HUSTERGS commented on PR #14772:
URL: https://github.com/apache/lucene/pull/14772#issuecomment-2973806605
> Interesting. I don't believe that being able to use `Arrays#fill` helps
much, but maybe the fact that this change helps reduce polymorphism does?
@jpountz Thanks for your reply! (apologies for late reply). Actually I'm not
very sure what's the root cause of this performance gain, My previous thought
is that this change can co-operate with other query which implemnts the
`nextDocsAndScores` and reduce the `score` function calls. Anyway, I did
another luceneutil `wikimediumall` and `taskCountPerCat=5` based on
`3064f4b2aac3526f5f5fe37a8798395511aabd46` (which is this pr checked out from)
to confirm the result again:
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
CombinedOrHighHigh 6.46 (2.9%) 6.35
(4.0%) -1.7% ( -8% - 5%) 0.124
CombinedOrHighMed 24.82 (4.6%) 24.42
(7.2%) -1.6% ( -12% - 10%) 0.405
Prefix3 94.43 (2.8%) 93.11
(3.9%) -1.4% ( -7% - 5%) 0.195
IntervalsOrdered 2.98 (2.7%) 2.94
(2.3%) -1.4% ( -6% - 3%) 0.085
TermMonthSort 2333.34 (3.4%) 2303.15
(3.4%) -1.3% ( -7% - 5%) 0.233
FilteredPrefix3 88.42 (3.0%) 87.33
(2.9%) -1.2% ( -6% - 4%) 0.184
IntSet 389.99 (5.2%) 385.58
(4.7%) -1.1% ( -10% - 9%) 0.470
TermDayOfYearSort 343.29 (3.4%) 339.52
(3.6%) -1.1% ( -7% - 6%) 0.321
TermDTSort 189.75 (4.9%) 187.77
(4.8%) -1.0% ( -10% - 9%) 0.494
CountTerm 6236.56 (5.6%) 6180.21
(4.8%) -0.9% ( -10% - 10%) 0.584
AndHighOrMedMed 17.22 (3.6%) 17.07
(3.3%) -0.9% ( -7% - 6%) 0.421
IntNRQ 48.24 (2.4%) 47.83
(2.8%) -0.8% ( -5% - 4%) 0.307
FilteredIntNRQ 47.76 (2.5%) 47.48
(2.7%) -0.6% ( -5% - 4%) 0.467
Wildcard 56.24 (1.8%) 55.94
(2.9%) -0.5% ( -5% - 4%) 0.480
CombinedTerm 13.58 (4.2%) 13.51
(4.2%) -0.5% ( -8% - 8%) 0.721
OrMany 5.44 (4.9%) 5.42
(4.7%) -0.3% ( -9% - 9%) 0.826
CombinedAndHighHigh 6.40 (3.0%) 6.39
(3.6%) -0.2% ( -6% - 6%) 0.834
Respell 43.41 (2.6%) 43.32
(3.0%) -0.2% ( -5% - 5%) 0.817
FilteredTerm 84.14 (2.3%) 83.97
(2.7%) -0.2% ( -5% - 4%) 0.802
CountFilteredIntNRQ 21.90 (1.7%) 21.86
(2.2%) -0.2% ( -4% - 3%) 0.781
CountPhrase 3.22 (2.9%) 3.22
(3.8%) -0.1% ( -6% - 6%) 0.909
TermTitleSort 63.89 (6.2%) 63.85
(6.8%) -0.1% ( -12% - 13%) 0.977
OrHighHigh 25.06 (2.4%) 25.05
(2.3%) -0.0% ( -4% - 4%) 0.964
OrStopWords 9.83 (6.1%) 9.84
(5.6%) 0.1% ( -11% - 12%) 0.969
CountOrHighHigh 62.03 (2.2%) 62.08
(2.3%) 0.1% ( -4% - 4%) 0.907
CountOrHighMed 93.40 (2.4%) 93.54
(2.5%) 0.2% ( -4% - 5%) 0.845
DismaxOrHighMed 63.11 (2.0%) 63.21
(2.7%) 0.2% ( -4% - 4%) 0.820
Fuzzy1 49.35 (3.2%) 49.43
(3.9%) 0.2% ( -6% - 7%) 0.881
CountFilteredOrMany 5.94 (2.3%) 5.95
(2.4%) 0.2% ( -4% - 4%) 0.799
SpanNear 3.05 (4.8%) 3.06
