fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-24 17:30:12 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

aco: Combine bit test to s_bitcmp.

Browse source 
Foz-DB Navi21:
Totals from 6396 (4.74% of 134913) affected shaders:
VGPRs: 483280 -> 483152 (-0.03%); split: -0.03%, +0.01%
SpillSGPRs: 8119 -> 7941 (-2.19%)
CodeSize: 63377880 -> 63268556 (-0.17%); split: -0.20%, +0.03%
MaxWaves: 86778 -> 86810 (+0.04%)
Instrs: 11745621 -> 11725857 (-0.17%); split: -0.20%, +0.03%
Latency: 162400148 -> 162282230 (-0.07%); split: -0.08%, +0.01%
InvThroughput: 29179429 -> 29133173 (-0.16%); split: -0.16%, +0.00%
VClause: 208032 -> 208100 (+0.03%); split: -0.01%, +0.05%
SClause: 431390 -> 430849 (-0.13%); split: -0.24%, +0.11%
Copies: 896222 -> 893285 (-0.33%); split: -0.62%, +0.30%
Branches: 349806 -> 348770 (-0.30%); split: -0.90%, +0.60%
PreSGPRs: 618908 -> 613773 (-0.83%); split: -0.83%, +0.00%
PreVGPRs: 482901 -> 482893 (-0.00%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18870>
This commit is contained in:
Georg Lehmann 2022-09-24 18:49:10 +02:00 • committed by Marge Bot
parent 853d2cb6f1
commit 73be938c48
1 changed files with 69 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

69
src/amd/compiler/aco_optimizer.cpp
View file

@ -2810,6 +2810,68 @@ use_absdiff:
return true;
}
/* s_cmp_{lg,eq}(s_and(a, s_lshl(1, b)), 0) -> s_bitcmp[10](a, b)*/
bool
combine_s_bitcmp(opt_ctx& ctx, aco_ptr<Instruction>& instr)
{
bool lg = false;
bool b64 = false;
switch (instr->opcode) {
case aco_opcode::s_cmp_lg_i32:
case aco_opcode::s_cmp_lg_u32: lg = true; break;
case aco_opcode::s_cmp_eq_i32:
case aco_opcode::s_cmp_eq_u32: break;
case aco_opcode::s_cmp_lg_u64: lg = true; FALLTHROUGH;
case aco_opcode::s_cmp_eq_u64: b64 = true; break;
default: return false;
}
aco_opcode s_and = b64 ? aco_opcode::s_and_b64 : aco_opcode::s_and_b32;
aco_opcode s_lshl = b64 ? aco_opcode::s_lshl_b64 : aco_opcode::s_lshl_b32;
for (unsigned cmp_idx = 0; cmp_idx < 2; cmp_idx++) {
Instruction* and_instr = follow_operand(ctx, instr->operands[cmp_idx], false);
if (!and_instr || and_instr->opcode != s_and)
continue;
for (unsigned and_idx = 0; and_idx < 2; and_idx++) {
Instruction* lshl_instr = follow_operand(ctx, and_instr->operands[and_idx], true);
if (!lshl_instr || lshl_instr->opcode != s_lshl ||
!lshl_instr->operands[0].constantEquals(1) ||
(lshl_instr->operands[1].isLiteral() && and_instr->operands[!and_idx].isLiteral()))
continue;
bool test1 = false;
if (instr->operands[!cmp_idx].constantEquals(0)) {
test1 = lg;
} else if (instr->operands[!cmp_idx].isTemp() &&
instr->operands[!cmp_idx].tempId() == lshl_instr->definitions[0].tempId()) {
test1 = !lg;
ctx.uses[lshl_instr->definitions[0].tempId()]--;
} else {
continue;
}
if (test1 && b64)
instr->opcode = aco_opcode::s_bitcmp1_b64;
else if (!test1 && b64)
instr->opcode = aco_opcode::s_bitcmp0_b64;
else if (test1 && !b64)
instr->opcode = aco_opcode::s_bitcmp1_b32;
else
instr->opcode = aco_opcode::s_bitcmp0_b32;
instr->operands[0] = copy_operand(ctx, and_instr->operands[!and_idx]);
instr->operands[1] = copy_operand(ctx, lshl_instr->operands[1]);
decrease_uses(ctx, and_instr);
decrease_op_uses_if_dead(ctx, lshl_instr);
return true;
}
}
return false;
}
bool
combine_add_sub_b2i(opt_ctx& ctx, aco_ptr<Instruction>& instr, aco_opcode new_op, uint8_t ops)
{
@ -4226,6 +4288,13 @@ combine_instruction(opt_ctx& ctx, aco_ptr<Instruction>& instr)
}
} else if (instr->opcode == aco_opcode::s_abs_i32) {
combine_sabsdiff(ctx, instr);
} else if (instr->opcode == aco_opcode::s_cmp_lg_i32 ||
instr->opcode == aco_opcode::s_cmp_lg_u32 ||
instr->opcode == aco_opcode::s_cmp_lg_u64 ||
instr->opcode == aco_opcode::s_cmp_eq_i32 ||
instr->opcode == aco_opcode::s_cmp_eq_u32 ||
instr->opcode == aco_opcode::s_cmp_eq_u64) {
combine_s_bitcmp(ctx, instr);
} else if (instr->opcode == aco_opcode::v_and_b32) {
combine_and_subbrev(ctx, instr);
} else if (instr->opcode == aco_opcode::v_fma_f32 || instr->opcode == aco_opcode::v_fma_f16) {