[aspect-devel] Aspect Timer

[Eric Heien]

Using the barrier method, I think I've tracked it down to the correct place (checkpoint_restart.cc:89):

      triangulation.save ((parameters.output_directory + "mesh").c_str());

This agrees with what I was seeing earlier (mesh file being slowly written).  It looks like this gets in to p4est_save() which I'm not familiar with.  Rather than try to fix this now, I'm going to put in a parameter to allow user specified checkpoint frequency (defaulting to 50).  We probably don't need checkpoints every 50 time steps, especially for smaller runs.

Related to this, should we have some sort of timer synchronization for parallel runs with Aspect?  Timo added the timer for checkpointing but it didn't record this section timing correctly until I added an MPI_Barrier.  This could potentially cause confusion in the future.

-Eric

On May 3, 2012, at 5:56 PM, Wolfgang Bangerth wrote:

> On 05/03/2012 07:51 PM, Eric Heien wrote:
>> I tracked this issue down to the following line (core.cc:312):
>> 
>>     DoFTools::count_dofs_per_block (dof_handler, system_dofs_per_block,
>>                                     system_sub_blocks);
>> 
>> This seems to take a long time after each checkpoint.  Any ideas?
> 
> I bet what you are seeing is this effect:
> 
> http://dealii.sourceforge.net/index.php/Deal.II_Questions_and_Answers#One_statement.2Fblock.2Ffunction_in_my_MPI_program_takes_a_long_time
> 
> W.
> 
> ------------------------------------------------------------------------
> Wolfgang Bangerth               email:            bangerth at math.tamu.edu
>                                 www: http://www.math.tamu.edu/~bangerth/
> 
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel