Graphics Processing Unit (GPU) has been around for a while. Although they are primarily used for high-end 3D graphics processing, their use is now acknowledged for general massive parallel computing. This paper presents an original technique based on to compute many instances of the longest common subsequence problem on a generic GPU architecture using classic DP-matching. Application of this algorithm has been found useful to address the problem of filtering false positives produced by metrics-based clone detection methods. Experimental results of this application are presented along with a discussion of possibilities of using GPUs for other cloning related problems.