BTW, I've been thinking lately about the inline CALL setup (even did a
LITTLE side code feasibility work,) and I think that's probably the way
to go for a simple 1st pass dynamic translation. I just haven't had the
chance to put any real work into it yet.
GREG
> On Wed, 2003-11-12 at 01:02, Johan Rydberg wrote:
> > -- Kevin Lawton wrote:
> >
> > Oops, only sent that last message to you. You can forward
> > if you want.
> >
> > -Kevin
> >
> > Begin forwarded message:
> >
> > Date: Sat, 8 Nov 2003 15:14:38 -0800 (PST)
> > From: Kevin Lawton <kevinlawton2001@xxxxxxxxx>
> > To: Johan Rydberg <jrydberg@xxxxxxxxxxxxxxxxx>
> > Subject: Re: [Bochs-developers] Dynamic code translation
> >
> > --- Johan Rydberg <jrydberg@xxxxxxxxxxxxxxxxx> wrote:
> >
> > > I'm sure John means "Threaded Code" as described by Bell in:
> > >
> > > http://portal.acm.org/citation.cfm?id=362270&jmp=abstract&dl=GUIDE&dl=ACM
> > >
> > > A technique commenly used in virtual machines and simulators such
> > > as SimICS or g88.
> >
> > Fwiw, I played with a beta threaded interpreter in bochs some
> > time ago using this style. After thinking about it a little I wondered
> > if it wouldn't be better to just use inline CALL instructions. As
> > it turns out on x86 as a host anyways, the CALL/RET mechanism is
> > extremely optimized and faster than indirect branches which don't
> > have the RET overhead.
> >
> > What I would do to make a really simple JIT for bochs is to just
> > generate CALL instructions and some inline data. This would be
> > easy to do for multiple host architectures, get a decent amount
> > of speed increase over the current bochs and be a lot more
> > manageable. It's certainly be a good 1st effort. Just
> > save/restore registers before entering tcode fragments and
> > forget about register scheduling. Just my 2c anyways.
> >
> > All those indirect branches in the "threaded code" interpreter
> > (in those Virtutech papers) kill performance.
> >
> > -Kevin
> >
> > __________________________________
> > Do you Yahoo!?
> > Protect your identity with Yahoo! Mail AddressGuard
> > http://antispam.yahoo.com/whatsnewfree
> >
>
>
-------------------------------------------------------
This SF. Net email is sponsored by: GoToMyPC
GoToMyPC is the fast, easy and secure way to access your computer from
any Web browser or wireless device. Click here to Try it Free!
https://www.gotomypc.com/tr/OSDN/AW/Q4_2003/t/g22lp?Target=mm/g22lp.tmpl
|