fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-25 06:08:21 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: fix packed 16bit fneg/fsat optimization

Browse source 
Make sure that the Operand is '1.0.xx'.

Fixes: b03be30e07 ('aco: optimize packed fneg')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17395>
(cherry picked from commit 66d46a23fb)
This commit is contained in:
Daniel Schürmann 2022-07-07 12:27:08 +02:00 • committed by Dylan Baker
parent 8a44563486
commit 9276b2f2f3
2 changed files with 9 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -472,7 +472,7 @@
"description": "aco: fix packed 16bit fneg/fsat optimization",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"main_sha": null,
"because_sha": "b03be30e07546b813acc1156407b964e68892beb"
},

10
src/amd/compiler/aco_optimizer.cpp
View file

@ -3326,7 +3326,8 @@ combine_vop3p(opt_ctx& ctx, aco_ptr<Instruction>& instr)
/* apply clamp */
if (instr->opcode == aco_opcode::v_pk_mul_f16 && instr->operands[1].constantEquals(0x3C00) &&
vop3p->clamp && instr->operands[0].isTemp() && ctx.uses[instr->operands[0].tempId()] == 1) {
vop3p->clamp && instr->operands[0].isTemp() && ctx.uses[instr->operands[0].tempId()] == 1 &&
!((vop3p->opsel_lo | vop3p->opsel_hi) & 2)) {
ssa_info& info = ctx.info[instr->operands[0].tempId()];
if (info.is_vop3p() && instr_info.can_use_output_modifiers[(int)info.instr->opcode]) {
@ -3350,6 +3351,12 @@ combine_vop3p(opt_ctx& ctx, aco_ptr<Instruction>& instr)
ssa_info& info = ctx.info[op.tempId()];
if (info.is_vop3p() && info.instr->opcode == aco_opcode::v_pk_mul_f16 &&
info.instr->operands[1].constantEquals(0x3C00)) {
VOP3P_instruction* fneg = &info.instr->vop3p();
if ((fneg->opsel_lo | fneg->opsel_hi) & 2)
continue;
Operand ops[3];
for (unsigned j = 0; j < instr->operands.size(); j++)
ops[j] = instr->operands[j];
@ -3357,7 +3364,6 @@ combine_vop3p(opt_ctx& ctx, aco_ptr<Instruction>& instr)
if (!check_vop3_operands(ctx, instr->operands.size(), ops))
continue;
VOP3P_instruction* fneg = &info.instr->vop3p();
if (fneg->clamp)
continue;
instr->operands[i] = fneg->operands[0];