harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mikhail Fursov (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HARMONY-5127) [drlvm][opt] Performance improvement for integer-to-float conversion
Date Wed, 20 Feb 2008 13:57:46 GMT

     [ https://issues.apache.org/jira/browse/HARMONY-5127?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Mikhail Fursov updated HARMONY-5127:

    Attachment: int2float.diff

int2float.diff: the fix for the problem
both helper-based and SSE based version of conversions supported.

Use -XX:jit.arg.hir2lir.useInternalHelpersForInteger2FloatConv=true to enable old mode.

I'm going to write add some boundary tests before committing the patch.

> [drlvm][opt] Performance improvement for integer-to-float conversion
> --------------------------------------------------------------------
>                 Key: HARMONY-5127
>                 URL: https://issues.apache.org/jira/browse/HARMONY-5127
>             Project: Harmony
>          Issue Type: Improvement
>          Components: DRLVM
>         Environment: All
>            Reporter: Evgueni Brevnov
>            Assignee: Mikhail Fursov
>            Priority: Minor
>         Attachments: int2float.diff
> Currently jitrino OPT optimizes float-to-integer conversion by injecting (inlining) required
code into managed method. Unfortunatelly it lacks such optimization for opposite (integer-to-float)
conversion. Such optimization may have noticable effect on SciMark workload.
> I also noticed that DRLVM helper VM_RT_F2I generates wrapper code (LIL stub) around C
function which implements float-to-integer conversion. It makes sense to get rid of such wrapper
and call C function directly. (or remove this helper at all and let jitrino implement it)
> Thanks
> Evgueni

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

View raw message