The ABINIT Users Mailing List ( CLOSED )

[braun@helios.phy.ohiou.edu]

Dear all,
I have run calculations for a molecule with one kpt and have tried to
exploit the band parallelism by using wfoptalg 1 and ndbblock 1,2,4 with
four procs and abinip. I am getting different overall cpu times and total 
energies depending on what ndbblock I am using. Furthermore the total cpu time
is two times bigger than by using the sequential version. Tests v1 - v4 run
fine (some precision differences) as well as the parallel tests. Should it not
be faster to use wfoptalg in parallel giving always the same energies ?
Thanks for helping
Kai