Abstract Multi-player game servers present an interesting computational challenge since the desire tosupport,large numbers ,of interacting players in real-time places strict constraints on server computation ,time. These ,high computational needs make ,these game ,servers attractive candidates for parallel processing ,because at high player ,counts the server becomes,a performance ,bottleneck. This work investigates the issues in parallelizing such servers by parallelizingone particular application, the QuakeWorld game server. The approach presented in this report focuses on utilizing the spatial locality of players within the game,environment,to achieve,more efficient parallelization. The main idea is to divide the players amongst all available threads in such a way as to only process clients sequentially if they are in physical proximity to one another. This is done by partitioning the virtual world into distinct areas and processing players within each area sequentially, while each distinct area of the virtual world is processed ,concurrently. To ,further increase the capacity, other stages of the server computation are also parallelized. This parallel implementation ,of the ,QuakeWorld game ,server was ,confirmed ,to be
更多
分享
引用
