Dear Forum,
I have built 2025.1 via the toolchain using the following input:
--install-all -j36 --with-openmpi=system --with-gcc=system --with-sirius=no --with-libtorch=no --enable-cuda --gpu-ver=A100 --no-check-certificate --with-quip=no --with-sirius=no --with-spfft=no --with-spla=no --with-pexsi=no --with-superlu=no --with-cusolvermp=no --with-dftd4=no --with-deepmd=no --with-elpa=no --with-cosma=no
The code compiled without an error. When I try to run a job, the job crashes with the following error:
ERROR: cuInit failed with error: 100 /cluster/home/tanoury/CP2K/cp2k-2025.1/src/offload/offload_library.c 57
My cuda version and the version supported by the GPUs are the same (12.3.0).
Thank you so much for any help you can provide,
Jerry