aboutsummaryrefslogtreecommitdiff
path: root/src/core/cpu
diff options
context:
space:
mode:
authorManuel Bottini <manuel.bottini@arm.com>2021-07-01 18:13:33 +0100
committerGeorgios Pinitas <georgios.pinitas@arm.com>2021-07-09 09:42:34 +0000
commit24b892072a2bd8190ba63d09fb0082113d7d032a (patch)
tree6376c4449f0d2d7f146496ab32604b3835acf976 /src/core/cpu
parentda816752cad76c8e1b367e8e9c648994a1af599a (diff)
downloadComputeLibrary-24b892072a2bd8190ba63d09fb0082113d7d032a.tar.gz
Port NECol2ImKernel
Resolves: COMPMID-4511 Change-Id: Id6335cb23ef22bba02083498025da0ecb1647714 Signed-off-by: Manuel Bottini <manuel.bottini@arm.com> Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5898 Tested-by: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/cpu')
-rw-r--r--src/core/cpu/kernels/CpuCol2ImKernel.cpp124
-rw-r--r--src/core/cpu/kernels/CpuCol2ImKernel.h87
2 files changed, 211 insertions, 0 deletions
diff --git a/src/core/cpu/kernels/CpuCol2ImKernel.cpp b/src/core/cpu/kernels/CpuCol2ImKernel.cpp
new file mode 100644
index 0000000000..f860825de6
--- /dev/null
+++ b/src/core/cpu/kernels/CpuCol2ImKernel.cpp
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2017-2021 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 "src/core/cpu/kernels/CpuCol2ImKernel.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/core/ITensor.h"
+#include "arm_compute/core/Size2D.h"
+#include "arm_compute/core/TensorInfo.h"
+#include "arm_compute/core/Types.h"
+#include "arm_compute/core/Validate.h"
+#include "arm_compute/core/utils/misc/ShapeCalculator.h"
+#include "src/core/helpers/AutoConfiguration.h"
+#include "src/core/helpers/WindowHelpers.h"
+
+namespace arm_compute
+{
+using namespace misc::shape_calculator;
+namespace cpu
+{
+namespace kernels
+{
+namespace
+{
+Status validate_arguments(const ITensorInfo *src, const ITensorInfo *dst, const Size2D &convolved_dims)
+{
+ //Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use CPU FP16 instructions.
+ ARM_COMPUTE_RETURN_ERROR_ON(src->data_type() == DataType::UNKNOWN);
+
+ // Validate configured output
+ if(dst->total_size() != 0)
+ {
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(dst->tensor_shape(), compute_col2im_shape(*src, convolved_dims, false));
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(src, dst);
+ ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(src, dst);
+ }
+
+ return Status{};
+}
+} // namespace
+
+void CpuCol2ImKernel::configure(const ITensorInfo *src, ITensorInfo *dst, const Size2D &convolved_dims)
+{
+ ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);
+ ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, dst, convolved_dims));
+
+ _convolved_dims = convolved_dims;
+
+ // Configure kernel window
+ // Output auto inizialitation if not yet initialized
+ auto_init_if_empty(*dst, src->clone()->set_tensor_shape(compute_col2im_shape(*src, convolved_dims, false)));
+
+ // Configure kernel window
+ Window win = calculate_max_window(*src, Steps());
+
+ ICpuKernel::configure(win);
+}
+
+Status CpuCol2ImKernel::validate(const ITensorInfo *src, const ITensorInfo *output, const Size2D &convolved_dims)
+{
+ ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, output, convolved_dims));
+ return Status{};
+}
+
+void CpuCol2ImKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info)
+{
+ ARM_COMPUTE_UNUSED(info);
+ ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
+ ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICpuKernel::window(), window);
