + */ + stfb->alpha_mask = + create_texture(pipe, PIPE_FORMAT_A8R8G8B8_UNORM, width, height); + + if (!stfb->alpha_mask) { + if (old_texture) + pipe_texture_reference(&old_texture, NULL); + return; + } + + vg_validate_state(ctx); + + /* alpha mask starts with 1.f alpha */ + mask_fill(0, 0, width, height, 1.f); + + /* if we had an old surface copy it over */ + if (old_texture) { + struct pipe_surface *surface = pipe->screen->get_tex_surface( + pipe->screen, + stfb->alpha_mask, + 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + struct pipe_surface *old_surface = pipe->screen->get_tex_surface( + pipe->screen, + old_texture, + 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ); + pipe->surface_copy(pipe, + surface, + 0, 0, + old_surface, + 0, 0, + MIN2(old_surface->width, width), + MIN2(old_surface->height, height)); + if (surface) + pipe_surface_reference(&surface, NULL); + if (old_surface) + pipe_surface_reference(&old_surface, NULL); + } + + /* Free the old texture + */ + if (old_texture) + pipe_texture_reference(&old_texture, NULL); +} + +void st_resize_framebuffer(struct st_framebuffer *stfb, + uint width, uint height) +{ + struct vg_context *ctx = vg_current_context(); + struct st_renderbuffer *strb = stfb->strb; + struct pipe_framebuffer_state *state; + + if (!ctx) + return; + + state = &ctx->state.g3d.fb; + + /* If this is a noop, exit early and don't do the clear, etc below. + */ + if (strb->width == width && + strb->height == height && + state->zsbuf) + return; + + if (strb->width != width || strb->height != height) + st_renderbuffer_alloc_storage(ctx, strb, + width, height); + + if (stfb->dsrb->width != width || stfb->dsrb->height != height) + st_renderbuffer_alloc_storage(ctx, stfb->dsrb, + width, height); + + { + VGuint i; + + memset(state, 0, sizeof(struct pipe_framebuffer_state)); + + state->width = width; + state->height = height; + + state->nr_cbufs = 1; + state->cbufs[0] = strb->surface; + for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i) + state->cbufs[i] = 0; + + state->zsbuf = stfb->dsrb->surface; + + cso_set_framebuffer(ctx->cso_context, state); + } + + ctx->state.dirty |= VIEWPORT_DIRTY; + ctx->state.dirty |= DEPTH_STENCIL_DIRTY;/*to reset the scissors*/ + + ctx->pipe->clear(ctx->pipe, PIPE_CLEAR_DEPTHSTENCIL, + NULL, 0.0, 0); + + /* we need all the other state already set */ + + setup_new_alpha_mask(ctx, stfb, width, height); + + pipe_texture_reference( &stfb->blend_texture, NULL ); + stfb->blend_texture = create_texture(ctx->pipe, PIPE_FORMAT_A8R8G8B8_UNORM, + width, height); +} + +void st_set_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface *surf) +{ + struct st_renderbuffer *rb = stfb->strb; + + /* unreference existing surfaces */ + pipe_surface_reference( &rb->surface, NULL ); + pipe_texture_reference( &rb->texture, NULL ); + + /* reference new ones */ + pipe_surface_reference( &rb->surface, surf ); + pipe_texture_reference( &rb->texture, surf->texture ); + + rb->width = surf->width; + rb->height = surf->height; +} + +int st_get_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface **surf) +{ + struct st_renderbuffer *rb = stfb->strb; + *surf = rb->surface; + return VG_TRUE; +} + +int st_get_framebuffer_texture(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_texture **tex) +{ + struct st_renderbuffer *rb = stfb->strb; + *tex = rb->texture; + return VG_TRUE; +} + +void * st_framebuffer_private(struct st_framebuffer *stfb) +{ + return stfb->privateData; +} + +void st_unreference_framebuffer(struct st_framebuffer *stfb) +{ + /* FIXME */ +} + +void st_make_current(struct vg_context *st, + struct st_framebuffer *draw, + struct st_framebuffer *read) +{ + vg_set_current_context(st); + if (st) { + st->draw_buffer = draw; + } +} + +struct vg_context *st_get_current(void) +{ + return vg_current_context(); +} + +void st_flush(struct vg_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence) +{ + st->pipe->flush(st->pipe, pipeFlushFlags, fence); +} + +void st_finish(struct vg_context *st) +{ + struct pipe_fence_handle *fence = NULL; + + st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); + + st->pipe->screen->fence_finish(st->pipe->screen, fence, 0); + st->pipe->screen->fence_reference(st->pipe->screen, &fence, NULL); +} + +void st_notify_swapbuffers(struct st_framebuffer *stfb) +{ + struct vg_context *ctx = vg_current_context(); + if (ctx && ctx->draw_buffer == stfb) { + st_flush(ctx, + PIPE_FLUSH_RENDER_CACHE | + PIPE_FLUSH_SWAPBUFFERS | + PIPE_FLUSH_FRAME, + NULL); + } +} + +void st_notify_swapbuffers_complete(struct st_framebuffer *stfb) +{ +} + +int st_bind_texture_surface(struct pipe_surface *ps, int target, int level, + enum pipe_format format) +{ + return 0; +} + +int st_unbind_texture_surface(struct pipe_surface *ps, int target, int level) +{ + return 0; +} diff --git a/src/gallium/state_trackers/vega/vg_tracker.h b/src/gallium/state_trackers/vega/vg_tracker.h new file mode 100644 index 00000000000..5457631106f --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_tracker.h @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef VG_TRACKER_H +#define VG_TRACKER_H + +#include "VG/openvg.h" + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" + +#define ST_SURFACE_FRONT_LEFT 0 +#define ST_SURFACE_BACK_LEFT 1 +#define ST_SURFACE_FRONT_RIGHT 2 +#define ST_SURFACE_BACK_RIGHT 3 +#define ST_SURFACE_DEPTH 8 + +struct vg_context; +struct st_framebuffer; +struct pipe_context; +struct pipe_fence_handle; +struct pipe_surface; + + +struct vg_context *st_create_context(struct pipe_context *pipe, + const void *visual, + struct vg_context *share); + +void st_destroy_context( struct vg_context *st ); + +void st_copy_context_state(struct vg_context *dst, struct vg_context *src, + uint mask); + +struct st_framebuffer *st_create_framebuffer(const void *visual, + enum pipe_format colorFormat, + enum pipe_format depthFormat, + enum pipe_format stencilFormat, + uint width, uint height, + void *privateData); + +void st_resize_framebuffer(struct st_framebuffer *stfb, + uint width, uint height); + +void st_set_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface *surf); + +void st_get_framebuffer_dimensions( struct st_framebuffer *stfb, + uint *width, uint *height); + +int st_bind_texture_surface(struct pipe_surface *ps, int target, int level, + enum pipe_format format); + +int st_unbind_texture_surface(struct pipe_surface *ps, int target, int level); + +int st_get_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface **surf); + +int st_get_framebuffer_texture(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_texture **tex); + +void *st_framebuffer_private(struct st_framebuffer *stfb); + +void st_unreference_framebuffer(struct st_framebuffer *stfb); + +void st_make_current(struct vg_context *st, + struct st_framebuffer *draw, + struct st_framebuffer *read); + +struct vg_context *st_get_current(void); + +void st_flush(struct vg_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence); +void st_finish(struct vg_context *st); + +void st_notify_swapbuffers(struct st_framebuffer *stfb); +void st_notify_swapbuffers_complete(struct st_framebuffer *stfb); + + +/** Generic function type */ +typedef void (*st_proc)(); + +st_proc st_get_proc_address(const char *procname); + +#endif diff --git a/src/gallium/state_trackers/vega/vg_translate.c b/src/gallium/state_trackers/vega/vg_translate.c new file mode 100644 index 00000000000..00e07647062 --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_translate.c @@ -0,0 +1,1030 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "vg_translate.h" + +#include "pipe/p_format.h" +#include "util/u_pack_color.h" + +void _vega_pack_rgba_span_float(struct vg_context *ctx, + VGuint n, VGfloat rgba[][4], + VGImageFormat dstFormat, + void *dstAddr) +{ + VGint i; + + switch (dstFormat) { + case VG_sRGBX_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = 255; + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + break; + case VG_sRGBA_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + break; + case VG_sRGBA_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + break; + case VG_sRGB_565: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + r = (r / 255.0) * 32; + g = (g / 255.0) * 32; + b = (b / 255.0) * 32; + + dst[i] = b | g << 5 | r << 11; + } + return; + } + break; + case VG_sRGBA_5551: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + r = (r / 255.0) * 32; + g = (g / 255.0) * 32; + b = (b / 255.0) * 32; + a = (a / 255.0); + + dst[i] = a | b << 1 | g << 6 | r << 11; + } + return; + } + break; + case VG_sRGBA_4444: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + r = (r / 255.0) * 16; + g = (g / 255.0) * 16; + b = (b / 255.0) * 16; + a = (a / 255.0) * 16; + + dst[i] = a | b << 4 | g << 8 | r << 12; + } + return; + } + break; + case VG_sL_8: { + VGubyte *dst = (VGubyte *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + + dst[i] = a; + } + return; + } + break; + case VG_lRGBX_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = 255; + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + break; + case VG_lRGBA_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + case VG_lRGBA_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = r << 24 | g << 16 | b << 8 | a; + } + return; + } + break; + case VG_lL_8: { + VGubyte *dst = (VGubyte *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a; + } + return; + } + break; + case VG_A_8: { + VGubyte *dst = (VGubyte *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + + dst[i] = a; + } + return; + } + break; + case VG_BW_1: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + VGubyte res; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + + res = (r + g + b + a)/4; + dst[i] = (res & (128)); + } + return; + } + break; +#ifdef OPENVG_VERSION_1_1 + case VG_A_1: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + + dst[i] = (a & (128)); + } + return; + } + break; + case VG_A_4: { + VGshort *dst = (VGshort *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b, a; + VGubyte res; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + + res = a/4; + dst[i] = (res & (128)); + } + return; + } + break; +#endif + case VG_sXRGB_8888: + break; + case VG_sARGB_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | r << 16 | g << 8 | b; + } + return; + } + break; + case VG_sARGB_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | r << 16 | g << 8 | b; + } + return; + } + break; + case VG_sARGB_1555: + break; + case VG_sARGB_4444: + break; + case VG_lXRGB_8888: + break; + case VG_lARGB_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | r << 16 | g << 8 | b; + } + return; + } + break; + case VG_lARGB_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | r << 16 | g << 8 | b; + } + return; + } + break; + case VG_sBGRX_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = 0xff; + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_sBGRA_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_sBGRA_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_sBGR_565: + break; + case VG_sBGRA_5551: + break; + case VG_sBGRA_4444: + break; + case VG_lBGRX_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = 0xff; + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_lBGRA_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_lBGRA_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = b << 24 | g << 16 | r << 8 | a; + } + return; + } + break; + case VG_sXBGR_8888: + break; + case VG_sABGR_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | b << 16 | g << 8 | r; + } + return; + } + break; + case VG_sABGR_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | b << 16 | g << 8 | r; + } + return; + } + break; + case VG_sABGR_1555: + break; + case VG_sABGR_4444: + break; + case VG_lXBGR_8888: + break; + case VG_lABGR_8888: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | b << 16 | g << 8 | r; + } + return; + } + break; + case VG_lABGR_8888_PRE: { + VGint *dst = (VGint *)dstAddr; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = float_to_ubyte(rgba[i][0]); + g = float_to_ubyte(rgba[i][1]); + b = float_to_ubyte(rgba[i][2]); + a = float_to_ubyte(rgba[i][3]); + dst[i] = a << 24 | b << 16 | g << 8 | r; + } + return; + } + break; + default: + assert(!"Unknown ReadPixels format"); + break; + } + assert(!"Not implemented ReadPixels format"); +} + +void _vega_unpack_float_span_rgba(struct vg_context *ctx, + VGuint n, + VGuint offset, + const void * data, + VGImageFormat dataFormat, + VGfloat rgba[][4]) +{ + VGint i; + + switch (dataFormat) { + case VG_sRGBX_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_sRGBA_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sRGBA_8888_PRE: { + VGint *src = (VGint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sRGB_565: { + VGshort *src = (VGshort *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGfloat clr[4]; + clr[0] = ((*src >> 10) & 31)/31.; + clr[1] = ((*src >> 5) & 95)/95.; + clr[2] = ((*src >> 0) & 31)/31.; + clr[3] = 1.f; + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_sRGBA_5551: { + VGshort *src = (VGshort *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGfloat clr[4]; + clr[0] = ((*src >> 10) & 31)/31.; + clr[1] = ((*src >> 5) & 31)/31.; + clr[2] = ((*src >> 1) & 31)/31.; + clr[3] = ((*src >> 0) & 1)/1.; + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_sRGBA_4444: { + VGshort *src = (VGshort *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGfloat clr[4]; + clr[0] = ((*src >> 12) & 15)/15.; + clr[1] = ((*src >> 8) & 15)/15.; + clr[2] = ((*src >> 4) & 15)/15.; + clr[3] = ((*src >> 0) & 15)/15.; + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_sL_8: { + VGubyte *src = (VGubyte *)data; + src += offset; + for (i = 0; i < n; ++i) { + util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_lRGBX_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_lRGBA_8888: { + VGint *src = (VGint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_lRGBA_8888_PRE: { + VGint *src = (VGint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + r = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + b = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_lL_8: { + VGubyte *src = (VGubyte *)data; + src += offset; + for (i = 0; i < n; ++i) { + util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_A_8: { + VGubyte *src = (VGubyte *)data; + src += offset; + for (i = 0; i < n; ++i) { + util_pack_color_ub(0xff, 0xff, 0xff, *src, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + } + return; + case VG_BW_1: { + VGubyte *src = (VGubyte *)data; + src += offset; + for (i = 0; i < n; i += 8) { + VGfloat clr[4]; + VGint j; + for (j = 0; j < 8 && j < n ; ++j) { + VGint shift = j; + clr[0] = (((*src) & (1<<shift)) >> shift); + clr[1] = clr[0]; + clr[2] = clr[0]; + clr[3] = 1.f; + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i+j]); + } + ++src; + } + } + return; +#ifdef OPENVG_VERSION_1_1 + case VG_A_1: { + VGubyte *src = (VGubyte *)data; + src += offset; + for (i = 0; i < n; i += 8) { + VGfloat clr[4]; + VGint j; + for (j = 0; j < 8 && j < n ; ++j) { + VGint shift = j; + clr[0] = 0.f; + clr[1] = 0.f; + clr[2] = 0.f; + clr[3] = (((*src) & (1<<shift)) >> shift); + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i+j]); + } + ++src; + } + } + return; + case VG_A_4: { + VGubyte *src = (VGubyte *)data; + src += offset/2; + for (i = 0; i < n; i += 2) { + VGfloat clr[4]; + VGint j; + for (j = 0; j < n && j < 2; ++j) { + VGint bitter, shift; + if (j == 0) { + bitter = 0x0f; + shift = 0; + } else { + bitter = 0xf0; + shift = 4; + } + clr[0] = 0.f; + clr[1] = 0.f; + clr[2] = 0.f; + clr[3] = ((*src) & (bitter)) >> shift; + + util_pack_color(clr, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i +j]); + } + ++src; + } + } + return; +#endif + case VG_sXRGB_8888: + break; + case VG_sARGB_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + r = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + b = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sARGB_8888_PRE: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + r = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + b = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sARGB_1555: + break; + case VG_sARGB_4444: + break; + case VG_lXRGB_8888: + break; + case VG_lARGB_8888: { + VGint *src = (VGint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + r = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + b = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_lARGB_8888_PRE: { + VGint *src = (VGint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + r = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + b = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sBGRX_8888: + break; + case VG_sBGRA_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + b = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + r = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sBGRA_8888_PRE: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + b = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + r = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sBGR_565: + break; + case VG_sBGRA_5551: + break; + case VG_sBGRA_4444: + break; + case VG_lBGRX_8888: + break; + case VG_lBGRA_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + b = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + r = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_lBGRA_8888_PRE: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + b = (*src >> 24) & 0xff; + g = (*src >> 16) & 0xff; + r = (*src >> 8) & 0xff; + a = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sXBGR_8888: + break; + case VG_sABGR_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + b = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + r = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sABGR_8888_PRE: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + b = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + r = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_sABGR_1555: + break; + case VG_sABGR_4444: + break; + case VG_lXBGR_8888: + break; + case VG_lABGR_8888: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + b = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + r = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + case VG_lABGR_8888_PRE: { + VGuint *src = (VGuint *)data; + src += offset; + for (i = 0; i < n; ++i) { + VGubyte r, g, b ,a; + a = (*src >> 24) & 0xff; + b = (*src >> 16) & 0xff; + g = (*src >> 8) & 0xff; + r = (*src >> 0) & 0xff; + + util_pack_color_ub(r, g, b, a, PIPE_FORMAT_R32G32B32A32_FLOAT, + rgba[i]); + ++src; + } + return; + } + break; + default: + assert(!"Unknown ReadPixels format"); + break; + } + assert(!"Not implemented ReadPixels format"); +} + +VGint _vega_size_for_format(VGImageFormat dataFormat) +{ + switch (dataFormat) { + case VG_sRGBX_8888: + case VG_sRGBA_8888: + case VG_sRGBA_8888_PRE: + return 4; + case VG_sRGB_565: + case VG_sRGBA_5551: + case VG_sRGBA_4444: + return 2; + case VG_sL_8: + return 1; + case VG_lRGBX_8888: + case VG_lRGBA_8888: + case VG_lRGBA_8888_PRE: + return 4; + case VG_lL_8: + return 1; + case VG_A_8: + return 1; + case VG_BW_1: + return 1; +#ifdef OPENVG_VERSION_1_1 + case VG_A_1: + break; + case VG_A_4: + break; +#endif + case VG_sXRGB_8888: + case VG_sARGB_8888: + case VG_sARGB_8888_PRE: + return 4; + case VG_sARGB_1555: + case VG_sARGB_4444: + return 2; + case VG_lXRGB_8888: + case VG_lARGB_8888: + case VG_lARGB_8888_PRE: + case VG_sBGRX_8888: + case VG_sBGRA_8888: + case VG_sBGRA_8888_PRE: + return 4; + case VG_sBGR_565: + case VG_sBGRA_5551: + case VG_sBGRA_4444: + return 2; + case VG_lBGRX_8888: + case VG_lBGRA_8888: + case VG_lBGRA_8888_PRE: + case VG_sXBGR_8888: + case VG_sABGR_8888: + case VG_sABGR_8888_PRE: + return 4; + case VG_sABGR_1555: + case VG_sABGR_4444: + return 2; + case VG_lXBGR_8888: + case VG_lABGR_8888: + case VG_lABGR_8888_PRE: + return 4; + default: + assert(!"Unknown ReadPixels format"); + break; + } + assert(!"Not implemented ReadPixels format"); + return 0; +} diff --git a/src/gallium/state_trackers/vega/vg_translate.h b/src/gallium/state_trackers/vega/vg_translate.h new file mode 100644 index 00000000000..70815bacbcb --- /dev/null +++ b/src/gallium/state_trackers/vega/vg_translate.