一种新颖的面向数据流量特征的片上网络设计

片上网络中处理器核之间、处理器与缓存之间通信具有不同的数据流量特征，因此需要不同的网络带宽和网络拓扑结构。针对上述两种通信流量的差异，设计了一种异构非对称双片上网络，分别用于处理器核之间和处理器与缓存之间的通信，并且每个网络都针对占主导地位的数据流量类型进行了优化。实验结果表明，与传统的单片上网络和同构的双片上网络相比，该设计不仅具有网络延迟低、网络带宽高的优点，而且提升了系统的整体性能。     

Hotel multiprocessor environments: guest and nonguest operated interfaces

Hotel property management systems (PMS) are very effective in monitoring guest transactions in the front office and financial transactions in the back office. There are, however, many other functional activities which are better handled outside their purview. While some of these applications might actually be included in a basic PMS design, the result would most likely be an unnecessarily complex and inefficient system; rather than an enhanced one. In addition, several candidate applications (e.g. point of sale, self check-in, call accounting, and electronic locking systems) require specialty hardware, software, and for operating systems. Realizing the scope of these requirements, vendors have developed independent, standalone systems capable of PMS interfacing. This design strategy enables increased fundamental PMS capabilities without enlarged structure. Connecting a secondary system (interface) to a hotel's primary computer system (PMS) creates a multiprocessor environment. Just as there are portions of the hotel that guests contact during a normal stay, there are PMS interfaces that require guest interaction to operate (guest operated interfaces); others that do not (nonguest operated interfaces). Popular guest operated interfaces are: self check-in/check-out systems, guest information systems, in-room movie systems, and in-room beverage systems. Nonguest operated interfaces include: point-of-sale systems, call accounting systems, electronic locking systems, and energy management systems. It is by virtue of their interface capability that guest and nonguest operated interfaces create a multiprocessor environment.