Heterogeneous Computing and The Real-World Applications

Abstract

In recent years, Graphics Processing Units (GPUs) in computational science have effectively driven the computing system to achieve higher computational power in various electronic devices. Combining with multi-core Central Processing Units (CPUs), the inevitable CPU-GPU integration is considered the new high-performance trend of the future. There have been multiple explorations on heterogeneous processing techniques, alongside the design of fused CPU and GPU chips. This survey paper will describe, summarize, and analyze some of the latest research from 2015 to map the techniques that can improve this computation hardware’s performance or energy efficiency. We then take a look at the collaborative CPU-GPU approaches with regards to runtime and applicability. We have certainty that this report can provide insights into the knowledge of CPU-GPU applications and help find future opportunities for these processors.