(4.7%) 0.2% ( -8% - 10%) 0.896
DismaxOrHighHigh 44.54 (2.8%) 44.64
(2.5%) 0.2% ( -4% - 5%) 0.794
Term1M 564.76 (3.8%) 566.06
(3.2%) 0.2% ( -6% - 7%) 0.835
OrHighMed 86.90 (2.3%) 87.10
(2.2%) 0.2% ( -4% - 4%) 0.743
CountFilteredOrHighHigh 24.82 (1.4%) 24.88
(1.2%) 0.2% ( -2% - 2%) 0.549
Term 564.57 (3.8%) 566.04
(3.2%) 0.3% ( -6% - 7%) 0.816
AndHighMed 66.11 (1.5%) 66.28
(1.6%) 0.3% ( -2% - 3%) 0.592
CountOrMany 5.96 (3.5%) 5.97
(4.1%) 0.3% ( -7% - 8%) 0.825
CountAndHighHigh 60.53 (2.2%) 60.70
(2.1%) 0.3% ( -3% - 4%) 0.685
CountFilteredOrHighMed 29.11 (1.3%) 29.19
(1.4%) 0.3% ( -2% - 3%) 0.484
FilteredOrHighHigh 17.45 (2.1%) 17.51
(1.7%) 0.3% ( -3% - 4%) 0.614
DismaxTerm 602.63 (2.9%) 604.59
(2.2%) 0.3% ( -4% - 5%) 0.689
FilteredOrMany 4.96 (2.3%) 4.97
(2.5%) 0.3% ( -4% - 5%) 0.655
Or3Terms 75.91 (3.6%) 76.17
(3.2%) 0.3% ( -6% - 7%) 0.748
Phrase 9.48 (3.5%) 9.51
(2.3%) 0.4% ( -5% - 6%) 0.707
CountAndHighMed 91.03 (2.7%) 91.37
(2.6%) 0.4% ( -4% - 5%) 0.661
AndHighHigh 26.15 (3.2%) 26.24
(3.3%) 0.4% ( -5% - 7%) 0.721
Term100 564.35 (4.0%) 566.60
(3.1%) 0.4% ( -6% - 7%) 0.725
FilteredOr3Terms 57.08 (2.1%) 57.32
(2.4%) 0.4% ( -4% - 5%) 0.548
Term10K 564.11 (3.7%) 566.57
(3.2%) 0.4% ( -6% - 7%) 0.693
AndStopWords 9.03 (5.3%) 9.07
(4.6%) 0.4% ( -9% - 10%) 0.775
Fuzzy2 44.51 (2.9%) 44.71
(3.7%) 0.5% ( -5% - 7%) 0.665
CountFilteredPhrase 11.37 (2.9%) 11.42
(3.3%) 0.5% ( -5% - 6%) 0.631
CombinedAndHighMed 24.50 (4.8%) 24.62
(7.5%) 0.5% ( -11% - 13%) 0.813
FilteredPhrase 12.18 (2.0%) 12.24
(2.0%) 0.5% ( -3% - 4%) 0.449
TermB1M1P 564.34 (3.9%) 567.28
(3.2%) 0.5% ( -6% - 7%) 0.643
SloppyPhrase 1.50 (3.9%) 1.51
(3.7%) 0.5% ( -6% - 8%) 0.667
AndMedOrHighHigh 20.39 (1.4%) 20.50
(2.1%) 0.6% ( -2% - 4%) 0.320
Or2Terms2StopWords 72.02 (5.1%) 72.42
(4.9%) 0.6% ( -8% - 11%) 0.721
FilteredOr2Terms2StopWords 63.56 (2.9%) 63.95
(3.2%) 0.6% ( -5% - 6%) 0.525
TermB1M 563.43 (4.0%) 566.92
(3.3%) 0.6% ( -6% - 8%) 0.594
FilteredOrStopWords 10.68 (2.2%) 10.76
(1.4%) 0.7% ( -2% - 4%) 0.238
FilteredOrHighMed 51.48 (2.9%) 51.87
(3.2%) 0.8% ( -5% - 7%) 0.431
FilteredAnd2Terms2StopWords 71.38 (4.7%) 71.95
(4.5%) 0.8% ( -8% - 10%) 0.585
And3Terms 83.40 (3.9%) 84.07
(3.4%) 0.8% ( -6% - 8%) 0.487
And2Terms2StopWords 69.70 (6.0%) 70.27
(5.5%) 0.8% ( -9% - 13%) 0.651
FilteredAnd3Terms 126.57 (3.0%) 127.78
(3.2%) 1.0% ( -5% - 7%) 0.324
OrHighRare 115.81 (6.0%) 117.21
(4.4%) 1.2% ( -8% - 12%) 0.465
FilteredAndHighMed 41.51 (2.1%) 42.89
(2.3%) 3.3% ( -1% - 7%) 0.000
FilteredAndStopWords 11.02 (2.2%) 11.50
(2.3%) 4.4% ( 0% - 9%) 0.000
FilteredAndHighHigh 13.88 (2.1%) 14.62
(2.0%) 5.3% ( 1% - 9%) 0.000
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
