[lovecraft22]

Hi all;
if you try and run the motorbike tutorial in parallel you'll get an error about the motorbike patch missing within the processors folders.

My error is this one:

keyword cilindro_patch0 is undefined in dictionary "/home/ubuntu/Desktop/condivisa/test/processor0/0/p::boundaryField"

It is actually missing within any of the boundary files held in the 0 folder of every processor folder after running decomposePar.

Now, I read somewhere on the forum that the problem is due to decomposePar which cannot handle all the "#include" inside the boundary files so I got rid of those but… again… the same error…

I enclose my test case. Is a simple cylinder made to speed up the whole process. I started from the motorbike tutorial.

Can anybody please help me understanding how a case needs to be set up to be run in parallel?

Thank you!

[lovecraft22]

The discussion I was referring is this one:
http://www.cfd-online.com/Forums/ope...omposepar.html

So is using changeDictionary the only way to achieve this? There's must be a way of defining the boundary conditions that doesn't require for the changeDictionary command to be run…


Also, I found out that if I run decomposePar after snappyHexMesh then the error disappears… I can run the solver in parallel then but it would be nice to run snappy as well in parallel…

[wyldckat]

Hi lovecraft22,

I can't go into detail, but you can study the tutorials provided here: http://code.google.com/p/bluecfd-sin...untimes202_211

Best regards,
Bruno

[lovecraft22]

Thank you Bruno;
I had a quick look and it seems that the boundaries are pretty similar to mine, apart for the "proc.*" entry.

There's a lot of stuff in the allRun instead so I need to have a better look at that.

Thank you!

[lovecraft22]

Hi Bruno;
could you please explain what's the difference between the method you advised and this one:
http://www.cfd-online.com/Forums/ope...tml#post363548

Thank you very much for your help!

[wyldckat]

Hi lovecraft22,

Originally Posted by lovecraft22

Hi Bruno;
could you please explain what's the difference between the method you advised and this one:
http://www.cfd-online.com/Forums/ope...tml#post363548

Thank you very much for your help!

If I remember correctly, the examples cases that I referenced above use the same "0" folder throughout most of the Allrun script. It only copies the folder "0.org" to "0" once. Again, this is if I remember correctly.

Another detail was to take into account the changes between scotch and ptscotch for decomposition and snappyHexMesh, respectively.

I believe that the method I used was more generic, since it would work for both 2.0 and 2.1, but in OpenFOAM 2.1 I think it doesn't need to reconstruct the mesh in between... the "motorBike" case in the "les" example demonstrates just that.

Best regards,
Bruno

[lovecraft22]

Thank you!