You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Michael McCandless (JIRA)" <ji...@apache.org> on 2014/03/25 19:38:16 UTC

[jira] [Commented] (LUCENE-5554) Add TermBulkScorer

    [ https://issues.apache.org/jira/browse/LUCENE-5554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13946920#comment-13946920 ] 

Michael McCandless commented on LUCENE-5554:
--------------------------------------------

Perf of patch vs trunk:

{noformat}
                    Task    QPS base      StdDev    QPS comp      StdDev                Pct diff
              AndHighLow      394.80      (2.4%)      390.51      (2.8%)   -1.1% (  -6% -    4%)
             LowSpanNear       10.98      (3.9%)       10.88      (3.4%)   -0.8% (  -7% -    6%)
            HighSpanNear       10.03      (7.5%)        9.95      (6.0%)   -0.8% ( -13% -   13%)
             MedSpanNear       33.49      (3.7%)       33.28      (3.2%)   -0.6% (  -7% -    6%)
              AndHighMed       34.43      (1.7%)       34.24      (1.3%)   -0.5% (  -3% -    2%)
         LowSloppyPhrase       44.55      (1.8%)       44.32      (2.3%)   -0.5% (  -4% -    3%)
             AndHighHigh       28.39      (1.5%)       28.26      (1.3%)   -0.5% (  -3% -    2%)
                 Respell       48.70      (3.3%)       48.58      (3.4%)   -0.2% (  -6% -    6%)
               LowPhrase       13.15      (2.1%)       13.14      (1.9%)   -0.1% (  -4% -    3%)
         MedSloppyPhrase        3.53      (3.7%)        3.52      (4.9%)   -0.0% (  -8% -    8%)
        HighSloppyPhrase        3.44      (5.5%)        3.44      (9.0%)    0.0% ( -13% -   15%)
              HighPhrase        4.21      (6.7%)        4.27      (7.0%)    1.2% ( -11% -   15%)
               MedPhrase      208.99      (6.0%)      211.61      (6.8%)    1.3% ( -10% -   14%)
                Wildcard       19.08      (3.4%)       19.39      (3.0%)    1.6% (  -4% -    8%)
                 Prefix3       85.31      (3.6%)       87.89      (1.6%)    3.0% (  -2% -    8%)
                  IntNRQ        3.37      (5.0%)        3.47      (2.2%)    3.1% (  -3% -   10%)
                  Fuzzy2       46.34      (3.8%)       48.72      (2.6%)    5.1% (  -1% -   11%)
                  Fuzzy1       58.95      (4.3%)       62.40      (2.3%)    5.9% (   0% -   13%)
                 LowTerm      323.43      (5.1%)      352.16      (4.9%)    8.9% (  -1% -   19%)
                 MedTerm      100.56      (8.8%)      118.56     (11.2%)   17.9% (  -1% -   41%)
            OrHighNotMed       31.20      (6.0%)       38.53      (3.1%)   23.5% (  13% -   34%)
                HighTerm       64.41     (11.6%)       80.45     (17.2%)   24.9% (  -3% -   60%)
               OrHighMed       26.99      (6.3%)       33.75      (3.5%)   25.0% (  14% -   37%)
               OrHighLow       19.16      (6.4%)       24.24      (3.7%)   26.5% (  15% -   39%)
            OrHighNotLow       25.28      (6.3%)       31.98      (3.7%)   26.5% (  15% -   38%)
              OrHighHigh        8.36      (6.3%)       10.60      (3.5%)   26.8% (  16% -   39%)
           OrHighNotHigh       11.28      (6.3%)       14.33      (3.1%)   27.0% (  16% -   38%)
           OrNotHighHigh        9.02      (6.4%)       11.57      (3.4%)   28.2% (  17% -   40%)
            OrNotHighMed       19.35      (7.2%)       24.84      (2.3%)   28.4% (  17% -   40%)
            OrNotHighLow       20.01      (7.4%)       25.84      (2.3%)   29.1% (  18% -   42%)
{noformat}

> Add TermBulkScorer
> ------------------
>
>                 Key: LUCENE-5554
>                 URL: https://issues.apache.org/jira/browse/LUCENE-5554
>             Project: Lucene - Core
>          Issue Type: Improvement
>          Components: core/search
>            Reporter: Michael McCandless
>            Assignee: Michael McCandless
>             Fix For: 4.8, 5.0
>
>         Attachments: LUCENE-5554.patch
>
>
> Hotspot was unhappy with the changes in LUCENE-5487, e.g.: http://people.apache.org/~mikemccand/lucenebench/OrHighHigh.html
> But it looks like we can get the performance back by making a dedicated BulkScorer for TermQuery.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org