Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
running_molpro_on_parallel_computers [2026/02/05 07:48] – [Running Molpro on parallel computers] wernerrunning_molpro_on_parallel_computers [2026/02/05 07:49] (current) – [Running Molpro on parallel computers] werner
Line 12: Line 12:
 Preallocating GA is not required with the ''mpi-pr'' runtime of GA or with the disk option. Preallocating GA is not required with the ''mpi-pr'' runtime of GA or with the disk option.
  
-Programs that use density fitting (e.g. df-hf, df-ks, df-multi, df-mp2) as well as pno Programs (pno-[lru]ccsd(t)[-F12|) can run on multiple nodes. However, a fast interconnect (e.g. Infiniband) is necessary to get reasonable performance. For most other methods, in particular conventional methods that use integrals stored on disk, multi-node calculations are disabled, since no performance gain can be expected and memory requirements could be exceedingly large.+Programs that use density fitting (e.g. df-hf, df-ks, df-multi, df-mp2) as well as pno Programs (pno-[lru]ccsd(t)[-F12|) can run on multiple computer nodes. However, a fast interconnect (e.g. Infiniband) is necessary to get reasonable performance. For most other methods, in particular conventional methods that use integrals stored on disk, multi-node calculations are disabled, since no performance gain can be expected and memory requirements could be exceedingly large.
 ===== GA Installation notes  ===== ===== GA Installation notes  =====