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 (at) codeplay.com> 9 // Benoit Steiner <benoit.steiner.goog (at) 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 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 } 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
