HUSTERGS commented on PR #14968:
URL: https://github.com/apache/lucene/pull/14968#issuecomment-3117985426
I ran the benchmark under identical setup on previous version (which add a
new api on `PostingEnum`), but removed related code under
`ConstantScoreScorer#applyAsRequiredClause`, here is the result:
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
DismaxOrHighHigh 35.29 (9.8%) 34.76
(10.2%) -1.5% ( -19% - 20%) 0.634
Phrase 7.49 (6.2%) 7.39
(5.7%) -1.3% ( -12% - 11%) 0.501
CombinedTerm 11.35 (5.3%) 11.21
(6.8%) -1.3% ( -12% - 11%) 0.517
OrHighMed 68.18 (17.4%) 67.53
(17.7%) -0.9% ( -30% - 41%) 0.864
OrHighHigh 21.91 (18.0%) 21.71
(18.0%) -0.9% ( -31% - 42%) 0.871
CountPhrase 2.64 (5.6%) 2.61
(5.2%) -0.9% ( -11% - 10%) 0.601
FilteredOrStopWords 7.93 (4.0%) 7.86
(6.3%) -0.9% ( -10% - 9%) 0.597
Or2Terms2StopWords 60.82 (13.1%) 60.30
(13.5%) -0.9% ( -24% - 29%) 0.837
CountFilteredOrMany 4.34 (5.4%) 4.32
(5.1%) -0.7% ( -10% - 10%) 0.683
AndHighOrMedMed 14.01 (5.0%) 13.92
(4.9%) -0.7% ( -10% - 9%) 0.678
CombinedOrHighMed 20.78 (7.3%) 20.65
(8.0%) -0.6% ( -14% - 15%) 0.796
AndHighHigh 22.70 (18.0%) 22.56
(18.4%) -0.6% ( -31% - 43%) 0.916
DismaxOrHighMed 47.41 (12.2%) 47.13
(11.8%) -0.6% ( -21% - 26%) 0.878
CountOrMany 4.89 (6.9%) 4.87
(5.7%) -0.5% ( -12% - 12%) 0.785
And2Terms2StopWords 58.89 (12.7%) 58.57
(13.3%) -0.5% ( -23% - 29%) 0.895
CountFilteredPhrase 8.96 (6.2%) 8.91
(5.2%) -0.5% ( -11% - 11%) 0.768
OrMany 4.49 (9.3%) 4.47
(8.5%) -0.5% ( -16% - 18%) 0.858
CountAndHighHigh 47.65 (4.9%) 47.41
(4.1%) -0.5% ( -9% - 8%) 0.725
Fuzzy1 41.70 (7.3%) 41.51
(6.6%) -0.4% ( -13% - 14%) 0.839
CombinedOrHighHigh 5.55 (5.4%) 5.52
(7.0%) -0.4% ( -12% - 12%) 0.837
CombinedAndHighHigh 5.60 (5.8%) 5.58
(5.3%) -0.4% ( -10% - 11%) 0.823
FilteredAnd3Terms 98.63 (6.1%) 98.27
(5.5%) -0.4% ( -11% - 11%) 0.843
IntSet 291.18 (6.7%) 290.32
(6.0%) -0.3% ( -12% - 13%) 0.882
CombinedAndHighMed 20.99 (7.6%) 20.93
(6.6%) -0.3% ( -13% - 15%) 0.910
CountOrHighMed 76.55 (4.4%) 76.38
(3.3%) -0.2% ( -7% - 7%) 0.860
DismaxTerm 488.06 (9.5%) 487.11
(10.9%) -0.2% ( -18% - 22%) 0.952
Or3Terms 64.30 (15.0%) 64.18
(15.6%) -0.2% ( -26% - 35%) 0.969
And3Terms 70.22 (14.3%) 70.10
(15.1%) -0.2% ( -25% - 34%) 0.971
CountFilteredOrHighMed 17.78 (2.2%) 17.75
(2.2%) -0.2% ( -4% - 4%) 0.819
FilteredOr2Terms2StopWords 49.06 (6.7%) 48.99
(5.9%) -0.1% ( -11% - 13%) 0.946
FilteredIntNRQ 41.04 (5.8%) 40.99
(5.4%) -0.1% ( -10% - 11%) 0.944
FilteredOr3Terms 42.88 (5.4%) 42.83
(5.3%) -0.1% ( -10% - 11%) 0.946
FilteredPhrase 9.63 (5.8%) 9.62
(4.8%) -0.1% ( -10% - 11%) 0.946
FilteredTerm 63.25 (5.5%) 63.21
(5.8%) -0.1% ( -10% - 11%) 0.972
