dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On 29 Sep 2011, at 00:46, Johan Hake <johan.hake@xxxxxxxxx> wrote:

> Nice!
> 
> Does this work for writing a Function to file in parallel too, like saving a 
> solution during a run?
> 

Yes.

Garth

> Johan
> 
> On Wednesday September 28 2011 16:34:57 Garth N. Wells wrote:
>> After a long absence, Function i/o to file has made a come back. It
>> was particularly necessary to support running code in parallel that
>> reads in data for Functions (such as the advection-diffusion demo). We
>> have more demos running in parallel now, and it's now possible to
>> restart parallel computations reliably (and even on a different number
>> of processes).
>> 
>> When sending a Function to an XML file, there is an entry for each
>> vector (dof) entry, e.g.
>> 
>>    <dof global_index="2" value="0" cell_index="3" local_dof_index="0" />
>> 
>> 'global_index' is the vector index, 'value' is the entry value,
>> 'cell_index' is a cell that the dof belongs to and 'local_dof_index'
>> is the local dof number relative to the given cell. With this
>> information, it's possible to read in and use function data
>> independently of the new dof map. If anyone has suggestions or
>> objections to this format, speak up now.
>> 
>> When reading in a Function, the Function must have a FunctionSpace,
>> and the vector must already have been sized (since the partitioning of
>> the vector in parallel will depend on the mesh partition).
>> 
>> Garth
>> 
>> _______________________________________________
>> Mailing list: https://launchpad.net/~dolfin
>> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~dolfin
>> More help   : https://help.launchpad.net/ListHelp