Compute Library
 21.11
CLSelectKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2018-2021 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"
32 #include "src/core/CL/CLValidate.h"
35 
36 #include "support/StringSupport.h"
37 
38 namespace arm_compute
39 {
40 namespace
41 {
42 Status validate_arguments(const ITensorInfo *c, const ITensorInfo *x, const ITensorInfo *y, const ITensorInfo *output)
43 {
44  ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(c, x, y, output);
50 
51  const bool is_same_rank = (c->tensor_shape().num_dimensions() == x->tensor_shape().num_dimensions());
52  ARM_COMPUTE_RETURN_ERROR_ON(is_same_rank && (x->tensor_shape() != c->tensor_shape()));
53  ARM_COMPUTE_RETURN_ERROR_ON(!is_same_rank && ((c->tensor_shape().num_dimensions() > 1) || (c->tensor_shape().x() != x->tensor_shape()[x->tensor_shape().num_dimensions() - 1])));
54 
55  if(output->total_size() != 0)
56  {
59  }
60 
61  return Status{};
62 }
63 } // namespace
64 
66  : _c(nullptr), _x(nullptr), _y(nullptr), _output(nullptr), _has_same_rank(false)
67 {
69 }
70 
71 void CLSelectKernel::configure(const CLCompileContext &compile_context, const ICLTensor *c, const ICLTensor *x, const ICLTensor *y, ICLTensor *output)
72 {
73  ARM_COMPUTE_ERROR_ON_NULLPTR(c, x, y, output);
74  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(c->info(), x->info(), y->info(), output->info()));
75 
76  _c = c;
77  _x = x;
78  _y = y;
79  _output = output;
80  _has_same_rank = (c->info()->tensor_shape().num_dimensions() == x->info()->tensor_shape().num_dimensions());
81 
82  auto padding_info = get_padding_info({ c, x, y, output });
83  const unsigned int vec_size_x = adjust_vec_size(16 / x->info()->element_size(), x->info()->dimension(0));
84  const int vec_size_x_leftovers = output->info()->dimension(0) % vec_size_x;
85 
86  // Set build options
87  CLBuildOptions build_opts;
88  build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(x->info()->element_size()));
89  build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(vec_size_x));
90  build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(vec_size_x_leftovers));
91 
92  // Create kernel
93  std::string kernel_name = "select";
94  if(_has_same_rank)
95  {
96  kernel_name += "_same_rank";
97  }
98  else
99  {
100  const bool is_input_rank_greater_than_two = x->info()->tensor_shape().num_dimensions() > 2;
101  if(is_input_rank_greater_than_two)
102  {
103  const size_t width = x->info()->tensor_shape().x();
104  const size_t height = x->info()->tensor_shape().y();
105  const size_t outer_size = x->info()->tensor_shape()[x->info()->tensor_shape().num_dimensions() - 1];
106  const size_t depth_size = x->info()->tensor_shape().total_size() / (width * height * outer_size);
107  build_opts.add_option("-DDEPTH_SIZE=" + support::cpp11::to_string(depth_size));
108  }
109  kernel_name += "_different_rank";
110  kernel_name += is_input_rank_greater_than_two ? "_n" : "_2";
111  }
112  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
113 
114  // Configure kernel window
115  auto_init_if_empty(*output->info(), *x->info()->clone());
116  Window win = calculate_max_window(*x->info(), Steps(vec_size_x));
117  ICLKernel::configure_internal(win);
118 
119  _config_id = "select_";
120  _config_id += string_from_data_type(x->info()->data_type());
121  _config_id += "_";
122  _config_id += support::cpp11::to_string(x->info()->dimension(0));
123  _config_id += "_";
124  _config_id += support::cpp11::to_string(x->info()->dimension(1));
125  _config_id += "_";
126  _config_id += support::cpp11::to_string(x->info()->dimension(2));
128 }
129 
130 Status CLSelectKernel::validate(const ITensorInfo *c, const ITensorInfo *x, const ITensorInfo *y, const ITensorInfo *output)
131 {
132  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(c, x, y, output));
133  return Status{};
134 }
135 
136 void CLSelectKernel::run(const arm_compute::Window &window, cl::CommandQueue &queue)
137 {
140 
142  Window slice = collapsed.first_slice_window_3D();
143 
144  if(!_has_same_rank)
145  {
146  Window vector_slice = window.first_slice_window_1D();
147  vector_slice.set(Window::DimX, Window::Dimension(0, 0, 0));
148  unsigned int idx = 0;
149  add_1D_tensor_argument(idx, _c, vector_slice);
150  }
151 
152  do
153  {
154  unsigned int idx = _has_same_rank ? 0 : num_arguments_per_1D_tensor();
155  if(_has_same_rank)
156  {
157  add_3D_tensor_argument(idx, _c, slice);
158  }
159  add_3D_tensor_argument(idx, _x, slice);
160  add_3D_tensor_argument(idx, _y, slice);
161  add_3D_tensor_argument(idx, _output, slice);
162 
