Compression ratio and decompression overhead tradeoffs in code compression for VLIW architectures

Abstract

Memory has been one of the most restricted resources in embedded system design. Therefore, code compression has been one of the approaches to reduce the system costs. Many code compression methods try to improve the compression ratio but ignore the decompression overhead. In this paper, we take advantage of the separation of coding and modeling in arithmetic coding and study the tradeoffs between compression ratio and decompression overhead by exploring different probability models. Designs of the decompression unit for different models are discussed. Our work shows that simpler models result in smaller decompression overhead in code compression for VLIW, at the expense of compression ratio.