A novel classified multi-dictionary code compression for embedded systems

This paper present a novel code compression method using classified multi-dictionary, which significantly improves the compression efficiency without introducing any decompression penalty. Our classified multi-dictionary code compression method separates the executable binary code into different classes, and each class of the binary code is compressed using its own dictionary and codeword. We use shorter codeword for the class which its binary code occurs more frequently than the frequency of the binary code in other classes. The appropriate classes are found to make sure that the compression efficiency is the best. Experimental results of MiBench benchmark for ARM show that a significant degree of compression can be achieved. Our approach outperforms the existing variance code compression techniques by an average of 15%, giving a compression ratio of 50%~55%. The impact on system performance is slight and for some memory implementations the reduced memory bandwidth actually increases performance.