UT theoretical chemistry code forum

I compiled VASP 6.2 with VTST (using the 180 version) and although it seemed to compile correctly, the VTST options were clearly ignored when running the code. This was using examples that worked on VASP 5.4.4. I don't exclude stupidity on my part, but has anyone got it working on 6.2?

Cheers,

Adam

Hi Adam,

First, great to hear from you!

We have not had any trouble with VTST and vasp 6.2 but I am eager to hear if anyone else does. Maybe start by greping the OUTCAR for VTST to see if the code was linked in. If not, there is likely a problem with how it was built. If you see a VTST statement in the OUTCAR and there is a subsequent problem with a calculation, please post more details and I will try to get to the bottom of it.

Graeme

Dear Graeme,

thanks and I hope all is well (as can be in these times) in Texas!

We got it built now, but have run into a fault in the VASP tests. Specifically it is runs fine (electron energy minimization), until it throws a segfault with the first tests that come with vasp. It looks like it segfaults in file chain.F, line 295 in routine chain_force. On that line it is using allocatable variable "Free". Our guess is that there is something wrong with that, chain_init is not called to allocate the array in time, or something else...we will keep checking to see if it relates to our machines etc.

Cheers,

Adam

Very strange; that is not new code. I would be very interested to hear what you find.

If you get stuck, I can try to reproduce the error here.

I should note that I actually only checked vasp 6.1.x; I'll check 6.2 tonight.

At least the difference we found:

vasp.6.1.2_patched/src/main.F:913: CALL chain_init( T_INFO, IO)
vasp.6.2.0/src/main.F:889: IF (LCHAIN) CALL chain_init( T_INFO, IO)

Just removing the conditional fixed the error and at least the first tests passed fine. But this is a rough fix, and maybe messes something else up?

Cheers,

Adam

Ah, that makes perfect sense, thank you! We rely on the call of chain_init to allocate variables. I can't think why removing the conditional and always calling chain_init shouldn't work (that was the logic previously) but I'll take a closer look at this again.

Again, really appreciate the heads up about this.