Run openFoam in Multithreads

chitharenjan · May 1, 2011, 04:20

Hi,

I am running OpenFoam in my PC. , which have following specifications

Operating System: Ubuntu 10.10 32 bit
Processor: i5, 2 cores and 4 threads

But whenever I am monitoring the openFoam processing, it shows it is only using single thread. I heard that there is a way to run openFoam simultaneously in all my cores. Please help me...

Thank you in advance.

santos · May 1, 2011, 06:57

Hi,

Make sure you run your solver with:

mpirun -np 4 <solvername> -parallel

Regards,
Jose

chitharenjan · May 1, 2011, 15:30

Thank you very much,

That worked perfectly.But I have some more doubts..

1, What is the sequence for running a parallel implementation. I mean for blockMesh and decomposePar, which one I should run first ?

2, I have used a stop watch to compare the speed gain in parallel processing from the normal processing. My processors were running at full load in parallel processing, but the speed gain was negligible. Why is it happening ? Do you have any suggestions in it ?

Waiting for your reply.

JunaidAhmad · May 2, 2011, 04:05

Can this Command works as it is on GPU base computers

Regards Junaid

akidess · May 2, 2011, 04:07

1. First mesh, then decompose.
2. There are lots of threads on the forum and the Internet in general on this topic, go have a look. In short, parallelization is no guarantee for speedup. Your problem has to be suitable for your parallel environment.

chitharenjan · May 2, 2011, 11:07

Thanks,

Comments were helpful.

Ashok

May 1, 2011, 04:20	Run openFoam in Multithreads	#1
chitharenjan New Member Ashok Join Date: Jul 2010 Location: India Posts: 20 Rep Power: 15	Hi, I am running OpenFoam in my PC. , which have following specifications Operating System: Ubuntu 10.10 32 bit Processor: i5, 2 cores and 4 threads But whenever I am monitoring the openFoam processing, it shows it is only using single thread. I heard that there is a way to run openFoam simultaneously in all my cores. Please help me... Thank you in advance.

May 1, 2011, 15:30	Sequence of Parallel Implimenation	#3
chitharenjan New Member Ashok Join Date: Jul 2010 Location: India Posts: 20 Rep Power: 15	Thank you very much, That worked perfectly.But I have some more doubts.. 1, What is the sequence for running a parallel implementation. I mean for *blockMesh* and *decomposePar*, which one I should run first ? 2, I have used a stop watch to compare the speed gain in parallel processing from the normal processing. My processors were running at full load in parallel processing, but the speed gain was negligible. Why is it happening ? Do you have any suggestions in it ? Waiting for your reply.

May 2, 2011, 11:07	Thanks	#6
chitharenjan New Member Ashok Join Date: Jul 2010 Location: India Posts: 20 Rep Power: 15	Thanks, Comments were helpful. Ashok

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
OpenFOAM 1.6.x - CentOS 5.3 x86_64	linnemann	OpenFOAM Installation	68	April 22, 2013 11:03
[mesh manipulation] createPatch / cyclicGgi / OpenFoam 1.5-dev	OFU	OpenFOAM Meshing & Mesh Conversion	0	June 16, 2010 04:36
Getting OpenFOAM to coexist with an existing JAVA VM	nik777	OpenFOAM Installation	5	February 22, 2007 07:21
Help I canbt run FoamX report Unable to contact name server error	cornerstone	OpenFOAM Installation	0	May 3, 2006 01:14
Run FOAM on a remote machine	luckyluke	OpenFOAM Installation	1	June 21, 2005 05:19

May 1, 2011, 06:57		#2
santos Senior Member Jose Luis Santos Join Date: Mar 2009 Location: Portugal Posts: 215 Rep Power: 18	Hi, Make sure you run your solver with: mpirun -np 4 <solvername> -parallel Regards, Jose

May 2, 2011, 04:05		#4
JunaidAhmad Member Junaid Ahmad Khan Join Date: Mar 2010 Location: Islamabad Posts: 43 Rep Power: 16	Can this Command works as it is on GPU base computers Regards Junaid

May 2, 2011, 04:07		#5
akidess Senior Member Anton Kidess Join Date: May 2009 Location: Germany Posts: 1,377 Rep Power: 29	1. First mesh, then decompose. 2. There are lots of threads on the forum and the Internet in general on this topic, go have a look. In short, parallelization is no guarantee for speedup. Your problem has to be suitable for your parallel environment.