This repo contains a collection of benchmark calculations for the MRChem program on different machines, using a variable number of MPI processes and OpenMP threads/process.
The following figure presents timings for a full SCF (PBE) optimization on alkane
molecules of increasing size. The calculations were performed in a weak-scaling manner,
adding resources as the system gets bigger, in particular one compute node per 10 carbon
atoms in the chain. The calculations were performed on Betzy, using 8 MPI tasks per node
and 16 threads per task. The numerical precision was set to world_prec = 1.0e-5, and
the orbitals were converged to orbital_thrs = 1.0e-3.
