harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavel Ozhdikhin (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HARMONY-3911) [DRLVM][JIT] wrong array element size calculated for allocArray helper inlining
Date Thu, 24 May 2007 10:26:16 GMT

     [ https://issues.apache.org/jira/browse/HARMONY-3911?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Pavel Ozhdikhin updated HARMONY-3911:
-------------------------------------

    Attachment: harmony-3911.patch

The attached patch fixes the problem in helper inlining. Thanks to Mikhail Fursov for the
hint where to look for the problem first - it speeded up the work.

With the attached patch I was able to pass 10 iterations of dacapo.hsqldb. Dacapo.xalan failed
for me running into lack of C heap. The problem is in the new register allocator - it requires
too much memory for big methods. The issue is being investigated. We can either switch back
to the old regalloc or dynamically use it only for big methofs. However, with the attached
server.emconf (it uses old regalloc) dacapo.xalan passed 10 iterations as well.


> [DRLVM][JIT] wrong array element size calculated for allocArray helper inlining
> -------------------------------------------------------------------------------
>
>                 Key: HARMONY-3911
>                 URL: https://issues.apache.org/jira/browse/HARMONY-3911
>             Project: Harmony
>          Issue Type: Bug
>          Components: DRLVM
>         Environment: windows and linux
>            Reporter: Yu-Nan He
>         Attachments: harmony-3911.patch, server.emconf
>
>
> During the investigation of HARMONY-3899, I found the failure is caused by a wrong array
element size passed to GC. 
> When running
> java -Xmx1500M -Xms1500M -Xem:server -XX:vm.dlls=gc_gen.dll -XX:gc.use_large_page=true
-cp . Harness -n 10 hsqldb 
> in some iteration, allocArray helper is invoked to create an array of [Z type with length
of 13. But the element size is set to be 4 (for one Byte). Then the calculated array object
size is wrong. 
> It would be great if some JIT guru can help to fix it. Thanks.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message