PAW & non-collinear & paral_kgb
Posted: Thu Feb 09, 2012 7:42 pm
Dear all
I’m performing several non-collinear calculations using PAW basis-sets. Since I’m dealing with fairly big systems I thought to speed up the calculations using the K-point, G-vectors and Bands parallelization that is available since version 6.10 has been released. I have found that the number of SCF iterations that abinit can perform depends on the amount of memory available. I mean that during the SCF loop the amount of memory requested by abinit increases with the number of the SCF iterations already done. It seems as if abint keeps in memory all the data since the beginning of the run. I experienced this problem only in non-collinear calculations with PAW and paral_kgb. Does someone know if this indeed happens, and if it does, what is the reason?
Thanks in advance for your help.
Bests.
Luca Sementa.
IPCF-CNR
Via G. Moruzzi 1
56124 Pisa (ITALY)
I’m performing several non-collinear calculations using PAW basis-sets. Since I’m dealing with fairly big systems I thought to speed up the calculations using the K-point, G-vectors and Bands parallelization that is available since version 6.10 has been released. I have found that the number of SCF iterations that abinit can perform depends on the amount of memory available. I mean that during the SCF loop the amount of memory requested by abinit increases with the number of the SCF iterations already done. It seems as if abint keeps in memory all the data since the beginning of the run. I experienced this problem only in non-collinear calculations with PAW and paral_kgb. Does someone know if this indeed happens, and if it does, what is the reason?
Thanks in advance for your help.
Bests.
Luca Sementa.
IPCF-CNR
Via G. Moruzzi 1
56124 Pisa (ITALY)