Abstract:

This paper finds the program’s bottleneck in inefficiency of loop’s operation by analyzing the traditional Fourier transform, besides, when the program is running, it is only distributed in one hardware core, so it doesn’t make the best of the multi-core. Aiming at the above problem, through researching the Intel Threading Building Blocks(TBB), the paper puts the loop’s operation into several mutually independent partition, and puts these partitions into multi-core, so it implements the traditional algorithm parallelization, the parallel program achieves an excellent effect.

Key words: Fourier transform, Intel software development tools, Intel threading building blocks, parallelization

摘要：

通过对传统傅里叶变换的分析，发现其运行的瓶颈主要是循环体的运算效率低下，并且程序执行时只会被分配到一个硬件核上，并没有充分利用多核。针对上述问题，通过对英特尔线程构建模块(TBB)的研究与应用，使得循环体内的运算被划分为各个相互独立的空间，并把这些空间的运算尽可能分配到多核上，实现了对传统傅里叶变换的并行化改造，并取得较好的效果。

关键词: 傅里叶变换, 英特尔软件开发套件, 英特尔线程构建模块, 并行化

CLC Number: 

• TP311.1