original-brownbear opened a new pull request, #13935:
URL: https://github.com/apache/lucene/pull/13935
We can be a little less verbose for one big conditional and can optimize a
few obvious spots in the logic (the hit comparison change is particularly
helpful here I think).
For wikimedium this is a small but clear win:
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
IntNRQ 88.51 (14.0%) 88.65
(14.7%) 0.2% ( -24% - 33%) 0.960
AndHighMed 173.56 (3.9%) 174.72
(3.8%) 0.7% ( -6% - 8%) 0.437
BrowseDateSSDVFacets 1.25 (6.1%) 1.26
(8.7%) 0.7% ( -13% - 16%) 0.672
OrHighHigh 85.29 (9.2%) 86.10
(8.1%) 0.9% ( -14% - 20%) 0.624
HighTermTitleBDVSort 27.17 (5.6%) 27.49
(6.2%) 1.2% ( -10% - 13%) 0.367
OrHighMed 277.10 (4.5%) 280.71
(4.6%) 1.3% ( -7% - 10%) 0.199
AndHighMedDayTaxoFacets 61.39 (2.4%) 62.32
(3.2%) 1.5% ( -3% - 7%) 0.016
BrowseRandomLabelSSDVFacets 3.33 (4.5%) 3.38
(2.8%) 1.5% ( -5% - 9%) 0.071
AndHighLow 1988.25 (4.8%) 2018.70
(4.4%) 1.5% ( -7% - 11%) 0.133
PKLookup 240.84 (1.8%) 245.08
(1.8%) 1.8% ( -1% - 5%) 0.000
AndHighHigh 52.61 (5.6%) 53.57
(4.8%) 1.8% ( -8% - 12%) 0.114
HighTermTitleSort 69.04 (3.8%) 70.37
(4.3%) 1.9% ( -5% - 10%) 0.034
HighTerm 479.90 (6.3%) 489.18
(5.7%) 1.9% ( -9% - 14%) 0.150
LowSloppyPhrase 86.07 (11.0%) 87.75
(11.3%) 2.0% ( -18% - 27%) 0.431
MedTermDayTaxoFacets 21.43 (3.6%) 21.85
(3.6%) 2.0% ( -5% - 9%) 0.015
Fuzzy2 85.21 (3.2%) 86.98
(3.9%) 2.1% ( -4% - 9%) 0.010
TermDTSort 177.37 (6.9%) 181.13
(6.8%) 2.1% ( -10% - 17%) 0.169
LowTerm 720.26 (5.4%) 735.66
(5.0%) 2.1% ( -7% - 13%) 0.068
MedPhrase 332.87 (5.3%) 340.01
(5.7%) 2.1% ( -8% - 13%) 0.082
HighSloppyPhrase 5.88 (8.2%) 6.02
(10.3%) 2.4% ( -14% - 22%) 0.242
LowIntervalsOrdered 99.26 (6.0%) 101.70
(5.5%) 2.5% ( -8% - 14%) 0.056
Prefix3 424.88 (4.6%) 435.41
(4.1%) 2.5% ( -5% - 11%) 0.011
AndHighHighDayTaxoFacets 18.44 (3.8%) 18.91
(3.5%) 2.5% ( -4% - 10%) 0.002
OrHighNotHigh 434.36 (5.4%) 445.63
(5.0%) 2.6% ( -7% - 13%) 0.026
HighPhrase 287.40 (5.3%) 294.89
(6.4%) 2.6% ( -8% - 15%) 0.047
MedIntervalsOrdered 39.67 (3.9%) 40.70
(3.7%) 2.6% ( -4% - 10%) 0.002
OrHighLow 688.03 (4.4%) 706.76
(4.0%) 2.7% ( -5% - 11%) 0.003
OrHighNotMed 523.00 (6.5%) 537.25
(7.3%) 2.7% ( -10% - 17%) 0.076
Fuzzy1 79.64 (2.1%) 81.81
(1.9%) 2.7% ( -1% - 6%) 0.000
BrowseMonthSSDVFacets 4.46 (7.0%) 4.59
(5.8%) 2.8% ( -9% - 16%) 0.055
LowSpanNear 145.66 (2.2%) 149.71
(2.0%) 2.8% ( -1% - 7%) 0.000
LowPhrase 181.90 (3.7%) 187.03
(4.5%) 2.8% ( -5% - 11%) 0.002
HighTermMonthSort 1467.83 (6.2%) 1510.29
(6.0%) 2.9% ( -8% - 16%) 0.034
OrNotHighMed 552.11 (6.8%) 568.79
(6.4%) 3.0% ( -9% - 17%) 0.040
Respell 35.74 (1.4%) 36.83
(2.0%) 3.0% ( 0% - 6%) 0.000
OrHighMedDayTaxoFacets 8.67 (5.7%) 8.94
(5.8%) 3.1% ( -7% - 15%) 0.015
MedSloppyPhrase 9.79 (6.6%) 10.10
(8.5%) 3.2% ( -11% - 19%) 0.062
OrNotHighLow 1115.78 (5.6%) 1151.15
(5.2%) 3.2% ( -7% - 14%) 0.009
MedSpanNear 84.96 (5.0%) 87.69
(4.7%) 3.2% ( -6% - 13%) 0.003
HighIntervalsOrdered 38.24 (5.5%) 39.48
(5.0%) 3.2% ( -6% - 14%) 0.006
BrowseRandomLabelTaxoFacets 4.34 (5.2%) 4.49
(5.1%) 3.3% ( -6% - 14%) 0.005
OrNotHighHigh 316.03 (5.8%) 326.59
(5.8%) 3.3% ( -7% - 15%) 0.010
MedTerm 628.56 (8.6%) 650.22
(7.2%) 3.4% ( -11% - 21%) 0.051
Wildcard 187.07 (4.3%) 193.66
(4.3%) 3.5% ( -4% - 12%) 0.000
HighSpanNear 18.01 (5.0%) 18.67
(4.9%) 3.6% ( -5% - 14%) 0.001
OrHighNotLow 580.06 (6.0%) 601.25
(7.4%) 3.7% ( -9% - 18%) 0.015
BrowseMonthTaxoFacets 12.43 (26.2%) 12.89
(27.5%) 3.7% ( -39% - 77%) 0.538
BrowseDayOfYearTaxoFacets 5.21 (9.6%) 5.40
(11.5%) 3.7% ( -15% - 27%) 0.115
HighTermDayOfYearSort 322.24 (8.0%) 334.40
(9.7%) 3.8% ( -12% - 23%) 0.057
BrowseDateTaxoFacets 5.11 (9.7%) 5.31
(11.6%) 4.0% ( -15% - 27%) 0.094
BrowseDayOfYearSSDVFacets 4.47 (4.1%) 4.65
(6.9%) 4.2% ( -6% - 15%) 0.001
```
--
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]
