Skip to content

restore: re-apply UDMA P2P perf demo (#40)#45

Open
hechangcheng2303 wants to merge 21 commits into
LingquLab:mainfrom
hechangcheng2303:udma-p2p-demo
Open

restore: re-apply UDMA P2P perf demo (#40)#45
hechangcheng2303 wants to merge 21 commits into
LingquLab:mainfrom
hechangcheng2303:udma-p2p-demo

Conversation

@hechangcheng2303

Copy link
Copy Markdown
Contributor

Summary

This PR restores the UDMA P2P perf demo that was merged in #40 and subsequently reverted in #44.

Re-applies the demo with the following changes:

  • UDMA transport/layout updates for P2P perf path
  • P2P perf config, demo kernel, and concurrency sweep scripts
  • Plotting and run scripts for P2P perf measurements
  • Unit tests for P2P perf config and transport layout
  • Perf guide and design/plan docs

Reverts the revert commit a02b856 (PR #44) so the demo is back on main.

Test

  • tests/udma/run_tests.sh
  • tests/udma/demo/run_tilexr_udma_p2p_perf.sh

hechangcheng2303 and others added 21 commits June 25, 2026 20:05
Introduce a new memory_consume transport that copies payload through the
IPC peer window and synchronizes via SyncCollectives outer flags, so both
ranks stay active. Wire the host launch path, perf config helpers
(P2PTransportUsesIpc / P2PTransportBothRanksActive), sweep defaults, unit
tests, and a source-guard test.
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