Module memcpy2d 

Runtime-API 2-D memory copies + pitched device allocations.

Mirrors [baracuda_driver::memcpy2d]. PitchedBuffer<T> owns a 2-D device allocation with driver-chosen row stride; free functions handle host ↔ pitched-device copies.

Structs

PitchedBuffer

A pitched device allocation — height × width_elems grid of Ts with driver-chosen pitch_bytes ≥ width_elems * size_of::<T>().

Functions

copy_d_to_h_2d

Synchronous pitched-device → host 2-D copy.

copy_h_to_d_2d

Synchronous host → pitched-device 2-D copy. src holds height rows of width_elems Ts starting src_host_pitch_bytes apart.

copy_h_to_d_2d_async

Async variant of copy_h_to_d_2d.

memset_2d

2-D memset: fill a pitched region with byte value.

memset_2d_async

Async 2-D memset on stream.