Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[CPU] Fixed Reduce kernel with bf16 destination precision #28731

Closed
wants to merge 1 commit into from
Closed

[CPU] Fixed Reduce kernel with bf16 destination precision #28731

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 22 additions & 14 deletions src/plugins/intel_cpu/src/nodes/reduce.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,8 +150,7 @@ struct jit_uni_reduce_kernel_f32 : public jit_uni_reduce_kernel, public jit_gene
data_type::f32);
}

if (mayiuse(avx512_core))
uni_vcvtneps2bf16 = std::make_shared<jit_uni_vcvtneps2bf16>(this, isa);
uni_vcvtneps2bf16 = std::make_shared<jit_uni_vcvtneps2bf16>(this, isa);

this->preamble();

Expand Down Expand Up @@ -185,8 +184,7 @@ struct jit_uni_reduce_kernel_f32 : public jit_uni_reduce_kernel, public jit_gene

this->postamble();

if (mayiuse(avx512_core))
uni_vcvtneps2bf16->emit_data();
uni_vcvtneps2bf16->emit_data();

if (jcp_.reduce_mode == Algorithm::ReduceAnd || jcp_.reduce_mode == Algorithm::ReduceL1 ||
jcp_.reduce_mode == Algorithm::ReduceMax || jcp_.reduce_mode == Algorithm::ReduceMin ||
Expand Down Expand Up @@ -1006,9 +1004,15 @@ struct jit_uni_reduce_kernel_f32 : public jit_uni_reduce_kernel, public jit_gene
uni_vmovups(op, vmm_dst);
break;
case memory::data_type::bf16:
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(ymm_dst.getIdx())});
vmovdqu16(op, ymm_dst);
if (isa == cpu::x64::avx512_core) {
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(ymm_dst.getIdx())});
vmovdqu16(op, ymm_dst);
} else {
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(xmm_dst.getIdx())});
uni_vmovdqu(op, xmm_dst);
}
break;
case memory::data_type::f16:
vcvtps2ph(op, vmm_dst, 0x4);
Expand Down Expand Up @@ -1237,8 +1241,7 @@ struct jit_uni_reduce_post_kernel_f32 : public jit_uni_reduce_post_kernel, publi
data_type::f32);
}

if (mayiuse(avx512_core))
uni_vcvtneps2bf16 = std::make_shared<jit_uni_vcvtneps2bf16>(this, isa);
uni_vcvtneps2bf16 = std::make_shared<jit_uni_vcvtneps2bf16>(this, isa);

this->preamble();

Expand Down Expand Up @@ -1292,8 +1295,7 @@ struct jit_uni_reduce_post_kernel_f32 : public jit_uni_reduce_post_kernel, publi

this->postamble();

if (mayiuse(avx512_core))
uni_vcvtneps2bf16->emit_data();
uni_vcvtneps2bf16->emit_data();

if (jcp_.reduce_mode == Algorithm::ReduceLogSum || jcp_.reduce_mode == Algorithm::ReduceLogSumExp) {
log_injector->prepare_table();
Expand Down Expand Up @@ -1723,9 +1725,15 @@ struct jit_uni_reduce_post_kernel_f32 : public jit_uni_reduce_post_kernel, publi
uni_vmovups(op, vmm_dst);
break;
case memory::data_type::bf16:
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(ymm_dst.getIdx())});
vmovdqu16(op, ymm_dst);
if (isa == cpu::x64::avx512_core) {
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(ymm_dst.getIdx())});
vmovdqu16(op, ymm_dst);
} else {
uni_vcvtneps2bf16->emit_code({static_cast<size_t>(vmm_dst.getIdx())},
{static_cast<size_t>(xmm_dst.getIdx())});
uni_vmovdqu(op, xmm_dst);
}
break;
case memory::data_type::f16:
vcvtps2ph(op, vmm_dst, 0x4);
Expand Down
Loading