[
https://issues.apache.org/jira/browse/LUCENE-5236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13793772#comment-13793772
]
Adrien Grand commented on LUCENE-5236:
--------------------------------------
bq. /** @return See {@link EliasFanoEncoder#sufficientlySmallerThanBitSet(long,
long)} */
We try to put text rather than just javadoc tags in javadocs (ie. "Return X"
rather than "@return X") so that the method summary doesn't appear empty on the
class-level javadocs.
bq. private static long numLongsForBits(long numBits) { // Note: int version in
FixedBitSet.bits2words()
I think you can directly use OpenBitSet.bits2words here? It already works with
longs.
There are some tabs mixed up with spaces and a broken javadoc link (see {{ant
precommit}}) but the patch otherwise looks really good!
{quote}
With index and broadword bit selection on my 32 bit machine, worst case
performance for load factor -1 (1/10) for any advance(), relative to
FixedBitSet, is -3.42 (2 log scale) for advance(3571).
The other advance() cases have better worst cases, so broadword bit selection
really helps there, too.
{quote}
Wow! This looks excellent! I will run the benchmark on my 64-bits machine
tomorrow to see if I get similar results.
> 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,
> 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]
