The cuda::memory::mapped::region_pair structure uses two raw pointers and a size field. Instead, let's make it a pair of memory::region_t's. Yes, it is a bit "wasteful" - another size_t value; but much more usable, and the pair might be optimized away, anyway.
The
cuda::memory::mapped::region_pair
structure uses two raw pointers and a size field. Instead, let's make it a pair ofmemory::region_t
's. Yes, it is a bit "wasteful" - another size_t value; but much more usable, and the pair might be optimized away, anyway.