jpountz opened a new pull request, #13957:
URL: https://github.com/apache/lucene/pull/13957
`LeafSimScorer` is a specialization of a `SimScorer` for a given segment. It
doesn't add much value, but benchmarks suggest that it adds measurable overhead
to queries sorted by score.
Here is a `luceneutil` run with `-searchConcurrency 0` on `wikibigall`:
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
CountAndHighMed 148.80 (3.6%) 146.79
(3.3%) -1.4% ( -8% - 5%) 0.219
Prefix3 210.12 (3.4%) 208.12
(3.1%) -1.0% ( -7% - 5%) 0.355
OrNotHighLow 930.49 (2.9%) 922.26
(2.8%) -0.9% ( -6% - 4%) 0.326
CountOrHighMed 104.34 (1.6%) 103.50
(1.5%) -0.8% ( -3% - 2%) 0.099
CountAndHighHigh 48.93 (3.6%) 48.55
(3.4%) -0.8% ( -7% - 6%) 0.485
HighTermMonthSort 3011.98 (2.9%) 2989.18
(4.1%) -0.8% ( -7% - 6%) 0.498
TermDTSort 342.40 (7.1%) 340.02
(6.1%) -0.7% ( -13% - 13%) 0.741
CountOrHighHigh 49.93 (1.6%) 49.76
(1.2%) -0.3% ( -3% - 2%) 0.451
HighTermTitleSort 111.58 (2.3%) 111.22
(3.1%) -0.3% ( -5% - 5%) 0.710
OrNotHighHigh 308.36 (3.1%) 307.70
(3.3%) -0.2% ( -6% - 6%) 0.835
Fuzzy2 71.17 (1.6%) 71.07
(2.2%) -0.1% ( -3% - 3%) 0.824
OrHighLow 726.98 (1.6%) 727.36
(2.5%) 0.1% ( -4% - 4%) 0.939
HighTermDayOfYearSort 764.56 (3.8%) 765.85
(3.4%) 0.2% ( -6% - 7%) 0.882
OrNotHighMed 350.64 (3.4%) 351.46
(4.3%) 0.2% ( -7% - 8%) 0.848
Fuzzy1 75.46 (1.9%) 75.80
(1.8%) 0.5% ( -3% - 4%) 0.448
IntNRQ 139.45 (13.7%) 140.08
(14.5%) 0.5% ( -24% - 33%) 0.918
HighTermTitleBDVSort 15.35 (5.7%) 15.42
(5.5%) 0.5% ( -10% - 12%) 0.781
PKLookup 265.51 (2.5%) 267.01
(1.6%) 0.6% ( -3% - 4%) 0.389
AndHighLow 989.77 (1.9%) 995.39
(2.2%) 0.6% ( -3% - 4%) 0.387
CountTerm 7984.92 (3.9%) 8051.09
(5.0%) 0.8% ( -7% - 10%) 0.557
OrHighNotHigh 321.43 (2.7%) 324.15
(3.1%) 0.8% ( -4% - 6%) 0.357
OrMany 18.24 (2.4%) 18.45
(2.1%) 1.1% ( -3% - 5%) 0.107
Wildcard 117.97 (3.2%) 119.40
(3.2%) 1.2% ( -5% - 7%) 0.230
OrHighRare 269.54 (5.3%) 273.78
(6.5%) 1.6% ( -9% - 14%) 0.401
OrHighMed 219.25 (2.5%) 222.89
(2.7%) 1.7% ( -3% - 7%) 0.044
And2Terms2StopWords 151.65 (1.8%) 154.21
(1.6%) 1.7% ( -1% - 5%) 0.002
Or2Terms2StopWords 153.46 (3.1%) 156.15
(2.8%) 1.8% ( -4% - 7%) 0.061
Or3Terms 164.81 (2.4%) 168.57
(2.9%) 2.3% ( -2% - 7%) 0.007
MedTerm 610.37 (3.5%) 625.30
(3.7%) 2.4% ( -4% - 10%) 0.032
OrHighNotMed 417.48 (2.8%) 427.78
(3.1%) 2.5% ( -3% - 8%) 0.008
LowTerm 981.78 (2.8%) 1008.35
(3.8%) 2.7% ( -3% - 9%) 0.010
And3Terms 165.41 (1.8%) 170.05
(1.7%) 2.8% ( 0% - 6%) 0.000
AndStopWords 30.15 (3.0%) 31.07
(3.8%) 3.0% ( -3% - 10%) 0.005
HighTerm 455.84 (3.4%) 469.91
(4.0%) 3.1% ( -4% - 10%) 0.009
OrHighHigh 68.52 (1.7%) 70.69
(3.7%) 3.2% ( -2% - 8%) 0.000
OrHighNotLow 412.63 (2.8%) 427.86
(3.5%) 3.7% ( -2% - 10%) 0.000
OrStopWords 33.50 (3.8%) 34.75
(5.1%) 3.7% ( -4% - 13%) 0.009
AndHighMed 165.41 (1.9%) 171.81
(1.7%) 3.9% ( 0% - 7%) 0.000
AndHighHigh 72.22 (1.7%) 76.11
(1.4%) 5.4% ( 2% - 8%) 0.000
```
I could reproduce these small speedups with a low p-value across several
runs.
--
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]
