Firefly and PC GAMESS-related discussion club


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



Re^4: Problem running XMCQDPT2 calculation with large (16e,16o) active space

Martial Boggio-Pasqua
martial.boggio@irsamc.ups-tlse.fr


Dear Alex,

I have now tried to run the XMCQDPT2 calculation with a (16e,16o) active space but for a 4-state problem. I am running into the same problem as before: the MQCACI complains that the MEMORY REQUEST EXCEEDS MAXIMUM. 783747472 WORDS REQUIRED † †300000000 AVAILABLE.
I am not sure if I can reduce any further the iselct(2) value without compromising the accuracy. Do you have any suggestion on how to run such a calculation? I attach the input and ouput files.

Best regards,
Martial

On Mon Jan 29 '18 5:18pm, Martial Boggio-Pasqua wrote
-----------------------------------------------------
>Dear Alex,

>many thanks for your answer and solution to my problem. I have run the job with your modified input file and it works. I suppose that I have to check the convergence of the transition energies I get depending on the value of the THRWGT variable which reduces the number of CSFs used in the XMCQDPT2 calculation. What is the meaning of the 2000000 number after iselct(1)=-4? Is it the maximum allowed number of CSFs to be selected?

>On Thu Jan 25 '18 2:59am, Alex Granovsky wrote
>----------------------------------------------
>>Dear Martial,

>>I forgot to mention that you may also need to reduce the number of threads to be used in calculations.

>>Kind regards,
>>Alex Granovsky

>>On Thu Jan 25 '18 2:50am, Alex Granovsky wrote
>>----------------------------------------------
>>>Dear Martial,
>>>
>>>Iíve attached the modified head of your input file to this post.

>>>Basically, to handle CAS(16e,16o) using current XMCQDPT2 implementation,
>>>you need to dramatically reduce MXBASE and to use CSFs selection.

>>>Note, the MQCACI (i.e. CAS CI) step is always performed using single
>>>CPU core and thus this step will time consuming. This is a limitation
>>>of present code which will be removed in the future releases.
>>>
>>>Kind regards,
>>>Alex Granovsky

>>>On Wed Jan 24 '18 6:53pm, Martial Boggio-Pasqua wrote
>>>-----------------------------------------------------
>>>>I have been using Firefly recently to run CASSCF and XMCQDPT2 calculations on a medium-sized organic photochromic system. This system requires a large active space: (14e,14o) to study its absorption properties and (16e,16o) to study its photoswitching mechanism. While I have been able to run XMCQDPT2 with a (14e,14o) active space, I am not able to run a similar job with a (16e,16o) active space. I cannot circumvent the memory limit in mwords. I get the following error message in MQCACI: ERROR: MEMORY REQUEST EXCEEDS MAXIMUM. 729888763 WORDS REQUIRED † †460000000 AVAILABLE.

>>>>Maybe someone can tell me if this calculation is possible or not. I have enclosed an input file in which the CASSCF(16,16) part is skipped as the CASSCF orbitals have been saved in a previous job.

>>>>I would be extremely grateful if I can get some assistance with this issue as the whole series of calculations I am planning to perform require this large (16e,16o) active space.

>>>>I am using the 8.2.0 version.

>>>>Many thanks in advance!
>>>>Martial
>>>>

This message contains the 4934 kb attachment
[ dhp_xmcqd16pt2_vtz_S0min_Ag_stv4.tar.gz ]


[ Previous ] [ Next ] [ Index ]           Mon Feb 5 '18 12:18pm
[ Reply ] [ Edit ] [ Delete ]           This message read 79 times