Bryce Allen

Device vendor dependent code should be refactored, so ifdefs only occur in one place (perhaps space.h. One idea is to have `gt::space::cuda`, `gt::space::hip`, etc, and have `gt::space::device` be a type...

The CUDA/HIP implementations are fragile in that there may be array sizes that overflow certain limits. Explore using linear launch indexing and mapping back to expression indexes. This requires integer...