Memory problems with LMP2 optimization
Nick Wilson
N.Wilson at bham.ac.uk
Wed Oct 29 11:45:26 GMT 2003
Dear Rodolfo,
Try putting the line
memory,100,m
at the beginning of your input. This will allocate 100 million words of
memory for the calculation. If you still get "NOT ENOUGH MEMORY" error
messages then increase the value until you get "Failure in attempting
memory allocation" error messages and you still have "NOT ENOUGH MEMORY"
errors then the problem is too large for your system.
Best wishes,
Nick Wilson
Rodolfo Briones wrote:
> Dear Molpro Users:
>
> I ve been trying to perform a geometry optmization through LMP2. (Dual
> Athlon MP Machine, 1GB RAM, 17GB partition Scratch, Molpro 2002_6 binary
> distribution)
> But It seems there is a memory problem I still cannot solve.
>
> The input:
> _________________________________
> file,2,St_LMP2_opt_FAVTZ.wf;
> punch,St_LMP2_opt_FAVTZ.pun;
> geomtyp=xyz
> geometry={nosym;noorient;angstroms;
> 33
> MP2_NW_05_Mini_Site01 geometry
> }
>
> basis=VDZ,F=AVTZ
> hf;WF,132,1,0
>
> mp2;
> local,LOCAL=3,MERGEDOM=1,DELBAS=1;
> PRINT,DOMAINS
> direct,DMP2=1,DTRAF=0,PAGE_DTRAF=0
> SAVE,4000.2
>
> optg
> coord,3N
> ACTIVE,X1,Y1,Z1;
> __________________________________
>
>
> As the manual recommends also tested direct: DMP2=2 and PAGE_DTRAF=1
> but before it finish the first LMP2 cycle, the output sends the error
> message:
>
>
> D2EXT transformation finished. Total time: 1392.10 sec (Integrals:
> 382.55 sec, Transformation: 985.89 sec, Sort: 23.60 sec)
>
> Amplitudes will be saved to record 4000.2
>
> NOT ENOUGH MEMORY IN CCSD: INCREASE BY******** WORDS
>
> ERROR EXIT
> CURRENT STACK: MAIN
>
> ...
> DISK USED * 16.90 GB
> SF USED * 2.60 GB
>
>
>
> After the error 2 files of 2 GB remains in disk (eaf_T2400018118.TMPx001
> and eaf_T2400018117.TMP)
>
> What must I do to perform the calculation?
>
> Increase the space in HD?
> Decrease Gradient Thresholds?
> Is there any INPUT keywords I've forgotten or misplaced?
>
> Thanks in advance
>
> Rodolfo Briones Jerez
> Universidad de Chile
>
More information about the Molpro-user
mailing list