节点文献
算法的相关性变换与划分带个数的关系
Relation between data dependency transformation and number of partitioning bands
【摘要】 使用相关性变换法,划分并映射循环算法到具有固定尺寸的Systolic阵列.下标集合被划分成若干条带,划分带的条数与变换后的Systolic阵列算法的执行时间成正比.指出了Moldovan给出的计算划分带条数的公式有很大局限性,给出了由空间变换计算划分带条数的方法.
【Abstract】 Using Data Dependency Method for partitioning and mapping algorithms into VLSIsystolic arrays,index set of algorithms must be partitioned into several bands.The processingtime of partitioned algorithm is proportional to the number of bands.The paper shows that theformula for computing the number of bands given by Moldovan is not correct and a method forcomputing it is presented by means of space transformation.
【关键词】 算法;
超大规模集成电路/算法变换;
算法划分;
【Key words】 algorithms; very large scale integrated circuits/algorithm transformation; algorithm partition;
【Key words】 algorithms; very large scale integrated circuits/algorithm transformation; algorithm partition;
- 【文献出处】 大连理工大学学报 ,JOURNAL OF DALIAN UNIVERSITY OF TECHNOLOGY , 编辑部邮箱 ,1995年03期
- 【分类号】O151.2
- 【被引频次】1
- 【下载频次】15