Performance Analysis of Arithmetic Code on Parallelized MPS
Wang Qian①②; Lü Dong-qiang③; Ge Bao-shan①
①School of Computer Science and Engineering, Beihang University, Beijing 100083, China; ②61081 army of PLA, Beijing 100094, China; ③The Fourth Institute of the Second Artillery Equipment Academe, Beijing 100085, China
Abstract：Arithmetic code on parallelized MPS(Most Probable Symbol) not only avoids complex operation of classical parallelized arithmetic code, but also does not inflect its basic probability estimation rule since utilizing statistic law of multidimensional binary coding. The relation between parallel degree, speedup ratio and coding efficiency is theoretically analyzed based on the theorem of complete probability and statistic average. It is pointed out the algorithm with 2 parallel degree is superior to others on the coding efficiency and speed, the algorithm of 3 parallel degree is equal to the one of 4 parallel degree on the coding efficiency. The result is verified by the experiment.
2009, Vol. 31 
