Skip to content

Add offline collective simulation tool#37

Open
Kur0x wants to merge 24 commits into
mainfrom
codex/collective-sim
Open

Add offline collective simulation tool#37
Kur0x wants to merge 24 commits into
mainfrom
codex/collective-sim

Conversation

@Kur0x

@Kur0x Kur0x commented Jun 20, 2026

Copy link
Copy Markdown
Member

Summary

  • Add a standalone offline collective simulation tool under tools/collective_sim
  • Model generic collective DAGs with communication-buffer validation, AllGather symbolic correctness, 1D-Clos topology timing, and SDMA/P2P/uplink calibration curves
  • Add 64P and 128P AllGather examples, CLI validate/run/sweep/report, and JSON/CSV/HTML reports
  • Document the no-hardware workflow and add boundary tests to prevent runtime hardware dependencies

Test plan

  • PYTHONPATH=tools/collective_sim python3 -m unittest discover -s tests/collective_sim -v (Ran 42 tests, OK)
  • CLI validate for 64P and 128P examples
  • CLI run/sweep/report generated expected artifacts
  • git diff --check 66e220f94a062f69c719ad3357e7179932f80301..HEAD

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant