Remove pass-local ExportPass replays (#22156) - #22156
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22156
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 Cancelled Job, 1 Unrelated FailureAs of commit 445bd6f with merge base f6f3f42 ( CANCELLED JOB - The following job was cancelled. Please retry:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@mcremon-meta has exported this pull request. If you are a Meta employee, you can view the originating Diff in D115500386. |
This PR needs a
|
Summary: - replace six pass-local blank `ExportPass` replays with local metadata ownership - preserve `GraphModule`/graph identity and unchanged no-op behavior - recompute bounded metadata frontiers for rewired Helios and Cadence chains - keep fbcode/xplat mirrors synchronized Performance: Avoids one whole-graph replay per modified pass, estimated at roughly 0.4-0.75 ms per graph node (about 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes). Reviewed By: ethansfng Differential Revision: D115500386
a03202a to
42e24cf
Compare
Summary: - replace six pass-local blank `ExportPass` replays with local metadata ownership - preserve `GraphModule`/graph identity and unchanged no-op behavior - recompute bounded metadata frontiers for rewired Helios and Cadence chains - keep fbcode/xplat mirrors synchronized Performance: Avoids one whole-graph replay per modified pass, estimated at roughly 0.4-0.75 ms per graph node (about 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes). Reviewed By: ethansfng Differential Revision: D115500386
42e24cf to
445bd6f
Compare
Summary:
ExportPassreplays with local metadata ownershipGraphModule/graph identity and unchanged no-op behaviorPerformance:
Avoids one whole-graph replay per modified pass, estimated at roughly 0.4-0.75 ms per graph node (about 4-8 ms for 10 nodes, 40-75 ms for 100 nodes, and 0.4-0.75 s for 1,000 nodes).
Reviewed By: ethansfng
Differential Revision: D115500386