一种应用于数据通道综合的定向搜索流水调度算法

A Directed Searching Pipeline Scheduling Algorithm for Datapath Synthesis

流水调度是专用数字信号处理器高层综合中一个困难而急待解决的问题。给出了一种定向搜索流水调度算法 ,目标是使全面考虑了运算单元、寄存器和互连的硬件代价最小化。它作为一种利用启发信息的迭代算法 ,一方面克服了确定性算法爬峰能力差、易于陷于局部极值的缺点 ,另一方面启发信息的利用加快了搜索过程。

In the high-level synthesis of application-specific DSP, pipeline scheduling is a difficult and urgent problem Concerning the global optimization of functional units, registers and interconnections, we present a novel algorithm called directed searching pipeline scheduling, which is an iterative algorithm utilizing heuristic information The heuristic information speeds up the searching process, and the weakness of deterministic algorithms, which are vulnerable to trip in local optimal, poor at hill-climbing, is overcome in this algorithm Typical design examples show that the performance of this transformation-based algorithm has reached or surpassed the best performance of pipeline scheduling algorithms reported so far