摘要
可重构计算分时复用有限的面积资源,实现更多的任务硬件加速运行,同时也给传统的软硬件协同设计带来了新的挑战.为此设计了一种基于顶点位置树的迭代协同ICS-VPT算法针对离线型、集中共享式可重构计算平台,综合软硬件划分、硬件布局和任务调度,提升系统性能:首次提出顶点位置树的数据结构,以较小的存储空间快速查找布局位置;迭代协同算法根据数据依赖图分组任务,结合通信代价获取软/硬件任务的优先级,进行合理划分和调度.实验结果表明,ICS-VPT算法在高效管理可重构资源和灵活处理通信代价的同时,保持了较低的系统运行时间.
In reconfigurable computing,more tasks can be executed at a higher speed in hardware by time multiplexing with the limited area resources.Meanwhile,it also brings new challenges to the traditional hardware software codesign.For off line scheduling,centralized shared structure reconfigurable platform,the ICS-VPT （iteration co synthesis based on vertex position tree） algorithm synthesized hardware-software partitioning,hardware placement and task scheduling to improve system performance：The VPT （vertex position tree） data structure was first proposed,which could find a placement position quickly with small storage space; The ICS （iteration co-synthesis） algorithm grouped tasks according to data dependence graph and obtained the hardware/software tasks＇ priorities by combining the communication cost,thus obtaining a reasonable partitioning and scheduling.Experimental results show that the ICS-VPT algorithm maintains a lower level of system running time by means of efficient reconfigurable resource management and flexible communication cost handling.
基金
国家自然科学基金(61202053)
江苏省自然科学基金(BK2012194)资助
关键词
可重构计算
顶点位置树
软硬件划分
迭代调度
硬件布局
reconfigurable computing
vertex position tree
hardware-software partitioning
iteration scheduling
hardware placement
