Dear siesta users, can anybody explain me what the " Maximum dynamic memory allocated" means just before the self-consistency cycle begins? In parallel calculations (ParallelOverK, k-points: 4 4 1) when I use
1 CPU Maximum dynamic memory allocated=158 MB 2 CPU Maximum dynamic memory allocated=113 MB 4 CPU Maximum dynamic memory allocated=66 MB 8 CPU Maximum dynamic memory allocated=58 MB Why the memory allocated reduces when the number of CPU increases? Is this memory allocated per CPU or what? No explanation in the manual... If I use ParallelOverK and k-points: 2 2 1 then SIESTA runs with 1 or 2 CPUs, but it fails with 4 or 8 CPUs without explanation. Why it can be? Parallel over orbitals (8 atoms, all DZP) does not work either with 2, 4, 8 CPUS. Kind regards Ruslan