aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp')
-rw-r--r--src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp104
1 files changed, 74 insertions, 30 deletions
diff --git a/src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp b/src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp
index 57f7af488b..f4c0839ad2 100644
--- a/src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp
+++ b/src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2018-2020 Arm Limited.
+ * Copyright (c) 2018-2021, 2023 Arm Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -26,6 +26,8 @@
#include "arm_compute/core/CL/CLHelpers.h"
#include "arm_compute/core/CL/ICLTensor.h"
#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "arm_compute/core/utils/StringUtils.h"
+
#include "src/core/CL/CLValidate.h"
#include "src/core/helpers/AutoConfiguration.h"
#include "src/core/helpers/WindowHelpers.h"
@@ -37,19 +39,22 @@ namespace arm_compute
{
namespace
{
-Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_info, const ITensorInfo *paddings, const ITensorInfo *output)
+Status validate_arguments(const ITensorInfo *input,
+ const ITensorInfo *block_info,
+ const ITensorInfo *paddings,
+ const ITensorInfo *output)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_info, paddings, output);
ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(block_info, 1, DataType::S32);
ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
ARM_COMPUTE_RETURN_ERROR_ON(block_info->num_dimensions() > 1);
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(block_info->tensor_shape(), TensorShape{ 2 });
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(block_info->tensor_shape(), TensorShape{2});
ARM_COMPUTE_RETURN_ERROR_ON(paddings->num_dimensions() > 2);
- ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(paddings->tensor_shape(), TensorShape{ 2, 2 });
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(paddings->tensor_shape(), TensorShape{2, 2});
// Validate output if initialized
- if(output->total_size() != 0)
+ if (output->total_size() != 0)
{
const DataLayout data_layout = input->data_layout();
const int idx_channel = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
@@ -60,7 +65,11 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_inf
return Status{};
}
-Status validate_arguments_static(const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
+Status validate_arguments_static(const ITensorInfo *input,
+ const int block_shape_x,
+ const int block_shape_y,
+ const Size2D &padding_left,
+ const Size2D &padding_right,
const ITensorInfo *output)
{
ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
@@ -69,9 +78,10 @@ Status validate_arguments_static(const ITensorInfo *input, const int block_shape
ARM_COMPUTE_RETURN_ERROR_ON(block_shape_x < 1 || block_shape_y < 1);
// Validate output if initialized
- if(output->total_size() != 0)
+ if (output->total_size() != 0)
{
- TensorShape expected_output_shape = misc::shape_calculator::compute_space_to_batch_shape(input, block_shape_x, block_shape_y, padding_left, padding_right);
+ TensorShape expected_output_shape = misc::shape_calculator::compute_space_to_batch_shape(
+ input, block_shape_x, block_shape_y, padding_left, padding_right);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), expected_output_shape);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(input, output);
@@ -84,18 +94,27 @@ Status validate_arguments_static(const ITensorInfo *input, const int block_shape
CLSpaceToBatchLayerKernel::CLSpaceToBatchLayerKernel()
: _input(nullptr), _block_shape(nullptr), _paddings(nullptr), _output(nullptr)
{
+ _type = CLKernelType::ELEMENTWISE;
}
-void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output)
+void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input,
+ const ICLTensor *block_shape,
+ const ICLTensor *paddings,
+ ICLTensor *output)
{
configure(CLKernelLibrary::get().get_compile_context(), input, block_shape, paddings, output);
}
-void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output)
+void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context,
+ const ICLTensor *input,
+ const ICLTensor *block_shape,
+ const ICLTensor *paddings,
+ ICLTensor *output)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(input, block_shape, paddings, output);
- ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), block_shape->info(), paddings->info(), output->info()));
- auto padding_info = get_padding_info({ input, block_shape, paddings, output });
+ ARM_COMPUTE_ERROR_THROW_ON(
+ validate_arguments(input->info(), block_shape->info(), paddings->info(), output->info()));
+ auto padding_info = get_padding_info({input, block_shape, paddings, output});
_input = input;
_block_shape = block_shape;
@@ -109,14 +128,17 @@ void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_contex
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
+ build_opts.add_option("-DDATA_TYPE=" +
