|
[Sponsors] |
the speed of parallel computation for over 3 million grid is very slow |
|
LinkBack | Thread Tools | Search this Thread | Display Modes |
June 17, 2014, 05:18 |
the speed of parallel computation for over 3 million grid is very slow
|
#1 |
Member
Jianming Liu
Join Date: Mar 2009
Location: China
Posts: 71
Rep Power: 17 |
Dear developer of SU2
I am doing a parallel computation with 3 million grid. As a start, I use 1 node with 8 cores to run. But the speed is so slow. After 5-6 hours, it only run 5 time steps. In fact I also test 5 nodes (per node with 8 cores) to run. And the speed is still so slow. So for SU2 _v3.0, which I used, is there any deficiency for parallel computation of SU2v3.0. Thank you very much. Jianming As a explanation, I also paste the PBS job file as follows: ############################################ #PBS -l select=1:ncpus=8:mem=14gb #Request hours of walltime # #PBS -l walltime=160:00:00 # #PBS -j oe #PBS -o BX.log #PBS -N BXcase1 # # cd /home/jamesliu/WingCase1 echo " " echo " " echo "Job started on `hostname` at `date`" /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_DDC turb_DW.cfg /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_CFD turb_DW.cfg /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_SOL turb_DW.cfg echo " " echo "Job Ended at `date`" echo " " Last edited by liujmljm; June 18, 2014 at 06:49. |
|
June 19, 2014, 17:44 |
|
#2 |
Member
Jianming Liu
Join Date: Mar 2009
Location: China
Posts: 71
Rep Power: 17 |
I am sorry. Maybe the reason is our cluster is too old.
|
|
June 26, 2014, 00:05 |
|
#3 |
Super Moderator
Francisco Palacios
Join Date: Jan 2013
Location: Long Beach, CA
Posts: 404
Rep Power: 15 |
There is an Intel Parallel Computing Center focused on SU2:
https://software.intel.com/en-us/art...cs-at-stanford We are working very hard to be sure that SU2 v4.0 will be a reference in CFD performance using the latest Intel Technology. Best, Francisco |
|
June 26, 2014, 11:44 |
|
#4 |
New Member
Giulio
Join Date: Apr 2014
Location: Milano
Posts: 17
Rep Power: 12 |
Hi,
I'm trying to run SU2 on my cluster and, like liujmljm, I'm using PBS too. With the same bash as him, I'm able to lunch the job without any error from PBS, but the job do not start. If I look at SU2_PRT log it says that SU2_PRT do not exist, but this is strange since I'm able to lunch simulations overriding the PBS. Any ideas? Best regards, Giulio
__________________
Giulio Gori Phd candidate, Politecnico di Milano |
|
June 26, 2014, 12:02 |
|
#5 | |
Member
Jianming Liu
Join Date: Mar 2009
Location: China
Posts: 71
Rep Power: 17 |
Quote:
The olde one (#PBS -l select=1:ncpus=8:mem=14gb) in our cluster, it doesn't work. So I use the following. ############################################ #PBS -l nodes=1: ppn=8 #Request hours of walltime # #PBS -l walltime=160:00:00 # #PBS -j oe #PBS -o BX.log #PBS -N BXcase1 # # cd /home/jamesliu/WingCase1 echo " " echo " " echo "Job started on `hostname` at `date`" /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_PRT turb_DW.cfg /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_CFD turb_DW.cfg /opt/sgi/mpt/mpt-2.07/bin/mpirun -np 8 SU2_SOL turb_DW.cfg echo " " echo "Job Ended at `date`" echo " " |
||
June 26, 2014, 12:13 |
|
#6 |
New Member
Giulio
Join Date: Apr 2014
Location: Milano
Posts: 17
Rep Power: 12 |
Thank you for your reply.
Unfortunately this not worked too, maybe is because I've installed SU2 locally. You've installed it locally too? Giulio
__________________
Giulio Gori Phd candidate, Politecnico di Milano |
|
June 26, 2014, 14:06 |
|
#7 |
Member
Jianming Liu
Join Date: Mar 2009
Location: China
Posts: 71
Rep Power: 17 |
||
June 27, 2014, 16:45 |
|
#8 |
Member
Eduardo Molina
Join Date: Sep 2010
Location: Brazil
Posts: 35
Rep Power: 16 |
Hi Giulio.
Can you try: ssh node mpirun -np 8 /pathtoSU2/SU2_PRT your.cfg If you are using sgi mpt (our case): mpirun node -np 8 SU2_PRT your.cfg mpirun node -np 8 SU2_CFD your.cfg (work for me) Both lines I put in our PBS job file and is working. Regards. |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Installation issues for parallel computation | Akash C | SU2 Installation | 1 | June 21, 2013 05:26 |
problem in the CFX12.1 parallel computation | BalanceChen | ANSYS | 2 | July 7, 2011 10:26 |
Parallel computation using NUMECA 6.1 | BalanceChen | Fidelity CFD | 1 | June 5, 2011 06:24 |
Increase speed of parallel computation | Purushothama | Siemens | 2 | November 30, 2010 14:51 |
Grid Independent Solution | Chuck Leakeas | Main CFD Forum | 2 | May 26, 2000 11:18 |