[Pw_forum] Nonlinear scaling with pool parallelization

Markus Meinert meinert at physik.uni-bielefeld.de
Tue Apr 5 19:54:35 CEST 2011

Sorry, I should go into a little bit more detail here. I used an 
_unshifted_ k-mesh. When I use a 8x8x8 mesh, yielding 58 points, I get a 
speedup of about 2 on three machines. With a 12x12x12 mesh with 144 
points the speedup becomes better (factor 2.6 for cpu time and 2.3 for 
WALL). I used to think that's because of communication. In the former 
case, an iteration takes less than a second. With more k points, the 
speedup converges slowly towards 3.

The slab has 20 k points. But, since a single iteration takes about 100 
seconds, I do not see where the time is being spent, when the k points 
are independent.


Dipl.-Phys. Markus Meinert

Thin Films and Physics of Nanostructures
Department of Physics
Bielefeld University
Universitätsstraße 25
33615 Bielefeld

Room D2-118
e-mail: meinert at physik.uni-bielefeld.de
Phone: +49 521 106 2661

More information about the users mailing list