Model physical trace occupancy after baseline routing

[AnasSarkiz]

Summary

• replace the dimensionless distinct-net density penalty with a physical per-layer occupancy model
• derive swept copper area from segment length and configurable trace pitch (minimum width + clearance, both defaulting to 0.1 mm)
• charge only shared occupancy by excluding the longest trace corridor on each layer
• activate occupancy scoring only after a complete baseline solution has been captured
• preserve occupancy state through snapshots, section solving, bus previews, and polygon routing

Why sample 6 regressed before

The previous cost used distinct-net count divided by the region's shortest dimension and applied it during the initial A* search. In narrow subdivided regions it could dominate the existing feasibility cost, changing route ownership and selective-rerip ordering until sample 6 exhausted its 2,000,000-iteration budget.

This implementation uses the actual swept trace length and region area instead. More importantly, it does not alter initial route construction: physical occupancy activates only after a complete baseline exists, so optimization can restore that completed solution rather than sacrificing reachability.

Validation

• bun test --timeout 9999999 (98 pass)
• bun run typecheck
• bun run format:check
• Pipeline 7 SRJ18 sample 6 focused selective-rerip test passes in 19.7s through the dependent autorouter PR

SRJ18 evidence through autorouter PR #1677

On samples 1, 5, 7, and 11, remaining relaxed DRC errors decrease from 13 on current main to 10 with this branch. Samples 1 and 5 remain DRC-clean; sample 7 improves from 7 to 6 errors and sample 11 from 6 to 4. Sample 6 completes end-to-end in 293.5s under a 600s cap instead of failing in TinyHypergraph near 120s.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
tiny-hypergraph	Ready	Preview, Comment	Jul 20, 2026 3:50pm

[tscircuitbot]

Benchmark This PR

Run benchmarks by commenting on this PR:

/benchmark [benchmark.sh args...]
/benchmark --dataset <hg07|srj18> --limit <n> --solver <core|poly> --families <default|default+deep|all> --concurrency <n|auto>

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 -> hg07, all samples, core solver
• /benchmark --dataset srj18 --limit 20 -> first 20 srj18 samples
• /benchmark --dataset srj18 --concurrency auto -> srj18 samples with runner CPU-count concurrency
• /benchmark --sample sample002 -> one hg07 sample
• /benchmark --limit 20 --solver poly -> first 20 hg07 samples with the poly solver
• /benchmark --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.

[AnasSarkiz]

Closing this unmerged occupancy dependency together with tscircuit-autorouter#1677 because the end-to-end srj18 relaxed-DRC pass rate did not improve.