[Pw_forum] memory problem

[Paolo Giannozzi]

Prasenjit Ghosh wrote:

> 507.1 Mb was the memory required at the beginning of one of the scf cycles.
> Then the job was exited because the calculation time was over.......Then 
> I restart the calculation and the memory requirement jumps to  1867.8 Mb

so there is something strange when restarting.

> Do you mean that from one scf cycle to another scf cycle, the memory can 
> increase while for different scf steps in an scf cycle it should take 
> more or less the same max. memory??

it depends on what you call a "cycle" and what you call a "step" ...
Anyway: actually the memory needed by self-consistency is allocated
at the beginning of the scf calculation, not during it, so the
allocated memory should never increase

P
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy