fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-02-02 15:20:26 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

aco/lower_to_hw: Don't use 2 SGPR operands before GFX10 in a single VOP3 instruction in do_pack_2x16()

Browse source 
Cc: mesa-stable
(cherry picked from commit 9f5996ae8a)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39462>
This commit is contained in:
Daniel Schürmann 2026-01-02 11:16:45 +01:00 • committed by Dylan Baker
parent 5407c3924e
commit fb1bd8bdf4
2 changed files with 11 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -664,7 +664,7 @@
"description": "aco/lower_to_hw: Don't use 2 SGPR operands before GFX10 in a single VOP3 instruction in do_pack_2x16()",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"main_sha": null,
"because_sha": null,
"notes": null

14
src/amd/compiler/aco_lower_to_hw_instr.cpp
View file

@ -1276,6 +1276,7 @@ create_bperm(Builder& bld, uint8_t swiz[4], Definition dst, Operand src1,
else if (!src0.isConstant())
src0 = Operand(PhysReg(src0.physReg().reg()), src0.regClass().resize(4));
assert(src0.isOfType(RegType::vgpr) || src1.isOfType(RegType::vgpr));
bld.vop3(aco_opcode::v_perm_b32, dst, src0, src1, Operand::c32(swiz_packed));
}
@ -1608,14 +1609,18 @@ do_pack_2x16(lower_context* ctx, Builder& bld, Definition def, Operand lo, Opera
return;
}
/* Whether both Operands can be used in a single VOP3 instruction. */
bool both_ops_are_sgpr = lo.isOfType(RegType::sgpr) && hi.isOfType(RegType::sgpr);
bool can_use_vop3 = ctx->program->gfx_level >= GFX10 ||
(!lo.isLiteral() && !hi.isLiteral() && !both_ops_are_sgpr);
/* v_pack_b32_f16 can be used for bit exact copies if:
* - fp16 input denorms are enabled, otherwise they get flushed to zero
* - signalling input NaNs are kept, which is the case with IEEE_MODE=0
* GFX12+ always quiets signalling NaNs, IEEE_MODE was removed
*/
bool can_use_pack = (ctx->block->fp_mode.denorm16_64 & fp_denorm_keep_in) &&
(ctx->program->gfx_level >= GFX10 ||
(ctx->program->gfx_level >= GFX9 && !lo.isLiteral() && !hi.isLiteral())) &&
ctx->program->gfx_level >= GFX9 && can_use_vop3 &&
ctx->program->gfx_level < GFX12;
if (can_use_pack) {
@ -1626,7 +1631,7 @@ do_pack_2x16(lower_context* ctx, Builder& bld, Definition def, Operand lo, Opera
}
/* a single alignbyte can be sufficient: hi can be a 32-bit integer constant */
if (lo.physReg().byte() == 2 && hi.physReg().byte() == 0 &&
if (lo.physReg().byte() == 2 && hi.physReg().byte() == 0 && can_use_vop3 &&
(!hi.isConstant() || (hi.constantValue() && (!Operand::c32(hi.constantValue()).isLiteral() ||
ctx->program->gfx_level >= GFX10)))) {
if (hi.isConstant())
@ -1637,7 +1642,8 @@ do_pack_2x16(lower_context* ctx, Builder& bld, Definition def, Operand lo, Opera
return;
}
if (ctx->program->gfx_level >= GFX10 && !lo.constantEquals(0) && !hi.constantEquals(0)) {
if (ctx->program->gfx_level >= GFX10 && !lo.constantEquals(0) && !hi.constantEquals(0) &&
!both_ops_are_sgpr) {
uint8_t swiz[4];
Operand ops[2] = {lo, hi};
for (unsigned i = 0; i < 2; i++) {