diff options
author | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-30 17:20:26 +0000 |
---|---|---|
committer | Anthony Barbier <Anthony.barbier@arm.com> | 2018-12-03 11:48:39 +0000 |
commit | 08a4517905da959b6e3401cc24f5e2018f9b51ac (patch) | |
tree | 9f6e2343ff22a53526a2f04e1ad22ecd91262767 /src/core/NEON/kernels/assembly | |
parent | 164a2727d3bbce0e575d24b7db787c85e2e2c203 (diff) | |
download | ComputeLibrary-08a4517905da959b6e3401cc24f5e2018f9b51ac.tar.gz |
COMPMID-1497: Use reshaped B strategy in GEMM by default
NELSTM, NEFullyConnectedLayer(For quantised types only), NERNN and NEWinogradLayer were all defaulting to on-the-fly reshaping of B
Fixed a bug in GemmInterleaved: it was ignoring the 'multis' dimension of the tensor to allocate the memory for B reshaped
Change-Id: I7b30f7f57fc65d6a03cccde0bf5515a811f17b54
Reviewed-on: https://review.mlplatform.org/323
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/assembly')
-rw-r--r-- | src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp b/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp index 6c201cedb3..41a031c1c7 100644 --- a/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp +++ b/src/core/NEON/kernels/assembly/NEGEMMInterleavedPrepareBWrapperKernel.cpp @@ -68,7 +68,7 @@ void for_each_element_in_window(const Window &window, const ITensor *b, ITensor // Calculate the size of transformed_b: template <typename To, bool use_dot> -unsigned int get_B_pretransposed_array_size(unsigned int N, unsigned int K, const BlockSizes &bs) +unsigned int get_B_pretransposed_array_size(unsigned int N, unsigned int K, const BlockSizes &bs, unsigned int multis) { using strategy = typename Kernel<To, use_dot>::strategy; @@ -89,6 +89,9 @@ unsigned int get_B_pretransposed_array_size(unsigned int N, unsigned int K, cons // Calculate the total size of the buffer: size_t total = num_full_k * normal_k_size * (num_full_x * normal_x_size + left_over_x_size); total += left_over_k_size * (left_over_x_size + num_full_x * normal_x_size); + + total *= multis; + return total; } @@ -114,7 +117,7 @@ void NEGEMMInterleavedPrepareBWrapperKernelTemplate<To, use_dot>::configure(cons _block_sizes = calculate_block_sizes<strategy>(ci, params.M, params.N, params.K); - auto_init_if_empty(*transformed_b->info(), b->info()->clone()->set_tensor_shape(TensorShape{ get_B_pretransposed_array_size<To, use_dot>(_Nsize, _Ksize, _block_sizes) })); + auto_init_if_empty(*transformed_b->info(), b->info()->clone()->set_tensor_shape(TensorShape{ get_B_pretransposed_array_size<To, use_dot>(_Nsize, _Ksize, _block_sizes, multis) })); Window window; window.set(Window::DimX, Window::Dimension(0, ceil_to_multiple(_Nsize, _block_sizes.x_block), _block_sizes.x_block)); |