OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aco: Use s_pack_ll for s_bfe operand on GFX9+.

Browse Source 
Foz-DB Navi21:
Totals from 1 (0.00% of 134913) affected shaders:
CodeSize: 340 -> 336 (-1.18%)
Instrs: 77 -> 76 (-1.30%)
Latency: 1065 -> 1063 (-0.19%)
InvThroughput: 4260 -> 4252 (-0.19%)

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18936>
This commit is contained in:
Georg Lehmann
2022-10-03 11:18:40 +02:00
committed by Marge Bot
parent 1cf1a94f97
commit cc06b7e00d
1 changed files with 6 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/amd/compiler/aco_instruction_selection.cpp
View File

@@ -3686,17 +3686,20 @@ visit_alu_instr(isel_context* ctx, nir_alu_instr* instr)
nir_const_value* const_offset = nir_src_as_const_value(instr->src[1].src); nir_const_value* const_offset = nir_src_as_const_value(instr->src[1].src);
nir_const_value* const_bits = nir_src_as_const_value(instr->src[2].src); nir_const_value* const_bits = nir_src_as_const_value(instr->src[2].src);
aco_opcode opcode =
instr->op == nir_op_ubfe ? aco_opcode::s_bfe_u32 : aco_opcode::s_bfe_i32;
if (const_offset && const_bits) { if (const_offset && const_bits) {
uint32_t extract = (const_bits->u32 << 16) | (const_offset->u32 & 0x1f); uint32_t extract = (const_bits->u32 << 16) | (const_offset->u32 & 0x1f);
aco_opcode opcode =
instr->op == nir_op_ubfe ? aco_opcode::s_bfe_u32 : aco_opcode::s_bfe_i32;
bld.sop2(opcode, Definition(dst), bld.def(s1, scc), base, Operand::c32(extract)); bld.sop2(opcode, Definition(dst), bld.def(s1, scc), base, Operand::c32(extract));
break; break;
} }
Temp offset = get_alu_src(ctx, instr->src[1]); Temp offset = get_alu_src(ctx, instr->src[1]);
Temp bits = get_alu_src(ctx, instr->src[2]); Temp bits = get_alu_src(ctx, instr->src[2]);
if (instr->op == nir_op_ubfe) { if (ctx->program->gfx_level >= GFX9) {
Temp extract = bld.sop2(aco_opcode::s_pack_ll_b32_b16, bld.def(s1), offset, bits);
bld.sop2(opcode, Definition(dst), bld.def(s1, scc), base, extract);
} else if (instr->op == nir_op_ubfe) {
Temp mask = bld.sop2(aco_opcode::s_bfm_b32, bld.def(s1), bits, offset); Temp mask = bld.sop2(aco_opcode::s_bfm_b32, bld.def(s1), bits, offset);
Temp masked = Temp masked =
bld.sop2(aco_opcode::s_and_b32, bld.def(s1), bld.def(s1, scc), base, mask); bld.sop2(aco_opcode::s_and_b32, bld.def(s1), bld.def(s1, scc), base, mask);