[
https://issues.apache.org/jira/browse/LUCENE-4690?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13556333#comment-13556333
]
Yonik Seeley commented on LUCENE-4690:
--------------------------------------
bq. Curious, how do you know that and/or measure that?
The number of cycles? It's all documented in various places. Of course one
needs a good sense of what assembly a compiler/hotspot will emit.
Integer multiply has been 3 cycles for quite a while for both Intel and AMD,
and shifts have been a single cycle (after the ill-fated P4).
http://gmplib.org/~tege/x86-timing.pdf
http://www.agner.org/optimize/instruction_tables.pdf
> remove hashing from NumericUtils.*ToPrefixCoded
> ------------------------------------------------
>
> Key: LUCENE-4690
> URL: https://issues.apache.org/jira/browse/LUCENE-4690
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Yonik Seeley
>
> As far as I can tell nothing actually uses the hash codes generated by these
> methods (not even any tests). If someone did want to generate a hash, it
> would be just as fast to do it on the BytesRef after the fact (or even faster
> from the input number itself).
--
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: [email protected]
For additional commands, e-mail: [email protected]
