Compute Library
 20.08
NEIm2ColKernel.h
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2017-2020 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_NEIM2COLKERNEL_H
25 #define ARM_COMPUTE_NEIM2COLKERNEL_H
26 
28 
29 namespace arm_compute
30 {
31 class ITensor;
32 class Size2D;
33 
34 /** Interface for the im2col reshape kernel.
35  *
36  * Rearranges image blocks into columns. It is used to strip out each convolution block to a single column.
37  * It is used to transform a convolution to a plain matrix multiplication.
38  *
39  * For example taking into account the image below and assuming 3x3 image blocks with stride of 1 we have:
40  *
41  * @f[
42  * \left( \begin{array}{cccc}
43  * a00 & a01 & a02 & a03 \\
44  * a10 & a11 & a12 & a13 \\
45  * a20 & a21 & a22 & a23 \\
46  * a30 & a31 & a32 & a33 \\
47  * \end{array} \right)
48  * \rightarrow
49  * \left( \begin{array}{ccccccccc}
50  * a00 & a01 & a02 & a10 & a11 & a12 & a20 & a21 & a22 \\
51  * a01 & a02 & a03 & a11 & a12 & a13 & a21 & a22 & a23 \\
52  * a10 & a11 & a12 & a20 & a21 & a22 & a30 & a31 & a32 \\
53  * a11 & a12 & a13 & a21 & a22 & a23 & a31 & a32 & a33 \\
54  * \end{array} \right)
55  * @f]
56  */
57 class NEIm2ColKernel : public INEKernel
58 {
59 public:
60  const char *name() const override
61  {
62  return "NEIm2ColKernel";
63  }
64  /** Default constructor */
66  /** Prevent instances of this class from being copied (As this class contains pointers) */
67  NEIm2ColKernel(const NEIm2ColKernel &) = delete;
68  /** Prevent instances of this class from being copied (As this class contains pointers) */
69  NEIm2ColKernel &operator=(const NEIm2ColKernel &) = delete;
70  /** Allow instances of this class to be moved */
71  NEIm2ColKernel(NEIm2ColKernel &&) = default;
72  /** Allow instances of this class to be moved */
74  /** Default destructor */
75  ~NEIm2ColKernel() = default;
76 
77  /** Set the input and output of the kernel.
78  *
79  * @param[in] input The input tensor to convert. 3 lower dimensions represent a single input [width, height, IFM],
80  * while every optional dimension from 4 and above represent a batch of inputs.
81  * Data types supported: QASYMM8/QASYMM8_SIGNED/BFLOAT16/F16/F32
82  * Note: QASYMM8/QASYMM8_SIGNED works only for has_bias = false
83  * @param[out] output The output tensor. Data types supported: Same as @p input
84  * @param[in] kernel_dims The kernel dimensions (width and height).
85  * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo.
86  * @param[in] has_bias In case biases are provided expands the matrix with 1.
87  * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
88  * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is not supported
89  */
90  void configure(const ITensor *input, ITensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info,
91  bool has_bias, const Size2D &dilation = Size2D(1U, 1U), unsigned int num_groups = 1);
92  /** Static function to check if given info will lead to a valid configuration of @ref NEIm2ColKernel
93  *
94  * @param[in] input The input tensor to convert. 3 lower dimensions represent a single input [width, height, IFM],
95  * while every optional dimension from 4 and above represent a batch of inputs.
96  * Data types supported: QASYMM8/QASYMM8_SIGNED/BFLOAT16/F16/F32
97  * Note: QASYMM8/QASYMM8_SIGNED works only for has_bias = false
98  * @param[in] output The output tensor. Data types supported: Same as @p input
99  * @param[in] kernel_dims The kernel dimensions (width and height).
100  * @param[in] conv_info Contains padding and stride information described in @ref PadStrideInfo.
101  * @param[in] has_bias In case biases are provided expands the matrix with 1.
102  * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1).
103  * @param[in] num_groups (Optional) Number of groups when performing a grouped convolution. num_groups != 1 is not supported
104  *
105  * @return a status
106  */
107  static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info,
108  bool has_bias, const Size2D &dilation = Size2D(1U, 1U), unsigned int num_groups = 1);
109 
110  // Inherited methods overridden:
111  void run(const Window &window, const ThreadInfo &info) override;
112 
113 private:
114  /** Template function to run im2col
115  *
116  * @param[in] window Region on which to execute the kernel. (Must be a valid region of the window returned by window()).
117  */
118  template <typename T, bool has_pads, bool is_nchw>
119  void run_im2col(const Window &window);
120 
121  /** Common signature for all the specialised im2col functions
122  *
123  * @param[in] window Region on which to execute the kernel.
124  */
125  using Im2ColFunctionPtr = void (NEIm2ColKernel::*)(const Window &window);
126 
127  Im2ColFunctionPtr _func;
128  const ITensor *_input;
129  ITensor *_output;
130  std::pair<unsigned int, unsigned int> _convolved_dims;
131  PadStrideInfo _conv_info;
132  unsigned int _kernel_width;
133  unsigned int _kernel_height;
134  bool _has_bias;
135  Size2D _dilation;
136  DataLayout _data_layout;
137 };
138 } // namespace arm_compute
139 #endif /*ARM_COMPUTE_NEIM2COLKERNEL_H */
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
const char * name() const override
Name of the kernel.
Common interface for all kernels implemented in C++.
Definition: ICPPKernel.h:38
Interface for the im2col reshape kernel.
Store the tensor's metadata.
Definition: ITensorInfo.h:40
NEIm2ColKernel & operator=(const NEIm2ColKernel &)=delete
Prevent instances of this class from being copied (As this class contains pointers)
Status class.
Definition: Error.h:52
Interface for NEON tensor.
Definition: ITensor.h:36
Copyright (c) 2017-2020 Arm Limited.
const unsigned int num_groups
Definition: Im2Col.cpp:148
NEIm2ColKernel()
Default constructor.
Padding and stride information class.
Definition: Types.h:689
void run(const Window &window, const ThreadInfo &info) override
Execute the kernel on the passed window.
Information about executing thread and CPU.
Definition: CPPTypes.h:235
~NEIm2ColKernel()=default
Default destructor.
Class for specifying the size of an image or rectangle.
Definition: Size2D.h:34
void configure(const ITensor *input, ITensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias, const Size2D &dilation=Size2D(1U, 1U), unsigned int num_groups=1)
Set the input and output of the kernel.
DataLayout
[DataLayout enum definition]
Definition: Types.h:120
static Status validate(const ITensorInfo *input, const ITensorInfo *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias, const Size2D &dilation=Size2D(1U, 1U), unsigned int num_groups=1)
Static function to check if given info will lead to a valid configuration of NEIm2ColKernel.
Describe a multidimensional execution window.
Definition: Window.h:39