Conversation
Reset per-branch streaks when the previous step was not cached. Remove redundant cap resets and simplify hit counting. Cover interleaved warmup, dynamic computation masks, both CFG orders, and repeated cache checks in regression tests.
This branch has not been deployed
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This extends #1114 to handle forced-compute steps that return before the consecutive-cache limit check.
With MC=1, a cache hit sets the counter to 1. If the next step is forced to compute by interleaved warmup or a dynamic computation mask, the counter remains 1 and incorrectly blocks the following cache opportunity.
The change resets each branch’s streak at the start of a step when the preceding step was not cached. It also simplifies hit counting to increment by one and removes the reset from the cap check.
Test coverage includes interleaved warmup, dynamic computation masks, both CFG execution orders, and repeated checks of the same step.