Improve Python bootstrap thread and process-group handling - #889
Open
Binyang Li (Binyang2014) wants to merge 3 commits into
Open
Improve Python bootstrap thread and process-group handling#889Binyang Li (Binyang2014) wants to merge 3 commits into
Binyang Li (Binyang2014) wants to merge 3 commits into
Conversation
Allow other Python threads to make progress while blocking bootstrap point-to-point operations are in flight. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
|
Azure Pipelines: There may be pipelines that require an authorized user to comment /azp run to run. |
Binyang Li (Binyang2014)
added a commit
that referenced
this pull request
Aug 27, 2026
The bootstrap binding update is now isolated in PR #889 so the EP implementation PR no longer carries this unrelated core change. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Use a group-aware object broadcast on the backend-compatible device instead of a fixed-size serialized tensor. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Binyang Li (Binyang2014)
changed the title
Let broadcast_object_list choose CUDA for NCCL process groups and CPU for other backends instead of duplicating that logic. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Updates Python bootstrap communication for safer threading and Torch subgroup support.
Changes:
- Releases the Python GIL during blocking bootstrap send/receive operations.
- Broadcasts unique IDs through the supplied Torch process group.
- Removes fixed-size serialization logic.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
python/mscclpp/_core/comm.py |
Uses Torch object broadcast with subgroup-aware root mapping. |
python/csrc/core_py.cpp |
Releases the GIL around bootstrap send/receive bindings. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Comment on lines
+56
to
+57
| group_root = dist.get_global_rank(torch_group, 0) | ||
| dist.broadcast_object_list(object_list, src=group_root, group=torch_group) |
| self->send(data, size, peer, tag); | ||
| }, | ||
| nb::arg("data"), nb::arg("size"), nb::arg("peer"), nb::arg("tag")) | ||
| nb::arg("data"), nb::arg("size"), nb::arg("peer"), nb::arg("tag"), nb::call_guard<nb::gil_scoped_release>()) |
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
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.
Summary
send/recvbindingsThis isolates the non-EP bootstrap changes from #852.
Validation
./tools/lint.shcmake --build build --target mscclpp_py -j$(nproc)python3 -m py_compile python/mscclpp/_core/comm.py