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef VG_TRANSLATE_H +#define VG_TRANSLATE_H + +#include "VG/openvg.h" +#include "vg_context.h" + +/*FIXME: we really should be using translate module + * but pipe_format can't express some of the VG formats + * (the premultiplied ones) so currently it won't work */ + +void _vega_pack_rgba_span_float(struct vg_context *ctx, + VGuint n, VGfloat rgba[][4], + VGImageFormat dstFormat, + void *dstAddr); +void _vega_unpack_float_span_rgba(struct vg_context *ctx, + VGuint n, + VGuint offset, + const void * data, + VGImageFormat dataFormat, + VGfloat rgba[][4]); +VGint _vega_size_for_format(VGImageFormat format); + +#endif diff --git a/src/gallium/state_trackers/vega/vgu.c b/src/gallium/state_trackers/vega/vgu.c new file mode 100644 index 00000000000..7dc51c5599c --- /dev/null +++ b/src/gallium/state_trackers/vega/vgu.c @@ -0,0 +1,450 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "VG/openvg.h" +#include "VG/vgu.h" + +#include "matrix.h" +#include "path.h" + +#include "util/u_debug.h" +#include "util/u_pointer.h" + +#include <math.h> +#include <assert.h> + +static VGboolean is_aligned_to(const void *ptr, VGbyte alignment) +{ + void *aligned = align_pointer(ptr, alignment); + return (ptr == aligned) ? VG_TRUE : VG_FALSE; +} + +static VGboolean is_aligned(const void *ptr) +{ + return is_aligned_to(ptr, 4); +} + +static void vgu_append_float_coords(VGPath path, + const VGubyte *cmds, + VGint num_cmds, + const VGfloat *coords, + VGint num_coords) +{ + VGubyte common_data[40 * sizeof(VGfloat)]; + struct path *p = (struct path *)path; + + vg_float_to_datatype(path_datatype(p), common_data, coords, num_coords); + vgAppendPathData(path, num_cmds, cmds, common_data); +} + +VGUErrorCode vguLine(VGPath path, + VGfloat x0, VGfloat y0, + VGfloat x1, VGfloat y1) +{ + static const VGubyte cmds[] = {VG_MOVE_TO_ABS, VG_LINE_TO_ABS}; + VGfloat coords[4]; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + + coords[0] = x0; + coords[1] = y0; + coords[2] = x1; + coords[3] = y1; + + vgu_append_float_coords(path, cmds, 2, coords, 4); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguPolygon(VGPath path, + const VGfloat * points, + VGint count, + VGboolean closed) +{ + VGubyte *cmds; + VGfloat *coords; + VGbitfield caps; + VGint i; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + + if (!points || count <= 0 || !is_aligned(points)) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + + cmds = malloc(sizeof(VGubyte) * count + 1); + coords = malloc(sizeof(VGfloat) * count * 2); + + cmds[0] = VG_MOVE_TO_ABS; + coords[0] = points[0]; + coords[1] = points[1]; + for (i = 1; i < count; ++i) { + cmds[i] = VG_LINE_TO_ABS; + coords[2*i + 0] = points[2*i + 0]; + coords[2*i + 1] = points[2*i + 1]; + } + + if (closed) { + cmds[i] = VG_CLOSE_PATH; + ++i; + } + + vgu_append_float_coords(path, cmds, i, coords, 2*i); + + free(cmds); + free(coords); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguRect(VGPath path, + VGfloat x, VGfloat y, + VGfloat width, VGfloat height) +{ + static const VGubyte cmds[] = {VG_MOVE_TO_ABS, + VG_HLINE_TO_REL, + VG_VLINE_TO_REL, + VG_HLINE_TO_REL, + VG_CLOSE_PATH + }; + VGfloat coords[5]; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + if (width <= 0 || height <= 0) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + + coords[0] = x; + coords[1] = y; + coords[2] = width; + coords[3] = height; + coords[4] = -width; + + vgu_append_float_coords(path, cmds, 5, coords, 5); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguRoundRect(VGPath path, + VGfloat x, VGfloat y, + VGfloat width, + VGfloat height, + VGfloat arcWidth, + VGfloat arcHeight) +{ + static const VGubyte cmds[] = {VG_MOVE_TO_ABS, + VG_HLINE_TO_REL, + VG_SCCWARC_TO_REL, + VG_VLINE_TO_REL, + VG_SCCWARC_TO_REL, + VG_HLINE_TO_REL, + VG_SCCWARC_TO_REL, + VG_VLINE_TO_REL, + VG_SCCWARC_TO_REL, + VG_CLOSE_PATH + }; + VGfloat c[26]; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + if (width <= 0 || height <= 0) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + + c[0] = x + arcWidth/2; c[1] = y; + + c[2] = width - arcWidth; + + c[3] = arcWidth/2; c[4] = arcHeight/2; c[5] = 0; + c[6] = arcWidth/2; c[7] = arcHeight/2; + + c[8] = height - arcHeight; + + c[9] = arcWidth/2; c[10] = arcHeight/2; c[11] = 0; + c[12] = -arcWidth/2; c[13] = arcHeight/2; + + c[14] = -(width - arcWidth); + + c[15] = arcWidth/2; c[16] = arcHeight/2; c[17] = 0; + c[18] = -arcWidth/2; c[19] = -arcHeight/2; + + c[20] = -(height - arcHeight); + + c[21] = arcWidth/2; c[22] = arcHeight/2; c[23] = 0; + c[24] = arcWidth/2; c[25] = -arcHeight/2; + + vgu_append_float_coords(path, cmds, 10, c, 26); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguEllipse(VGPath path, + VGfloat cx, VGfloat cy, + VGfloat width, + VGfloat height) +{ + static const VGubyte cmds[] = {VG_MOVE_TO_ABS, + VG_SCCWARC_TO_REL, + VG_SCCWARC_TO_REL, + VG_CLOSE_PATH + }; + VGfloat coords[12]; + VGbitfield caps; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + if (width <= 0 || height <= 0) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + + coords[0] = cx + width/2; coords[1] = cy; + + coords[2] = width/2; coords[3] = height/2; coords[4] = 0; + coords[5] = -width; coords[6] = 0; + + coords[7] = width/2; coords[8] = height/2; coords[9] = 0; + coords[10] = width; coords[11] = 0; + + vgu_append_float_coords(path, cmds, 4, coords, 11); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguArc(VGPath path, + VGfloat x, VGfloat y, + VGfloat width, VGfloat height, + VGfloat startAngle, + VGfloat angleExtent, + VGUArcType arcType) +{ + VGubyte cmds[11]; + VGfloat coords[40]; + VGbitfield caps; + VGfloat last = startAngle + angleExtent; + VGint i, c = 0; + + if (path == VG_INVALID_HANDLE) { + return VGU_BAD_HANDLE_ERROR; + } + caps = vgGetPathCapabilities(path); + if (!(caps & VG_PATH_CAPABILITY_APPEND_TO)) { + return VGU_PATH_CAPABILITY_ERROR; + } + if (width <= 0 || height <= 0) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + if (arcType != VGU_ARC_OPEN && + arcType != VGU_ARC_CHORD && + arcType != VGU_ARC_PIE) { + return VGU_ILLEGAL_ARGUMENT_ERROR; + } + + cmds[c] = VG_MOVE_TO_ABS; ++c; + coords[0] = x+cos(DEGREES_TO_RADIANS(startAngle))*width/2; + coords[1] = y+sin(DEGREES_TO_RADIANS(startAngle))*height/2; +#ifdef DEBUG_VGUARC + debug_printf("start [%f, %f]\n", coords[0], coords[1]); +#endif + i = 2; + if (angleExtent > 0) { + VGfloat angle = startAngle + 180; + while (angle < last) { + cmds[c] = VG_SCCWARC_TO_ABS; ++c; + coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0; + coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle))*width/2; + coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle))*height/2; +#ifdef DEBUG_VGUARC + debug_printf("1 [%f, %f]\n", coords[i+3], + coords[i+4]); +#endif + i += 5; + angle += 180; + } + cmds[c] = VG_SCCWARC_TO_ABS; ++c; + coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0; + coords[i+3] = x+cos(DEGREES_TO_RADIANS(last))*width/2; + coords[i+4] = y+sin(DEGREES_TO_RADIANS(last))*height/2; +#ifdef DEBUG_VGUARC + debug_printf("2 [%f, %f]\n", coords[i+3], + coords[i+4]); +#endif + i += 5; + } else { + VGfloat angle = startAngle - 180; + while (angle > last) { + cmds[c] = VG_SCWARC_TO_ABS; ++c; + coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0; + coords[i+3] = x + cos(DEGREES_TO_RADIANS(angle)) * width/2; + coords[i+4] = y + sin(DEGREES_TO_RADIANS(angle)) * height/2; +#ifdef DEBUG_VGUARC + debug_printf("3 [%f, %f]\n", coords[i+3], + coords[i+4]); +#endif + angle -= 180; + i += 5; + } + cmds[c] = VG_SCWARC_TO_ABS; ++c; + coords[i] = width/2; coords[i+1] = height/2; coords[i+2] = 0; + coords[i+3] = x + cos(DEGREES_TO_RADIANS(last)) * width/2; + coords[i+4] = y + sin(DEGREES_TO_RADIANS(last)) * height/2; +#ifdef DEBUG_VGUARC + debug_printf("4 [%f, %f]\n", coords[i+3], + coords[i+4]); +#endif + i += 5; + } + + if (arcType == VGU_ARC_PIE) { + cmds[c] = VG_LINE_TO_ABS; ++c; + coords[i] = x; coords[i + 1] = y; + i += 2; + } + if (arcType == VGU_ARC_PIE || arcType == VGU_ARC_CHORD) { + cmds[c] = VG_CLOSE_PATH; + ++c; + } + + assert(c < 11); + + vgu_append_float_coords(path, cmds, c, coords, i); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguComputeWarpQuadToSquare(VGfloat sx0, VGfloat sy0, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + VGfloat sx3, VGfloat sy3, + VGfloat * matrix) +{ + struct matrix mat; + + if (!matrix || !is_aligned(matrix)) + return VGU_ILLEGAL_ARGUMENT_ERROR; + + if (!matrix_quad_to_square(sx0, sy0, + sx1, sy1, + sx2, sy2, + sx3, sy3, + &mat)) + return VGU_BAD_WARP_ERROR; + + if (!matrix_is_invertible(&mat)) + return VGU_BAD_WARP_ERROR; + + memcpy(matrix, mat.m, sizeof(VGfloat) * 9); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguComputeWarpSquareToQuad(VGfloat dx0, VGfloat dy0, + VGfloat dx1, VGfloat dy1, + VGfloat dx2, VGfloat dy2, + VGfloat dx3, VGfloat dy3, + VGfloat * matrix) +{ + struct matrix mat; + + if (!matrix || !is_aligned(matrix)) + return VGU_ILLEGAL_ARGUMENT_ERROR; + + if (!matrix_square_to_quad(dx0, dy0, + dx1, dy1, + dx2, dy2, + dx3, dy3, + &mat)) + return VGU_BAD_WARP_ERROR; + + if (!matrix_is_invertible(&mat)) + return VGU_BAD_WARP_ERROR; + + memcpy(matrix, mat.m, sizeof(VGfloat) * 9); + + return VGU_NO_ERROR; +} + +VGUErrorCode vguComputeWarpQuadToQuad(VGfloat dx0, VGfloat dy0, + VGfloat dx1, VGfloat dy1, + VGfloat dx2, VGfloat dy2, + VGfloat dx3, VGfloat dy3, + VGfloat sx0, VGfloat sy0, + VGfloat sx1, VGfloat sy1, + VGfloat sx2, VGfloat sy2, + VGfloat sx3, VGfloat sy3, + VGfloat * matrix) +{ + struct matrix mat; + + if (!matrix || !is_aligned(matrix)) + return VGU_ILLEGAL_ARGUMENT_ERROR; + + if (!matrix_quad_to_quad(dx0, dy0, + dx1, dy1, + dx2, dy2, + dx3, dy3, + sx0, sy0, + sx1, sy1, + sx2, sy2, + sx3, sy3, + &mat)) + return VGU_BAD_WARP_ERROR; + + memcpy(matrix, mat.m, sizeof(VGfloat) * 9); + + return VGU_NO_ERROR; +} diff --git a/src/gallium/state_trackers/wgl/SConscript b/src/gallium/state_trackers/wgl/SConscript index 5bbcc7175f7..69b88618ecb 100644 --- a/src/gallium/state_trackers/wgl/SConscript +++ b/src/gallium/state_trackers/wgl/SConscript @@ -27,7 +27,9 @@ if env['platform'] in ['windows']: 'shared/stw_framebuffer.c', 'shared/stw_pixelformat.c', 'shared/stw_extensionsstring.c', + 'shared/stw_extswapinterval.c', 'shared/stw_getprocaddress.c', + 'shared/stw_extgallium.c', 'shared/stw_arbpixelformat.c', 'shared/stw_tls.c', ] diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c index 662b5fbcd2a..4968ecc692d 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_context.c +++ b/src/gallium/state_trackers/wgl/shared/stw_context.c @@ -47,6 +47,29 @@ #include "stw_context.h" #include "stw_tls.h" + +static INLINE struct stw_context * +stw_context(GLcontext *glctx) +{ + if(!glctx) + return NULL; + assert(glctx->DriverCtx); + return (struct stw_context *)glctx->DriverCtx; +} + +static INLINE struct stw_context * +stw_current_context(void) +{ + /* We must check if multiple threads are being used or GET_CURRENT_CONTEXT + * might return the current context of the thread first seen. */ + _glapi_check_multithread(); + + { + GET_CURRENT_CONTEXT( glctx ); + return stw_context(glctx); + } +} + BOOL stw_copy_context( UINT_PTR hglrcSrc, @@ -57,7 +80,7 @@ stw_copy_context( struct stw_context *dst; BOOL ret = FALSE; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); src = stw_lookup_context_locked( hglrcSrc ); dst = stw_lookup_context_locked( hglrcDst ); @@ -70,7 +93,7 @@ stw_copy_context( (void) mask; } - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); return ret; } @@ -84,7 +107,7 @@ stw_share_lists( struct stw_context *ctx2; BOOL ret = FALSE; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); ctx1 = stw_lookup_context_locked( hglrc1 ); ctx2 = stw_lookup_context_locked( hglrc2 ); @@ -94,11 +117,25 @@ stw_share_lists( ret = _mesa_share_state(ctx2->st->ctx, ctx1->st->ctx); } - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); return ret; } +static void +stw_viewport(GLcontext * glctx, GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct stw_context *ctx = (struct stw_context *)glctx->DriverCtx; + struct stw_framebuffer *fb; + + fb = stw_framebuffer_from_hdc( ctx->hdc ); + if(fb) { + stw_framebuffer_update(fb); + stw_framebuffer_release(fb); + } +} + UINT_PTR stw_create_layer_context( HDC hdc, @@ -158,10 +195,11 @@ stw_create_layer_context( goto no_st_ctx; ctx->st->ctx->DriverCtx = ctx; + ctx->st->ctx->Driver.Viewport = stw_viewport; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); ctx->hglrc = handle_table_add(stw_dev->ctx_table, ctx); - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); if (!ctx->hglrc) goto no_hglrc; @@ -188,17 +226,16 @@ stw_delete_context( if (!stw_dev) return FALSE; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); ctx = stw_lookup_context_locked(hglrc); handle_table_remove(stw_dev->ctx_table, hglrc); - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); if (ctx) { - GLcontext *glctx = ctx->st->ctx; - GET_CURRENT_CONTEXT( glcurctx ); - + struct stw_context *curctx = stw_current_context(); + /* Unbind current if deleting current context. */ - if (glcurctx == glctx) + if (curctx == ctx) st_make_current( NULL, NULL, NULL ); st_destroy_context(ctx->st); @@ -219,9 +256,9 @@ stw_release_context( if (!stw_dev) return FALSE; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); ctx = stw_lookup_context_locked( hglrc ); - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); if (!ctx) return FALSE; @@ -230,13 +267,8 @@ stw_release_context( * current for this thread. We should check that and return False * if not the case. */ - { - GLcontext *glctx = ctx->st->ctx; - GET_CURRENT_CONTEXT( glcurctx ); - - if (glcurctx != glctx) - return FALSE; - } + if (ctx != stw_current_context()) + return FALSE; if (stw_make_current( NULL, 0 ) == FALSE) return FALSE; @@ -248,14 +280,9 @@ stw_release_context( UINT_PTR stw_get_current_context( void ) { - GET_CURRENT_CONTEXT( glcurctx ); struct stw_context *ctx; - if(!glcurctx) - return 0; - - ctx = (struct stw_context *)glcurctx->DriverCtx; - assert(ctx); + ctx = stw_current_context(); if(!ctx) return 0; @@ -265,14 +292,9 @@ stw_get_current_context( void ) HDC stw_get_current_dc( void ) { - GET_CURRENT_CONTEXT( glcurctx ); struct stw_context *ctx; - if(!glcurctx) - return NULL; - - ctx = (struct stw_context *)glcurctx->DriverCtx; - assert(ctx); + ctx = stw_current_context(); if(!ctx) return NULL; @@ -284,36 +306,37 @@ stw_make_current( HDC hdc, UINT_PTR hglrc ) { - struct stw_context *ctx; - GET_CURRENT_CONTEXT( glcurctx ); - struct stw_framebuffer *fb; + struct stw_context *curctx = NULL; + struct stw_context *ctx = NULL; + struct stw_framebuffer *fb = NULL; if (!stw_dev) goto fail; - if (glcurctx != NULL) { - struct stw_context *curctx; - curctx = (struct stw_context *) glcurctx->DriverCtx; - + curctx = stw_current_context(); + if (curctx != NULL) { if (curctx->hglrc != hglrc) - st_flush(glcurctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + st_flush(curctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); /* Return if already current. */ - if (curctx->hglrc == hglrc && curctx->hdc == hdc) - return TRUE; + if (curctx->hglrc == hglrc && curctx->hdc == hdc) { + ctx = curctx; + fb = stw_framebuffer_from_hdc( hdc ); + goto success; + } } if (hdc == NULL || hglrc == 0) { return st_make_current( NULL, NULL, NULL ); } - pipe_mutex_lock( stw_dev->mutex ); - + pipe_mutex_lock( stw_dev->ctx_mutex ); ctx = stw_lookup_context_locked( hglrc ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); if(!ctx) goto fail; - fb = stw_framebuffer_from_hdc_locked( hdc ); + fb = stw_framebuffer_from_hdc( hdc ); if(!fb) { /* Applications should call SetPixelFormat before creating a context, * but not all do, and the opengl32 runtime seems to use a default pixel @@ -321,13 +344,11 @@ stw_make_current( */ int iPixelFormat = GetPixelFormat(hdc); if(iPixelFormat) - fb = stw_framebuffer_create_locked( hdc, iPixelFormat ); + fb = stw_framebuffer_create( hdc, iPixelFormat ); if(!fb) goto fail; } - pipe_mutex_unlock( stw_dev->mutex ); - if(fb->iPixelFormat != ctx->iPixelFormat) goto fail; @@ -344,11 +365,18 @@ stw_make_current( if(!st_make_current( ctx->st, fb->stfb, fb->stfb )) goto fail; - stw_framebuffer_resize(fb); +success: + assert(fb); + if(fb) { + stw_framebuffer_update(fb); + stw_framebuffer_release(fb); + } return TRUE; fail: + if(fb) + stw_framebuffer_release(fb); st_make_current( NULL, NULL, NULL ); return FALSE; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.c b/src/gallium/state_trackers/wgl/shared/stw_device.c index 1a6b29807d7..0b6954915a6 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.c +++ b/src/gallium/state_trackers/wgl/shared/stw_device.c @@ -30,6 +30,7 @@ #include "glapi/glthread.h" #include "util/u_debug.h" #include "pipe/p_screen.h" +#include "state_tracker/st_public.h" #ifdef DEBUG #include "trace/tr_screen.h" @@ -63,15 +64,36 @@ stw_flush_frontbuffer(struct pipe_screen *screen, { const struct stw_winsys *stw_winsys = stw_dev->stw_winsys; HDC hdc = (HDC)context_private; + struct stw_framebuffer *fb; + fb = stw_framebuffer_from_hdc( hdc ); + /* fb can be NULL if window was destroyed already */ + if (fb) { +#if DEBUG + { + struct pipe_surface *surface2; + + if(!st_get_framebuffer_surface( fb->stfb, ST_SURFACE_FRONT_LEFT, &surface2 )) + assert(0); + else + assert(surface2 == surface); + } +#endif + #ifdef DEBUG - if(stw_dev->trace_running) { - screen = trace_screen(screen)->screen; - surface = trace_surface(surface)->surface; - } + if(stw_dev->trace_running) { + screen = trace_screen(screen)->screen; + surface = trace_surface(surface)->surface; + } #endif + } stw_winsys->flush_frontbuffer(screen, surface, hdc); + + if(fb) { + stw_framebuffer_update(fb); + stw_framebuffer_release(fb); + } } @@ -113,7 +135,8 @@ stw_init(const struct stw_winsys *stw_winsys) stw_dev->screen->flush_frontbuffer = &stw_flush_frontbuffer; - pipe_mutex_init( stw_dev->mutex ); + pipe_mutex_init( stw_dev->ctx_mutex ); + pipe_mutex_init( stw_dev->fb_mutex ); stw_dev->ctx_table = handle_table_create(); if (!stw_dev->ctx_table) { @@ -133,20 +156,13 @@ error1: boolean stw_init_thread(void) { - if (!stw_tls_init_thread()) - return FALSE; - - if (!stw_framebuffer_init_thread()) - return FALSE; - - return TRUE; + return stw_tls_init_thread(); } void stw_cleanup_thread(void) { - stw_framebuffer_cleanup_thread(); stw_tls_cleanup_thread(); } @@ -161,7 +177,7 @@ stw_cleanup(void) if (!stw_dev) return; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->ctx_mutex ); { /* Ensure all contexts are destroyed */ i = handle_table_get_first_handle(stw_dev->ctx_table); @@ -171,11 +187,12 @@ stw_cleanup(void) } handle_table_destroy(stw_dev->ctx_table); } - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->ctx_mutex ); stw_framebuffer_cleanup(); - pipe_mutex_destroy( stw_dev->mutex ); + pipe_mutex_destroy( stw_dev->fb_mutex ); + pipe_mutex_destroy( stw_dev->ctx_mutex ); stw_dev->screen->destroy(stw_dev->screen); diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.h b/src/gallium/state_trackers/wgl/shared/stw_device.h index e097f1f71e0..e1bb9518dd1 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_device.h +++ b/src/gallium/state_trackers/wgl/shared/stw_device.h @@ -57,10 +57,10 @@ struct stw_device unsigned pixelformat_count; unsigned pixelformat_extended_count; - pipe_mutex mutex; - + pipe_mutex ctx_mutex; struct handle_table *ctx_table; + pipe_mutex fb_mutex; struct stw_framebuffer *fb_head; #ifdef DEBUG diff --git a/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c index 2660c591f9f..62c859e1f92 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c +++ b/src/gallium/state_trackers/wgl/shared/stw_extensionsstring.c @@ -38,6 +38,7 @@ static const char *stw_extension_string = "WGL_ARB_extensions_string " "WGL_ARB_multisample " "WGL_ARB_pixel_format " +/* "WGL_EXT_swap_interval " */ "WGL_EXT_extensions_string"; diff --git a/src/gallium/state_trackers/wgl/shared/stw_extgallium.c b/src/gallium/state_trackers/wgl/shared/stw_extgallium.c new file mode 100644 index 00000000000..fc22737d7e3 --- /dev/null +++ b/src/gallium/state_trackers/wgl/shared/stw_extgallium.