Reuse flat scope plans when assertions preserve local order

[charliermarsh]

We reuse flat function inputs during optimized compilation when removing assertions cannot change local ordering. We record whether each name first occurs inside an assertion without growing the stored name entries, retaining the existing collector when bound-name order or private aliases need it.

Matched Python 3.15 O2 profiles use about 0.26% fewer instructions for Mypy and 3.57% fewer for SymPy. Mypy improves by 0.28% at Python 3.13 O1; the O0 controls increase by about 0.04%. All measured outputs are byte-identical.

The 177-case instruction geometric mean improves by 0.09%, with individual changes from -13.23% to +2.78%. The largest regression is the O2 annotation microbenchmark. Native peak RSS falls about 1.45% for Mypy and increases about 0.3 MiB for SymPy; timing is noisy, so we make no native speed claim.

CodSpeed reports about 13% higher efficiency on the O1 and O2 large-dataset benchmarks, with the other 175 measured cases unchanged. It flags significant system time in the O1 async microbenchmark.

Full compatibility validation passes for Python 3.12 through 3.15 at all optimization levels, with unchanged results.

[codspeed]

Merging this PR will improve performance by 13.21%

⚠️ 1 benchmark spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• micro_async

⚡ 2 improved benchmarks
✅ 175 untouched benchmarks
⏩ 7 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	large_dataset	14 ms	12.3 ms	+13.24%
⚡	large_dataset	14 ms	12.4 ms	+13.19%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing charlie/codex-flat-scope-order-main-20261017 (72ea685) with main (2ecbc63)}

Footnotes

1. 7 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