gsmiller commented on PR #13568:
URL: https://github.com/apache/lucene/pull/13568#issuecomment-2289109976
I was already setup to run benchmarks so I did a quick run with
`wikimedium10m` that isolated this change. The results are a bit surprising.
@epotyom I'll be curious if you reproduce similar results.
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
MedTerm 551.00 (7.1%) 384.46
(5.7%) -30.2% ( -40% - -18%) 0.000
HighTerm 480.64 (7.6%) 348.90
(5.0%) -27.4% ( -37% - -16%) 0.000
LowTerm 786.47 (5.1%) 667.37
(4.6%) -15.1% ( -23% - -5%) 0.000
Prefix3 589.23 (18.0%) 555.57
(19.4%) -5.7% ( -36% - 38%) 0.334
PKLookup 124.78 (11.7%) 117.95
(7.3%) -5.5% ( -21% - 15%) 0.075
MedTermDayTaxoFacets 32.93 (8.7%) 31.23
(7.4%) -5.2% ( -19% - 12%) 0.043
IntNRQ 813.36 (3.2%) 772.66
(8.3%) -5.0% ( -15% - 6%) 0.012
BrowseDayOfYearSSDVFacets 10.35 (15.9%) 9.86
(16.5%) -4.7% ( -31% - 32%) 0.354
OrHighMedDayTaxoFacets 5.93 (6.6%) 5.65
(7.3%) -4.6% ( -17% - 9%) 0.035
MedPhrase 351.10 (8.2%) 338.39
(7.1%) -3.6% ( -17% - 12%) 0.135
OrNotHighHigh 336.05 (7.2%) 324.69
(5.8%) -3.4% ( -15% - 10%) 0.102
Fuzzy1 77.56 (6.8%) 75.18
(7.4%) -3.1% ( -16% - 11%) 0.172
OrHighHigh 169.49 (7.1%) 164.94
(6.4%) -2.7% ( -15% - 11%) 0.207
HighSloppyPhrase 41.51 (5.7%) 40.45
(5.8%) -2.5% ( -13% - 9%) 0.161
AndHighHighDayTaxoFacets 29.88 (5.9%) 29.14
(6.4%) -2.5% ( -13% - 10%) 0.204
AndHighHigh 104.87 (7.6%) 102.45
(5.9%) -2.3% ( -14% - 12%) 0.284
HighSpanNear 4.69 (3.1%) 4.61
(3.1%) -1.9% ( -7% - 4%) 0.063
AndHighMed 532.84 (4.6%) 525.69
(3.4%) -1.3% ( -8% - 7%) 0.297
OrHighLow 613.36 (5.2%) 605.70
(3.8%) -1.3% ( -9% - 8%) 0.387
Wildcard 572.65 (7.0%) 565.77
(6.5%) -1.2% ( -13% - 13%) 0.573
MedSloppyPhrase 149.60 (2.8%) 147.86
(3.1%) -1.2% ( -6% - 4%) 0.214
LowSloppyPhrase 159.98 (2.3%) 158.40
(4.2%) -1.0% ( -7% - 5%) 0.354
LowPhrase 62.66 (2.9%) 62.06
(3.9%) -1.0% ( -7% - 6%) 0.380
Fuzzy2 55.34 (6.9%) 54.87
(7.7%) -0.9% ( -14% - 14%) 0.714
HighPhrase 71.02 (3.0%) 70.47
(3.6%) -0.8% ( -7% - 6%) 0.458
AndHighMedDayTaxoFacets 54.98 (7.8%) 54.62
(6.1%) -0.7% ( -13% - 14%) 0.766
MedSpanNear 55.98 (2.7%) 55.64
(2.0%) -0.6% ( -5% - 4%) 0.424
OrNotHighLow 1382.70 (5.0%) 1376.34
(4.9%) -0.5% ( -9% - 9%) 0.769
HighTermDayOfYearSort 233.22 (9.8%) 232.28
(10.0%) -0.4% ( -18% - 21%) 0.898
AndHighLow 1429.55 (3.7%) 1427.29
(5.9%) -0.2% ( -9% - 9%) 0.920
HighTermTitleSort 107.20 (2.2%) 107.11
(2.8%) -0.1% ( -5% - 5%) 0.921
LowSpanNear 98.61 (3.0%) 98.83
(3.6%) 0.2% ( -6% - 7%) 0.832
MedIntervalsOrdered 36.23 (13.5%) 36.32
(12.5%) 0.3% ( -22% - 30%) 0.949
OrHighMed 400.68 (3.0%) 401.97
(4.6%) 0.3% ( -7% - 8%) 0.793
HighTermMonthSort 1074.59 (6.9%) 1079.54
(6.8%) 0.5% ( -12% - 15%) 0.831
OrNotHighMed 458.87 (6.6%) 461.40
(6.0%) 0.6% ( -11% - 14%) 0.784
BrowseMonthSSDVFacets 10.91 (15.7%) 10.97
(19.0%) 0.6% ( -29% - 41%) 0.919
LowIntervalsOrdered 38.84 (4.3%) 39.18
(3.8%) 0.9% ( -6% - 9%) 0.495
TermDTSort 191.60 (13.5%) 193.33
(10.6%) 0.9% ( -20% - 29%) 0.814
HighIntervalsOrdered 4.42 (6.5%) 4.46
(7.2%) 1.0% ( -11% - 15%) 0.656
HighTermTitleBDVSort 41.90 (7.7%) 42.44
(8.9%) 1.3% ( -14% - 19%) 0.626
BrowseRandomLabelSSDVFacets 6.94 (8.8%) 7.08
(8.3%) 2.0% ( -13% - 20%) 0.460
Respell 46.81 (7.8%) 47.76
(9.8%) 2.0% ( -14% - 21%) 0.469
BrowseDateSSDVFacets 1.93 (21.0%) 1.98
(22.3%) 2.9% ( -33% - 58%) 0.669
OrHighNotLow 420.46 (9.8%) 434.46
(8.8%) 3.3% ( -13% - 24%) 0.257
BrowseMonthTaxoFacets 22.93 (25.6%) 24.53
(17.5%) 7.0% ( -28% - 67%) 0.314
OrHighNotMed 361.50 (8.8%) 393.95
(7.1%) 9.0% ( -6% - 27%) 0.000
BrowseDateTaxoFacets 21.34 (28.0%) 23.32
(12.5%) 9.3% ( -24% - 69%) 0.174
BrowseRandomLabelTaxoFacets 14.44 (24.5%) 15.88
(10.4%) 9.9% ( -20% - 59%) 0.094
OrHighNotHigh 288.41 (8.4%) 327.71
(7.4%) 13.6% ( -2% - 32%) 0.000
BrowseDayOfYearTaxoFacets 21.28 (29.5%) 24.22
(14.6%) 13.8% ( -23% - 82%) 0.061
```
--
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]
