Journal of Systems Engineering and Electronics ›› 2010, Vol. 32 ›› Issue (11): 2459-2464.doi: 10.3969/j.issn.1001-506X.2010.11.43

• 软件、算法与仿真 • 上一篇    下一篇

可重构计算平台上软硬件任务划分与调度算法

马宏星1,2,周学海1,2,高妍妍1,2   

  1. 1. 中国科学技术大学计算机科学与技术学院, 安徽 合肥 230027;
    2. 中国科学技术大学苏州研究院, 江苏 苏州 215123
  • 出版日期:2010-11-23 发布日期:2010-01-03

Algorithm for hardware/software task partitioning and scheduling on reconfigurable computing platform

MA Hong-xing1,2,ZHOU Xue-hai1,2,GAO Yan-yan1,2   

  1. 1. Dept. of Computer Science and Technology, Univ. of Science and Technology of China, Hefei 230027, China;
    2. Suzhou Academy, Univ. of Science and Technology of China, Suzhou 215123, China
  • Online:2010-11-23 Published:2010-01-03

摘要:

软硬件任务划分和任务调度是保证任务集合在可重构计算平台上高效运行的重要技术。传统上软硬件任务划分与任务调度独立进行,划分方案和调度方案不一定匹配,加速效果难以保证。提出了一种将软硬件任务划分与任务调度相结合的算法。该算法综合考虑了任务的执行时间、任务间的通信开销、任务间的依赖关系以及计算资源间的并发关系,可以同时获得给定任务集的软硬件任务划分方案和任务调度方案,加速效果良好。算法开销与传统的任务调度算法相当,时间复杂度为O(V2log2V+PVlog2V)实验结果表明了该算法的可行性和有效性。

Abstract:

Hardware/software partitioning and scheduling are critical to minimizing the overall run-time of applications on reconfigurable computing platform. Traditionally, hardware/software task partitioning and task scheduling are independent.The partition result and scheduling result is not always match, so that the accelerating effect is difficult to be guaranteed. A high performance algorithm combining hardware/software task partitioning with scheduling is proposed. The executive time of tasks, dependent of tasks, communication overhead among tasks, and parallelism of computing resources are comprehensively considered in the algorithm. The algorithm can produce both partition result and schedule result simultaneously. The overhead of the algorithm is similar to traditional scheduling algorithms. The time complexity of the proposed algorithm is O(V2log2V+PVlog2V). Experimental results show the feasibleness and effectiveness of the proposed algorithm.