Evaluating the Efficiency of CPUs, GPUs and FPGAs on a Near-Duplicate Document Detection Via OpenCL