diff options
| author | Sanghoon Lee <sanghoon.lee@arm.com> | 2018-01-23 15:16:47 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:49:16 +0000 |
| commit | f47bfb97fa8bc928a7860b84b7b227f716f65e58 (patch) | |
| tree | 6623bc798f312e0f1836f5df0fe82d3bde3e2f95 /src/core/NEON/kernels | |
| parent | be1f4a7f12e41f4988d4157f35dcb951cf31b72d (diff) | |
| download | ComputeLibrary-f47bfb97fa8bc928a7860b84b7b227f716f65e58.tar.gz | |
COMPMID-594: Implement reference and CL/NEON validation for LocallyConnected
Change-Id: I01e7abcf3f1b19458128e277044af850ad9fa224
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/118610
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels')
| -rw-r--r-- | src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp b/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp index 58da0402bc..35beb82689 100644 --- a/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp +++ b/src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017, 2018 ARM Limited. + * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -186,7 +186,7 @@ void vector_matrix_multiply_f32(const ITensor *input0, const ITensor *input1, IT win_out.set(Window::DimX, Window::Dimension(window_start_x, window_end_x, window_step_x)); Window win_a(window); - win_a.set(Window::DimX, Window::Dimension(0, 1, 1)); + win_a.set(Window::DimX, Window::Dimension(0, 0, 0)); Iterator ina(input0, win_a); Iterator out(output, win_out); @@ -234,7 +234,7 @@ void vector_matrix_multiply_f32(const ITensor *input0, const ITensor *input1, IT asm volatile("PLD [%0, #128*1]" ::"r"(reinterpret_cast<const uint8_t *>(matrix_b + 2 * in_b_stride))); asm volatile("PLD [%0, #128*1]" ::"r"(reinterpret_cast<const uint8_t *>(matrix_b + 3 * in_b_stride))); asm volatile("PLD [%0, #128*1]" ::"r"(reinterpret_cast<const uint8_t *>(matrix_b + 4 * in_b_stride))); -#endif /* __arm __ */ +#endif /* __arm__ */ acc0 = vmlaq_lane_f32(acc0, b00, a0l, 0); acc1 = vmlaq_lane_f32(acc1, b01, a0l, 0); |