zacharymorn commented on PR #12194: URL: https://github.com/apache/lucene/pull/12194#issuecomment-1608695525
Thanks @jpountz for the feedback! On the second approach, I was actually thinking something simpler, such as this https://github.com/apache/lucene/pull/12194/commits/219beab1b1baea4950a3e47bc9bd46df32fef055. I ran the benchmark tests after the change, and got the following results: #### Index with sorting Result 1: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value AndMedFilterMonth 2142.90 (7.1%) 1996.67 (5.0%) -6.8% ( -17% - 5%) 0.000 AndHighHighDayTaxoFacets 13.99 (2.3%) 13.08 (2.3%) -6.5% ( -10% - -1%) 0.000 HighTermTitleSort 131.92 (2.5%) 124.03 (2.8%) -6.0% ( -10% - 0%) 0.000 AndHighMedDayTaxoFacets 91.57 (2.0%) 86.87 (3.2%) -5.1% ( -10% - 0%) 0.000 HighTermDayOfYearSort 235.50 (4.8%) 227.18 (4.1%) -3.5% ( -11% - 5%) 0.012 MedTermDayTaxoFacets 72.51 (2.3%) 70.17 (2.5%) -3.2% ( -7% - 1%) 0.000 HighTermMonthSort 3126.27 (3.2%) 3032.93 (3.8%) -3.0% ( -9% - 4%) 0.007 Wildcard 280.45 (5.7%) 272.30 (5.7%) -2.9% ( -13% - 8%) 0.107 OrHighMedDayTaxoFacets 22.17 (2.5%) 21.55 (2.4%) -2.8% ( -7% - 2%) 0.000 HighTermTitleBDVSort 23.23 (2.9%) 22.59 (2.9%) -2.7% ( -8% - 3%) 0.003 TermDTSort 198.40 (9.4%) 194.24 (9.1%) -2.1% ( -18% - 18%) 0.474 MedIntervalsOrdered 53.97 (13.6%) 52.88 (12.8%) -2.0% ( -24% - 28%) 0.629 MedSpanNear 172.23 (2.7%) 168.92 (2.4%) -1.9% ( -6% - 3%) 0.018 BrowseMonthSSDVFacets 20.44 (5.9%) 20.09 (6.0%) -1.7% ( -12% - 10%) 0.360 Prefix3 879.87 (4.4%) 866.61 (4.5%) -1.5% ( -9% - 7%) 0.284 HighSpanNear 50.36 (2.3%) 49.60 (1.9%) -1.5% ( -5% - 2%) 0.022 LowSpanNear 372.08 (2.3%) 366.84 (1.7%) -1.4% ( -5% - 2%) 0.030 AndHighLow 1852.63 (3.8%) 1829.62 (4.0%) -1.2% ( -8% - 6%) 0.318 LowIntervalsOrdered 9.32 (15.4%) 9.21 (15.4%) -1.2% ( -27% - 35%) 0.811 HighIntervalsOrdered 1.55 (20.5%) 1.53 (20.6%) -1.1% ( -35% - 50%) 0.872 BrowseDayOfYearSSDVFacets 14.47 (4.8%) 14.33 (5.4%) -1.0% ( -10% - 9%) 0.524 OrNotHighLow 1128.41 (4.1%) 1117.38 (3.9%) -1.0% ( -8% - 7%) 0.438 OrHighNotHigh 492.96 (3.9%) 488.47 (4.3%) -0.9% ( -8% - 7%) 0.486 MedPhrase 372.00 (2.7%) 368.75 (3.0%) -0.9% ( -6% - 4%) 0.325 OrNotHighMed 628.90 (2.8%) 623.55 (2.5%) -0.9% ( -5% - 4%) 0.310 OrHighNotLow 791.85 (3.8%) 785.12 (4.5%) -0.9% ( -8% - 7%) 0.520 OrNotHighHigh 447.73 (2.9%) 444.15 (3.1%) -0.8% ( -6% - 5%) 0.399 OrHighHigh 45.13 (4.0%) 44.78 (4.5%) -0.8% ( -8% - 8%) 0.575 OrHighNotMed 543.72 (3.6%) 539.69 (4.2%) -0.7% ( -8% - 7%) 0.545 HighTerm 946.86 (4.2%) 942.67 (3.8%) -0.4% ( -8% - 7%) 0.727 LowTerm 1446.44 (5.6%) 1442.00 (5.9%) -0.3% ( -11% - 11%) 0.866 MedTerm 1064.71 (3.8%) 1061.77 (3.7%) -0.3% ( -7% - 7%) 0.815 Fuzzy1 78.98 (6.5%) 78.77 (6.2%) -0.3% ( -12% - 13%) 0.896 BrowseDayOfYearTaxoFacets 14.60 (3.5%) 14.56 (3.0%) -0.2% ( -6% - 6%) 0.808 OrHighLow 383.35 (3.5%) 382.47 (4.1%) -0.2% ( -7% - 7%) 0.848 AndHighMed 419.38 (3.9%) 418.65 (3.4%) -0.2% ( -7% - 7%) 0.879 LowSloppyPhrase 45.89 (1.9%) 45.85 (1.8%) -0.1% ( -3% - 3%) 0.895 BrowseDateTaxoFacets 19.56 (2.2%) 19.55 (2.0%) -0.1% ( -4% - 4%) 0.925 OrHighMed 53.51 (4.0%) 53.49 (3.8%) -0.1% ( -7% - 8%) 0.965 MedSloppyPhrase 77.05 (1.8%) 77.03 (2.0%) -0.0% ( -3% - 3%) 0.962 AndHighHigh 119.95 (3.4%) 119.92 (3.2%) -0.0% ( -6% - 6%) 0.982 BrowseMonthTaxoFacets 18.33 (2.3%) 18.35 (2.3%) 0.1% ( -4% - 4%) 0.911 IntNRQ 119.39 (22.2%) 119.66 (22.4%) 0.2% ( -36% - 57%) 0.974 LowPhrase 105.06 (3.4%) 105.34 (4.0%) 0.3% ( -6% - 7%) 0.824 PKLookup 282.66 (3.6%) 283.54 (3.4%) 0.3% ( -6% - 7%) 0.782 BrowseDateSSDVFacets 5.42 (7.9%) 5.45 (7.3%) 0.5% ( -13% - 16%) 0.839 HighPhrase 72.39 (2.6%) 72.78 (2.6%) 0.5% ( -4% - 5%) 0.515 HighSloppyPhrase 1.41 (5.7%) 1.42 (3.5%) 0.9% ( -7% - 10%) 0.567 Respell 95.35 (5.5%) 96.22 (5.3%) 0.9% ( -9% - 12%) 0.593 Fuzzy2 102.31 (9.0%) 103.34 (9.6%) 1.0% ( -16% - 21%) 0.731 AndMedNotMonth 873.00 (3.1%) 984.05 (3.3%) 12.7% ( 6% - 19%) 0.000 AndHighFilterMonth 362.01 (2.8%) 484.47 (6.0%) 33.8% ( 24% - 43%) 0.000 AndHighNotMonth 