[Users] problem with GenericFD

[Ian Hinder]

On 1 Mar 2012, at 22:14, Bruno Giacomazzo wrote:

> Hi Roland,
> 	thanks for the prompt reply.
> 	
> On Mar 1, 2012, at 12:46 PM, Roland Haas wrote:
> 
>>> # Parameters of thorn CoordBase (implementing CoordBase)
>>> CoordBase::boundary_shiftout_x_lower = 1
>>> CoordBase::boundary_shiftout_z_lower = 1
>>> CoordBase::boundary_size_x_lower     = 3
>>> CoordBase::boundary_size_z_lower     = 3
>> Maybe
>> CoordBase::boundary_size_y_lower     = 3
>> CoordBase::boundary_size_x_upper     = 3
>> CoordBase::boundary_size_y_upper     = 3
>> CoordBase::boundary_size_z_upper     = 3
>> are missing? I certainly use those.
> 
> 
> 	this fixed the problem.
> 
> 	In the past I never used the *_upper and I used the*_ lower only for those directions where I was applying a symmetry condition. I think the *_upper just extends the grid from xmax to xmax+3*dx, while without them the boundary points are included in xmax. I'm not sure what is the effect of having added also y_lower, but the initial data look correct as before.

The boundary_size parameters identify a certain number of points as "boundary points".  Application thorns are free to interpret these as they see fit.  They default to 1, so you were using a boundary size of 1 for the physical boundaries.

WeylScal4, and generally all Kranc thorns, interpret the boundary points as those on which the standard stencil computation for the interior cannot be used, so it is consistent with other boundary conditions (e.g. symmetry boundary conditions) and interprocessor boundaries.  

I think that GenericFD could/should be modified to allow different numbers of boundary points on the different faces, but this would not change the outcome here, as if you got past this check, WeylScal4 would complain because it really *does* need to have a boundary size at least as large as its stencil size.

-- 
Ian Hinder
http://numrel.aei.mpg.de/people/hinder