diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2018-10-09 15:13:12 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:55:45 +0000 |
| commit | ecae3a14746fc7f678735b1a82347bd03f9a397f (patch) | |
| tree | 4c02d64cb9cbee8d270e2c498cc8ff1655bcdb36 /src/core/NEON/kernels/assembly | |
| parent | df3b5bb87296fdcde8ef88153f6365d693e80295 (diff) | |
| download | ComputeLibrary-ecae3a14746fc7f678735b1a82347bd03f9a397f.tar.gz | |
COMPMID-1451: Enable dot kernels in NEGEMMAssembly functions
Change-Id: I9dd26b80025ea3a4c66f5f0bf41b7a98dd0d3aa4
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/152549
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/assembly')
| -rw-r--r-- | src/core/NEON/kernels/assembly/NEGEMMInterleavedMatrixMultiplyWrapper.cpp | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/src/core/NEON/kernels/assembly/NEGEMMInterleavedMatrixMultiplyWrapper.cpp b/src/core/NEON/kernels/assembly/NEGEMMInterleavedMatrixMultiplyWrapper.cpp index 3d42f8a51f..2c9cd320f0 100644 --- a/src/core/NEON/kernels/assembly/NEGEMMInterleavedMatrixMultiplyWrapper.cpp +++ b/src/core/NEON/kernels/assembly/NEGEMMInterleavedMatrixMultiplyWrapper.cpp @@ -37,7 +37,7 @@ template <typename To, typename Tr, bool use_dot> void NEGEMMInterleavedMatrixMultiplyWrapperTemplate<To, Tr, use_dot>::configure(const ITensor *prepared_a, const ITensor *transformed_b, ITensor *tmp_c, ITensor *c, const Window &block_walker, const BlockSizes &block_sizes, const INEGEMMWrapperKernel::Params ¶ms, bool b_is_pretransposed, float alpha, float beta, unsigned int max_num_threads) { - using strategy = typename Kernel<To>::strategy; + using strategy = typename Kernel<To, use_dot>::strategy; _prepared_a = prepared_a; _transformed_b = transformed_b; @@ -57,7 +57,7 @@ template <typename To, typename Tr, bool use_dot> void NEGEMMInterleavedMatrixMultiplyWrapperTemplate<To, Tr, use_dot>::transform(const MatrixMultiplyWorkload &wl, const ThreadInfo &info, const Window &batch_window, const Coordinates &start_offset, const Coordinates &end_offset) { - using strategy = typename Kernel<To>::strategy; + using strategy = typename Kernel<To, use_dot>::strategy; strategy strat(info.cpu_info); TensorAccessor<To> prepared_a(*_prepared_a); @@ -98,7 +98,7 @@ void NEGEMMInterleavedMatrixMultiplyWrapperTemplate<To, Tr, use_dot>::transform( template <typename To, typename Tr, bool use_dot> void NEGEMMInterleavedMatrixMultiplyWrapperTemplate<To, Tr, use_dot>::create_workloads(std::vector<MatrixMultiplyWorkload> &workloads) { - using strategy = typename Kernel<To>::strategy; + using strategy = typename Kernel<To, use_dot>::strategy; unsigned int offset_transformed_b = 0; execute_window_loop(_block_walker, [&](const Coordinates & id) |