[jira] [Commented] (DRILL-1525) Use XXHash algorithm for hash functions

["Parth Chandra (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/DRILL-1525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14171677#comment-14171677 ] 

Parth Chandra commented on DRILL-1525:
--------------------------------------

Not sure how the hash32(long val, int seed) and hash32(int val, int seed) were derived. If we take the hash64 function and pass in a long/int sized buffer to the function, we will get, almost the same code as in these functions, but some initialization of the h64 variable is missing - 

  h64 = seed + PRIME64_5;
  h64 += len;

Also, xxhash.c has an implementation to compute a 32 bit hash. I would think that it might be better to use that (though maybe a tad bit slower) instead of computing the 64 bit hash and truncating it.

> Use XXHash algorithm for hash functions
> ---------------------------------------
>
>                 Key: DRILL-1525
>                 URL: https://issues.apache.org/jira/browse/DRILL-1525
>             Project: Apache Drill
>          Issue Type: Bug
>            Reporter: Mehant Baid
>            Assignee: Mehant Baid
>             Fix For: 0.7.0
>
>         Attachments: DRILL-1525.patch
>
>
> Currently we are using murmur hash algorithm for performing hash operations. We would like to move to XXHash which is a fast non-cryptographic  hash algorithm. More information can be found here: https://code.google.com/p/xxhash/
> The implementation should also eliminates an extra copy of the ByteBuf to be consumed by the hash function.



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