A fast complete deblocking filter on a coarse-grained reconfigurable processor supporting H.264 high profile decoding

This paper implements a fast complete deblocking filter on a novel coarse-grained reconfigurable processor, REMUS (REconfigurable Multimedia System). Reconfigurable processors have been proved to be the potential candidates for multimedia processing. However, their weakness in processing control-intensive tasks becomes a bottleneck in many reconfigurable applications. Mapping control-intensive tasks onto coarse-grained reconfigurable processors efficiently will advance relevant researches significantly. This paper proposes several methods and techniques to achieve this goal. With the proposed methods and techniques, the deblocking process, which represents one of the most control-intensive tasks in an H.264/AVC video decoder, consumes only 306 clock cycles per macroblock on REMUS. Comparing to the deblocking filters on ADRES, FloRA, and XPP-III, the deblocking filer on REMUS by this work achieves the speedup of 66.74%, 85.83%, and 55.52%, respectively. The implementing methods and techniques can also be extended into other applications.