internal error in: vhdf5.F

[md_ashraf_ali]

Hi,
I am facing this bug while running a test calculation after installing vasp 6.4.2.

[fabien_tran1]

Hi,

At the beginning of your screenshot, the line "running 1 mpi-ranks, with 36 threads/rank ..." is printed twice instead of only once as it should. This indicates that there is an inconsistency in your usage of MPI. For instance, this can occur if you attempt to run the calculation with e.g., OpenMPI, while VASP was compiled with the intel MPI library. Are you using the same MPI library for compilation and executing the calculation?

[md_ashraf_ali]

yes I am using the same libarary for compilation.I have followed the attached procedure for compilation.

[fabien_tran1]

You forgot the attachment.

[md_ashraf_ali]

Hi,
I am facing the following problem after installing my vasp while running a test atom.

[jonathan_lahnsteiner2]

Dear md_ashraf_ali,

Please supply the input files of your test calculation as stated in the vasp forum guidelines https://www.vasp.at/forum/viewtopic.php?t=17630.
I would need the INCAR, POSCAR, KPOINTS, and POTCAR file.
Could you tell me what compilers, mpi, and hdf5 library you were using? Then I can try to reproduce your problem.
Without this information, I am not able to help you.

All the best Jonathan

[md_ashraf_ali]

I have attached the zip file containing the necessary files and a screenshot of my compilation information.

[jonathan_lahnsteiner2]

Dear md_ashraf_ali,

I am not able to reproduce your problem with the supplied test case and using a similar environment as you.
In the first picture you supply you start the code with mpirun -np 2 vasp and get the output

Code: Select all

running 1 mpi-ranks

twice. And in the second picture, you supply, you get the output

Code: Select all

running 1 mpi-ranks.

But what you should get is

Code: Select all

running 2 mpi-ranks,

so this means that there are issues with the mpi. Are you sure that you are compiling and running vasp with the same mpi version.
You can verify this by using the ldd command.
First use

Code: Select all

ldd /path_to_vasp_bin/vasp_std | grep mpi

to check with which mpi version vasp was compiled. Then check with

Code: Select all

mpirun --version

the mpi version with which you execute vasp. Those version numbers should match.
The next step is to check to which hdf5 library vasp is linked. This can be obtained with

Code: Select all

ldd /path_to_vasp_bin/vasp_std | grep hdf5

Now you know the absolute path to your hdf5 library and the names of the library files. Now check which mpi version was used to compile the hdf5 library.
Therefore type

Code: Select all

ldd /absoulte_path_to_hdf5_from_previous_command/libhdf5_fortran.so

The filename libhdf5_fortran.so might be slightly different in your case. But you will figure the file name out from the ldd /path_to_vasp_bin/vasp_std | grep hdf5 command.
Now you have to compare all the mpi versions if they are the same. Please try doing this and post the output of all the
terminal commands here.

I hope this is of help

All the best Jonathan

[md_ashraf_ali]

here is the output

[jonathan_lahnsteiner2]

Dear md_ashraf_ali,

Thank you for submitting the output. Here the linked libraries look okay.
Could you please also supply the output of

Code: Select all

which mpirun

All the best Jonathan

[md_ashraf_ali]

which mpirun output

[jonathan_lahnsteiner2]

Dear md_ashraf_ali,

Thank you for posting the output.
So the problem with your calculation is the following:
You are compiling vasp properly, also with the hdf5. For this
compilation, you are using an mpi which is located in a folder /lib/x86_64-linux-gnu/

But for executing the code with mpirun you are using a different mpi.
You can see this because the output of which mpirun does not match the library path against you link
vasp. So all you have to do is use the mpirun belonging to the mpi with which you compile vasp.

All the best Jonathan

[md_ashraf_ali]

Could you please tell me how can I use the exact mpi which is linked with vasp?

[jonathan_lahnsteiner2]

Dear md_ashraf_ali,

I would first try to deactivate the conda environment, because the mpi that you are using for execution is the anaconda mpi.

Code: Select all

conda deactivate

The (base) in your command prompt should vanish. Then try to execute vasp.
You can also check again what mpirun version you are using with

Code: Select all

which mpirun

Please try this and post again the output in case vasp execution does not work.

All the best Jonathan

[md_ashraf_ali]

vasp calculation is running successfully without any bug error.But my vaspout.h5 is not opening.
More, I am facing the following problem in py4vasp.How can I fix the LORBIT setting issue?