From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2003-02-06 06:59:26
On Wednesday 05 February 2003 23:26, Matthias Troyer wrote:
> On Wednesday, February 5, 2003, at 04:52 PM, Toon Knapen wrote:
> > On Wednesday 05 February 2003 10:27, Matthias Troyer wrote:
> >> After we get the regrssion tests to work, there will be a special
> >> challenge for ublas or MTL-3: Calling the BLAS routine where
> >> appropriate will be essential in getting any decent performance on
> >> these types of CPUs. Does anybody know what the progress is on calling
> >> BLAS from ublas where appropriate?
> > The BLAS and LAPACK bindings for ublas are not complete yet but the
> > most important functions are available.
> Great. Is this part of the boost CVS
It's in the sandbox. If you're not familiar with the sandbox, drop me a line.
> > Recently, Joerg started adding blocked operations (which deteriorate
> > performance
> > on vector-processors) so in the future we might need configuration
> > macros to
> > indicate if we're on superscalar or vector machines.
> That will indeed be useful once we start running large production codes
> with sparse matrices on vector machines. For dense matrices the BLAS
> calls should take care of the optimization?
Again an advantage of using vendor-tune blas ;-)
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk