diff options
| author | Ioan-Cristian Szabo <ioan-cristian.szabo@arm.com> | 2017-11-13 13:34:08 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:35:24 +0000 |
| commit | 5edbd1c5dce43b66f30c903797a91e39369c5b62 (patch) | |
| tree | d4c20c1a92ff9a7e26ffc9b1f6976ee12a2e2ae0 /src/core/NEON/kernels/NESoftmaxLayerKernel.cpp | |
| parent | 84b51ad1aaa530d397761f2b6da65add9dc8a6b0 (diff) | |
| download | ComputeLibrary-5edbd1c5dce43b66f30c903797a91e39369c5b62.tar.gz | |
COMPMID-556: Add support to build arm64-v8.2-a for Android platform (clang compiler)
Change-Id: Ibb779dd3a8d10786da6d8f70590e654e14654d7b
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/95530
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NESoftmaxLayerKernel.cpp')
| -rw-r--r-- | src/core/NEON/kernels/NESoftmaxLayerKernel.cpp | 24 |
1 files changed, 12 insertions, 12 deletions
diff --git a/src/core/NEON/kernels/NESoftmaxLayerKernel.cpp b/src/core/NEON/kernels/NESoftmaxLayerKernel.cpp index 73aba284ca..f1027590e4 100644 --- a/src/core/NEON/kernels/NESoftmaxLayerKernel.cpp +++ b/src/core/NEON/kernels/NESoftmaxLayerKernel.cpp @@ -106,7 +106,7 @@ void logits_1d_max_qs16(const ITensor *in, ITensor *out, const Window &window) while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(max_slice)); } -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC void logits_1d_max_f16(const ITensor *in, ITensor *out, const Window &window) { Window in_slice = window.first_slice_window_1D(); @@ -138,7 +138,7 @@ void logits_1d_max_f16(const ITensor *in, ITensor *out, const Window &window) } while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(max_slice)); } -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ void logits_1d_max_f32(const ITensor *in, ITensor *out, const Window &window) { @@ -213,10 +213,10 @@ void NELogits1DMaxKernel::configure(const ITensor *input, ITensor *output) _func = &logits_1d_max_f32; break; case DataType::F16: -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC _func = &logits_1d_max_f16; break; -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ default: ARM_COMPUTE_ERROR("Unsupported data type."); } @@ -374,7 +374,7 @@ void logits_1d_shift_exp_sum_qs16(const ITensor *in, const ITensor *max, ITensor while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(max_slice)); } -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC void logits_1d_shift_exp_sum_f16(const ITensor *in, const ITensor *max, ITensor *out, ITensor *sum, const Window &window) { Window window_max(window); @@ -434,7 +434,7 @@ void logits_1d_shift_exp_sum_f16(const ITensor *in, const ITensor *max, ITensor } while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(max_slice)); } -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ void logits_1d_shift_exp_sum_f32(const ITensor *in, const ITensor *max, ITensor *out, ITensor *sum, const Window &window) { @@ -532,10 +532,10 @@ void NELogits1DShiftExpSumKernel::configure(const ITensor *input, const ITensor _func = &logits_1d_shift_exp_sum_f32; break; case DataType::F16: -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC _func = &logits_1d_shift_exp_sum_f16; break; -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ default: ARM_COMPUTE_ERROR("Unsupported data type."); break; @@ -637,7 +637,7 @@ void logits_1d_norm_qs16(const ITensor *in, const ITensor *sum, ITensor *out, co } while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(sum_slice)); } -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC void logits_1d_norm_f16(const ITensor *in, const ITensor *sum, ITensor *out, const Window &window) { Window window_sum(window); @@ -668,7 +668,7 @@ void logits_1d_norm_f16(const ITensor *in, const ITensor *sum, ITensor *out, con } while(window.slide_window_slice_1D(in_slice) && window.slide_window_slice_1D(sum_slice)); } -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ void logits_1d_norm_f32(const ITensor *in, const ITensor *sum, ITensor *out, const Window &window) { @@ -738,10 +738,10 @@ void NELogits1DNormKernel::configure(const ITensor *input, const ITensor *sum, I _func = &logits_1d_norm_f32; break; case DataType::F16: -#ifdef ARM_COMPUTE_AARCH64_V8_2 +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC _func = &logits_1d_norm_f16; break; -#endif /* ARM_COMPUTE_AARCH64_V8_2 */ +#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */ default: ARM_COMPUTE_ERROR("Unsupported data type."); break; |