The computing efficient of the code_saturne

Questions and remarks about code_saturne usage
Forum rules
Please read the forum usage recommendations before posting.
Post Reply
MartianDuan
Posts: 74
Joined: Fri Aug 02, 2013 4:26 pm

The computing efficient of the code_saturne

Post by MartianDuan » Thu Feb 06, 2014 3:20 pm

Hi Guys

I am new user of the code_saturne. I just did a simulation on the heated laminar flow in the pipe. It turned out the computing efficiency is too bad.

The simulation is based on the mesh with 1.35M elements. It is the serial simualtion as well. The following is the information of time consuming of 40 iteration. The 'listing' is attached as well.
User CPU time: 15093.529 s
System CPU time: 407.143 s

Elapsed time: 15509.198 s
CPU / elapsed time 0.999

It means that the simulation cost 4hrs to finish.

Do you have any ideas on this issue. And how can i fix it?

Thanks in advance,

Yu.
Attachments
listing.log
(217.3 KiB) Downloaded 68 times

Brian Angel
Posts: 98
Joined: Mon Jul 22, 2013 12:53 pm

Re: The computing efficient of the code_saturne

Post by Brian Angel » Thu Feb 06, 2014 4:10 pm

Hello Yu,

Code_Saturne has by default a tight convergence tolerance for all variables, 1x10-8. I have found that changing this tolerance to, for example, 1x10-5 can significantly speed-up a calculation with minimal impact on the quality of the results. Can you please try this and see what happens.

The listing file also shows nearly 873000 clips for the temperature between -9K and 300K (?). This indicates that the calculation might be slowed down by this clipping in the initial stages. Hence, you could try to start with an adiabatic calculation for a hundred iterations or so in order to establish a flow field and then restart with heat transfer turned on.

Can you also let me know how you compiled Code_Saturne (debug option, other)?

In the meantime I'll look to see what sort of performance we get from CS here for such a calculation.

Best regards,

Brian Angel.

Post Reply