Hi Ivo,
On 18 Jul., 12:15, Ivo Hedtke <hed...@me.com> wrote:
>
> Why do you wan't to use Strassen for Matrix Multiplication? This algorithms is
> not as stable as naive algorithm (if one uses floating point numbers, if you
> use
> it for symbol computations that is no problem).
I work over finite fields.
> It is also not very fast for
> small matrices.
That's why there is a cutoff.
> Why not Winograd's method.
Practical answer: Since I found that there is a framework for the
Strassen multiplication in Sage (see sage.matrix.strassen), but I
don't know whether a similar support exists for the Coppersmith-
Winograd algorithm.
search_src('Winograd') told me that the multiplication algorithm used
in sage.matrix.matrix_mod2_dense is called "Strassen-Winograd", but
the asymptotics seems to coincide with what I know as Strassen
algorithm.
> Or something like Tiling or Loop-Unrolling?
Would that change the asymptotics?
Cheers,
Simon
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