+ get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_width)));
build_opts.add_option("-DHEIGHT_IN=" + support::cpp11::to_string(input->info()->dimension(idx_height)));
build_opts.add_option("-DBATCH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_batch)));
- _kernel = create_kernel(compile_context, "space_to_batch_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
+ _kernel = create_kernel(compile_context,
+ "space_to_batch_" + lower_string(string_from_data_layout(input->info()->data_layout())),
+ build_opts.options());
// Configure kernel window
Window win = calculate_max_window(*output->info(), Steps());
@@ -124,22 +146,34 @@ void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_contex
ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));
}
-void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
- ICLTensor *output)
+void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input,
+ const int block_shape_x,
+ const int block_shape_y,
+ const Size2D &padding_left,
+ const Size2D &padding_right,
+ ICLTensor *output)
{
- configure(CLKernelLibrary::get().get_compile_context(), input, block_shape_x, block_shape_y, padding_left, padding_right, output);
+ configure(CLKernelLibrary::get().get_compile_context(), input, block_shape_x, block_shape_y, padding_left,
+ padding_right, output);
}
-void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left,
- const Size2D &padding_right,
- ICLTensor *output)
+void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context,
+ const ICLTensor *input,
+ const int block_shape_x,
+ const int block_shape_y,
+ const Size2D &padding_left,
+ const Size2D &padding_right,
+ ICLTensor *output)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
- TensorShape output_shape = misc::shape_calculator::compute_space_to_batch_shape(input->info(), block_shape_x, block_shape_y, padding_left, padding_right);
- auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->quantization_info());
+ TensorShape output_shape = misc::shape_calculator::compute_space_to_batch_shape(
+ input->info(), block_shape_x, block_shape_y, padding_left, padding_right);
+ auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(),
+ input->info()->quantization_info());
- ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_static(input->info(), block_shape_x, block_shape_y, padding_left, padding_right, output->info()));
+ ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_static(input->info(), block_shape_x, block_shape_y, padding_left,
+ padding_right, output->info()));
_input = input;
_output = output;
@@ -151,7 +185,8 @@ void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_contex
// Create kernel
CLBuildOptions build_opts;
- build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
+ build_opts.add_option("-DDATA_TYPE=" +
+ get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
@@ -164,22 +199,32 @@ void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_contex
build_opts.add_option("-DPAD_RIGHT_X=" + support::cpp11::to_string(padding_right.x()));
build_opts.add_option("-DPAD_LEFT_Y=" + support::cpp11::to_string(padding_left.y()));
build_opts.add_option("-DPAD_RIGHT_Y=" + support::cpp11::to_string(padding_right.y()));
- _kernel = create_kernel(compile_context, "space_to_batch_static_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
+ _kernel = create_kernel(
+ compile_context, "space_to_batch_static_" + lower_string(string_from_data_layout(input->info()->data_layout())),
+ build_opts.options());
// Configure kernel window
Window win = calculate_max_window(*output->info(), Steps());
ICLKernel::configure_internal(win);
}
-Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output)
+Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input,
+ const ITensorInfo *block_shape,
+ const ITensorInfo *paddings,
+ const ITensorInfo *output)
{
ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, block_shape, paddings, output));
return Status{};
}
-Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
+Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input,
+ const int block_shape_x,
+ const int block_shape_y,
+ const Size2D &padding_left,
+ const Size2D &padding_right,
const ITensorInfo *output)
{
- ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_static(input, block_shape_x, block_shape_y, padding_left, padding_right, output));
+ ARM_COMPUTE_RETURN_ON_ERROR(
+ validate_arguments_static(input, block_shape_x, block_shape_y, padding_left, padding_right, output));
return Status{};
}
@@ -216,7 +261,6 @@ void CLSpaceToBatchLayerKernel::run(const Window &window, cl::CommandQueue &queu
add_3D_tensor_argument(idx, _output, slice_out);
enqueue(queue, *this, slice_out, lws_hint());
++batch_id;
- }
- while(window.slide_window_slice_3D(slice_out));
+ } while (window.slide_window_slice_3D(slice_out));
}
} // namespace arm_compute
generated by cgit v1.2.1 (git 2.23.0) at 2024-05-03 17:19:17 +0000