You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Marshall Schor (JIRA)" <de...@uima.apache.org> on 2015/03/11 16:40:39 UTC

[jira] [Resolved] (UIMA-4283) tuning Binary Compressed Cas

     [ https://issues.apache.org/jira/browse/UIMA-4283?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Marshall Schor resolved UIMA-4283.
----------------------------------
    Resolution: Fixed
      Assignee: Marshall Schor

the compress part already had reduced buffer sizes, so the only change made was to reduce the 32k zip buffer for decompression to 1k for both form 4 and 6

> tuning Binary Compressed Cas
> ----------------------------
>
>                 Key: UIMA-4283
>                 URL: https://issues.apache.org/jira/browse/UIMA-4283
>             Project: UIMA
>          Issue Type: Improvement
>          Components: Core Java Framework
>            Reporter: Marshall Schor
>            Assignee: Marshall Schor
>            Priority: Minor
>             Fix For: 2.7.1SDK
>
>
> Measurements are showing some tuning is needed in buffer sizes for the binary compressed serialization/deserialization code.  Reducing one buffer from a max of 32k to 1k showed a speed up in deserialization of more than 2.5 x using an IBM JVM on an Intil i7 with 128K L1; using a Oracle JVM - there was a minor speedup.  This effect appears to be due to L1 cache flushing/dumping. 
> Check and tune both form 4 and 6.



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