Compute Library
 21.02
CLDepthConvertLayerKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016-2020 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"
33 #include "src/core/CL/CLValidate.h"
36 #include "support/StringSupport.h"
37 
38 #include <cstddef>
39 #include <set>
40 #include <string>
41 
42 namespace arm_compute
43 {
44 namespace
45 {
46 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, ConvertPolicy policy, uint32_t shift)
47 {
48  ARM_COMPUTE_UNUSED(policy);
50  ARM_COMPUTE_RETURN_ERROR_ON(input == output);
52  1,
57  1,
61  ARM_COMPUTE_RETURN_ERROR_ON_MSG(input->data_type() == output->data_type(), "Input and output data types must be different");
62  ARM_COMPUTE_RETURN_ERROR_ON_MSG(is_data_type_float(input->data_type()) && shift != 0, "Shift is used only with integer non-quantized inputs");
63  ARM_COMPUTE_RETURN_ERROR_ON_MSG(is_data_type_quantized(input->data_type()) && shift != 0, "Shift is used only with integer non-quantized inputs");
64  ARM_COMPUTE_RETURN_ERROR_ON(shift >= 8);
65 
66  // Validate in case of configured output
67  if(output->total_size() > 0)
68  {
70  }
71 
72  return Status{};
73 }
74 } // namespace
75 
76 void CLDepthConvertLayerKernel::configure(const ICLTensor *input, ICLTensor *output, ConvertPolicy policy, uint32_t shift)
77 {
78  configure(CLKernelLibrary::get().get_compile_context(), input, output, policy, shift);
79 }
80 
81 void CLDepthConvertLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, ICLTensor *output, ConvertPolicy policy, uint32_t shift)
82 {
83  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
84 
85  _input = input;
86  _output = output;
87 
88  // Auto initialize output shape if not initialized (We can only auto-configure the shape, datatype must be given)
89  set_shape_if_empty(*output->info(), input->info()->tensor_shape());
90 
91  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), policy, shift));
92 
93  auto padding_info = get_padding_info({ input, output });
94 
95  // Get data sizes
96  const size_t input_size = data_size_from_type(input->info()->data_type());
97  const size_t output_size = data_size_from_type(output->info()->data_type());
98 
99  // Get number of elements to process per iterations
100  const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / input->info()->element_size(), input->info()->dimension(0));
101 
102  // Set build options
103  CLBuildOptions build_opts;
104  build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
105  build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(input->info()->dimension(0) % num_elems_processed_per_iteration));
106  build_opts.add_option("-DDATA_TYPE_IN=" + get_cl_type_from_data_type(input->info()->data_type()));
107  build_opts.add_option("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(output->info()->data_type()));
108  // Conversions from float always SATURATE as out-of-bounds conversion from float->integer is implementation defined
109  build_opts.add_option_if(is_data_type_float(input->info()->data_type()) || policy == ConvertPolicy::SATURATE, "-DSATURATE");
110  build_opts.add_option_if(is_data_type_float(input->info()->data_type()) || is_data_type_float(output->info()->data_type()), "-DIS_DATA_TYPE_FLOAT");
111  build_opts.add_option_if(is_data_type_quantized(input->info()->data_type()), "-DIS_DATA_TYPE_QUANTIZED");
112 
113  // Create kernel
114  const std::string kernel_name = (input_size >= output_size) ? "convert_depth_down" : "convert_depth_up";
115  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
116 
117  // Set shift arg
118  unsigned int idx = 2 * num_arguments_per_3D_tensor(); // Skip the input and output parameters
119  _kernel.setArg(idx++, shift);
120 
121  // Since we have a leftover vector size calculated using the input tensor shape, it is required to
122  // have the input region equal to the tensor shape
123  ValidRegion input_valid_region = input->info()->valid_region();
124  input->info()->set_valid_region(ValidRegion(Coordinates(0, 0), input->info()->tensor_shape()));
125 
126  // Configure kernel
127  Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
128  ICLKernel::configure_internal(win);
129 
130  // Collapse window
131  const Window &full_window = window();
132  Window collapsed_window = full_window.collapse_if_possible(full_window, Window::DimZ);
133  ICLKernel::configure_internal(collapsed_window);
134 
135  // Restore the valid region
136  input->info()->set_valid_region(input_valid_region);
137 
139 
140  // Set config_id for enabling LWS tuning
141  _config_id = kernel_name;
142  _config_id += "_";
143  _config_id += lower_string(string_from_data_type(input->info()->data_type()));
144  _config_id += "_";
145  _config_id += support::cpp11::to_string(input->info()->dimension(0));
146  _config_id += "_";
147  _config_id += support::cpp11::to_string(input->info()->dimension(1));
148  _config_id += "_";
149  _config_id += support::cpp11::to_string(output->info()->dimension(0));
150  _config_id += "_";
151  _config_id += support::cpp11::to_string(output->info()->dimension(1));
152 }
153 
154 Status CLDepthConvertLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *output, ConvertPolicy policy, uint32_t shift)
155 {
156  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, policy, shift));
157 
158  return Status{};
159 }
160 } // namespace arm_compute
bool is_data_type_quantized(DataType dt)
Check if a given data type is of quantized type.
