diff options
Diffstat (limited to 'tensorflow/compiler/jit/partially_decluster_pass.h')
-rw-r--r-- | tensorflow/compiler/jit/partially_decluster_pass.h | 58 |
1 files changed, 58 insertions, 0 deletions
diff --git a/tensorflow/compiler/jit/partially_decluster_pass.h b/tensorflow/compiler/jit/partially_decluster_pass.h new file mode 100644 index 0000000000..6949b5028e --- /dev/null +++ b/tensorflow/compiler/jit/partially_decluster_pass.h @@ -0,0 +1,58 @@ +/* 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. +==============================================================================*/ + +#ifndef TENSORFLOW_COMPILER_JIT_PARTIALLY_DECLUSTER_PASS_H_ +#define TENSORFLOW_COMPILER_JIT_PARTIALLY_DECLUSTER_PASS_H_ + +#include "tensorflow/core/common_runtime/optimization_registry.h" + +namespace tensorflow { + +// Clones nodes from within a cluster to outside the cluster if profitable. +// +// Today this only clones to avoid device-to-host copies, but in the future we +// may consider other reasons to clone. For instance, we convert this: +// +// ..... +// | +// v +// A_Clustered ====> C_Unclustered +// | +// v +// B_Clustered +// +// to: +// +// ..... +// | | +// | +-------------+ +// | | +// v v +// A_Clustered A_Unclustered ====> C_Unclustered +// | +// v +// B_Clustered +// +// where the ===> arrow has a hostmem source and destination and would entail a +// device to host copy if the source and destination were not in the same XLA +// cluster. +class PartiallyDeclusterPass : public GraphOptimizationPass { + public: + Status Run(const GraphOptimizationPassOptions& options) override; +}; + +} // namespace tensorflow + +#endif // TENSORFLOW_COMPILER_JIT_PARTIALLY_DECLUSTER_PASS_H_ |