# Runtime baseline for `make check-quality` on laptop - see Makefile.
#
# MS_PER_FIXTURE: benchmark_optimal_solutions' own "Runtime: ... ms/fixture" line: the diffs
# alone, timed one at a time after the parallel grading pass.
# Informational only: a >2x jump warns, it never fails a deploy. Compared only on the
# machine that recorded it, since wall-clock time varies by machine far more than any
# real regression would. Per-fixture times are in laptop.csv beside this file.
#
# The accuracy gate does NOT live here. It is per-fixture, in ../quality_baseline.csv,
# because no single number over a corpus that keeps gaining hard fixtures can tell a
# real regression apart from new data - measured, see the quality-gate section in
# src/bin/benchmark_optimal_solutions.rs.
MS_PER_FIXTURE=15.3