Definition: Utils.h:1168
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
virtual size_t dimension(size_t index) const =0
Return the size of the requested dimension.
static Status validate(const ITensorInfo *input, const ITensorInfo *output, ConvertPolicy policy, uint32_t shift)
Static function to check if given info will lead to a valid configuration of CLDepthConvertLayerKerne...
1 channel, 1 U8 per channel
#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.
virtual DataType data_type() const =0
Data type used for each element of the tensor.
1 channel, 1 F32 per channel
#define ARM_COMPUTE_ERROR_ON(cond)
If the condition is true then an error message is printed and an exception thrown.
Definition: Error.h:466
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
1 channel, 1 U16 per channel
Status class.
Definition: Error.h:52
std::string lower_string(const std::string &val)
Lower a given string.
Definition: Utils.cpp:350
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
Copyright (c) 2017-2021 Arm Limited.
virtual void set_valid_region(const ValidRegion &valid_region)=0
Set the valid region of the tensor.
1 channel, 1 F16 per channel
virtual ValidRegion valid_region() const =0
Valid region of the tensor.
void configure(const ICLTensor *input, ICLTensor *output, ConvertPolicy policy, uint32_t shift)
Set the input and output of the kernel.
1 channel, 1 S32 per channel
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
const std::string & string_from_data_type(DataType dt)
Convert a data type identity into a string.
Definition: Utils.cpp:135
static constexpr unsigned int num_arguments_per_3D_tensor()
Returns the number of arguments enqueued per 3D tensor object.
Definition: ICLKernel.h:214
#define ARM_COMPUTE_UNUSED(...)
To avoid unused variables warnings.
Definition: Error.h:152
Window collapse_if_possible(const Window &full_window, size_t first, size_t last, bool *has_collapsed=nullptr) const
Collapse the dimensions between first and last if possible.
Definition: Window.inl:68
1 channel, 1 U32 per channel
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
quantized, asymmetric fixed-point 8-bit number unsigned
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
Coordinates of an item.
Definition: Coordinates.h:37
std::string kernel_name
std::string get_cl_type_from_data_type(const DataType &dt)
Translates a tensor data type to the appropriate OpenCL type.
Definition: CLHelpers.cpp:37
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
size_t data_size_from_type(DataType data_type)
The size in bytes of the data type.
Definition: Utils.h:106
bool set_shape_if_empty(ITensorInfo &info, const TensorShape &shape)
Set the shape to the specified value if the current assignment is empty.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
1 channel, 1 S16 per channel
bool has_padding_changed(const std::unordered_map< const ITensorInfo *, PaddingSize > &padding_map)
Check if the previously stored padding info has changed after configuring a kernel.
Definition: Utils.cpp:528
CLCompileContext class.
quantized, symmetric per channel fixed-point 8-bit number
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(...)
Definition: Validate.h:443
static constexpr size_t DimZ
Alias for dimension 2 also known as Z dimension.
Definition: Window.h:47
Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context...
#define ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:792
std::unordered_map< const ITensorInfo *, PaddingSize > get_padding_info(std::initializer_list< const ITensorInfo *> infos)
Stores padding information before configuring a kernel.
Definition: Utils.cpp:513
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
Wrapper to configure the Khronos OpenCL C++ header.
unsigned int num_elems_processed_per_iteration
#define ARM_COMPUTE_RETURN_ERROR_ON_MSG(cond, msg)
If the condition is true, an error is returned.
Definition: Error.h:244
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
unsigned int adjust_vec_size(unsigned int vec_size, size_t dim0)
Returns the adjusted vector size in case it is less than the input&#39;s first dimension, getting rounded down to its closest valid vector size.
Definition: Utils.h:1358
Container for valid region of a window.
Definition: Types.h:188
signed 8-bit number
Describe a multidimensional execution window.
Definition: Window.h:39
ConvertPolicy
Policy to handle overflow.
Definition: Types.h:385
bool is_data_type_float(DataType dt)
Check if a given data type is of floating point type.
Definition: Utils.h:1148