Compute Library
 22.11
ClCastKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016-2022 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 
37 #include "support/Cast.h"
38 #include "support/StringSupport.h"
39 
40 namespace arm_compute
41 {
42 namespace opencl
43 {
44 namespace kernels
45 {
46 namespace
47 {
48 Status validate_arguments(const ITensorInfo *src, const ITensorInfo *dst, ConvertPolicy policy)
49 {
50  ARM_COMPUTE_UNUSED(policy);
52  ARM_COMPUTE_RETURN_ERROR_ON(src == dst);
54  1,
59  1,
63  ARM_COMPUTE_RETURN_ERROR_ON_MSG(src->data_type() == dst->data_type(), "src and dst data types must be different");
64 
65  // Validate in case of configured dst
66  if(dst->total_size() > 0)
67  {
69  }
70 
71  return Status{};
72 }
73 } // namespace
74 
76 {
78 }
79 
80 void ClCastKernel::configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst, ConvertPolicy policy)
81 {
83 
84  // Auto initialize dst shape if not initialized (We can only auto-configure the shape, datatype must be given)
85  set_shape_if_empty(*dst, src->tensor_shape());
86 
88 
89  auto padding_info = get_padding_info({ src, dst });
90 
91  // Get data sizes
92  const size_t src_size = data_size_from_type(src->data_type());
93  const size_t dst_size = data_size_from_type(dst->data_type());
94 
95  // Get number of elements to process per iterations
96  const unsigned int num_elems_processed_per_iteration = adjust_vec_size(16 / src->element_size(), src->dimension(0));
97 
98  // Set build options
99  CLBuildOptions build_opts;
100  build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(num_elems_processed_per_iteration));
101  build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(src->dimension(0) % num_elems_processed_per_iteration));
102  build_opts.add_option("-DDATA_TYPE_IN=" + get_cl_type_from_data_type(src->data_type()));
103  build_opts.add_option("-DDATA_TYPE_OUT=" + get_cl_type_from_data_type(dst->data_type()));
104  // Conversions from float always SATURATE as out-of-bounds conversion from float->integer is implementation defined
105  build_opts.add_option_if(is_data_type_float(src->data_type()) || policy == ConvertPolicy::SATURATE, "-DSATURATE");
106  build_opts.add_option_if(is_data_type_float(src->data_type()) || is_data_type_float(dst->data_type()), "-DIS_DATA_TYPE_FLOAT");
107  build_opts.add_option_if(is_data_type_quantized(src->data_type()), "-DIS_DATA_TYPE_QUANTIZED");
108 
109  // Create kernel
110  const std::string kernel_name = (src_size >= dst_size) ? "cast_down" : "cast_up";
111  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
112 
113  // Configure kernel
114  Window win = calculate_max_window(*src, Steps(num_elems_processed_per_iteration));
115  ICLKernel::configure_internal(win);
116 
117  // Collapse window
118  const Window &full_window = window();
119  Window collapsed_window = full_window.collapse_if_possible(full_window, Window::DimZ);
120  ICLKernel::configure_internal(collapsed_window);
121 
123 
124  // Set config_id for enabling LWS tuning
125  _config_id = kernel_name;
126  _config_id += "_";
127  _config_id += lower_string(string_from_data_type(src->data_type()));
128  _config_id += "_";
129  _config_id += support::cpp11::to_string(src->dimension(0));
130  _config_id += "_";
131  _config_id += support::cpp11::to_string(src->dimension(1));
132  _config_id += "_";
133  _config_id += support::cpp11::to_string(dst->dimension(0));
134  _config_id += "_";
135  _config_id += support::cpp11::to_string(dst->dimension(1));
136 }
137 
139 {
141  return Status{};
142 }
143 
144 void ClCastKernel::run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue)
145 {
148 
149  const auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
150  auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
151 
153 
155  Window slice = collapsed.first_slice_window_3D();
156 
157  do
158  {
159  unsigned int idx = 0;
160  add_3D_tensor_argument(idx, src, slice);
161  add_3D_tensor_argument(idx, dst, slice);
162  enqueue(queue, *this, slice, lws_hint());
163  }
164  while(collapsed.slide_window_slice_3D(slice));
165 }
166 } // namespace kernels
167 } // namespace opencl
168 } // namespace arm_compute
bool is_data_type_quantized(DataType dt)
Check if a given data type is of quantized type.
Definition: Utils.h:1030
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:383
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
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
Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context...
Status class.
Definition: Error.h:52
std::string lower_string(const std::string &val)
Lower a given string.
Definition: Utils.cpp:353
Status validate_arguments(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *dst, const PadStrideInfo &conv_info)
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override
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:226
SimpleTensor< float > src
Definition: DFT.cpp:155
Copyright (c) 2017-2022 Arm Limited.
1 channel, 1 F16 per channel
1 channel, 1 S32 per channel
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:404
const std::string & string_from_data_type(DataType dt)
Convert a data type identity into a string.
Definition: Utils.cpp:135
#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
unsigned int num_elems_processed_per_iteration
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
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()
void configure(const CLCompileContext &compile_context, const ITensorInfo *src, ITensorInfo *dst, ConvertPolicy policy)
Set the src and dst of the kernel.
Elementwise CL kernel type.
Definition: CLTypes.h:85
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:349
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:915
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:603
CLCompileContext class.
quantized, symmetric per channel fixed-point 8-bit number
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_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:588
Wrapper to configure the Khronos OpenCL C++ header.
#define ARM_COMPUTE_RETURN_ERROR_ON_MSG(cond, msg)
If the condition is true, an error is returned.
Definition: Error.h:244
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:1222
quantized, asymmetric fixed-point 8-bit number signed
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:305
std::string kernel_name
signed 8-bit number
static Status validate(const ITensorInfo *src, const ITensorInfo *dst, ConvertPolicy policy)
Static function to check if given info will lead to a valid configuration.
Describe a multidimensional execution window.
Definition: Window.h:39
ConvertPolicy
Policy to handle integer overflow.
Definition: Types.h:404
bool is_data_type_float(DataType dt)
Check if a given data type is of floating point type.
Definition: Utils.h:1010
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:201
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)