Page 2 of 2

Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Fri Mar 21, 2025 2:30 pm
by karianaandrea_morenosader

Thanks Martin,

This is the OS details: Linux login-4 5.10.234-llgrid #20250211 SMP Tue Feb 11 11:49:53 EST 2025 x86_64 x86_64 x86_64 GNU/Linux
but I don't have sudo access in the SuperCloud but I can set up a conda environment?


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Fri Mar 21, 2025 3:22 pm
by martin.schlipf

Could you type

Code: Select all

module avail

and provide me with the output? I am wondering why you installed your own MPI and Scalapack instead of using the ones provided by your HPC site. Also just as an FYI, your HPC site can apply for a separate VASP license if they want to install VASP for their users. Usually, there is no additional fee if we have users at their site.


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Mon Mar 24, 2025 10:07 am
by karianaandrea_morenosader

Hi Martin,

Hope you had a good weekend! Here is the output:

Code: Select all

module avail
----------------------- /etc/environment-modules/modules -----------------------
anaconda/2022b                             intel/oneapi/mkl/2025.0
anaconda/2023a                             intel/oneapi/mkl/latest
anaconda/2023a-pytorch                     intel/oneapi/mpi/2021.14
anaconda/2023a-tensorflow                  intel/oneapi/mpi/latest
anaconda/2023b                             intel/oneapi/oclfpga/2025.0.0
anaconda/Python-ML-2023b                   intel/oneapi/oclfpga/latest
anaconda/Python-ML-2024b                   intel/oneapi/tbb/2022.0
anaconda/Python-ML-2025a                   intel/oneapi/tbb/latest
cuda/10.0                                  intel/oneapi/umf/0.9.1
cuda/10.1                                  intel/oneapi/umf/latest
cuda/10.2                                  intel/oneapi/vtune/2025.0
cuda/11.0                                  intel/oneapi/vtune/latest
cuda/11.1                                  julia/1.7.3
cuda/11.2                                  julia/1.8.5
cuda/11.3                                  julia/1.9.2
cuda/11.4                                  julia/1.10.1
cuda/11.6                                  julia/1.11.3
cuda/11.8                                  mpi/openmpi-4.1.3
cuda/12.2                                  mpi/openmpi-4.1.5
cuda/12.6                                  mpi/openmpi-5.0.5
graphblas/4.0.3                            nccl/2.5.6-cuda10.0
graphblas/5.0.5                            nccl/2.5.6-cuda10.1
graphblas/6.1.3                            nccl/2.5.6-cuda10.2
graphblas/7.2.0                            nccl/2.8.3-cuda11.0
graphblas/9.4.2                            nccl/2.8.3-cuda11.1
gurobi/gurobi-951                          nccl/2.8.3-cuda11.2
gurobi/gurobi-1000                         nccl/2.9.8-cuda11.3
gurobi/gurobi-1102                         nccl/2.10.3-cuda11.4
intel-oneapi/2023.1                        nccl/2.11.4-cuda11.6
intel/oneapi/advisor/2025.0                nccl/2.18.1-cuda11.8
intel/oneapi/advisor/latest                nccl/2.23.4-cuda12.2
:

A few other updates: I found out that the segmentation fault I encountered was reported as a bug in the OpenMPI version I was using. I switched to version 4.1.6 (which was already available in my lab's shared folder) and ran this simple example: https://vasp.at/tutorials/latest/bulk/part1/
When I ran this example with mpirun --np 1, I received the warning below, but the program still executed without any segmentation error

No OpenFabrics connection schemes reported that they were able to be
used on a specific port. As such, the openib BTL (OpenFabrics
support) will be disabled for this port.

Local host: login-4
Local device: mlx5_0
Local port: 1
CPCs attempted: rdmacm, udcm

When I tried --np 4, I also get an additional error at the end:

[login-4:57679] 3 more processes have sent help message help-mpi-btl-openib-cpc-base.txt / no cpcs for port
[login-4:57679] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages

And when I run the testsuite, I get a segmentation fault
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Another lab that holds a VASP license and runs jobs in Supercloud uses OpenMPI 4.1.5 with VASP 6.3.1, so maybe I should try setting mine with the available OpenMPI version on Supercloud? Regarding ScalaPack, I didn't see it in the module available so perhaps I should keep mine?

Thank you very much, Martin!! We are getting there.

[MS edit]: I reformatted the module avail output for better readability.


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Mon Mar 24, 2025 1:24 pm
by martin.schlipf

Can you try a different installation? Perhaps you copy the source to a different directory, so you don't lose any progress.

With the module list you provided, I would try the following

Code: Select all

module load intel-oneapi/2023.1
module load intel/oneapi/mkl/latest
module load intel/oneapi/mpi/latest

and then copy the appropriate makefile.include

Code: Select all

cp arch/makefile.include.oneapi makefile.include

and test if that works.


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Tue Mar 25, 2025 2:20 pm
by karianaandrea_morenosader

Hi Martin,

I completed the installation using onepi makefile, and the tests ran for 30 minutes. The .log file doesn't show any segmentation faults, but the summary says:

==================================================================
The following tests failed, please check the output file manually:
andersen_nve andersen_nve_constrain_fixed andersen_nve_constrain_fixed_RPR andersen_nve_efor andersen_nve_RPR andersen_nvt andersen_nvt_fixed andersen_nvt_RPR bulk_InP_SOC_DFT_ISYM=2 bulk_InP_SOC_DFT_ISYM=2_RPR bulk_InP_SOC_DFT_ISYM=3 bulk_InP_SOC_DFT_ISYM=3_RPR bulk_InP_SOC_G0W0_sym bulk_InP_SOC_G0W0_sym_RPR bulk_InP_SOC_PBE0_nosym bulk_InP_SOC_PBE0_sym bulk_InP_SOC_PBE0_sym_RPR H2Ostretch_efor H2_bare_HF NiOsLDAU=2_x NiOsLDAU=2_x_RPR NiOsLDAU=2_y NiOsLDAU=2_y_RPR NiOsLDAU=2_z NiOsLDAU=2_z_RPR Tl_x Tl_x_RPR Tl_y Tl_y_RPR Tl_z Tl_z_RPR

And looks like the errors come from not finding some files to start the runs. Could you please take a look at my .log file (see attached) to confirm whether my installation is all set? Thank you very much for all your help


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Tue Mar 25, 2025 3:19 pm
by marie-therese.huebsch

Looks like all the once where you need the vasp_ncl executable. Did you make all executables or just vasp_std?


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Tue Mar 25, 2025 3:30 pm
by karianaandrea_morenosader

Just std :) but I can do all if that is necessary to run the entire testsuite without such errors, but looks like I am good to go with std? thanks


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Tue Mar 25, 2025 5:00 pm
by martin.schlipf

I recommend building all executables now while the setup is fresh in your memory. Otherwise you may realize you need it in half a year and then need to recall all the steps you did now.

Let me know if any problems appear and have fun using VASP.


Re: Successful installation of VASP 6.5.1 but failed test run

Posted: Wed Mar 26, 2025 12:07 pm
by karianaandrea_morenosader

Thanks a lot Martin and Marie! I built all executables and ran the test suite without problem SUCCESS: ALL SELECTED TESTS PASSED

Thank you very much again for helping me with this