Repository navigation
Tags
Tags
lbzip2 2.6.5 Compression is 4.6 to 6.0% faster on real data with byte-identical output. The block checksum moves out of the run-length coder into a slice-by-eight pass over the raw input, and the symbol-frequency counters rotate through four arrays so neighbouring increments stop colliding.
Release version 2.6.4 A release build compiles without warnings on every supported platform, and CI keeps it that way with -Werror. Scaffolding left behind by the move from autotools to CMake is gone. -S is documented as what it has been since 2012: accepted and ignored. See CHANGES.md for detail.
Release version 2.6.3 Faster decompression through a vector shuffle in the inverse MTF transform, on both AArch64 and SSSE3 (kjn#31). A default build is optimised again (kjn#39). The worker thread count is bounded (kjn#24). Builds and passes its tests on Windows under MSYS2 (kjn#16), on musl, and on macOS. See CHANGES.md for detail.
PreviousNext