high performance computing on graphics processing units: hgpu.org

In this paper a novel approximate algorithm for massively-parallel decoding of trellis based error correcting codes (ECC) is presented. The potential effect of using such optimized decoder on acceleration of simulations of modern communication systems implementing the most recent communication standards, such as LTE-A (Long Term Evolution – Advanced) is evaluated quantitatively by presenting an original open source implementation in C running on a graphical processor (GPU). The focus of this design is to provide a seamless acceleration to Matlab simulations without breaking compatibility with existing CPU-based simulation frameworks. A quantitative throughput comparison with available open source and proprietary solutions is also presented.