A Methodology for Recursive Compression of the Compressed Data Files

The aim of this work is to search for the possibilities of the recursively recompressing of the compressed data files. We propose a series of operations for hierarchically partitioning a compressed input data file into blocks and compressing them by finding two memory locations with identical contents. These two locations, which have 16-bit words, are formed as a candidate pair if they are in the same core-level block. We keep one word of a candidate pair in the output file and compress the other word by skipping from the output file. There is an addressing information bit size problem for obtaining this candidate pair during decompressing. Therefore, we sacrifice unnecessary candidate pairs and keep the most suitable pairs that require minimal addressing information. The compression is carried out by saving all the parts of the compression metadata, which are status bits, addressing information and coded numbers, with the original data, except for some words that are skipped. Our goal is to try compressing a compressed file. Therefore, we can continue compression iterations recursively on the new compressed file versions using our application program without changing any parameters of it. The recompression iterations will be executed until there is no more compression possible. The blocks of a complete file or the file segments can be compressed independently from each other. Therefore, the parallel versions of the proposed method can be executed concurrently within a computer, as well as on a multi-computer system. An application program was implemented using the OpenMP and the GMP libraries. Our results show that we are able to obtain compression ratios up to 0.01% on certain compressed files.