parallel computing doesn't speed up

najunaju · March 15, 2013, 08:10

dear fomers,

greetings!

I am running kepsilon model using pisofoam on bluff body with 2D simulation in a Intel core 2 quad processor with 8gb ram. I have four local processors and the simulation is too heavy that to complete 1 iteration it takes 25 seconds. now,

1. if I decompose into 2 or 4 processors (like, 2 1 1 or 2 2 1) the speed is same. doesn't change at all.

2. another surprising thing is, if I decompose into 6 or 8 processors and run through mpi commad, it runs but the speed is same though I have 4 processors. how come it runs in 6 or 8 processors?

I have to have around 1223987 interations which would take 10 to 15 days if I can't use parallel computing effectively.

please help me, please. I will be waiting.

Hiroshiman · March 15, 2013, 10:35

Hi !
A few questions :
What is the size of your domain ?
What was the exact command you used to run your case ? I used to forget the -parallel parameter at first.

najunaju · March 16, 2013, 00:43

thank u hiroshiman for ur inquiry.

my domain is 41Dx25D and width is extruded by openfoam, in the width direction there are single cell. I have total 6,00,000 cells to maintain y plus around 1 to 3. D is the height of domain.

I used simple dicomposition with the following command,

mpirun -np 4 pisoFoam -parallel

please help me, if u can. I am running out of time.

Hiroshiman · March 18, 2013, 09:55

Hi,
with 6E6 cells, the problem should be large enough to be run over a few cpus.

Could you post the output of the solver on the few few first iterations to have a better idea of what's happening ?

najunaju · March 21, 2013, 03:39

Dear Hiroshiman,

Thank you for your query. I have reduced mesh number and running now, I think this allowable for me.

Thank you again.

March 15, 2013, 08:10	parallel computing doesn't speed up	#1
najunaju New Member Md. Naimul Haque Join Date: Jan 2013 Posts: 14 Rep Power: 13	dear fomers, greetings! I am running kepsilon model using pisofoam on bluff body with 2D simulation in a Intel core 2 quad processor with 8gb ram. I have four local processors and the simulation is too heavy that to complete 1 iteration it takes 25 seconds. now, 1. if I decompose into 2 or 4 processors (like, 2 1 1 or 2 2 1) the speed is same. doesn't change at all. 2. another surprising thing is, if I decompose into 6 or 8 processors and run through mpi commad, it runs but the speed is same though I have 4 processors. how come it runs in 6 or 8 processors? I have to have around 1223987 interations which would take 10 to 15 days if I can't use parallel computing effectively. please help me, please. I will be waiting.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
unchangeable continuity residuals in parallel computing	wlt_1985	FLUENT	0	August 1, 2011 12:15
Diffusion equation solved using Parallel Computing	Sachin Paramane	Main CFD Forum	0	June 11, 2007 23:48
Parallel Computing on Multi-Core Processors	Upgrading Hardware	CFX	6	June 7, 2007 15:54
Parallel Computing	peter	Main CFD Forum	7	May 15, 2006 09:53
Parallel Computing Classes at San Diego Supercomputer Center Jan. 20-22	Amitava Majumdar	Main CFD Forum	0	January 5, 1999 12:00

March 15, 2013, 10:35		#2
Hiroshiman Member Join Date: Sep 2012 Posts: 51 Rep Power: 13	Hi ! A few questions : What is the size of your domain ? What was the exact command you used to run your case ? I used to forget the -parallel parameter at first.

March 16, 2013, 00:43		#3
najunaju New Member Md. Naimul Haque Join Date: Jan 2013 Posts: 14 Rep Power: 13	thank u hiroshiman for ur inquiry. my domain is 41Dx25D and width is extruded by openfoam, in the width direction there are single cell. I have total 6,00,000 cells to maintain y plus around 1 to 3. D is the height of domain. I used simple dicomposition with the following command, mpirun -np 4 pisoFoam -parallel please help me, if u can. I am running out of time.

March 18, 2013, 09:55		#4
Hiroshiman Member Join Date: Sep 2012 Posts: 51 Rep Power: 13	Hi, with 6E6 cells, the problem should be large enough to be run over a few cpus. Could you post the output of the solver on the few few first iterations to have a better idea of what's happening ?

March 21, 2013, 03:39		#5
najunaju New Member Md. Naimul Haque Join Date: Jan 2013 Posts: 14 Rep Power: 13	Dear Hiroshiman, Thank you for your query. I have reduced mesh number and running now, I think this allowable for me. Thank you again.