摘要
在设计初期,估计粗粒度可重构结构的性能,对粗粒度可重构结构设计具有指导意义。在考虑局部数据存储器结构以及局部数据存储器与可重构阵列的接口结构的情况下,建立了粗粒度可重构结构的参数模型,使用改进的螺旋形绑定策略将应用算法DFG(Data Flow Graph)中的算子绑定到可重构阵列的处理单元上,提出了一种粗粒度可重构结构的性能估计方法。应用实例表明,在设计初期,该方法能得到周期精确的估计结果,有效地指导粗粒度可重构结构的设计。
In the first step of a design, performance estimation is a guider to developing a coarse-grained reconfigurable architecture. With considering local data memory and interface between local data memory and reconfigurable array, parameter model is established for coarse-grained reconfigurable architecture. Improved spiral binding strategy is applied to bind operators in DFG (Data Flow Graph) of application algorithm to processing elements in reconfigurable array. A new method for performance estimation of coarse-grained reconfigurable architecture is proposed. Experiment results show that it can produce cycle-accurate result and direct the design of coarse-grained reconfigurable architecture effectively at an early stage of design cycle.
出处
《电路与系统学报》
CSCD
北大核心
2007年第3期84-88,共5页
Journal of Circuits and Systems
基金
国家863计划资助课题(2003AA1410502003AA1Z1060)