Compute Library
 20.05
CLSelectKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2018-2020 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 
32 #include "arm_compute/core/Utils.h"
34 
35 #include "support/StringSupport.h"
36 
37 namespace arm_compute
38 {
39 namespace
40 {
41 Status validate_arguments(const ITensorInfo *c, const ITensorInfo *x, const ITensorInfo *y, const ITensorInfo *output)
42 {
49 
50  const bool is_same_rank = (c->tensor_shape().num_dimensions() == x->tensor_shape().num_dimensions());
51  ARM_COMPUTE_RETURN_ERROR_ON(is_same_rank && (x->tensor_shape() != c->tensor_shape()));
52  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])));
53 
54  if(output != nullptr && output->total_size() != 0)
55  {
58  }
59 
60  return Status{};
61 }
62 
63 std::pair<Status, Window> validate_and_configure_window(ITensorInfo *c, ITensorInfo *x, ITensorInfo *y, ITensorInfo *output)
64 {
65  if(output != nullptr)
66  {
67  // Output tensor auto initialization if not yet initialized
68  auto_init_if_empty(*output, *x->clone());
69  }
70 
71  const bool is_same_rank = (c->tensor_shape().num_dimensions() == x->tensor_shape().num_dimensions());
72 
73  const unsigned int num_elems_processed_per_iteration = 16 / x->element_size();
74 
75  // Configure kernel window
77  AccessWindowHorizontal x_access(x, 0, num_elems_processed_per_iteration);
78  AccessWindowHorizontal y_access(y, 0, num_elems_processed_per_iteration);
79  bool window_changed = update_window_and_padding(win, x_access, y_access);
80 
81  // Update window for condition
82  if(is_same_rank)
83  {
84  AccessWindowHorizontal c_access(c, 0, num_elems_processed_per_iteration);
85  window_changed = window_changed || update_window_and_padding(win, c_access);
86  }
87 
88  // Update window for output
89  if(output != nullptr)
90  {
91  AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration);
92  window_changed = window_changed || update_window_and_padding(win, output_access);
93  output_access.set_valid_region(win, x->valid_region());
94  }
95 
96  Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{};
97  return std::make_pair(err, win);
98 }
99 } // namespace
100 
102  : _c(nullptr), _x(nullptr), _y(nullptr), _output(nullptr), _has_same_rank(false)
103 {
104 }
105 void CLSelectKernel::configure(const ICLTensor *c, const ICLTensor *x, const ICLTensor *y, ICLTensor *output)
106 {
107  configure(CLKernelLibrary::get().get_compile_context(), c, x, y, output);
108 }
109 
110 void CLSelectKernel::configure(const CLCompileContext &compile_context, const ICLTensor *c, const ICLTensor *x, const ICLTensor *y, ICLTensor *output)
111 {
112  ARM_COMPUTE_ERROR_ON_NULLPTR(c, x, y, output);
113  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(c->info(), x->info(), y->info(), output->info()));
114 
115  _c = c;
116  _x = x;
117  _y = y;
118  _output = output;
119  _has_same_rank = (c->info()->tensor_shape().num_dimensions() == x->info()->tensor_shape().num_dimensions());
120 
121  const unsigned int num_elems_processed_per_iteration = 16 / x->info()->element_size();
122 
123  // Set build options
124  CLBuildOptions build_opts;
125  build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(x->info()->data_type()));
126  build_opts.add_option("-DSELECT_DATA_TYPE=" + get_cl_select_type_from_data_type(x->info()->data_type()));
128 
129  // Create kernel
130  std::string kernel_name = "select";
131  if(_has_same_rank)
132  {
133  kernel_name += "_same_rank";
134  }
135  else
136  {
137  const bool is_input_rank_greater_than_two = x->info()->tensor_shape().num_dimensions() > 2;
138  if(is_input_rank_greater_than_two)
139  {
140  const size_t width = x->info()->tensor_shape().x();
141  const size_t height = x->info()->tensor_shape().y();
142  const size_t outer_size = x->info()->tensor_shape()[x->info()->tensor_shape().num_dimensions() - 1];
143  const size_t depth_size = x->info()->tensor_shape().total_size() / (width * height * outer_size);
144  build_opts.add_option("-DDEPTH_SIZE=" + support::cpp11::to_string(depth_size));
145  }
146  kernel_name += "_different_rank";
147  kernel_name += is_input_rank_greater_than_two ? "_n" : "_2";
148  }
149  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
150 
151  // Configure kernel window
152  auto win_config = validate_and_configure_window(c->info(), x->info(), y->info(), output->info());
153  ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
154  ICLKernel::configure_internal(win_config.second);
155 
156  _config_id = "select_";
157  _config_id += string_from_data_type(x->info()->data_type());
158  _config_id += "_";
159  _config_id += support::cpp11::to_string(x->info()->dimension(0));
