• 2 min read
misa77 claims 2x faster decoding than LZ4
The open source misa77 codec targets write-once, read-many workloads with very fast decompression, modest ratios, and slow compression.

Image: Hacker News
A new open source codec called misa77 is pitching itself at a specific niche: write-once, read-many data that benefits from extremely fast single-threaded decompression. In its v0.2.0 release, the project says it can decode at roughly 2x LZ4's speed while often delivering better compression ratios, though the tradeoff is much slower compression.
misa77 is an LZ-based codec with no entropy backend, so its author positions it against tools like LZ4 rather than zstd. The design goals are straightforward:
- Very high decompression throughput
- Modest compression ratios
- Constant memory use regardless of input size
- Slow compression as the cost of those choices
The project says memory use stays at 5 MB or less across all compression modes, while decompression uses 0 MB of extra memory. It also claims a useful quirk: files compressed more aggressively can sometimes decompress faster, making higher-effort modes attractive despite the extra encode time.
On an Intel Core i7-14650HX with Turbo disabled, pinned to a single performance core, the posted benchmarks show misa77 well ahead of LZ4 in decode speed on two common test sets. On the Silesia Corpus, misa77 0.2.0 -0 posts 5219 MB/s decompression at a 42.64 ratio, versus LZ4 1.10.0 at 2505 MB/s and 47.59. On enwik8, misa77 0.2.0 -0 reaches 4802 MB/s at 48.59, compared with LZ4 at 2355 MB/s and 57.26.
The codec currently offers two main compression levels:

Recommended reading
Samsung 990 SSD hits 7,250 MB/s with a 2TB option
- Level 0: better decode throughput, slightly worse ratio
- Level 1: slightly worse decode throughput, better ratio
The author says misa77 is often on the Pareto frontier for decompression throughput versus compression ratio, especially on compressible files. In a file-level Silesia breakdown, level 0 reportedly decoded faster than LZ4 on all 12 files, while other levels beat LZ4 on 11 of 12. The outlier was x-ray, described as highly incompressible data where LZ4 effectively falls back to memcpy-like behavior.
Build requirements and current limitations
misa77 requires a C++20 compiler, CMake 3.20+, and a little-endian 64-bit system. The CLI targets Linux and macOS via POSIX. On x86-64, AVX2/SSE2 are selected at runtime; other architectures use a portable path that the author says auto-vectorizes on systems including Apple ARM.
The library exposes a small C++ API, and the dependency-free misa CLI supports three file-based commands:
- compress
- decompress
- suggest
There are caveats. The format may still change because the project is at v0.x.y, and as of v0.2.x the decoder assumes valid input. The author says an input-safe decoder is planned for v0.3.0; for now, invalid input is explicitly treated as undefined behavior. The project is released under the MIT license.
Computing Editor
Tomas lives in the terminal. He covers chips, laptops, and operating systems with a focus on performance and efficiency. He reads kernel changelogs the way other people read fiction, and he's always on the hunt for the perfect mechanical keyboard switch. If it processes data, Tomas has an opinion on it.
via Hacker News


