decomposing a grid in parallel

deeps · January 23, 2014, 08:09

Dear All,

If we want to run a simulation in parallel, we can do so in SU2 using the following :

parallel_computation.py -f file -p np

How about only partitioning a grid file in parallel? Lets say I have a 21M grid and I want to decompose it in 40 parts using 20 processors. Due to the large size of the file if I attempt doing this on a single processor, it would not run due to memory issues. Also, I tried:

mpirun -n 40 SU2_DDC file_name

but it still does the partitioning using 1 processor. (Although it shows 40 processors being used but entire memory is used only for 1 processor)

Is there a way of partitioning grids in parallel using n processors ? This would help in dealing with large grid sizes.

Thanks and regards,
Deepanshu.

rktchip · January 28, 2014, 15:32

SU2_DDC only works serially, although it can be called through mpirun as you show in order to carry a common interface. Parallel decomposition requires development to use a library called parmetis, in case anyone is interested in making this work!

deeps · January 29, 2014, 01:01

Dear Trent,

Thank you for your reply.

Is there any way to decompose large grids? Are there any variable settings that would help to partition large grids.

Regards,
Deepanshu.

January 23, 2014, 08:09	decomposing a grid in parallel	#1
deeps New Member Join Date: Nov 2009 Posts: 19 Rep Power: 16	Dear All, If we want to run a simulation in parallel, we can do so in SU2 using the following : parallel_computation.py -f file -p np How about only partitioning a grid file in parallel? Lets say I have a 21M grid and I want to decompose it in 40 parts using 20 processors. Due to the large size of the file if I attempt doing this on a single processor, it would not run due to memory issues. Also, I tried: mpirun -n 40 SU2_DDC file_name but it still does the partitioning using 1 processor. (Although it shows 40 processors being used but entire memory is used only for 1 processor) Is there a way of partitioning grids in parallel using n processors ? This would help in dealing with large grid sizes. Thanks and regards, Deepanshu.

January 29, 2014, 01:01	decomposing a grid in parallel	#3
deeps New Member Join Date: Nov 2009 Posts: 19 Rep Power: 16	Dear Trent, Thank you for your reply. Is there any way to decompose large grids? Are there any variable settings that would help to partition large grids. Regards, Deepanshu.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Can not run OpenFOAM in parallel in clusters, help!	ripperjack	OpenFOAM Running, Solving & CFD	5	May 6, 2014 15:25
parallel Grief: BoundaryFields ok in single CPU but NOT in Parallel	JR22	OpenFOAM Running, Solving & CFD	2	April 19, 2013 16:49
Parallel partion lines on grid	Allan Walsh	FLUENT	1	January 12, 2009 18:09
grid adaption on parallel	Nurul Murad	FLUENT	0	May 28, 2002 03:34
Combustion Convergence problems	Art Stretton	Phoenics	5	April 2, 2002 05:59

January 28, 2014, 15:32		#2
rktchip Member Trent Lukaczyk Join Date: Feb 2011 Location: Stanford, CA Posts: 75 Rep Power: 15	SU2_DDC only works serially, although it can be called through mpirun as you show in order to carry a common interface. Parallel decomposition requires development to use a library called parmetis, in case anyone is interested in making this work!