2017-02-09 Nathan Sidwell <nathan@codesourcery.com>

	    Cesar Philippidis  <cesar@codesourcery.com>
	    Joseph Myers  <joseph@codesourcery.com>
	    Chung-Lin Tang  <cltang@codesourcery.com>

	gcc/
	* gimplify.c (gimplify_scan_omp_clauses): No special handling for
	OMP_CLAUSE_TILE.
	(gimplify_adjust_omp_clauses): Don't delete TILE.
	(gimplify_omp_for): Deal with TILE.
	* internal-fn.c (expand_GOACC_TILE): New function.
	* internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
	(GOACC_TILE): New.
	* omp-expand.c (struct oacc_collapse): Add tile and outer fields.
	(expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
	element fields.
	(expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
	avoid DIV for outermost collapse var.
	(expand_oacc_for): Insert tile element loop as needed.  Adjust.
	Remove out of date comments, fix whitespace.
	* omp-general.c (omp_extract_for_data): Deal with tiling.
	* omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
	adjust OLF_DIM_BASE value.
	(struct omp_for_data): Add tiling field.
	* omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
	(lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
	for auto loops.  Remove default auto determining, moved to
	oacc_loop_fixed_partitions.
	* omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
	stmts, add e_mask field.
	(oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
	(oacc_thread_numbers): Use oacc_dim_call.
	(oacc_xform_tile): New.
	(new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
	(finish_oacc_loop): Adjust for ifns vector.
	(oacc_loop_discover_walk): Append loop abstraction sites to list,
	add case for GOACC_TILE fns.
	(oacc_loop_xform_loop): Delete.
	(oacc_loop_process): Iterate over call list directly, and add
	handling for GOACC_TILE fns.
	(oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
	dump partitioning.
	(oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
	vector partitioning to outer loops.  Assign 2 partitions to loops
	when available. Add TILE handling.
	(oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
	(execite_oacc_device_lower): Process GOACC_TILE fns, ignore unknown specs.
	* tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
	* tree.c (omp_clause_num_ops): Adjust TILE ops.
	* tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.

	gcc/c/
	* c-parser.c (c_parser_omp_clause_collapse): Disallow tile.
	(c_parser_oacc_clause_tile): Disallow collapse. Fix parsing and
	semantic checking.
	* c-parser.c (c_parser_omp_for_loop): Accept tiling constructs.

	gcc/cp/
	* parser.c (cp_parser_oacc_clause_tile): Disallow collapse.  Fix
	parsing.  Parse constant expression. Remove semantic checking.
	(cp_parser_omp_clause_collapse): Disallow tile.
	(cp_parser_omp_for_loop): Deal with tile clause.  Don't emit a parse
	error about missing for after already emitting one.  Use more
	conventional for idiom for unbounded loop.
	* pt.c (tsubst_omp_clauses): Handle OMP_CLAUSE_TILE.
	* semantics.c (finish_omp_clauses): Correct TILE semantic check.
	(finish_omp_for): Deal with tile clause.

	gcc/fortran/
	* openmp.c (resolve_omp_clauses): Error on directives
	containing both tile and collapse clauses.
	(resolve_oacc_loop_blocks): Represent '*' tile arguments as zero.
	* trans-openmp.c (gfc_trans_omp_do): Lower tiled loops like
	collapsed loops.

	gcc/testsuite/
	* c-c++-common/goacc/combined-directives.c: Remove xfail.
	* c-c++-common/goacc/loop-auto-1.c: Adjust and add additional case.
	* c-c++-common/goacc/loop-auto-2.c: New.
	* c-c++-common/goacc/tile.c: Include stdbool, fix expected errors.
	* c-c++-common/goacc/tile-2.c: New.
	* g++.dg/goacc/template.C: Test tile subst.  Adjust erroneous uses.
	* g++.dg/goacc/tile-1.C: New, check tile subst.
	* gcc.dg/goacc/loop-processing-1.c: Adjust dg-final pattern.
	* gfortran.dg/goacc/combined-directives.f90: Remove xfail.
	* gfortran.dg/goacc/tile-1.f90: New test.
	* gfortran.dg/goacc/tile-2.f90: New test.
	* gfortran.dg/goacc/tile-lowering.f95: New test.

