blob: 5e3d815d8c8508f0e1020a0756226f78da74df60 [file] [log] [blame]
Gian Marco Iodice7e4b2392018-02-22 16:17:20 +00001/*
2 * Copyright (c) 2018 ARM Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24#ifndef __ARM_COMPUTE_CLWINOGRADFILTERTRANSFORMKERNEL_H__
25#define __ARM_COMPUTE_CLWINOGRADFILTERTRANSFORMKERNEL_H__
26
27#include "arm_compute/core/CL/ICLKernel.h"
28
29namespace arm_compute
30{
31class ICLTensor;
32
33/** Interface for the Winograd filter transform kernel. */
34class CLWinogradFilterTransformKernel : public ICLKernel
35{
36public:
37 /** Default constructor */
38 CLWinogradFilterTransformKernel();
39 /** Prevent instances of this class from being copied (As this class contains pointers) */
40 CLWinogradFilterTransformKernel(const CLWinogradFilterTransformKernel &) = delete;
41 /** Prevent instances of this class from being copied (As this class contains pointers) */
42 CLWinogradFilterTransformKernel &operator=(const CLWinogradFilterTransformKernel &) = delete;
43 /** Allow instances of this class to be moved */
44 CLWinogradFilterTransformKernel(CLWinogradFilterTransformKernel &&) = default;
45 /** Allow instances of this class to be moved */
46 CLWinogradFilterTransformKernel &operator=(CLWinogradFilterTransformKernel &&) = default;
47 /** Default destructor */
48 ~CLWinogradFilterTransformKernel() = default;
49 /** Set the input and output tensor.
50 *
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000051 * @note Winograd filter transform supports the following configurations:
Gian Marco Iodicee52a3002018-04-11 15:59:10 +010052 * F(output tile, kernel size):F(2x2, 3x3), F(4x4, 3x3), F(4x4, 5x5)
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000053 * Strides: only unit strides
Giorgio Arena80d65d82018-06-08 16:30:00 +010054 * Data Layout: NCHW for all configurations, NHWC for F(4x4, 3x3) and F(4x4, 5x5)
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000055 *
Giorgio Arenadcb5b282018-04-25 12:07:29 +010056 * @param[in] input Source tensor. The input is a 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] (NCHW data layout) or [IFM, kernel_x, kernel_y, OFM] (NHWC data layout). Data types supported: F32.
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000057 * @param[out] output The output tensor. The shape for this tensor can be calculated using the utility function @p compute_winograd_filter_transform_shape. Data types supported: Same as @p input
58 * @param[in] winograd_info Contains Winograd's information described in @ref WinogradInfo
Gian Marco Iodice7e4b2392018-02-22 16:17:20 +000059 */
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000060 void configure(const ICLTensor *input, ICLTensor *output, const WinogradInfo &winograd_info);
Gian Marco Iodice7e4b2392018-02-22 16:17:20 +000061 /** Static function to check if given info will lead to a valid configuration of @ref CLWinogradFilterTransformKernel
62 *
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000063 * @note Winograd filter transform supports the following configurations:
Gian Marco Iodicee52a3002018-04-11 15:59:10 +010064 * F(output tile, kernel size):F(2x2, 3x3), F(4x4, 3x3), F(4x4, 5x5)
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000065 * Strides: only unit strides
Giorgio Arena80d65d82018-06-08 16:30:00 +010066 * Data Layout: NCHW for all configurations, NHWC for F(4x4, 3x3) and F(4x4, 5x5)
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000067 *
Giorgio Arenadcb5b282018-04-25 12:07:29 +010068 * @param[in] input Source tensor. The input is a 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM] (NCHW data layout) or [IFM, kernel_x, kernel_y, OFM] (NHWC data layout). Data types supported: F32.
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000069 * @param[out] output The output tensor. The shape for this tensor can be calculated using the utility function @p compute_winograd_filter_transform_shape. Data types supported: Same as @p input
70 * @param[in] winograd_info Contains Winograd's information described in @ref WinogradInfo
Gian Marco Iodice7e4b2392018-02-22 16:17:20 +000071 *
72 * @return a status
73 */
Gian Marco Iodice247f52c2018-03-22 11:24:56 +000074 static Status validate(const ITensorInfo *input, const ITensorInfo *output, const WinogradInfo &winograd_info);
Gian Marco Iodice7e4b2392018-02-22 16:17:20 +000075
76 // Inherited methods overridden:
77 void run(const Window &window, cl::CommandQueue &queue) override;
78
79private:
80 const ICLTensor *_input;
81 ICLTensor *_output;
82};
83} // namespace arm_compute
84#endif /*__ARM_COMPUTE_CLWINOGRADFILTERTRANSFORMKERNEL_H__ */