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

aco: use v_add_f{16,32} with clamp for fsat

Browse Source 
v_add can be dual issued on gfx11, v_med3 cannot.
Don't use v_add directly to still optimize omod(fsat(x)).

Foz-DB GFX1100:
Totals from 32702 (24.24% of 134913) affected shaders:
Latency: 475008203 -> 474928037 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 59226198 -> 59140787 (-0.14%); split: -0.14%, +0.00%

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
2023-02-22 17:31:06 +01:00
committed by Marge Bot
parent 3a0bc8f007
commit 177dba62a1
3 changed files with 47 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_optimizer.cpp
+42 -16
View File
@@ -1296,6 +1296,33 @@ is_scratch_offset_valid(opt_ctx& ctx, Instruction* instr, int64_t offset0, int64
return offset >= min && offset <= max;
}
bool
detect_clamp(Instruction* instr, unsigned* clamped_idx)
{
VALU_instruction& valu = instr->valu();
if (valu.omod != 0 || valu.opsel != 0)
return false;
unsigned idx = 0;
bool found_zero = false, found_one = false;
bool is_fp16 = instr->opcode == aco_opcode::v_med3_f16;
for (unsigned i = 0; i < 3; i++) {
if (!valu.neg[i] && instr->operands[i].constantEquals(0))
found_zero = true;
else if (!valu.neg[i] &&
instr->operands[i].constantEquals(is_fp16 ? 0x3c00 : 0x3f800000)) /* 1.0 */
found_one = true;
else
idx = i;
}
if (found_zero && found_one && instr->operands[idx].isTemp()) {
*clamped_idx = idx;
return true;
} else {
return false;
}
}
void
label_instruction(opt_ctx& ctx, aco_ptr<Instruction>& instr)
{
@@ -1882,22 +1909,8 @@ label_instruction(opt_ctx& ctx, aco_ptr<Instruction>& instr)
break;
case aco_opcode::v_med3_f16:
case aco_opcode::v_med3_f32: { /* clamp */
VALU_instruction& vop3 = instr->valu();
if (vop3.abs != 0 || vop3.neg != 0 || vop3.omod != 0 || vop3.opsel != 0)
break;
unsigned idx = 0;
bool found_zero = false, found_one = false;
bool is_fp16 = instr->opcode == aco_opcode::v_med3_f16;
for (unsigned i = 0; i < 3; i++) {
if (instr->operands[i].constantEquals(0))
found_zero = true;
else if (instr->operands[i].constantEquals(is_fp16 ? 0x3c00 : 0x3f800000)) /* 1.0 */
found_one = true;
else
idx = i;
}
if (found_zero && found_one && instr->operands[idx].isTemp())
unsigned idx;
if (detect_clamp(instr.get(), &idx) && !instr->valu().abs && !instr->valu().neg)
ctx.info[instr->operands[idx].tempId()].set_clamp(instr.get());
break;
}
@@ -4503,6 +4516,19 @@ combine_instruction(opt_ctx& ctx, aco_ptr<Instruction>& instr)
*/
ctx.mad_infos.emplace_back(nullptr, 0);
ctx.info[instr->definitions[0].tempId()].set_mad(ctx.mad_infos.size() - 1);
} else if (instr->opcode == aco_opcode::v_med3_f32 || instr->opcode == aco_opcode::v_med3_f16) {
unsigned idx;
if (detect_clamp(instr.get(), &idx)) {
instr->format = asVOP3(Format::VOP2);
instr->operands[0] = instr->operands[idx];
instr->operands[1] = Operand::zero();
instr->opcode =
instr->opcode == aco_opcode::v_med3_f32 ? aco_opcode::v_add_f32 : aco_opcode::v_add_f16;
instr->valu().clamp = true;
instr->valu().abs = (uint8_t)instr->valu().abs[idx];
instr->valu().neg = (uint8_t)instr->valu().neg[idx];
instr->operands.pop_back();
}
} else {
aco_opcode min, max, min3, max3, med3, minmax;
bool some_gfx9_only;