Improve performance of Graph.transitive_closure()

[ahal]

No description provided.

[codspeed-hq]

Merging this PR will regress 1 benchmark

⚡ 3 improved benchmarks
❌ 1 regressed benchmark
✅ 20 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	test_transitive_closure[fan]	19 ms	35.3 ms	-46.05%
⚡	test_transitive_closure[linear]	101,008.4 ms	40.2 ms	×2,500
⚡	test_transitive_closure[diamond]	153.6 s	2 s	×77
⚡	test_transitive_closure[btree]	46.3 ms	26.3 ms	+76.04%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ahal:ahal/xvuotksxtnmq (99a789c) with main (d7efa4b)¹}

Footnotes

1. No successful run was found on main (305a07f) during the generation of this report, so d7efa4b was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[ahal]

I don't think there's a way to not regress the fan test. This is a tradeoff, but should be worth it as long as the real graphs have linear / diamond segments in them (which Gecko does).

Ultimately none of this is going to have any measurable impact on real world graphs.. but this does at least make the codspeed task fast(ish) again.

I think we should take it.

[bhearsum]

If you could, maybe a quick before/after test with at least one gecko graph as a sanity check would be in order before this lands?

[ahal]

If you could, maybe a quick before/after test with at least one gecko graph as a sanity check would be in order before this lands?

There's too much noise to tell, especially due to that one 70s outlier. Seems like there's not a regression at least (as expected)

Before:

$ hyperfine "./mach taskgraph morphed" -w 1 -r 5
Benchmark 1: ./mach taskgraph morphed
  Time (mean ± σ):     51.263 s ± 11.194 s    [User: 114.538 s, System: 37.939 s]
  Range (min … max):   38.233 s … 69.037 s    5 runs

After:

$ hyperfine "./mach taskgraph morphed" -w 1 -r 5
Benchmark 1: ./mach taskgraph morphed
  Time (mean ± σ):     44.317 s ±  6.868 s    [User: 116.071 s, System: 34.583 s]
  Range (min … max):   34.272 s … 51.743 s    5 runs