diff options
| author |  A. Unique TensorFlower <gardener@tensorflow.org> | 2017-11-17 14:16:09 -0800 |
|---|---|---|
| committer | TensorFlower Gardener <gardener@tensorflow.org> | 2017-11-17 14:20:28 -0800 |
| commit | 3f888e1539db5551cfcf9ee837a0555c224e0018 (patch) | |
| tree | 5f2df45e666fc15e370e6c029bf0712ee65d53ed /tensorflow/compiler/xla/service/llvm_compiler.cc | |
| parent | d79dd4993061670c1ec5ea01db3022f28d72d0a3 (diff) | |
Add a Compiler::BuildExecutable interface that compiles the given Hlo module without optimizations.
PiperOrigin-RevId: 176158846
Diffstat (limited to 'tensorflow/compiler/xla/service/llvm_compiler.cc')
| -rw-r--r-- | tensorflow/compiler/xla/service/llvm_compiler.cc | 4 |
1 files changed, 3 insertions, 1 deletions
diff --git a/tensorflow/compiler/xla/service/llvm_compiler.cc b/tensorflow/compiler/xla/service/llvm_compiler.cc index ba0304fb8c..34f3419269 100644 --- a/tensorflow/compiler/xla/service/llvm_compiler.cc +++ b/tensorflow/compiler/xla/service/llvm_compiler.cc @@ -27,8 +27,10 @@ StatusOr<std::vector<std::unique_ptr<Executable>>> LLVMCompiler::Compile( "Model partitioning not implemented for the CPU/GPU compilers!"); } + TF_ASSIGN_OR_RETURN( + modules[i], RunHloPasses(std::move(modules[i]), stream_execs[i][0])); TF_ASSIGN_OR_RETURN(std::unique_ptr<Executable> executable, - Compile(std::move(modules[i]), stream_execs[i][0])); + RunBackend(std::move(modules[i]), stream_execs[i][0])); result.push_back(std::move(executable)); } |