fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-06 13:10:10 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco/statistics: improve v_fma_mix dual issuing detection

Browse source 
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21402>
This commit is contained in:
Georg Lehmann 2023-05-11 17:42:12 +02:00 • committed by Marge Bot
parent 0f023d90c0
commit 3a0bc8f007
1 changed files with 21 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

27
src/amd/compiler/aco_statistics.cpp
View file

@ -105,12 +105,12 @@ struct perf_info {
};
static bool
is_dual_issue_capable(const Program& program, const Instruction& instruction)
is_dual_issue_capable(const Program& program, const Instruction& instr)
{
if (program.gfx_level < GFX11 || !instruction.isVALU() || instruction.isDPP())
if (program.gfx_level < GFX11 || !instr.isVALU() || instr.isDPP())
return false;
switch (instruction.opcode) {
switch (instr.opcode) {
case aco_opcode::v_fma_f32:
case aco_opcode::v_fmac_f32:
case aco_opcode::v_fmaak_f32:
@ -122,9 +122,6 @@ is_dual_issue_capable(const Program& program, const Instruction& instruction)
case aco_opcode::v_mul_legacy_f32:
case aco_opcode::v_fma_legacy_f32:
case aco_opcode::v_fmac_legacy_f32:
case aco_opcode::v_fma_mix_f32:
case aco_opcode::v_fma_mixlo_f16:
case aco_opcode::v_fma_mixhi_f16:
case aco_opcode::v_fma_f16:
case aco_opcode::v_fmac_f16:
case aco_opcode::v_fmaak_f16:
@ -169,6 +166,24 @@ is_dual_issue_capable(const Program& program, const Instruction& instruction)
case aco_opcode::v_dot2_f16_f16:
case aco_opcode::v_dot2_f32_f16:
case aco_opcode::v_dot2c_f32_f16: return true;
case aco_opcode::v_fma_mix_f32:
case aco_opcode::v_fma_mixlo_f16:
case aco_opcode::v_fma_mixhi_f16: {
/* dst and acc type must match */
if (instr.valu().opsel_hi[2] == (instr.opcode == aco_opcode::v_fma_mix_f32))
return false;
/* If all operands are vgprs, two must be the same. */
for (unsigned i = 0; i < 3; i++) {
if (instr.operands[i].isConstant() || instr.operands[i].isOfType(RegType::sgpr))
return true;
for (unsigned j = 0; j < i; j++) {
if (instr.operands[i].physReg() == instr.operands[j].physReg())
return true;
}
}
return false;
}
default: return false;
}
}