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_fp16.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_fp16.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp index 93563a63d0..8e355c8f2c 100644 --- a/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp +++ b/src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp @@ -47,15 +47,15 @@ static const GemmImplementation<__fp16, __fp16> gemm_fp16_methods[] = { { GemmMethod::GEMM_HYBRID, "sve_hybrid_fp16_mla_6x4VL", - nullptr, - [](const GemmArgs &args) { return ((args._Ksize <= 256) && (args._Nsize <= 256)) || ((args._nmulti > 1) && ((args._Msize / args._maxthreads) < 8)); }, + [](const GemmArgs &args) { return args._ci->has_sve(); }, + [](const GemmArgs &args) { return args._ci->get_cpu_model() != CPUModel::KLEIN && (((args._Ksize <= 256) && (args._Nsize <= 256)) || ((args._nmulti > 1) && ((args._Msize / args._maxthreads) < 8))); }, [](const GemmArgs &args) { return new GemmHybridIndirect<cls_a64_hybrid_fp16_mla_6x32, __fp16, __fp16>(args); } }, { GemmMethod::GEMM_INTERLEAVED, "sve_interleaved_fp16_mla_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_fp16_mla_8x3VL, __fp16, __fp16>(args); } }, #endif |