harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Weldon Washburn" <weldon...@gmail.com>
Subject Re: [drlvm][gc/threading] need the two lsb's from object header for MMTk port
Date Mon, 09 Oct 2006 17:29:48 GMT
On 10/9/06, Salikh Zakirov <Salikh.Zakirov@intel.com> wrote:
>
> Weldon Washburn wrote:
> > Artem,
> > I'd like to confirm what you said. For GC_BIT_MASK, it looks like the
> > bottom
> > two bits of byte number one is available for exclusive GC use.  If the
> GC
> > needs to alter these two bits while the mutator(s) are running, the GC
> > needs
> > to use CAS on the entire 32 bits of obj_info.  Does this sound right?
>
> Doing 32-bit CAS while mutators are running looks like a safe solution.
> However, it can degrade performance considerably.


Actually, I only need these two bits for the MMTk port.  This code should
not affect any performance other than MMTk.  And MMTk performance is not a
concern at this stage.


Can we instead consider locking the object monitor before modifying its
> header bits?
>
>
> ---------------------------------------------------------------------
> Terms of use : http://incubator.apache.org/harmony/mailing.html
> To unsubscribe, e-mail: harmony-dev-unsubscribe@incubator.apache.org
> For additional commands, e-mail: harmony-dev-help@incubator.apache.org
>
>


-- 
Weldon Washburn
Intel Middleware Products Division

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