My Community

Recently, I have tried compiling VASP on Louisiana State University (LSU)'s SuperMIC computer which is part of the NSF's XSEDE funded machines. The configuration is that each node has 2.8GHz 10-Core Ivy Bridge-EP E5-2680 Xeon 64-bit Processors (as well as two co-processors which we aren't using) The ifort version is 14.0.2 and the mkl is provided as part of the same package with Cluster Studio XE 2013. For fftw, I'm linking against FFTW 3.3.3 version. During compilation, everything runs smoothly and I get the compiled versions for vasp_std, vasp_gam, and vasp_ncl.

However, when I go to test the compilation on even something like silicon, I immediately get the following error written to standard out and then nothing else (the job crashes):
Error reading item 'VCAIMAGES' from file INCAR.
(repeated for each processor on a node).

From searching the source code, I only find the term VCAIMAGES in two places: chain.F and main_mpi.F. I don't think the chain.F part should be causing it since I am not running an elastic band calculation and therefore am led to believe it is main_mpi.F causing it, perhaps during the setup of the calculation.
I have tried switching off the use of ScaLAPACK as well to no avail.

Since I can't find any reference to these in the manual or really anywhere else on-line, I was wondering if anybody else has encountered this error and has any possible suggestions to try for resolving it. I have included my INCAR file below as well.

Thank you,
Andy

INCAR file:
ISTART = 0
ICHARG = 2

ENCUT = 450
EDIFF = 1E-6
ALGO = Normal
NELMIN = 2
NELM = 80
LREAL = .FALSE.

NSW = 0
IBRION = -1
ISIF = 2
EDIFFG = -1E-2

ISMEAR = -5
LORBIT = 11
NEDOS = 1000

LWAVE = .FALSE.
LCHARG = .FALSE.
LVTOT = .FALSE.
LVHAR = .FALSE.

this VCAIMAGES error (if VCAIMAGES is not explicitely set in INCAR) usually indicates that INCAR is not available in the directory in which vasp is run

admin, thank you for your help - this pointed me in the correct direction. It turns out that the scheduler on this machine doesn't automatically switch into the directory that the script is submitted from while the other machine that I've been using does. Once I added the correct cd line, the examples run as expected.

On the topic of the VCAIMAGES error, I have once seen this another place too (when not explicitly setting it). I was once trying to set KPAR to a large number, but if I went over 12 I would get it. Is there a limit to the KPAR parallelization or some other factor when trying to really push parallelization that might also cause this?

Thanks again.