[Pw_forum] parallel run of phonon calculation (ph.x)

Huiqun Zhou hqzhou at nju.edu.cn
Mon Apr 7 13:47:44 CEST 2008


list-users,

I'm trying to run ph.x in the mode of q-point parallelization,
using 4 compute nodes with 4 cpu cores in each. The command is
sort of following 

    mpiexec -n 16 ph.x -npool 4 < input > output

Say, it determined there are 8 q points according to the q-mesh
I set. I'm expecting it could parallelly compute 4 q points per 
time (so get *.dyn1-4 simultaneously), but I saw it still compute
q point one by one. Is my understanding of q point parallelization
wrong?

Thanks!

Huiqun Zhou
@Earth Sciences, Nanjing University, China 


More information about the Pw_forum mailing list