基于GCC编译器向量化的数据结构布局优化研究  

Research on Data Structure Layout Optimization Based on GCC Compiler Vectorization

作　　者：朱广林 赖庆宽 何先波[1] 王博生 陈燕生 ZHU Guanglin;LAI Qingkuan;HE Xianbo;WANG Bosheng;CHEN Yansheng(School of Computer, China West Normal University, Nanchong, Sichuan 637009)

机构地区：[1]西华师范大学计算机学院,四川南充637009

出　　处：《绵阳师范学院学报》2021年第5期85-89,共5页Journal of Mianyang Teachers＇ College

基　　金：西华师范大学英才基金项目(17YC149).

摘　　要：有效利用单指令多数据(SIMD)技术能提高应用程序的性能,现代处理器和编译器都在不断地对SIMD技术进行扩展和改进,以生成向量化程度更高的代码.本文主要对现代x86多核处理器中,GCC编译器的数据结构布局进行优化,通过改善内存布局方式,提高向量化优化能力.最后在AMD平台上采用SPEC CPU2017进行实验,验证了该方法在GCC编译器上的有效性.Effective use of Single Instruction Multiple Data(SIMD)technology can improve application performance.Modern processors and compilers are constantly expanding and improving SIMD technology to generate more vectorized code.This paper mainly optimizes the data structure layout of the GCC compiler in modern x86 multi-core processors,and optimizes the vectorization optimization ability by improving the memory layout method.Finally,the SPEC CPU2017 was used on the AMD platform to conduct experiments to verify the effectiveness of the method on the GCC compiler.

关 键 词：单指令多数据 x86多核处理器 数据结构布局 SPEC CPU2017 GCC编译器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]