In this paper we explore the application of the QP water-marking algorithm proposed by G. Qu and M. Potkonjak to software watermarking. The algorithm was originally proposed as a technique for watermarking the graph coloring problem which can be applied to a variety of media such as FPGA designs and software through register allocation. We implemented the algorithm within the SandMark framework, a system that allows the study of watermarking, tamper-proofing, and obfuscation algorithms for Java bytecode. Through the use of this framework we were able to perform an empirical evaluation of the algorithm. In particular we demonstrate that the use of register allocation, while incurring no performance overhead and being stealthy, is in fact vulnerable to attacks such as decompile/recompile. We also demonstrate that the QP algorithm does not allow for accurate watermark recognition without significant modifications.