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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
|
//
// Copyright © 2017 Arm Ltd. All rights reserved.
// SPDX-License-Identifier: MIT
//
#include <boost/test/unit_test.hpp>
#include <boost/multi_array.hpp>
#include <armnn/ArmNN.hpp>
#include <backends/reference/RefWorkloadFactory.hpp>
#if ARMCOMPUTECL_ENABLED
#include <backends/cl/ClWorkloadFactory.hpp>
#endif
#if ARMCOMPUTENEON_ENABLED
#include <backends/neon/NeonWorkloadFactory.hpp>
#endif
#include <backends/CpuTensorHandle.hpp>
#include <test/TensorHelpers.hpp>
#include "TensorCopyUtils.hpp"
#include "WorkloadTestUtils.hpp"
#if ARMCOMPUTECL_ENABLED || ARMCOMPUTENEON_ENABLED
#include <backends/aclCommon/ArmComputeTensorUtils.hpp>
#endif
BOOST_AUTO_TEST_SUITE(MemCopyTestSuite)
void MemCopyTest(armnn::IWorkloadFactory& srcWorkloadFactory, armnn::IWorkloadFactory& dstWorkloadFactory,
bool withSubtensors)
{
const std::array<unsigned int, 4> shapeData = { { 1u, 1u, 6u, 5u } };
const armnn::TensorShape tensorShape(4, shapeData.data());
const armnn::TensorInfo tensorInfo(tensorShape, armnn::DataType::Float32);
boost::multi_array<float, 4> inputData = MakeTensor<float, 4>(tensorInfo, std::vector<float>(
{
1.0f, 2.0f, 3.0f, 4.0f, 5.0f,
6.0f, 7.0f, 8.0f, 9.0f, 10.0f,
11.0f, 12.0f, 13.0f, 14.0f, 15.0f,
16.0f, 17.0f, 18.0f, 19.0f, 20.0f,
21.0f, 22.0f, 23.0f, 24.0f, 25.0f,
26.0f, 27.0f, 28.0f, 29.0f, 30.0f,
})
);
boost::multi_array<float, 4> outputData(shapeData);
auto inputTensorHandle = srcWorkloadFactory.CreateTensorHandle(tensorInfo);
auto outputTensorHandle = dstWorkloadFactory.CreateTensorHandle(tensorInfo);
AllocateAndCopyDataToITensorHandle(inputTensorHandle.get(), inputData.data());
outputTensorHandle->Allocate();
armnn::MemCopyQueueDescriptor memCopyQueueDesc;
armnn::WorkloadInfo workloadInfo;
const unsigned int origin[4] = {};
auto workloadInput = (withSubtensors && srcWorkloadFactory.SupportsSubTensors())
? srcWorkloadFactory.CreateSubTensorHandle(*inputTensorHandle, tensorShape, origin)
: std::move(inputTensorHandle);
auto workloadOutput = (withSubtensors && dstWorkloadFactory.SupportsSubTensors())
? dstWorkloadFactory.CreateSubTensorHandle(*outputTensorHandle, tensorShape, origin)
: std::move(outputTensorHandle);
AddInputToWorkload(memCopyQueueDesc, workloadInfo, tensorInfo, workloadInput.get());
AddOutputToWorkload(memCopyQueueDesc, workloadInfo, tensorInfo, workloadOutput.get());
dstWorkloadFactory.CreateMemCopy(memCopyQueueDesc, workloadInfo)->Execute();
CopyDataFromITensorHandle(outputData.data(), workloadOutput.get());
BOOST_TEST(CompareTensors(inputData, outputData));
}
template <typename SrcWorkloadFactory, typename DstWorkloadFactory>
void MemCopyTest(bool withSubtensors)
{
SrcWorkloadFactory srcWorkloadFactory;
DstWorkloadFactory dstWorkloadFactory;
MemCopyTest(srcWorkloadFactory, dstWorkloadFactory, withSubtensors);
}
#if ARMCOMPUTECL_ENABLED || ARMCOMPUTENEON_ENABLED
BOOST_AUTO_TEST_CASE(AclTypeConversions)
{
arm_compute::Strides strides(1,2,3,4);
armnn::TensorShape convertedStrides = armnn::armcomputetensorutils::GetStrides(strides);
BOOST_TEST(convertedStrides[0] == 4);
BOOST_TEST(convertedStrides[1] == 3);
BOOST_TEST(convertedStrides[2] == 2);
BOOST_TEST(convertedStrides[3] == 1);
arm_compute::TensorShape shape(5,6,7,8);
armnn::TensorShape convertedshape = armnn::armcomputetensorutils::GetShape(shape);
BOOST_TEST(convertedshape[0] == 8);
BOOST_TEST(convertedshape[1] == 7);
BOOST_TEST(convertedshape[2] == 6);
BOOST_TEST(convertedshape[3] == 5);
}
#endif
#if ARMCOMPUTECL_ENABLED
BOOST_AUTO_TEST_CASE(CopyBetweenCpuAndGpu)
{
MemCopyTest<armnn::RefWorkloadFactory, armnn::ClWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenGpuAndCpu)
{
MemCopyTest<armnn::ClWorkloadFactory, armnn::RefWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenCpuAndGpuWithSubtensors)
{
MemCopyTest<armnn::RefWorkloadFactory, armnn::ClWorkloadFactory>(true);
}
BOOST_AUTO_TEST_CASE(CopyBetweenGpuAndCpuWithSubtensors)
{
MemCopyTest<armnn::ClWorkloadFactory, armnn::RefWorkloadFactory>(true);
}
#endif // ARMCOMPUTECL_ENABLED
#if ARMCOMPUTENEON_ENABLED
BOOST_AUTO_TEST_CASE(CopyBetweenCpuAndNeon)
{
MemCopyTest<armnn::RefWorkloadFactory, armnn::NeonWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenNeonAndCpu)
{
MemCopyTest<armnn::NeonWorkloadFactory, armnn::RefWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenCpuAndNeonWithSubtensors)
{
MemCopyTest<armnn::RefWorkloadFactory, armnn::NeonWorkloadFactory>(true);
}
BOOST_AUTO_TEST_CASE(CopyBetweenNeonAndCpuWithSubtensors)
{
MemCopyTest<armnn::NeonWorkloadFactory, armnn::RefWorkloadFactory>(true);
}
#endif // ARMCOMPUTENEON_ENABLED
#if ARMCOMPUTECL_ENABLED && ARMCOMPUTENEON_ENABLED
BOOST_AUTO_TEST_CASE(CopyBetweenNeonAndGpu)
{
MemCopyTest<armnn::NeonWorkloadFactory, armnn::ClWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenGpuAndNeon)
{
MemCopyTest<armnn::ClWorkloadFactory, armnn::NeonWorkloadFactory>(false);
}
BOOST_AUTO_TEST_CASE(CopyBetweenNeonAndGpuWithSubtensors)
{
MemCopyTest<armnn::NeonWorkloadFactory, armnn::ClWorkloadFactory>(true);
}
BOOST_AUTO_TEST_CASE(CopyBetweenGpuAndNeonWithSubtensors)
{
MemCopyTest<armnn::ClWorkloadFactory, armnn::NeonWorkloadFactory>(true);
}
#endif
BOOST_AUTO_TEST_SUITE_END()
|
