[
https://issues.apache.org/jira/browse/ORC-345?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16461667#comment-16461667
]
ASF GitHub Bot commented on ORC-345:
------------------------------------
Github user omalley commented on the issue:
https://github.com/apache/orc/pull/252
Ok, with this and ORC-344, the difference in writing small vs large
decimals to a null file system is huge:
Benchmark (precision) Mode Cnt Score Error Units
DecimalBench.write 8 avgt 30 36914.791 ± 866.408 us/op
DecimalBench.write 19 avgt 30 240789.318 ± 5146.880 us/op
So I'm seeing a 6x speed up when using precision = 8 with the new code path
instead of p = 19 and the old code path.
> Create a Decimal64StatisticsImpl
> --------------------------------
>
> Key: ORC-345
> URL: https://issues.apache.org/jira/browse/ORC-345
> Project: ORC
> Issue Type: Improvement
> Reporter: Owen O'Malley
> Assignee: Owen O'Malley
> Priority: Major
>
> We should create a fast path for handling decimal statistics where precision
> <= 18 where the values can be handled as longs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
