Subject: Initial- and boundary-value problems in Maxima.
From: Raymond Toy
Date: Fri, 16 Sep 2011 09:01:16 -0700
On Thu, Sep 15, 2011 at 4:15 PM, <talon at lpthe.jussieu.fr> wrote:
> Raymond Toy wrote:
>
> x
> >>
> >> I don't understand the discrepancy between 50% of the time and
> >> "negligible"
> >> between Michel and Ray;
> >> since I have some interest (as does Ray) in making lisp numerical
> >> programs run fast, I hope we can come to some agreement on this.
> >>
> >> Could be that sbcl did a better job of optimizing, especially if it's a
> > 64-bit version because then the 32-bit integer Fortran quantities will
> fit
> > nicely in a fixnum. Could also be a bug in cmucl's profiler.
> >
> > Ray
>
> My machine is a 32 bits machine. I have here, in another mail to you,
> profiles with both cmucl and sbcl which are consistent between themselves
> but not consistent with your profiles, so we can only agree to disagree.
> These profiles were done with you new colnew-if which allows to see
> directly the time spent in fsub and dfsub (basically 50 %): Note that dgesl
> and dgefa are much slower in cmucl than in sbcl. As always the boundary
> conditions gsub and dgsub are not relevant. This is because fsub and dfsub
> are computed on all mesh points while the boundary conditions are computed
> on 4 points only on this problem. And the number of mesh points rises up to
> around 80, plus there are several Newton iterations.
>
>
Before we do anything else, let's agree on what exactly is being tested.
The profile that I reported was using the current maxima sources using
cmucl. I just did load(colnew), then enabled profiling for the MAXIMA and
COLNEW packages and then load("share/colnew/prob4.mac"). What steps did
you use?
Also, perhaps prob4 is more complex that we really need. I think prob1.mac
is a simpler test.; let's use that for comparison.
Ray