1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
|
/*
* Copyright (c) 2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "arm_compute/runtime/BlobLifetimeManager.h"
#include "arm_compute/runtime/CL/CLBufferAllocator.h"
#include "arm_compute/runtime/CL/functions/CLConvolutionLayer.h"
#include "arm_compute/runtime/CL/functions/CLL2NormalizeLayer.h"
#include "arm_compute/runtime/MemoryGroup.h"
#include "arm_compute/runtime/MemoryManagerOnDemand.h"
#include "arm_compute/runtime/PoolManager.h"
#include "support/ToolchainSupport.h"
#include "tests/AssetsLibrary.h"
#include "tests/CL/CLAccessor.h"
#include "tests/Globals.h"
#include "tests/Utils.h"
#include "tests/framework/Asserts.h"
#include "tests/framework/Macros.h"
#include "tests/framework/datasets/Datasets.h"
#include "tests/validation/Validation.h"
#include "tests/validation/fixtures/UNIT/DynamicTensorFixture.h"
namespace arm_compute
{
namespace test
{
namespace validation
{
namespace
{
constexpr AbsoluteTolerance<float> absolute_tolerance_float(0.0001f); /**< Absolute Tolerance value for comparing reference's output against implementation's output for DataType::F32 */
RelativeTolerance<float> tolerance_f32(0.1f); /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */
constexpr float tolerance_num = 0.07f; /**< Tolerance number */
using CLL2NormLayerWrapper = SimpleFunctionWrapper<MemoryManagerOnDemand, CLL2NormalizeLayer, ICLTensor>;
template <>
void CLL2NormLayerWrapper::configure(ICLTensor *src, ICLTensor *dst)
{
_func.configure(src, dst, 0, 0.0001f);
}
} // namespace
TEST_SUITE(CL)
TEST_SUITE(UNIT)
TEST_SUITE(DynamicTensor)
using BlobMemoryManagementService = MemoryManagementService<CLBufferAllocator, BlobLifetimeManager, PoolManager, MemoryManagerOnDemand>;
using CLDynamicTensorType3SingleFunction = DynamicTensorType3SingleFunction<CLTensor, CLAccessor, BlobMemoryManagementService, CLL2NormLayerWrapper>;
/** Tests the memory manager with dynamic input and output tensors.
*
* Create and manage the tensors needed to run a simple function. After the function is executed,
* change the input and output size requesting more memory and go through the manage/allocate process.
* The memory manager should be able to update the inner structures and allocate the requested memory
* */
FIXTURE_DATA_TEST_CASE(DynamicTensorType3Single, CLDynamicTensorType3SingleFunction, framework::DatasetMode::ALL,
framework::dataset::zip(framework::dataset::make("Level0Shape", { TensorShape(12U, 11U, 3U), TensorShape(256U, 8U, 12U) }),
framework::dataset::make("Level1Shape", { TensorShape(67U, 31U, 15U), TensorShape(11U, 2U, 3U) })))
{
ARM_COMPUTE_EXPECT(internal_l0.size() == internal_l1.size(), framework::LogLevel::ERRORS);
ARM_COMPUTE_EXPECT(cross_l0.size() == cross_l1.size(), framework::LogLevel::ERRORS);
const unsigned int internal_size = internal_l0.size();
const unsigned int cross_size = cross_l0.size();
if(input_l0.total_size() < input_l1.total_size())
{
for(unsigned int i = 0; i < internal_size; ++i)
{
ARM_COMPUTE_EXPECT(internal_l0[i].size < internal_l1[i].size, framework::LogLevel::ERRORS);
}
for(unsigned int i = 0; i < cross_size; ++i)
{
ARM_COMPUTE_EXPECT(cross_l0[i].size < cross_l1[i].size, framework::LogLevel::ERRORS);
}
}
else
{
for(unsigned int i = 0; i < internal_size; ++i)
{
ARM_COMPUTE_EXPECT(internal_l0[i].size == internal_l1[i].size, framework::LogLevel::ERRORS);
}
for(unsigned int i = 0; i < cross_size; ++i)
{
ARM_COMPUTE_EXPECT(cross_l0[i].size == cross_l1[i].size, framework::LogLevel::ERRORS);
}
}
}
using CLDynamicTensorType3ComplexFunction = DynamicTensorType3ComplexFunction<CLTensor, CLAccessor, BlobMemoryManagementService, CLConvolutionLayer>;
/** Tests the memory manager with dynamic input and output tensors.
*
* Create and manage the tensors needed to run a complex function. After the function is executed,
* change the input and output size requesting more memory and go through the manage/allocate process.
* The memory manager should be able to update the inner structures and allocate the requested memory
* */
FIXTURE_DATA_TEST_CASE(DynamicTensorType3Complex, CLDynamicTensorType3ComplexFunction, framework::DatasetMode::ALL,
framework::dataset::zip(framework::dataset::zip(framework::dataset::zip(framework::dataset::zip(
framework::dataset::make("InputShape", { std::vector<TensorShape>{ TensorShape(12U, 12U, 16U), TensorShape(64U, 64U, 16U) } }),
framework::dataset::make("WeightsManager", { TensorShape(3U, 3U, 16U, 5U) })),
framework::dataset::make("BiasShape", { TensorShape(5U) })),
framework::dataset::make("OutputShape", { std::vector<TensorShape>{ TensorShape(12U, 12U, 5U), TensorShape(64U, 64U, 5U) } })),
framework::dataset::make("PadStrideInfo", { PadStrideInfo(1U, 1U, 1U, 1U) })))
{
for(unsigned int i = 0; i < num_iterations; ++i)
{
run_iteration(i);
validate(CLAccessor(dst_target), dst_ref, tolerance_f32, tolerance_num, absolute_tolerance_float);
}
}
using CLDynamicTensorType2PipelineFunction = DynamicTensorType2PipelineFunction<CLTensor, CLAccessor, BlobMemoryManagementService, CLConvolutionLayer>;
/** Tests the memory manager with dynamic input and output tensors.
*
* Create and manage the tensors needed to run a pipeline. After the function is executed, resize the input size and rerun.
*/
FIXTURE_DATA_TEST_CASE(DynamicTensorType2Pipeline, CLDynamicTensorType2PipelineFunction, framework::DatasetMode::ALL,
framework::dataset::make("InputShape", { std::vector<TensorShape>{ TensorShape(12U, 12U, 6U), TensorShape(128U, 128U, 6U) } }))
{
}
TEST_SUITE_END() // DynamicTensor
TEST_SUITE_END() // UNIT
TEST_SUITE_END() // CL
} // namespace validation
} // namespace test
} // namespace arm_compute
|
