Re: [rvm-research] PowerPC baseline compiler: fence for volatile

[David P G. <gr...@us...>]

Lei Zhao <lei...@gm...> wrote on 01/15/2013 02:36:13 PM:

> From: Lei Zhao <lei...@gm...>
> To: "General discussion of Jikes RVM design, implementation, issues,
> and plans" <jik...@li...>,
> Date: 01/15/2013 02:38 PM
> Subject: Re: [rvm-research] PowerPC baseline compiler: fence for volatile
>
> Hi David,
>
>   Sorry, please ignore the lack of instruction argument at the
> bottom of my previous email. It doesn't make sense as I was thinking
> about opt compiler. Baseline compiler apparently has nothing to do
> with the IRs whatsoever.
>
>   Here is a new patch where the assembler was also modified so that
> hwsync is used wherever needed. I would call this a fix for JMM. Thanks.
>

Thanks Lei.

I created RVM-1012 to track this patch
(http://jira.codehaus.org/browse/RVM-1012).  When you get a chance, could
you add a comment to the JIRA with the contribution agreement
(http://jikesrvm.org/Contributions,
http://docs.codehaus.org/download/attachments/74078/single-contrib.txt?version=2&modificationDate=1243890146719).

--dave