[Users] global integer coordinates

[Erik Schnetter]

Steve

I assume you refer to the integer coordinates where the coarse grid
points may have a large stride (e.g. 512).

Assume you have a grid point with index i,j,k in C. These are the
process-local indices.

First you calculate the global indices as gi=cctk_lbnd[0]+i,
gj=cctk_lbnd[1]+j, etc. These are global, but still per refinement
level. This convention was introduced by PUGH.

Next you need to scale by the refinement factor of the current level,
which is cctk_levfac[0], cctk_levfac[1], etc. This scale factor is 1
for the coarsest level.

Finally, you many need to add an offset. This offset can be a rational
number, since coarse and fine grid points do not coincide for
cell-centred refinement. This rational number is cctk_levoff[0] /
cctk_levoffdenom[0] etc. This offset also needs to be scaled by
cctk_levfac.

Given that you divide by cctk_levfac and cctk_levoffdenom, the result
may not be integer. Carpet thus multiplies everything by a large
integer to ensure that the result is integer. I don't think this
integer is stored in the cGH structure. This integer is probably
2^(maxreflevels-1) for vertex-centred and 2^maxreflevels for
cell-centred grids. It is stored e.g. in the gh structure as the
stride of the coarsest level. You can access this stride via

    const ibbox &baseext = vhh.AT(map)->baseextents.AT(mglevel).AT(reflevel);

where you set map, mglevel, reflevel = 0. Then baseext.stride()[0] is
the stride in the i direction.

-erik



On Mon, Jan 11, 2016 at 1:52 PM, Steven R. Brandt <sbrandt at cct.lsu.edu> wrote:
> How do I get the global integer coordinates for a grid point (i.e. the
> ones that take into account refinement level)?
>
> Cheers,
> Steve
> _______________________________________________
> Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users



-- 
Erik Schnetter <schnetter at cct.lsu.edu>
http://www.perimeterinstitute.ca/personal/eschnetter/