253.17 (1.8%) 596.55 (10.5%) 135.6% ( 121% - 150%) 0.000 ``` Result 2: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value AndHighMedDayTaxoFacets 66.69 (1.9%) 60.98 (4.1%) -8.6% ( -14% - -2%) 0.000 Prefix3 419.25 (4.7%) 387.06 (4.0%) -7.7% ( -15% - 1%) 0.000 AndHighHighDayTaxoFacets 38.85 (1.6%) 36.34 (3.7%) -6.5% ( -11% - -1%) 0.000 OrNotHighMed 418.16 (2.1%) 395.85 (3.8%) -5.3% ( -11% - 0%) 0.000 TermDTSort 151.83 (3.8%) 143.84 (3.5%) -5.3% ( -12% - 2%) 0.000 AndMedFilterMonth 1849.70 (4.7%) 1753.86 (5.3%) -5.2% ( -14% - 5%) 0.001 HighTermMonthSort 3079.34 (3.5%) 2932.00 (4.4%) -4.8% ( -12% - 3%) 0.000 HighTermTitleSort 161.65 (10.3%) 154.99 (7.7%) -4.1% ( -20% - 15%) 0.153 OrHighMedDayTaxoFacets 3.72 (3.8%) 3.59 (2.9%) -3.5% ( -9% - 3%) 0.001 HighTermDayOfYearSort 407.83 (3.0%) 394.86 (4.0%) -3.2% ( -9% - 3%) 0.005 MedTermDayTaxoFacets 69.43 (2.4%) 67.48 (2.0%) -2.8% ( -7% - 1%) 0.000 OrNotHighLow 1816.25 (3.9%) 1769.50 (3.9%) -2.6% ( -9% - 5%) 0.037 HighTermTitleBDVSort 36.84 (2.1%) 36.03 (1.0%) -2.2% ( -5% - 0%) 0.000 Wildcard 162.28 (6.6%) 158.82 (6.4%) -2.1% ( -14% - 11%) 0.297 LowTerm 1036.19 (5.4%) 1019.64 (5.9%) -1.6% ( -12% - 10%) 0.373 HighSpanNear 36.59 (5.4%) 36.09 (5.5%) -1.4% ( -11% - 10%) 0.428 LowSpanNear 103.37 (1.7%) 102.13 (1.9%) -1.2% ( -4% - 2%) 0.035 MedSpanNear 51.35 (1.6%) 50.86 (1.8%) -0.9% ( -4% - 2%) 0.083 OrHighLow 660.72 (3.4%) 656.05 (3.2%) -0.7% ( -7% - 6%) 0.494 MedIntervalsOrdered 21.71 (7.6%) 21.57 (7.6%) -0.6% ( -14% - 15%) 0.794 HighIntervalsOrdered 12.02 (9.4%) 11.96 (9.2%) -0.5% ( -17% - 19%) 0.854 LowIntervalsOrdered 55.04 (7.4%) 54.74 (7.4%) -0.5% ( -14% - 15%) 0.819 OrNotHighHigh 571.44 (2.6%) 568.54 (3.2%) -0.5% ( -6% - 5%) 0.581 MedTerm 1102.15 (4.2%) 1097.46 (4.4%) -0.4% ( -8% - 8%) 0.755 HighPhrase 108.28 (3.3%) 107.92 (2.9%) -0.3% ( -6% - 6%) 0.736 IntNRQ 81.85 (9.6%) 81.66 (9.6%) -0.2% ( -17% - 21%) 0.941 OrHighNotHigh 401.85 (3.9%) 401.24 (4.3%) -0.2% ( -8% - 8%) 0.907 HighSloppyPhrase 26.71 (4.1%) 26.67 (3.8%) -0.1% ( -7% - 8%) 0.908 Respell 110.22 (5.5%) 110.08 (5.1%) -0.1% ( -10% - 11%) 0.939 OrHighNotMed 635.92 (3.7%) 635.34 (4.1%) -0.1% ( -7% - 8%) 0.942 LowSloppyPhrase 11.69 (4.1%) 11.68 (3.9%) -0.1% ( -7% - 8%) 0.951 Fuzzy2 35.65 (7.7%) 35.65 (7.3%) -0.0% ( -13% - 16%) 0.995 OrHighMed 225.90 (2.9%) 225.90 (3.1%) 0.0% ( -5% - 6%) 0.998 PKLookup 282.08 (3.1%) 282.19 (3.6%) 0.0% ( -6% - 6%) 0.969 BrowseDateSSDVFacets 5.43 (9.8%) 5.44 (9.7%) 0.1% ( -17% - 21%) 0.981 AndHighMed 139.86 (2.5%) 139.97 (2.9%) 0.1% ( -5% - 5%) 0.929 OrHighHigh 42.67 (2.4%) 42.71 (2.7%) 0.1% ( -4% - 5%) 0.923 HighTerm 1193.77 (4.6%) 1194.76 (4.8%) 0.1% ( -8% - 10%) 0.956 AndHighHigh 72.75 (2.0%) 72.84 (3.0%) 0.1% ( -4% - 5%) 0.876 MedPhrase 17.85 (2.8%) 17.88 (2.5%) 0.2% ( -5% - 5%) 0.833 BrowseDateTaxoFacets 19.36 (2.5%) 19.40 (2.2%) 0.2% ( -4% - 5%) 0.794 AndHighLow 1612.73 (3.5%) 1616.44 (3.2%) 0.2% ( -6% - 7%) 0.826 MedSloppyPhrase 73.06 (2.5%) 73.26 (2.5%) 0.3% ( -4% - 5%) 0.740 BrowseDayOfYearTaxoFacets 14.68 (4.7%) 14.72 (4.3%) 0.3% ( -8% - 9%) 0.833 BrowseDayOfYearSSDVFacets 14.23 (7.7%) 14.27 (8.0%) 0.3% ( -14% - 17%) 0.900 BrowseMonthTaxoFacets 18.14 (2.6%) 18.20 (2.5%) 0.3% ( -4% - 5%) 0.695 OrHighNotLow 532.83 (4.4%) 535.13 (5.2%) 0.4% ( -8% - 10%) 0.778 LowPhrase 79.15 (2.1%) 79.50 (2.3%) 0.4% ( -3% - 4%) 0.528 Fuzzy1 103.09 (7.8%) 104.65 (7.6%) 1.5% ( -12% - 18%) 0.536 BrowseMonthSSDVFacets 19.62 (7.3%) 19.94 (12.2%) 1.6% ( -16% - 22%) 0.607 AndMedNotMonth 891.86 (2.8%) 1048.85 (5.4%) 17.6% ( 9% - 26%) 0.000 AndHighFilterMonth 685.23 (2.7%) 835.74 (6.1%) 22.0% ( 12% - 31%) 0.000 AndHighNotMonth 76.06 (0.9%) 657.26 (50.3%) 764.1% ( 706% - 822%) 0.000 ``` #### Index without sorting Result 1: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Prefix3 620.25 (3.8%) 572.56 (3.2%) -7.7% ( -14% - 0%) 0.000 Wildcard 94.62 (6.0%) 87.96 (5.2%) -7.0% ( -17% - 4%) 0.000 HighTermTitleSort 185.23 (15.8%) 172.39 (15.9%) -6.9% ( -33% - 29%) 0.167 