Make my solver to work in parallel

gionni · August 9, 2021, 07:11

Goodmorning,
I wrote a solver with some blocks of code where i manually alter some field values. Something like:

Code:

forAll(field,c) {
    if (something) field[c]=1;
}

When I do this, my code can't be run using mpirun: I was expecting the single thread to have its own information only, so field.size() being its own portion of the field, but apparently it doesn't work this way.
How can I have a code like this to work in parallel?

In my case I have 2 regions, and I need to couple them using a 1-halo layer. So I need to update the values at the boundary manually and impose the boundary conditions using boundaryFieldRef. Having the simulation running on a single core only would be very limiting.
Any help is appreciated. Thank you.

klausb · August 14, 2021, 16:14

Please be more specific about your requirements/setup, I am not convinced, that you need MPI operations at all.

When you talk about two regions, do you mean two areas which are part of your problem description or do you want to decompose your domain into two parts to run your simulation on two cores?

If you decompose you case to run it on multiple cores, are the settings (the values you want to introduce) for your "regions" influencing each other across processor boundaries which would require to transfer information from one core to another?

August 9, 2021, 07:11	Make my solver to work in parallel	#1
gionni New Member Giovanni Luddeni Join Date: Jan 2021 Posts: 14 Rep Power: 5	Goodmorning, I wrote a solver with some blocks of code where i manually alter some field values. Something like: Code: forAll(field,c) { if (something) field[c]=1; } When I do this, my code can't be run using mpirun: I was expecting the single thread to have its own information only, so field.size() being its own portion of the field, but apparently it doesn't work this way. How can I have a code like this to work in parallel? In my case I have 2 regions, and I need to couple them using a 1-halo layer. So I need to update the values at the boundary manually and impose the boundary conditions using boundaryFieldRef. Having the simulation running on a single core only would be very limiting. Any help is appreciated. Thank you.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Fail to converge when solving with a fabricated solution	zizhou	FLUENT	0	March 22, 2021 07:33
"Define_Profile" UDF for parallel solver	Antoine	Fluent UDF and Scheme Programming	9	February 29, 2016 07:09
Explicitly filtered LES	saeedi	Main CFD Forum	16	October 14, 2015 12:58
simpleFoam parallel	AndrewMortimer	OpenFOAM Running, Solving & CFD	12	August 7, 2015 19:45
RBF motion solver does not work well in parallel	lakeat	OpenFOAM Bugs	3	August 8, 2013 06:50

August 14, 2021, 16:14		#2
klausb Senior Member Klaus Join Date: Mar 2009 Posts: 281 Rep Power: 22	Please be more specific about your requirements/setup, I am not convinced, that you need MPI operations at all. When you talk about two regions, do you mean two areas which are part of your problem description or do you want to decompose your domain into two parts to run your simulation on two cores? If you decompose you case to run it on multiple cores, are the settings (the values you want to introduce) for your "regions" influencing each other across processor boundaries which would require to transfer information from one core to another?