Reduce vias with multiobjective hypergraph routing

[AnasSarkiz]

Summary

• track estimated vias, routing risk, and path length separately on routing labels
• keep one strict risk-first A* label per directed hop while preserving via/length data for goal and completed-solution comparison
• support multiple legal start/end ports and preserve the selected endpoint layer through reachability, replay, serialization, and section optimization
• compare completed and section-optimized solutions using via count with max/total region-risk constraints

Why

Endpoint layer choices and layer changes were previously collapsed into the existing region-cost scalar. That forced multi-layer endpoints onto one configured port and discarded lower-via completed solutions.

The first implementation also maintained a live Pareto archive that active search did not consume. CPU profiling showed that archive and intermediate candidate copying dominated the routing hot path. This PR now keeps the multiobjective data used by goal and completed-solution selection without maintaining the unused frontier.

Validation

• bun test: 98 passed, 0 failed
• bun run typecheck: passed
• bun run build: passed
• srj18 sample009 port pathing: 31.54s -> 14.47s with the same 1,686,408 iterations
• srj18 sample001: 28.1s -> 17.7s, 174 vias, completed, relaxed DRC passed
• downstream Dataset01 first 20 scenarios before the hot-path-only optimization:
  • completion: 100% -> 100%
  • relaxed DRC: 95% -> 100%
  • paired vias: 285 -> 268
  • 7 improved, 2 regressed, 11 unchanged
  • no new solver failures or timeouts

Follow-up

The companion autorouter PR remains draft while CI reruns the full benchmark matrix. The hot-path optimization does not change search iterations or the validated sample001 via result.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
tiny-hypergraph	Ready	Preview, Comment	Jul 15, 2026 11:29pm

[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]

/benchmark --dataset srj18

[AnasSarkiz]

/benchmark

[AnasSarkiz]

/benchmark --dataset srj18

[tscircuitbot]

Tiny Hypergraph Benchmark Results

Run	Dataset	Solver	Families	Success	P50	P95
Main	srj18	core	default	31.3%	2.389s	10.653s
PR	srj18	core	default	31.3%	5.111s	34.661s

Main Branch Details

Metric	Value
Dataset	srj18
Solver	core
Families	default
Samples	16
Success rate	31.3%
Improved rate	60.0%
Zero final max region cost rate	0.0%
Avg baseline max region cost	0.565
Avg final max region cost	0.499
Avg max region delta	0.066
Avg candidate count	5.200
Avg duration	3.034s
P50 duration	2.389s
P95 duration	10.653s

Sample	Status	Baseline	Final	Delta	Attempts	Duration	Error
sample001	success	0.670	0.579	0.090	4	3.193s
sample002	success	0.391	0.358	0.033	6	1.694s
sample003	success	0.429	0.429	0.000	5	0.796s
sample004	failed	n/a	n/a	n/a	0	0.029s	Error: Static reachability precheck failed: 2 route(s) have no legal path under the current reservation and start-reg...
sample005	success	0.423	0.423	0.000	7	1.309s
sample006	failed	n/a	n/a	n/a	0	0.346s	Error: Static reachability precheck failed: 9 route(s) have no legal path under the current reservation and start-reg...
sample007	failed	n/a	n/a	n/a	0	10.653s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample008	failed	n/a	n/a	n/a	0	2.389s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample009	failed	n/a	n/a	n/a	0	0.272s	Error: Static reachability precheck failed: 2 route(s) have no legal path under the current reservation and start-reg...
sample010	failed	n/a	n/a	n/a	0	4.343s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample011	failed	n/a	n/a	n/a	0	5.092s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample012	failed	n/a	n/a	n/a	0	7.087s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample013	failed	n/a	n/a	n/a	0	0.330s	Error: Static reachability precheck failed: 11 route(s) have no legal path under the current reservation and start-re...
sample014	failed	n/a	n/a	n/a	0	3.402s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample015	failed	n/a	n/a	n/a	0	5.069s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample016	success	0.912	0.707	0.205	4	2.547s

PR Details

Metric	Value
Dataset	srj18
Solver	core
Families	default
Samples	16
Success rate	31.3%
Improved rate	60.0%
Zero final max region cost rate	0.0%
Avg baseline max region cost	0.555
Avg final max region cost	0.489
Avg max region delta	0.066
Avg candidate count	5.000
Avg duration	9.577s
P50 duration	5.111s
P95 duration	34.661s

Sample	Status	Baseline	Final	Delta	Attempts	Duration	Error
sample001	success	0.670	0.579	0.090	4	7.411s
sample002	success	0.391	0.358	0.033	6	2.279s
sample003	success	0.378	0.378	0.000	4	1.271s
sample004	failed	n/a	n/a	n/a	0	0.053s	Error: Static reachability precheck failed: 2 route(s) have no legal path under the current reservation and start-reg...
sample005	success	0.423	0.423	0.000	7	2.142s
sample006	failed	n/a	n/a	n/a	0	0.470s	Error: Static reachability precheck failed: 9 route(s) have no legal path under the current reservation and start-reg...
sample007	failed	n/a	n/a	n/a	0	34.661s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample008	failed	n/a	n/a	n/a	0	7.410s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample009	failed	n/a	n/a	n/a	0	0.234s	Error: Static reachability precheck failed: 2 route(s) have no legal path under the current reservation and start-reg...
sample010	failed	n/a	n/a	n/a	0	14.697s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample011	failed	n/a	n/a	n/a	0	18.102s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample012	failed	n/a	n/a	n/a	0	32.795s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample013	failed	n/a	n/a	n/a	0	0.241s	Error: Static reachability precheck failed: 11 route(s) have no legal path under the current reservation and start-re...
sample014	failed	n/a	n/a	n/a	0	11.393s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample015	failed	n/a	n/a	n/a	0	14.967s	Error: TinyHyperGraphSectionPipelineSolver ran out of iterations
sample016	success	0.912	0.707	0.205	4	5.111s

Workflow: View run
Artifact: https://github.com/tscircuit/tiny-hypergraph/actions/runs/29456844290