1 /* Copyright 2016 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 #include "tensorflow/core/kernels/cast_op_impl.h" 17 18 #include "tensorflow/core/util/work_sharder.h" 19 20 namespace tensorflow { 21 22 typedef Eigen::ThreadPoolDevice CPUDevice; 23 typedef Eigen::GpuDevice GPUDevice; 24 25 std::function<void(OpKernelContext*, const Tensor&, Tensor*)> 26 GetCpuCastFromFloat(DataType dst_dtype) { 27 CURRY_TYPES3(CAST_CASE, CPUDevice, float); 28 if (dst_dtype == DT_BFLOAT16) { 29 return [](OpKernelContext* ctx, const Tensor& inp, Tensor* out) { 30 int64 N = out->NumElements(); 31 auto worker_threads = ctx->device()->tensorflow_cpu_worker_threads(); 32 auto work = [&inp, &out](int64 start, int64 end) { 33 FloatToBFloat16(inp.flat<float>().data() + start, 34 out->flat<bfloat16>().data() + start, end - start); 35 }; 36 Shard(worker_threads->num_threads, worker_threads->workers, N, 2, work); 37 }; 38 } 39 40 return nullptr; 41 } 42 43 #if GOOGLE_CUDA 44 std::function<void(OpKernelContext*, const Tensor&, Tensor*)> 45 GetGpuCastFromFloat(DataType dst_dtype) { 46 CURRY_TYPES3(CAST_CASE, GPUDevice, float); 47 CAST_CASE(GPUDevice, float, bfloat16); 48 return nullptr; 49 } 50 #endif // GOOGLE_CUDA 51 52 #ifdef TENSORFLOW_USE_SYCL 53 typedef Eigen::SyclDevice SYCLDevice; 54 std::function<void(OpKernelContext*, const Tensor&, Tensor*)> 55 GetSyclCastFromFloat(DataType dst_dtype) { 56 CURRY_TYPES3_NO_HALF(CAST_CASE, SYCLDevice, float); 57 return nullptr; 58 } 59 #endif // TENSORFLOW_USE_SYCL 60 61 } // namespace tensorflow 62
