new interFoam solver parallelization issue

gentodin · September 8, 2022, 03:00

Hi, I've created a new solver based on interfoam.

The solver will update cells with alpha values of 1 after completing certain conditions.

The problem now is that when I run it in a parallel case, the solver showed different results than when I ran it in serial (single core).

I am not sure if I am missing something such as updating face value or calculating face value based on updated cells with alpha values of 1.

forAll(alpha1, celli)
{

if (some condition)
{
alpha1[celli] = 1;
}

}

//Should I add something here? for updating boundary faces also?

kk415 · September 8, 2022, 06:29

If that condition is depending on cell face then it could be the reason. Maybe you can put another check than that whether the face is a processor boundary or not and modify your condition accordingly.

gentodin · September 13, 2022, 03:54

Quote:

Originally Posted by kk415

If that condition is depending on cell face then it could be the reason. Maybe you can put another check than that whether the face is a processor boundary or not and modify your condition accordingly.

Hello kk415,

Thank you for your reply, the condition is not dependant on cell face at all, it is always dependent on cell values. There will be cells that is being updated on processor boundaries which Im not sure if such a simple update will affect anything.

What do you think is the common error % between serial and parallel simulations?

kk415 · September 13, 2022, 04:29

Hello gentodin,

My personal experience is there is a negligible difference in the serial and parallel run. I have worked with single-phase and two-phase flow problems. I have worked on interpolation and gradient calculation techniques. There are a few functions that are different for serial and parallel runs like Sum(serial) and gSum(parallel). Be careful!

Thanks

September 8, 2022, 03:00	new interFoam solver parallelization issue	#1
gentodin Member Anonymous. Join Date: Sep 2020 Posts: 35 Rep Power: 6	Hi, I've created a new solver based on interfoam. The solver will update cells with alpha values of 1 after completing certain conditions. The problem now is that when I run it in a parallel case, the solver showed different results than when I ran it in serial (single core). I am not sure if I am missing something such as updating face value or calculating face value based on updated cells with alpha values of 1. forAll(alpha1, celli) { if (some condition) { alpha1[celli] = 1; } } //Should I add something here? for updating boundary faces also?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How to remove viscosity term from the governing equation in interFoam solver	Pasar	OpenFOAM Programming & Development	3	February 18, 2022 06:56
CAMWA special issue on open-source numerical solver	feixu2019	OpenFOAM Announcements from Other Sources	0	October 1, 2018 12:21
fluent divergence for no reason	sufjanst	FLUENT	2	March 23, 2016 17:08
interFoam solver needs pdRefCell?	openTom	OpenFOAM Running, Solving & CFD	2	May 10, 2009 11:20
Open Channel Flow using InterFoam type solver	sxhdhi	OpenFOAM Running, Solving & CFD	3	May 5, 2009 22:58

September 8, 2022, 06:29		#2
kk415 Senior Member krishna kant Join Date: Feb 2016 Location: Hyderabad, India Posts: 133 Rep Power: 10	If that condition is depending on cell face then it could be the reason. Maybe you can put another check than that whether the face is a processor boundary or not and modify your condition accordingly.

September 13, 2022, 04:29		#4
kk415 Senior Member krishna kant Join Date: Feb 2016 Location: Hyderabad, India Posts: 133 Rep Power: 10	Hello gentodin, My personal experience is there is a negligible difference in the serial and parallel run. I have worked with single-phase and two-phase flow problems. I have worked on interpolation and gradient calculation techniques. There are a few functions that are different for serial and parallel runs like Sum(serial) and gSum(parallel). Be careful! Thanks