My Community

Dear experts,
I recently installed VASP.6.4.2 and it was compiled successfully. While running calculations I get the following in the OUTCAR even after using multiple cores-
running 1 mpi-ranks, on 1 nodes
distrk: each k-point on 1 cores, 1 groups
distr: one band on NCORE= 1 cores, 1 groups

I have attached the OUTCAR, makefile.include and makefile files here. Please let me know if it was installed parallelly o serially and how to correct it. Thank you.

How did you start the calculation? You did use a variant of didn't you?

Another possible scenario is that the mpi library used to compile the code is not the same used to run the executable. In that case you would expect to see the VASP output number_of_processes times, every time using serial execution. This can happen on clusters or supercomputers if the modules loaded during compilation are different from the one during the execution.

If neither of the above solves the issue then please add the output of building VASP, your execution script or command, and the standard output of VASP to your reply. You may also try if prints 4 times Hello, World! to check the sanity of your MPI setup.

Thank you for the reply.

Yes I had used mpirun -np to execute the run. I have attached the makefile.include i have used to compile vasp and the output messages produced at that time (compile.txt). I should mention that, I have compiled vasp exactly like that in one of the HPCs of my institute and it runs perfectly fine. But I am not able to run the same parallelly in another HPC. I have also attached the job script I used to execute the calculation. Thank you.

Thank you for the output. This confirms that you are in this scenario.

martin.schlipf wrote: ↑Wed Oct 11, 2023 6:50 amAnother possible scenario is that the mpi library used to compile the code is not the same used to run the executable. In that case you would expect to see the VASP output number_of_processes times, every time using serial execution. This can happen on clusters or supercomputers if the modules loaded during compilation are different from the one during the execution.

Notice that you use mpiifort (Intel MPI) to compile your code but load OpenMPI in your jobscript. As a result you get 8 parallel executions of VASP on 1 rank instead of 1 parallel execution of VASP on 8 ranks. Please either recompile with OpenMPI or load the Intel MPI in your jobscript.

Thank you for the suggestion. The issue was finally fixed by not loading the openmpi module.