Recently, GPU has evolved into a highly parallel, multithreading, many core processor with tremendous computational capability and very high memory bandwidth. At the same time, multi-core CPU evolution continued and today's CPUs have 4-8 cores which offer dramatically increased performance and power savings characteristics. We are aware of very few works that consider both devices cooperating to solve general computations. The article tries to bring forward a method of similar master/ worker GPU-CPU cooperative computing to improve efficiency of Back-Propagation neural network-based image compression application even further than using either device independently.
展开▼
