fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-02 05:10:17 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

aco: don't set exec_hi for wave32 scan reductions

Browse source 
fossil-db (wave32):
Totals from 21 (0.02% of 133428) affected shaders:
Instrs: 10778 -> 10712 (-0.61%)
CodeSize: 56604 -> 56208 (-0.70%)
Latency: 168293 -> 168251 (-0.02%)
InvThroughput: 25256 -> 25253 (-0.01%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23745>
This commit is contained in:
Rhys Perry 2023-06-20 14:31:55 +01:00 • committed by Marge Bot
parent 5d03bbc91d
commit cfa7eec06c
1 changed files with 6 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/amd/compiler/aco_lower_to_hw_instr.cpp
View file

@ -649,7 +649,8 @@ emit_reduction(lower_context* ctx, aco_opcode op, ReduceOp reduce_op, unsigned c
/* fill in the gaps in rows 1 and 3 */
bld.sop1(aco_opcode::s_mov_b32, Definition(exec_lo, s1), Operand::c32(0x10000u));
bld.sop1(aco_opcode::s_mov_b32, Definition(exec_hi, s1), Operand::c32(0x10000u));
if (ctx->program->wave_size == 64)
bld.sop1(aco_opcode::s_mov_b32, Definition(exec_hi, s1), Operand::c32(0x10000u));
for (unsigned i = 0; i < src.size(); i++) {
Instruction* perm =
bld.vop3(aco_opcode::v_permlanex16_b32, Definition(PhysReg{vtmp + i}, v1),
@ -782,8 +783,10 @@ emit_reduction(lower_context* ctx, aco_opcode op, ReduceOp reduce_op, unsigned c
if (ctx->program->gfx_level >= GFX10) {
bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_lo, s1), Operand::c32(16u),
Operand::c32(16u));
bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_hi, s1), Operand::c32(16u),
Operand::c32(16u));
if (ctx->program->wave_size == 64) {
bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_hi, s1), Operand::c32(16u),
Operand::c32(16u));
}
for (unsigned i = 0; i < src.size(); i++) {
Instruction* perm =
bld.vop3(aco_opcode::v_permlanex16_b32, Definition(PhysReg{vtmp + i}, v1),