Compute Library
 21.11
CLReductionOperation.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2017-2021 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 
35 #include "src/runtime/Utils.h"
36 
37 #include "src/common/utils/Log.h"
38 
39 namespace arm_compute
40 {
41 CLReductionOperation::CLReductionOperation(std::shared_ptr<IMemoryManager> memory_manager)
42  : _memory_group(std::move(memory_manager)), _unreshaped_output(), _reduction_kernel(), _reshape(), _reduction_axis(), _is_reshape_required(false)
43 {
44 }
45 
47 
48 Status CLReductionOperation::validate(const ITensorInfo *input, const ITensorInfo *output, unsigned int axis, ReductionOperation op, bool keep_dims)
49 {
50  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
51  ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis >= TensorShape::num_max_dimensions, "Reduction axis greater than max number of dimensions");
52  ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis > 3, "Unsupported reduction axis");
53 
54  const bool is_reshape_required = !keep_dims;
55 
56  if(is_reshape_required && output->total_size() != 0)
57  {
58  const TensorInfo expected_output_shape = output->clone()->set_tensor_shape(arm_compute::misc::shape_calculator::compute_reduced_shape(input->tensor_shape(), axis, keep_dims));
59  ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(&expected_output_shape, output);
60  }
61 
62  auto *output_internal = output;
63 
64  TensorInfo output_before_reshape;
65  const auto input_shape = input->tensor_shape();
66  const auto input_num_channles = input->num_channels();
67  const auto input_qinfo = input->quantization_info();
68  const auto output_data_type = output->data_type();
69 
70  auto initialize_tensorinfo = [](TensorInfo & ti, TensorShape shape, DataType data_type, int num_channels, QuantizationInfo qinfo)
71  {
73  };
74 
75  if(is_reshape_required)
76  {
77  auto shape_before_reshape = input_shape;
78  shape_before_reshape.set(axis, 1);
79  initialize_tensorinfo(output_before_reshape, shape_before_reshape, output_data_type, input_num_channles, input_qinfo);
80  output_internal = &output_before_reshape;
81  }
82 
83  ARM_COMPUTE_RETURN_ON_ERROR(CLReductionOperationKernel::validate(input, output_internal, axis, op));
84 
85  if(is_reshape_required)
86  {
87  ARM_COMPUTE_RETURN_ON_ERROR(CLReshapeLayer::validate(output_internal, output));
88  }
89 
90  return Status{};
91 }
92 
93 ICLTensor *CLReductionOperation::configure_intermediate_result_vector(ICLTensor *input, ICLTensor *output)
94 {
95  if(!_is_reshape_required)
96  {
97  return output;
98  }
99 
100  auto shape = input->info()->tensor_shape();
101  shape.set(_reduction_axis, 1);
102  _unreshaped_output.allocator()->init(input->info()->clone()->set_tensor_shape(shape));
103  return &_unreshaped_output;
104 }
105 
106 void CLReductionOperation::configure(ICLTensor *input, ICLTensor *output, unsigned int axis, ReductionOperation op, bool keep_dims)
107 {
108  configure(CLKernelLibrary::get().get_compile_context(), input, output, axis, op, keep_dims);
109 }
110 
111 void CLReductionOperation::configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output, unsigned int axis, ReductionOperation op, bool keep_dims)
112 {
113  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
114  ARM_COMPUTE_LOG_PARAMS(input, output, axis, op, keep_dims);
115  _reduction_axis = axis;
116  _is_reshape_required = !keep_dims;
117 
118  auto *output_internal = configure_intermediate_result_vector(input, output);
119 
120  if(_is_reshape_required)
121  {
123  const auto output_data_type = input->info()->data_type();
124  auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape).set_data_type(output_data_type).reset_padding().set_is_resizable(true));
125 
126  _memory_group.manage(&_unreshaped_output);
127  }
128 
129  _reduction_kernel = std::make_unique<CLReductionOperationKernel>();
130  _reduction_kernel->configure(compile_context, input, output_internal, axis, op);
131 
132  if(_is_reshape_required)
133  {
134  _reshape.configure(compile_context, &_unreshaped_output, output);
135  _unreshaped_output.allocator()->allocate();
136  }
137 }
138 
140 {
141  MemoryGroupResourceScope scope_mg(_memory_group);
142 
143  CLScheduler::get().enqueue(*_reduction_kernel, false);
144 
145  if(_is_reshape_required)
146  {
147  _reshape.run();
148  }
149 }
150 } // namespace arm_compute
virtual ITensorInfo & set_num_channels(int num_channels)=0
Set the number of channels to the specified value.
static Status validate(const ITensorInfo *input, const ITensorInfo *output)
Static function to check if given info will lead to a valid configuration of CLReshapeLayer.
