Subject: Benchmarking the solar dynamo with Maxima
From: Stavros Macrakis
Date: Fri, 28 Nov 2008 13:57:53 -0500
On Fri, Nov 28, 2008 at 1:04 PM, Raymond Toy <toy.raymond at gmail.com> wrote:
> Adam Majewski wrote:
> > Valery Pipin pisze:
> >> compiled against the fastest lisps like sbcl or cmucl-sse2
> > 1. Why Maxima uses slower version of lisp ?
> >
> Because gcl runs "everywhere".
> > 2. How to compile it against the fastest lisps ?
> Last time I looked gcl ran the testsuite faster than all other lisps.
> Which is not to say gcl is fastest for all problems.
>
Just to expand a bit on Raymond's remark, there are vast differences in the
kinds of problems people use Maxima for. The biggest difference is between
purely symbolic and purely numeric problems, but there are many other kinds
of calculations in Maxima. Also, as Raymond implies, the testsuite is not
at all a good proxy for real problems. Not only does it mostly cover small
problems, but it is almost entirely interpreted (not compiled).
-s