I haven't looked at the code in depth, or even tried to run it for 
that matter.

The docs indicate that MMgc has the capability to call C++ object 
destructors for object finalization:

  http://developer.mozilla.org/en/docs/MMgc#GCFinalizedObject

which may explain why they chose to roll their own GC instead of
using Boehm, since Mozilla is written in C++.

Because MMgc is not (yet) threadsafe

  http://developer.mozilla.org/en/docs/MMgc#Threading

I would assume that collection/finalization would have to take place 
in the same thread during some sort of explicit yield or gc call.
But then again, my guesses are not that accurate. :-)

--- [EMAIL PROTECTED] wrote:
> They may have gone a custom route to avoid certain business relative  
> issues... Who knows what goes on in their slack times / meetings! From  
> what you can make of MMgc, does it look worthy?  Or is it young and  
> flightless?
> 
> Lee
> 
> Quoting Joe Wilson <[EMAIL PROTECTED]>:
> 
> > --- Joe Wilson <[EMAIL PROTECTED]> wrote:
> >> --- Matt Campbell <[EMAIL PROTECTED]> wrote:
> >> > I have taken a cursory look at the Tamarin source tree, and I observe
> >> > that Tamarin's GC is in a separate module called MMgc (the MM probably
> >> > stands for Macromedia).  MMgc is written in C++ and has a C++ API, but
> >> > it may still be a useful alternative to Boehm GC.  It begs the question
> >> > of why Macromedia/Adobe didn't use Boehm GC, given how long that package
> >> > has been around.
> >>
> >> I'm guessing that MMgc is a precise collector.
> >
> > ...and I guessed wrong - MMgc is a conservative mark/sweep collector:
> >
> >   http://developer.mozilla.org/en/docs/MMgc#Conservative_Collection
> >
> > go figure.



      
____________________________________________________________________________________
Fussy? Opinionated? Impossible to please? Perfect.  Join Yahoo!'s user panel 
and lay it on us. http://surveylink.yahoo.com/gmrs/yahoo_panel_invite.asp?a=7 


-- 
Neko : One VM to run them all
(http://nekovm.org)

Reply via email to