Hierarchically characterizing CUDA program behavior

Author

Yu, Zhibin ; Jin, Hai ; Goswami, Nilanjan ; Li, Tao ; John, Lizy K.

Author_Institution

Service Comput. Technol. & Syst. Lab., Huazhong Univ. of Sci. & Technol., Wuhan, China

fYear

2011

fDate

6-8 Nov. 2011

Firstpage

76

Lastpage

76

Abstract

CUDA has become a very popular programming paradigm in parallel computing area. However, very little work has been done for characterizing CUDA kernels. In this work, we measure the thread level performance, collect the basic block level characteristics, and glean the instruction level properties for about 35 programs from CUDA SDK, Parboil, and Rodinia benchmark suites. In addition, we define basic block vectors, synchronization vectors and thread similarity matrix to capture the characteristics of CUDA programs efficiently. We find that CUDA programs have some unique characteristics at each level compared to sequential programs.

Keywords

parallel architectures; parallel programming; CUDA SDK; CUDA kernels; CUDA program behavior; Parboil benchmark suites; Rodinia benchmark suites; instruction level properties; parallel computing; programming paradigm; sequential programs; synchronization vectors; Educational institutions; Graphics processing unit; Instruction sets; Kernel; Programming; USA Councils; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Workload Characterization (IISWC), 2011 IEEE International Symposium on

Conference_Location

Austin, TX

Print_ISBN

978-1-4577-2063-5

Electronic_ISBN

978-1-4577-2062-8

Type

conf

DOI

10.1109/IISWC.2011.6114201

Filename

6114201