1 /* Copyright 2015 The TensorFlow Authors. All Rights Reserved. 2 3 Licensed under the Apache License, Version 2.0 (the "License"); 4 you may not use this file except in compliance with the License. 5 You may obtain a copy of the License at 6 7 http://www.apache.org/licenses/LICENSE-2.0 8 9 Unless required by applicable law or agreed to in writing, software 10 distributed under the License is distributed on an "AS IS" BASIS, 11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 See the License for the specific language governing permissions and 13 limitations under the License. 14 ==============================================================================*/ 15 16 #if GOOGLE_CUDA 17 18 #define EIGEN_USE_GPU 19 20 #include "tensorflow/core/framework/bfloat16.h" 21 #define SPECIALIZE_FOR_GPUS 22 #include "tensorflow/core/kernels/cast_op.h" 23 #undef SPECIALIZE_FOR_GPUS 24 25 namespace tensorflow { 26 namespace functor { 27 28 typedef Eigen::GpuDevice GPUDevice; 29 30 CAST_FUNCTORS(GPUDevice); 31 32 #define DEFINE(O, I) template struct CastFunctor<GPUDevice, O, I> 33 34 #define DEFINE_ALL_FROM(in_type) \ 35 DEFINE(in_type, bool); \ 36 DEFINE(in_type, uint8); \ 37 DEFINE(in_type, uint16); \ 38 DEFINE(in_type, uint32); \ 39 DEFINE(in_type, uint64); \ 40 DEFINE(in_type, int8); \ 41 DEFINE(in_type, int16); \ 42 DEFINE(in_type, int32); \ 43 DEFINE(in_type, int64); \ 44 DEFINE(in_type, Eigen::half); \ 45 DEFINE(in_type, float); \ 46 DEFINE(in_type, double); \ 47 DEFINE(in_type, std::complex<float>); \ 48 DEFINE(in_type, std::complex<double>) 49 50 DEFINE_ALL_FROM(bool); 51 DEFINE_ALL_FROM(uint8); 52 DEFINE_ALL_FROM(uint16); 53 DEFINE_ALL_FROM(uint32); 54 DEFINE_ALL_FROM(uint64); 55 DEFINE_ALL_FROM(int8); 56 DEFINE_ALL_FROM(int16); 57 DEFINE_ALL_FROM(int32); 58 DEFINE_ALL_FROM(int64); 59 DEFINE_ALL_FROM(double); 60 DEFINE_ALL_FROM(std::complex<double>); 61 DEFINE(float, bfloat16); 62 63 #define DEFINE_ALL_TO_FLOAT(out_type) \ 64 DEFINE(out_type, bool); \ 65 DEFINE(out_type, uint8); \ 66 DEFINE(out_type, uint16); \ 67 DEFINE(out_type, uint32); \ 68 DEFINE(out_type, uint64); \ 69 DEFINE(out_type, int8); \ 70 DEFINE(out_type, int16); \ 71 DEFINE(out_type, int32); \ 72 DEFINE(out_type, int64); \ 73 DEFINE(out_type, Eigen::half); \ 74 DEFINE(out_type, float); \ 75 DEFINE(out_type, std::complex<float>) 76 77 #define DEFINE_ALL_TO_HALF(out_type) \ 78 DEFINE(out_type, bool); \ 79 DEFINE(out_type, uint8); \ 80 DEFINE(out_type, uint16); \ 81 DEFINE(out_type, uint32); \ 82 DEFINE(out_type, uint64); \ 83 DEFINE(out_type, int8); \ 84 DEFINE(out_type, int16); \ 85 DEFINE(out_type, int32); \ 86 DEFINE(out_type, int64); \ 87 DEFINE(out_type, Eigen::half) 88 89 DEFINE_ALL_TO_HALF(Eigen::half); 90 DEFINE_ALL_TO_HALF(bfloat16); 91 DEFINE_ALL_TO_FLOAT(float); 92 DEFINE_ALL_TO_FLOAT(std::complex<float>); 93 94 #undef DEFINE_ALL_TO_FLOAT 95 #undef DEFINE_ALL_TO_HALF 96 #undef DEFINE_ALL_FROM 97 #undef DEFINE 98 99 } // end namespace functor 100 } // end namespace tensorflow 101 102 #endif // GOOGLE_CUDA 103