Bzip3, described as a spiritual successor to BZip2, demonstrates improved compression ratios across benchmarks on text and code files, according to its GitHub repository.
The tool uses an order-0 context mixing entropy coder, a Burrows-Wheeler transform implementation via suffix arrays, and RLE with Lempel Ziv and PPM-style context modeling to achieve compression performance improvements over its predecessor.
In a benchmark test compressing all released versions of Perl5 into a single archive, bzip3 achieved notably smaller compressed file sizes than competing tools. When configured with a block size of 511 bytes and parallel decompression, bzip3 produced a 546 MB compressed file, compared to 2.06 GB for xz (LZMA), 3.44 GB for bzip2, and 3.08 GB for Zstandard.
Decompression times showed bzip3 performing competitively. The parallel version decompressed in 4 minutes 6 seconds on a WD Blue HDD, faster than xz’s 4 minutes 40 seconds and bzip2’s 9 minutes 22 seconds, though slightly slower than Zstandard’s 3 minutes 51 seconds.
When combined with lrzip for long-range deduplication preprocessing, bzip3 achieved even stronger results. The combined lrzip-plus-bzip3 approach produced a 60.7 MB file, outperforming lrzip with lzma (64.8 MB) and lrzip with bzip2 (75.7 MB).
According to the repository, bzip3’s performance varies significantly by platform and compiler. On x64 Linux with clang13, compression speeds can reach 17 MiB/s per thread and decompression 23 MiB/s per thread, though Windows and 32-bit builds may be considerably slower.
The tool is installable via package managers on macOS through Homebrew and supports multiple architectures including x86, x86_64, ARM variants, PowerPC, MIPS, SPARC, and s390x. The codebase is licensed under LGPLv3.
Key facts
- Bzip3 compressed all Perl5 versions to 546 MB versus 2.06 GB for xz and 3.44 GB for bzip2
- Decompression speed: bzip3 4:06, xz 4:40, bzip2 9:22, Zstandard 3:51
- Combined with lrzip preprocessing, bzip3 achieved 60.7 MB compression versus 64.8 MB for lrzip-lzma
- Compression speeds up to 17 MiB/s on x64 Linux with clang13; performance varies by platform
- Supports x86, x86_64, ARM, PowerPC, MIPS, SPARC, and s390x architectures