diff options
| author | Giorgio Arena <giorgio.arena@arm.com> | 2018-06-19 11:27:38 +0100 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:53:09 +0000 |
| commit | fa23f1102f3e2d41838b8a9b53ab74c24cea5b50 (patch) | |
| tree | 617dc6036d90d1fead6275389c0cac40a720a34f /src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp | |
| parent | 7282d562d459066dff3e27fd5299f71e0809990d (diff) | |
| download | ComputeLibrary-fa23f1102f3e2d41838b8a9b53ab74c24cea5b50.tar.gz | |
COMPMID-1246 CLDepthwiseConvolution QASYMM8 NHWC kernel cleanup
Change-Id: If9385e6bcbf2242b973f42d6979b16ebc39f2cb4
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/136159
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp')
| -rw-r--r-- | src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp | 12 |
1 files changed, 5 insertions, 7 deletions
diff --git a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp index a54e92c63a..d24ef0f496 100644 --- a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp +++ b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NHWCKernel.cpp @@ -77,7 +77,7 @@ std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITen const unsigned int num_rows_read_per_iteration = num_rows_processed_per_iteration + 2; const unsigned int num_rows_written_per_iteration = num_rows_processed_per_iteration / conv_info.stride().first; - const BorderSize border_size(conv_info.pad_left() + num_rows_read_per_iteration * std::max(conv_info.pad_top(), conv_info.pad_bottom()), 0, conv_info.pad_right(), 0); + const BorderSize border_size(std::max(conv_info.pad_left(), conv_info.pad_top()), 0, std::max(conv_info.pad_right(), conv_info.pad_bottom()), 0); // Configure kernel window Window win = calculate_max_window(*output, Steps(num_elems_accessed_per_iteration, num_rows_written_per_iteration)); @@ -140,13 +140,11 @@ void CLDepthwiseConvolutionLayer3x3NHWCKernel::configure(const ICLTensor *input, _weights = weights; _biases = biases; _conv_stride_y = conv_info.stride().second; - _conv_pad_left = conv_info.pad_left(); _num_rows_processed_per_iteration = 4; const unsigned int num_elems_accessed_per_iteration = 4; - const unsigned int num_rows_read_per_iteration = _num_rows_processed_per_iteration + 2; - _border_size = BorderSize(_conv_pad_left + num_rows_read_per_iteration * std::max(conv_info.pad_top(), conv_info.pad_bottom()), 0, conv_info.pad_right(), 0); + _border_size = BorderSize(std::max(conv_info.pad_left(), conv_info.pad_top()), 0, std::max(conv_info.pad_right(), conv_info.pad_bottom()), 0); float multiplier = _input->info()->quantization_info().scale * _weights->info()->quantization_info().scale / _output->info()->quantization_info().scale; int output_multiplier = 0; @@ -162,9 +160,10 @@ void CLDepthwiseConvolutionLayer3x3NHWCKernel::configure(const ICLTensor *input, build_opts.add_option("-DOUTPUT_MULTIPLIER=" + support::cpp11::to_string(output_multiplier)); build_opts.add_option("-DOUTPUT_SHIFT=" + support::cpp11::to_string(output_shift)); build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_accessed_per_iteration)); - build_opts.add_option("-DSRC_DEPTH=" + support::cpp11::to_string(_input->info()->dimension(2))); + build_opts.add_option("-DSRC_DIM_1=" + support::cpp11::to_string(_input->info()->dimension(1))); + build_opts.add_option("-DSRC_DIM_2=" + support::cpp11::to_string(_input->info()->dimension(2))); build_opts.add_option("-DCONV_PAD_TOP=" + support::cpp11::to_string(conv_info.pad_top())); - build_opts.add_option("-DROWS_READ=" + support::cpp11::to_string(num_rows_read_per_iteration)); + build_opts.add_option("-DCONV_PAD_LEFT=" + support::cpp11::to_string(conv_info.pad_left())); if(act_info.enabled()) { @@ -236,7 +235,6 @@ void CLDepthwiseConvolutionLayer3x3NHWCKernel::run(const Window &window, cl::Com // Create input window and adjust Window win_in = window; - win_in.adjust(Window::DimY, -_conv_pad_left, true); win_in.set_dimension_step(Window::DimY, _num_rows_processed_per_iteration); win_in.set_dimension_step(Window::DimZ, _conv_stride_y); |