diff options
| author | Pablo Tello <pablo.tello@arm.com> | 2017-06-28 11:32:05 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-09-17 14:15:39 +0100 |
| commit | 997aba2dc6a0b5a8c3b424a31ae8370b395cf0d8 (patch) | |
| tree | d0c12ba56607ae788c07880875bae39f8960030f /src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp | |
| parent | 7b7858df42fccefbe6eb086ad516d5c011becd07 (diff) | |
| download | ComputeLibrary-997aba2dc6a0b5a8c3b424a31ae8370b395cf0d8.tar.gz | |
COMPMID-421: Fixed a problem in Convolution Layer reference values for FP16.
All methods in std::numeric_limits<float16_t> return 0.
Change-Id: I2289e01853e1b2c38afdec119ef6fc8af8a9752e
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79312
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Moritz Pflanzer <moritz.pflanzer@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp')
| -rw-r--r-- | src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp b/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp index dcfbb13081..1db025723c 100644 --- a/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp @@ -639,6 +639,7 @@ template <bool multiply_alpha> void matrix_matrix_multiply_f16(const ITensor *input0, const ITensor *input1, ITensor *output, const Window &window, float alpha) { #ifdef ARM_COMPUTE_ENABLE_FP16 + const size_t in_b_stride = input1->info()->strides_in_bytes()[1] / data_size_from_type(input1->info()->data_type()); const size_t out_stride = output->info()->strides_in_bytes()[1] / data_size_from_type(output->info()->data_type()); |