VASP 6.2.0 GPU code, compiles but doesn't run

questions related to VASP with GPU support (vasp.5.4.1, version released in Feb 2016)

Moderators: Global Moderator, Moderator

Locked
Message
Author
james.almeida
Newbie
Newbie
Posts: 7
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

VASP 6.2.0 GPU code, compiles but doesn't run

#1 Post by james.almeida » Tue Oct 26, 2021 5:24 pm

Dear developers and users,

I have compiled VASP 6.2.0 with GPU support, using the attached makefile.include.
I am using the following compilers and libraries:
gcc 10.3.0
nvhpc-21.3
fftw-3.3.9-nvhpc-21.3

The software compiles, although, when I try to run, I get the following error:
ERROR: there must be 1 or 3 items on line 2 of POSCAR
----> I REFUSE TO CONTINUE WITH THIS SICK JOB ... BYE!!! <----


The same calculation runs well on the CPU, using the same files.
I am out of options to try, is this a problem with the version of the compilers/libraries used?
You do not have the required permissions to view the files attached to this post.

martin.schlipf
Global Moderator
Global Moderator
Posts: 440
Joined: Fri Nov 08, 2019 7:18 am

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#2 Post by martin.schlipf » Tue Oct 26, 2021 6:01 pm

Just to be clear, you use make std to build the GPU version, too, don't you? The make gpu command builds the CUDA port which is deprecated.

If that is the case, can you attach the inputs to the code (INCAR, KPOINTS, POSCAR) so that we can reproduce the issue.

james.almeida
Newbie
Newbie
Posts: 7
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#3 Post by james.almeida » Wed Oct 27, 2021 4:35 pm

Dear Martin Schlipf,

Thanks for the fast reply.

Yes, I have used the "make std" command.
The node as two Tesla P100 GPUs, thus, I am executing vasp with the following command: "mpirun -n 2 vasp_std"
I have also tried with only one process and the same error happens.
You do not have the required permissions to view the files attached to this post.

martin.schlipf
Global Moderator
Global Moderator
Posts: 440
Joined: Fri Nov 08, 2019 7:18 am

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#4 Post by martin.schlipf » Wed Oct 27, 2021 9:32 pm

Is it possible that you edited/generated the file on windows so that the line endings are different?

If you open your POSCAR file in vi and type
:set ff=unix
and save the file afterwards, does the issue disappear?

james.almeida
Newbie
Newbie
Posts: 7
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#5 Post by james.almeida » Thu Oct 28, 2021 11:22 am

Dear Martin Schlipf,

No I haven't eddited the file in windows.
I did the ":set ff=unix" anyway, and it still complains of the POSCAR.
I have basically run the simulation with another VASP compilation on CPU, everything was ok. Then I copied the input files to another directory and tried to run with the OpenACC compilation, and the POSCAR problem always shows up.

martin.schlipf
Global Moderator
Global Moderator
Posts: 440
Joined: Fri Nov 08, 2019 7:18 am

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#6 Post by martin.schlipf » Thu Oct 28, 2021 7:11 pm

Hmm, that is strange, for me that fixed the issue. Could you do one little test for me, please? I wrote a little test program that analyzes whether the line ending is indeed an issue. If you compile the code below with nvfortran and run it on your POSCAR file, what output do you see? Note that the expected output is "1.0", but on my machine before the line ending fix, I got "1.0 (with missing second quote).

Code: Select all

program main

    implicit none

    integer iunit
    character(len=1) char_
    character(len=256) line

    open(newunit=iunit, file='POSCAR', status='old')
    read(iunit, '(a1)') char_
    read(iunit, '(a)') line
    write(0,'(a,a,a)') '"', trim(line), '"'
    close(iunit)

end program main

james.almeida
Newbie
Newbie
Posts: 7
Joined: Sun Sep 29, 2019 11:43 pm
License Nr.: 5-426

Re: VASP 6.2.0 GPU code, compiles but doesn't run

#7 Post by james.almeida » Fri Oct 29, 2021 3:47 pm

Dear Martin.

I found the issue.
In the end the the problem was indeed with the POSCAR.
Thanks for providing the code to test it.

Now it is running fine!

Thanks for the assistance and sorry for the dumb mistake from my part.

Locked