Tightly coupled GPU/CPU: Architecture, Programming, and Performance

This session focuses on the latest and near future architecture and programming on GPU/CPU (and NIC) combination, especially considering the memory model where multiple devices access with each other, which is different and progressed from traditional memory system. Especially, three major GPU vendors; NVIDIA, AMD and Intel now have their own solutions in coupling of GPU and CPU, and it is very interesting information for accelerated computing users. We will discuss the advanced architecture, programming and performance provided by this new style of GPU-accelerated computing, based on the latest information by three vendors, followed by a free discussion as a panel discussion style.