unicode-org / icu4x / 16447821930

73%

Build Type

push

github

Committed by web-flow

Commit Message

Make ZeroVec.iter().collect() faster (#6764)

Original issue is https://bugzilla.mozilla.org/show_bug.cgi?id=1974993

Segmenter's LSTM data model uses ZeroVec.iter().collect() to extract a
model data. But this is slower after
https://github.com/unicode-org/icu4x/pull/5924.

`collect()` still uses `Iterator.size_hint()` even if
`ExactSizeIterator`. So we should implement it.

After applying this fix, this makes LSTM data model faster, like the
following:

```
Line Break/UTF8/Th/auto time:   [288.58 µs 288.88 µs 289.19 µs]
                        change: [-14.597% -12.338% -10.010%] (p = 0.00 < 0.05)
                        Performance has improved.
Line Break/UTF8/Th/lstm time:   [298.69 µs 300.79 µs 302.90 µs]
                        change: [-10.057% -8.6718% -6.7929%] (p = 0.00 < 0.05)
                        Performance has improved.
``

Run Details

3 of 3 new or added lines in 1 file covered. (100.0%)

2 existing lines in 1 file now uncovered.

59001 of 81282 relevant lines covered (72.59%)

460127.57 hits per line