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
retain each affected trace's unchanged prefix and suffix, reopening only a bounded window around its hottest segment
route the reopened window outside-in from both retained ends, with a hard maximum travel distance per frontier and a safe one-ended fallback
provide an inclusive route-count window so callers can use partial ripping only at graph scales where it wins
preserve the legacy hot path when partial ripping and outside-in routing are disabled
choose completed states region-cost-first, allowing lower-complexity alternatives only inside explicit max-region and total-region-cost envelopes anchored to the first completed state
expose routing/rip/frontier/cost telemetry and keep a checked-in log of the accepted and rejected experiments
The final /benchmark run on all eight committed SRJ18 Pipeline7 inputs reports:
8/8 cases and 2,001/2,001 routes complete on both main and this PR
107.536s total versus 256.486s on main: 2.39x faster
4.877s P50 versus 29.878s on main: 6.13x faster median
average max region cost 1.713 versus 2.333 on main: 26.6% better
average iterations 429,535 versus 1,335,850 on main: 67.8% fewer
The package benchmark's P95 is 66.033s versus 47.871s because its 361-route sample is outside the reliable scale range. The autorouter integration therefore enables the strategy only for 100–350 routes; graphs below and above that window use the compatibility path.
Integration policy
fewer than 100 routes: compatibility behavior
100–350 routes: one full-rip warmup, then bounded partial-rip attempts with region-cost-first selection
more than 350 routes: compatibility behavior
preloaded traces: partial ripping and outside-in routing disabled
complexity tie-breaking: capped relative to the first completed state at 20% max-region-cost growth and 10% total-region-cost growth
Everything after /benchmark is forwarded directly to ./benchmark.sh.
The result comment shows the matching dataset from the latest stored main-branch benchmark artifact beside the PR run; main is not rerun for each PR benchmark request.
Examples:
/benchmark -> all committed completed SRJ18 Pipeline7 cases, core solver
/benchmark --limit 4 -> first four committed Pipeline7 cases
/benchmark --dataset srj18 --concurrency auto -> srj18 samples with runner CPU-count concurrency
/benchmark --sample sample003 -> one committed SRJ18 Pipeline7 case
/benchmark --dataset hg07 --limit 20 --solver poly -> first 20 hg07 samples with the poly solver
/benchmark --dataset hg07 --limit 40 --families default+deep -> first 40 hg07 samples with deeper candidate families
Any PR whose title contains [BENCHMARK TEST] will automatically run the benchmark workflow on PR updates.
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.
Summary
This supersedes #158. The paired integration is tscircuit/tscircuit-autorouter#2051.
Official package benchmark
The final
/benchmarkrun on all eight committed SRJ18 Pipeline7 inputs reports:The package benchmark's P95 is 66.033s versus 47.871s because its 361-route sample is outside the reliable scale range. The autorouter integration therefore enables the strategy only for 100–350 routes; graphs below and above that window use the compatibility path.
Integration policy
Verification
bun run typecheckbun run buildgit diff --checkDetailed trial commands, timings, route completion, and region-cost results are recorded in
experiments/outside-in-partial-rip.md.