[
https://issues.apache.org/jira/browse/LUCENE-4022?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271312#comment-13271312
]
Simon Willnauer commented on LUCENE-4022:
-----------------------------------------
bq. How did you come up with the 10x factor though? Is it something off the top
of your head?
I wanted to differentiate between a significantly bigger "unallocated" heap to
force a grow if it makes sense so factor 10 seemed to be a good start. I mean
this automatic stuff should be a conservative default that gives you reasonable
performance. In the first place it should make sure your system is stable and
doesn't run into OOM etc. It might seem somewhat arbitrarily. I will add a
changes entry and commit this stuff. Seems like robert wants to roll a 3.6.1
soonish ;)
> Offline Sorter wrongly uses MIN_BUFFER_SIZE if there is more memory available
> -----------------------------------------------------------------------------
>
> Key: LUCENE-4022
> URL: https://issues.apache.org/jira/browse/LUCENE-4022
> Project: Lucene - Java
> Issue Type: Bug
> Components: modules/spellchecker
> Affects Versions: 3.6, 4.0
> Reporter: Simon Willnauer
> Assignee: Simon Willnauer
> Fix For: 4.0, 3.6.1
>
> Attachments: LUCENE-4022.patch
>
>
> The Sorter we use for offline sorting seems to use the MIN_BUFFER_SIZE as a
> upper bound even if there is more memory available. See this snippet:
> {code}
> long half = free/2;
> if (half >= ABSOLUTE_MIN_SORT_BUFFER_SIZE) {
> return new BufferSize(Math.min(MIN_BUFFER_SIZE_MB * MB, half));
> }
>
> // by max mem (heap will grow)
> half = (max - total) / 2;
> return new BufferSize(Math.min(MIN_BUFFER_SIZE_MB * MB, half));
> {code}
> use use use Math.max instead of min here.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
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]
