Firefly and PC GAMESS-related discussion club


 
Learn how to ask questions correctly  
 
 
We are NATO-free zone
 



Re: CASSCF and memory. Once again.

sanya
sanya@photonics.ru


You can try cistep=aldet, it requires less memory. In addition, new algorithm works better in parallel (use $DET distci=number_of_cores)

On Thu Jun 14 '12 6:39am, Solntsev Pasha wrote
----------------------------------------------
>Dear Alex.

>You already explained on this forum the problem with memory that rises from a big active space. I just want to clarify some things.

>I have a small molecule but a big active space - CASSCF(13,16).
>Some information from the output file is shown below:
>###################  Part From Output file ######################
>TOTAL NUMBER OF SHELLS              =   95
> TOTAL NUMBER OF BASIS FUNCTIONS     =  370
> NUMBER OF ELECTRONS                 =  119
> CHARGE OF MOLECULE                  =   -1
> STATE MULTIPLICITY                  =    2
> NUMBER OF OCCUPIED ORBITALS (ALPHA) =   60
> NUMBER OF OCCUPIED ORBITALS (BETA ) =   59
> TOTAL NUMBER OF ATOMS               =    6
> THE NUCLEAR REPULSION ENERGY IS     1097.1162359429
> TOTAL NUMBER OF CONTAMINANTS DROPPED=   61

> TOTAL NUMBER OF INTEGRALS =      10572
> NUMBER OF INTEGRALS/GROUP =      10572
> NUMBER OF INTEGRAL GROUPS =          1
> MAXIMUM RECORD SIZES ARE35395360 FOR UNIT 1135395360 FOR UNIT 12
>                         35395360 FOR UNIT 15   10572 FOR UNIT 16

>         -----------------------------------------
>          DAVIDSON METHOD CI-MATRIX DIAGONALIZATION
>                   WRITTEN BY STEVE ELBERT
>          -----------------------------------------
> NUMBER OF STATES REQUESTED =    20
> MAX. NUMB. OF EXPAN. VEC   =    41
> MAX. NUMB. IMPROVED STATES =    20
> MAX. NUMB. OF ITERATIONS   =   100
> CONVERGENCE CRITERION      =    1.0E-05

> CHOOSING OUT OF MEMORY STORAGE OF HAMILTONIAN,
> AND OUT OF MEMORY STORAGE OF TRIAL VECTORS.
> TO HOLD VECTORS IN MEMORY WOULD REQUIRE        -1 WORDS,
> SO FILES 12 AND 15 WILL BE USED TO STORE VECTORS.

> NUMBER OF WORDS AVAILABLE =  119989074
> NUMBER OF WORDS USED      = 1486611185
> ENERGY MATRIX BUFFER SIZE =        900
>***** ERROR: MEMORY REQUEST EXCEEDS MAXIMUM.
>  1486622111 WORDS REQUIRED    120000000 AVAILABLE.

>##################### end of the output file ####################

>Wave function contains 35395360 CSF.

>I understand that FF prints amount of memory per one core and amount i need really huge. But is it possible to decrease that number by increasing of the number of cores? Or may be we can play with parameters to overcome this problem. May be i can solve the problem with beta version of FF8. I will be appreciate for any help and suggestions.

>########### INPUT FILE ##################
> $SYSTEM MWORDS=120
>         fastf=.t.
>         nojac=100  $END
> $CONTRL wide=.t.
>         inttyp=hondo
>         icut=11
>         itol=30
>         d5=.t.
>         exetyp=run
>         maxit=200
>         SCFTYP=mcscf
>         ICHARG=-1  
>         MULT=2
>         ecp=read  $END
> $trans mptran=2
>        dirtrf=.t.
>        aoints=dist
>        altpar=.t.
>        mode=112  $end
> $smp call64=.t. $end
> $p2p p2p=.t. dlb=.t. $end
>! CASSCF(13,16) setup
> $drt  group=c1 nmcc=39 ndoc=6 nalp=1 nval=9 fors=.t.   $end  
> $mcscf cistep=guga soscf=.t. fors=.t. maxit=150 ntrack=15   $end
> $GUGEM pack2=.t. dirci=.t. fastci=.t.   $end
> $gugdia nstate=20 itermx=100   $end
> $gugdm iroot=1  $end
> $gugdm2 wstate(1)=1,1,1,1,1,1,1,1,1,1,-0  $end
>! end CASSCF
> $guess guess=moread norb=370 $end
> $SCF DIRSCF=.TRUE. FDIFF=.f. $END
> $DATA
>
>
>
>Pavel.
>
>
>


[ Previous ] [ Next ] [ Index ]           Thu Jun 14 '12 3:29pm
[ Reply ] [ Edit ] [ Delete ]           This message read 883 times