An Efficient Scheduling Algorithm for GPU-Attached Computing Systems

• Main Authors: Chien-Chen Lai, 賴建臣
• Other Authors: Cho-Chin Lin
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/50066465668284478784

碩士 === 國立宜蘭大學 === 電子工程學系碩士班 === 100 === H system is a computing system which consists of processing cores built in the CPUs of the system and attached processing devices such as GPUs. A complex task always has a tremendous computation need and it progresses by alternately performing work segments of different computation characteristics. In this thesis, an algorithm has been proposed for scheduling complex tasks to H system such that the device computing facilities can be effectively utilized. Our algorithm assigns priorities to complex tasks before mapping them to the cores of H system, and to reduce the amount of time all tasks need to spend in H system. The simulation has demonstrated that our algorithm is efficient than FFR algorithms for scheduling complex tasks to H system under various numbers of cores, and have obtained speedups as high as 19.4% on average compared to using FFR algorithm to scheduling tasks.