jfreden commented on PR #13036:
URL: https://github.com/apache/lucene/pull/13036#issuecomment-1916440041
I added code to only apply the optimization `if
count(term-with-less-docs)/count(term-with-more-docs) < 0.1` and it yielded a
way better result. Will investigate the term cache idea too since there is
still a slowdown of `CountHigHigh`.
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
CountOrHighHigh 43.48 (3.1%) 42.94
(3.1%) -1.2% ( -7% - 5%) 0.201
CountAndHighMed 117.33 (1.6%) 116.68
(2.2%) -0.6% ( -4% - 3%) 0.357
CountAndHighHigh 13.40 (1.4%) 13.34
(1.6%) -0.5% ( -3% - 2%) 0.305
CountOrHighMed 44.02 (2.4%) 85.10
(4.3%) 93.3% ( 84% - 102%) 0.000
```
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
HighIntervalsOrdered 2.45 (15.8%) 2.37
(16.8%) -3.3% ( -30% - 34%) 0.526
BrowseDateSSDVFacets 1.77 (9.7%) 1.71
(13.2%) -3.1% ( -23% - 21%) 0.394
IntNRQ 26.64 (9.5%) 25.91
(13.2%) -2.8% ( -23% - 22%) 0.447
HighTermMonthSort 4247.44 (5.4%) 4174.43
(6.8%) -1.7% ( -13% - 11%) 0.379
MedIntervalsOrdered 4.40 (9.0%) 4.34
(9.3%) -1.5% ( -18% - 18%) 0.597
OrNotHighLow 1013.74 (2.8%) 1000.22
(4.0%) -1.3% ( -7% - 5%) 0.223
CountOrHighHigh 43.48 (3.1%) 42.94
(3.1%) -1.2% ( -7% - 5%) 0.201
OrHighHigh 45.23 (7.6%) 44.71
(7.1%) -1.1% ( -14% - 14%) 0.624
LowIntervalsOrdered 75.80 (6.3%) 74.94
(7.0%) -1.1% ( -13% - 12%) 0.591
OrHighLow 490.67 (4.2%) 485.46
(3.7%) -1.1% ( -8% - 7%) 0.394
LowTerm 782.27 (3.7%) 774.48
(3.3%) -1.0% ( -7% - 6%) 0.366
OrNotHighMed 327.78 (2.4%) 325.10
(2.6%) -0.8% ( -5% - 4%) 0.295
OrHighMedDayTaxoFacets 7.49 (4.0%) 7.44
(4.3%) -0.7% ( -8% - 7%) 0.574
AndHighLow 1052.61 (4.4%) 1044.91
(4.7%) -0.7% ( -9% - 8%) 0.610
AndHighMedDayTaxoFacets 31.62 (1.4%) 31.39
(2.3%) -0.7% ( -4% - 3%) 0.245
MedSloppyPhrase 4.55 (2.2%) 4.52
(3.4%) -0.6% ( -6% - 5%) 0.494
HighSloppyPhrase 7.84 (2.6%) 7.80
(3.2%) -0.6% ( -6% - 5%) 0.535
CountAndHighMed 117.33 (1.6%) 116.68
(2.2%) -0.6% ( -4% - 3%) 0.357
HighTermDayOfYearSort 365.06 (3.2%) 363.03
(3.3%) -0.6% ( -6% - 6%) 0.590
Wildcard 35.27 (2.2%) 35.08
(2.6%) -0.6% ( -5% - 4%) 0.470
AndHighMed 149.03 (5.1%) 148.24
(4.7%) -0.5% ( -9% - 9%) 0.736
MedSpanNear 4.61 (2.7%) 4.58
(3.3%) -0.5% ( -6% - 5%) 0.588
MedPhrase 28.98 (4.8%) 28.83
