diff options
author | Michalis Spyrou <michalis.spyrou@arm.com> | 2018-08-21 18:03:58 +0100 |
---|---|---|
committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:54:54 +0000 |
commit | 16934a511ccd37a28000a9fabb3e6e5fc6f51ec9 (patch) | |
tree | 2c7875d92b4ede16ef6efc63033a2ad31ba68bc0 /arm_compute | |
parent | 0dd2391f9fb4dd464fac8c144a45cf24189079fa (diff) | |
download | ComputeLibrary-16934a511ccd37a28000a9fabb3e6e5fc6f51ec9.tar.gz |
COMPMID-1227 Implementing Space to Batch on OpenCL
Change-Id: I6fd83d6584c56a4fd2470948f1987e23237c16d3
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/145577
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: bsgcomp <bsgcomp@arm.com>
Diffstat (limited to 'arm_compute')
-rw-r--r-- | arm_compute/core/CL/CLKernels.h | 1 | ||||
-rw-r--r-- | arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h | 101 | ||||
-rw-r--r-- | arm_compute/core/utils/misc/ShapeCalculator.h | 10 | ||||
-rw-r--r-- | arm_compute/runtime/CL/CLFunctions.h | 1 | ||||
-rw-r--r-- | arm_compute/runtime/CL/functions/CLSpaceToBatchLayer.h | 103 |
5 files changed, 216 insertions, 0 deletions
diff --git a/arm_compute/core/CL/CLKernels.h b/arm_compute/core/CL/CLKernels.h index 9586d8cb34..2fd6ab2e1c 100644 --- a/arm_compute/core/CL/CLKernels.h +++ b/arm_compute/core/CL/CLKernels.h @@ -112,6 +112,7 @@ #include "arm_compute/core/CL/kernels/CLSobel5x5Kernel.h" #include "arm_compute/core/CL/kernels/CLSobel7x7Kernel.h" #include "arm_compute/core/CL/kernels/CLSoftmaxLayerKernel.h" +#include "arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h" #include "arm_compute/core/CL/kernels/CLStridedSliceKernel.h" #include "arm_compute/core/CL/kernels/CLTableLookupKernel.h" #include "arm_compute/core/CL/kernels/CLThresholdKernel.h" diff --git a/arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h b/arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h new file mode 100644 index 0000000000..81095f9aef --- /dev/null +++ b/arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h @@ -0,0 +1,101 @@ +/* + * Copyright (c) 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. + */ +#ifndef __ARM_COMPUTE_CLSPACETOBATCHLAYERKERNEL_H__ +#define __ARM_COMPUTE_CLSPACETOBATCHLAYERKERNEL_H__ + +#include "arm_compute/core/CL/ICLKernel.h" +#include "arm_compute/core/Types.h" + +namespace arm_compute +{ +class ICLTensor; + +/** Interface for the space to batch kernel */ +class CLSpaceToBatchLayerKernel : public ICLKernel +{ +public: + /** Default constructor */ + CLSpaceToBatchLayerKernel(); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLSpaceToBatchLayerKernel(const CLSpaceToBatchLayerKernel &) = delete; + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLSpaceToBatchLayerKernel &operator=(const CLSpaceToBatchLayerKernel &) = delete; + /** Allow instances of this class to be moved */ + CLSpaceToBatchLayerKernel(CLSpaceToBatchLayerKernel &&) = default; + /** Allow instances of this class to be moved */ + CLSpaceToBatchLayerKernel &operator=(CLSpaceToBatchLayerKernel &&) = default; + /** Default destructor */ + ~CLSpaceToBatchLayerKernel() = default; + /** Initialise the kernel's inputs and output. + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape 1-D tensor with shape [M]. Data types supported: S32 + * @param[in] paddings 2-D tensor with shape [2, M]. Data types supported: S32 + * @param[out] output Tensor output. Data types supported: same as @p input + */ + void configure(const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output); + /** Initialise the kernel's input and output. (Static block shape and paddings) + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape_x Block shape x value. + * @param[in] block_shape_y Block shape y value. + * @param[in] padding_left The left padding of the output tensor. + * @param[in] padding_right The right padding of the output tensor. + * @param[out] output Tensor output. Data types supported: same as @p input + */ + void configure(const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right, ICLTensor *output); + /** Static function to check if given info will lead to a valid configuration of @ref CLSpaceToBatchLayerKernel + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape 1-D tensor with shape [M]. Data types supported: S32 + * @param[in] paddings 2-D tensor with shape [2, M]. Data types supported: S32 + * @param[in] output Tensor output. Data types supported: same as @p input + * + * @return a status + */ + static Status validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output); + /** Static function to check if given info will lead to a valid configuration of @ref CLSpaceToBatchLayerKernel (Static block shape and paddings) + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape_x Block shape x value. + * @param[in] block_shape_y Block shape y value. + * @param[in] padding_left The left padding of the output tensor. + * @param[in] padding_right The right padding of the output tensor. + * @param[in] output Tensor output. Data types supported: same as @p input + * + * @return a status + */ + static Status 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); + + // Inherited methods overridden: + void run(const Window &window, cl::CommandQueue &queue) override; + +private: + const ICLTensor *_input; /**< Source tensor */ + const ICLTensor *_block_shape; /**< Block shape tensor */ + const ICLTensor *_paddings; /**< Paddings tensor */ + ICLTensor *_output; /**< Destination tensor */ +}; +} // namespace arm_compute +#endif /* __ARM_COMPUTE_CLSPACETOBATCHLAYERKERNEL_H__ */ diff --git a/arm_compute/core/utils/misc/ShapeCalculator.h