gautamworah96 commented on pull request #179:
URL: https://github.com/apache/lucene/pull/179#issuecomment-901547047
Hey @mikemccand , my apologies for leaving this PR hanging! The unexpected
benchmark results threw me off..
I've updated the PR to fix the bug you had previously discovered and updated
the PR to bring it on par with mainline.
I re-ran benchmarks after fixing the `leafReaderDocBase` not added to the
`leafReaderMaxDoc` bug:
TaskQPS baseline StdDevQPS my_modified_version
StdDev Pct diff p-value
OrHighLow 394.22 (4.1%) 390.60 (4.7%)
-0.9% ( -9% - 8%) 0.508
OrHighNotHigh 393.24 (5.0%) 393.11 (4.4%)
-0.0% ( -9% - 9%) 0.982
HighIntervalsOrdered 2.23 (1.3%) 2.23 (1.2%)
0.1% ( -2% - 2%) 0.768
HighPhrase 169.85 (4.9%) 170.09 (5.7%)
0.1% ( -9% - 11%) 0.932
BrowseDateTaxoFacets 0.57 (1.6%) 0.57 (1.6%)
0.2% ( -2% - 3%) 0.684
HighSpanNear 3.57 (2.2%) 3.58 (2.1%)
0.2% ( -3% - 4%) 0.711
LowSpanNear 11.32 (1.8%) 11.35 (1.4%)
0.3% ( -2% - 3%) 0.604
OrNotHighMed 398.53 (6.1%) 399.62 (7.5%)
0.3% ( -12% - 14%) 0.900
MedSpanNear 11.37 (1.8%) 11.41 (1.2%)
0.3% ( -2% - 3%) 0.522
LowTerm 949.41 (5.7%) 952.69 (5.2%)
0.3% ( -9% - 11%) 0.841
OrHighHigh 27.45 (2.5%) 27.55 (2.1%)
0.4% ( -4% - 5%) 0.615
IntNRQ 19.98 (30.1%) 20.06 (29.6%)
0.4% ( -45% - 85%) 0.967
Respell 24.48 (2.4%) 24.58 (2.5%)
0.4% ( -4% - 5%) 0.575
LowSloppyPhrase 19.15 (2.6%) 19.23 (2.3%)
0.4% ( -4% - 5%) 0.569
OrHighMed 23.02 (2.9%) 23.14 (1.8%)
0.5% ( -4% - 5%) 0.482
BrowseDayOfYearTaxoFacets 6311.77 (3.7%) 6349.39
(4.2%) 0.6% ( -7% - 8%) 0.634
Fuzzy1 42.80 (3.9%) 43.06 (5.0%)
0.6% ( -7% - 9%) 0.666
OrHighNotMed 446.39 (5.5%) 449.32 (5.6%)
0.7% ( -9% - 12%) 0.707
HighSloppyPhrase 3.57 (4.4%) 3.59 (3.8%)
0.7% ( -7% - 9%) 0.610
OrNotHighLow 418.72 (3.6%) 421.68 (3.6%)
0.7% ( -6% - 8%) 0.535
LowPhrase 12.47 (2.8%) 12.56 (2.8%)
0.8% ( -4% - 6%) 0.391
OrHighNotLow 494.57 (4.9%) 498.46 (5.4%)
0.8% ( -9% - 11%) 0.629
HighTermDayOfYearSort 7.07 (12.8%) 7.13 (10.2%)
0.9% ( -19% - 27%) 0.799
MedSloppyPhrase 21.85 (2.8%) 22.06 (2.5%)
0.9% ( -4% - 6%) 0.266
Fuzzy2 35.04 (3.6%) 35.43 (3.7%)
1.1% ( -5% - 8%) 0.330
HighTermTitleBDVSort 39.93 (8.7%) 40.40 (6.9%)
1.2% ( -13% - 18%) 0.638
Wildcard 38.16 (4.4%) 38.60 (3.9%)
1.2% ( -6% - 9%) 0.370
AndHighMed 47.87 (2.7%) 48.45 (3.6%)
1.2% ( -4% - 7%) 0.224
OrNotHighHigh 367.90 (4.8%) 372.56 (3.9%)
1.3% ( -7% - 10%) 0.362
AndHighHigh 20.63 (3.2%) 20.94 (4.6%)
1.5% ( -6% - 9%) 0.221
HighTermMonthSort 25.15 (12.1%) 25.54 (13.8%)
1.6% ( -21% - 31%) 0.703
MedPhrase 54.64 (6.1%) 55.54 (6.9%)
1.6% ( -10% - 15%) 0.424
TermDTSort 16.01 (14.6%) 16.28 (16.7%)
1.7% ( -25% - 38%) 0.737
BrowseMonthSSDVFacets 2.62 (0.9%) 2.66 (2.1%)
1.7% ( -1% - 4%) 0.001
AndHighLow 252.16 (3.3%) 256.71 (3.3%)
1.8% ( -4% - 8%) 0.086
HighTerm 998.70 (6.8%) 1020.66 (5.8%)
2.2% ( -9% - 15%) 0.271
BrowseMonthTaxoFacets 6354.02 (5.9%) 6495.78 (5.4%)
2.2% ( -8% - 14%) 0.211
PKLookup 113.56 (4.2%) 116.24 (4.4%)
2.4% ( -6% - 11%) 0.084
BrowseDayOfYearSSDVFacets 2.46 (0.5%) 2.52
(2.9%) 2.5% ( 0% - 5%) 0.000
MedTerm 815.18 (4.9%) 837.70 (6.0%)
2.8% ( -7% - 14%) 0.109
Prefix3 144.90 (10.0%) 149.12 (11.5%)
2.9% ( -16% - 27%) 0.392
The gain in taxonomy based facets is about 2%. I ran it multiple times
locally and it gave around 2% gain in those runs as well. SSDV facets still
show a 3% gain which is unexplained.
--
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: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
