Memory issue for spin-polarized calculation
Posted: Tue Aug 23, 2011 5:09 pm
I am trying to carry out a spin-polarized calculation for a largish supercell of 125 atoms. I am persistently running into problems during the first ionic step, as I am apparently running out of memory. Here is an example of the error which is written to standard output:
RMM: 18 -0.697604751458E+03 0.19404E-02 -0.14726E-03122011 0.698E-02 0.448E-01
RMM: 19 -0.697604761641E+03 -0.10183E-04 -0.52393E-04108118 0.416E-02
forrtl: severe (41): insufficient virtual memory
Image PC Routine Line Source
libintlc.so.5 00002BA99D6D23A6 Unknown Unknown Unknown
libintlc.so.5 00002BA99D6D1338 Unknown Unknown Unknown
libifcore.so.5 00002BA99D08A29E Unknown Unknown Unknown
libifcore.so.5 00002BA99D014DA4 Unknown Unknown Unknown
libifcore.so.5 00002BA99D0516EF Unknown Unknown Unknown
vasp 0000000000436354 Unknown Unknown Unknown
vasp 000000000040BB22 Unknown Unknown Unknown
libc.so.6 000000327BA1C3FB Unknown Unknown Unknown
vasp 000000000040BA6A Unknown Unknown Unknown
I am running on 128 processors, starting from a WAVECAR and a CHGCAR file from a converged non-spin polarized calculation. I have had a look at the page 'Not enough memory, what to do' in the manual, but so far I have not been able to resolve this issue. I am using a value for NPAR which is small compared to the number of processors (e.g. NPAR=4), but the calculation becomes prohibitively slow if I reduce NPAR to NPAR=1. Here is an example of the INCAR file I have been using for this calculation:
INCAR
SYSTEM=Ni fcc + P
PREC=Nor ISPIN=2
ICHARG=1 ISTART=1
MAGMOM=124*1.6 0
ISYM = 2
NPAR=4
LCHARG=.FALSE.
ENCUT=350.0
ENAUG=707.934
ISIF=3
IBRION=2
NSW=999
ISMEAR=1
SIGMA=0.05
ALGO=VERYFAST
LREAL=Auto
Is there some way I can reduce the memory requirements?
RMM: 18 -0.697604751458E+03 0.19404E-02 -0.14726E-03122011 0.698E-02 0.448E-01
RMM: 19 -0.697604761641E+03 -0.10183E-04 -0.52393E-04108118 0.416E-02
forrtl: severe (41): insufficient virtual memory
Image PC Routine Line Source
libintlc.so.5 00002BA99D6D23A6 Unknown Unknown Unknown
libintlc.so.5 00002BA99D6D1338 Unknown Unknown Unknown
libifcore.so.5 00002BA99D08A29E Unknown Unknown Unknown
libifcore.so.5 00002BA99D014DA4 Unknown Unknown Unknown
libifcore.so.5 00002BA99D0516EF Unknown Unknown Unknown
vasp 0000000000436354 Unknown Unknown Unknown
vasp 000000000040BB22 Unknown Unknown Unknown
libc.so.6 000000327BA1C3FB Unknown Unknown Unknown
vasp 000000000040BA6A Unknown Unknown Unknown
I am running on 128 processors, starting from a WAVECAR and a CHGCAR file from a converged non-spin polarized calculation. I have had a look at the page 'Not enough memory, what to do' in the manual, but so far I have not been able to resolve this issue. I am using a value for NPAR which is small compared to the number of processors (e.g. NPAR=4), but the calculation becomes prohibitively slow if I reduce NPAR to NPAR=1. Here is an example of the INCAR file I have been using for this calculation:
INCAR
SYSTEM=Ni fcc + P
PREC=Nor ISPIN=2
ICHARG=1 ISTART=1
MAGMOM=124*1.6 0
ISYM = 2
NPAR=4
LCHARG=.FALSE.
ENCUT=350.0
ENAUG=707.934
ISIF=3
IBRION=2
NSW=999
ISMEAR=1
SIGMA=0.05
ALGO=VERYFAST
LREAL=Auto
Is there some way I can reduce the memory requirements?