diff options
author | Pablo Tello <pablo.tello@arm.com> | 2017-06-26 17:17:42 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-09-17 14:16:42 +0100 |
commit | 0c34fe29c298057091d48cde332cb60bb14efee1 (patch) | |
tree | f77502715904a522b0e76f32eb1bcd6ce2f567dc /src/core/NEON/kernels/NEFillBorderKernel.cpp | |
parent | 2b26b850c0cff6a25f1012e9e4e7fe6654364a88 (diff) | |
download | ComputeLibrary-0c34fe29c298057091d48cde332cb60bb14efee1.tar.gz |
COMPMID-421: Added FP16 support in Pooling Layer
Change-Id: I6b6119c8770051c1656da40aa073c539c15b493e
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/78985
Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NEFillBorderKernel.cpp')
-rw-r--r-- | src/core/NEON/kernels/NEFillBorderKernel.cpp | 18 |
1 files changed, 17 insertions, 1 deletions
diff --git a/src/core/NEON/kernels/NEFillBorderKernel.cpp b/src/core/NEON/kernels/NEFillBorderKernel.cpp index bd99242b11..cd84e36aad 100644 --- a/src/core/NEON/kernels/NEFillBorderKernel.cpp +++ b/src/core/NEON/kernels/NEFillBorderKernel.cpp @@ -33,6 +33,10 @@ #include <algorithm> #include <cstdint> +#if ARM_COMPUTE_ENABLE_FP16 +#include <arm_fp16.h> // needed for float16_t +#endif /* ARM_COMPUTE_ENABLE_FP16 */ + using namespace arm_compute; namespace arm_compute @@ -47,7 +51,7 @@ NEFillBorderKernel::NEFillBorderKernel() void NEFillBorderKernel::configure(ITensor *tensor, BorderSize border_size, BorderMode border_mode, const PixelValue &constant_border_value) { - ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(tensor, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::U16, DataType::S16, DataType::U32, DataType::S32, DataType::F32); + ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(tensor, 1, DataType::U8, DataType::QS8, DataType::QS16, DataType::U16, DataType::S16, DataType::F16, DataType::U32, DataType::S32, DataType::F32); _tensor = tensor; _border_size = border_size; @@ -100,6 +104,12 @@ void NEFillBorderKernel::run(const Window &window) case DataType::S32: fill_constant_value_single_channel<int32_t>(window); break; +#ifdef ARM_COMPUTE_ENABLE_FP16 + case DataType::F16: + static_assert(sizeof(float16_t) == 2, "Float16_t must be 16 bit"); + fill_constant_value_single_channel<float16_t>(window); + break; +#endif /* ARM_COMPUTE_ENABLE_FP16 */ case DataType::F32: static_assert(sizeof(float) == 4, "Float must be 32 bit"); fill_constant_value_single_channel<float>(window); @@ -133,6 +143,12 @@ void NEFillBorderKernel::run(const Window &window) case DataType::S32: fill_replicate_single_channel<int32_t>(window); break; +#ifdef ARM_COMPUTE_ENABLE_FP16 + case DataType::F16: + static_assert(sizeof(float16_t) == 2, "Float16_t must be 16 bit"); + fill_replicate_single_channel<float16_t>(window); + break; +#endif /* ARM_COMPUTE_ENABLE_FP16 */ case DataType::F32: static_assert(sizeof(float) == 4, "Float must be 32 bit"); fill_replicate_single_channel<float>(window); |