feat: Implement zero-copy AsyncVectorEnv via shared memory for parallel execution - #1550
Open
shivansh023023 wants to merge 4 commits into
Open
feat: Implement zero-copy AsyncVectorEnv via shared memory for parallel execution#1550shivansh023023 wants to merge 4 commits into
shivansh023023 wants to merge 4 commits into
Conversation
Collaborator
|
Thanks! |
lanctot
self-requested a review
July 11, 2026 13:11
lanctot
requested changes
Jul 11, 2026
lanctot
left a comment
lanctot
left a comment
Collaborator
There was a problem hiding this comment.
- Can you add a test that shows it working on a simple game?
- Can you add the test to the python tests in python/CMakeLists.txt
lanctot
added
the
waiting
Contributor
Author
Done ! |
lanctot
removed
the
waiting
Collaborator
|
The tests failed. Can you take a look? Did they pass locally? |
lanctot
reviewed
Jul 12, 2026
| remote.recv() | ||
| for remote in self.remotes: | ||
| remote.send(("close", None)) | ||
| for p in self.processes: |
Collaborator
There was a problem hiding this comment.
Is there a race condition here?
Contributor
Author
There was a problem hiding this comment.
hi @lanctot there was indeed a potential race condition during shutdown if a worker was stuck on an IPC boundary when close() was triggered.
I've pushed a fix that handles process termination safely. I also resolved an underlying multi-player data alignment bug where info_state arrays were hitting dimension mismatches when unpacked at the master process level.
…_state dimension mapping
…ibility via dynamic short_name extraction
Contributor
Author
|
@lanctot i have implemented some chages , can you re run the checks |
Contributor
Author
|
hi @lanctot all the checks are passing now ! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Pull Request: feat: Implement zero-copy AsyncVectorEnv using shared memory
Description
This PR introduces
AsyncVectorEnvtoopen_spiel/python/vector_env.py, enabling true parallel execution of environments across multiple CPU cores.Key Architectural Details
multiprocessing.sharedctypes.RawArray.time_step.observations["info_state"]directly into pre-allocated shared NumPy views.multiprocessing.Pipeis strictly reserved for small control payloads (rewards, done flags, step types), completely eliminating the need to serialize massive multi-dimensional arrays.chessrunning 1,000 steps, this shared memory architecture achieved a ~2.87x speedup over the synchronousSyncVectorEnv, fully bypassing the IPC overhead that typically throttles complex state transmission.