The ABINIT Users Mailing List ( CLOSED )

[Bousquet Eric <eric.bousquet@ulg.ac.be>]

Dear all,

When I use a relaxation with ionmov in PAW and I look the process with an easy "top" command, I remark that the RAM memory increases with no limit until the end of calculation!
Apparently, the RAM used increase a litlle bit at each SCF!
For some jobs with long relaxation, the RAM used can be up to 2Go for a BaTiO3 bulk!!!
It's like if a "buffer" is not set to zero.

Is somebody observes the same behaviour or can correct an error in my procedure?

I run my calculation with ABINIT v 4.6.4 on MAC G5 (compiled with xlf90, using makefile_macros.macosx.xlf_nc).
My main parameters are:
ixc         3
iscf        13
ionmov  2

This behaviour is observed on LDA BaTiO3, Pb and PbTiO3 bulk (using the PAW atomic data of the Web site for Ba, Ti and O).
If I try to decrease the mkmem variable (inferior to kpt) the job stop with a "Bus error".

Regards,

Eric