fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-10 08:10:14 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

ac/llvm: fix umul_high

Browse source 
LLVM optimizes umul_hi with a constant to v_mul_hi_i32_i24_e32 which isn't
always what we need here. This causes miscalculations. To prevent LLVM to
apply this optimization, we insert a optimization barrier.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11761
Suggested-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30810>
This commit is contained in:
Karol Herbst 2024-08-22 19:39:46 +02:00 • committed by Marge Bot
parent 2162cf7476
commit 74dafa3c79
1 changed files with 8 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
src/amd/llvm/ac_nir_to_llvm.c
View file

@ -333,6 +333,14 @@ static LLVMValueRef emit_umul_high(struct ac_llvm_context *ctx, LLVMValueRef src
LLVMValueRef src1)
{
LLVMValueRef dst64, result;
#if LLVM_VERSION_MAJOR < 20
if (LLVMIsConstant(src0))
ac_build_optimization_barrier(ctx, &src1, false);
else
ac_build_optimization_barrier(ctx, &src0, false);
#endif
src0 = LLVMBuildZExt(ctx->builder, src0, ctx->i64, "");
src1 = LLVMBuildZExt(ctx->builder, src1, ctx->i64, "");