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
181
182
183
184
185
186
187
188
189
190
|
/*
* Copyright (C) 2017 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "NeuralNetworksWrapper.h"
#include <android/sharedmem.h>
//#include <android-base/logging.h>
#include <gtest/gtest.h>
#include <sys/mman.h>
#include <sys/types.h>
#include <unistd.h>
using namespace android::nn::wrapper;
namespace {
typedef float Matrix3x4[3][4];
// Tests the various ways to pass weights and input/output data.
class MemoryTest : public ::testing::Test {
protected:
virtual void SetUp() {}
const Matrix3x4 matrix1 = {{1.f, 2.f, 3.f, 4.f}, {5.f, 6.f, 7.f, 8.f}, {9.f, 10.f, 11.f, 12.f}};
const Matrix3x4 matrix2 = {{100.f, 200.f, 300.f, 400.f},
{500.f, 600.f, 700.f, 800.f},
{900.f, 1000.f, 1100.f, 1200.f}};
const Matrix3x4 matrix3 = {{20.f, 30.f, 40.f, 50.f},
{21.f, 22.f, 23.f, 24.f},
{31.f, 32.f, 33.f, 34.f}};
const Matrix3x4 expected3 = {{121.f, 232.f, 343.f, 454.f},
{526.f, 628.f, 730.f, 832.f},
{940.f, 1042.f, 1144.f, 1246.f}};
const Matrix3x4 expected3b = {{22.f, 34.f, 46.f, 58.f},
{31.f, 34.f, 37.f, 40.f},
{49.f, 52.f, 55.f, 58.f}};
};
// Check that the values are the same. This works only if dealing with integer
// value, otherwise we should accept values that are similar if not exact.
int CompareMatrices(const Matrix3x4& expected, const Matrix3x4& actual) {
int errors = 0;
for (int i = 0; i < 3; i++) {
for (int j = 0; j < 4; j++) {
if (expected[i][j] != actual[i][j]) {
printf("expected[%d][%d] != actual[%d][%d], %f != %f\n", i, j, i, j,
static_cast<double>(expected[i][j]), static_cast<double>(actual[i][j]));
errors++;
}
}
}
return errors;
}
// TODO: test non-zero offset.
TEST_F(MemoryTest, TestASharedMemory) {
// Layout where to place matrix2 and matrix3 in the memory we'll allocate.
// We have gaps to test that we don't assume contiguity.
constexpr uint32_t offsetForMatrix2 = 20;
constexpr uint32_t offsetForMatrix3 = offsetForMatrix2 + sizeof(matrix2) + 30;
constexpr uint32_t memorySize = offsetForMatrix3 + sizeof(matrix3) + 60;
int weightsFd = ASharedMemory_create("weights", memorySize);
ASSERT_GT(weightsFd, -1);
uint8_t* weightsData = (uint8_t*)mmap(nullptr, memorySize, PROT_READ | PROT_WRITE,
MAP_SHARED, weightsFd, 0);
ASSERT_NE(weightsData, nullptr);
memcpy(weightsData + offsetForMatrix2, matrix2, sizeof(matrix2));
memcpy(weightsData + offsetForMatrix3, matrix3, sizeof(matrix3));
Memory weights(memorySize, PROT_READ | PROT_WRITE, weightsFd, 0);
ASSERT_TRUE(weights.isValid());
Model model;
OperandType matrixType(Type::TENSOR_FLOAT32, {3, 4});
OperandType scalarType(Type::INT32, {});
int32_t activation(0);
auto a = model.addOperand(&matrixType);
auto b = model.addOperand(&matrixType);
auto c = model.addOperand(&matrixType);
auto d = model.addOperand(&matrixType);
auto e = model.addOperand(&matrixType);
auto f = model.addOperand(&scalarType);
model.setOperandValueFromMemory(e, &weights, offsetForMatrix2, sizeof(Matrix3x4));
model.setOperandValueFromMemory(a, &weights, offsetForMatrix3, sizeof(Matrix3x4));
model.setOperandValue(f, &activation, sizeof(activation));
model.addOperation(ANEURALNETWORKS_ADD, {a, c, f}, {b});
model.addOperation(ANEURALNETWORKS_ADD, {b, e, f}, {d});
model.identifyInputsAndOutputs({c}, {d});
ASSERT_TRUE(model.isValid());
model.finish();
// Test the two node model.
