• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra.
3 //
4 // Copyright (C) 2016
5 // Mehdi Goli    Codeplay Software Ltd.
6 // Ralph Potter  Codeplay Software Ltd.
7 // Luke Iwanski  Codeplay Software Ltd.
8 // Contact: <eigen@codeplay.com>
9 // Benoit Steiner <benoit.steiner.goog@gmail.com>
10 //
11 // This Source Code Form is subject to the terms of the Mozilla
12 // Public License v. 2.0. If a copy of the MPL was not distributed
13 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
14 
15 
16 #define EIGEN_TEST_NO_LONGDOUBLE
17 #define EIGEN_TEST_NO_COMPLEX
18 #define EIGEN_TEST_FUNC cxx11_tensor_sycl
19 #define EIGEN_DEFAULT_DENSE_INDEX_TYPE int
20 #define EIGEN_USE_SYCL
21 
22 #include "main.h"
23 #include <unsupported/Eigen/CXX11/Tensor>
24 
25 using Eigen::array;
26 using Eigen::SyclDevice;
27 using Eigen::Tensor;
28 using Eigen::TensorMap;
29 
test_sycl_cpu(const Eigen::SyclDevice & sycl_device)30 void test_sycl_cpu(const Eigen::SyclDevice &sycl_device) {
31 
32   int sizeDim1 = 100;
33   int sizeDim2 = 100;
34   int sizeDim3 = 100;
35   array<int, 3> tensorRange = {{sizeDim1, sizeDim2, sizeDim3}};
36   Tensor<float, 3> in1(tensorRange);
37   Tensor<float, 3> in2(tensorRange);
38   Tensor<float, 3> in3(tensorRange);
39   Tensor<float, 3> out(tensorRange);
40 
41   in2 = in2.random();
42   in3 = in3.random();
43 
44   float * gpu_in1_data  = static_cast<float*>(sycl_device.allocate(in1.dimensions().TotalSize()*sizeof(float)));
45   float * gpu_in2_data  = static_cast<float*>(sycl_device.allocate(in2.dimensions().TotalSize()*sizeof(float)));
46   float * gpu_in3_data  = static_cast<float*>(sycl_device.allocate(in3.dimensions().TotalSize()*sizeof(float)));
47   float * gpu_out_data =  static_cast<float*>(sycl_device.allocate(out.dimensions().TotalSize()*sizeof(float)));
48 
49   TensorMap<Tensor<float, 3>> gpu_in1(gpu_in1_data, tensorRange);
50   TensorMap<Tensor<float, 3>> gpu_in2(gpu_in2_data, tensorRange);
51   TensorMap<Tensor<float, 3>> gpu_in3(gpu_in3_data, tensorRange);
52   TensorMap<Tensor<float, 3>> gpu_out(gpu_out_data, tensorRange);
53 
54   /// a=1.2f
55   gpu_in1.device(sycl_device) = gpu_in1.constant(1.2f);
56   sycl_device.memcpyDeviceToHost(in1.data(), gpu_in1_data ,(in1.dimensions().TotalSize())*sizeof(float));
57   for (int i = 0; i < sizeDim1; ++i) {
58     for (int j = 0; j < sizeDim2; ++j) {
59       for (int k = 0; k < sizeDim3; ++k) {
60         VERIFY_IS_APPROX(in1(i,j,k), 1.2f);
61       }
62     }
63   }
64   printf("a=1.2f Test passed\n");
65 
66   /// a=b*1.2f
67   gpu_out.device(sycl_device) = gpu_in1 * 1.2f;
68   sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data ,(out.dimensions().TotalSize())*sizeof(float));
69   for (int i = 0; i < sizeDim1; ++i) {
70     for (int j = 0; j < sizeDim2; ++j) {
71       for (int k = 0; k < sizeDim3; ++k) {
72         VERIFY_IS_APPROX(out(i,j,k),
73                          in1(i,j,k) * 1.2f);
74       }
75     }
76   }
77   printf("a=b*1.2f Test Passed\n");
78 
79   /// c=a*b
80   sycl_device.memcpyHostToDevice(gpu_in2_data, in2.data(),(in2.dimensions().TotalSize())*sizeof(float));
