Code compression algorithms and architectures for embedded systems

摘要

Code compression techniques and decompression architectures for embedded systems are disclosed, providing good compression ratio while improving decompression time for VLIW instructions and reducing bus power consumption. The invention includes two fixed-to-variable (F2V) length code compression schemes based on a reduced arithmetic code compression algorithm combining arithmetic coding with probability models; a static probability model using static coding and semi-adaptive coding using a Markov model. Multi-bit decompression methods for the F2V techniques are presented, together with a parallel decompression scheme that tags and divides a compressed block into smaller sub-blocks. The Markov model provides better compression ratio, but the static model has a less complicated decompression unit design. The invention also includes two variable-to-fixed (V2F) length coding algorithms, one based on Tunstall coding and another on arithmetic coding. The V2F algorithms are also combined with a static model and a Markov model.