(5.1%) -0.5% ( -9% - 9%) 0.743
CountAndHighHigh 13.40 (1.4%) 13.34
(1.6%) -0.5% ( -3% - 2%) 0.305
HighPhrase 8.65 (4.5%) 8.61
(5.4%) -0.5% ( -9% - 9%) 0.765
LowSloppyPhrase 29.81 (2.4%) 29.70
(2.8%) -0.4% ( -5% - 4%) 0.639
OrNotHighHigh 294.82 (3.6%) 293.71
(2.9%) -0.4% ( -6% - 6%) 0.719
MedTerm 719.95 (6.3%) 717.27
(5.2%) -0.4% ( -11% - 11%) 0.839
Respell 70.19 (2.4%) 69.94
(2.1%) -0.4% ( -4% - 4%) 0.607
PKLookup 288.07 (2.7%) 287.18
(3.4%) -0.3% ( -6% - 5%) 0.748
Prefix3 411.30 (1.9%) 410.09
(2.5%) -0.3% ( -4% - 4%) 0.675
OrHighMed 226.84 (5.1%) 226.17
(4.0%) -0.3% ( -8% - 9%) 0.842
LowPhrase 48.24 (4.1%) 48.15
(4.3%) -0.2% ( -8% - 8%) 0.889
OrHighNotHigh 464.84 (3.1%) 464.02
(3.5%) -0.2% ( -6% - 6%) 0.866
Fuzzy1 108.46 (1.9%) 108.40
(1.9%) -0.1% ( -3% - 3%) 0.930
TermDTSort 290.54 (5.1%) 290.58
(5.5%) 0.0% ( -10% - 11%) 0.993
HighSpanNear 19.59 (2.1%) 19.60
(2.2%) 0.1% ( -4% - 4%) 0.916
Fuzzy2 107.38 (1.8%) 107.48
(2.3%) 0.1% ( -3% - 4%) 0.889
OrHighNotMed 452.75 (4.4%) 453.28
(4.4%) 0.1% ( -8% - 9%) 0.933
OrHighNotLow 384.90 (4.4%) 385.42
(5.1%) 0.1% ( -8% - 10%) 0.928
AndHighHighDayTaxoFacets 18.84 (1.8%) 18.87
(2.1%) 0.1% ( -3% - 4%) 0.816
LowSpanNear 57.00 (3.1%) 57.11
(2.9%) 0.2% ( -5% - 6%) 0.841
HighTermTitleSort 244.47 (2.3%) 244.98
(3.3%) 0.2% ( -5% - 5%) 0.821
HighTerm 715.29 (6.2%) 717.58
(5.0%) 0.3% ( -10% - 12%) 0.857
AndHighHigh 61.54 (4.0%) 61.84
(3.6%) 0.5% ( -6% - 8%) 0.687
MedTermDayTaxoFacets 38.45 (1.9%) 38.78
(2.3%) 0.9% ( -3% - 5%) 0.203
HighTermTitleBDVSort 10.48 (4.6%) 10.63
(7.4%) 1.4% ( -10% - 14%) 0.466
BrowseRandomLabelTaxoFacets 7.36 (3.7%) 7.48
(16.2%) 1.6% ( -17% - 22%) 0.675
BrowseMonthSSDVFacets 6.64 (2.7%) 6.78
(9.2%) 2.1% ( -9% - 14%) 0.330
BrowseDayOfYearTaxoFacets 7.77 (4.2%) 7.93
(14.0%) 2.2% ( -15% - 21%) 0.510
BrowseDateTaxoFacets 7.68 (3.8%) 7.87
(15.0%) 2.5% ( -15% - 22%) 0.468
BrowseRandomLabelSSDVFacets 6.05 (9.2%) 6.22
(11.2%) 2.8% ( -16% - 25%) 0.380
BrowseDayOfYearSSDVFacets 6.69 (2.5%) 6.92
(8.8%) 3.3% ( -7% - 15%) 0.103
BrowseMonthTaxoFacets 11.69 (24.5%) 12.24
(23.0%) 4.7% ( -34% - 69%) 0.534
CountOrHighMed 44.02 (2.4%) 85.10
(4.3%) 93.3% ( 84% - 102%) 0.000
```
--
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]