b/arm_compute/core/utils/misc/ShapeCalculator.h index cb04182c21..806149f83f 100644 --- a/arm_compute/core/utils/misc/ShapeCalculator.h +++ b/arm_compute/core/utils/misc/ShapeCalculator.h @@ -557,6 +557,16 @@ inline TensorShape compute_split_shape(const ITensorInfo *input, unsigned int ax return out_shape; } +inline TensorShape compute_space_to_batch_shape(const ITensorInfo *input, const int block_x, const int block_y, const Size2D &padding_left, const Size2D &padding_right) +{ + TensorShape output_shape{ input->tensor_shape() }; + output_shape.set(0, input->tensor_shape()[0] * block_x + padding_left.x() + padding_right.x()); + output_shape.set(1, input->tensor_shape()[1] * block_y + padding_left.y() + padding_right.y()); + output_shape.set(3, input->tensor_shape()[3] / (block_x * block_y)); + + return output_shape; +} + template <typename T> inline TensorShape extract_shape(T *data) { diff --git a/arm_compute/runtime/CL/CLFunctions.h b/arm_compute/runtime/CL/CLFunctions.h index 935a1ae5d0..cf8911b163 100644 --- a/arm_compute/runtime/CL/CLFunctions.h +++ b/arm_compute/runtime/CL/CLFunctions.h @@ -113,6 +113,7 @@ #include "arm_compute/runtime/CL/functions/CLSobel5x5.h" #include "arm_compute/runtime/CL/functions/CLSobel7x7.h" #include "arm_compute/runtime/CL/functions/CLSoftmaxLayer.h" +#include "arm_compute/runtime/CL/functions/CLSpaceToBatchLayer.h" #include "arm_compute/runtime/CL/functions/CLSplit.h" #include "arm_compute/runtime/CL/functions/CLStridedSlice.h" #include "arm_compute/runtime/CL/functions/CLTableLookup.h" diff --git a/arm_compute/runtime/CL/functions/CLSpaceToBatchLayer.h b/arm_compute/runtime/CL/functions/CLSpaceToBatchLayer.h new file mode 100644 index 0000000000..6478774701 --- /dev/null +++ b/arm_compute/runtime/CL/functions/CLSpaceToBatchLayer.h @@ -0,0 +1,103 @@ +/* + * Copyright (c) 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. + */ +#ifndef __ARM_COMPUTE_CLSPACETOBATCHLAYER_H__ +#define __ARM_COMPUTE_CLSPACETOBATCHLAYER_H__ + +#include "arm_compute/runtime/IFunction.h" + +#include "arm_compute/core/CL/kernels/CLSpaceToBatchLayerKernel.h" +#include "arm_compute/core/Types.h" +#include "arm_compute/runtime/CL/CLTensor.h" + +namespace arm_compute +{ +class ICLTensor; + +/** Basic function to run @ref CLSpaceToBatchLayerKernel. */ +class CLSpaceToBatchLayer : public IFunction +{ +public: + /** Default constructor */ + CLSpaceToBatchLayer(); + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLSpaceToBatchLayer(const CLSpaceToBatchLayer &) = delete; + /** Prevent instances of this class from being copied (As this class contains pointers) */ + CLSpaceToBatchLayer &operator=(const CLSpaceToBatchLayer &) = delete; + /** Allow instances of this class to be moved */ + CLSpaceToBatchLayer(CLSpaceToBatchLayer &&) = default; + /** Allow instances of this class to be moved */ + CLSpaceToBatchLayer &operator=(CLSpaceToBatchLayer &&) = default; + /** Default destructor */ + virtual ~CLSpaceToBatchLayer() = default; + /** Set the input and output tensors. + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape 1-D tensor with shape [M]. Data types supported: S32 + * @param[in] paddings 2-D tensor with shape [2, M]. Data types supported: S32 + * @param[out] output Tensor output. Data types supported: same as @p input + */ + void configure(const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output); + /** Set the input and output tensors. (Static block shape and paddings) + * + * @param[in] input Tensor input. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape_x Block shape x value. + * @param[in] block_shape_y Block shape y value. + * @param[in] padding_left The left padding of the output tensor. + * @param[in] padding_right The right padding of the output tensor. + * @param[out] output Tensor output. Data types supported: same as @p input + */ + void configure(const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right, ICLTensor *output); + /** Static function to check if given info will lead to a valid configuration of @ref CLSpaceToBatchLayer + * + * @param[in] input Tensor input info. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape block shape tensor info with shape [M]. Data types supported: S32 + * @param[in] paddings paddings tensor info with shape [2, M]. Data types supported: S32 + * @param[out] output Tensor output info. Data types supported: same as @p input + * + * @return a status + */ + static Status validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output); + /** Static function to check if given info will lead to a valid configuration of @ref CLSpaceToBatchLayer (Static block shape and paddings) + * + * @param[in] input Tensor input info. Supported tensor rank: 4. Data types supported: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32. + * @param[in] block_shape_x Block shape x value. + * @param[in] block_shape_y Block shape y value. + * @param[in] padding_left The left padding of the output tensor. + * @param[in] padding_right The right padding of the output tensor. + * @param[out] output Tensor output info. Data types supported: same as @p input + * + * @return a status + */ + static Status 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); + + // Inherited methods overridden: + void run() override; + +private: + CLSpaceToBatchLayerKernel _space_to_batch_kernel; /**< SpaceToBatch kernel to run */ + ICLTensor *_output; /**< Output tensor */ + bool _has_padding; /**< Flag to check if the output has padding */ +}; +} // namespace arm_compute +#endif /* __ARM_COMPUTE_CLSPACETOBATCHLAYER_H__ */ |