SchwarzDigits / hypermatch / 35089508624

96%

main: 97%

Pull #16

github

perf: evaluate conditions against a set of the matching leaves

A formula asked whether a leaf matched by searching the sorted list of
hits. It now tests a bit in a set of their ids, and that set is only
built once a formula needs it, so the many groups without one do not
pay for it.

Rule sets with many different anythingBut conditions on the same path
gain the most, because those conditions are evaluated for every event
that has their path: a workload whose exclusions all fail matches about
twice as fast, 603 ns instead of 1113 ns per event against 100,000
rules. Every other workload is unchanged, measured with interleaved A/B
runs against the previous commit.

The benchmark suite gets that workload, where every rule excludes the
region of the event, so the time goes into evaluating the conditions
rather than into the rules that match.

31 of 31 new or added lines in 1 file covered. (100.0%)

1931 of 2011 relevant lines covered (96.02%)

234268.92 hits per line