CUgrep: A GPU-based high performance multi-string matching system

String matching is one of the oldest and most pervasive problems in computer science. Nowadays applications related to string matching can be found everywhere. Meanwhile, the heterogeneous processing of CPU+GPU has become the popular parallel platform in solving high performance computing applications. This paper proposes a GPU-based multi-string matching algorithm, CUgrep, and uses this algorithm to implement the webpage matching system, which is almost 40 times peak performance to the original BNDM algorithm used in Nrgrep.