-
Hello, In the link provided, NVIDIA has benchmarked various MD simulation software. I kind of want to replicate the result in the same way with DeePMD-kit on V100. So, first, I attempted to replicate the same result that they had. LAMMPS Code --
Command to run Lammps mpirun -np ${gpu_count} lmp -k on g ${gpu_count} -sf kk -pk kokkos cuda/aware on neigh full comm device binsize 2.8 -var x 8 -var y 4 -var z 8 -in ${input} But they are using kokkos package to produce result. Is it possible to produce same result as kokkos package in ATOM-Time Steps/s as a metric by running Lammps normally like --- mpiexec -np ${gpu_count} lmp -in ${input} What changes are required to achieve the above-mentioned result? How do I add the kokkos package to an existing conda DeePMD-kit LAMMPS installation? Is it possible to add additional packages only if I install LAMMPS as a plugin? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 10 replies
-
Why do you want to use kokkos? It looks like kokkos does not provide a plugin, so if you want to use kokkos you have to recompile LAMMPS. |
Beta Was this translation helpful? Give feedback.
Why do you want to use kokkos?
It looks like kokkos does not provide a plugin, so if you want to use kokkos you have to recompile LAMMPS.