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
|
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2016
// Mehdi Goli Codeplay Software Ltd.
// Ralph Potter Codeplay Software Ltd.
// Luke Iwanski Codeplay Software Ltd.
// Contact: <eigen@codeplay.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_device_sycl
#define EIGEN_DEFAULT_DENSE_INDEX_TYPE int
#define EIGEN_USE_SYCL
#include "main.h"
#include <unsupported/Eigen/CXX11/Tensor>
#include<stdint.h>
void test_device_memory(const Eigen::SyclDevice &sycl_device) {
std::cout << "Running on: "
<< sycl_device.m_queue.get_device(). template get_info<cl::sycl::info::device::name>()
<< std::endl;
int sizeDim1 = 100;
array<int, 1> tensorRange = {{sizeDim1}};
Tensor<int, 1> in(tensorRange);
Tensor<int, 1> in1(tensorRange);
memset(in1.data(), 1,in1.size()*sizeof(int));
int* gpu_in_data = static_cast<int*>(sycl_device.allocate(in.size()*sizeof(int)));
sycl_device.memset(gpu_in_data, 1, in.size()*sizeof(int) );
sycl_device.memcpyDeviceToHost(in.data(), gpu_in_data, in.size()*sizeof(int) );
for (int i=0; i<in.size(); i++) {
VERIFY_IS_APPROX(in(i), in1(i));
}
sycl_device.deallocate(gpu_in_data);
}
void test_device_exceptions(const Eigen::SyclDevice &sycl_device) {
bool threw_exception = false;
array<int, 1> tensorDims = {{100}};
int* gpu_data = static_cast<int*>(sycl_device.allocate(100*sizeof(int)));
TensorMap<Tensor<int, 1>> in(gpu_data, tensorDims);
TensorMap<Tensor<int, 1>> out(gpu_data, tensorDims);
try {
out.device(sycl_device) = in / in.constant(0);
} catch(...) {
threw_exception = true;
}
VERIFY(threw_exception);
sycl_device.deallocate(gpu_data);
}
void test_cxx11_tensor_device_sycl() {
cl::sycl::gpu_selector s;
Eigen::SyclDevice sycl_device(s);
CALL_SUBTEST(test_device_memory(sycl_device));
// This deadlocks
// CALL_SUBTEST(test_device_exceptions(sycl_device));
}
|
