Ok, thank you both. Roy: yes, I'm willing to test the patch.

On Fri, Jan 10, 2014 at 1:46 AM, David Knezevic
<dkneze...@seas.harvard.edu>wrote:

>
> On 01/10/2014 02:09 AM, Roy Stogner wrote:
>
>>
>> On Thu, 9 Jan 2014, Francesco Ballarin wrote:
>>
>>  In particular, at line 346
>>> of src/reduced_basis/rb_eim_construction.C, the optimal_elem_id is
>>> broadcast among the processors, in order for the optimal element to
>>> be retrieved at line 356. However, when working in parallel on a
>>> ParallelMesh, the element may not be a local one, causing the
>>> assertion at line 318 of src/mesh/parallel_mesh.C to fail.
>>>
>>
>> David, this is the same place I'm seeing a failure in
>> reduced_basis_ex4.
>>
>>  I think that a copy of the optimal element should be broadcast
>>> between all the processors, possibly using a procedure similar to
>>> the one in MeshSerializer (only for the optimal element, not for the
>>> entire mesh), but my expertise with the ParallelMesh class is not
>>> sufficient for me to implement a bugfix on my own.
>>>
>>
>>  Do you have any suggestions?
>>>
>>
>> If this is the correct bug fix, I know enough about ParallelMesh to
>> come up with a patch for you to try out.  I don't know enough about RB
>> to be sure it's the correct bug fix (or to test my own patch), though.
>>
>
> This sounds like the correct fix to me.
>
> Regarding testing: If the parmesh version gives the same RB_sol.e as the
> serial mesh version, then it should be fine.
>
> David
>
>
------------------------------------------------------------------------------
CenturyLink Cloud: The Leader in Enterprise Cloud Services.
Learn Why More Businesses Are Choosing CenturyLink Cloud For
Critical Workloads, Development Environments & Everything In Between.
Get a Quote or Start a Free Trial Today. 
http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to