• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1LZ4 - Extremely fast compression
2================================
3
4LZ4 is lossless compression algorithm,
5providing compression speed > 500 MB/s per core,
6scalable with multi-cores CPU.
7It features an extremely fast decoder,
8with speed in multiple GB/s per core,
9typically reaching RAM speed limits on multi-core systems.
10
11Speed can be tuned dynamically, selecting an "acceleration" factor
12which trades compression ratio for faster speed.
13On the other end, a high compression derivative, LZ4_HC, is also provided,
14trading CPU time for improved compression ratio.
15All versions feature the same decompression speed.
16
17LZ4 is also compatible with [dictionary compression](https://github.com/facebook/zstd#the-case-for-small-data-compression),
18and can ingest any input file as dictionary,
19including those created by [Zstandard Dictionary Builder](https://github.com/facebook/zstd/blob/v1.3.5/programs/zstd.1.md#dictionary-builder).
20(note: only the final 64KB are used).
21
22LZ4 library is provided as open-source software using BSD 2-Clause license.
23
24
25|Branch      |Status   |
26|------------|---------|
27|master      | [![Build Status][travisMasterBadge]][travisLink] [![Build status][AppveyorMasterBadge]][AppveyorLink] [![coverity][coverBadge]][coverlink] |
28|dev         | [![Build Status][travisDevBadge]][travisLink]    [![Build status][AppveyorDevBadge]][AppveyorLink]                                         |
29
30[travisMasterBadge]: https://travis-ci.org/lz4/lz4.svg?branch=master "Continuous Integration test suite"
31[travisDevBadge]: https://travis-ci.org/lz4/lz4.svg?branch=dev "Continuous Integration test suite"
32[travisLink]: https://travis-ci.org/lz4/lz4
33[AppveyorMasterBadge]: https://ci.appveyor.com/api/projects/status/github/lz4/lz4?branch=master&svg=true "Windows test suite"
34[AppveyorDevBadge]: https://ci.appveyor.com/api/projects/status/github/lz4/lz4?branch=dev&svg=true "Windows test suite"
35[AppveyorLink]: https://ci.appveyor.com/project/YannCollet/lz4-1lndh
36[coverBadge]: https://scan.coverity.com/projects/4735/badge.svg "Static code analysis of Master branch"
37[coverlink]: https://scan.coverity.com/projects/4735
38
39> **Branch Policy:**
40> - The "master" branch is considered stable, at all times.
41> - The "dev" branch is the one where all contributions must be merged
42    before being promoted to master.
43>   + If you plan to propose a patch, please commit into the "dev" branch,
44      or its own feature branch.
45      Direct commit to "master" are not permitted.
46
47Benchmarks
48-------------------------
49
50The benchmark uses [lzbench], from @inikep
51compiled with GCC v7.3.0 on Linux 64-bits (Debian 4.15.17-1).
52The reference system uses a Core i7-6700K CPU @ 4.0GHz.
53Benchmark evaluates the compression of reference [Silesia Corpus]
54in single-thread mode.
55
56[lzbench]: https://github.com/inikep/lzbench
57[Silesia Corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia
58
59|  Compressor             | Ratio   | Compression | Decompression |
60|  ----------             | -----   | ----------- | ------------- |
61|  memcpy                 |  1.000  |13100 MB/s   |  13100 MB/s   |
62|**LZ4 default (v1.8.2)** |**2.101**|**730 MB/s** | **3900 MB/s** |
63|  LZO 2.09               |  2.108  |  630 MB/s   |    800 MB/s   |
64|  QuickLZ 1.5.0          |  2.238  |  530 MB/s   |    720 MB/s   |
65|  Snappy 1.1.4           |  2.091  |  525 MB/s   |   1750 MB/s   |
66|  [Zstandard] 1.3.4 -1   |  2.877  |  470 MB/s   |   1380 MB/s   |
67|  LZF v3.6               |  2.073  |  380 MB/s   |    840 MB/s   |
68| [zlib] deflate 1.2.11 -1|  2.730  |  100 MB/s   |    380 MB/s   |
69|**LZ4 HC -9 (v1.8.2)**   |**2.721**|   40 MB/s   | **3920 MB/s** |
70| [zlib] deflate 1.2.11 -6|  3.099  |   34 MB/s   |    410 MB/s   |
71
72[zlib]: http://www.zlib.net/
73[Zstandard]: http://www.zstd.net/
74
75LZ4 is also compatible and optimized for x32 mode,
76for which it provides additional speed performance.
77
78
79Installation
80-------------------------
81
82```
83make
84make install     # this command may require root permissions
85```
86
87LZ4's `Makefile` supports standard [Makefile conventions],
88including [staged installs], [redirection], or [command redefinition].
89It is compatible with parallel builds (`-j#`).
90
91[Makefile conventions]: https://www.gnu.org/prep/standards/html_node/Makefile-Conventions.html
92[staged installs]: https://www.gnu.org/prep/standards/html_node/DESTDIR.html
93[redirection]: https://www.gnu.org/prep/standards/html_node/Directory-Variables.html
94[command redefinition]: https://www.gnu.org/prep/standards/html_node/Utilities-in-Makefiles.html
95
96
97Documentation
98-------------------------
99
100The raw LZ4 block compression format is detailed within [lz4_Block_format].
101
102Arbitrarily long files or data streams are compressed using multiple blocks,
103for streaming requirements. These blocks are organized into a frame,
104defined into [lz4_Frame_format].
105Interoperable versions of LZ4 must also respect the frame format.
106
107[lz4_Block_format]: doc/lz4_Block_format.md
108[lz4_Frame_format]: doc/lz4_Frame_format.md
109
110
111Other source versions
112-------------------------
113
114Beyond the C reference source,
115many contributors have created versions of lz4 in multiple languages
116(Java, C#, Python, Perl, Ruby, etc.).
117A list of known source ports is maintained on the [LZ4 Homepage].
118
119[LZ4 Homepage]: http://www.lz4.org
120