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_screen.h" +#include "stw_public.h" +#include "stw_device.h" +#include "stw_winsys.h" + +#ifdef DEBUG +#include "trace/tr_screen.h" +#include "trace/tr_context.h" +#endif + + +struct pipe_screen * APIENTRY +wglGetGalliumScreenMESA(void) +{ + return stw_dev ? stw_dev->screen : NULL; +} + + +/* XXX: Unify with stw_create_layer_context */ +struct pipe_context * APIENTRY +wglCreateGalliumContextMESA(void) +{ + struct pipe_screen *screen = NULL; + struct pipe_context *pipe = NULL; + + if(!stw_dev) + return NULL; + + screen = stw_dev->screen; + +#ifdef DEBUG + /* Unwrap screen */ + if(stw_dev->trace_running) + screen = trace_screen(screen)->screen; +#endif + + pipe = stw_dev->stw_winsys->create_context( screen ); + if (pipe == NULL) + goto no_pipe; + +#ifdef DEBUG + /* Wrap context */ + if(stw_dev->trace_running) + pipe = trace_context_create(stw_dev->screen, pipe); +#endif + + return pipe; + +no_pipe: + return NULL; +} diff --git a/src/gallium/state_trackers/wgl/shared/stw_extgallium.h b/src/gallium/state_trackers/wgl/shared/stw_extgallium.h new file mode 100644 index 00000000000..cc35f2bb7fe --- /dev/null +++ b/src/gallium/state_trackers/wgl/shared/stw_extgallium.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STW_EXTGALLIUM_H_ +#define STW_EXTGALLIUM_H_ + + +#include <windows.h> + + +struct pipe_screen; +struct pipe_context; + + +struct pipe_screen * APIENTRY +wglGetGalliumScreenMESA(void); + + +struct pipe_context * APIENTRY +wglCreateGalliumContextMESA(void); + + +#endif /* STW_EXTGALLIUM_H_ */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_extswapinterval.c b/src/gallium/state_trackers/wgl/shared/stw_extswapinterval.c new file mode 100644 index 00000000000..9eac6a1d09d --- /dev/null +++ b/src/gallium/state_trackers/wgl/shared/stw_extswapinterval.c @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2009 VMware, Inc. + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include <windows.h> + +#define WGL_WGLEXT_PROTOTYPES + +#include <GL/gl.h> +#include <GL/wglext.h> +#include "util/u_debug.h" + +/* A dummy implementation of this extension. + * + * Required as some applications retrieve and call these functions + * regardless of the fact that we don't advertise the extension and + * further more the results of wglGetProcAddress are NULL. + */ +WINGDIAPI BOOL APIENTRY +wglSwapIntervalEXT(int interval) +{ + (void) interval; + debug_printf("%s: %d\n", __FUNCTION__, interval); + return TRUE; +} + +WINGDIAPI int APIENTRY +wglGetSwapIntervalEXT(void) +{ + return 0; +} + + diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c index a601fc5646b..b8956bb5509 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c +++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c @@ -45,20 +45,31 @@ #include "stw_tls.h" -struct stw_framebuffer * +/** + * Search the framebuffer with the matching HWND while holding the + * stw_dev::fb_mutex global lock. + */ +static INLINE struct stw_framebuffer * stw_framebuffer_from_hwnd_locked( HWND hwnd ) { struct stw_framebuffer *fb; for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next) - if (fb->hWnd == hwnd) + if (fb->hWnd == hwnd) { + pipe_mutex_lock(fb->mutex); break; + } return fb; } +/** + * Destroy this framebuffer. Both stw_dev::fb_mutex and stw_framebuffer::mutex + * must be held, by this order. Obviously no further access to fb can be done + * after this. + */ static INLINE void stw_framebuffer_destroy_locked( struct stw_framebuffer *fb ) @@ -74,17 +85,53 @@ stw_framebuffer_destroy_locked( st_unreference_framebuffer(fb->stfb); + pipe_mutex_unlock( fb->mutex ); + pipe_mutex_destroy( fb->mutex ); FREE( fb ); } +void +stw_framebuffer_release( + struct stw_framebuffer *fb) +{ + assert(fb); + pipe_mutex_unlock( fb->mutex ); +} + + +static INLINE void +stw_framebuffer_get_size( struct stw_framebuffer *fb ) +{ + unsigned width, height; + RECT rect; + + assert(fb->hWnd); + + GetClientRect( fb->hWnd, &rect ); + width = rect.right - rect.left; + height = rect.bottom - rect.top; + + if(width < 1) + width = 1; + if(height < 1) + height = 1; + + if(width != fb->width || height != fb->height) { + fb->must_resize = TRUE; + fb->width = width; + fb->height = height; + } +} + + /** * @sa http://msdn.microsoft.com/en-us/library/ms644975(VS.85).aspx * @sa http://msdn.microsoft.com/en-us/library/ms644960(VS.85).aspx */ -static LRESULT CALLBACK +LRESULT CALLBACK stw_call_window_proc( int nCode, WPARAM wParam, @@ -92,6 +139,7 @@ stw_call_window_proc( { struct stw_tls_data *tls_data; PCWPSTRUCT pParams = (PCWPSTRUCT)lParam; + struct stw_framebuffer *fb; tls_data = stw_tls_get_data(); if(!tls_data) @@ -100,51 +148,37 @@ stw_call_window_proc( if (nCode < 0) return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam); - if (pParams->message == WM_SIZE && pParams->wParam != SIZE_MINIMIZED) { - struct stw_framebuffer *fb; - - pipe_mutex_lock( stw_dev->mutex ); - fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd ); - pipe_mutex_unlock( stw_dev->mutex ); - - if(fb) { - unsigned width = LOWORD( pParams->lParam ); - unsigned height = HIWORD( pParams->lParam ); - - /* FIXME: The mesa statetracker makes the assumptions that only - * one context is using the framebuffer, and that that context is the - * current one. However neither holds true, as WGL allows more than - * one context to be bound to the same drawable, and this function can - * be called from any thread. - */ - pipe_mutex_lock( fb->mutex ); - if (fb->stfb) - st_resize_framebuffer( fb->stfb, width, height ); - pipe_mutex_unlock( fb->mutex ); + if (pParams->message == WM_WINDOWPOSCHANGED) { + /* We handle WM_WINDOWPOSCHANGED instead of WM_SIZE because according to + * http://blogs.msdn.com/oldnewthing/archive/2008/01/15/7113860.aspx + * WM_SIZE is generated from WM_WINDOWPOSCHANGED by DefWindowProc so it + * can be masked out by the application. */ + LPWINDOWPOS lpWindowPos = (LPWINDOWPOS)pParams->lParam; + if((lpWindowPos->flags & SWP_SHOWWINDOW) || + !(lpWindowPos->flags & SWP_NOSIZE)) { + fb = stw_framebuffer_from_hwnd( pParams->hwnd ); + if(fb) { + /* Size in WINDOWPOS includes the window frame, so get the size + * of the client area via GetClientRect. */ + stw_framebuffer_get_size(fb); + stw_framebuffer_release(fb); + } } } - - if (pParams->message == WM_DESTROY) { - struct stw_framebuffer *fb; - - pipe_mutex_lock( stw_dev->mutex ); - + else if (pParams->message == WM_DESTROY) { + pipe_mutex_lock( stw_dev->fb_mutex ); fb = stw_framebuffer_from_hwnd_locked( pParams->hwnd ); if(fb) stw_framebuffer_destroy_locked(fb); - - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->fb_mutex ); } return CallNextHookEx(tls_data->hCallWndProcHook, nCode, wParam, lParam); } -/** - * Create a new framebuffer object which will correspond to the given HDC. - */ struct stw_framebuffer * -stw_framebuffer_create_locked( +stw_framebuffer_create( HDC hdc, int iPixelFormat ) { @@ -169,51 +203,34 @@ stw_framebuffer_create_locked( stw_pixelformat_visual(&fb->visual, pfi); + stw_framebuffer_get_size(fb); + pipe_mutex_init( fb->mutex ); + /* This is the only case where we lock the stw_framebuffer::mutex before + * stw_dev::fb_mutex, since no other thread can know about this framebuffer + * and we must prevent any other thread from destroying it before we return. + */ + pipe_mutex_lock( fb->mutex ); + + pipe_mutex_lock( stw_dev->fb_mutex ); fb->next = stw_dev->fb_head; stw_dev->fb_head = fb; + pipe_mutex_unlock( stw_dev->fb_mutex ); return fb; } -static void -stw_framebuffer_get_size( struct stw_framebuffer *fb, GLuint *pwidth, GLuint *pheight ) -{ - GLuint width, height; - - if (fb->hWnd) { - RECT rect; - GetClientRect( fb->hWnd, &rect ); - width = rect.right - rect.left; - height = rect.bottom - rect.top; - } - else { - width = GetDeviceCaps( fb->hDC, HORZRES ); - height = GetDeviceCaps( fb->hDC, VERTRES ); - } - - if(width < 1) - width = 1; - if(height < 1) - height = 1; - - *pwidth = width; - *pheight = height; -} - - BOOL stw_framebuffer_allocate( struct stw_framebuffer *fb) { - pipe_mutex_lock( fb->mutex ); - + assert(fb); + if(!fb->stfb) { const struct stw_pixelformat_info *pfi = fb->pfi; enum pipe_format colorFormat, depthFormat, stencilFormat; - GLuint width, height; colorFormat = pfi->color_format; @@ -229,32 +246,50 @@ stw_framebuffer_allocate( else stencilFormat = PIPE_FORMAT_NONE; - stw_framebuffer_get_size(fb, &width, &height); - + assert(fb->must_resize); + assert(fb->width); + assert(fb->height); + fb->stfb = st_create_framebuffer( &fb->visual, colorFormat, depthFormat, stencilFormat, - width, - height, + fb->width, + fb->height, (void *) fb ); + + // to notify the context + fb->must_resize = TRUE; } - pipe_mutex_unlock( fb->mutex ); - return fb->stfb ? TRUE : FALSE; } +/** + * Update the framebuffer's size if necessary. + */ void -stw_framebuffer_resize( +stw_framebuffer_update( struct stw_framebuffer *fb) { - GLuint width, height; assert(fb->stfb); - stw_framebuffer_get_size(fb, &width, &height); - st_resize_framebuffer(fb->stfb, width, height); + assert(fb->height); + assert(fb->width); + + /* XXX: It would be nice to avoid checking the size again -- in theory + * stw_call_window_proc would have cought the resize and stored the right + * size already, but unfortunately threads created before the DllMain is + * called don't get a DLL_THREAD_ATTACH notification, and there is no way + * to know of their existing without using the not very portable PSAPI. + */ + stw_framebuffer_get_size(fb); + + if(fb->must_resize) { + st_resize_framebuffer(fb->stfb, fb->width, fb->height); + fb->must_resize = FALSE; + } } @@ -264,32 +299,47 @@ stw_framebuffer_cleanup( void ) struct stw_framebuffer *fb; struct stw_framebuffer *next; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->fb_mutex ); fb = stw_dev->fb_head; while (fb) { next = fb->next; + + pipe_mutex_lock(fb->mutex); stw_framebuffer_destroy_locked(fb); + fb = next; } stw_dev->fb_head = NULL; - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->fb_mutex ); } /** * Given an hdc, return the corresponding stw_framebuffer. */ -struct stw_framebuffer * +static INLINE struct stw_framebuffer * stw_framebuffer_from_hdc_locked( HDC hdc ) { + HWND hwnd; struct stw_framebuffer *fb; + /* + * Some applications create and use several HDCs for the same window, so + * looking up the framebuffer by the HDC is not reliable. Use HWND whenever + * possible. + */ + hwnd = WindowFromDC(hdc); + if(hwnd) + return stw_framebuffer_from_hwnd_locked(hwnd); + for (fb = stw_dev->fb_head; fb != NULL; fb = fb->next) - if (fb->hDC == hdc) + if (fb->hDC == hdc) { + pipe_mutex_lock(fb->mutex); break; + } return fb; } @@ -304,9 +354,26 @@ stw_framebuffer_from_hdc( { struct stw_framebuffer *fb; - pipe_mutex_lock( stw_dev->mutex ); + pipe_mutex_lock( stw_dev->fb_mutex ); fb = stw_framebuffer_from_hdc_locked(hdc); - pipe_mutex_unlock( stw_dev->mutex ); + pipe_mutex_unlock( stw_dev->fb_mutex ); + + return fb; +} + + +/** + * Given an hdc, return the corresponding stw_framebuffer. + */ +struct stw_framebuffer * +stw_framebuffer_from_hwnd( + HWND hwnd ) +{ + struct stw_framebuffer *fb; + + pipe_mutex_lock( stw_dev->fb_mutex ); + fb = stw_framebuffer_from_hwnd_locked(hwnd); + pipe_mutex_unlock( stw_dev->fb_mutex ); return fb; } @@ -326,22 +393,19 @@ stw_pixelformat_set( if (index >= count) return FALSE; - pipe_mutex_lock( stw_dev->mutex ); - fb = stw_framebuffer_from_hdc_locked(hdc); if(fb) { /* SetPixelFormat must be called only once */ - pipe_mutex_unlock( stw_dev->mutex ); + stw_framebuffer_release( fb ); return FALSE; } - fb = stw_framebuffer_create_locked(hdc, iPixelFormat); + fb = stw_framebuffer_create(hdc, iPixelFormat); if(!fb) { - pipe_mutex_unlock( stw_dev->mutex ); return FALSE; } - pipe_mutex_unlock( stw_dev->mutex ); + stw_framebuffer_release( fb ); /* Some applications mistakenly use the undocumented wglSetPixelFormat * function instead of SetPixelFormat, so we call SetPixelFormat here to @@ -358,13 +422,16 @@ int stw_pixelformat_get( HDC hdc ) { + int iPixelFormat = 0; struct stw_framebuffer *fb; fb = stw_framebuffer_from_hdc(hdc); - if(!fb) - return 0; + if(fb) { + iPixelFormat = fb->iPixelFormat; + stw_framebuffer_release(fb); + } - return fb->iPixelFormat; + return iPixelFormat; } @@ -380,10 +447,10 @@ stw_swap_buffers( if (fb == NULL) return FALSE; - if (!(fb->pfi->pfd.dwFlags & PFD_DOUBLEBUFFER)) + if (!(fb->pfi->pfd.dwFlags & PFD_DOUBLEBUFFER)) { + stw_framebuffer_release(fb); return TRUE; - - pipe_mutex_lock( fb->mutex ); + } /* If we're swapping the buffer associated with the current context * we have to flush any pending rendering commands first. @@ -394,7 +461,7 @@ stw_swap_buffers( if(!st_get_framebuffer_surface( fb->stfb, ST_SURFACE_BACK_LEFT, &surface )) { /* FIXME: this shouldn't happen, but does on glean */ - pipe_mutex_unlock( fb->mutex ); + stw_framebuffer_release(fb); return FALSE; } @@ -407,7 +474,8 @@ stw_swap_buffers( stw_dev->stw_winsys->flush_frontbuffer( screen, surface, hdc ); - pipe_mutex_unlock( fb->mutex ); + stw_framebuffer_update(fb); + stw_framebuffer_release(fb); return TRUE; } @@ -423,38 +491,3 @@ stw_swap_layer_buffers( return FALSE; } - - -boolean -stw_framebuffer_init_thread(void) -{ - struct stw_tls_data *tls_data; - - tls_data = stw_tls_get_data(); - if(!tls_data) - return FALSE; - - tls_data->hCallWndProcHook = SetWindowsHookEx(WH_CALLWNDPROC, - stw_call_window_proc, - NULL, - GetCurrentThreadId()); - if(tls_data->hCallWndProcHook == NULL) - return FALSE; - - return TRUE; -} - -void -stw_framebuffer_cleanup_thread(void) -{ - struct stw_tls_data *tls_data; - - tls_data = stw_tls_get_data(); - if(!tls_data) - return; - - if(tls_data->hCallWndProcHook) { - UnhookWindowsHookEx(tls_data->hCallWndProcHook); - tls_data->hCallWndProcHook = NULL; - } -} diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h index e7fa51c3a81..13d29f37e48 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h +++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h @@ -41,6 +41,23 @@ struct stw_pixelformat_info; */ struct stw_framebuffer { + /** + * This mutex has two purposes: + * - protect the access to the mutable data members below + * - prevent the the framebuffer from being deleted while being accessed. + * + * It is OK to lock this mutex while holding the stw_device::fb_mutex lock, + * but the opposite must never happen. + */ + pipe_mutex mutex; + + /* + * Immutable members. + * + * Note that even access to immutable members implies acquiring the mutex + * above, to prevent the framebuffer from being destroyed. + */ + HDC hDC; HWND hWnd; @@ -48,41 +65,84 @@ struct stw_framebuffer const struct stw_pixelformat_info *pfi; GLvisual visual; - pipe_mutex mutex; + /* + * Mutable members. + */ + struct st_framebuffer *stfb; - /** This is protected by stw_device::mutex, not the mutex above */ + /* FIXME: Make this work for multiple contexts bound to the same framebuffer */ + boolean must_resize; + unsigned width; + unsigned height; + + /** + * This is protected by stw_device::fb_mutex, not the mutex above. + * + * Deletions must be done by first acquiring stw_device::fb_mutex, and then + * acquiring the stw_framebuffer::mutex of the framebuffer to be deleted. + * This ensures that nobody else is reading/writing to the. + * + * It is not necessary to aquire the mutex above to navigate the linked list + * given that deletions are done with stw_device::fb_mutex held, so no other + * thread can delete. + */ struct stw_framebuffer *next; }; + +/** + * Create a new framebuffer object which will correspond to the given HDC. + * + * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release + * must be called when done + */ struct stw_framebuffer * -stw_framebuffer_create_locked( +stw_framebuffer_create( HDC hdc, int iPixelFormat ); +/** + * Search a framebuffer with a matching HWND. + * + * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release + * must be called when done + */ +struct stw_framebuffer * +stw_framebuffer_from_hwnd( + HWND hwnd ); + +/** + * Search a framebuffer with a matching HDC. + * + * This function will acquire stw_framebuffer::mutex. stw_framebuffer_release + * must be called when done + */ +struct stw_framebuffer * +stw_framebuffer_from_hdc( + HDC hdc ); + BOOL stw_framebuffer_allocate( struct stw_framebuffer *fb ); void -stw_framebuffer_resize( +stw_framebuffer_update( struct stw_framebuffer *fb); +/** + * Release stw_framebuffer::mutex lock. This framebuffer must not be accessed + * after calling this function, as it may have been deleted by another thread + * in the meanwhile. + */ void -stw_framebuffer_cleanup(void); - -struct stw_framebuffer * -stw_framebuffer_from_hdc_locked( - HDC hdc ); - -struct stw_framebuffer * -stw_framebuffer_from_hdc( - HDC hdc ); - -boolean -stw_framebuffer_init_thread(void); +stw_framebuffer_release( + struct stw_framebuffer *fb); +/** + * Cleanup any existing framebuffers when exiting application. + */ void -stw_framebuffer_cleanup_thread(void); +stw_framebuffer_cleanup(void); #endif /* STW_FRAMEBUFFER_H */ diff --git a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c index 4070cbd5c00..879ced925a5 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c +++ b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c @@ -34,6 +34,7 @@ #include "glapi/glapi.h" #include "stw_public.h" +#include "stw_extgallium.h" struct stw_extension_entry { @@ -56,6 +57,14 @@ static const struct stw_extension_entry stw_extension_entries[] = { /* WGL_EXT_extensions_string */ STW_EXTENSION_ENTRY( wglGetExtensionsStringEXT ), + /* WGL_EXT_swap_interval */ + STW_EXTENSION_ENTRY( wglGetSwapIntervalEXT ), + STW_EXTENSION_ENTRY( wglSwapIntervalEXT ), + + /* WGL_EXT_gallium ? */ + STW_EXTENSION_ENTRY( wglGetGalliumScreenMESA ), + STW_EXTENSION_ENTRY( wglCreateGalliumContextMESA ), + { NULL, NULL } }; @@ -65,13 +74,13 @@ stw_get_proc_address( { const struct stw_extension_entry *entry; - PROC p = (PROC) _glapi_get_proc_address( lpszProc ); - if (p) - return p; + if (lpszProc[0] == 'w' && lpszProc[1] == 'g' && lpszProc[2] == 'l') + for (entry = stw_extension_entries; entry->name; entry++) + if (strcmp( lpszProc, entry->name ) == 0) + return entry->proc; - for (entry = stw_extension_entries; entry->name; entry++) - if (strcmp( lpszProc, entry->name ) == 0) - return entry->proc; + if (lpszProc[0] == 'g' && lpszProc[1] == 'l') + return (PROC) _glapi_get_proc_address( lpszProc ); return NULL; } diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.c b/src/gallium/state_trackers/wgl/shared/stw_tls.c index 0c18a52352c..4bd6a9289c9 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_tls.c +++ b/src/gallium/state_trackers/wgl/shared/stw_tls.c @@ -51,9 +51,23 @@ stw_tls_data_create() data = CALLOC_STRUCT(stw_tls_data); if (!data) - return NULL; + goto no_data; + + data->hCallWndProcHook = SetWindowsHookEx(WH_CALLWNDPROC, + stw_call_window_proc, + NULL, + GetCurrentThreadId()); + if(data->hCallWndProcHook == NULL) + goto no_hook; + + TlsSetValue(tlsIndex, data); return data; + +no_hook: + FREE(data); +no_data: + return NULL; } boolean @@ -69,8 +83,6 @@ stw_tls_init_thread(void) if(!data) return FALSE; - TlsSetValue(tlsIndex, data); - return TRUE; } @@ -84,8 +96,16 @@ stw_tls_cleanup_thread(void) } data = (struct stw_tls_data *) TlsGetValue(tlsIndex); - TlsSetValue(tlsIndex, NULL); - FREE(data); + if(data) { + TlsSetValue(tlsIndex, NULL); + + if(data->hCallWndProcHook) { + UnhookWindowsHookEx(data->hCallWndProcHook); + data->hCallWndProcHook = NULL; + } + + FREE(data); + } } void @@ -110,12 +130,9 @@ stw_tls_get_data(void) if(!data) { /* DllMain is called with DLL_THREAD_ATTACH only by threads created after * the DLL is loaded by the process */ - data = stw_tls_data_create(); if(!data) return NULL; - - TlsSetValue(tlsIndex, data); } return data; diff --git a/src/gallium/state_trackers/wgl/shared/stw_tls.h b/src/gallium/state_trackers/wgl/shared/stw_tls.h index 6af8be70c9f..fbf8b1cbee4 100644 --- a/src/gallium/state_trackers/wgl/shared/stw_tls.h +++ b/src/gallium/state_trackers/wgl/shared/stw_tls.h @@ -50,4 +50,10 @@ stw_tls_cleanup(void); struct stw_tls_data * stw_tls_get_data(void); +LRESULT CALLBACK +stw_call_window_proc( + int nCode, + WPARAM wParam, + LPARAM lParam ); + #endif /* STW_TLS_H */ diff --git a/src/gallium/state_trackers/xorg/Makefile b/src/gallium/state_trackers/xorg/Makefile index a00ea3e2a4e..27a1990724d 100644 --- a/src/gallium/state_trackers/xorg/Makefile +++ b/src/gallium/state_trackers/xorg/Makefile @@ -1,29 +1,18 @@ -TARGET = libxorgtracker.a -CFILES = $(wildcard ./*.c) -OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) -GALLIUMDIR = ../.. -TOP = ../../../.. - +TOP = ../../../.. include $(TOP)/configs/current -CFLAGS = -DHAVE_CONFIG_H \ - -g -Wall -Wimplicit-function-declaration -fPIC \ - $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \ - -I$(GALLIUMDIR)/include \ - -I$(GALLIUMDIR)/auxiliary \ - -I$(TOP)/src/mesa/drivers/dri/common \ - -I$(TOP)/src/mesa \ - -I$(TOP)/include \ - -I$(TOP)/src/egl/main - -############################################# - -.PHONY = all clean +LIBNAME = xorgtracker -all: $(TARGET) +LIBRARY_INCLUDES = \ + -DHAVE_CONFIG_H \ + $(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/mesa/main -$(TARGET): $(OBJECTS) - ar rcs $(TARGET) $(OBJECTS) +C_SOURCES = $(wildcard ./*.c) -clean: - rm -rf $(OBJECTS) $(TARGET) +include ../../Makefile.template diff --git a/src/gallium/state_trackers/xorg/SConscript b/src/gallium/state_trackers/xorg/SConscript new file mode 100644 index 00000000000..15ebcf86733 --- /dev/null +++ b/src/gallium/state_trackers/xorg/SConscript @@ -0,0 +1,26 @@ +####################################################################### +# SConscript for xorg state_tracker + +Import('*') + +if 'xorg' in env['statetrackers']: + + env = env.Clone() + + env.Append(CPPPATH = [ + '#/src/mesa', + ]) + + env.ParseConfig('pkg-config --cflags --libs xorg-server') + + st_xorg = env.ConvenienceLibrary( + target = 'st_xorg', + source = [ 'xorg_composite.c', + 'xorg_crtc.c', + 'xorg_dri2.c', + 'xorg_driver.c', + 'xorg_exa.c', + 'xorg_output.c', + ] + ) + Export('st_xorg') diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c new file mode 100644 index 00000000000..bf9c82a7074 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_composite.c @@ -0,0 +1,115 @@ +#include "xorg_composite.h" + +struct xorg_composite_blend { + int op:8; + + unsigned rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ + + unsigned alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ +}; + +static const struct xorg_composite_blend xorg_blends[] = { + { PictOpClear, + PIPE_BLENDFACTOR_CONST_COLOR, PIPE_BLENDFACTOR_CONST_ALPHA, + PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO }, + + { PictOpSrc, + PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ONE, + PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO }, + + { PictOpDst, + PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO, + PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ONE }, + + { PictOpOver, + PIPE_BLENDFACTOR_SRC_ALPHA, PIPE_BLENDFACTOR_ONE, + PIPE_BLENDFACTOR_INV_SRC_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA }, + + { PictOpOverReverse, + PIPE_BLENDFACTOR_SRC_ALPHA, PIPE_BLENDFACTOR_ONE, + PIPE_BLENDFACTOR_INV_SRC_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA }, +}; + +struct acceleration_info { + int op : 16; + int with_mask : 1; + int component_alpha : 1; +}; +static const struct acceleration_info accelerated_ops[] = { + {PictOpClear, 1, 0}, + {PictOpSrc, 1, 0}, + {PictOpDst, 1, 0}, + {PictOpOver, 1, 0}, + {PictOpOverReverse, 1, 0}, + {PictOpIn, 1, 0}, + {PictOpInReverse, 1, 0}, + {PictOpOut, 1, 0}, + {PictOpOutReverse, 1, 0}, + {PictOpAtop, 1, 0}, + {PictOpAtopReverse, 1, 0}, + {PictOpXor, 1, 0}, + {PictOpAdd, 1, 0}, + {PictOpSaturate, 1, 0}, +}; + +static void +draw_texture(struct exa_context *exa) +{ +#if 0 + if (buf) { + util_draw_vertex_buffer(pipe, buf, 0, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe_buffer_reference(&buf, NULL); + } +#endif +} + +boolean xorg_composite_accelerated(int op, + PicturePtr pSrcPicture, + PicturePtr pMaskPicture, + PicturePtr pDstPicture) +{ + unsigned i; + unsigned accel_ops_count = + sizeof(accelerated_ops)/sizeof(struct acceleration_info); + + if (pSrcPicture) { + /* component alpha not supported */ + if (pSrcPicture->componentAlpha) + return FALSE; + /* fills not supported */ + if (pSrcPicture->pSourcePict) + return FALSE; + } + + for (i = 0; i < accel_ops_count; ++i) { + if (op == accelerated_ops[i].op) { + if (pMaskPicture && !accelerated_ops[i].with_mask) + return FALSE; + return TRUE; + } + } + return FALSE; +} + +boolean xorg_composite_bind_state(struct exa_context *exa, + int op, + PicturePtr pSrcPicture, + PicturePtr pMaskPicture, + PicturePtr pDstPicture) +{ + return FALSE; +} + +void xorg_composite(struct exa_context *exa, + struct exa_pixmap_priv *dst, + int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height) +{ +} + diff --git a/src/gallium/state_trackers/xorg/xorg_composite.h b/src/gallium/state_trackers/xorg/xorg_composite.h new file mode 100644 index 00000000000..a52e0e6dff0 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_composite.h @@ -0,0 +1,22 @@ +#ifndef XORG_COMPOSITE_H +#define XORG_COMPOSITE_H + +#include "xorg_exa.h" + +boolean xorg_composite_accelerated(int op, + PicturePtr pSrcPicture, + PicturePtr pMaskPicture, + PicturePtr pDstPicture); + +boolean xorg_composite_bind_state(struct exa_context *exa, + int op, + PicturePtr pSrcPicture, + PicturePtr pMaskPicture, + PicturePtr pDstPicture); + +void xorg_composite(struct exa_context *exa, + struct exa_pixmap_priv *dst, + int srcX, int srcY, int maskX, int maskY, + int dstX, int dstY, int width, int height); + +#endif diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c index 7304113a659..2235961b031 100644 --- a/src/gallium/state_trackers/xorg/xorg_crtc.c +++ b/src/gallium/state_trackers/xorg/xorg_crtc.c @@ -171,7 +171,6 @@ crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data) static void crtc_destroy(xf86CrtcPtr crtc) { - modesettingPtr ms = modesettingPTR(crtc->scrn); struct crtc_private *crtcp = crtc->driver_private; if (crtcp->cursor_buf) @@ -194,9 +193,10 @@ crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image) PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_GPU_READ, 64*64*4); - drm_api_hooks.handle_from_buffer(ms->screen, - crtcp->cursor_buf, - &crtcp->cursor_handle); + ms->api->handle_from_buffer(ms->api, + ms->screen, + crtcp->cursor_buf, + &crtcp->cursor_handle); } ptr = pipe_buffer_map(ms->screen, crtcp->cursor_buf, PIPE_BUFFER_USAGE_CPU_WRITE); diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c index 401bd39dac0..7d6954016fe 100644 --- a/src/gallium/state_trackers/xorg/xorg_dri2.c +++ b/src/gallium/state_trackers/xorg/xorg_dri2.c @@ -39,10 +39,13 @@ #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_rect.h" + typedef struct { PixmapPtr pPixmap; struct pipe_texture *tex; struct pipe_buffer *buf; + struct pipe_fence_handle *fence; } *BufferPrivatePtr; static DRI2BufferPtr @@ -57,6 +60,7 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) DRI2BufferPtr buffers; PixmapPtr pPixmap; unsigned stride, handle; + boolean have_depth = FALSE, have_stencil = FALSE; int i; buffers = xcalloc(count, sizeof *buffers); @@ -67,6 +71,16 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) if (!privates) goto fail_privates; + for (i = 0; i < count; i++) { + if (attachments[i] == DRI2BufferDepth) + have_depth = TRUE; + else if (attachments[i] == DRI2BufferStencil) + have_stencil = TRUE; + } + + if (have_stencil && !have_depth) + FatalError("Doesn't support only stencil yet\n"); + depth = NULL; for (i = 0; i < count; i++) { pPixmap = NULL; @@ -83,33 +97,40 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count) pipe_texture_reference(&tex, depth); } else if (attachments[i] == DRI2BufferDepth) { struct pipe_texture template; - memset(&template, 0, sizeof(template)); template.target = PIPE_TEXTURE_2D; - template.format = PIPE_FORMAT_S8Z24_UNORM; + if (have_stencil) + template.format = ms->ds_depth_bits_last ? + PIPE_FORMAT_S8Z24_UNORM : PIPE_FORMAT_Z24S8_UNORM; + else + template.format = ms->d_depth_bits_last ? + PIPE_FORMAT_X8Z24_UNORM : PIPE_FORMAT_Z24X8_UNORM; pf_get_block(template.format, &template.block); template.width[0] = pDraw->width; template.height[0] = pDraw->height; template.depth[0] = 1; template.last_level = 0; - template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; tex = ms->screen->texture_create(ms->screen, &template); + depth = tex; + } else if (attachments[i] == DRI2BufferFakeFrontLeft && + pDraw->type == DRAWABLE_PIXMAP) { + pPixmap = (PixmapPtr) pDraw; + pPixmap->refcnt++; + tex = xorg_exa_get_texture(pPixmap); } else { - struct pipe_texture template; - memset(&template, 0, sizeof(template)); - template.target = PIPE_TEXTURE_2D; - template.format = PIPE_FORMAT_A8R8G8B8_UNORM; - pf_get_block(template.format, &template.block); - template.width[0] = pDraw->width; - template.height[0] = pDraw->height; - template.depth[0] = 1; - template.last_level = 0; - template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; - tex = ms->screen->texture_create(ms->screen, &template); + pPixmap = (*pScreen->CreatePixmap)(pScreen, pDraw->width, + pDraw->height, + pDraw->depth, + 0); + tex = xorg_exa_get_texture(pPixmap); } - drm_api_hooks.buffer_from_texture(tex, &buf, &stride); - drm_api_hooks.global_handle_from_buffer(ms->screen, buf, &handle); + if (!tex) + FatalError("NO TEXTURE IN DRI2\n"); + + ms->api->buffer_from_texture(ms->api, tex, &buf, &stride); + ms->api->global_handle_from_buffer(ms->api, ms->screen, buf, &handle); buffers[i].name = handle; buffers[i].attachment = attachments[i]; @@ -138,15 +159,17 @@ driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count) modesettingPtr ms = modesettingPTR(pScrn); BufferPrivatePtr private; int i; + (void)ms; for (i = 0; i < count; i++) { private = buffers[i].driverPrivate; - if (private->pPixmap) - (*pScreen->DestroyPixmap)(private->pPixmap); - pipe_texture_reference(&private->tex, NULL); pipe_buffer_reference(&private->buf, NULL); + ms->screen->fence_reference(ms->screen, &private->fence, NULL); + + if (private->pPixmap) + (*pScreen->DestroyPixmap)(private->pPixmap); } if (buffers) { @@ -164,19 +187,83 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion, modesettingPtr ms = modesettingPTR(pScrn); BufferPrivatePtr dst_priv = pDestBuffer->driverPrivate; BufferPrivatePtr src_priv = pSrcBuffer->driverPrivate; + PixmapPtr src_pixmap; + PixmapPtr dst_pixmap; + GCPtr gc; + RegionPtr copy_clip; - struct pipe_surface *dst_surf = - ms->screen->get_tex_surface(ms->screen, dst_priv->tex, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_WRITE); - struct pipe_surface *src_surf = - ms->screen->get_tex_surface(ms->screen, src_priv->tex, 0, 0, 0, - PIPE_BUFFER_USAGE_GPU_READ); + /* + * In driCreateBuffers we dewrap windows into the + * backing pixmaps in order to get to the texture. + * We need to use the real drawable in CopyArea + * so that cliprects and offsets are correct. + */ + src_pixmap = src_priv->pPixmap; + dst_pixmap = dst_priv->pPixmap; + if (pSrcBuffer->attachment == DRI2BufferFrontLeft) + src_pixmap = (PixmapPtr)pDraw; + if (pDestBuffer->attachment == DRI2BufferFrontLeft) + dst_pixmap = (PixmapPtr)pDraw; - ms->ctx->surface_copy(ms->ctx, dst_surf, 0, 0, src_surf, - 0, 0, pDraw->width, pDraw->height); + /* + * The clients implements glXWaitX with a copy front to fake and then + * waiting on the server to signal its completion of it. While + * glXWaitGL is a client side flush and a copy from fake to front. + * This is how it is done in the DRI2 protocol, how ever depending + * which type of drawables the server does things a bit differently + * then what the protocol says as the fake and front are the same. + * + * for pixmaps glXWaitX is a server flush. + * for pixmaps glXWaitGL is a client flush. + * for windows glXWaitX is a copy from front to fake then a server flush. + * for windows glXWaitGL is a client flush then a copy from fake to front. + * + * XXX in the windows case this code always flushes but that isn't a + * must in the glXWaitGL case but we don't know if this is a glXWaitGL + * or a glFlush/glFinish call. + */ + if (dst_pixmap == src_pixmap) { + /* pixmap glXWaitX */ + if (pSrcBuffer->attachment == DRI2BufferFrontLeft && + pDestBuffer->attachment == DRI2BufferFakeFrontLeft) { + ms->ctx->flush(ms->ctx, PIPE_FLUSH_SWAPBUFFERS, NULL); + return; + } + /* pixmap glXWaitGL */ + if (pDestBuffer->attachment == DRI2BufferFrontLeft && + pSrcBuffer->attachment == DRI2BufferFakeFrontLeft) { + return; + } else { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "copying between the same pixmap\n"); + } + } + + gc = GetScratchGC(pDraw->depth, pScreen); + copy_clip = REGION_CREATE(pScreen, NULL, 0); + REGION_COPY(pScreen, copy_clip, pRegion); + (*gc->funcs->ChangeClip) (gc, CT_REGION, copy_clip, 0); + ValidateGC(&dst_pixmap->drawable, gc); + + /* If this is a full buffer swap, throttle on the previous one */ + if (dst_priv->fence && REGION_NUM_RECTS(pRegion) == 1) { + BoxPtr extents = REGION_EXTENTS(pScreen, pRegion); - pipe_surface_reference(&dst_surf, NULL); - pipe_surface_reference(&src_surf, NULL); + if (extents->x1 == 0 && extents->y1 == 0 && + extents->x2 == pDraw->width && extents->y2 == pDraw->height) { + ms->screen->fence_finish(ms->screen, dst_priv->fence, 0); + ms->screen->fence_reference(ms->screen, &dst_priv->fence, NULL); + } + } + + (*gc->ops->CopyArea)(&src_pixmap->drawable, &dst_pixmap->drawable, gc, + 0, 0, pDraw->width, pDraw->height, 0, 0); + + FreeScratchGC(gc); + + ms->ctx->flush(ms->ctx, PIPE_FLUSH_SWAPBUFFERS, + pDestBuffer->attachment == DRI2BufferFrontLeft ? + &dst_priv->fence : NULL); } Bool @@ -199,6 +286,15 @@ driScreenInit(ScreenPtr pScreen) dri2info.DestroyBuffers = driDestroyBuffers; dri2info.CopyRegion = driCopyRegion; + ms->d_depth_bits_last = + ms->screen->is_format_supported(ms->screen, PIPE_FORMAT_X8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + ms->ds_depth_bits_last = + ms->screen->is_format_supported(ms->screen, PIPE_FORMAT_S8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0); + return DRI2ScreenInit(pScreen, &dri2info); } diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c index 45e831f0c28..923662b24a9 100644 --- a/src/gallium/state_trackers/xorg/xorg_driver.c +++ b/src/gallium/state_trackers/xorg/xorg_driver.c @@ -54,6 +54,7 @@ #include <pciaccess.h> +#include "pipe/p_context.h" #include "xorg_tracker.h" #include "xorg_winsys.h" @@ -179,8 +180,10 @@ CreateFrontBuffer(ScrnInfoPtr pScrn) modesettingPtr ms = modesettingPTR(pScrn); ScreenPtr pScreen = pScrn->pScreen; PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen); + unsigned handle, stride; ms->noEvict = TRUE; + xorg_exa_set_displayed_usage(rootPixmap); pScreen->ModifyPixmapHeader(rootPixmap, pScrn->virtualX, pScrn->virtualY, pScrn->depth, pScrn->bitsPerPixel, @@ -188,13 +191,16 @@ CreateFrontBuffer(ScrnInfoPtr pScrn) NULL); ms->noEvict = FALSE; + handle = xorg_exa_get_pixmap_handle(rootPixmap, &stride); + drmModeAddFB(ms->fd, pScrn->virtualX, pScrn->virtualY, pScrn->depth, pScrn->bitsPerPixel, - pScrn->displayWidth * pScrn->bitsPerPixel / 8, - xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id); + stride, + handle, + &ms->fb_id); pScrn->frameX0 = 0; pScrn->frameY0 = 0; @@ -300,6 +306,7 @@ PreInit(ScrnInfoPtr pScrn, int flags) ms->PciInfo->dev, ms->PciInfo->func ); + ms->api = drm_api_create(); ms->fd = drmOpen(NULL, BusID); if (ms->fd < 0) @@ -418,6 +425,44 @@ RestoreHWState(ScrnInfoPtr pScrn) return TRUE; } +static void xorgBlockHandler(int i, pointer blockData, pointer pTimeout, + pointer pReadmask) +{ + ScreenPtr pScreen = screenInfo.screens[i]; + modesettingPtr ms = modesettingPTR(xf86Screens[pScreen->myNum]); + + pScreen->BlockHandler = ms->blockHandler; + pScreen->BlockHandler(i, blockData, pTimeout, pReadmask); + pScreen->BlockHandler = xorgBlockHandler; + + ms->ctx->flush(ms->ctx, PIPE_FLUSH_RENDER_CACHE, NULL); + +#ifdef DRM_MODE_FEATURE_DIRTYFB + { + RegionPtr dirty = DamageRegion(ms->damage); + unsigned num_cliprects = REGION_NUM_RECTS(dirty); + + if (num_cliprects) { + drmModeClip *clip = alloca(num_cliprects * sizeof(drmModeClip)); + BoxPtr rect = REGION_RECTS(dirty); + int i; + + for (i = 0; i < num_cliprects; i++, rect++) { + clip[i].x = rect->x1; + clip[i].y = rect->y1; + clip[i].width = rect->x2 - rect->x1; + clip[i].height = rect->y2 - rect->y1; + } + + /* TODO query connector property to see if this is needed */ + drmModeDirtyFB(ms->fd, ms->fb_id, clip, num_cliprects); + + DamageEmpty(ms->damage); + } + } +#endif +} + static Bool CreateScreenResources(ScreenPtr pScreen) { @@ -425,6 +470,7 @@ CreateScreenResources(ScreenPtr pScreen) modesettingPtr ms = modesettingPTR(pScrn); PixmapPtr rootPixmap; Bool ret; + unsigned handle, stride; ms->noEvict = TRUE; @@ -434,21 +480,40 @@ CreateScreenResources(ScreenPtr pScreen) rootPixmap = pScreen->GetScreenPixmap(pScreen); + xorg_exa_set_displayed_usage(rootPixmap); if (!pScreen->ModifyPixmapHeader(rootPixmap, -1, -1, -1, -1, -1, NULL)) FatalError("Couldn't adjust screen pixmap\n"); ms->noEvict = FALSE; + handle = xorg_exa_get_pixmap_handle(rootPixmap, &stride); + drmModeAddFB(ms->fd, pScrn->virtualX, pScrn->virtualY, pScrn->depth, pScrn->bitsPerPixel, - pScrn->displayWidth * pScrn->bitsPerPixel / 8, - xorg_exa_get_pixmap_handle(rootPixmap), &ms->fb_id); + stride, + handle, + &ms->fb_id); AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0); +#ifdef DRM_MODE_FEATURE_DIRTYFB + ms->damage = DamageCreate(NULL, NULL, DamageReportNone, TRUE, + pScreen, rootPixmap); + + if (ms->damage) { + DamageRegister(&rootPixmap->drawable, ms->damage); + + xf86DrvMsg(pScrn->scrnIndex, X_INFO, "Damage tracking initialized\n"); + } else { + xf86DrvMsg(pScrn->scrnIndex, X_ERROR, + "Failed to create screen damage record\n"); + return FALSE; + } +#endif + return ret; } @@ -476,7 +541,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv) } if (!ms->screen) { - ms->screen = drm_api_hooks.create_screen(ms->fd, NULL); + ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL); if (!ms->screen) { FatalError("Could not init pipe_screen\n"); @@ -525,6 +590,8 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv) fbPictureInit(pScreen, NULL, 0); + ms->blockHandler = pScreen->BlockHandler; + pScreen->BlockHandler = xorgBlockHandler; ms->createScreenResources = pScreen->CreateScreenResources; pScreen->CreateScreenResources = CreateScreenResources; @@ -622,6 +689,10 @@ LeaveVT(int scrnIndex, int flags) RestoreHWState(pScrn); + if (drmDropMaster(ms->fd)) + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "drmDropMaster failed: %s\n", strerror(errno)); + pScrn->vtSema = FALSE; } @@ -634,6 +705,17 @@ EnterVT(int scrnIndex, int flags) ScrnInfoPtr pScrn = xf86Screens[scrnIndex]; modesettingPtr ms = modesettingPTR(pScrn); + if (drmSetMaster(ms->fd)) { + if (errno == EINVAL) { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "drmSetMaster failed: 2.6.29 or newer kernel required for " + "multi-server DRI\n"); + } else { + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "drmSetMaster failed: %s\n", strerror(errno)); + } + } + /* * Only save state once per server generation since that's what most * drivers do. Could change this to save state at each VT enter. @@ -673,11 +755,22 @@ CloseScreen(int scrnIndex, ScreenPtr pScreen) driCloseScreen(pScreen); #endif + pScreen->BlockHandler = ms->blockHandler; pScreen->CreateScreenResources = ms->createScreenResources; +#ifdef DRM_MODE_FEATURE_DIRTYFB + if (ms->damage) { + DamageUnregister(&pScreen->GetScreenPixmap(pScreen)->drawable, ms->damage); + DamageDestroy(ms->damage); + ms->damage = NULL; + } +#endif + if (ms->exa) xorg_exa_close(pScrn); + ms->api->destroy(ms->api); + ms->api = NULL; drmClose(ms->fd); ms->fd = -1; diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c index 79131743542..3c90c2c9429 100644 --- a/src/gallium/state_trackers/xorg/xorg_exa.c +++ b/src/gallium/state_trackers/xorg/xorg_exa.c @@ -28,34 +28,23 @@ * */ -#include "xorg-server.h" -#include "xf86.h" +#include "xorg_exa.h" #include "xorg_tracker.h" +#include "xorg_composite.h" + +#include <xorg-server.h> +#include <xf86.h> +#include <picturestr.h> +#include <picture.h> #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" -#include "util/u_rect.h" - -struct exa_entity -{ - ExaDriverPtr pExa; - struct pipe_context *ctx; - struct pipe_screen *scrn; -}; - -struct PixmapPriv -{ - int flags; - struct pipe_texture *tex; - unsigned int color; - struct pipe_surface *src_surf; /* for copies */ +#include "cso_cache/cso_context.h" - struct pipe_transfer *map_transfer; - unsigned map_count; -}; +#include "util/u_rect.h" /* * Helper functions @@ -113,8 +102,8 @@ ExaPrepareAccess(PixmapPtr pPix, int index) ScreenPtr pScreen = pPix->drawable.pScreen; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; - struct PixmapPriv *priv; + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv; priv = exaGetPixmapDriverPrivate(pPix); @@ -145,8 +134,8 @@ ExaFinishAccess(PixmapPtr pPix, int index) ScreenPtr pScreen = pPix->drawable.pScreen; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; - struct PixmapPriv *priv; + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv; priv = exaGetPixmapDriverPrivate(pPix); if (!priv) @@ -168,8 +157,8 @@ ExaDone(PixmapPtr pPixmap) { ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); - struct exa_entity *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_context *exa = ms->exa; if (!priv) return; @@ -190,8 +179,8 @@ ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg) { ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); - struct exa_entity *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_context *exa = ms->exa; if (1) return FALSE; @@ -221,8 +210,8 @@ ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1) { ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); struct pipe_surface *surf = exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | @@ -240,9 +229,9 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir, { ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap); - struct PixmapPriv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap); + struct exa_pixmap_priv *src_priv = exaGetPixmapDriverPrivate(pSrcPixmap); if (1) return FALSE; @@ -279,8 +268,8 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY, { ScrnInfoPtr pScrn = xf86Screens[pDstPixmap->drawable.pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pDstPixmap); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDstPixmap); struct pipe_surface *surf = exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | @@ -296,13 +285,25 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture, PicturePtr pDstPicture, PixmapPtr pSrc, PixmapPtr pMask, PixmapPtr pDst) { - return FALSE; + ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_context *exa = ms->exa; + + return xorg_composite_bind_state(exa, op, pSrcPicture, pMaskPicture, + pDstPicture); } static void ExaComposite(PixmapPtr pDst, int srcX, int srcY, int maskX, int maskY, int dstX, int dstY, int width, int height) { + ScrnInfoPtr pScrn = xf86Screens[pDst->drawable.pScreen->myNum]; + modesettingPtr ms = modesettingPTR(pScrn); + struct exa_context *exa = ms->exa; + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pDst); + + xorg_composite(exa, priv, srcX, srcY, maskX, maskY, + dstX, dstY, width, height); } static Bool @@ -310,15 +311,18 @@ ExaCheckComposite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture, PicturePtr pDstPicture) { - return FALSE; + return xorg_composite_accelerated(op, + pSrcPicture, + pMaskPicture, + pDstPicture); } static void * ExaCreatePixmap(ScreenPtr pScreen, int size, int align) { - struct PixmapPriv *priv; + struct exa_pixmap_priv *priv; - priv = xcalloc(1, sizeof(struct PixmapPriv)); + priv = xcalloc(1, sizeof(struct exa_pixmap_priv)); if (!priv) return NULL; @@ -328,7 +332,7 @@ ExaCreatePixmap(ScreenPtr pScreen, int size, int align) static void ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv) { - struct PixmapPriv *priv = (struct PixmapPriv *)dPriv; + struct exa_pixmap_priv *priv = (struct exa_pixmap_priv *)dPriv; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); @@ -344,7 +348,7 @@ ExaDestroyPixmap(ScreenPtr pScreen, void *dPriv) static Bool ExaPixmapIsOffscreen(PixmapPtr pPixmap) { - struct PixmapPriv *priv; + struct exa_pixmap_priv *priv; priv = exaGetPixmapDriverPrivate(pPixmap); @@ -357,13 +361,53 @@ ExaPixmapIsOffscreen(PixmapPtr pPixmap) return FALSE; } +int +xorg_exa_set_displayed_usage(PixmapPtr pPixmap) +{ + struct exa_pixmap_priv *priv; + priv = exaGetPixmapDriverPrivate(pPixmap); + + if (!priv) { + FatalError("NO PIXMAP PRIVATE\n"); + return 0; + } + + if (priv->flags & ~PIPE_TEXTURE_USAGE_PRIMARY) { + FatalError("BAD FLAGS\n"); + return 0; + } + priv->flags = PIPE_TEXTURE_USAGE_PRIMARY; + + return 0; +} + +int +xorg_exa_set_shared_usage(PixmapPtr pPixmap) +{ + struct exa_pixmap_priv *priv; + priv = exaGetPixmapDriverPrivate(pPixmap); + + if (!priv) { + FatalError("NO PIXMAP PRIVATE\n"); + return 0; + } + + if (priv->flags & ~PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + FatalError("BAD FLAGS\n"); + return 0; + } + priv->flags = PIPE_TEXTURE_USAGE_DISPLAY_TARGET; + + return 0; +} + unsigned -xorg_exa_get_pixmap_handle(PixmapPtr pPixmap) +xorg_exa_get_pixmap_handle(PixmapPtr pPixmap, unsigned *stride_out) { ScreenPtr pScreen = pPixmap->drawable.pScreen; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; modesettingPtr ms = modesettingPTR(pScrn); - struct PixmapPriv *priv; + struct exa_pixmap_priv *priv; struct pipe_buffer *buffer = NULL; unsigned handle; unsigned stride; @@ -380,9 +424,12 @@ xorg_exa_get_pixmap_handle(PixmapPtr pPixmap) return 0; } - drm_api_hooks.buffer_from_texture(priv->tex, &buffer, &stride); - drm_api_hooks.handle_from_buffer(ms->screen, buffer, &handle); + ms->api->buffer_from_texture(ms->api, priv->tex, &buffer, &stride); + ms->api->handle_from_buffer(ms->api, ms->screen, buffer, &handle); pipe_buffer_reference(&buffer, NULL); + if (stride_out) + *stride_out = stride; + return handle; } @@ -393,9 +440,9 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, { ScreenPtr pScreen = pPixmap->drawable.pScreen; ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum]; - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; + struct exa_context *exa = ms->exa; if (!priv) return FALSE; @@ -419,11 +466,17 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, bitsPerPixel, devKind, NULL); /* Deal with screen resize */ - if (priv->tex && (priv->tex->width[0] != width || priv->tex->height[0] != height)) { + if (priv->tex && (priv->tex->width[0] != width || + priv->tex->height[0] != height || + priv->tex_flags != priv->flags)) { pipe_texture_reference(&priv->tex, NULL); } - if (!priv->tex) { + if (!priv->tex +#ifdef DRM_MODE_FEATURE_DIRTYFB + && priv->flags +#endif + ) { struct pipe_texture template; memset(&template, 0, sizeof(template)); @@ -434,16 +487,25 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, template.height[0] = height; template.depth[0] = 1; template.last_level = 0; - template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags; + priv->tex_flags = priv->flags; priv->tex = exa->scrn->texture_create(exa->scrn, &template); } +#ifdef DRM_MODE_FEATURE_DIRTYFB + if (!priv->tex) { + if (pPixData) + pPixmap->devPrivate.ptr = pPixData; + else + pPixmap->devPrivate.ptr = xalloc(pPixmap->drawable.height * pPixmap->devKind); + } else +#endif if (pPixData) { struct pipe_transfer *transfer = exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0, PIPE_TRANSFER_WRITE, 0, 0, width, height); - pipe_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), + util_copy_rect(exa->scrn->transfer_map(exa->scrn, transfer), &priv->tex->block, transfer->stride, 0, 0, width, height, pPixData, pPixmap->devKind, 0, 0); exa->scrn->transfer_unmap(exa->scrn, transfer); @@ -456,7 +518,7 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height, struct pipe_texture * xorg_exa_get_texture(PixmapPtr pPixmap) { - struct PixmapPriv *priv = exaGetPixmapDriverPrivate(pPixmap); + struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap); struct pipe_texture *tex = NULL; pipe_texture_reference(&tex, priv->tex); return tex; @@ -466,7 +528,12 @@ void xorg_exa_close(ScrnInfoPtr pScrn) { modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa = ms->exa; + struct exa_context *exa = ms->exa; + + if (exa->cso) { + cso_release_all(exa->cso); + cso_destroy_context(exa->cso); + } if (exa->ctx) exa->ctx->destroy(exa->ctx); @@ -480,10 +547,10 @@ void * xorg_exa_init(ScrnInfoPtr pScrn) { modesettingPtr ms = modesettingPTR(pScrn); - struct exa_entity *exa; + struct exa_context *exa; ExaDriverPtr pExa; - exa = xcalloc(1, sizeof(struct exa_entity)); + exa = xcalloc(1, sizeof(struct exa_context)); if (!exa) return NULL; @@ -493,33 +560,35 @@ xorg_exa_init(ScrnInfoPtr pScrn) } memset(pExa, 0, sizeof(*pExa)); - pExa->exa_major = 2; - pExa->exa_minor = 2; - pExa->memoryBase = 0; - pExa->memorySize = 0; - pExa->offScreenBase = 0; + + pExa->exa_major = 2; + pExa->exa_minor = 2; + pExa->memoryBase = 0; + pExa->memorySize = 0; + pExa->offScreenBase = 0; pExa->pixmapOffsetAlign = 0; - pExa->pixmapPitchAlign = 1; - pExa->flags = EXA_OFFSCREEN_PIXMAPS | EXA_HANDLES_PIXMAPS; - pExa->maxX = 8191; /* FIXME */ - pExa->maxY = 8191; /* FIXME */ - pExa->WaitMarker = ExaWaitMarker; - pExa->MarkSync = ExaMarkSync; - pExa->PrepareSolid = ExaPrepareSolid; - pExa->Solid = ExaSolid; - pExa->DoneSolid = ExaDone; - pExa->PrepareCopy = ExaPrepareCopy; - pExa->Copy = ExaCopy; - pExa->DoneCopy = ExaDone; - pExa->CheckComposite = ExaCheckComposite; - pExa->PrepareComposite = ExaPrepareComposite; - pExa->Composite = ExaComposite; - pExa->DoneComposite = ExaDoneComposite; - pExa->PixmapIsOffscreen = ExaPixmapIsOffscreen; - pExa->PrepareAccess = ExaPrepareAccess; - pExa->FinishAccess = ExaFinishAccess; - pExa->CreatePixmap = ExaCreatePixmap; - pExa->DestroyPixmap = ExaDestroyPixmap; + pExa->pixmapPitchAlign = 1; + pExa->flags = EXA_OFFSCREEN_PIXMAPS | EXA_HANDLES_PIXMAPS; + pExa->maxX = 8191; /* FIXME */ + pExa->maxY = 8191; /* FIXME */ + + pExa->WaitMarker = ExaWaitMarker; + pExa->MarkSync = ExaMarkSync; + pExa->PrepareSolid = ExaPrepareSolid; + pExa->Solid = ExaSolid; + pExa->DoneSolid = ExaDone; + pExa->PrepareCopy = ExaPrepareCopy; + pExa->Copy = ExaCopy; + pExa->DoneCopy = ExaDone; + pExa->CheckComposite = ExaCheckComposite; + pExa->PrepareComposite = ExaPrepareComposite; + pExa->Composite = ExaComposite; + pExa->DoneComposite = ExaDoneComposite; + pExa->PixmapIsOffscreen = ExaPixmapIsOffscreen; + pExa->PrepareAccess = ExaPrepareAccess; + pExa->FinishAccess = ExaFinishAccess; + pExa->CreatePixmap = ExaCreatePixmap; + pExa->DestroyPixmap = ExaDestroyPixmap; pExa->ModifyPixmapHeader = ExaModifyPixmapHeader; if (!exaDriverInit(pScrn->pScreen, pExa)) { @@ -527,10 +596,12 @@ xorg_exa_init(ScrnInfoPtr pScrn) } exa->scrn = ms->screen; - exa->ctx = drm_api_hooks.create_context(exa->scrn); + exa->ctx = ms->api->create_context(ms->api, exa->scrn); /* Share context with DRI */ ms->ctx = exa->ctx; + exa->cso = cso_create_context(exa->ctx); + return (void *)exa; out_err: diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h new file mode 100644 index 00000000000..0189eabaa40 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_exa.h @@ -0,0 +1,32 @@ +#ifndef XORG_EXA_H +#define XORG_EXA_H + +#include "xorg_tracker.h" + +struct cso_context; + +struct exa_context +{ + ExaDriverPtr pExa; + struct pipe_context *ctx; + struct pipe_screen *scrn; + struct cso_context *cso; +}; + + +struct exa_pixmap_priv +{ + int flags; + int tex_flags; + + struct pipe_texture *tex; + unsigned int color; + struct pipe_surface *src_surf; /* for copies */ + + struct pipe_transfer *map_transfer; + unsigned map_count; +}; + + + +#endif diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c new file mode 100644 index 00000000000..04d8977a4cc --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c @@ -0,0 +1,225 @@ +#include "xorg_exa_tgsi.h" + +/*### stupidity defined in X11/extensions/XI.h */ +#undef Absolute + +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_memory.h" +#include "util/u_simple_shaders.h" + +#include "tgsi/tgsi_ureg.h" + +#include "cso_cache/cso_context.h" +#include "cso_cache/cso_hash.h" + +struct xorg_shaders { + struct exa_context *exa; + + struct cso_hash *vs_hash; + struct cso_hash *fs_hash; +}; + +/* SAMP[0] = dst + * SAMP[1] = src + * SAMP[2] = mask + * IN[0] = pos dst + * IN[1] = pos src + * IN[2] = pos mask + * CONST[0] = (0, 0, 0, 1) + */ + +static const char over_op[] = + "SUB TEMP[3], CONST[0].wwww, TEMP[1].wwww\n" + "MUL TEMP[3], TEMP[0], TEMP[3]\n" + "ADD TEMP[0], TEMP[3], TEMP[0]\n"; + + +static INLINE void +create_preamble(struct ureg_program *ureg) +{ +} + + +static INLINE void +src_in_mask(struct ureg_program *ureg, + struct ureg_dst dst, + struct ureg_src src, + struct ureg_src mask) +{ + /* MUL dst, src, mask.wwww */ + ureg_MUL(ureg, dst, src, + ureg_scalar(mask, TGSI_SWIZZLE_W)); +} + +static void * +create_vs(struct pipe_context *pipe, + unsigned vs_traits) +{ + struct ureg_program *ureg; + struct ureg_src src; + struct ureg_dst dst; + + ureg = ureg_create(TGSI_PROCESSOR_VERTEX); + if (ureg == NULL) + return 0; + + if ((vs_traits & VS_COMPOSITE)) { + src = ureg_DECL_vs_input(ureg, + TGSI_SEMANTIC_POSITION, 1); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 1); + ureg_MOV(ureg, dst, src); + } + if ((vs_traits & VS_MASK)) { + src = ureg_DECL_vs_input(ureg, + TGSI_SEMANTIC_POSITION, 2); + dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 2); + ureg_MOV(ureg, dst, src); + } + + ureg_END(ureg); + + return ureg_create_shader_and_destroy(ureg, pipe); +} + +static void * +create_fs(struct pipe_context *pipe, + unsigned fs_traits) +{ + struct ureg_program *ureg; + struct ureg_src dst_sampler, src_sampler, mask_sampler; + struct ureg_src dst_pos, src_pos, mask_pos; + struct ureg_src src, mask; + struct ureg_dst out; + + ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT); + if (ureg == NULL) + return 0; + +#if 0 /* unused right now */ + dst_sampler = ureg_DECL_sampler(ureg); + dst_pos = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_POSITION, + 0, + TGSI_INTERPOLATE_PERSPECTIVE); +#endif + + src_sampler = ureg_DECL_sampler(ureg); + src_pos = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_POSITION, + 1, + TGSI_INTERPOLATE_PERSPECTIVE); + + out = ureg_DECL_output(ureg, + TGSI_SEMANTIC_COLOR, + 0); + + if ((fs_traits & FS_MASK)) { + mask_sampler = ureg_DECL_sampler(ureg); + mask_pos = ureg_DECL_fs_input(ureg, + TGSI_SEMANTIC_POSITION, + 2, + TGSI_INTERPOLATE_PERSPECTIVE); + } + + if ((fs_traits & FS_MASK)) { + ureg_TEX(ureg, ureg_dst(mask), + TGSI_TEXTURE_2D, mask_pos, mask_sampler); + /* src IN mask */ + src_in_mask(ureg, out, src, mask); + } else { + ureg_TEX(ureg, out, + TGSI_TEXTURE_2D, src_pos, src_sampler); + } + + ureg_END(ureg); + + return ureg_create_shader_and_destroy(ureg, pipe); +} + +struct xorg_shaders * xorg_shaders_create(struct exa_context *exa) +{ + struct xorg_shaders *sc = CALLOC_STRUCT(xorg_shaders); + + sc->exa = exa; + sc->vs_hash = cso_hash_create(); + sc->fs_hash = cso_hash_create(); + + return sc; +} + +static void +cache_destroy(struct cso_context *cso, + struct cso_hash *hash, + unsigned processor) +{ + struct cso_hash_iter iter = cso_hash_first_node(hash); + while (!cso_hash_iter_is_null(iter)) { + void *shader = (void *)cso_hash_iter_data(iter); + if (processor == PIPE_SHADER_FRAGMENT) { + cso_delete_fragment_shader(cso, shader); + } else if (processor == PIPE_SHADER_VERTEX) { + cso_delete_vertex_shader(cso, shader); + } + iter = cso_hash_erase(hash, iter); + } + cso_hash_delete(hash); +} + +void xorg_shaders_destroy(struct xorg_shaders *sc) +{ + cache_destroy(sc->exa->cso, sc->vs_hash, + PIPE_SHADER_VERTEX); + cache_destroy(sc->exa->cso, sc->fs_hash, + PIPE_SHADER_FRAGMENT); + + free(sc); +} + +static INLINE void * +shader_from_cache(struct pipe_context *pipe, + unsigned type, + struct cso_hash *hash, + unsigned key) +{ + void *shader = 0; + + struct cso_hash_iter iter = cso_hash_find(hash, key); + + if (cso_hash_iter_is_null(iter)) { + if (type == PIPE_SHADER_VERTEX) + shader = create_vs(pipe, key); + else + shader = create_fs(pipe, key); + cso_hash_insert(hash, key, shader); + } else + shader = (void *)cso_hash_iter_data(iter); + + return shader; +} + +struct xorg_shader xorg_shaders_get(struct xorg_shaders *sc, + unsigned vs_traits, + unsigned fs_traits) +{ + struct xorg_shader