+
+ auto src = tensors.get_const_tensor(TensorType::ACL_SRC);
+ auto dst = tensors.get_tensor(TensorType::ACL_DST);
+
+ const uint8_t el_size = src->info()->element_size();
+ const int output_stride_x = dst->info()->strides_in_bytes().x();
+ const int output_stride_y = dst->info()->strides_in_bytes().y();
+ const int output_stride_z = dst->info()->strides_in_bytes().z();
+
+ Window window_out(window);
+ window_out.set(Window::DimX, Window::Dimension(0, 0, 0));
+ window_out.set(Window::DimY, Window::Dimension(0, 0, 0));
+ window_out.set(Window::DimZ, Window::Dimension(0, 0, 0));
+
+ // Create iterators
+ Iterator in(src, window);
+ Iterator out(dst, window_out);
+
+ execute_window_loop(window, [&](const Coordinates & id)
+ {
+ const int hidx = id.y();
+ const int idx = id.x() * output_stride_z + (hidx / _convolved_dims.width) * output_stride_y + (hidx % _convolved_dims.width) * output_stride_x;
+ std::memcpy(out.ptr() + idx, in.ptr(), el_size);
+ },
+ in, out);
+}
+
+const char *CpuCol2ImKernel::name() const
+{
+ return "CpuCol2ImKernel";
+}
+} // namespace kernels
+} // namespace cpu
+} // namespace arm_compute \ No newline at end of file
diff --git a/src/core/cpu/kernels/CpuCol2ImKernel.h b/src/core/cpu/kernels/CpuCol2ImKernel.h
new file mode 100644
index 0000000000..3c1802230b
--- /dev/null
+++ b/src/core/cpu/kernels/CpuCol2ImKernel.h
@@ -0,0 +1,87 @@
+/*
+ * Copyright (c) 2017-2021 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.
+ */
+#ifndef ARM_COMPUTE_CPU_COL2IM_KERNEL_H
+#define ARM_COMPUTE_CPU_COL2IM_KERNEL_H
+
+#include "arm_compute/core/Size2D.h"
+#include "src/core/common/Macros.h"
+#include "src/core/cpu/ICpuKernel.h"
+
+namespace arm_compute
+{
+namespace cpu
+{
+namespace kernels
+{
+/** Kernel to perform col2im reshaping.
+ *
+ * Rearranges each matrix column into image blocks. It's the inverse operation of @ref CpuIm2ColKernel.
+ *
+ * For example, a vector of 9 elements can be reshaped to a block(image) of 3x3:
+ *
+ * @f[
+ * \left( \begin{array}{ccccccccc}
+ * a0 & a1 & a2 & a3 & a4 & a5 & a6 & a7 & a8 \\
+ * \end{array} \right)
+ * \rightarrow
+ * \left( \begin{array}{ccc}
+ * a0 & a1 & a2 \\
+ * a3 & a4 & a5 \\
+ * a6 & a7 & a8 \\
+ * \end{array} \right)
+ * @f]
+ */
+class CpuCol2ImKernel : public ICpuKernel
+{
+public:
+ /** Default constructor */
+ CpuCol2ImKernel() = default;
+ ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuCol2ImKernel);
+ /** Set the input and output of the kernel.
+ *
+ * @param[in] src The input tensor info to convert. Data types supported: All
+ * @param[out] dst The output tensor info. 3 lower dimensions represent a single output [width, height, OFM],
+ * while the rest represent batch of outputs. Data types supported: Same as @p input
+ * @param[in] convolved_dims Output convolved dimensions.
+ */
+ void configure(const ITensorInfo *src, ITensorInfo *dst, const Size2D &convolved_dims);
+ /** Static function to check if given info will lead to a valid configuration
+ *
+ * Similar to CpuCol2ImKernel::configure()
+ *
+ * @return a status
+ */
+ static Status validate(const ITensorInfo *src, const ITensorInfo *dst, const Size2D &convolved_dims);
+
+ // Inherited methods overridden:
+ void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
+ const char *name() const override;
+
+private:
+ Size2D _convolved_dims{};
+};
+} // namespace kernels
+} // namespace cpu
+} // namespace arm_compute
+#endif /*ARM_COMPUTE_CPU_COL2IM_KERNEL_H */