Re: Introducing Jitgen

[Andy Wingo <wingo@pobox.com>]

Heya Noah,

On Fri 10 Dec 2010 10:09, ludo@gnu.org (Ludovic Courtès) writes:

> Better late than never...

Indeed, and my apologies for only getting back to you now.

> Noah Lavine <noah.b.lavine@gmail.com> writes:
>
>> My biggest question is, is this something that you would be interested
>> in having in Guile?
>
> I think so, yes.  Not in 2.0, I guess, but when it starts stabilizing it
> could go into a branch.

I am skeptical.  I think that the jitgen language sounds like a fine
intermediate thing, but I don't know if it will be good enough for
describing the whole VM.

OTOH something like it does seem useful for implementing all of the VM
ops in the JIT engine.

Have you considered going the other way?  That's to say, writing a
limited C -> Jitgen parser, and thus keeping the primary representation
of the VM in C?  It doesn't have to be a full parser.  It doesn't have
to expand macros -- i.e. it can interpret NEXT() directly, without
relying on details of its expansion.  If there is an idiom that the
parser is having a hard time with, we can change the C VM implementation
to make the meanings more explicit, via more macro use.

In this way we would keep the VM's canonical representation in a
well-defined language.  This makes it easier for anyone to maintain and
fix the VM.  It also frees you up to make changes in jitgen without
worrying about impacting the non-jit case.

What do you think?

Regards,

Andy
-- 
http://wingolog.org/