HUSTERGS opened a new pull request, #14991:
URL: https://github.com/apache/lucene/pull/14991
### Description
This is a tiny modification about how to loop through the `IntRef` during
the `PointRangeQuery` visit process, which is a hot path.
I ran the luceneutil on `wikimediumall` with `searchConcurrency=0,
taskCountPerCat=5, taskRepeatCount=50`, here is the result after 20 iterations:
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
AndHighOrMedMed 14.22 (3.4%) 14.05
(2.8%) -1.2% ( -7% - 5%) 0.242
Prefix3 74.69 (4.2%) 73.87
(3.7%) -1.1% ( -8% - 7%) 0.383
CountTerm 6021.67 (4.7%) 5959.35
(5.5%) -1.0% ( -10% - 9%) 0.522
FilteredAndStopWords 8.22 (2.5%) 8.14
(2.8%) -1.0% ( -6% - 4%) 0.227
IntSet 294.24 (5.3%) 291.38
(4.7%) -1.0% ( -10% - 9%) 0.541
FilteredPrefix3 69.88 (4.1%) 69.24
(3.4%) -0.9% ( -8% - 6%) 0.441
FilteredAndHighHigh 10.17 (2.6%) 10.08
(2.7%) -0.9% ( -6% - 4%) 0.283
SpanNear 2.50 (3.0%) 2.47
(4.6%) -0.9% ( -8% - 6%) 0.481
CountPhrase 2.66 (2.9%) 2.63
(3.4%) -0.8% ( -6% - 5%) 0.410
FilteredPhrase 9.75 (2.8%) 9.68
(2.8%) -0.8% ( -6% - 5%) 0.384
AndStopWords 9.14 (2.1%) 9.07
(2.2%) -0.7% ( -4% - 3%) 0.293
AndMedOrHighHigh 16.93 (2.5%) 16.81
(2.3%) -0.7% ( -5% - 4%) 0.371
DismaxOrHighMed 50.64 (4.2%) 50.31
(4.0%) -0.7% ( -8% - 7%) 0.611
FilteredAnd3Terms 100.95 (3.9%) 100.28
(3.4%) -0.7% ( -7% - 6%) 0.568
Respell 36.47 (3.6%) 36.25
(5.0%) -0.6% ( -8% - 8%) 0.653
FilteredOrMany 3.97 (2.9%) 3.95
(2.9%) -0.6% ( -6% - 5%) 0.520
IntervalsOrdered 2.43 (2.0%) 2.42
(3.5%) -0.6% ( -5% - 5%) 0.513
FilteredAndHighMed 31.57 (2.6%) 31.39
(2.5%) -0.6% ( -5% - 4%) 0.467
SloppyPhrase 1.11 (5.1%) 1.10
(4.1%) -0.6% ( -9% - 9%) 0.706
And3Terms 74.71 (4.4%) 74.32
(4.3%) -0.5% ( -8% - 8%) 0.703
FilteredOrStopWords 8.05 (2.3%) 8.01
(2.4%) -0.5% ( -5% - 4%) 0.489
OrMany 4.66 (4.4%) 4.64
(4.8%) -0.5% ( -9% - 9%) 0.727
Fuzzy2 38.17 (4.0%) 37.97
(5.3%) -0.5% ( -9% - 9%) 0.734
CountAndHighMed 74.45 (3.0%) 74.11
(2.9%) -0.5% ( -6% - 5%) 0.615
DismaxOrHighHigh 37.52 (3.8%) 37.36
(3.2%) -0.5% ( -7% - 6%) 0.683
TermMonthSort 2148.50 (2.6%) 2139.58
(3.2%) -0.4% ( -6% - 5%) 0.655
FilteredOr3Terms 43.11 (3.9%) 42.93
(3.8%) -0.4% ( -7% - 7%) 0.737
And2Terms2StopWords 61.27 (7.3%) 61.03
(7.3%) -0.4% ( -14% - 15%) 0.865
FilteredOr2Terms2StopWords 49.27 (5.0%) 49.08
(5.1%) -0.4% ( -10% - 10%) 0.810
CountOrHighMed 77.15 (2.9%) 76.88
(2.6%) -0.4% ( -5% - 5%) 0.680
Or2Terms2StopWords 63.59 (6.8%) 63.36
(6.6%) -0.4% ( -12% - 14%) 0.866
FilteredAnd2Terms2StopWords 60.51 (5.3%) 60.31
(5.3%) -0.3% ( -10% - 10%) 0.847
OrStopWords 9.88 (2.4%) 9.85
(2.4%) -0.3% ( -4% - 4%) 0.673
CombinedOrHighHigh 5.60 (3.4%) 5.58
