Compute Library
 21.11
ClBatchConcatenateKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2019-2021 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 
30 #include "arm_compute/core/Utils.h"
31 #include "src/core/CL/CLValidate.h"
33 #include "support/Cast.h"
34 
35 #include "support/StringSupport.h"
36 
37 namespace arm_compute
38 {
39 namespace opencl
40 {
41 namespace kernels
42 {
43 namespace
44 {
45 Status validate_arguments(const ITensorInfo *src, unsigned int batch_offset, const ITensorInfo *dst)
46 {
51 
52  ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimX) != dst->dimension(Window::DimX));
53  ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimY) != dst->dimension(Window::DimY));
54  ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(Window::DimZ) != dst->dimension(Window::DimZ));
55  ARM_COMPUTE_RETURN_ERROR_ON(src->dimension(3) + batch_offset > dst->dimension(3));
57 
58  return Status{};
59 }
60 } // namespace
61 
63 {
65 }
66 
67 void ClBatchConcatenateKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int batch_offset, ITensorInfo *dst)
68 {
70  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, batch_offset, dst));
71 
72  auto padding_info = get_padding_info({ src, dst });
73 
74  _batch_offset = batch_offset;
75 
76  const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / src->element_size(), src->dimension(0));
77 
78  // Add build options
79  CLBuildOptions build_opts;
80  build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->data_type()));
81  build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
82  build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(src->dimension(0) % num_elems_processed_per_iteration));
84  {
85  const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
86  const UniformQuantizationInfo oq_info = dst->quantization_info().uniform();
87 
88  build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq_info.offset));
89  build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
90  build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq_info.scale));
91  build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
92  }
93 
94  // Create kernel
95  _kernel = create_kernel(compile_context, "concatenate", build_opts.options());
96 
97  // Configure kernel window
98  auto win = calculate_max_window(*dst, Steps(num_elems_processed_per_iteration));
99  win.set(3, Window::Dimension(0, src->tensor_shape()[3], 1));
100  ICLKernel::configure_internal(win);
101 
102  // Set config_id for enabling LWS tuning
103  _config_id = "concatenate_";
104  _config_id += support::cpp11::to_string(3);
105  _config_id += "_";
106  _config_id += support::cpp11::to_string(batch_offset);
107  _config_id += "_";
108  _config_id += support::cpp11::to_string(src->dimension(0));
109  _config_id += "_";
110  _config_id += support::cpp11::to_string(src->dimension(1));
111  _config_id += "_";
112  _config_id += support::cpp11::to_string(src->dimension(2));
113  _config_id += "_";
114  _config_id += support::cpp11::to_string(src->dimension(3));
115 
117 }
118 
120  unsigned int batch_offset,
121  const arm_compute::ITensorInfo *dst)
122 {
123  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, batch_offset, dst));
124  return Status{};
125 }
126 
127 void ClBatchConcatenateKernel::run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue)
128 {
131 
132  const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
133  auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
134 
136 
137  const int offset_to_first_elements_in_bytes = _batch_offset * dst->info()->strides_in_bytes()[3];
138 
139  unsigned int idx = 2 * num_arguments_per_3D_tensor(); // Skip the src and dst parameters
140  _kernel.setArg<cl_int>(idx, offset_to_first_elements_in_bytes);
141 
142  do
143  {
144  unsigned int idx = 0;
145  add_3D_tensor_argument(idx, src, slice);
146  add_3D_tensor_argument(idx, dst, slice);
147  enqueue(queue, *this, slice, lws_hint());
148  }
149  while(window.slide_window_slice_3D(slice));
150 }
151 } // namespace opencl
152 } // namespace kernels
153 } // 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
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
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:318
#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&#39;s metadata.
Definition: ITensorInfo.h:40
static Status validate(const ITensorInfo *src, unsigned int batch_offset, const ITensorInfo *dst)
Static function to check if given info will lead to a valid configuration.
void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Quantization info when assuming per layer quantization.
Describe one of the image&#39;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&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:214
SimpleTensor< float > src
Definition: DFT.cpp:155
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.
const ITensor * get_const_tensor(int id) const
Get constant tensor of a given id.
Definition: ITensorPack.cpp:54
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
static constexpr size_t DimX
Alias for dimension 0 also known as X dimension.
Definition: Window.h:43
static constexpr unsigned int num_arguments_per_3D_tensor()
Returns the number of arguments enqueued per 3D tensor object.
Definition: ICLKernel.h:256
std::string float_to_string_with_full_precision(float val)
Create a string with the float in full precision.
Definition: Utils.h:1075
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
unsigned int num_elems_processed_per_iteration
UniformQuantizationInfo uniform() const
Return per layer quantization info.
std::string get_cl_type_from_data_type(const DataType &dt)
Translates a tensor data type to the appropriate OpenCL type.
Definition: CLHelpers.cpp:39
void configure(const CLCompileContext &compile_context, ITensorInfo *src, unsigned int batch_offset, ITensorInfo *dst)
Initialise the kernel&#39;s source and destination.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
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
virtual QuantizationInfo quantization_info() const =0
Get the quantization settings (scale and offset) of the tensor.
#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.
bool is_data_type_quantized_asymmetric(DataType dt)
Check if a given data type is of asymmetric quantized type.
Definition: Utils.h:1003
static constexpr size_t DimY
Alias for dimension 1 also known as Y dimension.
Definition: Window.h:45
ITensor * get_tensor(int id)
Get tensor of a given id from the pac.
Definition: ITensorPack.cpp:64
#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
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:541
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
Tensor packing service.
Definition: ITensorPack.h:39
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
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
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:291
Describe a multidimensional execution window.
Definition: Window.h:39
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:201
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)