|
[Sponsors] |
April 16, 2013, 03:56 |
parallel ReconstructPar with mpi?
|
#1 |
Member
Luca
Join Date: Mar 2013
Posts: 68
Rep Power: 13 |
Hi everybody.
I'm running a quite big case on a supercomputer. The case runs pretty fast, but now i'm stucked in a bottle-neck: reconstructPar. I need to reconstruct about 900 timesteps with 3.7 milion cells. With serial computing this takes more times that the solution! I found a this topic that should help in my situation. But i can't understand how to use it with MPIRUN and PBS. Can someone help me please? Thank you very much |
|
April 16, 2013, 05:19 |
|
#2 |
Senior Member
ATM
Join Date: May 2009
Location: United States
Posts: 104
Rep Power: 16 |
Hello Luca,
If you are asking about how to run a script in batch mode, I think this would work #PBS -N Jobname #PBS -l walltime=02:00:00 #PBS -l nodes=1 : ppn=1 #PBS -o $PBS_JOBNAME.log cd $PBS_O_WORKDIR ./scriptname The "scriptname" should be the script you should use from that thread you posted. Make sure its in your working directory. BEfore you submit the job I recommend testing if the script runs in your home directory directly. If it runs there, it should do so in batch mode as well. I have used the above for the .exe files I had created - So I guess this should work for a script/any executable file too. Let me know. Regards Atm |
|
April 16, 2013, 05:27 |
|
#3 |
Senior Member
Pablo Higuera
Join Date: Jan 2011
Location: Auckland
Posts: 627
Rep Power: 19 |
Hi Luca,
do you really need to reconstruct the case? I don't know if you are aware that most of the decomposed cases can be postprocessed while still being decomposed. If the issue is that you cannot open it with paraview try: "paraFoam -builtin". Having 3.7 million cells is not such a huge case to impede paraview from loading it at once ;-) Regarding the reconstructPar question, you can see it from a simpler point of view, it only runs in serial, but you can send any simulataneous number of processes to shorten the time. For example, if you send 2 processes the wait time will become more or less a half. You can easily do this with the -time instruction: Code:
reconstructPar -time '0:450' reconstructPar -time '451:900' Best, Pablo |
|
April 16, 2013, 05:27 |
|
#4 |
Member
Luca
Join Date: Mar 2013
Posts: 68
Rep Power: 13 |
Hi Atm,
I'm sorry but that's not my problem. I already know how to run scripts in batch mode. What i'm asking is how to reconstruct big cases with more than one CPU. I found a script wrote by a user but i don't understand how to run it. Since that thread was 3 years old i started a new one. I'm not experienced enough to write a batch script to automatize it by my own. Is there any solution? Is it possible that i can run a case with 1200 CPU and then i have to reconstruct it with only 1? My situation right know is that i need 4h to run a case, and then i take 1 day to reconstruct it. Quite silly... |
|
April 16, 2013, 05:31 |
|
#5 |
Member
Luca
Join Date: Mar 2013
Posts: 68
Rep Power: 13 |
Thanks Pablo,
No: i didn't know i could run paraView on a decomposed case. How can i do it? I haven't seen nothing pointing in that direction neither on the User Guide, now typing paraFoam -help in the bash. Can you please explain me how to do it or just give me a link? Thank you very much |
|
April 16, 2013, 05:33 |
|
#6 |
Senior Member
Pablo Higuera
Join Date: Jan 2011
Location: Auckland
Posts: 627
Rep Power: 19 |
paraFoam -builtin
;-) |
|
April 16, 2013, 09:08 |
|
#7 |
Member
Luca
Join Date: Mar 2013
Posts: 68
Rep Power: 13 |
As soon as i'll be back to work i'll check this.
Thank you very much |
|
February 5, 2018, 06:13 |
|
#8 |
New Member
Ali Aghaei
Join Date: Oct 2014
Posts: 12
Rep Power: 11 |
Hi! using this command, I got the following error message:
unknown option/argument: '-builtin' |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
A question about HP MPI Local Parallel | nucfusion | CFX | 2 | March 15, 2013 03:07 |
HP MPI warning...Distributed parallel processing | Peter | CFX | 10 | May 14, 2011 07:17 |
Error using LaunderGibsonRSTM on SGI ALTIX 4700 | jaswi | OpenFOAM | 2 | April 29, 2008 11:54 |
Is Testsuite on the way or not | lakeat | OpenFOAM Installation | 6 | April 28, 2008 12:12 |
MPI and parallel computation | Wang | Main CFD Forum | 7 | April 15, 2004 12:25 |