diff options
author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2020-01-02 12:11:13 +0000 |
---|---|---|
committer | Giorgio Arena <giorgio.arena@arm.com> | 2020-01-10 18:08:22 +0000 |
commit | 13ec5f0a09e038f12cbe0f3b119a215934b72b42 (patch) | |
tree | 9a574b7d5227176d1fb9216587d30b60a97125b3 /src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp | |
parent | 781d727a741eb264f0e6614b41780c05050972e3 (diff) | |
download | ComputeLibrary-13ec5f0a09e038f12cbe0f3b119a215934b72b42.tar.gz |
COMPMID-2800: Add support for QASYMM8_SIGNED in NEDepthwiseConvolutionLayer3x3Kernel
Change-Id: Ia5d23ff2c9e59c80ded2fac5ca02704214f0a01a
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/2537
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp')
-rw-r--r-- | src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp b/src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp index bcf70b3ad8..4a71c1edea 100644 --- a/src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp +++ b/src/core/NEON/kernels/NEDirectConvolutionLayerKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2019 ARM Limited. + * Copyright (c) 2017-2020 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -651,7 +651,7 @@ public: const int output_w = output->info()->dimension(0); const int output_h = output->info()->dimension(1); const int num_planes_z = window.z().end() - window.z().start(); - const int delta_input = get_input_num_elems_processed<stridex>(num_elems_written_per_iteration); + const int delta_input = get_input_num_elems_processed(num_elems_written_per_iteration, stridex); const int kernel_depth = weights->info()->dimension(Window::DimZ); const unsigned int conv_stride_y = std::get<1>(conv_info.stride()); const unsigned int conv_pad_left = conv_info.pad_left(); @@ -718,7 +718,7 @@ public: for(int ow = 0; ow < output_w; ow += num_elems_written_per_iteration, in_top += delta_input, in_mid += delta_input, in_low += delta_input, p_out += num_elems_written_per_iteration) { - auto vres = convolve_3x3<stridex>(in_top, in_mid, in_low, vk_r0, vk_r1, vk_r2); + auto vres = convolve_3x3(in_top, in_mid, in_low, vk_r0, vk_r1, vk_r2, stridex); store_results<stridex>(p_out, vres); } } @@ -743,7 +743,7 @@ public: for(int ow = 0; ow < output_w; ow += num_elems_written_per_iteration, in_top += delta_input, in_mid += delta_input, in_low += delta_input, p_out += num_elems_written_per_iteration) { - auto vres = convolve_3x3<stridex>(in_top, in_mid, in_low, vk_r0, vk_r1, vk_r2); + auto vres = convolve_3x3(in_top, in_mid, in_low, vk_r0, vk_r1, vk_r2, stridex); accumulate_results<stridex>(p_out, vres); } } @@ -774,7 +774,7 @@ public: const int output_w = output->info()->dimension(0); const int output_h = output->info()->dimension(1); const int num_planes_z = window.z().end() - window.z().start(); - const int delta_input = get_input_num_elems_processed<stridex>(num_elems_written_per_iteration); + const int delta_input = get_input_num_elems_processed(num_elems_written_per_iteration, stridex); const int kernel_depth = weights->info()->dimension(Window::DimZ); const unsigned int conv_stride_y = std::get<1>(conv_info.stride()); const unsigned int conv_pad_left = conv_info.pad_left(); |