Dear William,
On Mon, 24 Mar 2008, William Stein wrote:
> I would much prefer speeding up Sage matrices rather than incorporating
> MTX into Sage, since the result will in the former case will be much
> easier for users to understand and lots of other codes benefits.
Sure. As a user, i wouldn't like to have completely different types of
matrices for small or big fields *on the surface*. I wouln't care, though,
what is done internally.
> Why? It's worth seeing if Linbox properly used can beat MTX -- I mean
> dense nullspace of matrices of that sort of size is I think exactly where
> linbox should be beating everything else, at least if one has a properly
> optimized BLAS.
I wondered, myself. Does Sage properly optimize BLAS (via ATLAS)? I hope
so!
> > - MeatAxe (at least version 2.2.3 of 1997 that i am using) relies on
> ^^^^^^^^^^^^^^^^^^^^^^^^^^
> That's a big (!!!) warning to me, unless you want to become the
> official maintainer and make new releases...
:-/
I guess that my wrapper would work with the new MeatAxe as well. Do you
think it is worth trying?
In my project (computation of cohomology rings of finite p-groups), i
wouldn't easily put down the "good old" MeatAxe, because my Sage programs
rely on C programs of David Green (my boss), and he said that they rely on
*old* MeatAxe versions and wouldn't work with the new (i didn't test
myself, though).
So, unless Sage matrices are not *vastly* faster in 1. sum/difference of
matrices, 2. product of matrix by scalar, 3. computing hash, 4. computing
nullspace, i wouldn't change it in my programs.
But that doesn't mean it is good for Sage, in general. I would be very
glad if my comments would initiate improvements on Sage matrices!
Yours
Simon
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
