典型编译器自动并行化效果和评估  

Effect and Evaluation of Automatic Parallelization of Typical Compiler

作　　者：王鹏翔 韩林 丁丽丽 张素平 王冬 WANG Pengxiang;HAN Lin;DING Lili;ZHANG Suping;WANG Dong(State Key Laboratory of Matliematical Engineering and Advanced Computing,Zhengzhou 450001,China)

机构地区：[1]数学工程与先进计算国家重点实验室,河南郑州450001

出　　处：《信息工程大学学报》2018年第2期186-190,共5页Journal of Information Engineering University

摘　　要：采用Spec CPU2006基准测试程序,对Intel编译器、Open64编译器和GCC编译器3个典型编译器自动并行化的效果进行评估,同时对GCC编译器进行深入分析,用程序片段评估GCC编译器自动并行化的能力,分析GCC编译器中现有自动并行化能力的不足,为提升典型编译器的自动并行化能力提供参考。Automatic parallelization technique is an important means of transplanting serial programs to parallel architectures. A lot of typical compilers implement automatic parallelization modules and produce good acceleration effect. Automatic parallelization modules provide parallel analysis,exploration and parallel code generation of serial program. Compiler＇s automatic parallelization is an important way to improve application performance. Using compilers that support automatic parallelization to evaluate the performance of automatic parallelization is very helpful to compilers analysis and improvement. We test Spec CPU2006 benchmark programs on Intel compiler,Open64 compiler and the GCC compiler to evaluate automatic parallelization performance of these three typical compilers.Then an in-depth analysis of GCC compiler is conducted by using program fragments to evaluate the ability of GCC compiler automatic parallelization. Finally,we analyze the capabilities and limitation of automatic parallelization functions in GCC compiler,to provide a useful reference for further enhancement to typical compilers.

关 键 词：自动并行化 典型编译器 并行编译技术 

分 类 号：TP314[自动化与计算机技术—计算机软件与理论]