constexpr uint32_t offsetForMatrix1 = 20;
int inputFd = ASharedMemory_create("input", offsetForMatrix1 + sizeof(Matrix3x4));
ASSERT_GT(inputFd, -1);
uint8_t* inputData = (uint8_t*)mmap(nullptr, offsetForMatrix1 + sizeof(Matrix3x4),
PROT_READ | PROT_WRITE, MAP_SHARED, inputFd, 0);
ASSERT_NE(inputData, nullptr);
memcpy(inputData + offsetForMatrix1, matrix1, sizeof(Matrix3x4));
Memory input(offsetForMatrix1 + sizeof(Matrix3x4), PROT_READ, inputFd, 0);
ASSERT_TRUE(input.isValid());
constexpr uint32_t offsetForActual = 32;
int outputFd = ASharedMemory_create("output", offsetForActual + sizeof(Matrix3x4));
ASSERT_GT(outputFd, -1);
uint8_t* outputData = (uint8_t*)mmap(nullptr, offsetForActual + sizeof(Matrix3x4),
PROT_READ | PROT_WRITE, MAP_SHARED, outputFd, 0);
ASSERT_NE(outputData, nullptr);
memset(outputData, 0, offsetForActual + sizeof(Matrix3x4));
Memory actual(offsetForActual + sizeof(Matrix3x4), PROT_READ | PROT_WRITE, outputFd, 0);
ASSERT_TRUE(actual.isValid());
Compilation compilation2(&model);
ASSERT_EQ(compilation2.finish(), Result::NO_ERROR);
Execution execution2(&compilation2);
ASSERT_EQ(execution2.setInputFromMemory(0, &input, offsetForMatrix1, sizeof(Matrix3x4)),
Result::NO_ERROR);
ASSERT_EQ(execution2.setOutputFromMemory(0, &actual, offsetForActual, sizeof(Matrix3x4)),
Result::NO_ERROR);
ASSERT_EQ(execution2.compute(), Result::NO_ERROR);
ASSERT_EQ(CompareMatrices(expected3, *reinterpret_cast<Matrix3x4*>(outputData + offsetForActual)), 0);
}
TEST_F(MemoryTest, TestFd) {
// Create a file that contains matrix2 and matrix3.
char path[] = "/data/local/tmp/TestMemoryXXXXXX";
int fd = mkstemp(path);
const uint32_t offsetForMatrix2 = 20;
const uint32_t offsetForMatrix3 = 200;
static_assert(offsetForMatrix2 + sizeof(matrix2) < offsetForMatrix3, "matrices overlap");
lseek(fd, offsetForMatrix2, SEEK_SET);
write(fd, matrix2, sizeof(matrix2));
lseek(fd, offsetForMatrix3, SEEK_SET);
write(fd, matrix3, sizeof(matrix3));
fsync(fd);
Memory weights(offsetForMatrix3 + sizeof(matrix3), PROT_READ, fd, 0);
ASSERT_TRUE(weights.isValid());
Model model;
OperandType matrixType(Type::TENSOR_FLOAT32, {3, 4});
OperandType scalarType(Type::INT32, {});
int32_t activation(0);
auto a = model.addOperand(&matrixType);
auto b = model.addOperand(&matrixType);
auto c = model.addOperand(&matrixType);
auto d = model.addOperand(&matrixType);
auto e = model.addOperand(&matrixType);
auto f = model.addOperand(&scalarType);
model.setOperandValueFromMemory(e, &weights, offsetForMatrix2, sizeof(Matrix3x4));
model.setOperandValueFromMemory(a, &weights, offsetForMatrix3, sizeof(Matrix3x4));
model.setOperandValue(f, &activation, sizeof(activation));
model.addOperation(ANEURALNETWORKS_ADD, {a, c, f}, {b});
model.addOperation(ANEURALNETWORKS_ADD, {b, e, f}, {d});
model.identifyInputsAndOutputs({c}, {d});
ASSERT_TRUE(model.isValid());
model.finish();
// Test the three node model.
Matrix3x4 actual;
memset(&actual, 0, sizeof(actual));
Compilation compilation2(&model);
ASSERT_EQ(compilation2.finish(), Result::NO_ERROR);
Execution execution2(&compilation2);
ASSERT_EQ(execution2.setInput(0, matrix1, sizeof(Matrix3x4)), Result::NO_ERROR);
ASSERT_EQ(execution2.setOutput(0, actual, sizeof(Matrix3x4)), Result::NO_ERROR);
ASSERT_EQ(execution2.compute(), Result::NO_ERROR);
ASSERT_EQ(CompareMatrices(expected3, actual), 0);
close(fd);
unlink(path);
}
} // end namespace
|
