On Sun, Jul 03, 2022 at 03:41:50AM +0200, Kurt Pagani wrote:
>
> There is also a very useful fortran to lisp translator (f2cl) available:
>
> https://trac.common-lisp.net/f2cl/
> https://trac.common-lisp.net/f2cl/wiki/GettingF2cl
> http://quickdocs.org/f2cl/
>
> """
> The easiest way to get F2CL is to use Quicklisp. You can just say
> (ql:quickload :f2cl)
> to get f2cl.
> """
>
> I translated some examples (*)
> (https://github.com/nilqed/f2cl/tree/master/packages/) some years ago (e.g.
> BLAS) and it worked well, i.e. it seems easier to me to do lisp->spad than
> c->spad.
>
> (*) fortran sources partly from netlib.org
Well, from my point of view main point of BLAS is speed. Currently
speed that we can get from Spad code is limited by code generator
in underlying Lisp. It is not clear for me if f2cl gives us any
gain for real (double computations). There is probably gain for
complex due to incompatiblity of Spad Complex with Lisp complex
types. But I expect system BLAS to have much higher speed
(IIRC in the past optimized BLAS had 4-10 time better performance
than "reference BLAS" which is probably base for translation to
Lisp). On BLAS type code I would expect sbcl to get 2-4 times
slower code that Fortran compiler. And since optimized BLAS
uses special machine instructions and newer machines introduced
more such instructions I expect that gap between optimized BLAS
and f2cl result will be bigger.
--
Waldek Hebisch
--
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/fricas-devel/20220707015844.GB12604%40fricas.math.uni.wroc.pl.
