high performance computing on graphics processing units: hgpu.org

In this paper we propose a dynamic model for parallel H.264/AVC video encoding on hybrid GPU/CPU systems. Entire inter-loop is parallelized on both CPU and GPU and computationally light and efficient model is proposed to dynamically distribute computation load among simultaneously processing devices. This model includes both dependency aware task scheduling and load balancing algorithm based on dynamic performance model. Experimental results show that with proposed model it is possible to achieve a speedup of 1.5 regarding the state-of-the-art solutions. Moreover, up to 40 frames per second implementation is achieved, for HD 1920×1080 resolution, considering exhaustive motion estimation.