libCEED
libCEED copied to clipboard
AtPoints Operations for GPU Backends
We should add initial GPU capability for AtPoints Bases, Restrictions, and Operators
Ok, moving this up on priority.
I think the main question is how to handle different numbers of points per cell. My instinct is to give each cell the size of scratch space as the maximum number of points in the cell and pad out the data by replicating the last value in the cell.
Ok, need a multi-element apply method and to tweak the tests for restrictions to handle different packing strategies of the e-vec