Compute Library
 21.02
CLDepthwiseConvolutionLayerReshapeWeightsKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2019-2020 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  */
25 
31 #include "arm_compute/core/Utils.h"
34 #include "src/core/CL/CLValidate.h"
35 #include "src/core/CL/ICLKernel.h"
38 #include "support/StringSupport.h"
39 
40 namespace arm_compute
41 {
42 namespace
43 {
44 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
45 {
47  const size_t idx_w = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH);
48  const size_t idx_h = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT);
49 
52  ARM_COMPUTE_RETURN_ERROR_ON(info.c0 != 4);
53  ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(idx_h) != 3);
54  ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(idx_w) != 3);
55  ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
56 
57  if(output->total_size() != 0)
58  {
61  ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), reshaped_weights_shape);
63  }
64 
65  return Status{};
66 }
67 
68 std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
69 {
71  auto_init_if_empty(*output, reshaped_input_shape, 1, input->data_type(), input->quantization_info());
72 
73  Window win = calculate_max_window(*input, Steps(info.c0));
74  AccessWindowHorizontal weights_access(input, 0, info.c0);
75  const bool window_changed = update_window_and_padding(win, weights_access);
76 
77  output->set_valid_region(ValidRegion(Coordinates(), output->tensor_shape()));
78  Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
79  return std::make_pair(err, win);
80 }
81 } // namespace
82 
84  : _input(nullptr), _output(nullptr)
85 {
86 }
87 
89 {
90  configure(CLKernelLibrary::get().get_compile_context(), input, output, info);
91 }
92 
94 {
95  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
97  auto win_config = validate_and_configure_window(input->info(), output->info(), info);
98  ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
99 
100  ICLKernel::configure_internal(win_config.second);
101 
102  _input = input;
103  _output = output;
104 
105  // Build the kernel
106  CLBuildOptions build_opts;
107  build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(info.c0));
108  build_opts.add_option("-DDST_WIDTH=" + support::cpp11::to_string(_output->info()->dimension(0)));
109  build_opts.add_option_if(info.transpose, "-DTRANSPOSE");
110  build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(input->info()->element_size()));
111 
112  _kernel = create_kernel(compile_context, "depthwise_convolution_reshape_weights", build_opts.options());
113 }
114 
116 {
117  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, info));
118  ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get(), info).first);
119  return Status{};
120 }
121 
123 {
126 
127  unsigned int idx = 0;
128  add_3D_tensor_argument(idx, _input, window);
129  add_2D_tensor_argument(idx, _output, window);
130  enqueue(queue, *this, window, lws_hint());
131 }
132 } // namespace arm_compute
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
#define ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(tensor)
Definition: CLValidate.h:35
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
#define ARM_COMPUTE_RETURN_ERROR_ON_DATA_LAYOUT_NOT_IN(t,...)
Definition: Validate.h:746
void enqueue(IGCKernel &kernel, const Window &window, const gles::NDRange &lws=gles::NDRange(1U, 1U, 1U))
Add the kernel to the command queue with the given window.
Definition: IGCKernel.cpp:41
virtual size_t dimension(size_t index) const =0
Return the size of the requested dimension.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(...)
Definition: Validate.h:610
const StringSet & options() const
Gets the current options list set.
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:276
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
std::string to_string(T &&value)
Convert integer and float values to string.
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
Store the tensor&#39;s metadata.
Definition: ITensorInfo.h:40
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
static Status validate(const ITensorInfo *input, const ITensorInfo *output, const DepthwiseConvolutionReshapeInfo &info)
Static function to check if given info will lead to a valid configuration of CLDepthwiseConvolutionLa...
Status class.
Definition: Error.h:52
TensorShape compute_reshaped_depthwise_weights_shape(const ITensorInfo &input, const DepthwiseConvolutionReshapeInfo &info)
Calculate the reshaped shape of the weights to use in depthwise convolution.
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(...)
Definition: Validate.h:288
void add_3D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 3D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:172
Copyright (c) 2017-2021 Arm Limited.
#define ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(...)
Definition: Validate.h:163
void add_option(std::string option)
Adds option to the existing build option list.
cl::Kernel create_kernel(const CLCompileContext &ctx, const std::string &kernel_name, const std::set< std::string > &build_opts=std::set< std::string >())
Creates an opencl kernel using a compile context.
Definition: CLHelpers.cpp:403
bool update_window_and_padding(Window &win, Ts &&... patterns)
Update window and padding size for each of the access patterns.
Definition: WindowHelpers.h:46
bool auto_init_if_empty(ITensorInfo &info, const TensorShape &shape, int num_channels, DataType data_type, QuantizationInfo quantization_info=QuantizationInfo())
Auto initialize the tensor info (shape, number of channels and data type) if the current assignment i...
virtual std::unique_ptr< T > clone() const =0
Provide a clone of the current object of class T.
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
void add_option_if(bool cond, std::string option)
Adds option if a given condition is true;.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
unsigned int c0
Number of channels processed by the depth-wise convolution.
Definition: Types.h:1938
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
CLCompileContext class.
void configure(const ICLTensor *input, ICLTensor *output, const DepthwiseConvolutionReshapeInfo &info)
Initialize the function&#39;s source and destination.
void add_2D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 2D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:148
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
#define ARM_COMPUTE_CREATE_ERROR(error_code, msg)
Creates an error with a given message.
Definition: Error.h:159
Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context...
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:545
Num samples, height, width, channels.
std::string get_cl_unsigned_type_from_element_size(size_t element_size)
Translates the element size to an unsigned integer data type.
Definition: CLHelpers.cpp:103
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
size_t get_data_layout_dimension_index(const DataLayout data_layout, const DataLayoutDimension data_layout_dimension)
Get the index of the given dimension.
Definition: Helpers.inl:193
Describe a multidimensional execution window.
Definition: Window.h:39
void run(const Window &window, cl::CommandQueue &queue) override
Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue...
bool transpose
True if the block MxC0 (where M is the area of the filter i.e.
Definition: Types.h:1939
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205