Problems with Spin Polerized Calculations, while Running in Parallel
Posted: Fri Sep 05, 2008 9:52 pm
Hello,
I'm not a Unix guru and I didn't compile the parallel version of VASP on our Unix machine. Unfortunately, nobody in my group can help me get a properly compiled version of Parallel VASP. The machine we use is a SGI Altix 3000.
Here is the problem. I'm trying to benchmark iron for the machine. The benchmark run for Serial VASP gives good results for spin polarized calculations. However, Parallel VASP gives positive energies and a lattice constant way too large (an energy of 494 eV and a lattice constant of 16.38 Ang.) Interestingly enough, when I switch to non-spin polerized calculations the energy is negative and the lattice constant is more reasonable.
Does anyone have any ideas about what might be going on. I can send a Makefile if necessary. Thank you.
Best Regards,
Modey3
I'm not a Unix guru and I didn't compile the parallel version of VASP on our Unix machine. Unfortunately, nobody in my group can help me get a properly compiled version of Parallel VASP. The machine we use is a SGI Altix 3000.
Here is the problem. I'm trying to benchmark iron for the machine. The benchmark run for Serial VASP gives good results for spin polarized calculations. However, Parallel VASP gives positive energies and a lattice constant way too large (an energy of 494 eV and a lattice constant of 16.38 Ang.) Interestingly enough, when I switch to non-spin polerized calculations the energy is negative and the lattice constant is more reasonable.
Does anyone have any ideas about what might be going on. I can send a Makefile if necessary. Thank you.
Best Regards,
Modey3