分簇VLIW DSP上支持单双字模式选择的SIMD编译优化

doi:10.11772/j.issn.1001-9081.2015.08.2371

计算机应用 ›› 2015, Vol. 35 ›› Issue (8): 2371-2374.DOI: 10.11772/j.issn.1001-9081.2015.08.2371

分簇VLIW DSP上支持单双字模式选择的SIMD编译优化

黄胜兵^1,2, 郑启龙^1,2, 郭连伟^1,2

1. 中国科学技术大学计算机科学与技术学院, 合肥 230027;
2. 安徽省高性能计算重点实验室(中国科学技术大学), 合肥 230027

收稿日期:2015-03-23 修回日期:2015-06-02 出版日期:2015-08-10 发布日期:2015-08-14
通讯作者: 黄胜兵(1990-),男,安徽安庆人,硕士研究生,主要研究方向:并行编译,huangsb@mail.ustc.edu.cn
作者简介:郑启龙(1969-),男,四川成都人,副教授,硕士,主要研究方向:并行编译; 郭连伟(1990-),男,安徽阜阳人,硕士研究生,主要研究方向:并行编译。
基金资助:
国家"核高基"重大专项(2012ZX01034-001-001)。

SIMD compiler optimization by selecting single or double word mode for clustered VLIW DSP

HUANG Shengbing^1,2, ZHENG Qilong^1,2, GUO Lianwei^1,2

1. School of Computer Science and Technology, University of Science and Technology of China, Hefei Anhui 230027, China;
2. Anhui Province Key Laboratory of High Performance Computing (University of Science and Technology of China), Hefei Anhui 230027, China

Received:2015-03-23 Revised:2015-06-02 Online:2015-08-10 Published:2015-08-14

摘要/Abstract

摘要：

BWDSP100是一款采用超长指令字(VLIW)和单指令多数据流(SIMD)架构的针对高性能计算领域而设计的32位静态标量数字信号处理器,其指令级并行(ILP)主要是通过其特殊的分簇体系结构和SIMD指令来实现,然而现有的编译框架无法对这些特殊的SIMD指令提供支持。由于BWDSP100拥有丰富的SIMD向量化资源,且其所运用的雷达数字信号处理领域对程序的性能要求极高,因此针对BWDSP100结构的特点,在传统Open64编译器中SIMD编译优化框架的基础上提出并实现了一种支持单双字模式选择的SIMD编译优化算法,通过该算法可以显著提高一些在DSP上有着广泛运用计算密集型程序的性能。实验结果表明,与优化前相比,该算法方案在BWDSP编译器上的实现能够平均取得5.66的加速比。

关键词: 编译优化, 指令级并行, 分簇体系数字信号处理器, 超长指令字, 单指令多数据流, Open64编译器

Abstract:

BWDSP100 is a 32-bit static scalar Digital Signal Processor (DSP) with Very Long Instruction Word (VLIW) and Single Instruction Multiple Data (SIMD) features, which is designed for high-performance computing. Its Instruction Level Parallelism (ILP) is acquired though clustering and special SIMD instructions. However, the existing compiler framework can not provide support for these SIMD instructions. Since BWDSP100 has much SIMD vectorization resources and there are very high requirements in radar digital signal processing for the program performance, an SIMD optimization which surpported the selection of single or double word mode was put forward based on the traditional Open64 compiler according to the characteristics of BWDSP100 structure, and it can significantly improve the performance of some compute-intensive programs which are widely used in DSP field. The experimental results show that this algorithm can achieve speedup of 5.66 on average compared with before optimization.

Key words: compiler optimization, Instruction Level Parallelism (ILP), multi-cluster Digital Signal Processor (DSP), Very Long Instruction Word (VLIW), Single Instruction Multiple Data (SIMD), Open64 compiler

中图分类号:

TP311.54

黄胜兵, 郑启龙, 郭连伟. 分簇VLIW DSP上支持单双字模式选择的SIMD编译优化[J]. 计算机应用, 2015, 35(8): 2371-2374.

HUANG Shengbing, ZHENG Qilong, GUO Lianwei. SIMD compiler optimization by selecting single or double word mode for clustered VLIW DSP[J]. Journal of Computer Applications, 2015, 35(8): 2371-2374.

