gsmiller commented on pull request #597:
URL: https://github.com/apache/lucene/pull/597#issuecomment-1011241519
Benchmark results with all three other faceting changes (omitting this one)
reverted (using `wikimedium10m`). None of the results are significant but there
is a trend with "browse" taxo-faceting tasks hinting at a performance
improvement with all of these changes reverted. I'll start testing individual
reverts to see if there's a pattern.
```
TaskQPS baseline StdDevQPS candidate
StdDev Pct diff p-value
BrowseMonthSSDVFacets 14.20 (20.0%) 13.69
(17.1%) -3.6% ( -33% - 41%) 0.544
HighSpanNear 8.89 (5.7%) 8.76
(4.9%) -1.5% ( -11% - 9%) 0.358
LowSloppyPhrase 4.58 (6.0%) 4.53
(7.4%) -1.3% ( -13% - 12%) 0.547
BrowseDayOfYearSSDVFacets 12.80 (15.5%) 12.68
(14.5%) -0.9% ( -26% - 34%) 0.843
AndHighHighDayTaxoFacets 6.24 (3.1%) 6.19
(3.2%) -0.9% ( -6% - 5%) 0.384
AndHighMedDayTaxoFacets 33.63 (2.1%) 33.35
(2.5%) -0.8% ( -5% - 3%) 0.249
MedTermDayTaxoFacets 24.37 (4.2%) 24.17
(4.4%) -0.8% ( -9% - 8%) 0.550
OrHighMedDayTaxoFacets 4.85 (4.6%) 4.82
(5.8%) -0.7% ( -10% - 10%) 0.651
MedSpanNear 7.52 (2.2%) 7.49
(2.4%) -0.4% ( -4% - 4%) 0.603
IntNRQ 435.95 (1.9%) 434.54
(1.6%) -0.3% ( -3% - 3%) 0.559
HighPhrase 80.03 (2.8%) 79.82
(2.8%) -0.3% ( -5% - 5%) 0.766
OrHighMed 58.92 (4.2%) 58.77
(4.1%) -0.3% ( -8% - 8%) 0.847
Fuzzy2 94.44 (1.7%) 94.20
(1.3%) -0.3% ( -3% - 2%) 0.592
Fuzzy1 90.72 (1.6%) 90.53
(1.2%) -0.2% ( -2% - 2%) 0.633
MedPhrase 769.64 (2.6%) 768.03
(2.7%) -0.2% ( -5% - 5%) 0.804
LowPhrase 53.31 (2.8%) 53.26
(2.8%) -0.1% ( -5% - 5%) 0.910
AndHighLow 673.62 (4.4%) 673.14
(2.9%) -0.1% ( -7% - 7%) 0.952
OrHighLow 602.85 (3.7%) 602.49
(2.7%) -0.1% ( -6% - 6%) 0.954
OrHighHigh 40.93 (3.5%) 40.93
(3.5%) -0.0% ( -6% - 7%) 0.997
LowSpanNear 83.92 (2.4%) 83.94
(1.5%) 0.0% ( -3% - 4%) 0.967
Respell 62.10 (1.5%) 62.12
(1.2%) 0.0% ( -2% - 2%) 0.950
Wildcard 94.65 (4.1%) 94.73
(4.0%) 0.1% ( -7% - 8%) 0.946
MedSloppyPhrase 123.05 (3.1%) 123.23
(3.4%) 0.1% ( -6% - 6%) 0.889
AndHighMed 170.12 (4.0%) 170.42
(2.9%) 0.2% ( -6% - 7%) 0.872
Prefix3 123.75 (14.1%) 124.06
(14.8%) 0.2% ( -25% - 33%) 0.957
OrHighNotMed 1193.23 (3.8%) 1196.67
(3.1%) 0.3% ( -6% - 7%) 0.791
HighSloppyPhrase 17.02 (2.8%) 17.08
(3.1%) 0.3% ( -5% - 6%) 0.713
OrNotHighMed 1090.09 (3.2%) 1094.62
(2.8%) 0.4% ( -5% - 6%) 0.661
PKLookup 170.36 (2.9%) 171.22
(3.0%) 0.5% ( -5% - 6%) 0.589
AndHighHigh 62.54 (3.2%) 62.89
(3.1%) 0.6% ( -5% - 7%) 0.580
LowTerm 2069.63 (3.0%) 2082.84
(3.3%) 0.6% ( -5% - 7%) 0.525
OrNotHighLow 887.97 (3.1%) 894.86
(2.6%) 0.8% ( -4% - 6%) 0.386
HighIntervalsOrdered 17.71 (6.3%) 17.88
(6.0%) 1.0% ( -10% - 14%) 0.624
MedIntervalsOrdered 40.94 (5.0%) 41.35
(5.0%) 1.0% ( -8% - 11%) 0.528
LowIntervalsOrdered 15.18 (3.2%) 15.33
(3.4%) 1.0% ( -5% - 7%) 0.337
OrHighNotLow 1254.36 (4.0%) 1268.04
(3.6%) 1.1% ( -6% - 9%) 0.370
OrNotHighHigh 856.87 (3.8%) 869.56
(3.5%) 1.5% ( -5% - 9%) 0.199
OrHighNotHigh 886.15 (4.8%) 899.30
(4.2%) 1.5% ( -7% - 10%) 0.296
BrowseRandomLabelSSDVFacets 9.42 (2.3%) 9.57
(3.0%) 1.7% ( -3% - 7%) 0.048
BrowseRandomLabelTaxoFacets 23.50 (31.2%) 23.95
(28.4%) 1.9% ( -43% - 89%) 0.840
MedTerm 1700.47 (4.9%) 1735.12
(4.9%) 2.0% ( -7% - 12%) 0.189
HighTerm 1620.80 (4.4%) 1654.63
(4.2%) 2.1% ( -6% - 11%) 0.127
BrowseDateTaxoFacets 30.09 (31.4%) 30.88
(29.0%) 2.6% ( -43% - 91%) 0.785
BrowseDayOfYearTaxoFacets 30.48 (31.7%) 31.32
(29.2%) 2.7% ( -44% - 93%) 0.776
BrowseMonthTaxoFacets 30.20 (21.2%) 31.22
(17.0%) 3.4% ( -28% - 52%) 0.578
TermDTSort 136.80 (15.4%) 141.79
(21.5%) 3.6% ( -28% - 47%) 0.538
HighTermDayOfYearSort 92.40 (7.2%) 97.28
(14.6%) 5.3% ( -15% - 29%) 0.148
HighTermTitleBDVSort 47.88 (13.0%) 50.64
(19.1%) 5.8% ( -23% - 43%) 0.264
HighTermMonthSort 132.96 (14.1%) 145.01
(17.0%) 9.1% ( -19% - 46%) 0.067
```
--
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]
