- 學位論文
平行化無預掃H.264解碼器
Coarse Grain Parallelization of the H.264 Decoder without a Start-Code Scanner
摘要
Fine grain methods for parallelization of the H.264 decoder have good latency perfor- mance and less memory usage. However, they could not reach the scalability of coarse grain approaches although assuming a well-designed entropy decoder which can feed the increasing number of parallel working cores. We would like to introduce a GOP (Group of Pictures) level approach due to its high scalability, mentioning solution approaches for the well-known memory and latency issues. Our design revokes the need to a scanner for GOP start-codes which was used in the earlier methods. This approach lets all the cores work on the decoding task. Our experiments showed that the memory initialization op- erations may degrade the scalability of parallel applications substantially. The multicore cache architecture appeared to be a critical point for getting the desired speedup. For FHD resolution video, we observed a speedup of 7.51 with 8 processors having separate caches, and a speedup of 14.46 using 15 processors when a cache is shared by 2 processors.
並列摘要
Fine grain methods for parallelization of the H.264 decoder have good latency perfor- mance and less memory usage. However, they could not reach the scalability of coarse grain approaches although assuming a well-designed entropy decoder which can feed the increasing number of parallel working cores. We would like to introduce a GOP (Group of Pictures) level approach due to its high scalability, mentioning solution approaches for the well-known memory and latency issues. Our design revokes the need to a scanner for GOP start-codes which was used in the earlier methods. This approach lets all the cores work on the decoding task. Our experiments showed that the memory initialization op- erations may degrade the scalability of parallel applications substantially. The multicore cache architecture appeared to be a critical point for getting the desired speedup. For FHD resolution video, we observed a speedup of 7.51 with 8 processors having separate caches, and a speedup of 14.46 using 15 processors when a cache is shared by 2 processors.