aboutsummaryrefslogtreecommitdiff
path: root/src/core/CL
diff options
context:
space:
mode:
Diffstat (limited to 'src/core/CL')
-rw-r--r--src/core/CL/CLKernelLibrary.cpp7
-rw-r--r--src/core/CL/cl_kernels/deconvolution_layer.cl50
-rw-r--r--src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp117
3 files changed, 173 insertions, 1 deletions
diff --git a/src/core/CL/CLKernelLibrary.cpp b/src/core/CL/CLKernelLibrary.cpp
index de75518a05..352b89baa5 100644
--- a/src/core/CL/CLKernelLibrary.cpp
+++ b/src/core/CL/CLKernelLibrary.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2016, 2017 ARM Limited.
+ * Copyright (c) 2016, 2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -186,6 +186,7 @@ const std::map<std::string, std::string> CLKernelLibrary::_kernel_program_map =
{ "copy_plane", "channel_extract.cl" },
{ "copy_planes_3p", "channel_combine.cl" },
{ "copy_to_keypoint", "fast_corners.cl" },
+ { "deconvolution_upsample", "deconvolution_layer.cl" },
{ "depthwise_convolution_3x3", "depthwise_convolution.cl" },
{ "depthwise_convolution_3x3_quantized", "depthwise_convolution_quantized.cl" },
{ "depthwise_im2col", "depthwise_convolution.cl" },
@@ -421,6 +422,10 @@ const std::map<std::string, std::string> CLKernelLibrary::_program_source_map =
#include "./cl_kernels/convolution_rectangle.clembed"
},
{
+ "deconvolution_layer.cl",
+#include "./cl_kernels/deconvolution_layer.clembed"
+ },
+ {
"depth_convert.cl",
#include "./cl_kernels/depth_convert.clembed"
},
diff --git a/src/core/CL/cl_kernels/deconvolution_layer.cl b/src/core/CL/cl_kernels/deconvolution_layer.cl
new file mode 100644
index 0000000000..2514ddc8cc
--- /dev/null
+++ b/src/core/CL/cl_kernels/deconvolution_layer.cl
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2017, 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "helpers.h"
+
+/** This function applies upsample on an input image.
+ *
+ * @param[in] src_ptr Pointer to the source image. Supported data types: F32
+ * @param[in] src_stride_x Stride of the source image in X dimension (in bytes)
+ * @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in] src_stride_y Stride of the source image in Y dimension (in bytes)
+ * @param[in] src_step_y src_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in] src_offset_first_element_in_bytes The offset of the first element in the source image
+ * @param[out] dst_ptr Pointer to the destination image. Supported data types: F32
+ * @param[in] dst_stride_x Stride of the destination image in X dimension (in bytes)
+ * @param[in] dst_step_x dst_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in] dst_stride_y Stride of the destination image in Y dimension (in bytes)
+ * @param[in] dst_step_y dst_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in] dst_offset_first_element_in_bytes The offset of the first element in the destination image
+ */
+__kernel void deconvolution_upsample(
+ IMAGE_DECLARATION(src),
+ IMAGE_DECLARATION(dst))
+{
+ Image src = CONVERT_TO_IMAGE_STRUCT(src);
+ Image dst = CONVERT_TO_IMAGE_STRUCT(dst);
+
+ // Store result
+ *((__global float *)dst.ptr) = *((__global float *)src.ptr);
+}
diff --git a/src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp b/src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp
new file mode 100644
index 0000000000..5c08d5bee2
--- /dev/null
+++ b/src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp
@@ -0,0 +1,117 @@
+/*
+ * Copyright (c) 2017, 2018 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "arm_compute/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.h"
+
+#include "arm_compute/core/CL/CLHelpers.h"
+#include "arm_compute/core/CL/CLKernelLibrary.h"
+#include "arm_compute/core/CL/ICLTensor.h"
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/Window.h"
+
+using namespace arm_compute;
+
+CLDeconvolutionLayerUpsampleKernel::CLDeconvolutionLayerUpsampleKernel()
+ : _input(nullptr), _output(nullptr), _inner_border(), _info()
+{
+}
+
+Status CLDeconvolutionLayerUpsampleKernel::validate(const ITensorInfo *input, const ITensorInfo *output, const BorderSize &inner_border,
+ const PadStrideInfo &info)
+{
+ ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
+ ARM_COMPUTE_UNUSED(info);
+
+ ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+ ARM_COMPUTE_RETURN_ERROR_ON(output->dimension(0) == 0);
+ ARM_COMPUTE_RETURN_ERROR_ON(output->dimension(1) == 0);
+
+ for(size_t i = 2; i < Coordinates::num_max_dimensions; ++i)
+ {
+ ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(i) != output->dimension(i));
+ }
+
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(inner_border.right > info.stride().first - 1, "inner_border_right must be smaller that stride_x");
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(inner_border.top > info.stride().second - 1, "inner_border_top must be smaller that stride_y");
+
+ return Status{};
+}
+
+void CLDeconvolutionLayerUpsampleKernel::configure(const ICLTensor *input, ICLTensor *output, const BorderSize &inner_border,
+ const PadStrideInfo &info)
+{
+ ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
+
+ _input = input;
+ _output = output;
+ _inner_border = inner_border;
+ _info = info;
+
+ // Perform validation step
+ ARM_COMPUTE_ERROR_THROW_ON(CLDeconvolutionLayerUpsampleKernel::validate(input->info(), output->info(), inner_border, info));
+
+ // Create kernel
+ _kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("deconvolution_upsample"));
+
+ constexpr unsigned int num_elems_processed_per_iteration = 1;
+
+ // Configure kernel window
+ Window win = calculate_max_window(*output->info(), Steps(num_elems_processed_per_iteration));
+
+ AccessWindowHorizontal output_access(output->info(), 0, 0, num_elems_processed_per_iteration);
+ output_access.set_valid_region(win, ValidRegion(Coordinates(), output->info()->tensor_shape()));
+
+ ICLKernel::configure(win);
+}
+
+void CLDeconvolutionLayerUpsampleKernel::run(const Window &window, cl::CommandQueue &queue)
+{
+ ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+ ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);
+
+ const int out_start_x = _info.pad().first;
+ const int out_end_x = _output->info()->dimension(0) - _inner_border.right - _info.pad().first + _info.stride().first - 1;
+ const int out_step_x = _info.stride().first;
+
+ const int out_start_y = _inner_border.top + _info.pad().second;
+ const int out_end_y = _output->info()->dimension(1) - _info.pad().second + _info.stride().second - 1;
+ const int out_step_y = _info.stride().second;
+
+ Window slice_out = window.first_slice_window_2D();
+ slice_out.set(Window::DimX, Window::Dimension(out_start_x, out_end_x, out_step_x));
+ slice_out.set(Window::DimY, Window::Dimension(out_start_y, out_end_y, out_step_y));
+
+ Window slice_in = window.first_slice_window_2D();
+
+ do
+ {
+ unsigned int idx = 0;
+ add_2D_tensor_argument(idx, _input, slice_in);
+ add_2D_tensor_argument(idx, _output, slice_out);
+ enqueue(queue, *this, slice_out);
+ }
+ while(window.slide_window_slice_2D(slice_in) && window.slide_window_slice_2D(slice_out));
+}