.boundary() in parallel

nlinder · January 8, 2013, 07:50

Hi,

in my solver I have a loop through all my boundaries to extract those of type wall. This is realised by

Code:

forAll (mesh.boundary(), patchI)
{
    if (mesh.boundary()[patchI].type() == "wall")
    {
        wallPatchID = patchI;
        
        Info << "U.boundaryField()[wallPatchID].size(): " << U.boundaryField()[wallPatchID].size() << endl;
...

The output of the Info statement on a single core is correct:

Code:

U.boundaryField()[wallPatchID].size(): 2500

If I run it in parallel with two domains the output is

Code:

U.boundaryField()[wallPatchID].size(): 0

How can I make that work in parallel? What changes when I run this in parallel concerning the numbering of the boundary-patches?

Thanks in advance!
Nicklas

nlinder · January 11, 2013, 08:18

After playing around with the decomposition methods, I looks as if the code only "runs" on the processor0. Anyone knows what to do?

Thanks

MartinB · January 11, 2013, 08:26

Hi Nicklas,

use "Pout<< " instead of "Info<< ", then each processor will output its local size.

With "Info<< " only processor0 will print its value, and you must use a gSum command to gather the information from all processors to processor0 first.

Martin

nlinder · January 11, 2013, 08:33

Hi Martin,

thanks for the hint.. Ok, for the output it is clear. But if i e.g. want to set U as U^2 on a boundary I first have to "gSum" the velocity fields, then square them, and redistribute?

Nicklas

MartinB · January 11, 2013, 08:36

Since every processor calls the gSum, too, the sum is distributed already, I think. So every processor can calculate the U^2 value and set it locally.
You can check it with the Pout command again.

Martin

nlinder · January 11, 2013, 08:37

Allright, I'll try that, thanks for your help!

nlinder · January 11, 2013, 08:59

Sorry to bother again, but...
gSum returns only the sum.. I actually want a vector(or scalar)Field of a patch which is split to several processors. That seems to be sth different?!

Greetings
Nicklas

MartinB · January 11, 2013, 09:25

May be you can find the answer in this code snippet, where slave processors are sending data to the master processor with Pstream:
http://www.cfd-online.com/Forums/ope...tml#post282446

Martin

nlinder · January 11, 2013, 10:00

Hi Martin,

I think I now understood what was going wrong. I just added a condition, that the operations on the boundary are only performed, if it's size is greater than 0. No all warnings and Errors disappeared and it seems to work!
Thanks for your help!

Nicklas

January 8, 2013, 07:50	.boundary() in parallel	#1
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	Hi, in my solver I have a loop through all my boundaries to extract those of type wall. This is realised by Code: forAll (mesh.boundary(), patchI) { if (mesh.boundary()[patchI].type() == "wall") { wallPatchID = patchI; Info << "U.boundaryField()[wallPatchID].size(): " << U.boundaryField()[wallPatchID].size() << endl; ... The output of the Info statement on a single core is correct: Code: U.boundaryField()[wallPatchID].size(): 2500 If I run it in parallel with two domains the output is Code: U.boundaryField()[wallPatchID].size(): 0 How can I make that work in parallel? What changes when I run this in parallel concerning the numbering of the boundary-patches? Thanks in advance! Nicklas

January 11, 2013, 08:26		#3
MartinB Senior Member Martin Join Date: Oct 2009 Location: Aachen, Germany Posts: 255 Rep Power: 21	Hi Nicklas, use "Pout<< " instead of "Info<< ", then each processor will output its local size. With "Info<< " only processor0 will print its value, and you must use a gSum command to gather the information from all processors to processor0 first. Martin Loki Kerry likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Script to Run Parallel Jobs in Rocks Cluster	asaha	OpenFOAM Running, Solving & CFD	12	July 4, 2012 22:51
CFX parallel multi-node jobs fail w/ SLURM on Ubuntu 10.04	danieru	CFX	0	February 17, 2012 06:20
parallel performance on BX900	uzawa	OpenFOAM Installation	3	September 5, 2011 15:52
IcoFoam parallel woes	msrinath80	OpenFOAM Running, Solving & CFD	9	July 22, 2007 02:58
Parallel Computing Classes at San Diego Supercomputer Center Jan. 20-22	Amitava Majumdar	Main CFD Forum	0	January 5, 1999 12:00

January 11, 2013, 08:18		#2
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	After playing around with the decomposition methods, I looks as if the code only "runs" on the processor0. Anyone knows what to do? Thanks

January 11, 2013, 08:33		#4
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	Hi Martin, thanks for the hint.. Ok, for the output it is clear. But if i e.g. want to set U as U^2 on a boundary I first have to "gSum" the velocity fields, then square them, and redistribute? Nicklas

January 11, 2013, 08:36		#5
MartinB Senior Member Martin Join Date: Oct 2009 Location: Aachen, Germany Posts: 255 Rep Power: 21	Since every processor calls the gSum, too, the sum is distributed already, I think. So every processor can calculate the U^2 value and set it locally. You can check it with the Pout command again. Martin

January 11, 2013, 08:37		#6
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	Allright, I'll try that, thanks for your help!

January 11, 2013, 08:59		#7
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	Sorry to bother again, but... gSum returns only the sum.. I actually want a vector(or scalar)Field of a patch which is split to several processors. That seems to be sth different?! Greetings Nicklas

January 11, 2013, 09:25		#8
MartinB Senior Member Martin Join Date: Oct 2009 Location: Aachen, Germany Posts: 255 Rep Power: 21	May be you can find the answer in this code snippet, where slave processors are sending data to the master processor with Pstream: http://www.cfd-online.com/Forums/ope...tml#post282446 Martin

January 11, 2013, 10:00		#9
nlinder Member Nicklas Linder Join Date: Jul 2012 Location: Germany Posts: 35 Rep Power: 13	Hi Martin, I think I now understood what was going wrong. I just added a condition, that the operations on the boundary are only performed, if it's size is greater than 0. No all warnings and Errors disappeared and it seems to work! Thanks for your help! Nicklas