Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer software > Program design,software engineering > Software Engineering > Software Development

CUDA-based regular expression matching system design and implementation

Author ChenWeiMin
Tutor LiuWei
School Huazhong University of Science and Technology
Course Electronics and Communication Engineering
Keywords Web Content Management String matching CUDA Regular Expressions Pattern Matching
CLC TP311.52
Type Master's thesis
Year 2011
Downloads 30
Quotes 0
Download Dissertation

As the network influence on public life significantly improve and strengthen the building of web content management healthy and harmonious network environment plays a crucial role. String matching web content management is to achieve one of the key technologies . In many implementations of string matching , regular expressions as ability, easy to use, good scalability and other factors which have been widely used . However , regular expressions are more commonly implemented in software , as the network bandwidth and improve its processing performance is increasingly becoming a bottleneck restricting the development of related applications ; while in a particular hardware platform , such as : FPGA and ASIC implemented on platforms such as matching system hardware and software co costly , scalability is poor, it can not effectively meet the needs of web content management . Therefore, to find a fusion of performance and flexibility for high-speed string matching method to become industry-wide concern. In this paper, based on the CUDA platform for regular expression matching system solves the problem of matching performance and scalability , and lower costs of hardware and software to work together , Web content management is to achieve a more realistic solution. This major work include: ① the entire match tasks into serial and parallel sections , and then use the classes provided by CUDA C and CUDA API parallel part of the code will be ported to the GPU. ② studied a transplant for GPU parallel pattern matching algorithm and implements the algorithm. The algorithm , leveraging multi-threaded parallel processing GPU hide memory access latency , on the other hand use the GPU parallel computing capabilities powerful pattern matching process accelerated significantly improved matching system performance and efficiency. ③ matching files in the CPU side towards segmentation process , so that each thread load balancing. After testing , CUDA-based regular expression matching system relative to the regular expression library has a great performance boost.

Related Dissertations
More Dissertations