PR #10811: Restore ReshapeDecomposer And LayoutNormalization Passes in GPU Compiler
Imported from GitHub PR https://github.com/openxla/xla/pull/10811 Restores individual ReshapeDecomposer and LayoutNormalization passes in the GPU compiler previously removed in #9852 and fixes failures in cudnn_norm_rewriter_test.cc. Copybara import of the project: -- 52065712a6bd4f6bb874575d82b4d99b8f2d16dd by Philipp Hack <phack@nvidia.com>: Restore ReshapeDecomposer and LayoutNormalization passes. Merging this change closes #10811 FUTURE_COPYBARA_INTEGRATE_REVIEW=https://github.com/openxla/xla/pull/10811 from philipphack:u_layer_reshape_decomposer_xla 52065712a6bd4f6bb874575d82b4d99b8f2d16dd PiperOrigin-RevId: 617990157
Showing
- tensorflow/compiler/mlir/tfrt/ir/mlrt/tf_mlrt_ops.td 18 additions, 0 deletionstensorflow/compiler/mlir/tfrt/ir/mlrt/tf_mlrt_ops.td
- tensorflow/compiler/mlir/tfrt/tests/mlrt/inline.mlir 1 addition, 1 deletiontensorflow/compiler/mlir/tfrt/tests/mlrt/inline.mlir
- tensorflow/compiler/mlir/tfrt/tests/mlrt/tf_to_mlrt.mlir 5 additions, 6 deletionstensorflow/compiler/mlir/tfrt/tests/mlrt/tf_to_mlrt.mlir
- tensorflow/compiler/mlir/tfrt/transforms/mlrt/BUILD 3 additions, 0 deletionstensorflow/compiler/mlir/tfrt/transforms/mlrt/BUILD
- tensorflow/compiler/mlir/tfrt/transforms/mlrt/tf_to_mlrt.cc 15 additions, 0 deletionstensorflow/compiler/mlir/tfrt/transforms/mlrt/tf_to_mlrt.cc
- tensorflow/core/tfrt/mlrt/kernel/BUILD 3 additions, 0 deletionstensorflow/core/tfrt/mlrt/kernel/BUILD
- tensorflow/core/tfrt/mlrt/kernel/kernel.cc 37 additions, 0 deletionstensorflow/core/tfrt/mlrt/kernel/kernel.cc
- third_party/xla/xla/service/gpu/gpu_compiler.cc 4 additions, 0 deletionsthird_party/xla/xla/service/gpu/gpu_compiler.cc
Please register or sign in to comment