GitHub: <a href="https://github.com/Piezoid/pugz" rel="nofollow">https://github.com/Piezoid/pugz</a>.<p>From the readme:<p>"Contrary to the pigz program which does single-threaded decompression (see <a href="https://github.com/madler/pigz/blob/master/pigz.c#L232" rel="nofollow">https://github.com/madler/pigz/blob/master/pigz.c#L232</a>), pugz found a way to do truly parallel decompression. In a nutshell: the compressed file is splitted into consecutive sections, processed one after the other. Sections are in turn splitted into chunks (one chunk per thread) and will be decompressed in parallel. A first pass decompresses chunks and keeps track of back-references (see e.g. our paper for the definition of that term), but is unable to resolve them. Then, a quick sequential pass is done to resolve the contexts of all chunks. A final parallel pass translates all unresolved back-references and outputs the file."
Somewhat related, for bzip2, I use pbzip2 which uses all the cores, or as many as you specify. [1] It is in the EPEL repo for RHEL/CentOS/Fedora.<p>[1] - <a href="https://linux.die.net/man/1/pbzip2" rel="nofollow">https://linux.die.net/man/1/pbzip2</a>