Compute Library
 21.02
CLL2NormalizeLayer.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2017-2020 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 
27 #include "arm_compute/core/Error.h"
35 
36 namespace arm_compute
37 {
38 namespace
39 {
40 constexpr int max_input_tensor_dim = 3;
41 } // namespace
42 
43 CLL2NormalizeLayer::CLL2NormalizeLayer(std::shared_ptr<IMemoryManager> memory_manager)
44  : _memory_group(std::move(memory_manager)),
45  _reduce_func(),
46  _normalize_kernel(std::make_unique<CLL2NormalizeLayerKernel>()),
47  _sumsq()
48 {
49 }
50 
52 
54 {
55  configure(CLKernelLibrary::get().get_compile_context(), input, output, axis, epsilon);
56 }
57 
58 void CLL2NormalizeLayer::configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output, int axis, float epsilon)
59 {
60  // Reset auxiliary tensor
61  _sumsq.allocator()->init(TensorInfo());
62 
63  // Manage intermediate buffers
64  _memory_group.manage(&_sumsq);
65 
66  // Configure kernels
67  const uint32_t actual_axis = wrap_around(axis, max_input_tensor_dim);
68  _reduce_func.configure(compile_context, input, &_sumsq, actual_axis, ReductionOperation::SUM_SQUARE);
69  _normalize_kernel->configure(compile_context, input, &_sumsq, output, axis, epsilon);
70 
71  // Allocate intermediate tensor
72  _sumsq.allocator()->allocate();
73 }
74 
76 {
77  TensorShape shape(input->tensor_shape());
78 
79  // Create intermediate tensor info
80  TensorInfo sum_sq;
81  sum_sq.set_data_type(input->data_type());
82  sum_sq.set_tensor_shape(shape);
83 
84  const uint32_t actual_axis = wrap_around(axis, max_input_tensor_dim);
86 
87  // Reduce shape on axis
88  shape.set(actual_axis, 1);
89  sum_sq.set_tensor_shape(shape);
90 
91  ARM_COMPUTE_RETURN_ON_ERROR(CLL2NormalizeLayerKernel::validate(input, &sum_sq, output, axis, epsilon));
92 
93  return Status{};
94 }
95 
97 {
98  MemoryGroupResourceScope scope_mg(_memory_group);
99 
100  _reduce_func.run();
101  CLScheduler::get().enqueue(*_normalize_kernel, true);
102 }
103 } // namespace arm_compute
static Status validate(const ITensorInfo *input, const ITensorInfo *output, int axis, float epsilon=1e-12f)
Static function to check if given info will lead to a valid configuration of CLL2NormalizeLayer.
Shape of a tensor.
Definition: TensorShape.h:39
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:321
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 manage(IMemoryManageable *obj) override
Sets a object to be managed by the given memory group.
Definition: MemoryGroup.h:79
CLL2NormalizeLayer(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Constructor.
Interface to enqueue OpenCL kernels and get/set the OpenCL CommandQueue and ICLTuner.
T wrap_around(T x, T m)
Wrap-around a number within the range 0 <= x < m.
Definition: Helpers.h:231
~CLL2NormalizeLayer()
Default Destructor.
void configure(ICLTensor *input, ICLTensor *output, int axis, float epsilon=1e-12f)
Set the input and output tensors.
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
void run() override
Run the kernels contained in the function.
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.
void run() override
Run the kernels contained in the function.
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
Interface for performing a L2 normalize on a given axis given the square sum of it in this axis...
static Status validate(const ITensorInfo *input, const ITensorInfo *sum, const ITensorInfo *output, int axis, float epsilon)
Static function to check if given info will lead to a valid configuration of CLL2NormalizeLayerKernel...
Store the tensor&#39;s metadata.
Definition: TensorInfo.h:45
void configure(ICLTensor *input, ICLTensor *output, unsigned int axis, ReductionOperation op, bool keep_dims=true)
Set the input and output tensors.