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
|
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2016 Benoit Steiner <benoit.steiner.goog@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#define EIGEN_TEST_NO_LONGDOUBLE
#define EIGEN_TEST_NO_COMPLEX
#define EIGEN_TEST_FUNC cxx11_tensor_sycl
#define EIGEN_DEFAULT_DENSE_INDEX_TYPE int
#define EIGEN_USE_SYCL
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
using Eigen::array;
using Eigen::SyclDevice;
using Eigen::Tensor;
using Eigen::TensorMap;
// Types used in tests:
using TestTensor = Tensor<float, 3>;
using TestTensorMap = TensorMap<Tensor<float, 3>>;
void test_sycl_cpu() {
cl::sycl::gpu_selector s;
cl::sycl::queue q(s, [=](cl::sycl::exception_list l) {
for (const auto& e : l) {
try {
std::rethrow_exception(e);
} catch (cl::sycl::exception e) {
std::cout << e.what() << std::endl;
}
}
});
SyclDevice sycl_device(q);
int sizeDim1 = 100;
int sizeDim2 = 100;
int sizeDim3 = 100;
array<int, 3> tensorRange = {{sizeDim1, sizeDim2, sizeDim3}};
TestTensor in1(tensorRange);
TestTensor in2(tensorRange);
TestTensor in3(tensorRange);
TestTensor out(tensorRange);
in1 = in1.random();
in2 = in2.random();
in3 = in3.random();
TestTensorMap gpu_in1(in1.data(), tensorRange);
TestTensorMap gpu_in2(in2.data(), tensorRange);
TestTensorMap gpu_in3(in3.data(), tensorRange);
TestTensorMap gpu_out(out.data(), tensorRange);
/// a=1.2f
gpu_in1.device(sycl_device) = gpu_in1.constant(1.2f);
sycl_device.deallocate(in1.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(in1(i,j,k), 1.2f);
}
}
}
printf("a=1.2f Test passed\n");
/// a=b*1.2f
gpu_out.device(sycl_device) = gpu_in1 * 1.2f;
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i,j,k),
in1(i,j,k) * 1.2f);
}
}
}
printf("a=b*1.2f Test Passed\n");
/// c=a*b
gpu_out.device(sycl_device) = gpu_in1 * gpu_in2;
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i,j,k),
in1(i,j,k) *
in2(i,j,k));
}
}
}
printf("c=a*b Test Passed\n");
/// c=a+b
gpu_out.device(sycl_device) = gpu_in1 + gpu_in2;
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i,j,k),
in1(i,j,k) +
in2(i,j,k));
}
}
}
printf("c=a+b Test Passed\n");
/// c=a*a
gpu_out.device(sycl_device) = gpu_in1 * gpu_in1;
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i,j,k),
in1(i,j,k) *
in1(i,j,k));
}
}
}
printf("c= a*a Test Passed\n");
//a*3.14f + b*2.7f
gpu_out.device(sycl_device) = gpu_in1 * gpu_in1.constant(3.14f) + gpu_in2 * gpu_in2.constant(2.7f);
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i,j,k),
in1(i,j,k) * 3.14f
+ in2(i,j,k) * 2.7f);
}
}
}
printf("a*3.14f + b*2.7f Test Passed\n");
///d= (a>0.5? b:c)
gpu_out.device(sycl_device) =(gpu_in1 > gpu_in1.constant(0.5f)).select(gpu_in2, gpu_in3);
sycl_device.deallocate(out.data());
for (int i = 0; i < sizeDim1; ++i) {
for (int j = 0; j < sizeDim2; ++j) {
for (int k = 0; k < sizeDim3; ++k) {
VERIFY_IS_APPROX(out(i, j, k), (in1(i, j, k) > 0.5f)
? in2(i, j, k)
: in3(i, j, k));
}
}
}
printf("d= (a>0.5? b:c) Test Passed\n");
}
void test_cxx11_tensor_sycl() {
CALL_SUBTEST(test_sycl_cpu());
}
|
