CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM (http://www.cfd-online.com/Forums/openfoam/)
-   -   Floating Point Exception (http://www.cfd-online.com/Forums/openfoam/76875-floating-point-exception.html)

pi06jl6 June 7, 2010 16:52

Floating Point Exception
 
Greetings all,

My program halts with the old classical "Floating Point Exception", at snGrad.

Some relevant details:

1. Operative system is Debian 64 bits
2. OpenFoam 1.6
3. Single thread, icoFoam+DragLift.H
4. This code worked on OpenFoam 1.5 and 32 bits
5. Same test case for 64 and 32 bits
6. Tutorials like icoFoam works fine.

So basically i moved from a 32 bit to 64 bit OS and changed on the way OF1.5 to OF1.6.

My Q:

What is the cause of the of problem and how to resolve this?

Would appriciate a swift answer to this important problem.

Best Regards

Johan

wyldckat June 8, 2010 18:54

Greetings Johan,

Swift answer: HowTo debugging in openfoamwiki.net :D

Not so swift: Possibly something changed in the definitions of boundary conditions between OF 1.5 and 1.6. Check the official --> release notes <-- or compare the tutorial you used as reference in OF 1.5 with the one in OF 1.6 :)

Best regards,
Bruno

pi06jl6 June 16, 2010 06:12

Floating point exception
 
Quote:

Originally Posted by wyldckat (Post 262179)
Greetings Johan,

Swift answer: HowTo debugging in openfoamwiki.net :D

Not so swift: Possibly something changed in the definitions of boundary conditions between OF 1.5 and 1.6. Check the official --> release notes <-- or compare the tutorial you used as reference in OF 1.5 with the one in OF 1.6 :)

Best regards,
Bruno

Thank you for the suggestion, the link was indeed interesting!

I tried toggle of these FOAM_SIGFPE and FOAM_SETNAN, but i didnt see any difference. One has to recompile the whole package? I though wmake of the case itself was enought.

Anyone? Would like to avoid recompile the package.

Important additional information, the program doesnt halt at first "encounter", rather randomly, cant see any pattern and it would provide anything good to paste the logs. From what i see it, its about i am using corrected method in sngrad and i can understand that this is floating point sensitive for special case flow as it appears flow around a body.

The crucial difference is that i have same case from 32 bits, which worked excellent, when i run the same case in 64 bits, i get this floating point exception.

Anyone who could provide me with usefil tip in how to twink to improve the solver i am using, should i go to uncorrected? The drag contribution to my FSI case is very low so have a robust drag estimator is to prefer.

In worst case scenario i will write an own routine. would like to avoid this so i appriciate any input to resolve this urgent matter.

Best Regards

Johan


All times are GMT -4. The time now is 11:35.