AndHighHighDayTaxoFacets 13.33 (3.2%) 12.53 (3.2%) -6.0% ( -11% - 0%) 0.000 AndMedFilterMonth 780.10 (3.2%) 750.30 (3.5%) -3.8% ( -10% - 2%) 0.000 AndHighMedDayTaxoFacets 78.48 (1.9%) 75.69 (2.5%) -3.6% ( -7% - 0%) 0.000 OrHighMedDayTaxoFacets 9.81 (6.9%) 9.48 (4.9%) -3.4% ( -14% - 9%) 0.077 MedTermDayTaxoFacets 91.99 (2.5%) 89.02 (2.2%) -3.2% ( -7% - 1%) 0.000 AndHighFilterMonth 458.58 (4.4%) 444.81 (4.4%) -3.0% ( -11% - 6%) 0.032 HighTermTitleBDVSort 28.60 (2.0%) 27.83 (1.3%) -2.7% ( -5% - 0%) 0.000 IntNRQ 84.30 (9.3%) 82.10 (11.6%) -2.6% ( -21% - 20%) 0.430 HighTermMonthSort 2392.61 (7.2%) 2330.41 (5.4%) -2.6% ( -14% - 10%) 0.198 OrNotHighLow 1069.43 (3.6%) 1042.84 (4.4%) -2.5% ( -10% - 5%) 0.050 LowIntervalsOrdered 227.98 (2.7%) 222.64 (4.2%) -2.3% ( -8% - 4%) 0.035 TermDTSort 212.12 (3.3%) 207.21 (3.9%) -2.3% ( -9% - 5%) 0.044 HighIntervalsOrdered 38.94 (1.8%) 38.11 (3.3%) -2.1% ( -7% - 3%) 0.011 OrNotHighMed 593.33 (3.4%) 581.36 (3.7%) -2.0% ( -8% - 5%) 0.074 MedIntervalsOrdered 21.30 (1.3%) 20.88 (2.9%) -2.0% ( -6% - 2%) 0.005 MedSpanNear 25.77 (2.5%) 25.41 (2.4%) -1.4% ( -6% - 3%) 0.062 HighSpanNear 17.69 (2.7%) 17.45 (2.7%) -1.4% ( -6% - 4%) 0.107 LowSpanNear 128.36 (1.4%) 126.90 (1.3%) -1.1% ( -3% - 1%) 0.008 OrNotHighHigh 548.99 (3.4%) 545.38 (2.9%) -0.7% ( -6% - 5%) 0.508 HighTermDayOfYearSort 424.25 (5.8%) 421.73 (3.8%) -0.6% ( -9% - 9%) 0.703 OrHighNotHigh 465.31 (3.7%) 462.73 (4.0%) -0.6% ( -7% - 7%) 0.648 PKLookup 292.46 (4.0%) 290.85 (3.8%) -0.6% ( -7% - 7%) 0.652 HighPhrase 170.22 (3.6%) 169.76 (3.5%) -0.3% ( -7% - 7%) 0.808 BrowseMonthTaxoFacets 40.22 (23.2%) 40.14 (23.1%) -0.2% ( -37% - 60%) 0.979 BrowseDateTaxoFacets 45.02 (10.1%) 44.99 (10.0%) -0.1% ( -18% - 22%) 0.982 BrowseDayOfYearTaxoFacets 45.36 (7.5%) 45.33 (7.3%) -0.1% ( -13% - 15%) 0.978 LowSloppyPhrase 17.25 (5.0%) 17.25 (4.9%) -0.0% ( -9% - 10%) 0.999 Fuzzy1 83.95 (4.1%) 84.05 (4.2%) 0.1% ( -7% - 8%) 0.928 BrowseDayOfYearSSDVFacets 26.61 (26.3%) 26.64 (26.3%) 0.1% ( -41% - 71%) 0.988 MedPhrase 47.60 (3.3%) 47.66 (3.3%) 0.1% ( -6% - 7%) 0.900 MedSloppyPhrase 25.42 (4.8%) 25.45 (4.8%) 0.1% ( -9% - 10%) 0.930 OrHighNotMed 533.40 (3.8%) 534.12 (3.9%) 0.1% ( -7% - 8%) 0.912 LowPhrase 72.95 (2.5%) 73.07 (2.5%) 0.2% ( -4% - 5%) 0.836 AndHighMed 331.56 (5.4%) 332.11 (5.5%) 0.2% ( -10% - 11%) 0.923 LowTerm 1279.72 (7.0%) 1282.14 (5.1%) 0.2% ( -11% - 13%) 0.922 Respell 94.36 (6.2%) 94.70 (5.9%) 0.4% ( -11% - 13%) 0.850 BrowseMonthSSDVFacets 26.44 (25.9%) 26.53 (25.7%) 0.4% ( -40% - 70%) 0.964 HighSloppyPhrase 38.93 (4.1%) 39.09 (4.0%) 0.4% ( -7% - 8%) 0.750 AndMedNotMonth 891.59 (3.9%) 895.75 (2.6%) 0.5% ( -5% - 7%) 0.657 AndHighLow 1747.72 (3.8%) 1756.48 (4.2%) 0.5% ( -7% - 8%) 0.693 AndHighHigh 77.63 (4.2%) 78.03 (4.1%) 0.5% ( -7% - 9%) 0.697 OrHighNotLow 646.76 (4.5%) 650.22 (4.7%) 0.5% ( -8% - 10%) 0.712 Fuzzy2 69.85 (7.1%) 70.28 (7.4%) 0.6% ( -12% - 16%) 0.789 OrHighMed 251.43 (4.0%) 252.98 (4.3%) 0.6% ( -7% - 9%) 0.637 BrowseDateSSDVFacets 5.52 (2.8%) 5.56 (4.1%) 0.7% ( -6% - 7%) 0.534 OrHighLow 509.73 (5.2%) 516.04 (4.7%) 1.2% ( -8% - 11%) 0.432 HighTerm 716.67 (5.0%) 725.95 (4.2%) 1.3% ( -7% - 10%) 0.372 MedTerm 1009.75 (4.5%) 1023.79 (4.1%) 1.4% ( -6% - 10%) 0.306 OrHighHigh 40.11 (3.8%) 40.67 (3.4%) 1.4% ( -5% - 8%) 0.217 AndHighNotMonth 443.44 (4.1%) 467.71 (3.5%) 5.5% ( -2% - 13%) 0.000 ``` Result 2: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Prefix3 379.27 (4.2%) 350.63 (3.5%) -7.6% ( -14% - 0%) 0.000 HighTermMonthSort 2644.50 (10.0%) 2467.84 (7.9%) -6.7% ( -22% - 12%) 0.020 Wildcard 111.47 (5.8%) 105.48 (5.2%) -5.4% ( -15% - 6%) 0.002 HighTermTitleSort 163.42 (8.4%) 155.35 (6.4%) -4.9% ( -18% - 10%) 0.036 AndHighMedDayTaxoFacets 106.98 (2.2%) 101.81 (3.1%) -4.8% ( -9% - 0%) 0.000 TermDTSort 200.05 (4.0%) 192.23 (6.4%) -3.9% ( -13% - 6%) 0.021 AndHighHighDayTaxoFacets 39.09 (1.8%) 37.65 (1.5%) -3.7% ( -6% - 0%) 0.000 AndHighFilterMonth 205.53 (2.8%) 198.29 (3.4%) -3.5% ( -9% - 2%) 0.000 HighTermTitleBDVSort 