AMD interview question

Cuda processing and alternative ways to achieve the same , parallel processing, semaphores etc