[Carpet] Recovering the grid structure

[Peter Diener]

Hi,

The problem may be related to a hdf5 issue. I was experimenting recently
with a alpha version of hdf5 version 1.8, that seems to use a lot less 
memory when reading in Carpet data. If I get some time in the next couple
of days I'll try and compile this version on pelican, so we can experiment
with this.

Cheers,

  Peter

On Mon, 8 May 2006, Ryoji Takahashi wrote:

> Currently, i have problem to run in pelican at LSU.
> which is AIX machine.   This looks like due to memory leak ... within 
> couple hours, the codes terminate.
> 
> How i can ask the help for solving this problem.
> 
> it is great pity for me to waste huge local rescues.
> 
> 
> ryoji
> 
> Erik Schnetter wrote:
> 
> > I have just pushed a set of patches that make it possible to recover  
> > the grid structure from checkpoint files.  This requires setting two  
> > new parameters
> >
> >     Carpet::regrid_during_recovery = no
> >     CarpetIOHDF5::use_grid_structure_from_checkpoint = yes
> >
> > These settings may become the default in the future.  The existing  
> > methods for deciding the grid structure while recovering are not  
> > used; instead, the grid structure is taken directly from the  
> > checkpoint file.  This works only for checkpoint files that were  
> > written after these patches have been applied.
> >
> > The first evolution step after recovering starts, as usual, with a  
> > regridding procedure, so that it is possible to change the grid  
> > layout after recovering.
> >
> > Please give feedback.
> >
> > -erik
> >
> >------------------------------------------------------------------------
> >
> >_______________________________________________
> >developers mailing list
> >developers at lists.carpetcode.org
> >http://lists.carpetcode.org/listinfo/developers
> >  
> >
> 
>