Use an unbounded thread pool for StreamExecutor PjRt clients #36030
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Outdated suggestions cannot be applied.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.
Use an unbounded thread pool for StreamExecutor PjRt clients
If
tsl::thread::ThreadPoolruns out of threads, it runs the callback inline on the thread that calledSchedule(). This pattern is known to be prone to deadlocks (e.g., ifScheduleis called while holding a mutex that the callable needs to acquire, it will deadlock) and may harm the asynchrony in the tail case, so switching to an unbounded thread pool based ontsl::UnboundedWorkQueue.Since
tsl::thread::ThreadPoolis not extensible, this requires replacing the executor type fromtsl::thread::ThreadPooltoAsyncWorkRunner. The existing thread pool is still kept becauseExecutableBuildOptions::set_compile_thread_pool()requirestsl::thread::ThreadPool.Reverts 485e912