摘要
编译器由于程序分析能力不足,无法自动实现循环向量化或者会造成盲目自动向量化。为此,提出一种基于编译指示的向量化方法。通过在代码中插入向量化编译指示语句,指导自动向量化编译工具的处理过程,自动生成高效的向量化代码。测试结果表明,该方法能够有效提高目标代码的运行性能。
Aiming at the problems that vectorization compiler tool can not perform auto-vectorization or does blind auto-vectorization because of the insufficient of program analysis ability,this paper proposes a vectorization method based on the compiler directive.The compiler directive is used to guide vectorization compiler tool to perform some special actions and produce high efficiency vectorization code automaticly when it is inserted into the program.Test results show that the method can effectively improve the performance of the object code.
出处
《计算机工程》
CAS
CSCD
2012年第12期272-275,共4页
Computer Engineering
基金
"核高基"重大专项"支持国产CPU的编译系统及工具链"分课题"自动并行化与二进制翻译系统"(2009ZX01036-001-001-2)
关键词
单指令多数据流
自动向量化
数据依赖关系
编译指示
循环不变量
向量寄存器
Single Instruction Multiple Data（SIMD）
automatic vectorization
data dependence relationship
compiler directive
loop invariant
vector register
