Problem with a DGEGV parameter

Message

smarocchi · #1 Post by **smarocchi** » Tue May 21, 2013 7:34 pm

Dear VASP support, I run a scf calculation on a quite large system (171 atoms) and I decided to exploit the full potential of my machine (blugene Q) so I tried launch my calculation on 1024 core (each one with 1 GB of RAM) using the RMM-DIIS instead of Davidson to save memory. The run started well but unexpectedly at the electronic iteration 16 I received this message

** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value
** On entry to DGEGV parameter number 16 had an illegal value

I already sent you the input and output file at the email address vasp.materialphysik@univie.ac.at.

Thanks for your help,
Simone

annmattsson · #2 Post by **annmattsson** » Thu Oct 31, 2013 12:51 am

Try setting MAXMIX to something larger than 8 in INCAR.

annmattsson · #3 Post by **annmattsson** » Thu Oct 31, 2013 7:26 pm

I think it can also be that you run out of memory. Bluegene does not have a lot of memory per node.