New implementation: Emacs Common Lisp

[lars-zgYzP9v7iJcdnm+ (Lars Brinkhoff)]

[Cc to mailing list, because of general interest.]

First thing I noticed is that it's unfortunately *extremely*
slow for the few tests I ran, so I'm not sure if I'll actually
be able to use it for anything other than simple tasks

Maybe not. Make sure
1. Emacs CL itself is compiled
2. your own code is compiled
for less slowness.

That helped, and it also helps to run tests that don't run into bignum
territory :-)
So it's looking very promising after all. Thanks, and sorry about the
silly mistake.

Oh, no problem at all. Also, my "maybe not" was meant as an
agreement, i.e. "yes, maybe it's not usable".

It's good to know it's not THAT bad after all. Bignum performace CAN
be improved a lot; my current implementation is really bad. And if
it's really important to you, you'll be glad to know that XEmacs 21.5
will include fast support for bignums (and ratios).

[Klaus Weidner]

Post by lars-zgYzP9v7iJcdnm+ (Lars Brinkhoff)

That helped, and it also helps to run tests that don't run into bignum
territory :-)

It's good to know it's not THAT bad after all. Bignum performace CAN
be improved a lot; my current implementation is really bad. And if
it's really important to you, you'll be glad to know that XEmacs 21.5
will include fast support for bignums (and ratios).

Well, I didn't really *need* the bignums, I just accidentally picked some
tests which exceeded the fixnum range. I'll see how it works with real
programs once I'm less overloaded.

Just having access to #+emacs conditional loading, lexical closures and
standard CL function names would already be enough for what I need - I'm
glad that I'm not a CLOS addict, that would make porting harder ;-)

-Klaus
P.S.: sorry, messed up my mail address again. I do think I've now
convinced Mutt to do this right from now on...

[lars-zgYzP9v7iJcdnm+ (Lars Brinkhoff)]

Post by Klaus Weidner
I didn't really *need* the bignums, I just accidentally picked some
tests which exceeded the fixnum range. I'll see how it works with
real programs once I'm less overloaded.

For now, a function written in Emacs Lisp will be faster than the
corresponding Common Lisp code. In particular, the compiler needs
more work to remove redundant multiple value bookkeeping. But I don't
think it's exceedingly bad.

I've made no attempt at all to make the interpreter fast, and it will
continue to be a very low priority.