diff options
| author | Pablo Marquez Tello <pablo.tello@arm.com> | 2021-03-08 17:27:05 +0000 |
|---|---|---|
| committer | Pablo Marquez Tello <pablo.tello@arm.com> | 2021-03-17 12:45:26 +0000 |
| commit | a50f19346c5b79e2743f882ce0c691c07076f207 (patch) | |
| tree | 40141711eae786bc65738f04baa4e17cd6a20d97 /src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp | |
| parent | d0c9cb808f674ce8bbfbdf0e66c5b8451f6af0f2 (diff) | |
| download | ComputeLibrary-a50f19346c5b79e2743f882ce0c691c07076f207.tar.gz | |
Updated cpu detection
* Added the case in the cpu detection code for Klein cores
* Added has_sve() and set_sve() methods in CpuInfo
* Detection code checks for presence of SVE via HWCAP_SVE
* Updated the heuristic in sve kernels to check for the absence of Klein
* Resolves: COMPMID-4085
Change-Id: I0b8c72ff19dc5a3a81628d121a1afa836e724b4f
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5257
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp | 12 |
1 files changed, 6 insertions, 6 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp b/src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp index 96b9734221..d8134c4bb5 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp @@ -49,22 +49,22 @@ static const GemmImplementation<bfloat16, float> gemm_bf16_methods[] = { // gemm_bf16_interleaved GemmMethod::GEMM_INTERLEAVED, "sve_interleaved_bf16fp32_mmla_8x3VL", - [](const GemmArgs &args) { return (args._Ksize>4); }, - nullptr, + [](const GemmArgs &args) { return args._ci->has_sve() && (args._Ksize>4); }, + [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN; }, [](const GemmArgs &args) { return new GemmInterleaved<cls_sve_interleaved_bf16fp32_mmla_8x3VL, bfloat16, float>(args); } }, { GemmMethod::GEMM_HYBRID, "sve_hybrid_bf16fp32_dot_6x4VL", - nullptr, - [](const GemmArgs &args) { return ((args._Ksize <= 128) && (args._Nsize <= 128)); }, + [](const GemmArgs &args) { return args._ci->has_sve(); }, + [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN && ((args._Ksize <= 128) && (args._Nsize <= 128)); }, [](const GemmArgs &args) { return new GemmHybridIndirect<cls_sve_hybrid_bf16fp32_dot_6x4VL, bfloat16, float>(args); } }, { // gemm_bf16_interleaved GemmMethod::GEMM_INTERLEAVED, "sve_interleaved_bf16fp32_dot_8x3VL", - [](const GemmArgs &args) { return (args._Ksize>2); }, - nullptr, + [](const GemmArgs &args) { return args._ci->has_sve() && (args._Ksize>2); }, + [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN; }, [](const GemmArgs &args) { return new GemmInterleaved<cls_sve_interleaved_bf16fp32_dot_8x3VL, bfloat16, float>(args); } }, # endif // SVE |