参考文献

[1] CETC38. BWDSP100 hardware user manual [R]. Heifei: China Electronics Technology Group Corporation No.38 Research Institute, 2011: 1-2. (CETC38.BWDSP100硬件用户手册[R].合肥:中国电子科技集团第三十八研究所,2011:1-2.)
[2] CETC38. BWDSP100 software user manual [R]. Heifei: China Electronics Technology Group Corporation No.38 Research Institute, 2011: 181-191. (CETC38.BWDSP100软件用户手册[R].合肥:中国电子科技集团第三十八研究所,2011:181-191.)
[3] SUI Y. Open64 introduction [EB/OL].[2015-03-17]. http://www.cse.unsw.edu.au/~ysui/saber/open64.pdf.
[4] Open64. Open64 compiler WHIRL intermediate representation [EB/OL].[2015-03-17]. http://www.mcs.anl.gov/OpenAD/open64A.pdf.
[5] Open64. Using the x86 Open64 compiler suite [EB/OL].[2015-03-17]. http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2012/10/open64.pdf.
[6] SIMD [EB/OL].[2015-03-17]. http://en.wikipedia.org/wiki/SIMD.
[7] CHENG G, LAM M S. An optimizer for multimedia instruction sets [C/OL]//Proceedings of the 2nd SUIF Compiler Workshop. Stanford: Stanford University, 1997 [2015-01-16]. http://www-suif.stanford.edu/suifconf/suifconf2/.
[8] KRALL A, LELAIT S. Compilation techniques for multimedia processors [J]. International Journal of Parallel Programming, 2000, 28(4): 347-361.
[9] LARSEN S, AMARASINGHE S. Exploiting superword level parallelism with multimedia instruction sets [J]. ACM SIGPLAN Notices, 2000, 35(5): 145-156.
[10] LI Y. Rearch on SIMD vectorization and optimization of non-multimedia applications [D]. Heifei: University of Science and Technology of China, 2008:42-44. (李玉祥.面向非多媒体程序的SIMD向量化方法及优化技术研究[D].合肥:中国科学技术大学,2008:42-44.)
[11] ZHANG J. Register allocation for machines with connected and partitioned register banks [D]. Beijing: Chinese Academy of Sciences, Institute of Computing Technology, 2005: 92-94. (张军超.相连多寄存器组体系结构上的寄存器分配技术[D].北京:中国科学院计算技术研究所,2005:92-94.)
[12] XUE L. Research of global register allocation based on Godson 1 [D]. Beijing: Chinese Academy of Sciences, Institute of Computing Technology, 2001: 21-24. (薛丽萍.基于龙芯1的全局寄存器分配研究[D].北京:中国科学院计算技术研究所,2004:21-24.)
[13] WANG H, HUANG G, WANG X. Research and implementation of propagation cluster algorithm based on BWDSP100[J]. China Integrated Circuit, 2014, 23(8): 24-28. (王昊,黄光红,王向前.基于BWDSP100的传播分簇算法研究与实现[J].中国集成电路,2014,23(8):24-28.)
[14] JIANG J, WANG C, WEI H. An optimisation strategy for local register allocation [J]. Computer Applications and Software, 2013, 30(12): 216-217. (姜军,王超,尉红梅.一种局部寄存器分配的优化策略[J].计算机应用与软件,2013,30(12):216-217.)
[15] Integrated Systems for Signal Processing. DSPstone: a DSP-oriented benchmarking methodology [EB/OL]. [2015-03-12]. https://www.ice.rwth-aachen.de/fileadmin/publications/Zivojnovic94icspat.pdf

分簇VLIW DSP上支持单双字模式选择的SIMD编译优化

SIMD compiler optimization by selecting single or double word mode for clustered VLIW DSP

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 9

编辑推荐

Metrics

[1]	龚鸣清, 叶煌, 张鉴, 卢兴敬, 陈伟. 基于ARMv8架构的面向机器翻译的单精度浮点通用矩阵乘法优化[J]. 计算机应用, 2019, 39(6): 1557-1562.
[2]	王寿成, 徐进辉, 严迎建, 李功丽, 贾永旺. 面向密码流处理器的AES算法软件流水实现方法[J]. 计算机应用, 2017, 37(6): 1620-1624.
[3]	张拥军, 陈艇. 基于软件无线电的并行多输入多输出均衡技术[J]. 计算机应用, 2015, 35(4): 1179-1184.
[4]	闫国昌何炎祥李清安. 降低寄存器软错误的静态寄存器重分配方法[J]. 计算机应用, 2014, 34(9): 2730-2733.
[5]	路冬冬何军杨剑新王飙. 基于指令距离的存储相关性预测方法[J]. 计算机应用, 2013, 33(07): 1903-1907.
[6]	索维毅赵荣彩姚远刘鹏. 面向DSP的超字并行指令分析和冗余优化算法[J]. 计算机应用, 2012, 32(12): 3303-3307.
[7]	许杰麻军平何虎. 基于VLIW DSP 加密与认证算法的实现[J]. 计算机应用, 2012, 32(06): 1650-1653.
[8]	邱鹏飞洪一耿锐徐云. 基于数据流图的异构VLIW DSP分簇方法[J]. 计算机应用, 2011, 31(04): 935-937.
[9]	雷一鸣洪一徐云姜海涛. 一种基于寄存器压力的VLIW DSP分簇算法[J]. 计算机应用, 2010, 30(1): 274-276.