On 10/08/2013 07:08 PM, Roy Stogner wrote:
>
> On Tue, 8 Oct 2013, David Knezevic wrote:
>
>> It wasn't a segfault actually. The solver was crashing, reporting
>> that there was a NaN in the matrix.
>>
>> I reverted to e1167380e6872c6dbf2e487d4908be192caa21e3 and it still
>> didn't work. Then I reverted to
>> 99edb484698f647358c539df5a229503147bcdfd and it worked again.
>>
>> I'll continue with the binary search...
>
> Hmm... not sure a binary search is going to be helpful here. This
> change required a major refactor, and then lots of the subsequent
> changes are small bugfixes in that.
>
> Testing feature by feature might help. I assume your failure case
> uses AMR and/or Dirichlet BCs? Does it fail in serial? If you print
> the dof constraints for a working case vs the branch head, are there
> differences?
False alarm. The problem was at my end, sorry! Looks like it's working
fine for me for the real-valued case.
However, I still get compilation errors with --enable-complex (on
eb4d88f). Have you tested this?
Tomorrow I'll start looking at the topic in the subject line:
DirichletBoundary on nodes. Do you have any suggestions on how I should
get started? I think the key point from your original email was this:
"For parallel consistency we actually do:
1. Interpolate at nodes
2. L2 project any edge degrees of freedom left over when holding node
dofs fixed
3. L2 project any face dofs left over when holding node and edge dofs
fixed "
Thanks,
David
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60134071&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
