doubling memory requirements near the end of a computation
Posted: Wed Jul 23, 2014 7:12 am
we have observed that for a particular kind of computation vasp doubles the used memory for a couple of seconds at the end of the computation for a couple of seconds. Typical examples are when some non trivial is written at the end (volumetric data, energy spectra, etc.). As for 99.9% of the time this amount of memory is not needed it is a waste of resources. Usually, the process crashes and needs to be restarted with more memory.
is there any workaround or practical approach to avoid such issues?
any help or advice is well appreciated.
if needed the input files of an example of such a calculation can be provided.
is there any workaround or practical approach to avoid such issues?
any help or advice is well appreciated.
if needed the input files of an example of such a calculation can be provided.