TURBOMOLE Users Forum

TURBOMOLE Modules => Mpgrad and Rimp2 => Topic started by: golden on April 15, 2013, 05:47:43 PM

Title: parallel rimp2 optimization
Post by: golden on April 15, 2013, 05:47:43 PM
Hi All,
  As specified in the tutorial i prepared my input using Define,  then ran a dscf calculation to get the MO's.

Then I ran
mp2prep -e -p 2000 -m 700

After that to optimize the structure I ran
nohup jobex -ri  -c 1000 -energy 8 -level mp2 > jobexmp2.out &

during the optimization I checked the process, and when dscf step is running it uses 10 procs,
  PID USER      PR  NI  VIRT  RES  SHR S  %CPU %MEM     TIME+ COMMAND
9591 golden  20   0  328m 185m 3084 R 991.6  1.5  38:38.26 dscf_smp

But when the mp2 is running it only uses 1 proc
  PID USER      PR  NI  VIRT  RES  SHR S  %CPU %MEM     TIME+ COMMAND                                                                     
 9798 golden  20   0  900m 575m 2452 R 104.1  4.8   7:00.48 rimp2     


I assumed ri-mp2 program can be parallel, how can I make it to run in 10 procs or multiple procs?

Thanks




Title: Re: parallel rimp2 optimization
Post by: Arnim on April 15, 2013, 11:51:41 PM
Hi,

the rimp2 program does not run in parallel, the ricc2 program does.

BTW: The functionalities of mpgrad, rimp2, and ricc2 are summarized in Chapter 8.1 of the manual.

Cheers,

Arnim