摘要
多处理机调度问题是影响并行处理速度与效率的因素之一.通过对多处理机的调度环境、调度对象和任务模型的详细分析,在阐述了通过编译实现指导式自调度GSS算法基础上,提出了调度迭代块容量小于N/P(N是循环限值,P是处理机数),同步次数少于或等于GSS情况的改进算法RGSS,以求提高并行处理的速度和效率.本文提出的改进算法RGSS,克服了GSS算法极端分配调度块的现象,在大多数情况下减少了调度迭代块的子数,获得了负载平衡和减少系统开销的两个目标.最后文中给出了RGSS算法的数学证明与仿真结果.
Multiprocessor scheduling is one of the factors having influences on parallel processing speed and efficiency. By analysing the schedule environment, schedule object and task model of multiprocessor, expounding on the foundation realizing the guide scheduling strategy (GSS) algorithm through compiling, a reformed RGSS algorithm is put forward for improving speed and efficiency of parallel processing when the scheduling iteration piece capacity is smaller than N/P (N is the limit value circulating, and P is the processor number), and the synchronization time is less or equal to GSS circumstances.The improved algorithm RGSS proposed in this paper overcomes the shortcoming of GSS algorithm distributing scheduling piece, reduces the number of scheduling iteration pieces, and gains two objects of workload balance and reducing system overhead. Finally the mathematical proof and simulation results of RGSS algorithm are given.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
2002年第6期92-96,共5页
Journal of Harbin Engineering University
关键词
并行循环
负载平衡
指导式动态自调度
系统开销
parallel circular
workload balance
guide self-scheduling strategy
system overhead
