A stream is a sequence of commands that relate to the GPU in CUDA programming. In other words, all the kernel calls and data transfers are handled by the CUDA stream. By default, CUDA provides a default stream, and all the commands use the stream implicitly. Therefore, we do not have to handle this ourselves.

CUDA supports additional streams, created explicitly. While the operations in a stream are sequential, CUDA can execute multiple operations concurrently by using the multiple streams. Let's learn how to handle streams, and what features they have.