xtb --ohess on certain computers

[longdragoon]

Describe the bug xTB (ver 6.7.1) --ohess fails on NERSC (https://docs.nersc.gov/) xTB --opt followed by --hess works properly xTB --ohess works on my PC and Grace (https://docs.ycrc.yale.edu/clusters/grace/)

To Reproduce xtb --gfn 2 --ohess --input xcontrol mol_10.xyz > xtbopt.out & xcontrol and xtbopt.out files are attached (renamed to .txt due to github requirement) xcontrol.txt xtbopt.txt

[marcelmbn]

Can you post the coordinate file mol_10.xyz in addition to the outputs?

[longdragoon]

It's simply H2. The coordinates are as below (xyz format):

2

H 0.349273 0.000000 0.000000 H -0.349273 0.000000 0.000000

[marcelmbn]

I cannot reproduce the behavior on the following machines:

• macOS@arm64: self-compiled via gcc
• macOS@arm64: brew installation
• Ubuntu-20.04: self-compiled via ifort
• openSUSE: self-compiled via ifort or gcc

If you have news on that, or additional output or if you can narrow down the error, we can definitely look into that but up to now I will close.

P.S.: Can you describe how you compiled the program? Maybe, we note some errors with the LAPACK/BLAS interface or similar.

[toxtran]

I can reproduce, but I do not know why it is happening

[marcelmbn]

Then we reopen. How can you reproduce it? Can you give detailed information on

• setup (machine, OS)
• compiler and compiler settings
• setup and build commands used
• CLI command used (xtb --ohess struc.xyz?)

[foxtran]

So, it happens for precompiled xtb 6.7.0 and xtb-bleed. Manually compiled for specific env works fine.

I'm using Hungarian Komondor supercomputer. See the doc here. It is CRAY machine.

Some system info:

$ cat /etc/os-release
NAME="Red Hat Enterprise Linux"
VERSION="8.6 (Ootpa)"
$ uname -a
Linux vn01 4.18.0-372.9.1.el8.x86_64

Compilers:

$ cc --version
gcc (GCC) 11.2.1 20220127 (Red Hat 11.2.1-9)
$ CC --version
g++ (GCC) 11.2.1 20220127 (Red Hat 11.2.1-9)

I swapped envs for avoiding compiler bugs in Cray compilers:

module swap PrgEnv-cray/8.4.0 PrgEnv-gnu/8.4.0

Now, any running of xtb lead to SegFault.

If I do not swap, xtb works. With configured python environments, which is not enabled by default, it also works.

[foxtran]

Looks like it is some CRAY specific problems :D NERSC has a Perlmutter cluster by CRAY :D