My Community

When I run a job in vasp 4.6.27, the files CHGCAR and WAVCAR are not written at the end of the job, although CHG is written. It appears to be a bug in the program. Thanks for any help resolving this.

Craig Plaisance
University of Virginia
Chemical Engineering

please check
1) were LWAVE and LCHARG really set to True
2) did the run really finish (last ionic step completed)?
CHG is written after each ionic step whereas CHGCAR and
WAVECAR are written after the last ionic step (NSW or
convergence reached)

yes and yes
other people in my group have also encountered this problem, so it is not due to something unusual that I am doing. Our other versions of vasp do not have this problem.

I can clarify what Craig is saying. We have 4.6.27 compiled on a few different clusters in our group. On this one particular cluster, opteron with intel fortran compiler, the WAVECAR and CHGCAR are not produced, even when the tags are set. The CHG file is produced, however, when I look at the file, it is formatted and contains the same data as the CHGCAR. It is quite mysterious to us why this is occurring; and it is completely reproducible for other systems.

I am running tests on other machines to see if this is reproducible with the binaries we have compiled on other architectures.

Chris

Update: Two other clusters (a Pentium IV cluster, and an AMD Athlon cluster) show the same bug. VASP was compiled with the intel fortran compiler in both cases. Is this compiler known to be problematic for VASP?

Chris

please try if it helps to upgrade to the latest vasp release (4.6.28) in such cases. I am not aware of any such troubles for vasp.4.6.28.

I was wondering if this problem had ever been actually solved. We're having the same problem (but with 4.6.28) on our cluster (Intel x86_64, MPI, Myrinet, RHEL WS 4.0). Some of our calcs (but not all) are stalling at the end of the run. The final energies, forces, etc. are written out after the last ionic step, but the program stalls at the "writing wavefunctions" statement in both the OUTCAR and stdout. Strangely enough, though, not all of our calcs are having this problem, only some of them (and sometimes it's the same one over and over again, and sometimes one that worked before stalls on the continuation run). Very strange...

Turning off the optimization for fileio.F seems to help (pgf90).
The ALLOCATE() call seems to have trouble, since it doesent return zero after succesfully allocating when compilation is done with -O3. It returns something like 1E-154 which gets written on an integer ISTAT. ISTAT then has some nonsense value, which causes a return after the conditional IF(ISTAT>0)...return.

Hope I could help.

I also have seen all and even worse problems with vasp.4.6.27 on our AMD cluster (ifort 9.0 + lam+ Goto). Note that the same jobs ran perfectly using vasp.4.26 on a HP (Intel Itanium2 1.5 GHz) clusters.

I think the best way to sort the problem out is to run the same job on the same machine compiled with 4.6.26, 4.6.27 and 4.6.28. (and vasp.5.0, hehe!)

However I cannot find the 4.6.26 and 4.6.28 on the ftp web site!

Can anyone point me where I can find them?

Thanks in advance for all of your kind help

--yi-ren

vasp cannot be downloaded from the web site at all.
please ask the person in your group who is responsible for the maintainance of vasp to download the latest version 4.6 from our ftp server for you.
vasp.5 is not released yet