With the advent of accelerator-based heterogeneous parallel systems, the need for a solution of the task-device matching problem is increasing. Due to the enormously growing diversity in existing computing architectures, optimal matching promises to deliver high performance at reduced energy costs. By means of OpenCL and particularly the LLVM compiler infrastructure, our approach makes the task-device matching decisions taking into account the characteristics and particularities of the different processing hardware. We evaluate our approach using a set of OpenCL based real-world applications and well established benchmarks, which are run on different hardware platforms and architectures. Our results indicate highly accurate predictions made by our model during the matching procedure.