Driven by performance demands and energy-efficiency requirements, computa-tional systems of the future will most likely be composed of many heterogeneous processing units. In these systems, only a small set of workloads can benefit from a custom functional unit. FPGA-like reconfigurable architectures have the ability to adapt and, thereby, meet the specific computational needs of a wider range of workloads. As current research efforts make these devices more accessible to application develo-pers, we expect these devices to find a place in the computational paradigm in many different domains, integrated in a variety of different ways. Hence, to cover the wide range of possibilities, we need tools that automate the process of selecting, retargeting and executing parts of an application on these recon-figurable devices. Furthermore, there is also a need for domain specific customization of the reconfigu-rable fabric, which will help to maximize the benefits from reconfigurability.

Категория: Научные статьи | 1 Комментарий »

Комментарии