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-5554) Add TermBulkScorer
Date Thu, 27 Mar 2014 16:05:15 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-5554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13949484#comment-13949484
] 

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

bq. I am also unsure if the current "performance degradation" is real at all. Its a benchmark
and the methods in question are compiled with OSR. It seems like it may just be a benchmark
issue.

We should improve the luceneutil benchmark if we can, but just because hotspot chose to use
OSR doesn't mean the benchmark is a microbenchmark (the reverse is often true I guess).

The benchmark takes a large task file (many different queries in each of the categories),
randomly picks one query per category, executes each query 100 times, but shuffles the order
that all these queries will run, and runs with 3 threads.

I think there are clear improvements we could make, e.g. using "real world" queries (these
queries were generated by walking the index to find "good"/hard queries; the source for that
is CreateQueries.java in luceneutil), also do faceting/grouping/joining/sort by field/highlighting
as well, etc.  Basically make it closer to a real search app ...

But net/net I don't think this is a microbenchmark; I think it's a fair test.  Please don't
shoot the messenger :)

If anyone else has different benchmarks to try, please do so!  Even just running luceneutil
on your env is great because maybe my env has some strange behavior or something...

> 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, 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


Mime
View raw message