10.04 (2.4%) 9.74 (2.1%) -3.0% ( -7% - 1%) 0.000 MedTermDayTaxoFacets 89.94 (2.0%) 87.38 (1.8%) -2.8% ( -6% - 1%) 0.000 HighTermDayOfYearSort 418.97 (6.0%) 408.95 (6.9%) -2.4% ( -14% - 11%) 0.241 AndMedFilterMonth 639.08 (2.2%) 624.53 (2.5%) -2.3% ( -6% - 2%) 0.002 OrHighMedDayTaxoFacets 7.17 (4.0%) 7.02 (4.4%) -2.1% ( -10% - 6%) 0.112 AndHighNotMonth 606.76 (4.7%) 597.90 (4.7%) -1.5% ( -10% - 8%) 0.328 OrNotHighMed 624.52 (2.5%) 616.48 (3.1%) -1.3% ( -6% - 4%) 0.149 MedTerm 883.90 (5.7%) 872.74 (4.5%) -1.3% ( -10% - 9%) 0.436 AndMedNotMonth 1015.43 (3.5%) 1003.04 (2.7%) -1.2% ( -7% - 5%) 0.216 MedSpanNear 11.88 (0.9%) 11.75 (0.8%) -1.1% ( -2% - 0%) 0.000 OrHighHigh 42.30 (4.6%) 41.85 (3.4%) -1.1% ( -8% - 7%) 0.406 OrHighNotMed 573.99 (4.2%) 568.16 (4.5%) -1.0% ( -9% - 8%) 0.463 MedIntervalsOrdered 38.88 (6.8%) 38.50 (6.9%) -1.0% ( -13% - 13%) 0.656 HighTerm 1015.56 (5.6%) 1005.98 (4.9%) -0.9% ( -10% - 10%) 0.571 LowIntervalsOrdered 154.19 (8.0%) 152.80 (8.2%) -0.9% ( -15% - 16%) 0.724 OrHighNotLow 844.90 (4.9%) 837.29 (4.9%) -0.9% ( -10% - 9%) 0.561 LowSpanNear 78.74 (4.1%) 78.11 (4.1%) -0.8% ( -8% - 7%) 0.541 OrHighMed 153.42 (4.0%) 152.21 (3.4%) -0.8% ( -7% - 6%) 0.504 HighSpanNear 2.96 (0.7%) 2.94 (0.7%) -0.8% ( -2% - 0%) 0.001 MedPhrase 376.73 (2.0%) 374.16 (1.9%) -0.7% ( -4% - 3%) 0.272 HighIntervalsOrdered 29.22 (7.3%) 29.03 (7.4%) -0.7% ( -14% - 15%) 0.769 OrHighNotHigh 498.90 (4.1%) 496.29 (4.3%) -0.5% ( -8% - 8%) 0.695 LowSloppyPhrase 132.03 (3.4%) 131.46 (3.1%) -0.4% ( -6% - 6%) 0.674 PKLookup 293.57 (3.3%) 292.52 (3.3%) -0.4% ( -6% - 6%) 0.735 LowTerm 1274.81 (7.5%) 1270.44 (6.3%) -0.3% ( -13% - 14%) 0.876 BrowseDayOfYearSSDVFacets 25.40 (20.8%) 25.33 (20.8%) -0.3% ( -34% - 52%) 0.968 AndHighMed 185.94 (4.1%) 185.51 (4.1%) -0.2% ( -8% - 8%) 0.859 AndHighLow 1532.23 (4.7%) 1529.57 (4.7%) -0.2% ( -9% - 9%) 0.908 OrNotHighHigh 433.52 (3.2%) 433.05 (3.2%) -0.1% ( -6% - 6%) 0.914 OrNotHighLow 2173.75 (2.4%) 2171.48 (4.1%) -0.1% ( -6% - 6%) 0.922 HighSloppyPhrase 60.06 (4.6%) 60.02 (4.1%) -0.1% ( -8% - 8%) 0.966 IntNRQ 96.82 (0.8%) 96.78 (0.8%) -0.0% ( -1% - 1%) 0.867 BrowseDateTaxoFacets 43.83 (12.3%) 43.82 (12.3%) -0.0% ( -21% - 27%) 0.992 BrowseMonthSSDVFacets 25.69 (20.6%) 25.69 (20.6%) -0.0% ( -34% - 51%) 0.998 MedSloppyPhrase 160.44 (3.8%) 160.44 (3.6%) -0.0% ( -7% - 7%) 0.998 BrowseMonthTaxoFacets 35.59 (27.4%) 35.60 (27.3%) 0.0% ( -42% - 75%) 0.996 Fuzzy1 86.58 (7.6%) 86.62 (6.8%) 0.1% ( -13% - 15%) 0.981 OrHighLow 1145.19 (3.4%) 1146.01 (3.5%) 0.1% ( -6% - 7%) 0.948 BrowseDateSSDVFacets 5.47 (2.3%) 5.48 (2.3%) 0.1% ( -4% - 4%) 0.917 BrowseDayOfYearTaxoFacets 44.34 (11.1%) 44.38 (11.3%) 0.1% ( -20% - 25%) 0.976 Respell 90.09 (4.5%) 90.25 (4.4%) 0.2% ( -8% - 9%) 0.899 AndHighHigh 113.61 (3.3%) 113.84 (3.5%) 0.2% ( -6% - 7%) 0.849 HighPhrase 103.91 (4.5%) 104.21 (4.5%) 0.3% ( -8% - 9%) 0.837 LowPhrase 103.85 (3.5%) 104.28 (3.2%) 0.4% ( -6% - 7%) 0.697 Fuzzy2 92.09 (8.4%) 92.55 (10.2%) 0.5% ( -16% - 20%) 0.865 ``` Overall, it was able to prevent severe performance drop when the index is unsorted, and only had around 5% negative impact to `AndMedFilterMonth` task. However, I also noticed that somehow with sorted index, `AndMedFilterMonth` was also negatively impacted, so I also made some changes to the task definition below to use the same set of months from `AndHighFilterMonth`, just to see if query / index characteristics might have affected this. ```diff --- a/tasks/wikimedium.10M.nostopwords.tasks +++ b/tasks/wikimedium.10M.nostopwords.tasks @@ -17381,8 +17381,8 @@ AndHighFilterMonth: +united +filter=monthPostings:feb # freq=1185528 AndHighFilterMonth: +year +filter=monthPostings:mar # freq=1098425 AndHighFilterMonth: +its +filter=monthPostings:apr # freq=1160703 AndHighFilterMonth: +but +filter=monthPostings:may # freq=1484398 -AndMedFilterMonth: +mostly +filter=monthPostings:jun # freq=89401 -AndMedFilterMonth: +interview +filter=monthPostings:jul # freq=94736 -AndMedFilterMonth: +9 +filter=monthPostings:aug # freq=541405 -AndMedFilterMonth: +hard +filter=monthPostings:sep # freq=92045 -AndMedFilterMonth: +bay +filter=monthPostings:oct # freq=117167 \ No newline at end of file +AndMedFilterMonth: +mostly +filter=monthPostings:jan # freq=89401 +AndMedFilterMonth: +interview +filter=monthPostings:feb # freq=94736 +AndMedFilterMonth: +9 +filter=monthPostings:mar # freq=541405 +AndMedFilterMonth: +hard +filter=monthPostings:apr # freq=92045 +AndMedFilterMonth: +bay +filter=monthPostings:may # freq=117167 ``` and got the following results: #### Index with sorting Result 1: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Prefix3 283.26 (5.8%) 253.88 (4.9%) -10.4% ( -19% - 0%) 0.000 AndHighHighDayTaxoFacets 20.44 (2.7%) 19.15 (2.4%) -6.3% ( -11% - -1%) 0.000 HighTermTitleSort 120.21 (4.1%) 114.08 (3.7%) -5.1% ( -12% - 2%) 0.000 AndHighMedDayTaxoFacets 170.82 (3.3%) 162.82 (3.8%) -4.7% ( -11% - 2%) 0.000 HighTermMonthSort 3189.35 (3.2%) 3057.15 (3.0%) -4.1% ( -10% - 2%) 0.000 MedTermDayTaxoFacets 53.47 (2.7%) 51.33 (2.3%) -4.0% ( -8% - 1%) 0.000 TermDTSort 195.70 (3.6%) 188.61 (3.8%) -3.6% ( -10% - 4%) 0.002 HighTermDayOfYearSort 373.45 (2.5%) 360.41 (3.2%) -3.5% ( -9% - 2%) 0.000 OrNotHighLow 1406.59 (3.4%) 1357.71 (3.8%) -3.5% ( -10% - 3%) 0.002 OrHighMedDayTaxoFacets 19.41 (5.5%) 18.90 (4.5%) -2.7% ( -11% - 7%) 0.092 Wildcard 147.47 (6.6%) 144.35 (6.7%) -2.1% ( -14% - 11%) 0.314 LowSloppyPhrase 351.23 (4.5%) 344.15 (4.5%) -2.0% ( -10% - 7%) 0.156 HighTermTitleBDVSort 30.79 (4.5%) 30.22 (4.2%) -1.8% ( -10% - 7%) 0.177 BrowseDateSSDVFacets 5.64 (7.3%) 5.55 (10.2%) -1.7% ( -17% - 17%) 0.553 LowTerm 1412.43 (5.4%) 1393.26 (4.4%) -1.4% ( -10% - 8%) 0.382 HighSpanNear 35.54 (1.5%) 35.08 (1.1%) -1.3% ( -3% - 1%) 0.003 HighIntervalsOrdered 6.83 (13.5%) 6.74 (13.5%) -1.3% ( -24% - 29%) 0.764 OrHighLow 715.40 (3.0%) 706.45 (4.5%) -1.3% ( -8% - 6%) 0.304 OrNotHighMed 667.12 (2.4%) 659.49 (2.9%) -1.1% ( -6% - 4%) 0.173 LowIntervalsOrdered 16.55 (9.1%) 16.37 (9.1%) -1.1% ( -17% - 18%) 0.707 MedSpanNear 137.56 (2.3%) 136.08 (2.1%) -1.1% ( -5% - 3%) 0.125 LowSpanNear 126.90 (1.7%) 125.66 (1.4%) -1.0% ( -3% - 2%) 0.046 MedPhrase 259.38 (5.0%) 257.29 (5.4%) -0.8% ( -10% - 10%) 0.625 MedIntervalsOrdered 94.29 (13.2%) 93.61 (13.3%) -0.7% ( -24% - 29%) 0.863 OrHighNotLow 736.08 (4.6%) 731.29 (4.4%) -0.7% ( -9% - 8%) 0.645 PKLookup 277.80 (4.1%) 276.28 (4.9%) -0.5% ( -9% - 8%) 0.702 OrNotHighHigh 505.64 (3.7%) 503.51 (3.3%) -0.4% ( -7% - 6%) 0.701 Fuzzy2 85.81 (4.3%) 85.49 (7.0%) -0.4% ( -11% - 11%) 0.837 BrowseDayOfYearSSDVFacets 14.25 (6.6%) 14.20 (6.0%) -0.4% ( -12% - 13%) 0.853 AndHighMed 123.40 (3.7%) 122.95 (3.0%) -0.4% ( -6% - 6%) 0.729 HighPhrase 105.62 (2.9%) 105.25 (3.0%) -0.4% ( -6% - 5%) 0.704 HighSloppyPhrase 25.14 (5.8%) 25.06 (5.8%) -0.3% ( -11% - 11%) 0.862 OrHighNotHigh 488.98 (3.4%) 487.73 (3.4%) -0.3% ( -6% - 6%) 0.812 MedTerm 2015.74 (3.9%) 2010.82 (3.7%) -0.2% ( -7% - 7%) 0.839 AndHighHigh 83.67 (3.7%) 83.48 (3.0%) -0.2% ( -6% - 6%) 0.831 IntNRQ 101.13 (6.3%) 100.91 (6.3%) -0.2% ( -12% - 13%) 0.914 LowPhrase 39.11 (3.0%) 39.03 (3.3%) -0.2% ( -6% - 6%) 0.845 MedSloppyPhrase 106.20 (2.6%) 106.07 (2.4%) -0.1% ( -5% - 5%) 0.875 Fuzzy1 79.48 (7.4%) 79.42 (7.4%) -0.1% ( -13% - 15%) 0.974 BrowseDateTaxoFacets 19.43 (2.5%) 19.42 (2.2%) -0.1% ( -4% - 4%) 0.938 Respell 83.42 (7.4%) 83.45 (7.9%) 0.0% ( -14% - 16%) 0.987 HighTerm 819.26 (5.4%) 820.06 (4.9%) 0.1% ( -9% - 10%) 0.952 OrHighHigh 34.32 (4.6%) 34.36 (3.9%) 0.1% ( -8% - 9%) 0.940 OrHighMed 92.58 (3.8%) 92.67 (3.1%) 0.1% ( -6% - 7%) 0.924 OrHighNotMed 572.17 (4.5%) 572.87 (4.5%) 0.1% ( -8% - 9%) 0.931 BrowseMonthTaxoFacets 18.19 (2.6%) 18.24 (2.4%) 0.3% ( -4% - 5%) 0.709 BrowseDayOfYearTaxoFacets 14.46 (3.5%) 14.55 (3.0%) 0.7% ( -5% - 7%) 0.531 BrowseMonthSSDVFacets 20.07 (7.4%) 20.23 (9.7%) 0.8% ( -15% - 19%) 0.766 AndHighLow 1259.85 (4.0%) 1273.60 (3.8%) 1.1% ( -6% - 9%) 0.373 AndMedNotMonth 1075.74 (3.8%) 1172.98 (3.2%) 9.0% ( 1% - 16%) 0.000 AndMedFilterMonth 1140.24 (6.0%) 1289.34 (5.3%) 13.1% ( 1% - 25%) 0.000 AndHighFilterMonth 467.70 (3.7%) 596.35 (4.9%) 27.5% ( 18% - 37%) 0.000 AndHighNotMonth 364.62 (3.0%) 627.22 (7.8%) 72.0% ( 59% - 85%) 0.000 ``` Result 2: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Prefix3 378.42 (4.0%) 353.70 (3.3%) -6.5% ( -13% - 0%) 0.000 AndHighMedDayTaxoFacets 46.31 (2.7%) 43.95 (3.0%) -5.1% ( -10% - 0%) 0.000 TermDTSort 195.09 (4.0%) 186.41 (4.1%) -4.5% ( -12% - 3%) 0.001 HighTermTitleSort 110.04 (3.9%) 105.25 (4.3%) -4.4% ( -12% - 4%) 0.001 AndHighHighDayTaxoFacets 28.36 (2.3%) 27.13 (2.2%) -4.3% ( -8% - 0%) 0.000 Wildcard 220.25 (4.3%) 212.79 (3.8%) -3.4% ( -11% - 4%) 0.008 OrHighLow 430.09 (5.2%) 420.41 (6.3%) -2.3% ( -13% - 9%) 0.218 HighTermDayOfYearSort 322.71 (3.0%) 315.77 (3.1%) -2.2% ( -7% - 4%) 0.025 HighTermTitleBDVSort 31.31 (2.2%) 30.70 (1.6%) -2.0% ( -5% - 1%) 0.001 MedTermDayTaxoFacets 91.69 (3.4%) 89.93 (2.8%) -1.9% ( -7% - 4%) 0.050 OrHighNotLow 542.35 (6.1%) 532.91 (4.4%) -1.7% ( -11% - 9%) 0.301 OrHighMedDayTaxoFacets 8.60 (4.3%) 8.46 (3.7%) -1.7% ( -9% - 6%) 0.188 HighSpanNear 48.01 (1.6%) 47.32 (1.3%) -1.4% ( -4% - 1%) 0.002 OrHighNotMed 847.43 (4.6%) 835.55 (4.1%) -1.4% ( -9% - 7%) 0.310 MedTerm 962.30 (6.4%) 950.63 (5.4%) -1.2% ( -12% - 11%) 0.517 Fuzzy1 81.00 (6.5%) 80.08 (6.0%) -1.1% ( -12% - 12%) 0.566 OrNotHighHigh 552.98 (3.7%) 546.73 (3.0%) -1.1% ( -7% - 5%) 0.286 MedSpanNear 12.59 (0.8%) 12.45 (0.7%) -1.1% ( -2% - 0%) 0.000 LowTerm 1599.39 (6.3%) 1582.98 (3.8%) -1.0% ( -10% - 9%) 0.536 OrHighHigh 45.70 (4.6%) 45.27 (4.6%) -1.0% ( -9% - 8%) 0.512 OrNotHighLow 1445.93 (3.8%) 1433.84 (3.9%) -0.8% ( -8% - 7%) 0.494 OrHighNotHigh 561.36 (4.2%) 557.15 (3.0%) -0.7% ( -7% - 6%) 0.514 HighTermMonthSort 2730.23 (3.9%) 2711.82 (3.6%) -0.7% ( -7% - 7%) 0.566 MedIntervalsOrdered 80.30 (5.8%) 79.80 (5.8%) -0.6% ( -11% - 11%) 0.736 Fuzzy2 87.54 (6.0%) 87.09 (5.7%) -0.5% ( -11% - 11%) 0.780 HighIntervalsOrdered 6.28 (5.5%) 6.25 (5.5%) -0.5% ( -10% - 11%) 0.776 Respell 71.34 (4.4%) 71.00 (4.3%) -0.5% ( -8% - 8%) 0.727 MedPhrase 214.30 (4.9%) 213.27 (4.7%) -0.5% ( -9% - 9%) 0.754 LowIntervalsOrdered 10.88 (5.0%) 10.83 (5.1%) -0.5% ( -10% - 10%) 0.769 OrNotHighMed 503.93 (3.9%) 501.66 (3.0%) -0.5% ( -7% - 6%) 0.683 IntNRQ 53.53 (5.3%) 53.33 (4.6%) -0.4% ( -9% - 10%) 0.810 HighTerm 788.21 (6.7%) 785.26 (5.1%) -0.4% ( -11% - 12%) 0.843 LowPhrase 581.98 (4.8%) 579.89 (5.2%) -0.4% ( -9% - 10%) 0.822 OrHighMed 265.72 (3.6%) 264.97 (3.6%) -0.3% ( -7% - 7%) 0.803 PKLookup 281.97 (2.5%) 281.25 (2.3%) -0.3% ( -4% - 4%) 0.736 LowSpanNear 97.42 (1.4%) 97.18 (1.8%) -0.2% ( -3% - 2%) 0.620 MedSloppyPhrase 57.32 (3.4%) 57.19 (2.9%) -0.2% ( -6% - 6%) 0.816 HighSloppyPhrase 30.68 (2.5%) 30.63 (2.6%) -0.1% ( -5% - 5%) 0.865 LowSloppyPhrase 32.06 (2.8%) 32.02 (2.2%) -0.1% ( -5% - 5%) 0.871 BrowseMonthSSDVFacets 20.53 (4.6%) 20.52 (4.7%) -0.1% ( -9% - 9%) 0.956 BrowseDayOfYearSSDVFacets 14.53 (2.6%) 14.52 (2.5%) -0.0% ( -5% - 5%) 0.963 BrowseDayOfYearTaxoFacets 14.63 (2.8%) 14.66 (2.9%) 0.2% ( -5% - 6%) 0.808 BrowseDateTaxoFacets 19.45 (2.4%) 19.51 (2.2%) 0.3% ( -4% - 5%) 0.671 BrowseMonthTaxoFacets 18.23 (2.5%) 18.31 (2.5%) 0.4% ( -4% - 5%) 0.611 HighPhrase 161.02 (4.2%) 162.00 (3.8%) 0.6% ( -7% - 8%) 0.634 AndHighHigh 75.68 (3.6%) 76.22 (3.8%) 0.7% ( -6% - 8%) 0.540 AndHighLow 1653.05 (4.5%) 1665.61 (3.5%) 0.8% ( -6% - 9%) 0.550 AndHighMed 317.63 (4.7%) 321.08 (3.9%) 1.1% ( -7% - 10%) 0.428 BrowseDateSSDVFacets 5.47 (12.7%) 5.53 (14.0%) 1.1% ( -22% - 31%) 0.792 AndMedNotMonth 907.12 (5.8%) 1006.04 (5.5%) 10.9% ( 0% - 23%) 0.000 AndMedFilterMonth 826.00 (3.8%) 1014.21 (6.3%) 22.8% ( 12% - 34%) 0.000 AndHighFilterMonth 484.41 (3.8%) 618.97 (5.9%) 27.8% ( 17% - 39%) 0.000 AndHighNotMonth 76.61 (1.0%) 690.62 (48.2%) 801.4% ( 745% - 858%) 0.000 ``` #### Index without sorting Result 1: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Wildcard 35.96 (7.1%) 