COMPMID-1048 Add NHWC data format support to Winograd filter transform 4x4_3x3

Change-Id: Ifd125fcb5451dbac3c28b15a9471048a74fee0ad
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/128987
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>

1 files changed, 7 insertions, 8 deletions

diff --git a/src/core/CL/kernels/CLWinogradFilterTransformKernel.cpp b/src/core/CL/kernels/CLWinogradFilterTransformKernel.cpp
index 41b3ac50b5..cf4d73fbc1 100644
--- a/src/core/CL/kernels/CLWinogradFilterTransformKernel.cpp
+++ b/src/core/CL/kernels/CLWinogradFilterTransformKernel.cpp
@@ -47,7 +47,6 @@ namespace
 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const WinogradInfo &winograd_info)
 {
     ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
-    ARM_COMPUTE_RETURN_ERROR_ON(input->data_layout() != DataLayout::NCHW);
 
     const Size2D kernel_size      = winograd_info.kernel_size;
     const Size2D output_tile_size = winograd_info.output_tile_size;
@@ -56,6 +55,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, c
     const size_t idx_h = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT);
 
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size != Size2D(3U, 3U) && kernel_size != Size2D(5U, 5U), "Winograd filter transform only supports 3x3 and 5x5 kernels");
+    ARM_COMPUTE_RETURN_ERROR_ON(input->data_layout() == DataLayout::NHWC && (output_tile_size != Size2D(4U, 4U) || kernel_size != Size2D(3U, 3U)));
     ARM_COMPUTE_RETURN_ERROR_ON_MSG(kernel_size == Size2D(3U, 3U) && output_tile_size != Size2D(2U, 2U)
                                     && output_tile_size != Size2D(4U, 4U),
                                     "Winograd filter transform only supports 2x2 or 4x4 output tile for 3x3 kernels");
@@ -79,10 +79,11 @@ std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITen
 {
     ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
 
-    const unsigned int num_elems_processed_per_iteration_x = input->dimension(get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH));
-    const unsigned int num_elems_processed_per_iteration_y = input->dimension(get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT));
+    const unsigned int num_elems_processed_per_iteration_x = input->data_layout() == DataLayout::NCHW ? input->dimension(0) : 1;
+    const unsigned int num_elems_processed_per_iteration_y = input->dimension(1);
+    const unsigned int num_elems_read_per_iteration_z      = input->data_layout() == DataLayout::NCHW ? 1 : input->dimension(2);
 
-    Window win            = calculate_max_window(*input, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y));
+    Window win            = calculate_max_window(*input, Steps(num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y, num_elems_read_per_iteration_z));
     bool   window_changed = false;
 
     AccessWindowRectangle input_access(input, 0, 0, num_elems_processed_per_iteration_x, num_elems_processed_per_iteration_y);
@@ -111,17 +112,15 @@ void CLWinogradFilterTransformKernel::configure(const ICLTensor *input, ICLTenso
 
     ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), winograd_info));
 
-    const size_t idx_c = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::CHANNEL);
-
     // Set build options
     CLBuildOptions build_opts;
-    build_opts.add_option("-DNUM_CHANNELS=" + support::cpp11::to_string(input->info()->dimension(idx_c)));
+    build_opts.add_option("-DSRC_DIM_Z=" + support::cpp11::to_string(input->info()->dimension(2)));
 
     const Size2D kernel_size      = winograd_info.kernel_size;
     const Size2D output_tile_size = winograd_info.output_tile_size;
 
     // Create kernel
-    std::string kernel_name = "winograd_filter_transform_" + output_tile_size.to_string() + "_" + kernel_size.to_string() + "_nchw";
+    std::string kernel_name = "winograd_filter_transform_" + output_tile_size.to_string() + "_" + kernel_size.to_string() + "_" + lower_string(string_from_data_layout(input->info()->data_layout()));
     _kernel                 = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel(kernel_name, build_opts.options()));
 
     _input  = input;