[K3d-development] CUDA subdivide edges and next steps

[Bart J. <bar...@li...>]

Hi Evan,

I've been going over the SubdivideEdges code again, and (as you
probably know) there are still some missing steps:
- k3d::create_edge_adjacency_lookup. This is used to look up the
companions. This should be implemented on the device, as it will be
used in many plugins and it is a fairly expensive step.
- The edge_index_calculator. This is a sequential step, but keeping it
on the CPU would result in unneeded copying overhead when working with
device meshes.
- varying data copying: I propose postponing this one. For me,
handling the named arrays on the GPU in general does not need to be
part of the GSoc project.

After this, we can move on to the next plugin. Possible suggestions are:
- A mesh source, i.e. PolyGrid
- CalculateNormals
- The "CenterToMidpoints" part of SubdivideFaces. This is
topologically equivalent to Catmull-Clark subdivision, so it should be
relatively easy to reuse in SDS later.

Cheers,

-- 
Bart