CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Floating point exception during parallel mode of fluent (http://www.cfd-online.com/Forums/fluent/83783-floating-point-exception-during-parallel-mode-fluent.html)

aamer January 11, 2011 08:01

Floating point exception during parallel mode of fluent
 
Dear all

i am doing simulation on 3 D wing which is having 6.5 million cells. my OS is windows 7 64 bit with 8 gb RAM and 8 cores (i am using 64 bit version of fluent). when i run the case by using 4 cores of the computer (i.e by starting fluent with command fluent 3d -t4) i get error as "primitive error at node 1, node 2 and node 3: floating point exception" after few iterations. As remedial measure i reduced my under relaxation factor and time step but the problem remained the same......
when i run the same case with single processor, the case is running okie under same settings. In order to save the computational time i have to exploit parallelism, i am trying to resolve this issue...... any body having any idea why this error is coming with parallel mode????

mohammad March 5, 2011 06:45

Quote:

Originally Posted by aamer (Post 289993)
Dear all

i am doing simulation on 3 D wing which is having 6.5 million cells. my OS is windows 7 64 bit with 8 gb RAM and 8 cores (i am using 64 bit version of fluent). when i run the case by using 4 cores of the computer (i.e by starting fluent with command fluent 3d -t4) i get error as "primitive error at node 1, node 2 and node 3: floating point exception" after few iterations. As remedial measure i reduced my under relaxation factor and time step but the problem remained the same......
when i run the same case with single processor, the case is running okie under same settings. In order to save the computational time i have to exploit parallelism, i am trying to resolve this issue...... any body having any idea why this error is coming with parallel mode????

Hi Aamer, I'm not going to answer your quesetion, rather to ask you a question.
I have a 4-core computer and I want to use Ansys11 on windows7( 64bit)..is it possible? I will be really thankful if you can help me?

Shamoon Jamshed March 5, 2011 12:43

Hello Mohammad and Amir

@Amir : I must say that your error appears because of divergence. Actually what Fluent does is that it divided your comp. domain to all the cores or (nodes) in case of more than one PC. And when divergence appears with minor error the data that is shared b/w the cores also bulges out. This gets handled very easily in case of shared cores in single PC where delay time is negligible b/w cores. May be you get divergence after some more time as compared to distributed parallel. Try to use Density based solver with low CFL at first.

* Shared parallel : Same pc and parallelism is between cores
** Distributed parallel : Different (more than one) PC and parallelism between nodes (containing multiple cores)

@ Muhammad : Better to check compatibilty on ANSYS website, you should shift to version 12 I think cos ANSYS 11 came prior to Windows 7.

sheikh nasir February 7, 2012 23:55

Floating point error :invalid number
 
Hello.
I am getting floating point error in my thesis ie the train moving through tunnel. Can any body help me. My email is
sheikhnasir39@gmail.com
please help me
thanks

ahmadimo September 8, 2012 01:51

Floating point exception during parallel mode of fluent
 
Quote:

Originally Posted by aamer (Post 289993)
Dear all

i am doing simulation on 3 D wing which is having 6.5 million cells. my OS is windows 7 64 bit with 8 gb RAM and 8 cores (i am using 64 bit version of fluent). when i run the case by using 4 cores of the computer (i.e by starting fluent with command fluent 3d -t4) i get error as "primitive error at node 1, node 2 and node 3: floating point exception" after few iterations. As remedial measure i reduced my under relaxation factor and time step but the problem remained the same......
when i run the same case with single processor, the case is running okie under same settings. In order to save the computational time i have to exploit parallelism, i am trying to resolve this issue...... any body having any idea why this error is coming with parallel mode????

Hi aamer,

I am now having the same issue getting floating point errors when using SMP with 8 cores. However, things are going well when running with single processor. I was wondering if you had any progress on this issue or anybody else could help me with it.

I am modeling an Eulerian-Eulerian gas-solid fluidized bed. The approach proposed by Shamon Jamshed does not work for me as I am not allowed to use density-based solver with MultiPhase models in Fluent.

Regards,
Amir

victoryv October 30, 2012 21:14

I don't know if this information helps you or not. I was getting the same error. I reduced the relaxation factors as I assumed it was due to divergence. It worked perfect. Btw, I was also running in parallel.

mypergrfr May 8, 2013 05:47

Quote:

Originally Posted by victoryv (Post 389369)
I don't know if this information helps you or not. I was getting the same error. I reduced the relaxation factors as I assumed it was due to divergence. It worked perfect. Btw, I was also running in parallel.

Exactly, I get a convergence result by using your solution. But why this happens?

jipaz November 30, 2013 14:35

What parameters should I reduce relaxation factor and until what values?

Thanks in advance

mypergrfr December 1, 2013 23:00

Quote:

Originally Posted by jipaz (Post 464173)
What parameters should I reduce relaxation factor and until what values?

Thanks in advance

It depends on which solving method you are using,and which problem you are solving. I am using the coupled method to solve the natural convection problem. I changed all there under-relaxation factors from 1 to 0.3.

Regards.


All times are GMT -4. The time now is 03:19.