[ 
https://issues.apache.org/jira/browse/LUCENE-6028?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Adrien Grand closed LUCENE-6028.
--------------------------------
    Resolution: Won't Fix

Wow this actually resulted in a massive slowdown:

{noformat}
                    TaskQPS baseline      StdDev   QPS patch      StdDev        
        Pct diff
               OrHighMed       25.54      (9.3%)       20.96      (3.6%)  
-17.9% ( -28% -   -5%)
              OrHighHigh       14.48      (8.9%)       12.10      (3.5%)  
-16.4% ( -26% -   -4%)
               OrHighLow       42.25     (10.8%)       35.33      (3.6%)  
-16.4% ( -27% -   -2%)
                  Fuzzy1       82.31      (9.0%)       81.73      (6.2%)   
-0.7% ( -14% -   15%)
                  Fuzzy2       84.84     (13.0%)       84.93     (13.9%)    
0.1% ( -23% -   31%)
            OrHighNotMed      122.51      (9.6%)      123.04      (5.0%)    
0.4% ( -12% -   16%)
             AndHighHigh       94.76      (7.0%)       95.49      (3.5%)    
0.8% (  -9% -   12%)
        HighSloppyPhrase       15.26     (16.1%)       15.40     (11.4%)    
0.9% ( -22% -   33%)
                  IntNRQ        7.75     (10.0%)        7.83      (5.5%)    
0.9% ( -13% -   18%)
         LowSloppyPhrase      117.30      (9.3%)      118.59      (3.6%)    
1.1% ( -10% -   15%)
            OrHighNotLow       35.89      (9.9%)       36.29      (5.7%)    
1.1% ( -13% -   18%)
                 Prefix3      170.95      (8.0%)      173.04      (4.2%)    
1.2% ( -10% -   14%)
                Wildcard       39.14      (9.4%)       39.63      (4.5%)    
1.2% ( -11% -   16%)
           OrHighNotHigh       38.28      (8.6%)       38.81      (4.5%)    
1.4% ( -10% -   15%)
               MedPhrase      110.04      (9.2%)      111.64      (3.1%)    
1.5% ( -10% -   15%)
            OrNotHighMed      198.34      (9.1%)      201.25      (3.7%)    
1.5% ( -10% -   15%)
             MedSpanNear        8.02      (9.4%)        8.14      (4.2%)    
1.5% ( -11% -   16%)
              HighPhrase       35.05      (8.6%)       35.62      (2.1%)    
1.6% (  -8% -   13%)
           OrNotHighHigh       65.17      (8.7%)       66.24      (3.6%)    
1.6% (  -9% -   15%)
                PKLookup      297.53      (8.6%)      302.41      (3.6%)    
1.6% (  -9% -   15%)
            OrNotHighLow     1286.49     (10.3%)     1310.19      (7.0%)    
1.8% ( -14% -   21%)
         MedSloppyPhrase       18.45     (11.1%)       18.81      (3.2%)    
1.9% ( -11% -   18%)
             LowSpanNear       10.14      (9.4%)       10.35      (3.2%)    
2.1% (  -9% -   16%)
               LowPhrase       24.20      (9.8%)       24.70      (2.2%)    
2.1% (  -9% -   15%)
                 Respell       79.70      (8.3%)       81.37      (4.0%)    
2.1% (  -9% -   15%)
              AndHighMed      207.24     (11.0%)      211.83      (3.2%)    
2.2% ( -10% -   18%)
                 LowTerm      743.41     (10.1%)      761.93      (6.5%)    
2.5% ( -12% -   21%)
              AndHighLow      836.05     (13.3%)      858.54      (6.7%)    
2.7% ( -15% -   26%)
                 MedTerm      235.88      (8.9%)      243.39     (10.0%)    
3.2% ( -14% -   24%)
            HighSpanNear       37.28     (11.7%)       38.71      (3.8%)    
3.8% ( -10% -   21%)
                HighTerm       83.60     (10.4%)       87.17     (11.8%)    
4.3% ( -16% -   29%)
{noformat}


> Cut over DisjunctionScorer to oal.util.PriorityQueue
> ----------------------------------------------------
>
>                 Key: LUCENE-6028
>                 URL: https://issues.apache.org/jira/browse/LUCENE-6028
>             Project: Lucene - Core
>          Issue Type: Task
>            Reporter: Adrien Grand
>            Assignee: Adrien Grand
>            Priority: Minor
>             Fix For: 5.0
>
>         Attachments: LUCENE-6028.patch
>
>
> DisjunctionScorer maintains its own implementation of a priority queue, I 
> think it should just use oal.util.PriorityQueue?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to