	libgomp/
	* testsuite/libgomp.oacc-c-c++-common/tile-1.c: New.
	* testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c: Adjust and
	add additional case.
	* testsuite/libgomp.oacc-c-c++-common/vprop.c: XFAIL under
	"openacc_nvidia_accel_selected".
	* libgomp.oacc-fortran/nested-function-1.f90 (test2):
	Add num_workers(8) clause.



git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@245300 138bc75d-0d04-0410-961f-82ee72b054a4

3 files changed, 24 insertions, 2 deletions

diff --git a/gcc/fortran/ChangeLog b/gcc/fortran/ChangeLog
index 0b3279667dd..3488f01e7cd 100644
--- a/gcc/fortran/ChangeLog
+++ b/gcc/fortran/ChangeLog
@@ -1,3 +1,12 @@
+2017-02-09  Cesar Philippidis  <cesar@codesourcery.com>
+	    Joseph Myers  <joseph@codesourcery.com>
+
+	* openmp.c (resolve_omp_clauses): Error on directives
+	containing both tile and collapse clauses.
+	(resolve_oacc_loop_blocks): Represent '*' tile arguments as zero.
+	* trans-openmp.c (gfc_trans_omp_do): Lower tiled loops like
+	collapsed loops.
+
 2017-02-07  Steven G. Kargl  <kargl@gcc.gnu.org>
 
 	* trans-types.c	(gfc_get_int_kind_from_width_isofortranen):  Choose
diff --git a/gcc/fortran/openmp.c b/gcc/fortran/openmp.c
index d19ee948330..3ca23493251 100644
--- a/gcc/fortran/openmp.c
+++ b/gcc/fortran/openmp.c
@@ -4754,6 +4754,8 @@ resolve_omp_clauses (gfc_code *code, gfc_omp_clauses *omp_clauses,
     if (omp_clauses->wait_list)
       for (el = omp_clauses->wait_list; el; el = el->next)
 	resolve_scalar_int_expr (el->expr, "WAIT");
+  if (omp_clauses->collapse && omp_clauses->tile_list)
+    gfc_error ("Incompatible use of TILE and COLLAPSE at %L", &code->loc);
   if (omp_clauses->depend_source && code->op != EXEC_OMP_ORDERED)
     gfc_error ("SOURCE dependence type only allowed "
 	       "on ORDERED directive at %L", &code->loc);
@@ -5900,11 +5902,11 @@ resolve_oacc_loop_blocks (gfc_code *code)
 	  if (el->expr == NULL)
 	    {
 	      /* NULL expressions are used to represent '*' arguments.
-		 Convert those to a -1 expressions.  */
+		 Convert those to a 0 expressions.  */
 	      el->expr = gfc_get_constant_expr (BT_INTEGER,
 						gfc_default_integer_kind,
 						&code->loc);
-	      mpz_set_si (el->expr->value.integer, -1);
+	      mpz_set_si (el->expr->value.integer, 0);
 	    }
 	  else
 	    {
diff --git a/gcc/fortran/trans-openmp.c b/gcc/fortran/trans-openmp.c
index 4f525feb502..662036f514d 100644
--- a/gcc/fortran/trans-openmp.c
+++ b/gcc/fortran/trans-openmp.c
@@ -3488,6 +3488,17 @@ gfc_trans_omp_do (gfc_code *code, gfc_exec_op op, stmtblock_t *pblock,
   dovar_init *di;
   unsigned ix;
   vec<tree, va_heap, vl_embed> *saved_doacross_steps = doacross_steps;
+  gfc_expr_list *tile = do_clauses ? do_clauses->tile_list : clauses->tile_list;
+
+  /* Both collapsed and tiled loops are lowered the same way.  In
+     OpenACC, those clauses are not compatible, so prioritize the tile
+     clause, if present.  */
+  if (tile)
+    {
+      collapse = 0;
+      for (gfc_expr_list *el = tile; el; el = el->next)
+	collapse++;
+    }
 
   doacross_steps = NULL;
   if (clauses->orderedc)