shader = {0}; + void *vs, *fs; + + vs = shader_from_cache(sc->exa->ctx, PIPE_SHADER_VERTEX, + sc->vs_hash, vs_traits); + fs = shader_from_cache(sc->exa->ctx, PIPE_SHADER_FRAGMENT, + sc->fs_hash, fs_traits); + + debug_assert(vs && fs); + if (!vs || !fs) + return shader; + + shader.vs = vs; + shader.fs = fs; + + return shader; +} diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h new file mode 100644 index 00000000000..003e5d8caf4 --- /dev/null +++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h @@ -0,0 +1,35 @@ +#ifndef XORG_EXA_TGSI_H +#define XORG_EXA_TGSI_H + +#include "xorg_exa.h" + +enum xorg_vs_traits { + VS_COMPOSITE = 1 << 0, + VS_MASK = 1 << 1, + VS_FILL = 1 << 2 + /*VS_TRANSFORM = 1 << 3*/ +}; + +enum xorg_fs_traits { + FS_COMPOSITE = 1 << 0, + FS_MASK = 1 << 1, + FS_FILL = 1 << 2, + FS_LINEAR_GRADIENT = 1 << 3, + FS_RADIAL_GRADIENT = 1 << 4 +}; + +struct xorg_shader { + void *fs; + void *vs; +}; + +struct xorg_shaders; + +struct xorg_shaders *xorg_shaders_create(struct exa_context *exa); +void xorg_shaders_destroy(struct xorg_shaders *shaders); + +struct xorg_shader xorg_shaders_get(struct xorg_shaders *shaders, + unsigned vs_traits, + unsigned fs_traits); + +#endif diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h index 82c3890dfbd..8c60d50927d 100644 --- a/src/gallium/state_trackers/xorg/xorg_tracker.h +++ b/src/gallium/state_trackers/xorg/xorg_tracker.h @@ -31,11 +31,19 @@ #ifndef _XORG_TRACKER_H_ #define _XORG_TRACKER_H_ +#include <stddef.h> +#include <stdint.h> #include <errno.h> #include <drm.h> #include <xf86drm.h> #include <xf86drmMode.h> -#include "exa.h" +#include <xorg-server.h> +#include <xf86.h> +#include <exa.h> + +#ifdef DRM_MODE_FEATURE_DIRTYFB +#include <damage.h> +#endif #include "pipe/p_screen.h" #include "state_tracker/drm_api.h" @@ -72,16 +80,23 @@ typedef struct _modesettingRec unsigned int SaveGeneration; + void (*blockHandler)(int, pointer, pointer, pointer); CreateScreenResourcesProcPtr createScreenResources; /* gallium */ + struct drm_api *api; struct pipe_screen *screen; struct pipe_context *ctx; + boolean d_depth_bits_last; + boolean ds_depth_bits_last; /* exa */ void *exa; Bool noEvict; +#ifdef DRM_MODE_FEATURE_DIRTYFB + DamagePtr damage; +#endif } modesettingRec, *modesettingPtr; #define modesettingPTR(p) ((modesettingPtr)((p)->driverPrivate)) @@ -94,7 +109,13 @@ struct pipe_texture * xorg_exa_get_texture(PixmapPtr pPixmap); unsigned -xorg_exa_get_pixmap_handle(PixmapPtr pPixmap); +xorg_exa_get_pixmap_handle(PixmapPtr pPixmap, unsigned *stride); + +int +xorg_exa_set_displayed_usage(PixmapPtr pPixmap); + +int +xorg_exa_set_shared_usage(PixmapPtr pPixmap); void * xorg_exa_init(ScrnInfoPtr pScrn); diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template index 985e5a861fc..9635c3c50e9 100644 --- a/src/gallium/winsys/drm/Makefile.template +++ b/src/gallium/winsys/drm/Makefile.template @@ -83,7 +83,9 @@ default: depend symlinks $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME) $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template $(MKLIB) -noprefix -o $@ \ - $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS) + $(OBJECTS) $(PIPE_DRIVERS) \ + -Wl,--start-group $(MESA_MODULES) -Wl,--end-group \ + $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS) $(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS) $(MKLIB) -o $(LIBNAME_EGL) \ diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile index de39e759d87..5e212b62a46 100644 --- a/src/gallium/winsys/drm/intel/dri/Makefile +++ b/src/gallium/winsys/drm/intel/dri/Makefile @@ -8,6 +8,7 @@ PIPE_DRIVERS = \ $(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \ $(TOP)/src/gallium/drivers/trace/libtrace.a \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/identity/libidentity.a \ $(TOP)/src/gallium/drivers/i915simple/libi915simple.a diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript index e14e96e32fd..6c00861f517 100644 --- a/src/gallium/winsys/drm/intel/dri/SConscript +++ b/src/gallium/winsys/drm/intel/dri/SConscript @@ -2,11 +2,14 @@ Import('*') env = drienv.Clone() +env.ParseConfig('pkg-config --cflags --libs libdrm_intel') + drivers = [ + st_dri, + inteldrm, softpipe, i915simple, trace, - inteldrm ] env.SharedLibrary( diff --git a/src/gallium/winsys/drm/intel/egl/Makefile b/src/gallium/winsys/drm/intel/egl/Makefile index c5217ad2d63..490baded66b 100644 --- a/src/gallium/winsys/drm/intel/egl/Makefile +++ b/src/gallium/winsys/drm/intel/egl/Makefile @@ -8,6 +8,7 @@ PIPE_DRIVERS = \ $(TOP)/src/gallium/state_trackers/egl/libegldrm.a \ $(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ $(TOP)/src/gallium/drivers/i915simple/libi915simple.a DRIVER_SOURCES = diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.c b/src/gallium/winsys/drm/intel/gem/intel_be_api.c index a74be13bf7a..dd5f814da3c 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.c @@ -1,21 +1,37 @@ #include "intel_be_api.h" #include "i915simple/i915_winsys.h" +#include "identity/id_drm.h" +#include "trace/tr_drm.h" -#ifdef DEBUG -#include "trace/trace_drm.h" +static void destroy(struct drm_api *api) +{ -struct drm_api hooks = -#else -struct drm_api drm_api_hooks = -#endif +} + +struct drm_api intel_be_drm_api = { /* intel_be_context.c */ .create_context = intel_be_create_context, /* intel_be_device.c */ .create_screen = intel_be_create_screen, - .buffer_from_texture = i915_get_texture_buffer, + .buffer_from_texture = intel_be_get_texture_buffer, .buffer_from_handle = intel_be_buffer_from_handle, .handle_from_buffer = intel_be_handle_from_buffer, .global_handle_from_buffer = intel_be_global_handle_from_buffer, + .destroy = destroy, }; + +struct drm_api * +drm_api_create() +{ +#ifdef DEBUG +#if 0 + return identity_drm_create(&intel_be_drm_api); +#else + return trace_drm_create(&intel_be_drm_api); +#endif +#else + return &intel_be_drm_api; +#endif +} diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_api.h b/src/gallium/winsys/drm/intel/gem/intel_be_api.h index 1c622f3b978..f286b62eb8c 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_api.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_api.h @@ -8,8 +8,11 @@ #include "intel_be_device.h" -struct pipe_screen *intel_be_create_screen(int drmFD, +extern struct drm_api intel_be_drm_api; + +struct pipe_screen *intel_be_create_screen(struct drm_api *api, int drmFD, struct drm_create_screen_arg *arg); -struct pipe_context *intel_be_create_context(struct pipe_screen *screen); +struct pipe_context *intel_be_create_context(struct drm_api *api, + struct pipe_screen *screen); #endif diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c index d5e63c3bae5..c4a79586e68 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c @@ -21,12 +21,10 @@ intel_be_batchbuffer_alloc(struct intel_be_context *intel) batch->base.size = 0; batch->base.actual_size = intel->device->max_batch_size; batch->base.relocs = 0; - batch->base.max_relocs = 500;/*INTEL_DEFAULT_RELOCS;*/ + batch->base.max_relocs = 100;/*INTEL_DEFAULT_RELOCS;*/ - batch->base.map = malloc(batch->base.actual_size); - memset(batch->base.map, 0, batch->base.actual_size); - - batch->base.ptr = batch->base.map; + batch->intel = intel; + batch->device = intel->device; intel_be_batchbuffer_reset(batch); @@ -41,16 +39,17 @@ intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) if (batch->bo) drm_intel_bo_unreference(batch->bo); + batch->bo = drm_intel_bo_alloc(dev->pools.gem, + "gallium3d_batch_buffer", + batch->base.actual_size, + 4096); + drm_intel_bo_map(batch->bo, TRUE); + batch->base.map = batch->bo->virtual; memset(batch->base.map, 0, batch->base.actual_size); batch->base.ptr = batch->base.map; batch->base.size = batch->base.actual_size - BATCH_RESERVED; - batch->base.relocs = 0; - - batch->bo = drm_intel_bo_alloc(dev->pools.gem, - "gallium3d_batch_buffer", - batch->base.actual_size, 0); } int @@ -88,6 +87,7 @@ intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch, struct i915_batchbuffer *i915 = &batch->base; unsigned used = 0; int ret = 0; + int i; assert(i915_batchbuffer_space(i915) >= 0); @@ -105,11 +105,29 @@ intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch, used = batch->base.ptr - batch->base.map; - drm_intel_bo_subdata(batch->bo, 0, used, batch->base.map); - ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0); + drm_intel_bo_unmap(batch->bo); + /* Do the sending to HW */ + ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0); assert(ret == 0); + if (batch->device->dump_cmd) { + unsigned *ptr; + drm_intel_bo_map(batch->bo, FALSE); + ptr = (unsigned*)batch->bo->virtual; + + debug_printf("%s:\n", __func__); + for (i = 0; i < used / 4; i++, ptr++) { + debug_printf("\t%08x: %08x\n", i*4, *ptr); + } + + drm_intel_bo_unmap(batch->bo); + } else { + /* TODO figgure out why the gpu hangs if we don't run sync */ + drm_intel_bo_map(batch->bo, FALSE); + drm_intel_bo_unmap(batch->bo); + } + intel_be_batchbuffer_reset(batch); if (fence) { @@ -134,6 +152,5 @@ intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) if (batch->bo) drm_intel_bo_unreference(batch->bo); - free(batch->base.map); free(batch); } diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c index fe0b138fbea..ff4518f8680 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_context.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c @@ -36,7 +36,7 @@ intel_be_batch_reloc(struct i915_winsys *sws, } if (access_flags & I915_BUFFER_ACCESS_READ) { - read |= I915_GEM_DOMAIN_VERTEX; + read |= I915_GEM_DOMAIN_SAMPLER; } ret = intel_be_offset_relocation(intel->batch, @@ -57,9 +57,6 @@ intel_be_batch_flush(struct i915_winsys *sws, struct intel_be_context *intel = intel_be_context(sws); struct intel_be_fence **f = (struct intel_be_fence **)fence; - if (fence && *fence) - assert(0); - intel_be_batchbuffer_flush(intel->batch, f); } @@ -97,7 +94,7 @@ intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *de } struct pipe_context * -intel_be_create_context(struct pipe_screen *screen) +intel_be_create_context(struct drm_api *api, struct pipe_screen *screen) { struct intel_be_context *intel; struct pipe_context *pipe; diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c index 907ac86637f..5312865a039 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.c +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c @@ -7,6 +7,7 @@ #include "pipe/p_inlines.h" #include "util/u_memory.h" #include "util/u_debug.h" +#include "util/u_math.h" #include "intel_be_fence.h" @@ -16,6 +17,8 @@ #include "intel_be_api.h" #include <stdio.h> +#define I915_TILING_X 1 + /* * Buffer */ @@ -25,9 +28,10 @@ intel_be_buffer_map(struct pipe_winsys *winsys, struct pipe_buffer *buf, unsigned flags) { + struct intel_be_buffer *buffer = intel_be_buffer(buf); drm_intel_bo *bo = intel_bo(buf); int write = 0; - int ret; + int ret = 0; if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { /* Remove this when drm_intel_bo_map supports DONTBLOCK @@ -38,19 +42,37 @@ intel_be_buffer_map(struct pipe_winsys *winsys, if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) write = 1; - ret = drm_intel_bo_map(bo, write); + if (buffer->map_count) + goto out; + + if (buffer->map_gtt) + ret = drm_intel_gem_bo_map_gtt(bo); + else + ret = drm_intel_bo_map(bo, write); + buffer->ptr = bo->virtual; + +out: if (ret) return NULL; - return bo->virtual; + buffer->map_count++; + return buffer->ptr; } static void intel_be_buffer_unmap(struct pipe_winsys *winsys, struct pipe_buffer *buf) { - drm_intel_bo_unmap(intel_bo(buf)); + struct intel_be_buffer *buffer = intel_be_buffer(buf); + + if (--buffer->map_count) + return; + + if (buffer->map_gtt) + drm_intel_gem_bo_unmap_gtt(intel_bo(buf)); + else + drm_intel_bo_unmap(intel_bo(buf)); } static void @@ -80,8 +102,13 @@ intel_be_buffer_create(struct pipe_winsys *winsys, buffer->base.size = size; buffer->flinked = FALSE; buffer->flink = 0; + buffer->map_gtt = FALSE; - if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { + if (usage & I915_BUFFER_USAGE_SCANOUT) { + /* Scanout buffer */ + name = "gallium3d_scanout"; + pool = dev->pools.gem; + } else if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { /* Local buffer */ name = "gallium3d_local"; pool = dev->pools.gem; @@ -96,6 +123,12 @@ intel_be_buffer_create(struct pipe_winsys *winsys, } buffer->bo = drm_intel_bo_alloc(pool, name, size, alignment); + if (usage & I915_BUFFER_USAGE_SCANOUT) { + unsigned tiling = I915_TILING_X; + unsigned stride = 2048 * 4; /* TODO do something smarter here */ + drm_intel_bo_set_tiling(buffer->bo, &tiling, stride); + buffer->map_gtt = TRUE; + } if (!buffer->bo) goto err; @@ -142,8 +175,61 @@ err: return NULL; } +static struct pipe_buffer * +intel_be_surface_buffer_create(struct pipe_winsys *winsys, + unsigned width, unsigned height, + enum pipe_format format, + unsigned usage, + unsigned tex_usage, + unsigned *stride) +{ + struct pipe_format_block block; + unsigned buf_usage = 0; + unsigned buf_stride = 0; + unsigned buf_size = 0; + + pf_get_block(format, &block); + buf_stride = pf_get_stride(&block, width); + buf_stride = align(buf_stride, 64); + + if (tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) { + /* TODO more checks */ + assert(buf_stride <= 2048*4); + assert(height % 8 == 0); + buf_stride = 2048 * 4; + buf_usage |= I915_BUFFER_USAGE_SCANOUT; + } + + buf_size = buf_stride * height; + *stride = buf_stride; + + return intel_be_buffer_create(winsys, + 0, + buf_usage, + buf_size); +} + +boolean +intel_be_get_texture_buffer(struct drm_api *api, + struct pipe_texture *texture, + struct pipe_buffer **buffer, + unsigned *stride) +{ + struct intel_be_device *dev; + + if (!texture) + return FALSE; + + dev = intel_be_device(texture->screen->winsys); + if (dev->softpipe) + return softpipe_get_texture_buffer(texture, buffer, stride); + else + return i915_get_texture_buffer(texture, buffer, stride); +} + struct pipe_buffer * -intel_be_buffer_from_handle(struct pipe_screen *screen, +intel_be_buffer_from_handle(struct drm_api *api, + struct pipe_screen *screen, const char* name, unsigned handle) { struct intel_be_device *dev = intel_be_device(screen->winsys); @@ -174,7 +260,8 @@ err: } boolean -intel_be_handle_from_buffer(struct pipe_screen *screen, +intel_be_handle_from_buffer(struct drm_api *api, + struct pipe_screen *screen, struct pipe_buffer *buffer, unsigned *handle) { @@ -186,7 +273,8 @@ intel_be_handle_from_buffer(struct pipe_screen *screen, } boolean -intel_be_global_handle_from_buffer(struct pipe_screen *screen, +intel_be_global_handle_from_buffer(struct drm_api *api, + struct pipe_screen *screen, struct pipe_buffer *buffer, unsigned *handle) { @@ -204,6 +292,7 @@ intel_be_global_handle_from_buffer(struct pipe_screen *screen, *handle = buf->flink; return TRUE; } + /* * Fence */ @@ -275,8 +364,8 @@ intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) dev->base.buffer_unmap = intel_be_buffer_unmap; dev->base.buffer_destroy = intel_be_buffer_destroy; - /* Not used anymore */ - dev->base.surface_buffer_create = NULL; + /* Used by softpipe */ + dev->base.surface_buffer_create = intel_be_surface_buffer_create; dev->base.fence_reference = intel_be_fence_refunref; dev->base.fence_signalled = intel_be_fence_signalled; @@ -287,6 +376,7 @@ intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size); dev->softpipe = debug_get_bool_option("INTEL_SOFTPIPE", FALSE); + dev->dump_cmd = debug_get_bool_option("INTEL_DUMP_CMD", FALSE); return true; } @@ -296,6 +386,7 @@ intel_be_get_device_id(unsigned int *device_id) { char path[512]; FILE *file; + void *shutup_gcc; /* * FIXME: Fix this up to use a drm ioctl or whatever. @@ -307,13 +398,14 @@ intel_be_get_device_id(unsigned int *device_id) return; } - fgets(path, sizeof(path), file); + shutup_gcc = fgets(path, sizeof(path), file); sscanf(path, "%x", device_id); fclose(file); } struct pipe_screen * -intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg) +intel_be_create_screen(struct drm_api *api, int drmFD, + struct drm_create_screen_arg *arg) { struct intel_be_device *dev; struct pipe_screen *screen; @@ -339,7 +431,6 @@ intel_be_create_screen(int drmFD, struct drm_create_screen_arg *arg) if (dev->softpipe) { screen = softpipe_create_screen(&dev->base); - drm_api_hooks.buffer_from_texture = softpipe_get_texture_buffer; } else screen = i915_create_screen(&dev->base, deviceID); diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h index b32637ece29..c397048f8c5 100644 --- a/src/gallium/winsys/drm/intel/gem/intel_be_device.h +++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h @@ -8,6 +8,8 @@ #include "drm.h" #include "intel_bufmgr.h" +struct drm_api; + /* * Device */ @@ -17,6 +19,7 @@ struct intel_be_device struct pipe_winsys base; boolean softpipe; + boolean dump_cmd; int fd; /**< Drm file discriptor */ @@ -45,18 +48,32 @@ intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); struct intel_be_buffer { struct pipe_buffer base; + void *ptr; + unsigned map_count; + boolean map_gtt; + drm_intel_bo *bo; boolean flinked; unsigned flink; }; +/* + * Wrapper for driver get_texture_buffer functions. + */ +boolean +intel_be_get_texture_buffer(struct drm_api *api, + struct pipe_texture *texture, + struct pipe_buffer **buffer, + unsigned *stride); + /** * Create a be buffer from a drm bo handle. * * Takes a reference. */ struct pipe_buffer * -intel_be_buffer_from_handle(struct pipe_screen *screen, +intel_be_buffer_from_handle(struct drm_api *api, + struct pipe_screen *screen, const char* name, unsigned handle); /** @@ -65,7 +82,8 @@ intel_be_buffer_from_handle(struct pipe_screen *screen, * If buffer is destroyed handle may become invalid. */ boolean -intel_be_handle_from_buffer(struct pipe_screen *screen, +intel_be_handle_from_buffer(struct drm_api *api, + struct pipe_screen *screen, struct pipe_buffer *buffer, unsigned *handle); @@ -75,7 +93,8 @@ intel_be_handle_from_buffer(struct pipe_screen *screen, * If buffer is destroyed handle may become invalid. */ boolean -intel_be_global_handle_from_buffer(struct pipe_screen *screen, +intel_be_global_handle_from_buffer(struct drm_api *api, + struct pipe_screen *screen, struct pipe_buffer *buffer, unsigned *handle); diff --git a/src/gallium/winsys/drm/intel/xorg/Makefile b/src/gallium/winsys/drm/intel/xorg/Makefile index d9aa5d54e1d..9e56853b021 100644 --- a/src/gallium/winsys/drm/intel/xorg/Makefile +++ b/src/gallium/winsys/drm/intel/xorg/Makefile @@ -1,36 +1,39 @@ TARGET = modesetting_drv.so CFILES = $(wildcard ./*.c) OBJECTS = $(patsubst ./%.c,./%.o,$(CFILES)) -GALLIUMDIR = ../../../.. TOP = ../../../../../.. -include ${TOP}/configs/current +include $(TOP)/configs/current -CFLAGS = -DHAVE_CONFIG_H \ - -g -Wall -Wimplicit-function-declaration -fPIC \ - $(shell pkg-config --cflags pixman-1 xorg-server libdrm xproto) \ - -I../gem \ - -I${GALLIUMDIR}/include \ - -I${GALLIUMDIR}/drivers \ - -I${GALLIUMDIR}/auxiliary \ - -I${TOP}/src/mesa \ - -I$(TOP)/include \ - -I$(TOP)/src/egl/main +INCLUDES = \ + $(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \ + -I../gem \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main LIBS = \ - $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a \ - $(GALLIUMDIR)/winsys/drm/intel/gem/libinteldrm.a \ + $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a \ + $(TOP)/src/gallium/winsys/drm/intel/gem/libinteldrm.a \ $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + $(TOP)/src/gallium/drivers/trace/libtrace.a \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ $(GALLIUM_AUXILIARIES) +DRIVER_DEFINES = \ + -DHAVE_CONFIG_H + + ############################################# all default: $(TARGET) -$(TARGET): $(OBJECTS) Makefile $(GALLIUMDIR)/state_trackers/xorg/libxorgtracker.a $(LIBS) +$(TARGET): $(OBJECTS) Makefile $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a $(LIBS) $(TOP)/bin/mklib -noprefix -o $@ \ $(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel @@ -41,4 +44,14 @@ install: $(INSTALL) -d $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) $(MINSTALL) -m 755 $(TARGET) $(DESTDIR)/$(XORG_DRIVER_INSTALL_DIR) + +############################################## + + +.c.o: + $(CC) -c $(CFLAGS) $(INCLUDES) $(DRIVER_DEFINES) $< -o $@ + + +############################################## + .PHONY = all clean install diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index 395b21ec7ab..3167ccfdc16 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -12,7 +12,7 @@ #include "nouveau/nouveau_screen.h" static struct pipe_surface * -dri_surface_from_handle(struct pipe_screen *screen, +dri_surface_from_handle(struct drm_api *api, struct pipe_screen *screen, unsigned handle, enum pipe_format format, unsigned width, @@ -24,7 +24,7 @@ dri_surface_from_handle(struct pipe_screen *screen, struct pipe_texture templat; struct pipe_buffer *buf = NULL; - buf = drm_api_hooks.buffer_from_handle(screen, "front buffer", handle); + buf = api->buffer_from_handle(api, screen, "front buffer", handle); if (!buf) return NULL; @@ -70,7 +70,8 @@ static struct dri1_api nouveau_dri1_api = { }; static struct pipe_screen * -nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg) +nouveau_drm_create_screen(struct drm_api *api, int fd, + struct drm_create_screen_arg *arg) { struct dri1_create_screen_arg *dri1 = (void *)arg; struct nouveau_winsys *nvws; @@ -134,13 +135,12 @@ nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg) else format = PIPE_FORMAT_A8R8G8B8_UNORM; - nvws->front = dri_surface_from_handle(nvws->pscreen, - nvdri->front_offset, - format, - nvdri->width, - nvdri->height, - nvdri->front_pitch * - (nvdri->bpp / 8)); + nvws->front = dri_surface_from_handle(api, nvws->pscreen, + nvdri->front_offset, + format, nvdri->width, + nvdri->height, + nvdri->front_pitch * + (nvdri->bpp / 8)); if (!nvws->front) { debug_printf("%s: error referencing front buffer\n", __func__); @@ -155,7 +155,7 @@ nouveau_drm_create_screen(int fd, struct drm_create_screen_arg *arg) } static struct pipe_context * -nouveau_drm_create_context(struct pipe_screen *pscreen) +nouveau_drm_create_context(struct drm_api *api, struct pipe_screen *pscreen) { struct nouveau_winsys *nvws = nouveau_winsys_screen(pscreen); struct pipe_context *(*init)(struct pipe_screen *, unsigned); @@ -206,15 +206,15 @@ nouveau_drm_create_context(struct pipe_screen *pscreen) } static boolean -nouveau_drm_pb_from_pt(struct pipe_texture *pt, struct pipe_buffer **ppb, - unsigned *stride) +nouveau_drm_pb_from_pt(struct drm_api *api, struct pipe_texture *pt, + struct pipe_buffer **ppb, unsigned *stride) { return false; } static struct pipe_buffer * -nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name, - unsigned handle) +nouveau_drm_pb_from_handle(struct drm_api *api, struct pipe_screen *pscreen, + const char *name, unsigned handle) { struct nouveau_device *dev = nouveau_screen(pscreen)->device; struct pipe_buffer *pb; @@ -242,8 +242,8 @@ nouveau_drm_pb_from_handle(struct pipe_screen *pscreen, const char *name, } static boolean -nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, - unsigned *handle) +nouveau_drm_handle_from_pb(struct drm_api *api, struct pipe_screen *pscreen, + struct pipe_buffer *pb, unsigned *handle) { struct nouveau_bo *bo = nouveau_bo(pb); @@ -255,8 +255,8 @@ nouveau_drm_handle_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, } static boolean -nouveau_drm_name_from_pb(struct pipe_screen *pscreen, struct pipe_buffer *pb, - unsigned *handle) +nouveau_drm_name_from_pb(struct drm_api *api, struct pipe_screen *pscreen, + struct pipe_buffer *pb, unsigned *handle) { struct nouveau_bo *bo = nouveau_bo(pb); @@ -275,3 +275,8 @@ struct drm_api drm_api_hooks = { .global_handle_from_buffer = nouveau_drm_name_from_pb, }; +struct drm_api * +drm_api_create() { + return &drm_api_hooks; +} + diff --git a/src/gallium/winsys/drm/radeon/SConscript b/src/gallium/winsys/drm/radeon/SConscript index 8f99055b2f7..b2dfd504d42 100644 --- a/src/gallium/winsys/drm/radeon/SConscript +++ b/src/gallium/winsys/drm/radeon/SConscript @@ -4,4 +4,4 @@ SConscript(['core/SConscript',]) if 'mesa' in env['statetrackers']: - SConscript(['dri2/SConscript']) + SConscript(['dri/SConscript']) diff --git a/src/gallium/winsys/drm/radeon/core/SConscript b/src/gallium/winsys/drm/radeon/core/SConscript index 578174e32ba..2ad68e403fe 100644 --- a/src/gallium/winsys/drm/radeon/core/SConscript +++ b/src/gallium/winsys/drm/radeon/core/SConscript @@ -11,7 +11,9 @@ radeon_sources = [ env.Append(CPPPATH = '#/src/gallium/drivers/r300') -env.ConvenienceLibrary( +radeonwinsys = env.ConvenienceLibrary( target ='radeonwinsys', source = radeon_sources, ) + +Export('radeonwinsys') diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 0d0fdc5bd86..07551e7cd16 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -1,8 +1,8 @@ -/* +/* * Copyright © 2008 Jérôme Glisse * 2009 Corbin Simpson * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,14 +10,14 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. * * The above copyright notice and this permission notice (including the @@ -72,6 +72,7 @@ static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws, alignment, domain, 0); if (radeon_buffer->bo == NULL) { FREE(radeon_buffer); + return NULL; } return &radeon_buffer->base; } @@ -133,16 +134,17 @@ static void *radeon_buffer_map(struct pipe_winsys *ws, (struct radeon_pipe_buffer*)buffer; int write = 0; - if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { - /* XXX Remove this when radeon_bo_map supports DONTBLOCK */ - return NULL; + if (!(flags & PIPE_BUFFER_USAGE_DONTBLOCK)) { + radeon_bo_wait(radeon_buffer->bo); } if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) { write = 1; } - if (radeon_bo_map(radeon_buffer->bo, write)) + if (radeon_bo_map(radeon_buffer->bo, write)) { return NULL; + } + return radeon_buffer->bo->ptr; } @@ -198,6 +200,7 @@ struct radeon_winsys* radeon_pipe_winsys(int fd) return NULL; } + radeon_ws->priv->fd = fd; radeon_ws->priv->bom = radeon_bo_manager_gem_ctor(fd); radeon_ws->base.flush_frontbuffer = radeon_flush_frontbuffer; diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h index ca8bbb3c117..f5153b06af5 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.h @@ -1,7 +1,7 @@ -/* +/* * Copyright © 2008 Jérôme Glisse * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -9,14 +9,14 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. * * The above copyright notice and this permission notice (including the @@ -48,19 +48,19 @@ struct radeon_pipe_buffer { struct pipe_buffer base; struct radeon_bo *bo; + boolean flinked; + uint32_t flink; }; #define RADEON_MAX_BOS 24 struct radeon_winsys_priv { + /* DRM FD */ + int fd; + /* Radeon BO manager. */ struct radeon_bo_manager* bom; - /* Radeon BO space checker. */ - struct radeon_cs_space_check sc[RADEON_MAX_BOS]; - /* Current BO count. */ - unsigned bo_count; - /* Radeon CS manager. */ struct radeon_cs_manager* csm; diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c index 5406d2bbea0..8d818cf8301 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c @@ -1,7 +1,7 @@ -/* +/* * Copyright © 2009 Corbin Simpson * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -9,14 +9,14 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. * * The above copyright notice and this permission notice (including the @@ -30,12 +30,9 @@ #include "radeon_drm.h" -#ifdef DEBUG -#include "trace/trace_drm.h" -#endif - /* Create a pipe_screen. */ -struct pipe_screen* radeon_create_screen(int drmFB, +struct pipe_screen* radeon_create_screen(struct drm_api* api, + int drmFB, struct drm_create_screen_arg *arg) { struct radeon_winsys* winsys = radeon_pipe_winsys(drmFB); @@ -50,25 +47,30 @@ struct pipe_screen* radeon_create_screen(int drmFB, } /* Create a pipe_context. */ -struct pipe_context* radeon_create_context(struct pipe_screen* screen) +struct pipe_context* radeon_create_context(struct drm_api* api, + struct pipe_screen* screen) { if (getenv("RADEON_SOFTPIPE")) { return radeon_create_softpipe(screen->winsys); } else { - return r300_create_context(screen, screen->winsys); + return r300_create_context(screen, + (struct r300_winsys*)screen->winsys); } } -boolean radeon_buffer_from_texture(struct pipe_texture* texture, +boolean radeon_buffer_from_texture(struct drm_api* api, + struct pipe_texture* texture, struct pipe_buffer** buffer, unsigned* stride) { - return FALSE; + /* XXX fix this */ + return r300_get_texture_buffer(texture, buffer, stride); } /* Create a buffer from a handle. */ /* XXX what's up with name? */ -struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, +struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api, + struct pipe_screen* screen, const char* name, unsigned handle) { @@ -95,7 +97,8 @@ struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, return &radeon_buffer->base; } -boolean radeon_handle_from_buffer(struct pipe_screen* screen, +boolean radeon_handle_from_buffer(struct drm_api* api, + struct pipe_screen* screen, struct pipe_buffer* buffer, unsigned* handle) { @@ -105,27 +108,50 @@ boolean radeon_handle_from_buffer(struct pipe_screen* screen, return TRUE; } -boolean radeon_global_handle_from_buffer(struct pipe_screen* screen, +boolean radeon_global_handle_from_buffer(struct drm_api* api, + struct pipe_screen* screen, struct pipe_buffer* buffer, unsigned* handle) { - /* XXX WTF is the difference here? global? */ + int retval, fd; + struct drm_gem_flink flink; struct radeon_pipe_buffer* radeon_buffer = (struct radeon_pipe_buffer*)buffer; - *handle = radeon_buffer->bo->handle; + + if (!radeon_buffer->flinked) { + fd = ((struct radeon_winsys*)screen->winsys)->priv->fd; + + flink.handle = radeon_buffer->bo->handle; + + retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink); + if (retval) { + debug_printf("radeon: DRM_IOCTL_GEM_FLINK failed, error %d\n", + retval); + return FALSE; + } + + radeon_buffer->flink = flink.name; + radeon_buffer->flinked = TRUE; + } + + *handle = radeon_buffer->flink; return TRUE; } -#ifdef DEBUG -struct drm_api hooks = { -#else struct drm_api drm_api_hooks = { -#endif .create_screen = radeon_create_screen, .create_context = radeon_create_context, - /* XXX fix this */ - .buffer_from_texture = r300_get_texture_buffer, + .buffer_from_texture = radeon_buffer_from_texture, .buffer_from_handle = radeon_buffer_from_handle, .handle_from_buffer = radeon_handle_from_buffer, .global_handle_from_buffer = radeon_global_handle_from_buffer, }; + +struct drm_api* drm_api_create() +{ +#ifdef DEBUG + return trace_drm_create(&drm_api_hooks); +#else + return &drm_api_hooks; +#endif +} diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h index 049f9984dba..88a5c82b284 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h @@ -30,8 +30,13 @@ #ifndef RADEON_DRM_H #define RADEON_DRM_H +#include <sys/ioctl.h> + +#include "xf86drm.h" + #include "pipe/p_screen.h" +#include "trace/tr_drm.h" #include "util/u_memory.h" #include "state_tracker/drm_api.h" @@ -40,25 +45,35 @@ #include "radeon_r300.h" #include "radeon_winsys_softpipe.h" -struct pipe_screen* radeon_create_screen(int drmFB, +/* XXX */ +#include "r300_screen.h" + +struct pipe_screen* radeon_create_screen(struct drm_api* api, + int drmFB, struct drm_create_screen_arg *arg); -struct pipe_context* radeon_create_context(struct pipe_screen* screen); +struct pipe_context* radeon_create_context(struct drm_api* api, + struct pipe_screen* screen); -boolean radeon_buffer_from_texture(struct pipe_texture* texture, +boolean radeon_buffer_from_texture(struct drm_api* api, + struct pipe_texture* texture, struct pipe_buffer** buffer, unsigned* stride); -struct pipe_buffer* radeon_buffer_from_handle(struct pipe_screen* screen, +struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api, + struct pipe_screen* screen, const char* name, unsigned handle); -boolean radeon_handle_from_buffer(struct pipe_screen* screen, +boolean radeon_handle_from_buffer(struct drm_api* api, + struct pipe_screen* screen, struct pipe_buffer* buffer, unsigned* handle); -boolean radeon_global_handle_from_buffer(struct pipe_screen* screen, +boolean radeon_global_handle_from_buffer(struct drm_api* api, + struct pipe_screen* screen, struct pipe_buffer* buffer, unsigned* handle); +void radeon_destroy_drm_api(struct drm_api* api); #endif diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index 63aa3179ac8..d7238762219 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -27,66 +27,25 @@ static boolean radeon_r300_add_buffer(struct r300_winsys* winsys, uint32_t rd, uint32_t wd) { - int i; struct radeon_winsys_priv* priv = (struct radeon_winsys_priv*)winsys->radeon_winsys; - struct radeon_cs_space_check* sc = priv->sc; struct radeon_bo* bo = ((struct radeon_pipe_buffer*)pbuffer)->bo; - /* Check to see if this BO is already in line for validation; - * find a slot for it otherwise. */ - for (i = 0; i < priv->bo_count; i++) { - if (sc[i].bo == bo) { - sc[i].read_domains |= rd; - sc[i].write_domain |= wd; - return; - } - } - - if (priv->bo_count >= RADEON_MAX_BOS) { - /* Dohoho. Not falling for that one again. Request a flush. */ - return FALSE; - } - - sc[priv->bo_count].bo = bo; - sc[priv->bo_count].read_domains = rd; - sc[priv->bo_count].write_domain = wd; - priv->bo_count++; - + radeon_cs_space_add_persistent_bo(priv->cs, bo, rd, wd); return TRUE; } static boolean radeon_r300_validate(struct r300_winsys* winsys) { - int retval, i; struct radeon_winsys_priv* priv = (struct radeon_winsys_priv*)winsys->radeon_winsys; - struct radeon_cs_space_check* sc = priv->sc; - - retval = radeon_cs_space_check(priv->cs, sc, priv->bo_count); - - if (retval == RADEON_CS_SPACE_OP_TO_BIG) { - /* We might as well HCF, since this is not going to fit in the card, - * period. */ - /* XXX just drop it on the floor instead */ - exit(1); - } else if (retval == RADEON_CS_SPACE_FLUSH) { - /* We must flush before more rendering can commence. */ - return TRUE; - } - /* XXX should probably be its own function */ - for (i = 0; i < priv->bo_count; i++) { - if (sc[i].read_domains && sc[i].write_domain) { - /* Cute, cute. We need to flush first. */ - debug_printf("radeon: BO %p can't be read and written; " - "requesting flush.\n", sc[i].bo); - return TRUE; - } + if (radeon_cs_space_check(priv->cs) < 0) { + return FALSE; } /* Things are fine, we can proceed as normal. */ - return FALSE; + return TRUE; } static boolean radeon_r300_check_cs(struct r300_winsys* winsys, int size) @@ -151,8 +110,7 @@ static void radeon_r300_flush_cs(struct r300_winsys* winsys) { struct radeon_winsys_priv* priv = (struct radeon_winsys_priv*)winsys->radeon_winsys; - struct radeon_cs_space_check* sc = priv->sc; - int retval = 1; + int retval; /* Emit the CS. */ retval = radeon_cs_emit(priv->cs); @@ -160,44 +118,48 @@ static void radeon_r300_flush_cs(struct r300_winsys* winsys) debug_printf("radeon: Bad CS, dumping...\n"); radeon_cs_print(priv->cs, stderr); } - radeon_cs_erase(priv->cs); /* Clean out BOs. */ - memset(sc, 0, sizeof(struct radeon_cs_space_check) * RADEON_MAX_BOS); - priv->bo_count = 0; + radeon_cs_space_reset_bos(priv->cs); + + /* Reset CS. + * Someday, when we care about performance, we should really find a way + * to rotate between two or three CS objects so that the GPU can be + * spinning through one CS while another one is being filled. */ + radeon_cs_erase(priv->cs); } /* Helper function to do the ioctls needed for setup and init. */ static void do_ioctls(struct r300_winsys* winsys, int fd) { struct drm_radeon_gem_info gem_info = {0}; - drm_radeon_getparam_t gp = {0}; struct drm_radeon_info info = {0}; int target = 0; int retval; info.value = ⌖ - gp.value = ⌖ - /* First, get PCI ID */ + /* First, get the number of pixel pipes */ + info.request = RADEON_INFO_NUM_GB_PIPES; + retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)); + if (retval) { + fprintf(stderr, "%s: Failed to get GB pipe count, " + "error number %d\n", __FUNCTION__, retval); + exit(1); + } + winsys->gb_pipes = target; + + /* Then, get PCI ID */ info.request = RADEON_INFO_DEVICE_ID; retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)); if (retval) { - fprintf(stderr, "%s: New ioctl for PCI ID failed " - "(error number %d), trying classic ioctl...