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 */
24 #include "arm_compute/runtime/CL/functions/CLL2NormalizeLayer.h"
25
26 #include "arm_compute/core/CL/ICLTensor.h"
27 #include "arm_compute/core/Error.h"
28 #include "arm_compute/core/PixelValue.h"
29 #include "arm_compute/core/TensorInfo.h"
30 #include "arm_compute/core/Validate.h"
31 #include "arm_compute/runtime/CL/CLScheduler.h"
32 #include "src/core/CL/kernels/CLFillBorderKernel.h"
33 #include "src/core/CL/kernels/CLL2NormalizeLayerKernel.h"
34 #include "src/core/CL/kernels/CLReductionOperationKernel.h"
35 #include "support/MemorySupport.h"
36
37 namespace arm_compute
38 {
39 namespace
40 {
41 constexpr int max_input_tensor_dim = 3;
42 } // namespace
43
CLL2NormalizeLayer(std::shared_ptr<IMemoryManager> memory_manager)44 CLL2NormalizeLayer::CLL2NormalizeLayer(std::shared_ptr<IMemoryManager> memory_manager)
45 : _memory_group(std::move(memory_manager)),
46 _reduce_func(),
47 _normalize_kernel(support::cpp14::make_unique<CLL2NormalizeLayerKernel>()),
48 _sumsq()
49 {
50 }
51
52 CLL2NormalizeLayer::~CLL2NormalizeLayer() = default;
53
configure(ICLTensor * input,ICLTensor * output,int axis,float epsilon)54 void CLL2NormalizeLayer::configure(ICLTensor *input, ICLTensor *output, int axis, float epsilon)
55 {
56 configure(CLKernelLibrary::get().get_compile_context(), input, output, axis, epsilon);
57 }
58
configure(const CLCompileContext & compile_context,ICLTensor * input,ICLTensor * output,int axis,float epsilon)59 void CLL2NormalizeLayer::configure(const CLCompileContext &compile_context, ICLTensor *input, ICLTensor *output, int axis, float epsilon)
60 {
61 // Reset auxiliary tensor
62 _sumsq.allocator()->init(TensorInfo());
63
64 // Manage intermediate buffers
65 _memory_group.manage(&_sumsq);
66
67 // Configure kernels
68 const uint32_t actual_axis = wrap_around(axis, max_input_tensor_dim);
69 _reduce_func.configure(compile_context, input, &_sumsq, actual_axis, ReductionOperation::SUM_SQUARE);
70 _normalize_kernel->configure(compile_context, input, &_sumsq, output, axis, epsilon);
71
72 // Allocate intermediate tensor
73 _sumsq.allocator()->allocate();
74 }
75
validate(const ITensorInfo * input,const ITensorInfo * output,int axis,float epsilon)76 Status CLL2NormalizeLayer::validate(const ITensorInfo *input, const ITensorInfo *output, int axis, float epsilon)
77 {
78 TensorShape shape(input->tensor_shape());
79
80 // Create intermediate tensor info
81 TensorInfo sum_sq;
82 sum_sq.set_data_type(input->data_type());
83 sum_sq.set_tensor_shape(shape);
84
85 const uint32_t actual_axis = wrap_around(axis, max_input_tensor_dim);
86 ARM_COMPUTE_RETURN_ON_ERROR(CLReductionOperation::validate(input, &sum_sq, actual_axis, ReductionOperation::SUM_SQUARE));
87
88 // Reduce shape on axis
89 shape.set(actual_axis, 1);
90 sum_sq.set_tensor_shape(shape);
91
92 ARM_COMPUTE_RETURN_ON_ERROR(CLL2NormalizeLayerKernel::validate(input, &sum_sq, output, axis, epsilon));
93
94 return Status{};
95 }
96
run()97 void CLL2NormalizeLayer::run()
98 {
99 MemoryGroupResourceScope scope_mg(_memory_group);
100
101 _reduce_func.run();
102 CLScheduler::get().enqueue(*_normalize_kernel, true);
103 }
104 } // namespace arm_compute
105