Tests for the VTST

aab · Post by **aab** » Sun Jan 06, 2008 10:57 pm

Hi,

I am trying to install the VTST tools. I downloaded the code from the
web page. Rebuild the VASP. Unfortunately, the code doesn't seems
to work in parallel. It dies in ZHEGV routine (called from eddav) with the
code suggesting that the matrix is not positively defined.
I tried to debug it a little and it seems to me that the wavefunction
(W1%CR in EECP called from davidson.F) doesn't look right (elements
are too large, about 200-400).

My question: is there any test suit for the VTST tools I can try?
I am currently interested in the NEB method.

Best wishes,

Andrey

Post by **graeme** » Mon Jan 07, 2008 1:35 am

There is no test suite, but this problem is also probably not due to the VTST code - I would be very surprised if it was. If you can run the normal vasp benchmarks, do a regular NEB calculation (without the VTST code), and get this error only when you link in the VTST code, then we would certainly like to debug it. Most likely, though, it is the typical problem of getting your math library, fft routines, and mpi all working together, and not related to our code.