Re: [Libmesh-devel] XDA Restart With Subdomain Restricted Variables

[Derek G. <der...@in...>]

Did you not use EquationSystems::read() then?  And I think that what you are proposing will only work as long as your parallel partitioning is exactly the same between the two runs (which actually, in the case where we caught this is _not_ true!).

Derek

On Jun 22, 2011, at 4:05 PM, Kirk, Benjamin (JSC-EG311) wrote:

> I know we've done this recently - but maybe we are declaring the system first and then just reading the values?
> 
> Anyway, I'll have a look but it will have to be next week...
> 
> -Ben
> 
> 
> 
> On Jun 22, 2011, at 4:53 PM, "Derek Gaston" <der...@in...> wrote:
> 
>> It doesn't work.
>> 
>> Derek
>> 
>> 
>> 
>> ;-)
>> 
>> But really... it doesn't work.  It appears that there are assumptions in System::read_serialized_vector() and in read_serialized_block_dof_objects() that are assuming that if a variable has nodal dofs then it has dofs at every node.
>> 
>> John and I have been talking and it seems like we should be doing something different here.  If the variable is subdomain restricted it seems like we need to be passing different arguments into read_serialized_block_dof_objects() around line 906 in system_io.C.... like maybe we should be passing the number of nodes in the subdomain and iterators that just go over those nodes?  Is there something else we need to do beyond that?
>> 
>> Or maybe we need to rethink the whole way this is designed... and try to use the dof_map a little bit more directly (for instance some of the calculations of number of things to read in the file inside read_serialized_block_dof_objects() seem like they should be redundant when we could just ask the dof_map...).  But I'm sure I'm missing some of the edge cases that are already thought of in the current implementation (one John and I were talking about is maybe restarting with a different partitioning pattern... that might screw up being able to get info from the dof_map about what to read.... I don't know).
>> 
>> Anwyay... I just wanted to open the discussion to see if anyone had a bright idea.....
>> 
>> Derek
>> ------------------------------------------------------------------------------
>> Simplify data backup and recovery for your virtual environment with vRanger.
>> Installation's a snap, and flexible recovery options mean your data is safe,
>> secure and there when you need it. Data protection magic?
>> Nope - It's vRanger. Get your free trial download today.
>> http://p.sf.net/sfu/quest-sfdev2dev
>> _______________________________________________
>> Libmesh-devel mailing list
>> Lib...@li...
>> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
> 
> ------------------------------------------------------------------------------
> Simplify data backup and recovery for your virtual environment with vRanger.
> Installation's a snap, and flexible recovery options mean your data is safe,
> secure and there when you need it. Data protection magic?
> Nope - It's vRanger. Get your free trial download today.
> http://p.sf.net/sfu/quest-sfdev2dev
> _______________________________________________
> Libmesh-devel mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-devel