jpountz commented on PR #12444:
URL: https://github.com/apache/lucene/pull/12444#issuecomment-1637792504
Here is the usual set of queries, still on wikimedium10m. Sparser
disjunctive queries like `Fuzzy1`, `Fuzzy2` and `OrHighLow` can get a slowdown
when the majority of clauses have very few matches per window of 2048 doc IDs,
so the bitset adds more overhead than it removes.
```
TaskQPS baseline StdDevQPS
my_modified_version StdDev Pct diff p-value
Fuzzy1 114.52 (2.1%) 101.50
(1.3%) -11.4% ( -14% - -8%) 0.000
BrowseDayOfYearTaxoFacets 44.09 (4.4%) 41.73
(16.0%) -5.3% ( -24% - 15%) 0.150
BrowseDateTaxoFacets 43.48 (3.1%) 41.17
(16.0%) -5.3% ( -23% - 14%) 0.145
BrowseRandomLabelTaxoFacets 36.20 (5.4%) 34.54
(15.0%) -4.6% ( -23% - 16%) 0.199
OrHighLow 504.34 (3.3%) 483.98
(5.4%) -4.0% ( -12% - 4%) 0.004
Fuzzy2 73.82 (2.3%) 71.64
(1.0%) -3.0% ( -6% - 0%) 0.000
OrNotHighLow 1132.98 (4.6%) 1100.07
(4.5%) -2.9% ( -11% - 6%) 0.044
AndHighLow 1468.55 (2.9%) 1435.02
(3.3%) -2.3% ( -8% - 4%) 0.020
BrowseMonthTaxoFacets 28.59 (7.9%) 27.99
(12.2%) -2.1% ( -20% - 19%) 0.514
BrowseDateSSDVFacets 5.66 (8.2%) 5.56
(7.3%) -1.7% ( -15% - 15%) 0.501
HighSloppyPhrase 22.58 (4.2%) 22.21
(4.5%) -1.6% ( -9% - 7%) 0.241
HighSpanNear 42.83 (5.2%) 42.14
(7.2%) -1.6% ( -13% - 11%) 0.416
AndHighHigh 47.72 (6.7%) 47.10
(4.9%) -1.3% ( -12% - 11%) 0.485
AndHighMed 337.19 (4.8%) 333.34
(3.7%) -1.1% ( -9% - 7%) 0.399
HighTermTitleSort 103.35 (6.8%) 102.27
(6.3%) -1.0% ( -13% - 12%) 0.615
LowSloppyPhrase 32.10 (2.7%) 31.78
(2.3%) -1.0% ( -5% - 4%) 0.203
HighTermDayOfYearSort 381.91 (2.9%) 378.63
(2.8%) -0.9% ( -6% - 5%) 0.346
OrNotHighMed 552.91 (4.2%) 549.11
(4.5%) -0.7% ( -9% - 8%) 0.617
TermDTSort 181.43 (2.3%) 180.22
(2.3%) -0.7% ( -5% - 3%) 0.353
BrowseRandomLabelSSDVFacets 15.36 (9.1%) 15.28
(9.5%) -0.5% ( -17% - 19%) 0.864
LowPhrase 538.67 (4.0%) 536.18
(3.8%) -0.5% ( -7% - 7%) 0.710
BrowseMonthSSDVFacets 20.59 (1.6%) 20.49
(1.7%) -0.5% ( -3% - 2%) 0.394
MedSpanNear 15.50 (5.1%) 15.44
(4.5%) -0.4% ( -9% - 9%) 0.788
Respell 58.48 (1.9%) 58.25
(1.4%) -0.4% ( -3% - 3%) 0.465
HighTermMonthSort 3272.95 (5.1%) 3260.51
(3.8%) -0.4% ( -8% - 8%) 0.789
AndHighMedDayTaxoFacets 54.85 (2.3%) 54.70
(1.6%) -0.3% ( -4% - 3%) 0.681
LowSpanNear 37.04 (4.4%) 36.95
(3.8%) -0.2% ( -8% - 8%) 0.855
OrHighNotMed 572.03 (5.9%) 571.26
(5.5%) -0.1% ( -10% - 11%) 0.941
AndHighHighDayTaxoFacets 13.89 (4.7%) 13.89
(2.7%) -0.0% ( -7% - 7%) 0.974
MedSloppyPhrase 6.27 (4.2%) 6.27
(5.4%) -0.0% ( -9% - 9%) 0.980
Prefix3 647.76 (2.6%) 648.17
(1.9%) 0.1% ( -4% - 4%) 0.929
PKLookup 242.95 (5.0%) 243.25
(4.4%) 0.1% ( -8% - 9%) 0.934
HighPhrase 80.44 (3.6%) 80.57
(4.1%) 0.2% ( -7% - 8%) 0.899
Wildcard 57.52 (3.0%) 57.61
(2.6%) 0.2% ( -5% - 5%) 0.852
MedPhrase 73.08 (2.7%) 73.28
(3.4%) 0.3% ( -5% - 6%) 0.773
OrHighMedDayTaxoFacets 19.03 (3.1%) 19.09
(2.9%) 0.3% ( -5% - 6%) 0.723
OrNotHighHigh 524.34 (5.4%) 527.14
(5.7%) 0.5% ( -10% - 12%) 0.761
MedIntervalsOrdered 128.67 (4.8%) 129.38
(3.9%) 0.6% ( -7% - 9%) 0.688
MedTermDayTaxoFacets 52.89 (6.6%) 53.31
(4.4%) 0.8% ( -9% - 12%) 0.655
OrHighNotLow 432.03 (6.4%) 435.93
(7.6%) 0.9% ( -12% - 15%) 0.685
OrHighNotHigh 240.53 (6.9%) 242.89
(7.2%) 1.0% ( -12% - 16%) 0.660
MedTerm 684.80 (5.7%) 691.78
(6.3%) 1.0% ( -10% - 13%) 0.591
HighIntervalsOrdered 2.08 (5.0%) 2.10
(5.1%) 1.1% ( -8% - 11%) 0.509
HighTermTitleBDVSort 18.90 (4.9%) 19.11
(3.0%) 1.1% ( -6% - 9%) 0.390
BrowseDayOfYearSSDVFacets 20.08 (8.3%) 20.44
(11.7%) 1.8% ( -16% - 23%) 0.574
OrHighMed 175.91 (3.4%) 179.48
(4.0%) 2.0% ( -5% - 9%) 0.085
LowIntervalsOrdered 324.63 (5.4%) 331.22
(4.5%) 2.0% ( -7% - 12%) 0.199
HighTerm 566.09 (6.0%) 577.68
(6.8%) 2.0% ( -10% - 15%) 0.310
IntNRQ 1518.60 (6.5%) 1555.96
(7.7%) 2.5% ( -10% - 17%) 0.273
LowTerm 965.29 (5.9%) 992.04
(6.1%) 2.8% ( -8% - 15%) 0.146
OrHighHigh 53.17 (4.6%) 55.48
(5.6%) 4.3% ( -5% - 15%) 0.007
```
--
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]