\n", - __FUNCTION__, retval); - gp.param = RADEON_PARAM_DEVICE_ID; - retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, - sizeof(gp)); - if (retval) { - fprintf(stderr, "%s: Failed to get PCI ID, " - "error number %d\n", __FUNCTION__, retval); - exit(1); - } + fprintf(stderr, "%s: Failed to get PCI ID, " + "error number %d\n", __FUNCTION__, retval); + exit(1); } winsys->pci_id = target; - /* Then, retrieve MM info */ + /* Finally, retrieve MM info */ retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO, &gem_info, sizeof(gem_info)); if (retval) { diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.h b/src/gallium/winsys/drm/radeon/core/radeon_r300.h index a2e0e582485..741c1371889 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.h +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.h @@ -20,6 +20,9 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#ifndef RADEON_R300_H +#define RADEON_R300_H + /* XXX WTF is this! I shouldn't have to include those first three! FUCK! */ #include <stdint.h> #include <stdlib.h> @@ -31,18 +34,9 @@ #include "radeon_buffer.h" -/* protect us from bonghits */ -#ifndef RADEON_INFO_DEVICE_ID -#define RADEON_INFO_DEVICE_ID 0 -#endif -#ifndef DRM_RADEON_INFO -#define DRM_RADEON_INFO 0x1 -struct drm_radeon_info { - uint32_t request; - uint32_t pad; - uint64_t value; -}; -#endif +struct radeon_winsys; struct r300_winsys* radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys); + +#endif /* RADEON_R300_H */ diff --git a/src/gallium/winsys/drm/radeon/dri/SConscript b/src/gallium/winsys/drm/radeon/dri/SConscript index f2cdee97d92..aea987a3aca 100644 --- a/src/gallium/winsys/drm/radeon/dri/SConscript +++ b/src/gallium/winsys/drm/radeon/dri/SConscript @@ -2,7 +2,10 @@ Import('*') env = drienv.Clone() +env.ParseConfig('pkg-config --cflags --libs libdrm_radeon') + drivers = [ + trace, softpipe, r300 ] @@ -10,5 +13,5 @@ drivers = [ env.SharedLibrary( target ='radeon_dri.so', source = COMMON_GALLIUM_SOURCES, - LIBS = drivers + mesa + auxiliaries + env['LIBS'], + LIBS = st_dri + radeonwinsys + mesa + drivers + auxiliaries + env['LIBS'], ) diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile index a33a50ec22b..3efb7ed4afa 100644 --- a/src/gallium/winsys/egl_xlib/Makefile +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -37,7 +37,7 @@ UNUSED_LIBS = \ $(TOP)/src/mesa/libmesagallium.a \ -LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 +LOCAL_CFLAGS = .c.o: diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index c10e3c00ffc..d02f8250478 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -33,6 +33,7 @@ #include <dlfcn.h> +#include <X11/Xlib.h> #include <X11/Xutil.h> #include "pipe/p_compiler.h" @@ -61,6 +62,14 @@ struct xlib_egl_driver { _EGLDriver Base; /**< base class */ + EGLint apis; +}; + + +/** driver data of _EGLDisplay */ +struct xlib_egl_display +{ + Display *Dpy; struct pipe_winsys *winsys; struct pipe_screen *screen; @@ -82,6 +91,7 @@ struct xlib_egl_surface { _EGLSurface Base; /**< base class */ + /* These are set for window surface */ Display *Dpy; /**< The X Display of the window */ Window Win; /**< The user-created window ID */ GC Gc; @@ -93,6 +103,12 @@ struct xlib_egl_surface }; +static void +flush_frontbuffer(struct pipe_winsys *pws, + struct pipe_surface *psurf, + void *context_private); + + /** cast wrapper */ static INLINE struct xlib_egl_driver * xlib_egl_driver(_EGLDriver *drv) @@ -101,19 +117,24 @@ xlib_egl_driver(_EGLDriver *drv) } -static struct xlib_egl_surface * -lookup_surface(EGLSurface surf) +static INLINE struct xlib_egl_display * +xlib_egl_display(_EGLDisplay *dpy) { - _EGLSurface *surface = _eglLookupSurface(surf); - return (struct xlib_egl_surface *) surface; + return (struct xlib_egl_display *) dpy->DriverData; } -static struct xlib_egl_context * -lookup_context(EGLContext surf) +static INLINE struct xlib_egl_surface * +lookup_surface(_EGLSurface *surf) { - _EGLContext *context = _eglLookupContext(surf); - return (struct xlib_egl_context *) context; + return (struct xlib_egl_surface *) surf; +} + + +static INLINE struct xlib_egl_context * +lookup_context(_EGLContext *ctx) +{ + return (struct xlib_egl_context *) ctx; } @@ -132,19 +153,18 @@ bitcount(unsigned int n) * Create the EGLConfigs. (one per X visual) */ static void -create_configs(_EGLDriver *drv, EGLDisplay dpy) +create_configs(struct xlib_egl_display *xdpy, _EGLDisplay *disp) { static const EGLint all_apis = (EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT | EGL_OPENVG_BIT | EGL_OPENGL_BIT); - _EGLDisplay *disp = _eglLookupDisplay(dpy); XVisualInfo *visInfo, visTemplate; int num_visuals, i; /* get list of all X visuals, create an EGL config for each */ - visTemplate.screen = DefaultScreen(disp->Xdpy); - visInfo = XGetVisualInfo(disp->Xdpy, VisualScreenMask, + visTemplate.screen = DefaultScreen(xdpy->Dpy); + visInfo = XGetVisualInfo(xdpy->Dpy, VisualScreenMask, &visTemplate, &num_visuals); if (!visInfo) { printf("egl_xlib.c: couldn't get any X visuals\n"); @@ -180,10 +200,14 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy) SET_CONFIG_ATTRIB(config, EGL_NATIVE_RENDERABLE, EGL_FALSE); SET_CONFIG_ATTRIB(config, EGL_CONFORMANT, all_apis); SET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE, all_apis); - SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT); + SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT | EGL_PBUFFER_BIT); + SET_CONFIG_ATTRIB(config, EGL_BIND_TO_TEXTURE_RGBA, EGL_TRUE); + SET_CONFIG_ATTRIB(config, EGL_BIND_TO_TEXTURE_RGB, EGL_TRUE); _eglAddConfig(disp, config); } + + XFree(visInfo); } @@ -191,16 +215,47 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy) * Called via eglInitialize(), drv->API.Initialize(). */ static EGLBoolean -xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, - EGLint *minor, EGLint *major) +xlib_eglInitialize(_EGLDriver *drv, _EGLDisplay *dpy, + EGLint *major, EGLint *minor) { - create_configs(drv, dpy); + struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); + struct xlib_egl_display *xdpy; + + xdpy = CALLOC_STRUCT(xlib_egl_display); + if (!xdpy) + return _eglError(EGL_BAD_ALLOC, "eglInitialize"); + + xdpy->Dpy = (Display *) dpy->NativeDisplay; + if (!xdpy->Dpy) { + xdpy->Dpy = XOpenDisplay(NULL); + if (!xdpy->Dpy) { + free(xdpy); + return EGL_FALSE; + } + } - drv->Initialized = EGL_TRUE; + /* create winsys and pipe screen */ + xdpy->winsys = create_sw_winsys(); + if (!xdpy->winsys) { + free(xdpy); + return _eglError(EGL_BAD_ALLOC, "eglInitialize"); + } + xdpy->winsys->flush_frontbuffer = flush_frontbuffer; + xdpy->screen = softpipe_create_screen(xdpy->winsys); + if (!xdpy->screen) { + free(xdpy->winsys); + free(xdpy); + return _eglError(EGL_BAD_ALLOC, "eglInitialize"); + } + + dpy->DriverData = (void *) xdpy; + dpy->ClientAPIsMask = xdrv->apis; + + create_configs(xdpy, dpy); /* we're supporting EGL 1.4 */ - *minor = 1; - *major = 4; + *major = 1; + *minor = 4; return EGL_TRUE; } @@ -210,8 +265,20 @@ xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, * Called via eglTerminate(), drv->API.Terminate(). */ static EGLBoolean -xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) +xlib_eglTerminate(_EGLDriver *drv, _EGLDisplay *dpy) { + struct xlib_egl_display *xdpy = xlib_egl_display(dpy); + + _eglReleaseDisplayResources(drv, dpy); + _eglCleanupDisplay(dpy); + + xdpy->screen->destroy(xdpy->screen); + free(xdpy->winsys); + + if (!dpy->NativeDisplay) + XCloseDisplay(xdpy->Dpy); + free(xdpy); + return EGL_TRUE; } @@ -264,7 +331,13 @@ static void check_and_update_buffer_size(struct xlib_egl_surface *surface) { uint width, height; - get_drawable_size(surface->Dpy, surface->Win, &width, &height); + if (surface->Base.Type == EGL_PBUFFER_BIT) { + width = surface->Base.Width; + height = surface->Base.Height; + } + else { + get_drawable_size(surface->Dpy, surface->Win, &width, &height); + } st_resize_framebuffer(surface->Framebuffer, width, height); surface->Base.Width = width; surface->Base.Height = height; @@ -281,6 +354,9 @@ display_surface(struct pipe_winsys *pws, XImage *ximage; void *data; + if (xsurf->Base.Type == EGL_PBUFFER_BIT) + return; + ximage = XCreateImage(xsurf->Dpy, xsurf->VisInfo.visual, xsurf->VisInfo.depth, @@ -330,24 +406,23 @@ flush_frontbuffer(struct pipe_winsys *pws, /** * Called via eglCreateContext(), drv->API.CreateContext(). */ -static EGLContext -xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, - EGLContext share_list, const EGLint *attrib_list) +static _EGLContext * +xlib_eglCreateContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf, + _EGLContext *share_list, const EGLint *attrib_list) { - struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); - _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); + struct xlib_egl_display *xdpy = xlib_egl_display(dpy); struct xlib_egl_context *ctx; struct st_context *share_ctx = NULL; /* XXX fix */ __GLcontextModes visual; ctx = CALLOC_STRUCT(xlib_egl_context); if (!ctx) - return EGL_NO_CONTEXT; + return NULL; /* let EGL lib init the common stuff */ - if (!_eglInitContext(drv, dpy, &ctx->Base, config, attrib_list)) { + if (!_eglInitContext(drv, &ctx->Base, conf, attrib_list)) { free(ctx); - return EGL_NO_CONTEXT; + return NULL; } /* API-dependent context creation */ @@ -359,7 +434,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, /* fall-through */ case EGL_OPENGL_API: /* create a softpipe context */ - ctx->pipe = softpipe_create(xdrv->screen); + ctx->pipe = softpipe_create(xdpy->screen); /* Now do xlib / state tracker inits here */ _eglConfigToContextModesRec(conf, &visual); ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); @@ -367,43 +442,33 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, default: _eglError(EGL_BAD_MATCH, "eglCreateContext(unsupported API)"); free(ctx); - return EGL_NO_CONTEXT; + return NULL; } - _eglSaveContext(&ctx->Base); - - return _eglGetContextHandle(&ctx->Base); + return &ctx->Base; } static EGLBoolean -xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) +xlib_eglDestroyContext(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx) { struct xlib_egl_context *context = lookup_context(ctx); - if (context) { - if (context->Base.IsBound) { - context->Base.DeletePending = EGL_TRUE; - } - else { - /* API-dependent clean-up */ - switch (context->Base.ClientAPI) { - case EGL_OPENGL_ES_API: - case EGL_OPENVG_API: - /* fall-through */ - case EGL_OPENGL_API: - st_destroy_context(context->Context); - break; - default: - assert(0); - } - free(context); + + if (!_eglIsContextBound(&context->Base)) { + /* API-dependent clean-up */ + switch (context->Base.ClientAPI) { + case EGL_OPENGL_ES_API: + case EGL_OPENVG_API: + /* fall-through */ + case EGL_OPENGL_API: + st_destroy_context(context->Context); + break; + default: + assert(0); } - return EGL_TRUE; - } - else { - _eglError(EGL_BAD_CONTEXT, "eglDestroyContext"); - return EGL_TRUE; + free(context); } + return EGL_TRUE; } @@ -411,16 +476,25 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) * Called via eglMakeCurrent(), drv->API.MakeCurrent(). */ static EGLBoolean -xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, - EGLSurface draw, EGLSurface read, EGLContext ctx) +xlib_eglMakeCurrent(_EGLDriver *drv, _EGLDisplay *dpy, + _EGLSurface *draw, _EGLSurface *read, _EGLContext *ctx) { struct xlib_egl_context *context = lookup_context(ctx); struct xlib_egl_surface *draw_surf = lookup_surface(draw); struct xlib_egl_surface *read_surf = lookup_surface(read); + struct st_context *oldcontext = NULL; + _EGLContext *oldctx; + + oldctx = _eglGetCurrentContext(); + if (oldctx && _eglIsContextLinked(oldctx)) + oldcontext = st_get_current(); - if (!_eglMakeCurrent(drv, dpy, draw, read, context)) + if (!_eglMakeCurrent(drv, dpy, draw, read, ctx)) return EGL_FALSE; + /* Flush before switching context. Check client API? */ + if (oldcontext) + st_flush(oldcontext, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, NULL); st_make_current((context ? context->Context : NULL), (draw_surf ? draw_surf->Framebuffer : NULL), (read_surf ? read_surf->Framebuffer : NULL)); @@ -474,39 +548,34 @@ choose_stencil_format(const __GLcontextModes *visual) /** * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface(). */ -static EGLSurface -xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, +static _EGLSurface * +xlib_eglCreateWindowSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf, NativeWindowType window, const EGLint *attrib_list) { - struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); - _EGLDisplay *disp = _eglLookupDisplay(dpy); - _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); - + struct xlib_egl_display *xdpy = xlib_egl_display(disp); struct xlib_egl_surface *surf; __GLcontextModes visual; uint width, height; surf = CALLOC_STRUCT(xlib_egl_surface); if (!surf) - return EGL_NO_SURFACE; + return NULL; /* Let EGL lib init the common stuff */ - if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_WINDOW_BIT, - config, attrib_list)) { + if (!_eglInitSurface(drv, &surf->Base, EGL_WINDOW_BIT, + conf, attrib_list)) { free(surf); - return EGL_NO_SURFACE; + return NULL; } - _eglSaveSurface(&surf->Base); - /* * Now init the Xlib and gallium stuff */ surf->Win = (Window) window; /* The X window ID */ - surf->Dpy = disp->Xdpy; /* The X display */ + surf->Dpy = xdpy->Dpy; /* The X display */ surf->Gc = XCreateGC(surf->Dpy, surf->Win, 0, NULL); - surf->winsys = xdrv->winsys; + surf->winsys = xdpy->winsys; _eglConfigToContextModesRec(conf, &visual); get_drawable_size(surf->Dpy, surf->Win, &width, &height); @@ -525,35 +594,172 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, st_resize_framebuffer(surf->Framebuffer, width, height); - return _eglGetSurfaceHandle(&surf->Base); + return &surf->Base; +} + + +static _EGLSurface * +xlib_eglCreatePbufferSurface(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf, + const EGLint *attrib_list) +{ + struct xlib_egl_display *xdpy = xlib_egl_display(disp); + struct xlib_egl_surface *surf; + __GLcontextModes visual; + uint width, height; + EGLBoolean bind_texture; + + surf = CALLOC_STRUCT(xlib_egl_surface); + if (!surf) { + _eglError(EGL_BAD_ALLOC, "eglCreatePbufferSurface"); + return NULL; + } + + if (!_eglInitSurface(drv, &surf->Base, EGL_PBUFFER_BIT, + conf, attrib_list)) { + free(surf); + return NULL; + } + if (surf->Base.Width < 0 || surf->Base.Height < 0) { + _eglError(EGL_BAD_PARAMETER, "eglCreatePbufferSurface"); + free(surf); + return NULL; + } + + bind_texture = (surf->Base.TextureFormat != EGL_NO_TEXTURE); + width = (uint) surf->Base.Width; + height = (uint) surf->Base.Height; + if ((surf->Base.TextureTarget == EGL_NO_TEXTURE && bind_texture) || + (surf->Base.TextureTarget != EGL_NO_TEXTURE && !bind_texture)) { + _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface"); + free(surf); + return NULL; + } + /* a framebuffer of zero width or height confuses st */ + if (width == 0 || height == 0) { + _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface"); + free(surf); + return NULL; + } + /* no mipmap generation */ + if (surf->Base.MipmapTexture) { + _eglError(EGL_BAD_MATCH, "eglCreatePbufferSurface"); + free(surf); + return NULL; + } + + surf->winsys = xdpy->winsys; + + _eglConfigToContextModesRec(conf, &visual); + + /* Create GL statetracker framebuffer */ + surf->Framebuffer = st_create_framebuffer(&visual, + choose_color_format(&visual), + choose_depth_format(&visual), + choose_stencil_format(&visual), + width, height, + (void *) surf); + st_resize_framebuffer(surf->Framebuffer, width, height); + + return &surf->Base; } static EGLBoolean -xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +xlib_eglDestroySurface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface) { struct xlib_egl_surface *surf = lookup_surface(surface); - if (surf) { - _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface); - if (surf->Base.IsBound) { - surf->Base.DeletePending = EGL_TRUE; - } - else { + if (!_eglIsSurfaceBound(&surf->Base)) { + if (surf->Base.Type != EGL_PBUFFER_BIT) XFreeGC(surf->Dpy, surf->Gc); - st_unreference_framebuffer(surf->Framebuffer); - free(surf); - } - return EGL_TRUE; + st_unreference_framebuffer(surf->Framebuffer); + free(surf); } - else { - _eglError(EGL_BAD_SURFACE, "eglDestroySurface"); - return EGL_FALSE; + return EGL_TRUE; +} + + +static EGLBoolean +xlib_eglBindTexImage(_EGLDriver *drv, _EGLDisplay *dpy, + _EGLSurface *surface, EGLint buffer) +{ + struct xlib_egl_surface *xsurf = lookup_surface(surface); + struct xlib_egl_context *xctx; + struct pipe_surface *psurf; + enum pipe_format format; + int target; + + if (!xsurf || xsurf->Base.Type != EGL_PBUFFER_BIT) + return _eglError(EGL_BAD_SURFACE, "eglBindTexImage"); + if (buffer != EGL_BACK_BUFFER) + return _eglError(EGL_BAD_PARAMETER, "eglBindTexImage"); + if (xsurf->Base.BoundToTexture) + return _eglError(EGL_BAD_ACCESS, "eglBindTexImage"); + + /* this should be updated when choose_color_format is */ + switch (xsurf->Base.TextureFormat) { + case EGL_TEXTURE_RGB: + format = PIPE_FORMAT_R8G8B8_UNORM; + break; + case EGL_TEXTURE_RGBA: + format = PIPE_FORMAT_A8R8G8B8_UNORM; + break; + default: + return _eglError(EGL_BAD_MATCH, "eglBindTexImage"); + } + + switch (xsurf->Base.TextureTarget) { + case EGL_TEXTURE_2D: + target = ST_TEXTURE_2D; + break; + default: + return _eglError(EGL_BAD_MATCH, "eglBindTexImage"); + } + + /* flush properly */ + if (eglGetCurrentSurface(EGL_DRAW) == surface) { + xctx = lookup_context(_eglGetCurrentContext()); + st_flush(xctx->Context, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, + NULL); + } + else if (_eglIsSurfaceBound(&xsurf->Base)) { + xctx = lookup_context(xsurf->Base.Binding); + if (xctx) + st_finish(xctx->Context); } + + st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT, + &psurf); + st_bind_texture_surface(psurf, target, xsurf->Base.MipmapLevel, format); + xsurf->Base.BoundToTexture = EGL_TRUE; + + return EGL_TRUE; +} + + +static EGLBoolean +xlib_eglReleaseTexImage(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surface, + EGLint buffer) +{ + struct xlib_egl_surface *xsurf = lookup_surface(surface); + struct pipe_surface *psurf; + + if (!xsurf || xsurf->Base.Type != EGL_PBUFFER_BIT || + !xsurf->Base.BoundToTexture) + return _eglError(EGL_BAD_SURFACE, "eglReleaseTexImage"); + if (buffer != EGL_BACK_BUFFER) + return _eglError(EGL_BAD_PARAMETER, "eglReleaseTexImage"); + + st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT, + &psurf); + st_unbind_texture_surface(psurf, ST_TEXTURE_2D, xsurf->Base.MipmapLevel); + xsurf->Base.BoundToTexture = EGL_FALSE; + + return EGL_TRUE; } static EGLBoolean -xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +xlib_eglSwapBuffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *draw) { /* error checking step: */ if (!_eglSwapBuffers(drv, dpy, draw)) @@ -610,12 +816,20 @@ find_supported_apis(void) } +static void +xlib_Unload(_EGLDriver *drv) +{ + struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); + free(xdrv); +} + + /** * This is the main entrypoint into the driver. * Called by libEGL to instantiate an _EGLDriver object. */ _EGLDriver * -_eglMain(_EGLDisplay *dpy, const char *args) +_eglMain(const char *args) { struct xlib_egl_driver *xdrv; @@ -625,10 +839,6 @@ _eglMain(_EGLDisplay *dpy, const char *args) if (!xdrv) return NULL; - if (!dpy->Xdpy) { - dpy->Xdpy = XOpenDisplay(NULL); - } - _eglInitDriverFallbacks(&xdrv->Base); xdrv->Base.API.Initialize = xlib_eglInitialize; xdrv->Base.API.Terminate = xlib_eglTerminate; @@ -636,27 +846,24 @@ _eglMain(_EGLDisplay *dpy, const char *args) xdrv->Base.API.CreateContext = xlib_eglCreateContext; xdrv->Base.API.DestroyContext = xlib_eglDestroyContext; xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface; + xdrv->Base.API.CreatePbufferSurface = xlib_eglCreatePbufferSurface; xdrv->Base.API.DestroySurface = xlib_eglDestroySurface; + xdrv->Base.API.BindTexImage = xlib_eglBindTexImage; + xdrv->Base.API.ReleaseTexImage = xlib_eglReleaseTexImage; xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent; xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers; - xdrv->Base.ClientAPIsMask = find_supported_apis(); - if (xdrv->Base.ClientAPIsMask == 0x0) { + xdrv->apis = find_supported_apis(); + if (xdrv->apis == 0x0) { /* the app isn't directly linked with any EGL-supprted APIs * (such as libGLESv2.so) so use an EGL utility to see what * APIs might be loaded dynamically on this system. */ - xdrv->Base.ClientAPIsMask = _eglFindAPIs(); - } + xdrv->apis = _eglFindAPIs(); + } xdrv->Base.Name = "Xlib/softpipe"; - - /* create one winsys and use it for all contexts/surfaces */ - xdrv->winsys = create_sw_winsys(); - xdrv->winsys->flush_frontbuffer = flush_frontbuffer; - - xdrv->screen = softpipe_create_screen(xdrv->winsys); + xdrv->Base.Unload = xlib_Unload; return &xdrv->Base; } - diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index aa1bfa8e883..79ff2cc985d 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -166,6 +166,7 @@ surface_buffer_create(struct pipe_winsys *winsys, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { const unsigned alignment = 64; diff --git a/src/gallium/winsys/xlib/xlib_brw_screen.c b/src/gallium/winsys/xlib/xlib_brw_screen.c index fe8dfff7672..6f3861e2cd6 100644 --- a/src/gallium/winsys/xlib/xlib_brw_screen.c +++ b/src/gallium/winsys/xlib/xlib_brw_screen.c @@ -249,6 +249,7 @@ aub_i915_surface_buffer_create(struct pipe_winsys *winsys, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { const unsigned alignment = 64; diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c index 44b8464518a..67fea023a3b 100644 --- a/src/gallium/winsys/xlib/xlib_softpipe.c +++ b/src/gallium/winsys/xlib/xlib_softpipe.c @@ -303,6 +303,7 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, */ XMesaContext xmctx = (XMesaContext) context_private; xlib_softpipe_display_surface(xmctx->xm_buffer, surf); + xmesa_check_and_update_buffer_size(xmctx, xmctx->xm_buffer); } @@ -375,6 +376,7 @@ xm_surface_buffer_create(struct pipe_winsys *winsys, unsigned width, unsigned height, enum pipe_format format, unsigned usage, + unsigned tex_usage, unsigned *stride) { const unsigned alignment = 64; |