北京科技大学学报2012，Vol.34Issue(12)：1469-1477,9.
基于超图的非规则应用局部性优化
Hypergraph-based irregular application locality optimization
摘要
Abstract
Muhiple indirection arrays often exist in one iteration, which is involved in irregular loop applications. A formal description of the hypergraph arrays was presented to solve this problem. Besides, three hypergraph-based data reordering algorithms (hypergraph-based non-repetitive coding data reordering algorithm, hypergraph-based backtracking search data reordering algorithm, and hypergraph-based partition first and then backtracking data reordering algorithm ) and two hypergraph-based iteration reordering algorithms (hypergraph-based non-repetitive coding iteration reordering algorithm and hypergraph-based backtracking search iteration reordering algorithm) were put forward. Experiments were performed on computational fluid dynamics, which was a representative irregular application. It is indicated that data locality is improved by the single reordering algorithm, with the execution speed increasing by 25. d%. The combination of the data reordering algorithm and the iteration reordering algorithm demonstrates the best performance, with the average hit rates of level-1 and level-2 cache reaching 91.7% and 96. 5% , respectively.关键词
数据局部性/高速缓冲存储器/重排/非规则/编译Key words
data locality/cache memory/reordering/irregularity/compiling分类
信息技术与安全科学引用本文复制引用
曹倩,刘立红,颉斌,陈洪菊..基于超图的非规则应用局部性优化[J].北京科技大学学报,2012,34(12):1469-1477,9.基金项目
国家自然科学基金资助项目 （）
国家重点基础研究发展规划资助项目 （）
北京工商大学青年教师科研启动基金项目 （）
北京市大学生科学研究与创业行动计划建设项目 （）
