AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aco/optimizer: fix signed extract of sub dword temps with SDWA

Browse Source 
If an instruction didn't already use SDWA convert_to_SDWA in apply_extract
will add ubyte0/uword0 selections for v1b/v2b operands. This loses information
that the instruction doesn't care about the high bits and makes the next
apply_extract_twice fail.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>

Fixes: 6cb9d39bc2 ("aco: combine extracts with sub-dword definitions")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32803>
This commit is contained in:
Georg Lehmann
2024-12-30 16:13:25 +01:00
committed by Marge Bot
parent 346f4d3c11
commit 3da2d96bc5
2 changed files with 24 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_optimizer.cpp
+7 -3
View File
@@ -1149,9 +1149,13 @@ apply_extract(opt_ctx& ctx, aco_ptr<Instruction>& instr, unsigned idx, ssa_info&
instr.reset(mad);
} else if (can_use_SDWA(ctx.program->gfx_level, instr, true) &&
(tmp.type() == RegType::vgpr || ctx.program->gfx_level >= GFX9)) {
convert_to_SDWA(ctx.program->gfx_level, instr);
instr->sdwa().sel[idx] = apply_extract_twice(sel, instr->operands[idx].getTemp(),
instr->sdwa().sel[idx], Temp(0, v1));
if (instr->isSDWA()) {
instr->sdwa().sel[idx] = apply_extract_twice(sel, instr->operands[idx].getTemp(),
instr->sdwa().sel[idx], Temp(0, v1));
} else {
convert_to_SDWA(ctx.program->gfx_level, instr);
instr->sdwa().sel[idx] = sel;
}
} else if (instr->isVALU()) {
if (sel.offset()) {
instr->valu().opsel[idx] = true;