fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-14 05:48:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

aco: optimize v_pk_fma_f16 -> v_pk_fmac_f16 on GFX10

Browse source 
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6680>
(cherry picked from commit 6ecbccfb23)
This commit is contained in:
Daniel Schürmann 2020-09-16 10:32:29 +01:00 • committed by Dylan Baker
parent 4562f9f894
commit 71a58f02e5
2 changed files with 7 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -292,7 +292,7 @@
"description": "aco: optimize v_pk_fma_f16 -> v_pk_fmac_f16 on GFX10",
"nominated": false,
"nomination_type": null,
"resolution": 4,
"resolution": 1,
"master_sha": null,
"because_sha": null
},

7
src/amd/compiler/aco_register_allocation.cpp
View file

@ -2083,7 +2083,8 @@ void register_allocation(Program *program, std::vector<IDSet>& live_out_per_bloc
(instr->opcode == aco_opcode::v_fma_f32 && program->chip_class >= GFX10) ||
instr->opcode == aco_opcode::v_mad_f16 ||
instr->opcode == aco_opcode::v_mad_legacy_f16 ||
(instr->opcode == aco_opcode::v_fma_f16 && program->chip_class >= GFX10)) &&
(instr->opcode == aco_opcode::v_fma_f16 && program->chip_class >= GFX10) ||
(instr->opcode == aco_opcode::v_pk_fma_f16 && program->chip_class >= GFX10)) &&
instr->operands[2].isTemp() &&
instr->operands[2].isKillBeforeDef() &&
instr->operands[2].getTemp().type() == RegType::vgpr &&
@ -2113,6 +2114,9 @@ void register_allocation(Program *program, std::vector<IDSet>& live_out_per_bloc
case aco_opcode::v_fma_f16:
instr->opcode = aco_opcode::v_fmac_f16;
break;
case aco_opcode::v_pk_fma_f16:
instr->opcode = aco_opcode::v_pk_fmac_f16;
break;
default:
break;
}
@ -2125,6 +2129,7 @@ void register_allocation(Program *program, std::vector<IDSet>& live_out_per_bloc
instr->opcode == aco_opcode::v_fmac_f32 ||
instr->opcode == aco_opcode::v_mac_f16 ||
instr->opcode == aco_opcode::v_fmac_f16 ||
instr->opcode == aco_opcode::v_pk_fmac_f16 ||
instr->opcode == aco_opcode::v_writelane_b32 ||
instr->opcode == aco_opcode::v_writelane_b32_e64) {
instr->definitions[0].setFixed(instr->operands[2].physReg());