dolfin team mailing list archive

[Johan Hake <johan.hake@xxxxxxxxx>]

Cutejeff!

It looks like the memory problem arise within the umfpack solver. I also get 
the same error at iteration 5. Changing the linear solver of the NewtonSolver 
to mumps, solved that problem. Not sure what the issue with umfpack is. Maybe 
it is not compiled for 64 bits?

Your problem still does not converge and mumps tells this through some 
filtered PETSc message.

Johan

On Friday October 1 2010 13:51:45 bin hu wrote:
> Hello Anders,
> 
> Thank you for your helps.
> 
> I have attached the code. Instead of making a 10-lines code with the
> same error, I changed a little bit the Cahn Hilliard demo code for my
> problem, since I basically followed this code to develop my original
> code.  Hopefully, it's easier for you to dig through.
> 
> I also attached a PDF file to explain how to split the original
> equation into 4 mixed equations. It's another problem. I want to use
> less mixed equations, therefore use less memories. But I couldn't
> figure out a way to do that.  If you have time, please look at my
> problem a bit. Any suggestion would be a great help.
> 
> If you run this code, the error message should show up after 5
> iterations, and the residual is also getting bigger during the first 5
> iterations, which is converging if using a coarser mesh.
> 
> I have got lots helps from the launchpad. I really like to help back.
> I hope I can contribute more after I am getting more familiar with
> Fenics. Please let me know if you do have some work that I can help
> with. BTW, I am a graduate student in University of Kansas, I know
> some programming stuff.
> 
> Thanks a lot.
> 
> Regards,
> 
> Bin (cutejeff)