aboutsummaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
authorSamuel Pitoiset <[email protected]>2020-05-26 09:38:27 +0200
committerSamuel Pitoiset <[email protected]>2020-06-03 19:48:42 +0200
commitf31c9b4edf6e8d972f26461c49c5e193bf6d9a13 (patch)
tree8ca5ecc3406bbc8b9bf7a6375ead1438f1ccd9ea /src
parenta521c67d223c6e32ea9040643b1bd5a3cdfea04e (diff)
aco: fix subdword copies on GFX6-GFX7
SDWA is only GFX8+. Use v_mov_b32 since the upper 16 bits don't matter. Signed-off-by: Samuel Pitoiset <[email protected]> Reviewed-by: Rhys Perry <[email protected]> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5227>
Diffstat (limited to 'src')
-rw-r--r--src/amd/compiler/aco_builder_h.py18
1 files changed, 11 insertions, 7 deletions
diff --git a/src/amd/compiler/aco_builder_h.py b/src/amd/compiler/aco_builder_h.py
index 2b56e04e501..eb655471c90 100644
--- a/src/amd/compiler/aco_builder_h.py
+++ b/src/amd/compiler/aco_builder_h.py
@@ -382,13 +382,17 @@ public:
} else if (op.bytes() > 2) {
return pseudo(aco_opcode::p_create_vector, dst, op);
} else if (dst.regClass().is_subdword()) {
- aco_ptr<SDWA_instruction> sdwa{create_instruction<SDWA_instruction>(aco_opcode::v_mov_b32, asSDWA(Format::VOP1), 1, 1)};
- sdwa->operands[0] = op;
- sdwa->definitions[0] = dst;
- sdwa->sel[0] = op.bytes() == 1 ? sdwa_ubyte : sdwa_uword;
- sdwa->dst_sel = dst.bytes() == 1 ? sdwa_ubyte : sdwa_uword;
- sdwa->dst_preserve = true;
- return insert(std::move(sdwa));
+ if (program->chip_class >= GFX8) {
+ aco_ptr<SDWA_instruction> sdwa{create_instruction<SDWA_instruction>(aco_opcode::v_mov_b32, asSDWA(Format::VOP1), 1, 1)};
+ sdwa->operands[0] = op;
+ sdwa->definitions[0] = dst;
+ sdwa->sel[0] = op.bytes() == 1 ? sdwa_ubyte : sdwa_uword;
+ sdwa->dst_sel = dst.bytes() == 1 ? sdwa_ubyte : sdwa_uword;
+ sdwa->dst_preserve = true;
+ return insert(std::move(sdwa));
+ } else {
+ return vop1(aco_opcode::v_mov_b32, dst, op);
+ }
} else {
unreachable("Unhandled case in bld.copy()");
}