Firefly and PC GAMESS-related discussion club

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

Re^3: Speeding up ALTTRF during XMCQDPT2 calculations


Dear Alex,

Thank you for this information. I'll try if parallel execution will improve performance.

I have one question regarding skipping the MQCACI step. Normally, I run the MCSCF part in parallel with

to punch orbitals with extra accuracy, and then do the XMCQDPT2 part separately in multithreading mode with
to read in the converged orbitals and skip the MCSCF step. Am I correct in assuming that in this scenario it is not possible to skip the MQCACI step?

Kind regards,

On Sun Sep 16 '12 4:49pm, Alex Granovsky wrote
>Dear Thom,

>>Finally, with Firefly v. 8.0.0 it is possible to get rid of the
>>first MQCACI and integral transformation procedures for some types
>>of XMCQDPT2 jobs. These types are, however, the most typically
>>used ones. I'll document this in my next post to this thread.

>Assume the averaging is exactly the same in both SA-CASSCF and
>XMCQDPT2 procedures i.e. the weights defined by the WSTATE arrays
>are the exactly the same and refer to the same states. Provided
>this is indeed the case, one can avoid the first integral transformation
>and CASCI procedure (performed by the routine MQCACI) by overlapping
>part of calculations with those performed at the MCSCF stage and
>reusing them at XMCQDPT2 stage.

>To use this feature, one needs to direct MCSCF code to form
>semi-canonical Fock orbitals and to save core and active Fock
>operators. This can be done as follows:


 $mcscf iforb=.t. $end

>One also needs to instruct QDPT code to reuse this information:


 $xmcqdpt iforb(1)=-1,1,1 $end

>This will do the trick.  

>Kind regards,


[ Previous ] [ Next ] [ Index ]           Sun Sep 16 '12 7:57pm
[ Reply ] [ Edit ] [ Delete ]           This message read 756 times