32.01 (4.6%) -11.0% ( -21% - 0%) 0.000 Prefix3 560.43 (3.0%) 526.43 (3.1%) -6.1% ( -11% - 0%) 0.000 AndMedFilterMonth 264.61 (2.4%) 249.82 (2.9%) -5.6% ( -10% - 0%) 0.000 AndHighMedDayTaxoFacets 61.28 (1.6%) 58.14 (3.0%) -5.1% ( -9% - 0%) 0.000 AndHighHighDayTaxoFacets 24.07 (2.2%) 22.87 (2.5%) -5.0% ( -9% - 0%) 0.000 HighTermTitleSort 229.96 (3.0%) 220.60 (3.1%) -4.1% ( -9% - 2%) 0.000 AndHighFilterMonth 294.96 (3.8%) 283.28 (4.5%) -4.0% ( -11% - 4%) 0.002 OrHighMedDayTaxoFacets 11.38 (4.1%) 10.93 (3.8%) -3.9% ( -11% - 4%) 0.002 HighTermDayOfYearSort 412.03 (3.4%) 397.78 (2.8%) -3.5% ( -9% - 2%) 0.000 TermDTSort 226.90 (2.8%) 219.54 (2.7%) -3.2% ( -8% - 2%) 0.000 MedTermDayTaxoFacets 60.64 (3.1%) 58.70 (2.5%) -3.2% ( -8% - 2%) 0.000 HighTermTitleBDVSort 21.84 (4.4%) 21.15 (4.5%) -3.2% ( -11% - 5%) 0.023 OrNotHighLow 1026.62 (3.2%) 994.09 (3.8%) -3.2% ( -9% - 3%) 0.004 OrNotHighMed 545.06 (2.5%) 530.44 (4.1%) -2.7% ( -8% - 3%) 0.012 HighIntervalsOrdered 44.88 (6.0%) 44.07 (5.3%) -1.8% ( -12% - 10%) 0.315 LowIntervalsOrdered 108.36 (3.1%) 106.48 (3.8%) -1.7% ( -8% - 5%) 0.110 LowSpanNear 49.51 (1.9%) 48.78 (2.0%) -1.5% ( -5% - 2%) 0.016 HighTermMonthSort 2328.38 (7.3%) 2294.68 (6.9%) -1.4% ( -14% - 13%) 0.518 Fuzzy1 150.33 (6.2%) 148.18 (6.3%) -1.4% ( -13% - 11%) 0.467 HighSpanNear 29.48 (1.2%) 29.09 (1.1%) -1.3% ( -3% - 1%) 0.000 MedIntervalsOrdered 179.17 (5.8%) 176.92 (5.5%) -1.3% ( -11% - 10%) 0.484 MedSpanNear 15.41 (2.1%) 15.24 (2.0%) -1.1% ( -5% - 3%) 0.086 OrHighNotHigh 525.87 (4.7%) 520.47 (3.8%) -1.0% ( -9% - 7%) 0.449 OrHighLow 567.25 (4.0%) 563.56 (4.5%) -0.7% ( -8% - 8%) 0.627 AndHighNotMonth 585.89 (5.8%) 582.21 (5.9%) -0.6% ( -11% - 11%) 0.736 LowTerm 1183.22 (5.6%) 1176.56 (4.9%) -0.6% ( -10% - 10%) 0.735 AndMedNotMonth 860.00 (4.3%) 855.40 (4.2%) -0.5% ( -8% - 8%) 0.689 Respell 73.80 (5.1%) 73.55 (5.1%) -0.3% ( -10% - 10%) 0.833 HighPhrase 80.32 (2.8%) 80.13 (3.1%) -0.2% ( -5% - 5%) 0.797 PKLookup 288.52 (4.4%) 287.84 (4.8%) -0.2% ( -8% - 9%) 0.869 OrNotHighHigh 512.10 (4.5%) 510.97 (3.9%) -0.2% ( -8% - 8%) 0.868 HighTerm 655.55 (7.1%) 654.16 (6.5%) -0.2% ( -12% - 14%) 0.922 MedTerm 1090.42 (6.0%) 1088.33 (5.3%) -0.2% ( -10% - 11%) 0.915 HighSloppyPhrase 8.83 (3.0%) 8.81 (3.2%) -0.1% ( -6% - 6%) 0.884 BrowseDateSSDVFacets 5.36 (6.7%) 5.36 (6.7%) -0.1% ( -12% - 14%) 0.956 BrowseDayOfYearTaxoFacets 45.49 (5.4%) 45.47 (5.3%) -0.0% ( -10% - 11%) 0.986 OrHighNotMed 602.62 (4.7%) 602.56 (4.8%) -0.0% ( -9% - 9%) 0.995 LowSloppyPhrase 13.54 (1.9%) 13.54 (1.9%) 0.0% ( -3% - 3%) 0.989 BrowseDateTaxoFacets 45.52 (5.4%) 45.53 (5.4%) 0.0% ( -10% - 11%) 0.992 MedSloppyPhrase 22.22 (2.2%) 22.23 (2.2%) 0.1% ( -4% - 4%) 0.941 BrowseMonthTaxoFacets 37.52 (24.4%) 37.56 (24.5%) 0.1% ( -39% - 64%) 0.987 LowPhrase 81.07 (1.9%) 81.19 (2.2%) 0.1% ( -3% - 4%) 0.826 BrowseDayOfYearSSDVFacets 26.54 (27.0%) 26.59 (26.7%) 0.2% ( -42% - 73%) 0.982 BrowseMonthSSDVFacets 25.92 (26.6%) 25.98 (26.2%) 0.2% ( -41% - 72%) 0.977 MedPhrase 128.80 (1.7%) 129.15 (2.0%) 0.3% ( -3% - 4%) 0.635 OrHighHigh 49.81 (3.4%) 49.95 (3.6%) 0.3% ( -6% - 7%) 0.787 AndHighHigh 78.84 (3.7%) 79.09 (3.8%) 0.3% ( -6% - 8%) 0.784 OrHighNotLow 729.06 (5.0%) 732.65 (5.2%) 0.5% ( -9% - 11%) 0.761 OrHighMed 117.86 (3.2%) 118.47 (3.2%) 0.5% ( -5% - 7%) 0.611 Fuzzy2 104.46 (5.4%) 105.20 (5.9%) 0.7% ( -10% - 12%) 0.690 IntNRQ 87.18 (2.2%) 87.88 (1.2%) 0.8% ( -2% - 4%) 0.142 AndHighLow 1166.89 (3.4%) 1183.39 (3.3%) 1.4% ( -5% - 8%) 0.179 AndHighMed 367.61 (4.9%) 373.01 (5.2%) 1.5% ( -8% - 12%) 0.355 ``` Result 2: ``` TaskQPS baseline StdDevQPS my_modified_version StdDev Pct diff p-value Prefix3 147.12 (6.2%) 129.27 (4.0%) -12.1% ( -21% - -2%) 0.000 AndMedFilterMonth 271.72 (3.3%) 257.32 (3.4%) -5.3% ( -11% - 1%) 0.000 AndHighMedDayTaxoFacets 142.77 (2.1%) 135.63 (3.0%) -5.0% ( -9% - 0%) 0.000 HighTermTitleSort 202.56 (2.2%) 194.53 (3.0%) -4.0% ( -8% - 1%) 0.000 AndHighHighDayTaxoFacets 39.70 (0.9%) 38.17 (1.2%) -3.8% ( -5% - -1%) 0.000 