160  _config_id += "_";
161  _config_id += support::cpp11::to_string(x->info()->dimension(1));
162  _config_id += "_";
163  _config_id += support::cpp11::to_string(x->info()->dimension(2));
164 }
165 
166 Status CLSelectKernel::validate(const ITensorInfo *c, const ITensorInfo *x, const ITensorInfo *y, const ITensorInfo *output)
167 {
169  ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(c->clone().get(), x->clone().get(), y->clone().get(), output->clone().get()).first);
170  return Status{};
171 }
172 
173 void CLSelectKernel::run(const arm_compute::Window &window, cl::CommandQueue &queue)
174 {
177 
179  Window slice = collapsed.first_slice_window_3D();
180 
181  if(!_has_same_rank)
182  {
183  Window vector_slice = window.first_slice_window_1D();
184  vector_slice.set(Window::DimX, Window::Dimension(0, 0, 0));
185  unsigned int idx = 0;
186  add_1D_tensor_argument(idx, _c, vector_slice);
187  }
188 
189  do
190  {
191  unsigned int idx = _has_same_rank ? 0 : num_arguments_per_1D_tensor();
192  if(_has_same_rank)
193  {
194  add_3D_tensor_argument(idx, _c, slice);
195  }
196  add_3D_tensor_argument(idx, _x, slice);
197  add_3D_tensor_argument(idx, _y, slice);
198  add_3D_tensor_argument(idx, _output, slice);
199 
200  enqueue(queue, *this, slice, lws_hint());
201  }
202  while(collapsed.slide_window_slice_3D(slice));
203 }
204 } // 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:184
CLSelectKernel()
Default constructor.
#define ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(tensor)
Definition: CLValidate.h:34
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:39
const StringSet & options() const
Gets the current options list set.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:545
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:247
std::string get_cl_select_type_from_data_type(const DataType &dt)
Translates a tensor data type to the appropriate OpenCL select type.
Definition: CLHelpers.cpp:139
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_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:792
void configure(const ICLTensor *c, const ICLTensor *x, const ICLTensor *y, ICLTensor *output)
Initialise the kernel's inputs and output.
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
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:75
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
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps=Steps(), bool skip_border=false, BorderSize border_size=BorderSize())
Calculate the maximum window for a given tensor shape and border setting.
Definition: Helpers.cpp:28
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:158
Copyright (c) 2017-2020 ARM Limited.
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...
Definition: Helpers.inl:202
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:81
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:387
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
bool update_window_and_padding(Window &win, Ts &&... patterns)
Update window and padding size for each of the access patterns.
Definition: Helpers.h:437
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
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(...)
Definition: Validate.h:443
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.
std::string kernel_name
size_t total_size() const
Collapses all dimensions to a single linear total size.
Definition: TensorShape.h:171
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 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'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
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:333
CLCompileContext class.
#define ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(...)
Definition: Validate.h:163
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
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
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:122
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
unsigned int num_elems_processed_per_iteration
T y() const
Alias to access the size of the second dimension.
Definition: Dimensions.h:86
void add_1D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 1D tensor's parameters to the object's kernel's arguments starting from the index idx.
Definition: ICLKernel.h:110
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:289
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205
Describe a multidimensional execution window.
Definition: Window.h:39
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
Window first_slice_window_1D() const
First 1D slice of the window.
Definition: Window.h:273
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)