An Approach of Mining Parallelizable Points for Multi-core Computing
With the rapid and wide applying of multi-core CPU in two years, how to take full advantage of the ability of parallel processing for multi-core computers and program the parallel programs is the issue in the research and application field of multi-core. Exploring parallelizable points in the source codes is the first step of parallel compiling technologies and parallelizable points of source code are mainly consisted in loops and inter-procedurals. In this paper, we explore intra-procedural and interprocedural parallelizable information to mine parallelizable points comprehensively. We carry out the intra-procedural parallelizable information exploration by using SVM to extract parallelizable characteristics from loops. The inter-procedural parallelizable information is extracted by analyzing control flow and data flow of inter-procedurals, after which we can get a new intermediate expression of the inter-procedural dependence relationship called Inter-procedural Dependence Graph. The experimental results show that our approach can explore parallelizable information of source code comprehensively, with which we can get considerable speed-up by parallelization.
Par allelizable points SVM Inter-procedural Intra-procedural
Yuhua Zhang Peifeng LI Guo Shen Qiaoming ZHU
School of Computer Science and Technology Soochow University,Suzhou,China
国际会议
太原
英文
529-533
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）