[Wien] Wien2k with load-distributing systems

[Peter Blaha]

> I'd like to know what you think about the multithreading of the ifc+mkl
> on shared memory systems.  They claim to have "excellent scaling",
> whatever that may mean.

I don't have a dual (or more) P4 machine, so everything is speculation or
comes from some spare info of other sources.

However, I would doubt that compiler multithreading can spead up a single
lapw1 job on a dual P4 by a factor of 2. I'd expect more a factor of 1.1 or
1.2

However, with two lapw1 running on this dual P4 I expect a speedup of 1.8,
provided you have enough memory for the two lapw1. It is most likely not a
factor of 2, since the memory bus gets congested by the two lapw1 jobs on
these "non-balenced" dual P4 architecture.

On two singel P4 Pcs I'd expect a speedup of 1.99.

Regards

                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------