This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: unflattening fortran matrices
Jack Howarth wrote:
> Paul,
> This question of what we are potentially losing is exactly
> why I suggested we try to find a common set of benchmarks that
> exist as both c and fortran versions with relatively equivalent
> levels of source code optimization. This might give us some
> concept of what gains are available from vectorization and
> any other optimizations that are stumbling on the problem of
> calculating the stride of the flat fortran matrices. I believe
> vectorization, as currently implemented will only gain us a
> few percent, but vectorization is only at its early stages and
> is supposed to be significantly improved over time.
> Jack
>
Do the Levine-Callahan-Dongarra "vector" tests, including f95 and C++
versions, fit your description?