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

aco: emit float8 wmma

Browse Source 
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35434>
This commit is contained in:
Georg Lehmann
2025-04-09 14:04:18 +02:00
committed by Marge Bot
parent a1b88fcd8c
commit 65650cfef8
2 changed files with 23 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_optimizer.cpp
+9 -1
View File
@@ -467,7 +467,11 @@ can_apply_sgprs(opt_ctx& ctx, aco_ptr<Instruction>& instr)
instr->opcode != aco_opcode::v_wmma_f16_16x16x16_f16 &&
instr->opcode != aco_opcode::v_wmma_bf16_16x16x16_bf16 &&
instr->opcode != aco_opcode::v_wmma_i32_16x16x16_iu8 &&
instr->opcode != aco_opcode::v_wmma_i32_16x16x16_iu4;
instr->opcode != aco_opcode::v_wmma_i32_16x16x16_iu4 &&
instr->opcode != aco_opcode::v_wmma_f32_16x16x16_fp8_fp8 &&
instr->opcode != aco_opcode::v_wmma_f32_16x16x16_fp8_bf8 &&
instr->opcode != aco_opcode::v_wmma_f32_16x16x16_bf8_fp8 &&
instr->opcode != aco_opcode::v_wmma_f32_16x16x16_bf8_bf8;
}
/* only covers special cases */
@@ -529,6 +533,10 @@ alu_can_accept_constant(const aco_ptr<Instruction>& instr, unsigned operand)
case aco_opcode::v_dot2_bf16_bf16: /* TODO */
case aco_opcode::v_wmma_f32_16x16x16_f16:
case aco_opcode::v_wmma_f32_16x16x16_bf16:
case aco_opcode::v_wmma_f32_16x16x16_fp8_fp8:
case aco_opcode::v_wmma_f32_16x16x16_fp8_bf8:
case aco_opcode::v_wmma_f32_16x16x16_bf8_fp8:
case aco_opcode::v_wmma_f32_16x16x16_bf8_bf8:
case aco_opcode::v_wmma_f16_16x16x16_f16:
case aco_opcode::v_wmma_bf16_16x16x16_bf16:
case aco_opcode::v_wmma_i32_16x16x16_iu8:
src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp
+14
View File
@@ -3762,6 +3762,20 @@ visit_cmat_muladd(isel_context* ctx, nir_intrinsic_instr* instr)
neg_lo[0] = type_a == GLSL_TYPE_INT8;
neg_lo[1] = type_b == GLSL_TYPE_INT8;
break;
case GLSL_TYPE_FLOAT_E4M3FN:
switch (type_b) {
case GLSL_TYPE_FLOAT_E4M3FN: opcode = aco_opcode::v_wmma_f32_16x16x16_fp8_fp8; break;
case GLSL_TYPE_FLOAT_E5M2: opcode = aco_opcode::v_wmma_f32_16x16x16_fp8_bf8; break;
default: unreachable("invalid cmat_muladd_amd type");
}
break;
case GLSL_TYPE_FLOAT_E5M2:
switch (type_b) {
case GLSL_TYPE_FLOAT_E4M3FN: opcode = aco_opcode::v_wmma_f32_16x16x16_bf8_fp8; break;
case GLSL_TYPE_FLOAT_E5M2: opcode = aco_opcode::v_wmma_f32_16x16x16_bf8_bf8; break;
default: unreachable("invalid cmat_muladd_amd type");
}
break;
}
default: unreachable("invalid cmat_muladd_amd type");
}