as in a previous topic 'Sudden memory increase on Efield calculations'
I' m also experiencing such a sudden memoryincrease when simulating 4096
Si atoms.
No problem with 512 atoms. Siesta starts with about 1 GB memory, then
suddenly increases to about 10 GB. All the same running on 1,2,4 or 8 procs
(memory seems to be rather well distributed, with little overhead).
I compiled siesta with mpich2 and usual blacs scalapacks lapack gfortran
etc on a dell precision t7500 with
8 xeon cores. I' m testing on Si to check use on large systems. Run all
the tests ok.
Probably some large memory leakage ? Any idea how to get out of it ??
