> Gib Bogle ihug.too.much.spam.co.nz> wrote:

>> The reason I did this
>> quick test is that I have a student who is solving a nonlinear system of
>> PDEs

>
> Why on earth do you then benchmark with a matrix-matrix multiplly?
> Operations on PDEs (which I assume are all sparse matrices) can not be
> more unlike a mmm. You should have benchmarked the sparse matrix vector
> product, and you would actually see that matlab was much slower since
> its sparse data format is really inefficient.
>
> Victor.
> --
> Victor Eijkhout -- eijkhout at tacc utexas edu