[molpro-user] How Much Memory is Needed for Molpro?

[Kirk Peterson]

Dear Liming,

please remember that the memory you request on the memory directive is essentially a per-core request, so requesting more cores or nodes with the same memory request will never improve things. You have a fairly large calculation below and are only asking for about 7.5 Gb of RAM. I would suggest doubling what you’ve requested (i.e., 2000 Mwords) but only run it on 4 cores since that corresponds to a total aggregate memory request of about 60 Gb. Of course you could get more by running across nodes rather than on a single node, but I don’t know exactly what to suggest there without knowing more on how the particular queue system is configured. I think you have the information you need now to figure that out.

regards,

-Kirk

> On Mar 5, 2016, at 7:32 PM, 王黎明 <wanglm at scut.edu.cn> wrote:
> 
> Dear All,
> 
> We are having MEMORY problem when trying CCSD(T)-F12 on an open-shell species C6H7O3.
> 
> We have tried the calculations on nodes with 64 GB memory (12 cores or 24 cores). 
> In the input file, we have specified
>      MEMORY,1000,M
> and run molpro as
>      molpro -n 6 xx.inp 
>      molpro -n 4 xx.inp 
> or    molpro -n 4 -t 3 xx.inp
> We tried the calcs on SINGLE node. The working nodes were randomly assigned from a large array of nodes from a supercomputer system.
> 
> Here is my input file. We are using Molpro 2015 Binary version (Version 2015.1 linked Jan 11 2016):
> 
> ***,C6H6-OH-2OOs CCSD(T)-F12/cc-pVDZ-F12
> MEMORY,1000,M
> ANGSTROM
> GEOMTYP=xyz
> geometry={
> 16           ! Number of Atoms
> The C6H6-1OH-2OOs @ BH&HLYP/6-311++G(2df,2p)
> C,  0.4896916042,  0.957210998,  -0.5274539138
> C,  -0.5220129911,  1.4593284657,  0.4548488539
> O,  1.798982791,  1.3481087669,  -0.2327248195
> C,  0.3795090277,  -0.5419399373,  -0.8054683779
> C,  -1.6990294152,  0.8682102219,  0.5700955585
> C,  -1.0328015783,  -1.0219854803,  -0.8108838961
> O,  1.1911345017,  -1.3117952708,  0.1244774923
> C,  -1.9813663576,  -0.3546568177,  -0.1700943447
> O,  0.7106684183,  -1.3358694996,  1.3257252382
> H,  0.2582067535,  1.4334500447,  -1.4783897973
> H,  -0.2697031013,  2.3521132212,  0.9965739711
> H,  2.1187148014,  0.8644543054,  0.5212382213
> H,  0.8821806298,  -0.7552332209,  -1.737738857
> H,  -2.4566695723,  1.2630433973,  1.2206075858
> H,  -1.2415583588,  -1.9549910075,  -1.299754911
> H,  -2.985861153,  -0.736941187,  -0.1583860036
> }
> BASIS=cc-pVDZ-F12
> {UHF,MaxIT=512;}
> {RHF,MaxIT=512;}
> {UCCSD(T)-F12 SCALE_TRIP=1,MaxIT=200;}
> 
> The calculation finishes well in UHF and RHF, but in F12 part, we always got the following error message:
> ? Error
> ? actual memory layout of target op is different from the expected one
> ? The problem occurs in LoadF12Op
> 
> Another calc on a C10H17O system exits when starting CCSD(T) as
> insufficient memory available - require              103156065  have
>              63563208
> the request was for real wordsoo
> with "MEMORY,800,M" and "molpro -n 6 xx.inp" on nodes with 64GB RAM. It looks like I don't get enough memory.
> 
> My question is: How much memory do I need to run such calculations?
> 
> I notice no matter how much memory I put in "MEMORY, xxx, M", the calculations qucikly takes almost all the memory available.
> It looks to me that Molpro simply ignores the restriction from "MEMORY, xxx, M". Is there something here I did not set up correctly ? Can molpro run on a "memory reduced" mode?
> 
> Suggestions appreciated!
> 
> Liming Wang
> 
> --
> Liming Wang, Ph D
> School of Chemistry & Chemical Engineering
> South China University of Technology
> 381, Wushan Rd.
> Guangzhou, China 510640
> E-Mail: wanglm at scut.edu.cn
> ResearchGate: https://urldefense.proofpoint.com/v2/url?u=https-3A__www.researchgate.net_profile_Liming-5FWang6&d=CwICAg&c=C3yme8gMkxg_ihJNXS06ZyWk4EJm8LdrrvxQb-Je7sw&r=mqaXWsZ0ghQR-pXvPLjY_g&m=4cqFeuMYkQcDCM7eItVRyp7GjmEwXRwUszMhChuYrVU&s=b0e010nHBmOOVsC_Qyn__CmZM0i698AXQ5s-CEjWb3k&e= 
> 
> 
> 
> 
> _______________________________________________
> 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=CwICAg&c=C3yme8gMkxg_ihJNXS06ZyWk4EJm8LdrrvxQb-Je7sw&r=mqaXWsZ0ghQR-pXvPLjY_g&m=4cqFeuMYkQcDCM7eItVRyp7GjmEwXRwUszMhChuYrVU&s=9Irdx1ZSQgcBayAl6GpvQKKuxFZcxdOoGbu3Ljod2H8&e=