DKCHER算法是基于超扩展规则的求差知识编译算法,也是目前为止表现最好的EPCCL理论编译算法.本文通过研究DKCHER算法的执行流程,设计了一种新的启发式策略MOVR(maximum occurrence number of variables in middle result),用于动态地从输入子句集中选择所包含变量在中间结果中出现次数最多的子句.将MOVR启发式策略与DKCHER算法相结合,设计了MOVR_DKCHER算法.实验结果表明,MOVR启发式策略能够显著提高DKCHER算法的编译效率和编译质量,编译效率平均可提升70倍左右,最高可以提高237倍.
Abstract
DKCHER is a knowledge compilation algorithm of computing difference based on hyper extension rule
which is the best EPCCL compiler so far. We research the executing process of DKCHER algorithm in this paper
and design MOVR (maximum occurrence number of variables in middle result) heuristics
which is used to dynamically select the clause with maximum occurrence number of variables in middle result. By combining MOVR heuristics with DKCHER
MOVR_DKCHER algorithm is designed. Experimentally
MOVR heuristics can significantly improve the compilation efficiency and compilation quality of DKCHER
in which the compilation efficiency of DKCHER can be averagely improved about 70 times
and the improvement can be up to 237 times at most.