FilteredOrHighMed 38.02 (5.9%) 38.00
(5.5%) -0.0% ( -10% - 12%) 0.983
Fuzzy2 37.27 (6.6%) 37.25
(6.1%) -0.0% ( -12% - 13%) 0.987
CountOrHighHigh 48.89 (6.1%) 48.89
(3.6%) -0.0% ( -9% - 10%) 0.998
CountFilteredOrHighHigh 15.65 (2.9%) 15.66
(2.2%) 0.0% ( -5% - 5%) 0.966
FilteredOrMany 3.94 (4.6%) 3.94
(5.1%) 0.1% ( -9% - 10%) 0.974
AndHighMed 53.22 (17.1%) 53.27
(17.4%) 0.1% ( -29% - 41%) 0.987
TermDTSort 132.49 (5.8%) 132.67
(6.0%) 0.1% ( -11% - 12%) 0.941
IntNRQ 41.28 (6.4%) 41.34
(5.9%) 0.1% ( -11% - 13%) 0.943
IntervalsOrdered 2.37 (5.1%) 2.37
(4.8%) 0.2% ( -9% - 10%) 0.921
CountFilteredIntNRQ 16.07 (3.0%) 16.09
(2.8%) 0.2% ( -5% - 6%) 0.860
SloppyPhrase 1.08 (6.9%) 1.08
(6.4%) 0.2% ( -12% - 14%) 0.923
Term1M 454.48 (12.1%) 455.58
(14.4%) 0.2% ( -23% - 30%) 0.954
FilteredOrHighHigh 12.66 (5.2%) 12.69
(4.6%) 0.2% ( -9% - 10%) 0.876
TermB1M 455.08 (11.9%) 456.22
(13.9%) 0.3% ( -22% - 29%) 0.951
CountAndHighMed 73.55 (5.1%) 73.75
(3.7%) 0.3% ( -8% - 9%) 0.849
TermB1M1P 453.43 (12.2%) 455.20
(14.1%) 0.4% ( -23% - 30%) 0.925
Term10K 453.54 (12.3%) 455.78
(13.9%) 0.5% ( -22% - 30%) 0.905
SpanNear 2.45 (6.7%) 2.46
(7.1%) 0.5% ( -12% - 15%) 0.816
TermDayOfYearSort 254.18 (7.0%) 255.72
(6.1%) 0.6% ( -11% - 14%) 0.772
AndMedOrHighHigh 16.35 (7.9%) 16.46
(8.3%) 0.7% ( -14% - 18%) 0.790
Term 452.67 (12.4%) 455.81
(14.0%) 0.7% ( -22% - 30%) 0.868
Respell 35.60 (5.5%) 35.85
(4.8%) 0.7% ( -9% - 11%) 0.667
OrHighRare 91.77 (8.3%) 92.47
(8.6%) 0.8% ( -14% - 19%) 0.778
OrStopWords 9.15 (15.6%) 9.22
(15.7%) 0.8% ( -26% - 38%) 0.864
FilteredAnd2Terms2StopWords 58.95 (8.8%) 59.47
(9.3%) 0.9% ( -15% - 20%) 0.759
TermTitleSort 48.51 (8.0%) 48.95
(8.3%) 0.9% ( -14% - 18%) 0.727
Term100 450.57 (13.0%) 456.09
(14.2%) 1.2% ( -22% - 32%) 0.776
FilteredPrefix3 67.50 (6.2%) 68.46
(5.2%) 1.4% ( -9% - 13%) 0.428
Wildcard 45.25 (6.3%) 45.94
(4.9%) 1.5% ( -9% - 13%) 0.390
Prefix3 72.16 (5.9%) 73.29
(5.3%) 1.6% ( -9% - 13%) 0.374
TermMonthSort 2010.81 (9.9%) 2053.91
(10.6%) 2.1% ( -16% - 25%) 0.508
CountTerm 5564.36 (13.2%) 5695.14
(13.3%) 2.4% ( -21% - 33%) 0.576
FilteredAndHighMed 30.63 (6.8%) 31.78
(6.8%) 3.8% ( -9% - 18%) 0.080
AndStopWords 8.59 (13.0%) 8.95
(13.5%) 4.2% ( -19% - 35%) 0.322
FilteredAndStopWords 8.16 (5.2%) 8.66
(3.5%) 6.1% ( -2% - 15%) 0.000
FilteredAndHighHigh 10.07 (6.0%) 10.78
(3.2%) 7.0% ( -2% - 17%) 0.000
```
It produced similar speedup, but some queries like `OrStopWords` and
`AndStopWords` do not have performance gain anymore
Maybe the speedup do not worth the complicity of new added apis ? I'm little
bit confused
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