163  enqueue(queue, *this, slice, lws_hint());
164  }
165  while(collapsed.slide_window_slice_3D(slice));
166 }
167 } // namespace arm_compute
static constexpr unsigned int num_arguments_per_1D_tensor()
Returns the number of arguments enqueued per 1D tensor object.
Definition: ICLKernel.h:240
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
CLSelectKernel()
Default constructor.
#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.
void enqueue(cl::CommandQueue &queue, ICLKernel &kernel, const Window &window, const cl::NDRange &lws_hint=CLKernelLibrary::get().default_ndrange(), bool use_dummy_work_items=false)
Add the kernel to the command queue with the given window.
Definition: ICLKernel.cpp:32
const StringSet & options() const
Gets the current options list set.
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:318
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.
#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
Store the tensor's metadata.
Definition: ITensorInfo.h:40
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Describe one of the image's dimensions with a start, end and step.
Definition: Window.h:77
Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context...
Status class.
Definition: Error.h:52
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
void add_3D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 3D tensor's parameters to the object's kernel's arguments starting from the index idx...
Definition: ICLKernel.h:214
Copyright (c) 2017-2021 Arm Limited.
#define ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(...)
Definition: Validate.h:159
void add_option(std::string option)
Adds option to the existing build option list.
T x() const
Alias to access the size of the first dimension.
Definition: Dimensions.h:87
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:391
const std::string & string_from_data_type(DataType dt)
Convert a data type identity into a string.
Definition: Utils.cpp:135
static constexpr size_t DimX
Alias for dimension 0 also known as X dimension.
Definition: Window.h:43
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
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
void run(const Window &window, cl::CommandQueue &queue) override
Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue...
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
size_t total_size() const
Collapses all dimensions to a single linear total size.
Definition: TensorShape.h:172
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...
void configure(const CLCompileContext &compile_context, const ICLTensor *c, const ICLTensor *x, const ICLTensor *y, ICLTensor *output)
Initialise the kernel&#39;s inputs and output.
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.
static Status validate(const ITensorInfo *c, const ITensorInfo *x, const ITensorInfo *y, const ITensorInfo *output)
Static function to check if given info will lead to a valid configuration of CLSelectKernel.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
void set(size_t dimension, const Dimension &dim)
Set the values of a given dimension.
Definition: Window.inl:49
Elementeise CL kernel type.
Definition: CLTypes.h:84
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:335
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:915
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:533
CLCompileContext class.
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(...)
Definition: Validate.h:439
static constexpr size_t DimZ
Alias for dimension 2 also known as Z dimension.
Definition: Window.h:47
unsigned int num_dimensions() const
Returns the effective dimensionality of the tensor.
Definition: Dimensions.h:143
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:541
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:105
#define ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:788
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:518
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
T y() const
Alias to access the size of the second dimension.
Definition: Dimensions.h:92
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:1171
void add_1D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 1D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:166
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:291
std::string kernel_name
Describe a multidimensional execution window.
Definition: Window.h:39
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:201
Window first_slice_window_1D() const
First 1D slice of the window.
Definition: Window.h:275
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)