lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael McCandless (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LUCENE-4577) Nuke TFIDFSim's cache
Date Wed, 28 Nov 2012 21:53:58 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-4577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13505938#comment-13505938
] 

Michael McCandless commented on LUCENE-4577:
--------------------------------------------

+1

I tested patch on 10M Wikipedia index:

{noformat}
                    Task    QPS base      StdDev    QPS comp      StdDev                Pct
diff
                HighTerm       27.63     (18.2%)       25.45      (7.9%)   -7.9% ( -28% -
  22%)
                 MedTerm      226.33     (16.0%)      211.09      (7.8%)   -6.7% ( -26% -
  20%)
               MedPhrase        7.14      (5.4%)        6.94      (4.4%)   -2.8% ( -11% -
   7%)
              HighPhrase        4.79      (5.0%)        4.66      (5.5%)   -2.7% ( -12% -
   8%)
                 LowTerm      505.56      (7.8%)      492.74      (5.7%)   -2.5% ( -14% -
  11%)
               LowPhrase        9.60      (3.6%)        9.39      (3.4%)   -2.2% (  -8% -
   4%)
         MedSloppyPhrase       29.15      (3.3%)       28.72      (2.3%)   -1.5% (  -6% -
   4%)
        HighSloppyPhrase        4.05      (5.0%)        4.00      (3.4%)   -1.3% (  -9% -
   7%)
         LowSloppyPhrase        9.15      (4.1%)        9.07      (2.9%)   -0.9% (  -7% -
   6%)
              AndHighMed       86.43      (1.2%)       85.99      (1.6%)   -0.5% (  -3% -
   2%)
                PKLookup      197.83      (1.3%)      197.12      (1.1%)   -0.4% (  -2% -
   2%)
             AndHighHigh       25.58      (1.5%)       25.53      (2.2%)   -0.2% (  -3% -
   3%)
              AndHighLow      657.57      (1.0%)      658.40      (1.5%)    0.1% (  -2% -
   2%)
             LowSpanNear       11.99      (2.1%)       12.02      (2.2%)    0.2% (  -3% -
   4%)
                 Prefix3       17.52      (5.7%)       17.60      (5.3%)    0.4% ( -10% -
  12%)
            HighSpanNear        5.64      (2.4%)        5.66      (3.5%)    0.5% (  -5% -
   6%)
             MedSpanNear       12.44      (2.9%)       12.52      (2.7%)    0.6% (  -4% -
   6%)
                Wildcard       91.63      (3.8%)       92.33      (3.3%)    0.8% (  -6% -
   8%)
                 Respell       82.47      (2.9%)       83.58      (2.2%)    1.3% (  -3% -
   6%)
                  IntNRQ       10.65     (10.7%)       10.79     (10.8%)    1.3% ( -18% -
  25%)
                  Fuzzy2       67.84      (3.0%)       69.46      (2.7%)    2.4% (  -3% -
   8%)
                  Fuzzy1       76.30      (2.4%)       78.46      (2.2%)    2.8% (  -1% -
   7%)
              OrHighHigh        9.30      (6.7%)        9.96      (8.6%)    7.2% (  -7% -
  24%)
               OrHighMed       41.01      (7.3%)       44.63      (9.3%)    8.8% (  -7% -
  27%)
               OrHighLow       25.19      (7.8%)       27.45      (9.9%)    9.0% (  -8% -
  28%)
{noformat}


                
> Nuke TFIDFSim's cache
> ---------------------
>
>                 Key: LUCENE-4577
>                 URL: https://issues.apache.org/jira/browse/LUCENE-4577
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Robert Muir
>         Attachments: LUCENE-4577.patch
>
>
> This is the old termscorer cache. 
> This helps nothing, and maybe hurts: I removed it and here are the results:
> {noformat}
> Chart saved to out.png... (wd: /home/rmuir/workspace/lucene-trunk/lucene/benchmark)
>                     Task    QPS base      StdDev   QPS patch      StdDev            
   Pct diff
>              TermGroup1M       52.87      (2.2%)       52.62      (2.4%)   -0.5% (  -4%
-    4%)
>               AndHighMed       34.82      (2.8%)       34.70      (3.6%)   -0.3% (  -6%
-    6%)
>                 SpanNear        6.28      (5.3%)        6.26      (3.9%)   -0.2% (  -8%
-    9%)
>                   IntNRQ       13.24     (11.0%)       13.24      (9.9%)    0.0% ( -18%
-   23%)
>                  Prefix3       42.19      (7.6%)       42.21      (7.0%)    0.1% ( -13%
-   15%)
>                 Wildcard       36.90      (6.8%)       37.02      (5.9%)    0.3% ( -11%
-   13%)
>              AndHighHigh       25.68      (4.5%)       25.79      (3.2%)    0.5% (  -6%
-    8%)
>                   Phrase        9.28      (4.7%)        9.35      (4.4%)    0.7% (  -8%
-   10%)
>             TermBGroup1M       45.76      (6.3%)       46.10      (3.2%)    0.7% (  -8%
-   10%)
>             SloppyPhrase       10.25      (3.9%)       10.33      (4.4%)    0.8% (  -7%
-    9%)
>               OrHighHigh        8.87      (6.4%)        8.97      (6.7%)    1.1% ( -11%
-   15%)
>                   Fuzzy1       70.28      (4.3%)       71.24      (7.1%)    1.4% (  -9%
-   13%)
>                OrHighMed       10.70      (7.0%)       10.86      (6.4%)    1.5% ( -11%
-   15%)
>                   Fuzzy2       27.79      (6.1%)       28.31      (5.1%)    1.9% (  -8%
-   13%)
>                  Respell       71.72      (6.8%)       73.39      (3.7%)    2.3% (  -7%
-   13%)
>                     Term      209.49      (4.4%)      214.58      (3.7%)    2.4% (  -5%
-   11%)
>           TermBGroup1M1P        7.10      (5.1%)        7.48      (7.8%)    5.3% (  -7%
-   19%)
> {noformat}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

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


Mime
View raw message