[
https://issues.apache.org/jira/browse/HADOOP-10838?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14066428#comment-14066428
]
Todd Lipcon commented on HADOOP-10838:
--------------------------------------
Cool, nice CPU improvement!
A couple small nits remaining, then I think this is good to go:
- like Colin suggested above, you don't need the wrapper function that calls
through to the native function
- the indentation on the inner loop is one level too deep
- this would infinite loop if bytesPerChecksum > 1MB. We probably should add a
MAX(1, ...) in the calculation of checksumsPerMB, even if it means we might go
over our target in this edge case.
> Byte array native checksumming
> ------------------------------
>
> Key: HADOOP-10838
> URL: https://issues.apache.org/jira/browse/HADOOP-10838
> Project: Hadoop Common
> Issue Type: Improvement
> Components: performance
> Reporter: James Thomas
> Assignee: James Thomas
> Attachments: HADOOP-10838.2.patch, HADOOP-10838.patch,
> HDFS-3528.patch, HDFS-6560.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
