A GPGPU Implementation of Approximate String Matching with Regular Expression Operators and Comparison with Its FPGA Implementation

In this paper, we propose an efficient GPGPU implementation of an algorithm for approximate string matching with regular expression operators, originally implemented on an FPGA, and compare the GPGPU, FPGA and CPU implementations by experiments. Approximate string matching with regular expression operators is used in various applications, such as full text database search and DNA sequence analysis. To efficiently handle a long text in the matching, a hardware algorithm for FPGA implementation has been proposed. However, due to the limitation of FPGAs’ capacity, it cannot handle long patterns. In contrast, our proposed GPGPU implementation is able to handle long patterns efficiently, utilizing the scalability of GPGPU programming. Experimental results showed that the GPU implementation is more than 18 times as fast as the CPU one when the pattern length is greater than 3200, while the FPGA one could not handle such a long pattern.