More on CUDA and graphics processing unit computing
Mark Harris and Isaac Gelado
Abstract
This chapter presents a brief history of the evolution of and future outlook for CUDA. It clarifies the capabilities and limitations in several generations of the host-device interaction model, including zero-copy, unified virtual address space, and unified memory. It further presents the additional functionalities enabled by the recent ability to handle page faults during kernel execution. The chapter also presents the evolution of kernel execution efficiency and multiple kernel execution support. It concludes with an update on the recent advancement in programming, profiling, and debugging environments
Keywords
Unified memory
Get Programming Massively Parallel Processors, 3rd Edition now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
