Is it possible to accelerate the ode solver of reactingFoam?

pajofego · May 15, 2011, 11:59

Hello at all,

I'm doing a lot of combustion simulations with the reactingFoam solver. The problem is portioned -depending on the model size - up to 12 cpu cores. My experience is that the ode solver (sibs, KRR4, ...) for the chemistry is the most time consuming part of one time iteration. I haven't taken a deeper look into the code of the ode solver as well as I'm not a expert for Runge Kutta problems. So, did anyone know if it possible to accelerate the ode solver with other special techniques or to use alternative solvers? Maybe someone have think about it to export the ode solver to a GPU?

If you have some suggestions and ideas for me it would be appreciate.

Thanks in advance.
pajofego

Alish1984 · September 10, 2011, 18:33

Hi Paul

I also have the same problem,as u know when u use detailed chemistry,the solver solves a system of Ns(Number of species) ODE in each cell and each time step and it is so time consuming.
I tried to solve chemical kinetics sequentially instead of representing it in system of ODEs (by choosing sequential in chemistryProperties,chemistrySolver part)but it sometimes works with acceptable error and sometimes not! and i dont know why?
u can learn a little about the basics of sequentially solving reactions with reference species technique in "Complex Chemistry Modeling of Diesel Spray Combustion" by Niklas Nordin.as Nordin said this technique is also used in Kiva codes.I couldn't find any other article to explain more about sequentially solving reactions technique yet. maybe it works for ur case, but if it works,the speedup is noticeable.

Regards

Ali Shamooni

Aleksey_R · August 10, 2014, 06:38

Dear colleagues,

did you succeed in solving ODEs on GPU? I'm also interested in this problem.

Best regards,
Aleksey.

May 15, 2011, 11:59	Is it possible to accelerate the ode solver of reactingFoam?	#1
pajofego New Member Paul Join Date: Mar 2009 Posts: 19 Rep Power: 17	Hello at all, I'm doing a lot of combustion simulations with the reactingFoam solver. The problem is portioned -depending on the model size - up to 12 cpu cores. My experience is that the ode solver (sibs, KRR4, ...) for the chemistry is the most time consuming part of one time iteration. I haven't taken a deeper look into the code of the ode solver as well as I'm not a expert for Runge Kutta problems. So, did anyone know if it possible to accelerate the ode solver with other special techniques or to use alternative solvers? Maybe someone have think about it to export the ode solver to a GPU? If you have some suggestions and ideas for me it would be appreciate. Thanks in advance. pajofego Alish1984 likes this.

September 10, 2011, 18:33		#2
Alish1984 Member Ali Shamooni Join Date: Oct 2010 Posts: 44 Rep Power: 16	Hi Paul I also have the same problem,as u know when u use detailed chemistry,the solver solves a system of Ns(Number of species) ODE in each cell and each time step and it is so time consuming. I tried to solve chemical kinetics sequentially instead of representing it in system of ODEs (by choosing sequential in chemistryProperties,chemistrySolver part)but it sometimes works with acceptable error and sometimes not! and i dont know why? u can learn a little about the basics of sequentially solving reactions with reference species technique in "Complex Chemistry Modeling of Diesel Spray Combustion" by Niklas Nordin.as Nordin said this technique is also used in Kiva codes.I couldn't find any other article to explain more about sequentially solving reactions technique yet. maybe it works for ur case, but if it works,the speedup is noticeable. Regards Ali Shamooni Last edited by Alish1984; September 10, 2011 at 19:03.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
smoothSolver diverges - solution in using PBiCG solver?	makaveli_lcf	OpenFOAM Running, Solving & CFD	3	September 11, 2013 13:44
Working directory via command line	Luiz	CFX	4	March 6, 2011 21:02
Getting too many iterations by velocity solving (aborting). Changing U - Solver?	suitup	OpenFOAM Running, Solving & CFD	0	January 20, 2010 08:45
why the solver reject it? Anyone with experience?	bearcat	CFX	6	April 28, 2008 15:08
Error during Solver	cfd guy	CFX	4	May 8, 2001 07:04

August 10, 2014, 06:38		#3
Aleksey_R Member Aleksey Rukhlenko Join Date: Nov 2009 Location: Moscow Posts: 55 Rep Power: 17	Dear colleagues, did you succeed in solving ODEs on GPU? I'm also interested in this problem. Best regards, Aleksey.