[molpro-user] Functions of the new MRCIC program and efficient use of the older MRCI program

[Peterson, Kirk]

Dear Christian,

at least in regards to point 3 below, it looks like your job is needing about 80GB of RAM (~134 MW = 1 GB).  Note that this is 80GB per core (the memory you request with the memory directive is also per core). So if you have 256 GB of RAM on your node, you probably don’t want to request more than 2-3 cores and then make sure to allocate the whole node for yourself (or run across nodes if you have something like infiniband).

best,

-Kirk

PS - perhaps for historical reasons, I always use the Davidson correction and not the Pople correction. 


> On Oct 28, 2016, at 9:07 AM, Christian Wiebeler <christian.wiebeler at mail.huji.ac.il> wrote:
> 
> Dear Molpro developers and users,
> 
> At the moment, I am using Molpro for MRCI calculations and have a couple of questions regarding the proper use and efficiency:
> 
> 1.    Is the MRCIC program still restricted to ground state calculations for each given symmetry, i.e. spatial and spin symmetry?
> 
> 
> 2.    After a MRCI calculation, several energy corrections are printed out. The manual and the paper discuss the different Davidson corrections, but the Pople corrections are not mentioned. Which correction do you recommend for the final excitation energy evaluation?
> 
> 
> 3.    I managed to run the MRCI calculations with 176 basis functions, 1232 configurations in reference space and 207 basis functions, 2076 configurations. However, a calculation with 239 basis functions and 6765 configurations is constantly crashing with the following error message:
> “insufficient memory available - require 10572865380 have 100680817 the request was for real words” or it just freezes without any output for several days. I have compute nodes with 40 cores and 256 GB memory available. How should I choose the optimal conditions for such a calculation to finish successfully? Maybe the memory per core is not sufficient and I should choose less cores?
> 
> 
> 
> 
> Best regards,
> Christian
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.molpro.net_mailman_listinfo_molpro-2Duser&d=DQIGaQ&c=C3yme8gMkxg_ihJNXS06ZyWk4EJm8LdrrvxQb-Je7sw&r=mqaXWsZ0ghQR-pXvPLjY_g&m=5ntYH_KGTOGMyAB5Glxwgr74TlA4ZyriqN28cjwBVno&s=IgaWGavS7dv3dEF29ocDfFx3dDNllQwOI2WbJaSYKrU&e=