Shape of a tensor.
Definition: TensorShape.h:39
ReductionOperation
Available reduction operations.
Definition: Types.h:463
static CLScheduler & get()
Access the scheduler singleton.
virtual ITensorInfo & set_tensor_shape(const TensorShape &shape)=0
Set the shape of an already initialized tensor.
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
virtual DataType data_type() const =0
Data type used for each element of the tensor.
ITensorInfo & set_data_type(DataType data_type) override
Set the data type to the specified value.
Definition: TensorInfo.cpp:287
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
Store the tensor&#39;s metadata.
Definition: ITensorInfo.h:40
CLTensorAllocator * allocator()
Return a pointer to the tensor&#39;s allocator.
Definition: CLTensor.cpp:61
Status class.
Definition: Error.h:52
void init(const TensorInfo &input, size_t alignment=0)
Initialize a tensor based on the passed TensorInfo.
Copyright (c) 2017-2021 Arm Limited.
void run() override
Run the kernels contained in the function.
void manage(IMemoryManageable *obj) override
Sets a object to be managed by the given memory group.
Definition: MemoryGroup.h:79
const DataType data_type
Definition: Im2Col.cpp:150
Interface to enqueue OpenCL kernels and get/set the OpenCL CommandQueue and ICLTuner.
Quantization information.
const auto input_shape
Validate test suite is to test ARM_COMPUTE_RETURN_ON_* macros we use to check the validity of given a...
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
void run() override
Run the kernels contained in the function.
TensorShape compute_reduced_shape(const TensorShape &input, unsigned int axis, bool keep_dims=true)
Calculate the reduced shape of a tensor given an axis.
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...
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.
virtual ITensorInfo & set_quantization_info(const QuantizationInfo &quantization_info)=0
Set the quantization settings (scale and offset) of the tensor.
CLReductionOperation(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Default Constructor.
static Status validate(const ITensorInfo *input, const ITensorInfo *output, unsigned int axis, ReductionOperation op, bool keep_dims=true)
Static function to check if given info will lead to a valid configuration of CLReductionOperation.
virtual QuantizationInfo quantization_info() const =0
Get the quantization settings (scale and offset) of the tensor.
void enqueue(ICLKernel &kernel, bool flush=true)
Schedule the execution of the passed kernel if possible.
CLCompileContext class.
void allocate() override
Allocate size specified by TensorInfo of OpenCL memory.
Memory group resources scope handling class.
Definition: IMemoryGroup.h:82
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
virtual size_t total_size() const =0
Returns the total size of the tensor in bytes.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(...)
Definition: Validate.h:439
const QuantizationInfo qinfo
Definition: Im2Col.cpp:155
#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_LOG_PARAMS(...)
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
Store the tensor&#39;s metadata.
Definition: TensorInfo.h:43
void configure(ICLTensor *input, ICLTensor *output, unsigned int axis, ReductionOperation op, bool keep_dims=true)
Set the input and output tensors.
void configure(const ICLTensor *input, ICLTensor *output)
Initialise the kernel&#39;s inputs and outputs.
static Status validate(const ITensorInfo *input, const ITensorInfo *output, unsigned int axis, ReductionOperation op)
Static function to check if given info will lead to a valid configuration of CLReductionOperationKern...
~CLReductionOperation()
Default Destructor.
static constexpr size_t num_max_dimensions
Number of dimensions the tensor has.
Definition: Dimensions.h:46
DataType
Available data types.
Definition: Types.h:79
virtual size_t num_channels() const =0
The number of channels for each tensor element.