Propagate intra_op_parallelism_threads from SessionOptions to xla::LocalClientOptions #30996
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
With CPU Tensorflow, when we set intra_op_parallelism_threads to 1 in the SessionOptions config, the XLA CPU backend still spawns a huge number of threads because xla::LocalClientOptions defaults this setting to -1, which causes the CPU backend to spawn one thread per core.
Fix is to propagate this setting from the SessionOptions into xla::LocalClientOptions when creating the xla::LocalClient.