81   gpu_out.device(sycl_device) = gpu_in1 * gpu_in2;
82   sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
83   for (int i = 0; i < sizeDim1; ++i) {
84     for (int j = 0; j < sizeDim2; ++j) {
85       for (int k = 0; k < sizeDim3; ++k) {
86         VERIFY_IS_APPROX(out(i,j,k),
87                          in1(i,j,k) *
88                              in2(i,j,k));
89       }
90     }
91   }
92   printf("c=a*b Test Passed\n");
93 
94   /// c=a+b
95   gpu_out.device(sycl_device) = gpu_in1 + gpu_in2;
96   sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
97   for (int i = 0; i < sizeDim1; ++i) {
98     for (int j = 0; j < sizeDim2; ++j) {
99       for (int k = 0; k < sizeDim3; ++k) {
100         VERIFY_IS_APPROX(out(i,j,k),
101                          in1(i,j,k) +
102                              in2(i,j,k));
103       }
104     }
105   }
106   printf("c=a+b Test Passed\n");
107 
108   /// c=a*a
109   gpu_out.device(sycl_device) = gpu_in1 * gpu_in1;
110   sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
111   for (int i = 0; i < sizeDim1; ++i) {
112     for (int j = 0; j < sizeDim2; ++j) {
113       for (int k = 0; k < sizeDim3; ++k) {
114         VERIFY_IS_APPROX(out(i,j,k),
115                          in1(i,j,k) *
116                              in1(i,j,k));
117       }
118     }
119   }
120   printf("c= a*a Test Passed\n");
121 
122   //a*3.14f + b*2.7f
123   gpu_out.device(sycl_device) =  gpu_in1 * gpu_in1.constant(3.14f) + gpu_in2 * gpu_in2.constant(2.7f);
124   sycl_device.memcpyDeviceToHost(out.data(),gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
125   for (int i = 0; i < sizeDim1; ++i) {
126     for (int j = 0; j < sizeDim2; ++j) {
127       for (int k = 0; k < sizeDim3; ++k) {
128         VERIFY_IS_APPROX(out(i,j,k),
129                          in1(i,j,k) * 3.14f
130                        + in2(i,j,k) * 2.7f);
131       }
132     }
133   }
134   printf("a*3.14f + b*2.7f Test Passed\n");
135 
136   ///d= (a>0.5? b:c)
137   sycl_device.memcpyHostToDevice(gpu_in3_data, in3.data(),(in3.dimensions().TotalSize())*sizeof(float));
138   gpu_out.device(sycl_device) =(gpu_in1 > gpu_in1.constant(0.5f)).select(gpu_in2, gpu_in3);
139   sycl_device.memcpyDeviceToHost(out.data(), gpu_out_data,(out.dimensions().TotalSize())*sizeof(float));
140   for (int i = 0; i < sizeDim1; ++i) {
141     for (int j = 0; j < sizeDim2; ++j) {
142       for (int k = 0; k < sizeDim3; ++k) {
143         VERIFY_IS_APPROX(out(i, j, k), (in1(i, j, k) > 0.5f)
144                                                 ? in2(i, j, k)
145                                                 : in3(i, j, k));
146       }
147     }
148   }
149   printf("d= (a>0.5? b:c) Test Passed\n");
150   sycl_device.deallocate(gpu_in1_data);
151   sycl_device.deallocate(gpu_in2_data);
152   sycl_device.deallocate(gpu_in3_data);
153   sycl_device.deallocate(gpu_out_data);
154 }
test_cxx11_tensor_sycl()155 void test_cxx11_tensor_sycl() {
156   cl::sycl::gpu_selector s;
157   Eigen::SyclDevice sycl_device(s);
158   CALL_SUBTEST(test_sycl_cpu(sycl_device));
159 }
160