Warp, execution context are GPU implementation details, but not an abstraction in CUDA.

Block is an abstraction in CUDA.

CUDA threads are both hardware implementation detail and abstraction in CUDA.

A block can be consisted of multiple warps.