Split cwise igammas GPU code into multiple files

This file was a bottleneck during compilation, often taking many minutes to compile. In local testing this change reduces the wall-clock build time for the affected kernels from 141s to 87s.

PiperOrigin-RevId: 229179013

tensorflow/core/kernels/cwise_op_gpu_igammas_double.cu.cc

@@ -0,0 +1,29 @@
+/* Copyright 2018 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#if GOOGLE_CUDA
+
+#include "tensorflow/core/kernels/cwise_ops_gpu_common.cu.h"
+#include "tensorflow/core/kernels/cwise_ops_gpu_gradients.cu.h"
+
+namespace tensorflow {
+namespace functor {
+DEFINE_BINARY1(igamma, double);
+DEFINE_BINARY1(igamma_grad_a, double);
+DEFINE_BINARY1(igammac, double);
+}  // namespace functor
+}  // namespace tensorflow
+
+#endif  // GOOGLE_CUDA

tensorflow/core/kernels/cwise_op_gpu_igammas_float.cu.cc

@@ -20,9 +20,9 @@ limitations under the License.
 
 namespace tensorflow {
 namespace functor {
-DEFINE_BINARY2(igamma, float, double);
+DEFINE_BINARY1(igamma, float);
-DEFINE_BINARY2(igamma_grad_a, float, double);
+DEFINE_BINARY1(igamma_grad_a, float);
-DEFINE_BINARY2(igammac, float, double);
+DEFINE_BINARY1(igammac, float);
 }  // namespace functor
 }  // namespace tensorflow