You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We compute the uncovered region directly from a marker cover, adapting EXTRA's fused area/cover operation from section 7.2. Recursive extraction caches the cover without eagerly constructing its function. Subtraction follows shared cofactors and rebuilds each uncovered decision once. Native range merging still materializes a suffix's function when needed to combine equivalent branches.
This is an experiment, not a demonstrated optimization. Against f9df0b826c, repeated conversion with 16 and 32 conflicting extras is 24.5% and 14.6% slower. Small warm conversions are within 3%. We should not merge this version for runtime performance.
Warm medians in microseconds; lower is better. format includes conflict absorption and serialization.
Fresh-process results are mixed: two-clause conversion improves from 17.2 to 13.8 µs, single-clause conversion regresses from 13.1 to 15.0 µs, and 32-extra conversion regresses from 89.7 to 98.5 µs.
These measurements compare 60031437c4 directly with its parent on a Linux devbox, using three randomized warm cohorts and 200 fresh processes per small case (30 per larger case). An additional CPU confirmed the conflict-heavy regression, though absolute timings varied across cores. These are component measurements, not end-to-end lock timings.
System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.
Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.
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
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.
Based on #22651.
We compute the uncovered region directly from a marker cover, adapting EXTRA's fused area/cover operation from section 7.2. Recursive extraction caches the cover without eagerly constructing its function. Subtraction follows shared cofactors and rebuilds each uncovered decision once. Native range merging still materializes a suffix's function when needed to combine equivalent branches.
This is an experiment, not a demonstrated optimization. Against
f9df0b826c, repeated conversion with 16 and 32 conflicting extras is 24.5% and 14.6% slower. Small warm conversions are within 3%. We should not merge this version for runtime performance.Warm medians in microseconds; lower is better.
formatincludes conflict absorption and serialization.Fresh-process results are mixed: two-clause conversion improves from 17.2 to 13.8 µs, single-clause conversion regresses from 13.1 to 15.0 µs, and 32-extra conversion regresses from 89.7 to 98.5 µs.
These measurements compare
60031437c4directly with its parent on a Linux devbox, using three randomized warm cohorts and 200 fresh processes per small case (30 per larger case). An additional CPU confirmed the conflict-heavy regression, though absolute timings varied across cores. These are component measurements, not end-to-end lock timings.Full 25-case comparison, raw samples, and reproduction scripts.
All 66,044 exhaustive Boolean/range cases render identically to the base. The relevant unit and lock/conflict integration tests pass.