diff options
| author | Danylo Piliaiev <dpiliaiev@igalia.com> | 2021-06-16 14:43:19 +0300 |
|---|---|---|
| committer | Marge Bot <eric+marge@anholt.net> | 2021-06-25 15:39:51 +0000 |
| commit | fdc0f489e098d320593a1c6837a19726c84d90e9 (patch) | |
| tree | be5a34ed76eb7ae31b475ca8a6508a33999c398f /src/freedreno/ir3/tests/disasm.c | |
| parent | 4b06db0548603a719c836070b7b8959bce45a2c2 (diff) | |
| download | mesa-fdc0f489e098d320593a1c6837a19726c84d90e9.tar.gz | |
ir3: add ldg.a,stg.a which allow complex in-place offset calculation
The full form for ldg.a/stg.a offset is:
g[reg_address + reg_offset << (imm_shift + 2) + imm_offset << 2]
where imm_shift is in [0, 3] and imm_offset is in [0, 3]
a6xx blob was found to produce a bit simplier offset calculations
for TES/TCS shaders in GTA V:
[c002000a_03c14215] ldg.a.f32 r2.z, g[r1.y+((r2.z+1)<<2)], 3;
[c0020004_01c14609] ldg.a.f32 r1.x, g[r1.y+((r1.x+3)<<2)], 1;
Our new syntax:
stg.a.u32 g[r2.x+(r1.x+1)<<2], r5.x, 1
stg.a.u32 g[r2.x+r1.x<<4+3<<2], r5.x, 1
ldg.a.f32 r1.w, g[r1.y+(r1.w+1)<<2], 3
ldg.a.f32 r1.w, g[r1.y+r1.w<<5+2<<2], 3
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11431>
Diffstat (limited to 'src/freedreno/ir3/tests/disasm.c')
| -rw-r--r-- | src/freedreno/ir3/tests/disasm.c | 22 |
1 files changed, 16 insertions, 6 deletions
diff --git a/src/freedreno/ir3/tests/disasm.c b/src/freedreno/ir3/tests/disasm.c index 6f500f0253e..31fcf97bf4a 100644 --- a/src/freedreno/ir3/tests/disasm.c +++ b/src/freedreno/ir3/tests/disasm.c @@ -153,20 +153,30 @@ static const struct test { // TODO is this a real instruction? Or float -6.0 ? // INSTR_6XX(c0c00000_00000000, "stg.f16 g[hr0.x], hr0.x, hr0.x", .parse_fail=true), /* dEQP-GLES31.functional.tessellation.invariance.outer_edge_symmetry.isolines_equal_spacing_ccw */ - INSTR_6XX(c0d20906_02800004, "stg.f32 g[r1.x+r1.z], r0.z, 2"), /* stg.a.f32 g[r1.x+(r1.z<<2)], r0.z, 2 */ - INSTR_6XX(c0da052e_01800042, "stg.s32 g[r0.z+r11.z], r8.y, 1"), /* stg.a.s32 g[r0.z+(r11.z<<2)], r8.y, 1 */ + INSTR_6XX(c0d20906_02800004, "stg.a.f32 g[r1.x+(r1.z)<<2], r0.z, 2"), /* stg.a.f32 g[r1.x+(r1.z<<2)], r0.z, 2 */ + INSTR_6XX(c0da052e_01800042, "stg.a.s32 g[r0.z+(r11.z)<<2], r8.y, 1"), /* stg.a.s32 g[r0.z+(r11.z<<2)], r8.y, 1 */ INSTR_6XX(c0ca0505_03800042, "stg.s32 g[r0.z+5], r8.y, 3"), INSTR_6XX(c0ca0500_03800042, "stg.s32 g[r0.z], r8.y, 3"), INSTR_6XX(c0ca0531_03800242, "stg.s32 g[r0.z+305], r8.y, 3"), - INSTR_6XX(c0020011_04c08023, "ldg.f32 r4.y, g[r0.z+r4.y], 4"), /* ldg.a.f32 r4.y, g[r0.z+(r4.y<<2)], 4 */ - INSTR_6XX(c0060006_01c18017, "ldg.u32 r1.z, g[r1.z+r2.w], 1"), /* ldg.a.u32 r1.z, g[r1.z+(r2.w<<2)], 1 */ + /* Customely crafted */ + INSTR_6XX(c0d61104_01800228, "stg.a.u32 g[r2.x+(r1.x+1)<<2], r5.x, 1"), + INSTR_6XX(c0d61104_01802628, "stg.a.u32 g[r2.x+r1.x<<4+3<<2], r5.x, 1"), + + INSTR_6XX(c0020011_04c08023, "ldg.a.f32 r4.y, g[r0.z+(r4.y)<<2], 4"), /* ldg.a.f32 r4.y, g[r0.z+(r4.y<<2)], 4 */ + INSTR_6XX(c0060006_01c18017, "ldg.a.u32 r1.z, g[r1.z+(r2.w)<<2], 1"), /* ldg.a.u32 r1.z, g[r1.z+(r2.w<<2)], 1 */ INSTR_6XX(c0060006_0181800f, "ldg.u32 r1.z, g[r1.z+7], 1"), INSTR_6XX(c0060006_01818001, "ldg.u32 r1.z, g[r1.z], 1"), INSTR_6XX(c0060003_0180c269, "ldg.u32 r0.w, g[r0.w+308], 1"), - INSTR_6XX(c0020011_04c08023, "ldg.f32 r4.y, g[r0.z+r4.y], 4"), /* ldg.a.f32 r4.y, g[r0.z+(r4.y<<2)], 4 */ - INSTR_6XX(c0060006_01c18017, "ldg.u32 r1.z, g[r1.z+r2.w], 1"), /* ldg.a.u32 r1.z, g[r1.z+(r2.w<<2)], 1 */ + /* Found in TCS/TES shaders of GTA V */ + INSTR_6XX(c0020007_03c1420f, "ldg.a.f32 r1.w, g[r1.y+(r1.w+1)<<2], 3"), /* ldg.a.f32 r1.w, g[r1.y+((r1.w+1)<<2)], 3 */ + + /* Customely crafted */ + INSTR_6XX(c0020007_03c1740f, "ldg.a.f32 r1.w, g[r1.y+r1.w<<5+2<<2], 3"), + + INSTR_6XX(c0020011_04c08023, "ldg.a.f32 r4.y, g[r0.z+(r4.y)<<2], 4"), /* ldg.a.f32 r4.y, g[r0.z+(r4.y<<2)], 4 */ + INSTR_6XX(c0060006_01c18017, "ldg.a.u32 r1.z, g[r1.z+(r2.w)<<2], 1"), /* ldg.a.u32 r1.z, g[r1.z+(r2.w<<2)], 1 */ INSTR_6XX(c0060006_0181800f, "ldg.u32 r1.z, g[r1.z+7], 1"), INSTR_6XX(c0060006_01818001, "ldg.u32 r1.z, g[r1.z], 1"), |