docs(perf): close out the despeckle optimization round with final baselines#40
Merged
Conversation
…elines Regenerated both committed baselines on the full optimization stack: - pipeline (200-page fixture): conv 9.03s at -j8 vs the original 14.48s baseline (-37.6%); despeckle stage 5.21s vs 10.19s (-49%), its share down from 71.6% to 57.7%. At -j1: conv 35.63s vs 49.85s (-28.5%). - cleaner: clean() without component stats 106.8ms vs the 174.9ms round-opening baseline (-39%); the remaining distribution is the selectBySize block (~70%), dilate 11%, write/read ~7%. Also records the round's negative results, measured rather than assumed: the smallHoles De Morgan flip was implemented, measured at zero effect (the inverted-select penalty lives in pixConnComp's extraction of the giant background component, not in the re-render), and reverted; the single-labeling fusion was not pursued (predicted -4..6% conv sits at the acceptance gate with the round's largest FFM surface cost). The next levers, should they ever be needed: the extraction cost inside pixSelectBySize, or the register stage (now 35% of conv). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The round's result (200-page fixture, committed baselines)
Delivered by #31 (op-level benchmark), #32 (skip the metrics-only counting passes), #33 (exact DWA morphology routing).
Negative results, measured rather than assumed
pixConnComp's extraction of the giant background component, not in the re-render the flip eliminated.If more is ever needed
The remaining clean() distribution is the
selectBySizeblock (~70%); the next levers are the extraction cost insidepixSelectBySize, or the register stage (now 35% of conv).🤖 Generated with Claude Code