prt1dm problem with fft parallelism

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
asorini
Posts: 13
Joined: Wed Apr 07, 2010 12:23 am

prt1dm problem with fft parallelism

Post by asorini » Tue Aug 31, 2010 9:50 pm

Hi,

I'm using abinit 6.0.3. For a parallel run at the gamma point parallelized only over bands (npband=30), I used prt1dm to print the density and potential with reasonable results. But when I parallelize over bands and fft (npband=30 npfft=4), I get a lot of large negative numbers for the density in the 1DM output file.

Cheers,

Adam

mverstra
Posts: 655
Joined: Wed Aug 19, 2009 12:01 pm

Re: prt1dm problem with fft parallelism

Post by mverstra » Sat Sep 04, 2010 10:14 am

could you try abinit 6.2? It sounds like you have a proper bug here, someone will have to have a look at making sure this output is correctly coded in the parallel fft case.

cheers

matthieu
Matthieu Verstraete
University of Liege, Belgium

Locked