AndHighFilterMonth 207.32 (4.1%) 199.45 (3.7%) -3.8% ( -11% - 4%) 0.002 HighTermDayOfYearSort 428.44 (5.8%) 413.79 (5.2%) -3.4% ( -13% - 8%) 0.050 OrHighMedDayTaxoFacets 7.64 (9.2%) 7.41 (7.3%) -3.0% ( -17% - 14%) 0.249 TermDTSort 194.56 (4.1%) 188.70 (3.8%) -3.0% ( -10% - 5%) 0.016 MedTermDayTaxoFacets 93.00 (1.7%) 90.24 (1.3%) -3.0% ( -5% - 0%) 0.000 BrowseDayOfYearTaxoFacets 46.35 (0.6%) 45.27 (8.7%) -2.3% ( -11% - 7%) 0.234 OrNotHighMed 696.99 (3.3%) 681.56 (3.0%) -2.2% ( -8% - 4%) 0.027 Fuzzy1 96.77 (7.0%) 94.90 (8.2%) -1.9% ( -16% - 14%) 0.423 BrowseDateSSDVFacets 5.37 (6.3%) 5.27 (8.8%) -1.8% ( -15% - 14%) 0.449 HighTermMonthSort 2656.62 (5.9%) 2610.82 (5.8%) -1.7% ( -12% - 10%) 0.352 MedSpanNear 177.92 (1.9%) 174.98 (2.0%) -1.7% ( -5% - 2%) 0.008 OrHighNotMed 637.79 (4.7%) 627.39 (4.7%) -1.6% ( -10% - 8%) 0.274 HighTermTitleBDVSort 35.13 (1.6%) 34.58 (1.9%) -1.6% ( -5% - 1%) 0.005 AndHighNotMonth 602.88 (6.3%) 593.45 (6.2%) -1.6% ( -13% - 11%) 0.426 HighSpanNear 6.98 (1.4%) 6.88 (1.5%) -1.4% ( -4% - 1%) 0.002 OrHighNotHigh 442.50 (4.3%) 437.29 (4.1%) -1.2% ( -9% - 7%) 0.380 OrNotHighHigh 634.69 (4.0%) 627.34 (4.8%) -1.2% ( -9% - 8%) 0.409 OrHighNotLow 518.26 (4.7%) 513.28 (5.1%) -1.0% ( -10% - 9%) 0.537 LowSpanNear 16.59 (1.2%) 16.44 (0.9%) -0.9% ( -2% - 1%) 0.004 BrowseMonthTaxoFacets 42.47 (18.0%) 42.10 (19.3%) -0.9% ( -32% - 44%) 0.881 LowIntervalsOrdered 196.39 (6.2%) 194.69 (6.6%) -0.9% ( -12% - 12%) 0.669 OrNotHighLow 1353.91 (4.2%) 1342.18 (4.3%) -0.9% ( -9% - 8%) 0.522 AndMedNotMonth 975.83 (6.2%) 967.70 (5.7%) -0.8% ( -11% - 11%) 0.658 HighIntervalsOrdered 26.70 (7.2%) 26.48 (7.5%) -0.8% ( -14% - 15%) 0.729 LowTerm 1227.46 (5.2%) 1217.96 (5.9%) -0.8% ( -11% - 10%) 0.658 Wildcard 314.75 (5.1%) 312.45 (5.1%) -0.7% ( -10% - 9%) 0.648 LowPhrase 110.34 (5.4%) 109.54 (5.4%) -0.7% ( -10% - 10%) 0.674 MedIntervalsOrdered 66.86 (10.3%) 66.39 (10.9%) -0.7% ( -19% - 22%) 0.834 BrowseDateTaxoFacets 45.81 (5.6%) 45.50 (8.6%) -0.7% ( -14% - 14%) 0.766 AndHighMed 354.45 (4.5%) 352.40 (4.9%) -0.6% ( -9% - 9%) 0.697 AndHighHigh 91.95 (3.7%) 91.49 (3.6%) -0.5% ( -7% - 7%) 0.665 Fuzzy2 83.49 (7.0%) 83.14 (7.1%) -0.4% ( -13% - 14%) 0.852 OrHighMed 203.18 (3.6%) 202.61 (3.1%) -0.3% ( -6% - 6%) 0.793 IntNRQ 100.90 (11.1%) 100.67 (11.0%) -0.2% ( -20% - 24%) 0.948 MedPhrase 22.39 (2.7%) 22.34 (2.7%) -0.2% ( -5% - 5%) 0.804 HighPhrase 324.29 (3.2%) 323.68 (2.5%) -0.2% ( -5% - 5%) 0.838 PKLookup 293.51 (3.5%) 293.02 (3.3%) -0.2% ( -6% - 6%) 0.878 HighTerm 1125.52 (4.8%) 1124.54 (5.0%) -0.1% ( -9% - 10%) 0.955 OrHighHigh 70.53 (5.3%) 70.47 (5.6%) -0.1% ( -10% - 11%) 0.965 BrowseMonthSSDVFacets 27.34 (30.0%) 27.32 (29.9%) -0.1% ( -46% - 85%) 0.994 Respell 75.69 (5.6%) 75.65 (5.5%) -0.1% ( -10% - 11%) 0.975 BrowseDayOfYearSSDVFacets 27.94 (29.7%) 27.93 (29.8%) -0.0% ( -45% - 84%) 0.997 MedSloppyPhrase 5.29 (4.0%) 5.29 (3.9%) -0.0% ( -7% - 8%) 0.997 HighSloppyPhrase 41.15 (2.4%) 41.18 (2.8%) 0.1% ( -4% - 5%) 0.932 AndHighLow 3122.83 (5.9%) 3125.14 (5.3%) 0.1% ( -10% - 11%) 0.966 LowSloppyPhrase 36.10 (2.4%) 36.17 (2.4%) 0.2% ( -4% - 5%) 0.808 OrHighLow 426.69 (5.1%) 428.21 (4.9%) 0.4% ( -9% - 10%) 0.820 MedTerm 909.25 (5.3%) 912.72 (4.8%) 0.4% ( -9% - 11%) 0.811 ``` `AndMedFilterMonth` is now getting modest improvement with the same sorted index and code. So as expected, the index and query also play a big part in the benchmark results as well. --- With regard to the bulk scorer approach, it is interesting! However, if we were to split the window based on certain size and only call `peexNextNonMatchingDocID` when advancing to a new window, I felt it might not be as effective, since for unsorted index, the matching doc ids might have large gaps between ids to begin with? My implementation above would temporarily pause calling `peexNextNonMatchingDocID` for the next 128 docs encountered, so it should be more aggressive in terms of skipping the calls. -- 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