Improve general dense singleton bus ordering

[seveibar]

Summary

• prioritize direct center-exit singleton buses using source-field geometry
• keep same-layer embedded singletons behind the wide bus whose source field contains them
• avoid edge-specific or dataset-specific checks; the ordering uses relative geometry, topology, and assigned layers
• update the two AM62L SVG snapshots whose valid route ordering changes

Dataset31 benchmark

Rebased commit 07cff8f, standard configuration: concurrency 4, 120s per sample, sample-default assignment budget.

• 4/12 solved (previous saved baseline: 2/12)
• solved: 02-top-center, 05-right-center, 06-right-bottom-offset, 11-left-center
• every solved case routes and validates 135/135 connections
• eight remaining cases time out at 120s; no partial or error results

Verification

• full suite: 155 passed before snapshot refresh; the only two failures were expected SVG route changes
• refreshed and visually inspected both changed snapshots
• refreshed snapshot tests: 2 passed
• new main northeast failure regression test: passed
• TypeScript typecheck: passed
• git diff check: passed
• post-rebase dataset31 benchmark: 4/12

[vercel]

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

Project	Deployment	Actions	Updated
fanout-solver	Ready	Preview	Sep 4, 2026 4:52pm UTC

[seveibar]

/benchmark

[seveibar]

Benchmark follow-up: the verified 8-vCPU Blacksmith run at commit c4290c7 completed all 12 dataset-31 samples with concurrency 4 and the unchanged 120s per-sample deadline. Result: 3/12 solved on Blacksmith (05, 06, 11); the same routing commit solves 4/12 locally because sample 02 finishes in 71s locally but exceeds 120s on the ARM runner. Run: https://github.com/tscircuit/fanout-solver/actions/runs/33870208474

[seveibar]

/benchmark

[github-actions]

Dataset 31 — AM62L fanout benchmark

Commit: 63f6701. Blacksmith job: success.

Full JSON/Markdown reports and logs

Solved 3/12 selected samples. Completed 12/12; partial 0; errors 0; timeouts 9.

Concurrency: 4; per-sample deadline: 120s; assignment budget: sample defaults.

Only dataset-fanout31-am62l is benchmarked. Solved means validated AM62L fanout with the original constraints, not RAM fanout or inter-chip routing.

Dataset revision: 8c73bef.

Per-sample results

Sample	Status	Routed	Seconds
01-top-left-offset	timeout	0/135	120.08
02-top-center	timeout	0/135	120.09
03-top-right-offset	timeout	0/135	120.07
04-right-top-offset	timeout	0/135	120.07
05-right-center	solved	135/135	8.10
06-right-bottom-offset	solved	135/135	15.31
07-bottom-right-offset	timeout	0/135	120.10
08-bottom-center	timeout	0/135	120.07
09-bottom-left-offset	timeout	0/135	120.04
10-left-bottom-offset	timeout	0/135	120.04
11-left-center	solved	135/135	13.22
12-left-top-offset	timeout	0/135	120.06

[seveibar]

/benchmark

[github-actions]

Dataset 31 — AM62L fanout benchmark

Commit: 9c472e0. Blacksmith job: success.

Full JSON/Markdown reports and logs

Solved 4/12 selected samples. Completed 12/12; partial 0; errors 0; timeouts 8.

Concurrency: 4; per-sample deadline: 120s; assignment budget: sample defaults.

Only dataset-fanout31-am62l is benchmarked. Solved means validated AM62L fanout with the original constraints, not RAM fanout or inter-chip routing.

Dataset revision: 8c73bef.

Per-sample results

Sample	Status	Routed	Seconds
01-top-left-offset	timeout	0/135	120.04
02-top-center	solved	135/135	114.67
03-top-right-offset	timeout	0/135	120.07
04-right-top-offset	timeout	0/135	120.06
05-right-center	solved	135/135	7.92
06-right-bottom-offset	solved	135/135	15.51
07-bottom-right-offset	timeout	0/135	120.08
08-bottom-center	timeout	0/135	120.08
09-bottom-left-offset	timeout	0/135	120.06
10-left-bottom-offset	timeout	0/135	120.05
11-left-center	solved	135/135	13.82
12-left-top-offset	timeout	0/135	120.07