Breakwater

Backtest / #1256

perf(utils): precompile the wheel project-name pattern

pypa/packaging#1256 · merged 2026-06-13 · base 54c697714e → head d2ac6ba3f8 · ran on Docker

PASS

No behaviour divergence found by any probe.

† Merged after Nemotron 3 Ultra was released (2026-06-04): the model cannot have seen this PR or its fixes.

Performance · parse_wheel_filename

No realistic input family measured slower.

Measured in backtest packaging-v2 (same commits, run bt-1256-85d56f, image bw-pair:0e393c2af8d05bae). The semantic results above are from a later run; perf doesn't use the semantic planner, so it was not re-measured.

head slowerhead faster0.84×1×input size (characters)F3F3: 0.91× at ~30 charsF3: 0.91× at ~30 charsF3: 0.91× at ~32 charsF3: 0.91× at ~32 charsF3: 0.91× at ~32 charsF3: 0.91× at ~32 charsF3: 0.91× at ~32 chars
  • F3
Head/base cachegrind instructions per call, by input size (log-log). Each line is one input family Nemotron proposed; points above 1× are where the "faster" code does more work.
FamilyInputsRatio range Slope base → headVerdict
F3valid wheel filenames with varying project namesclaimed speedup · plan0.91× – 0.91×0.00 → 0.00faster

Receipt

TargetBackendImageRun AuditAdjudication
packaging.utils:parse_wheel_filenamelocal Dockerbw-pair:4fb0fb81fd63864abt-1256-74a808$0.11$0.000

Nemotron cost on Nebius Token Factory. Backend: where the untrusted code ran. local Docker (an image cited by its content-addressed tag). Every image is content-addressed (repo at each commit + harness), so every card re-runs from the image and seed 0. Raw data: audit.