Re: [Texmacs-dev] Efficiency of scheme implementation

[Joris van der Hoeven]

On Tue, 17 Feb 2004, David Allouche wrote:
> On Mon, Feb 16, 2004 at 05:05:57PM +0100, Joris van der Hoeven wrote:
> > Here are some (approximate) timings for (fib 35):
> >
> >     Guile           37 sec
> >     Scheme 48       17 sec
> >     MzScheme         8 sec
> >
> > Whenever we have time to implement another Scheme backend,
> > it thus seems that MzScheme remains the best choice.
> > Maybe someone knows of some more classical benchmarks for
> > Scheme implementations?
>
> The Ackerman function? ;)
>
> Maybe you should try running some of your experimental non-deterministic
> evaluation code. Continuations are such a sexy feature for tree
> traversals, it is a shame to be forced out of using them because of
> performance problems.

I agree. But I already know that MzScheme will be 1000 times faster;
I might try Scheme 48 though.

> Also, you might want to keep an eye out for developments in Pika. That's
> the new scheme system from Tom Lord, currently still pre-alpha, but it
> looks promising and its certainly more lightweight than MzScheme.

Thanks for the pointer. This will probably still take some time though.
In any case, I feel that the Scheme interface should be modified in such
a way that it is reasonably easy to plug-in abitrary Scheme implementations.
Moreover, this modification can probably be carried out by someone else
than me, since the Scheme glue has already been cleaned up quite a deal.