diff options
| author | Dana Zlotnik <dana.zlotnik@arm.com> | 2021-11-25 09:58:27 +0200 |
|---|---|---|
| committer | Dana Zlotnik <dana.zlotnik@arm.com> | 2021-12-20 12:56:22 +0000 |
| commit | 3229171402dcb9a63d63380865ba18477b81ff89 (patch) | |
| tree | bc9da509a853307184c4d55c7874376adfdcf13d /src/cpu/kernels/CpuActivationKernel.cpp | |
| parent | 4d44ac8685662984386b65869c3ed6af1144a419 (diff) | |
| download | ComputeLibrary-3229171402dcb9a63d63380865ba18477b81ff89.tar.gz | |
Decouple CpuActivationKernel
1- Data types were already decoupled. This commit arrange the folder struct of the activation kernel.
2- Refactor NEON CpuActivationKernel for floating-point cases.
Resolves COMPMID-4636
Change-Id: Ia4527244c84260dce1dd1d4bd4a9e3cfe2486d85
Signed-off-by: Dana Zlotnik <dana.zlotnik@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6739
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Giorgio Arena <giorgio.arena@arm.com>
Diffstat (limited to 'src/cpu/kernels/CpuActivationKernel.cpp')
| -rw-r--r-- | src/cpu/kernels/CpuActivationKernel.cpp | 30 |
1 files changed, 13 insertions, 17 deletions
diff --git a/src/cpu/kernels/CpuActivationKernel.cpp b/src/cpu/kernels/CpuActivationKernel.cpp index aed73d1fec..3af379d8af 100644 --- a/src/cpu/kernels/CpuActivationKernel.cpp +++ b/src/cpu/kernels/CpuActivationKernel.cpp @@ -65,57 +65,57 @@ static const ActivationKernel available_kernels[] = { "sve_fp16_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::F16 && data.ci.has_sve(); }, - REGISTER_FP16_SVE(arm_compute::cpu::fp16_sve_activation) + REGISTER_FP16_SVE(arm_compute::cpu::sve_fp16_activation) }, { "sve_fp32_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::F32 && data.ci.has_sve(); }, - REGISTER_FP32_SVE(arm_compute::cpu::fp32_sve_activation) + REGISTER_FP32_SVE(arm_compute::cpu::sve_fp32_activation) }, #endif /* defined(ARM_COMPUTE_ENABLE_SVE) */ #if defined(ARM_COMPUTE_ENABLE_NEON) { "neon_fp16_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::F16; }, - REGISTER_FP16_NEON(arm_compute::cpu::fp16_neon_activation) + REGISTER_FP16_NEON(arm_compute::cpu::neon_fp16_activation) }, { "neon_fp32_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::F32; }, - REGISTER_FP32_NEON(arm_compute::cpu::fp32_neon_activation) + REGISTER_FP32_NEON(arm_compute::cpu::neon_fp32_activation) }, #endif /* defined(ARM_COMPUTE_ENABLE_NEON) */ #if defined(ARM_COMPUTE_ENABLE_SVE2) { "sve_qu8_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QASYMM8 && data.ci.has_sve2(); }, - REGISTER_QASYMM8_SVE(arm_compute::cpu::qasymm8_sve_activation) + REGISTER_QASYMM8_SVE2(arm_compute::cpu::sve2_qasymm8_activation) }, { "sve_qs8_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QASYMM8_SIGNED && data.ci.has_sve2(); }, - REGISTER_QASYMM8_SIGNED_SVE(arm_compute::cpu::qasymm8_signed_sve_activation) + REGISTER_QASYMM8_SIGNED_SVE2(arm_compute::cpu::sve2_qasymm8_signed_activation) }, { "sve_qs16_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QSYMM16 && data.ci.has_sve2(); }, - REGISTER_QSYMM16_SVE(arm_compute::cpu::qsymm16_sve_activation) + REGISTER_QSYMM16_SVE2(arm_compute::cpu::sve2_qsymm16_activation) }, #endif /* defined(ARM_COMPUTE_ENABLE_SVE2) */ { "neon_qu8_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QASYMM8; }, - REGISTER_QASYMM8_NEON(arm_compute::cpu::qasymm8_neon_activation) + REGISTER_QASYMM8_NEON(arm_compute::cpu::neon_qasymm8_activation) }, { "neon_qs8_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QASYMM8_SIGNED; }, - REGISTER_QASYMM8_SIGNED_NEON(arm_compute::cpu::qasymm8_signed_neon_activation) + REGISTER_QASYMM8_SIGNED_NEON(arm_compute::cpu::neon_qasymm8_signed_activation) }, { "neon_qs16_activation", [](const ActivationSelectorData & data) { return data.dt == DataType::QSYMM16; }, - REGISTER_QSYMM16_NEON(arm_compute::cpu::qsymm16_neon_activation) + REGISTER_QSYMM16_NEON(arm_compute::cpu::neon_qsymm16_activation) }, }; @@ -233,18 +233,14 @@ Status CpuActivationKernel::validate(const ITensorInfo *src, const ITensorInfo * size_t CpuActivationKernel::get_mws(const CPUInfo &platform, size_t thread_count) const { ARM_COMPUTE_UNUSED(thread_count); - // Tuning results that gave optimized results in performance investigation - if (platform.get_cpu_model() == CPUModel::A73 ) + // Tuning results that gave optimized results in performance investigation + if(platform.get_cpu_model() == CPUModel::A73) { return 10240; } - else if (platform.get_cpu_model() == CPUModel::A76) - { - return 9216; - } else { - return ICPPKernel::default_mws; + return 9216; } } |