[ 
https://issues.apache.org/jira/browse/LUCENE-5236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13793814#comment-13793814
 ] 

Paul Elschot commented on LUCENE-5236:
--------------------------------------

The second patch of 13 Oct 2013 improves the javadocs of EliasFanoDocIdSet, and 
adds the condition to sufficientlySmallerThanBitSet to always prefer a bit set 
when it uses no more than 4 longs.

ant validate
does not complain about tabs anymore here.

I could not find a broken link in the javadocs, maybe because I'm using java 
1.7:
ant documentation-lint
fails with the error message "Linting documentation HTML is not supported on 
this Java version (1.7)".
There is an http:// reference to archiv.org in the javadoc as plain text (it is 
not marked up as a link), and just now it worked fine as a link.

OpenBitSet.bits2words will fail when the number of bits gets just over the 
maximum int value, so I prefer not to use it for now.

I am looking forward to the benchmark results on a 64 bits machine.

In the decoder there is a test to use naive bit selection when the needed rank 
is less then or equal to 8, and otherwise use broadword bit selection.
I expect this constant 8 to be close to the optimal value on a current 64 bit 
machine, but some tuning may be needed.
See also my first comment of 11 July 2013 at LUCENE-5098.


> Use broadword bit selection in EliasFanoDecoder
> -----------------------------------------------
>
>                 Key: LUCENE-5236
>                 URL: https://issues.apache.org/jira/browse/LUCENE-5236
>             Project: Lucene - Core
>          Issue Type: Improvement
>            Reporter: Paul Elschot
>            Assignee: Adrien Grand
>            Priority: Minor
>         Attachments: LUCENE-5236.patch, LUCENE-5236.patch, LUCENE-5236.patch, 
> LUCENE-5236.patch, TestDocIdSetBenchmark.java
>
>
> Try and speed up decoding



--
This message was sent by Atlassian JIRA
(v6.1#6144)

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

Reply via email to