[Pw_forum] band parallelization

Lorenzo Paulatto lorenzo.paulatto at impmc.upmc.fr
Tue Sep 22 13:35:23 CEST 2015

On Tuesday, September 22, 2015 10:42:13 AM Tiana Davide wrote:
> is the principle correct or it should be done in another way?

Yes, but as the calculation is becoming slower with more CPU it means that 
communication is becoming too expensive. You have to check at the end of the 
output file to see what scales and what does not, in you case it is probably 
the Fourier transform that is killing you scaling, you can try to use some 
task groups (-ntg XX if I remember correctly) to see if you can scale a bit 


Dr. Lorenzo Paulatto
IdR @ IMPMC -- CNRS & Université Paris 6
+33 (0)1 44 275 084 / skype: paulatz
23-24/4é16 Boîte courrier 115, 
4 place Jussieu 75252 Paris Cédex 05

More information about the users mailing list