You basically only need to learn about "zstd -T0". Better and faster algorithm and multi-threading directly included.
Zstd also has experimental code in the `contrib` dir for parallel (and, more importantly, seekable) decompression: https://github.com/facebook/zstd/issues/395