On Fri, Jul 01, 2022 at 08:30:46AM +0800, Qian Yun wrote:
>
>
> On 7/1/22 04:32, Waldek Hebisch wrote:
> >
> >I personally see no need for such option, but it is hard to
> >know what other people want.
>
> So no objection here. I think it is a good thing to properly
> layer a project -- it is good abstraction. In our case,
> it is the Lisp layer and C layer, and OS layer (shell stuff).
>
> >
> >Also, without FFI we would have no gmp for sbcl. With ECL
> >no C compiler means no compiler at all.
>
> It is just an option for sbcl.
>
> For ECL I think it will call C compiler regardless of our CC setting.
>
> >Let me add that I consider making FFI mandatory and using
> >foreign code to speed up algebra.
> >
>
> The "speed up" is GMP right?
Not only. We could make more use of GMP than we make now,
but there are also other possibilities. One is to have
fast routines for polynomials mod p. We could use BLAS
to speed up linear algebra.
> Currently only available for
> sbcl and clozurecl.
>
> On the same subject, what's your opinion on providing GMP
> for all Lisps via CFFI?
ECL and GCL use GMP to implement their fixnums. This is
actually a big point of our GMP support: for sbcl and Clozure
CL we speed up _native_ bignums. So avoiding _explict_
use of FFI is a feature. We could could make more use
of GMP and in such case FFI binding would be useful. But
it is likely that any such use would go via C wrapper.
I must admit that for clisp I decided that GMP binding
is too much work for the effect: clisp is slow anyway.
--
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/20220701012142.GA27101%40fricas.math.uni.wroc.pl.
