experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Remove left-over declaration for buffer assignment pass that has since been removed.

Browse Source 
PiperOrigin-RevId: 316841221
Change-Id: I9c38becf498cfaaa7edb9490948890d52e49e737
This commit is contained in:
Stephan Herhut 2020-06-17 00:55:02 -07:00 committed by TensorFlower Gardener
parent af92698487
commit 02aad07710
1 changed files with 0 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
tensorflow/compiler/mlir/xla/transforms/passes.h
View File

@ -115,24 +115,6 @@ std::unique_ptr<Pass> createLhloCopyRemovalPass();
std::unique_ptr<OperationPass<FuncOp>> createLegalizeLhloToParallelLoopsPass();
} // namespace xla_lhlo
namespace xla {
/// Moves alloc nodes (and their associated dealloc nodes - if any) into the
/// right positions. If there is no associated dealloc node for a given alloc
/// node, this pass will automatically insert a proper dealloc node in the right
/// place. The intended use case of this pass is to store SSA values into
/// buffers using load/store operations. For this purpose, you need to know
/// proper positions to place the required allocs and deallocs.
/// 1) Note that the function signatures and all types for which buffers should
/// be allocated need to be converted in advance.
/// 2) All required alloc nodes have the be inserted in advance.
/// 3) Note that the current implementation does not support loops.
/// Refer to the class mlir::xla::BufferAssignmentLegalizer for more
/// information.
std::unique_ptr<OperationPass<FuncOp>> createBufferAssignmentPass();
} // namespace xla
} // namespace mlir
#endif // TENSORFLOW_COMPILER_MLIR_XLA_TRANSFORMS_PASSES_H_