On June 18, after more than a month of development, version 0.12.0 of the library and cross-platform console utility was released. ZXC (github.com), implementing high-performance multithreading asymmetric compression without loss and optimized for game resources, firmware, and application packages. The format is designed based on the principle of 'write once, read multiple times' (WORM).
Unlike codecs such as LZ4, ZXC sacrifices compression speed for maximum throughput during decompression.
It is declared that the unpacking speed is 10-47% faster than LZ4 with the default compression level, with an equal or higher compression ratio.
The main change in this release is the ability to compress using a pre-trained dictionary based on representative samples, which significantly reduces the size of archives with small blocks or a large number of small files.
The container format version has been updated to v6, and the library SOVERSION has increased from 3 to 4.
The project is written in C and released under the BSD 3 license.
Key Changes:
- compression with a pre-trained dictionary and a global Huffman table;
- changes in the console utility:
- a new parameter —train has been added for training the dictionary based on input files (the path to output files is specified with the -o parameter; by default — dictionary_.zxd). This parameter replaces the previous parameter —train-dict PATH;
- -D, —dict FILE — compresses or uncompresses a file using the dictionary FILE; -l, —list — displays the contents of the archive or information about the .zxd dictionary;
- a new alias for unzipping unzxc has been added — it is installed as a symbolic link to zxc and, by default, works in unpacking mode (equivalent to the command zxc -d).
- compression at level 6 is approximately 10% faster than version 0.11.0;
- decompression at levels 1 and 2 is approximately 3% faster than version 0.11.0;
- other improvements and bug fixes.
Source: linux.org.ru
