Skip to content

ACEAPEX v1.0.0

Choose a tag to compare

View all tags
@yasha1971-coder yasha1971-coder released this 18 May 08:13
· 699 commits to main since this release
v1.0.0
4f7cc0b

Lossless parallel-decode LZ77. First stable release.
Included in lzbench (PR #276 + PR #277, May 2026).

What's new

  • Parallel block decode: N threads decode N blocks simultaneously
  • Zero cross-block dependencies at decode time
  • BIT-PERFECT verified (XXH3-64)
  • Builds on Linux and macOS with libzstd

Benchmarks (nci, AMD EPYC 4344P, lzbench 2.2.1)

Threads Compress Decompress Ratio
I=1 150 MB/s 2574 MB/s 8.56%
I=8 862 MB/s 10192 MB/s 8.56%
I=16 1057 MB/s 11674 MB/s 8.56%

Decode scales with threads. zstd and lz4 do not.

Known limitations

  • Single-thread decode slower than zstd
  • Encode requires ~2.8GB RAM (ACEPX2 format)
  • Encode 7x slower than zstd
  • Research-grade — no ABI stability guarantee
  • Poor compression on binary data (sao: 1.29x ratio)

Notes

Tested on WSL (Windows Subsystem for Linux) — works.
BIT-PERFECT verified on nci, xml, dickens, sao.

Built with Claude AI assistance. MIT license.

Assets 3
Loading