diff options
| author | Jonathan Deakin <jonathan.deakin@arm.com> | 2022-08-23 11:44:18 +0100 |
|---|---|---|
| committer | Pablo Marquez Tello <pablo.tello@arm.com> | 2022-09-26 08:28:46 +0000 |
| commit | d6b8a71714361881a249a6f6ed67125f290f4a83 (patch) | |
| tree | 63fb3c5fba04980816e41a11240511b3a587a5ad /src/cpu/kernels/activation/generic/sve/fp16.cpp | |
| parent | ce79ac6297e6eb2407abd24846b8504dee43770f (diff) | |
| download | ComputeLibrary-d6b8a71714361881a249a6f6ed67125f290f4a83.tar.gz | |
Add FP32 Neon™ swish activation
Change-Id: Id37b59adbc8c4cbe218d1652aeb02a0b4ce42c66
Signed-off-by: Jonathan Deakin <jonathan.deakin@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8256
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez Tello <pablo.tello@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/kernels/activation/generic/sve/fp16.cpp')
| -rw-r--r-- | src/cpu/kernels/activation/generic/sve/fp16.cpp | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/src/cpu/kernels/activation/generic/sve/fp16.cpp b/src/cpu/kernels/activation/generic/sve/fp16.cpp index 5730a361d9..6e9648b5bf 100644 --- a/src/cpu/kernels/activation/generic/sve/fp16.cpp +++ b/src/cpu/kernels/activation/generic/sve/fp16.cpp @@ -114,6 +114,9 @@ void sve_fp16_activation(const ITensor *src, ITensor *dst, const ActivationLayer case ActivationLayerInfo::ActivationFunction::HARD_SWISH: tmp = svmul_f16_z(pg, vin, svmul_f16_z(pg, const_inv_6, svmin_f16_z(pg, const_6, svmax_f16_z(pg, const_0, svadd_f16_z(pg, vin, const_3))))); break; + case ActivationLayerInfo::ActivationFunction::SWISH: + tmp = svmul_f16_z(pg, vin, svinv_f16_z(pg, svadd_f16_z(pg, const_1, svexp_f16_z(pg, svneg_f16_z(pg, svmul_f16_z(pg, va, vin)))))); + break; default: ARM_COMPUTE_ERROR("Unsupported activation function"); } @@ -129,4 +132,4 @@ void sve_fp16_activation(const ITensor *src, ITensor *dst, const ActivationLayer } } // namespace cpu } // namespace arm_compute -#endif /* defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */
\ No newline at end of file +#endif /* defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */ |