Building LAMMPS with GPU single precision support on v100 nodes

LAMMPS LAMMPS Installation
1

Hello,
I would like to build lammps with libgpu support in single precision on V100 nodes. I am using the following cmake command:cmake -DPKG_GPU=on
-DPKG_REAXFF=on
-DPKG_MANYBODY=on
-DPKG_ML-SNAP=on
-DPKG_ASPHERE=on
-DPKG_RIGID=on
-DPKG_REPLICA=on
-DPKG_kspace=on
-DPKG_MOLECULE=on
-DPKG_EXTRA-MOLECULE=on
-DBUILD_MPI=on
-DCMAKE_INSTALL_PREFIX=…/install_v100_gpu_single
-DMPI_CXX_COMPILER=(which mpicxx) \ -DCMAKE_BUILD_TYPE=Release \ -DCMAKE_CXX_STANDARD=14 \ -DCMAKE_CXX_COMPILER=(which mpicxx)
-DGPU_API=cuda
-DGPU_PREC=single
-DGPU_ARCH=sm_70
…/lammps/cmake

and I am running into the following cuda driver error:
Cuda driver error 4 in call at file ‘/nobackupnfs1/gjost/lammpsJun2023-cmake/lammps/lib/gpu/geryon/nvd_timer.h’ in line 98.
Cuda driver error 4 in call at file ‘/nobackupnfs1/gjost/lammpsJun2023-cmake/lammps/lib/gpu/geryon/nvd_timer.h’ in line 99.

It works fine with Kokkos in double precision. Hase anybody seen this problem before or is there a glaring mistake in my cmake command?
Thanks in advance for any pointers, Gabriele

2

a) please properly quote any cut-n-paste text with triple backquotes (```) as it otherwise becomes very difficult to read since the forum software otherwise will interpret the special characters as markup.
b) please provide all relevant details (LAMMPS version, compiler version, input files that produce the error, command lines etc.).

If you have not done so, please check out the forum guidelines where this is explained in details and the motivation for it and some more, related things (Please Read This First: Guidelines and Suggestions for posting LAMMPS questions)