Alex Granovsky
gran@classic.chem.msu.su
>In general, it is advisable to perform CASSCF calculation in parallel on as many nodes as possible (32 cores for 1000 basis functions and up to 100000 CSFs is OK), because MCQDPT can run only on one node in multi-threading mode, while CASSCF in this mode will be very slow. Next, MOREAD the OPTIMIZED ORBITALS for MCQDPT run. MCQDPT will use settings from $DET of $DRT groups, so that the input in $MCQDPT group is reduced to INORB=1 (means use of CASSCF MOs as is) KSTATE(1)=.... EDSHFT=0.02 (reasonable value of ISA energy denominator shift).
One more comment here. If you split calculations, it is advisable
to set $contrl wide=.t. $end to punch orbitals with extra high
precision.
Regards,
Alex Granovsky
![[ Previous ]](/ceilidh-icons/ceil_prv.gif){kind=icon}
![[ Next ]](/ceilidh-icons/ceil_nxt.gif){kind=icon}
![[ Index ]](/ceilidh-icons/ceil_ind.gif){kind=icon}
Tue Oct 19 '10 10:00pm![[ Reply ]](/ceilidh-icons/ceil_rep.gif){kind=icon}
![[ Edit ]](/ceilidh-icons/ceil_edt.gif){kind=icon}
![[ Delete ]](/ceilidh-icons/ceil_del.gif){kind=icon}
This message read 897 times