mikemccand commented on PR #13472:
URL: https://github.com/apache/lucene/pull/13472#issuecomment-2182966557
Again, with BMW enabled again (i.e. Lucene's default):
```
se 18.329667||P999 Cmp 17.666751||Pct Diff -3.6166287145314704||P100 Base
43.345448||P100 Cmp 38.763535||Pct Diff -10.570690144902875
TaskQPS main-conc StdDevQPS branch-conc
StdDev Pct diff p-value
Prefix3 1132.20 (9.7%) 1093.01
(7.9%) -3.5% ( -19% - 15%) 0.217
HighSpanNear 8.39 (4.7%) 8.13
(4.7%) -3.1% ( -11% - 6%) 0.036
HighIntervalsOrdered 11.05 (5.4%) 10.78
(4.4%) -2.4% ( -11% - 7%) 0.126
OrNotHighLow 518.80 (10.0%) 508.31
(7.3%) -2.0% ( -17% - 17%) 0.467
HighTermMonthSort 720.24 (32.8%) 707.34
(38.4%) -1.8% ( -55% - 103%) 0.874
OrHighHigh 74.56 (3.6%) 73.66
(2.5%) -1.2% ( -7% - 5%) 0.218
AndHighMedDayTaxoFacets 98.52 (2.3%) 97.64
(1.0%) -0.9% ( -4% - 2%) 0.105
OrHighMedDayTaxoFacets 7.39 (1.2%) 7.34
(1.8%) -0.7% ( -3% - 2%) 0.156
Fuzzy1 131.47 (1.4%) 130.92
(1.0%) -0.4% ( -2% - 1%) 0.273
BrowseDateTaxoFacets 4.86 (0.4%) 4.84
(0.5%) -0.2% ( -1% - 0%) 0.078
HighTermDayOfYearSort 299.45 (5.5%) 298.80
(4.7%) -0.2% ( -9% - 10%) 0.893
BrowseDayOfYearSSDVFacets 23.33 (0.4%) 23.30
(0.4%) -0.1% ( 0% - 0%) 0.287
BrowseMonthTaxoFacets 5.12 (0.4%) 5.11
(0.4%) -0.1% ( 0% - 0%) 0.270
Fuzzy2 90.36 (1.7%) 90.24
(9.7%) -0.1% ( -11% - 11%) 0.952
BrowseDayOfYearTaxoFacets 18.49 (0.3%) 18.48
(0.7%) -0.1% ( -1% - 0%) 0.564
BrowseMonthSSDVFacets 8.77 (0.3%) 8.76
(0.2%) -0.0% ( 0% - 0%) 0.579
BrowseRandomLabelSSDVFacets 5.90 (0.3%) 5.90
(0.2%) 0.0% ( 0% - 0%) 0.869
MedSloppyPhrase 17.21 (2.7%) 17.23
(1.8%) 0.1% ( -4% - 4%) 0.854
AndHighHighDayTaxoFacets 6.95 (0.4%) 6.96
(0.4%) 0.2% ( 0% - 0%) 0.101
HighSloppyPhrase 25.81 (3.7%) 25.87
(3.2%) 0.2% ( -6% - 7%) 0.848
LowSpanNear 30.77 (3.7%) 30.89
(2.2%) 0.4% ( -5% - 6%) 0.683
LowIntervalsOrdered 44.40 (2.0%) 44.65
(2.3%) 0.6% ( -3% - 5%) 0.406
Wildcard 124.82 (7.6%) 126.19
(6.6%) 1.1% ( -12% - 16%) 0.626
OrHighNotLow 218.46 (3.1%) 221.07
(3.1%) 1.2% ( -4% - 7%) 0.225
HighTermTitleSort 34.91 (3.4%) 35.41
(1.5%) 1.4% ( -3% - 6%) 0.086
MedTermDayTaxoFacets 22.46 (1.6%) 22.81
(2.0%) 1.6% ( -2% - 5%) 0.007
AndHighHigh 102.09 (3.1%) 104.00
(2.5%) 1.9% ( -3% - 7%) 0.033
IntNRQ 502.04 (3.5%) 511.50
(3.6%) 1.9% ( -5% - 9%) 0.092
MedIntervalsOrdered 45.14 (4.2%) 46.12
(6.9%) 2.2% ( -8% - 13%) 0.230
OrNotHighHigh 98.43 (7.0%) 100.64
(7.2%) 2.2% ( -11% - 17%) 0.317
OrNotHighMed 139.86 (7.1%) 143.10
(9.0%) 2.3% ( -12% - 19%) 0.367
Respell 45.34 (10.4%) 46.39
(12.3%) 2.3% ( -18% - 27%) 0.519
MedPhrase 39.65 (0.7%) 40.61
(0.9%) 2.4% ( 0% - 4%) 0.000
LowSloppyPhrase 58.51 (1.7%) 59.94
(1.8%) 2.4% ( -1% - 6%) 0.000
MedSpanNear 59.70 (5.8%) 61.56
(5.7%) 3.1% ( -7% - 15%) 0.084
HighTerm 253.49 (5.8%) 261.54
(7.9%) 3.2% ( -9% - 17%) 0.148
OrHighLow 348.72 (23.4%) 361.75
(23.9%) 3.7% ( -35% - 66%) 0.617
HighPhrase 281.12 (1.8%) 293.95
(0.9%) 4.6% ( 1% - 7%) 0.000
OrHighNotMed 132.32 (4.9%) 138.67
(6.0%) 4.8% ( -5% - 16%) 0.005
BrowseRandomLabelTaxoFacets 4.53 (10.0%) 4.76
(2.7%) 5.2% ( -6% - 19%) 0.025
LowPhrase 84.50 (1.6%) 89.33
(1.4%) 5.7% ( 2% - 8%) 0.000
OrHighMed 229.07 (5.8%) 242.42
(7.7%) 5.8% ( -7% - 20%) 0.007
OrHighNotHigh 93.10 (8.6%) 99.11
(10.4%) 6.4% ( -11% - 27%) 0.032
AndHighMed 368.55 (2.8%) 394.64
(6.7%) 7.1% ( -2% - 17%) 0.000
MedTerm 189.05 (8.7%) 203.14
(10.6%) 7.5% ( -10% - 29%) 0.015
LowTerm 281.21 (10.1%) 305.74
(7.0%) 8.7% ( -7% - 28%) 0.002
AndHighLow 425.29 (4.6%) 481.36
(36.4%) 13.2% ( -26% - 56%) 0.108
HighTermTitleBDVSort 10.21 (5.4%) 11.56
(3.2%) 13.2% ( 4% - 23%) 0.000
TermDTSort 118.33 (2.8%) 136.60
(2.2%) 15.4% ( 10% - 21%) 0.000
```
I think I trust these results ... and they do show some nice speedups with
high confidence, and maybe a bit of slowndown but also with less confidence.
It's interesting how BMW alters the intra-query concurrency game ...
--
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]
