harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Simon Chow" <simon.harm...@gmail.com>
Subject Re: [drlvm][gc] difference between compiled code and object data
Date Mon, 25 Feb 2008 03:19:35 GMT
Egor, thanks for your answer!
So the memory for compiled code can be allocated by PoolManager, is it true?
And I do not find the routines for releasing those memory, does it means
that the compiled code is always there until VM is shutdown?

On 21 Feb 2008 15:36:17 +0300, Egor Pasko <egor.pasko@gmail.com> wrote:
>
> On the 0x3F1 day of Apache Harmony Xiao-Feng Li wrote:
> > On Thu, Feb 21, 2008 at 11:17 AM, Simon Chow <simon.harmony@gmail.com>
> wrote:
> > > Thanks. It is every helpful!
> > >  So how does Harmony manage JIT generated code at runtime? Is there
> some
> > >  consideration for code locality improvement?
> > >
> >
> > Simon, let the JIT people to answer the question.
>
>
> Simon, within each method Jitrino performs some code layout
> optimizations. But Jitrino does not bother with cross-method code
> locality, just asks VM to allocate code with
> method_allocate_code_block().
>
> Does that answer your question?
>
>
> > For your reference,
> > please check the paper below:
> > Dynamic code management: improving whole program code locality in
> > managed runtimes. by Xianglong Huang, Brian T. Lewis, Kathryn S.
> > McKinley: VEE 2006: 133-143.
> >
> > >
> > >  On 21/02/2008, Xiao-Feng Li <xiaofeng.li@gmail.com> wrote:
> > >  >
> > >  > Simon, currently, Harmony manages VM data and Java heap separately.
> > >  > Java Heap is a specific term for the space that is managed by GC
> for
> > >  > Java objects. VM data such as VM data structures and  JIT generated
> > >  > code have their respective management, not managed by GC. They are
> > >  > interacting in certain contexts, such as class loading/unloading,
> > >  > etc., where native data have pointers to Java objects or the
> reverse.
> > >  > But basically they are independent at the moment.
> > >  >
> > >  > Thanks,
> > >  > xiaofeng
> > >  >
> > >  >
> > >  > On Thu, Feb 21, 2008 at 9:57 AM, Simon Chow <
> simon.harmony@gmail.com>
> > >  > wrote:
> > >  > > I find the compiled code produced by JIT is placed in the JVM
> heap space
> > >  > >  (0xA0000000~0xB0000000 by default) as well as object data, is
> that
> > >  > true?
> > >  > >  What's the difference of memory management policies used for
> object
> > >  > data and
> > >  > >  compiled code?
> > >  > >
> > >  > >
> > >  > >  --
> > >  > >  From : Simon.Chow@Software School of Fudan University
> > >  > >
> > >  >
> > >  >
> > >  >
> > >  >
> > >  > --
> > >  > http://xiao-feng.blogspot.com
> > >  >
> > >
> > >
> > >
> > >  --
> > >
> > >
> > > From : Simon.Chow@Software School of Fudan University
> > >
> >
> >
> >
> > --
> > http://xiao-feng.blogspot.com
> >
>
>
> --
>
> Egor Pasko
>
>


-- 
>From : Simon.Chow@Software School of Fudan University

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message