(5.3%) -0.3% ( -8% - 8%) 0.823
AndHighMed 57.66 (3.9%) 57.50
(3.9%) -0.3% ( -7% - 7%) 0.817
FilteredOrHighHigh 12.86 (2.9%) 12.82
(3.1%) -0.3% ( -6% - 5%) 0.763
Phrase 7.51 (3.5%) 7.49
(2.6%) -0.3% ( -6% - 5%) 0.772
CombinedOrHighMed 20.83 (5.2%) 20.78
(6.3%) -0.3% ( -11% - 11%) 0.883
CountFilteredOrMany 4.40 (3.1%) 4.39
(2.9%) -0.2% ( -6% - 5%) 0.807
CountOrHighHigh 49.43 (2.8%) 49.32
(2.4%) -0.2% ( -5% - 5%) 0.785
CountOrMany 4.96 (3.3%) 4.95
(2.9%) -0.2% ( -6% - 6%) 0.827
Or3Terms 68.39 (4.4%) 68.24
(4.5%) -0.2% ( -8% - 9%) 0.879
FilteredOrHighMed 38.42 (4.0%) 38.33
(4.0%) -0.2% ( -7% - 8%) 0.866
Fuzzy1 42.60 (4.3%) 42.51
(5.6%) -0.2% ( -9% - 10%) 0.897
CountAndHighHigh 48.02 (2.4%) 47.93
(2.4%) -0.2% ( -4% - 4%) 0.805
OrHighMed 73.36 (4.9%) 73.28
(4.9%) -0.1% ( -9% - 10%) 0.943
CountFilteredOrHighHigh 15.82 (1.1%) 15.80
(0.9%) -0.1% ( -2% - 1%) 0.745
CountFilteredPhrase 9.01 (3.1%) 9.01
(3.5%) -0.1% ( -6% - 6%) 0.941
AndHighHigh 24.76 (2.8%) 24.74
(2.4%) -0.1% ( -5% - 5%) 0.945
OrHighHigh 23.88 (2.8%) 23.87
(2.8%) -0.1% ( -5% - 5%) 0.953
CountFilteredOrHighMed 17.90 (0.9%) 17.90
(0.8%) -0.0% ( -1% - 1%) 0.897
FilteredTerm 64.07 (4.3%) 64.08
(4.1%) 0.0% ( -8% - 8%) 0.989
CombinedAndHighHigh 5.67 (2.4%) 5.68
(2.7%) 0.1% ( -4% - 5%) 0.856
TermDayOfYearSort 260.32 (2.6%) 260.72
(2.9%) 0.2% ( -5% - 5%) 0.858
DismaxTerm 516.56 (4.5%) 517.71
(4.2%) 0.2% ( -8% - 9%) 0.871
TermB1M1P 486.46 (5.3%) 487.97
(4.1%) 0.3% ( -8% - 10%) 0.836
CombinedAndHighMed 21.11 (5.0%) 21.18
(5.5%) 0.3% ( -9% - 11%) 0.846
Term1M 486.31 (5.3%) 488.01
(4.0%) 0.3% ( -8% - 10%) 0.813
Term 486.88 (5.2%) 488.60
(3.9%) 0.4% ( -8% - 10%) 0.810
Term10K 486.28 (5.4%) 488.06
(4.1%) 0.4% ( -8% - 10%) 0.809
Wildcard 46.68 (3.5%) 46.86
(3.9%) 0.4% ( -6% - 8%) 0.749
TermDTSort 135.05 (2.4%) 135.58
(2.8%) 0.4% ( -4% - 5%) 0.638
Term100 486.10 (5.3%) 488.34
(4.0%) 0.5% ( -8% - 10%) 0.755
TermTitleSort 49.03 (4.3%) 49.31
(6.3%) 0.6% ( -9% - 11%) 0.740
TermB1M 485.85 (5.3%) 488.61
(4.0%) 0.6% ( -8% - 10%) 0.703
CountFilteredIntNRQ 16.19 (1.2%) 16.31
(1.3%) 0.8% ( -1% - 3%) 0.062
CombinedTerm 11.39 (3.6%) 11.48
(4.2%) 0.8% ( -6% - 8%) 0.523
OrHighRare 94.52 (6.4%) 96.19
(3.8%) 1.8% ( -7% - 12%) 0.287
FilteredIntNRQ 41.34 (2.9%) 42.21
(3.1%) 2.1% ( -3% - 8%) 0.027
IntNRQ 41.60 (2.9%) 42.60
(3.0%) 2.4% ( -3% - 8%) 0.009
```
It's not a big speedup, nor is the p-value zero, but it should do no harm
and is a easy win I guess?
<!--
If this is your first contribution to Lucene, please make sure you have
reviewed the contribution guide.
https://github.com/apache/lucene/blob/main/CONTRIBUTING.md
-->
--
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]
