fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-04 22:49:13 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

aco/gfx6: allow vgpr offset for global access with sgpr address

Browse source 
No reason why we can't use offen like normal buffer loads.

fossil-db (pitcairn):
Totals from 122 (0.20% of 62069) affected shaders:
MaxWaves: 521 -> 525 (+0.77%)
Instrs: 238341 -> 237228 (-0.47%); split: -0.57%, +0.10%
CodeSize: 1196260 -> 1188076 (-0.68%); split: -0.78%, +0.09%
SGPRs: 8752 -> 8760 (+0.09%); split: -0.64%, +0.73%
VGPRs: 10456 -> 10440 (-0.15%); split: -0.88%, +0.73%
Latency: 3958385 -> 3946186 (-0.31%); split: -0.38%, +0.07%
InvThroughput: 3097193 -> 3084417 (-0.41%); split: -0.42%, +0.01%
VClause: 4058 -> 4500 (+10.89%); split: -0.02%, +10.92%
SClause: 4511 -> 4500 (-0.24%); split: -0.42%, +0.18%
Copies: 31228 -> 31718 (+1.57%); split: -0.38%, +1.95%
PreSGPRs: 7211 -> 7461 (+3.47%)
PreVGPRs: 8174 -> 8147 (-0.33%); split: -0.34%, +0.01%
VALU: 174779 -> 173294 (-0.85%); split: -0.87%, +0.02%
SALU: 29138 -> 29641 (+1.73%); split: -0.09%, +1.82%

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35465>
This commit is contained in:
Rhys Perry 2025-06-02 14:53:02 +01:00 • committed by Marge Bot
parent 09a5af121f
commit 684943bd1f
1 changed files with 29 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

39
src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp
View file

@ -855,10 +855,11 @@ lower_global_address(Builder& bld, uint32_t offset_in, Temp* address_inout,
}
if (bld.program->gfx_level == GFX6) {
/* GFX6 (MUBUF): (SGPR address, SGPR offset) or (VGPR address, SGPR offset) */
/* GFX6 (MUBUF): (SGPR address, SGPR offset) or (SGPR address, VGPR offset) */
/* GFX6 (MUBUF-addr64): (VGPR address, SGPR offset) */
/* Disallow SGPR address with both a const_offset and offset because of possible overflow. */
if (offset.id() && (offset.type() != RegType::sgpr ||
(address.type() == RegType::sgpr && const_offset > 0))) {
if (offset.id() &&
(address.type() == RegType::vgpr ? offset.type() != RegType::sgpr : const_offset > 0)) {
address = add64_32(bld, address, Operand(offset));
offset = Temp();
}
@ -937,8 +938,14 @@ global_load_callback(Builder& bld, const LoadEmitInfo& info, unsigned bytes_need
if (use_mubuf) {
aco_ptr<Instruction> mubuf{create_instruction(op, Format::MUBUF, 3, 1)};
mubuf->operands[0] = Operand(get_gfx6_global_rsrc(bld, addr));
mubuf->operands[1] = addr.type() == RegType::vgpr ? Operand(addr) : Operand(v1);
mubuf->operands[2] = Operand(offset);
if (addr.type() == RegType::vgpr)
mubuf->operands[1] = Operand(addr);
else if (offset.type() == RegType::vgpr)
mubuf->operands[1] = Operand(offset);
else
mubuf->operands[1] = Operand(v1);
mubuf->operands[2] = offset.type() == RegType::sgpr ? Operand(offset) : Operand::c32(0);
mubuf->mubuf().offen = offset.type() == RegType::vgpr;
mubuf->mubuf().cache = info.cache;
mubuf->mubuf().offset = const_offset;
mubuf->mubuf().addr64 = addr.type() == RegType::vgpr;
@ -2564,10 +2571,16 @@ visit_store_global(isel_context* ctx, nir_intrinsic_instr* instr)
aco_ptr<Instruction> mubuf{create_instruction(op, Format::MUBUF, 4, 0)};
mubuf->operands[0] = Operand(rsrc);
mubuf->operands[1] =
write_address.type() == RegType::vgpr ? Operand(write_address) : Operand(v1);
mubuf->operands[2] = Operand(write_offset);
if (write_address.type() == RegType::vgpr)
mubuf->operands[1] = Operand(write_address);
else if (write_offset.type() == RegType::vgpr)
mubuf->operands[1] = Operand(write_offset);
else
mubuf->operands[1] = Operand(v1);
mubuf->operands[2] =
write_offset.type() == RegType::sgpr ? Operand(write_offset) : Operand::c32(0);
mubuf->operands[3] = Operand(write_datas[i]);
mubuf->mubuf().offen = write_offset.type() == RegType::vgpr;
mubuf->mubuf().cache = get_cache_flags(ctx, access);
mubuf->mubuf().offset = write_const_offset;
mubuf->mubuf().addr64 = write_address.type() == RegType::vgpr;
@ -2699,13 +2712,19 @@ visit_global_atomic(isel_context* ctx, nir_intrinsic_instr* instr)
aco_ptr<Instruction> mubuf{create_instruction(op, Format::MUBUF, 4, return_previous ? 1 : 0)};
mubuf->operands[0] = Operand(rsrc);
mubuf->operands[1] = addr.type() == RegType::vgpr ? Operand(addr) : Operand(v1);
mubuf->operands[2] = Operand(offset);
if (addr.type() == RegType::vgpr)
mubuf->operands[1] = Operand(addr);
else if (offset.type() == RegType::vgpr)
mubuf->operands[1] = Operand(offset);
else
mubuf->operands[1] = Operand(v1);
mubuf->operands[2] = offset.type() == RegType::sgpr ? Operand(offset) : Operand::c32(0);
mubuf->operands[3] = Operand(data);
Definition def =
return_previous ? (cmpswap ? bld.def(data.regClass()) : Definition(dst)) : Definition();
if (return_previous)
mubuf->definitions[0] = def;
mubuf->mubuf().offen = offset.type() == RegType::vgpr;
mubuf->mubuf().cache = get_atomic_cache_flags(ctx, return_previous);
mubuf->mubuf().offset = const_offset;
mubuf->mubuf().addr64 = addr.type() == RegType::vgpr;