commit | 7173a631db61ab9535bd0d6e5e00e9dc081d4df3 | [log] [download] |
---|---|---|
author | Yann Collet <cyan@fb.com> | Sat Feb 24 11:47:53 2018 -0800 |
committer | Yann Collet <cyan@fb.com> | Sat Feb 24 11:47:53 2018 -0800 |
tree | 4701389655ac1f1b10eea0e38aa7fcd6fc37336c | |
parent | 99c26729b59d0389734973a9e3c55f7ef8408efb [diff] |
edge case : compress up to end-mflimit (12 bytes) The LZ4 block format specification states that the last match must start at a minimum distance of 12 bytes from the end of the block. However, out of an abundance of caution, the reference implementation would actually stop searching matches at 13 bytes from the end of the block. This patch fixes this small detail. The new version is now able to properly compress a limit case such as `aaaaaaaabaaa\n` as reported by Gao Xiang (@hsiangkao). Obviously, it doesn't change a lot of things. This is just one additional match candidate per block, with a maximum match length of 7 (since last 5 bytes must remain literals). With default policy, blocks are 4 MB long, so it doesn't happen too often Compressing silesia.tar at default level 1 saves 5 bytes (100930101 -> 100930096). At max level 12, it saves a grand 16 bytes (77389871 -> 77389855). The impact is a bit more visible when blocks are smaller, hence more numerous. For example, compressing silesia with blocks of 64 KB (using -12 -B4D) saves 543 bytes (77304583 -> 77304040). So the smaller the packet size, the more visible the impact. And it happens we have a ton of scenarios with little blocks using LZ4 compression ... And a useless "hooray" sidenote : the patch improves the LZ4 compression record of silesia (using -12 -B7D --no-frame-crc) by 16 bytes (77270672 -> 77270656) and the record on enwik9 by 44 bytes (371680396 -> 371680352) (previously claimed by [smallz4](http://create.stephan-brumme.com/smallz4/) ).
LZ4 is lossless compression algorithm, providing compression speed at 400 MB/s per core, scalable with multi-cores CPU. It features an extremely fast decoder, with speed in multiple GB/s per core, typically reaching RAM speed limits on multi-core systems.
Speed can be tuned dynamically, selecting an “acceleration” factor which trades compression ratio for more speed up. On the other end, a high compression derivative, LZ4_HC, is also provided, trading CPU time for improved compression ratio. All versions feature the same decompression speed.
LZ4 library is provided as open-source software using BSD 2-Clause license.
Branch | Status |
---|---|
master | |
dev |
Branch Policy:
- The “master” branch is considered stable, at all times.
- The “dev” branch is the one where all contributions must be merged before being promoted to master.
- If you plan to propose a patch, please commit into the “dev” branch, or its own feature branch. Direct commit to “master” are not permitted.
The benchmark uses lzbench, from @inikep compiled with GCC v6.2.0 on Linux 64-bits. The reference system uses a Core i7-3930K CPU @ 4.5GHz. Benchmark evaluates the compression of reference Silesia Corpus in single-thread mode.
Compressor | Ratio | Compression | Decompression |
---|---|---|---|
memcpy | 1.000 | 7300 MB/s | 7300 MB/s |
LZ4 fast 8 (v1.7.3) | 1.799 | 911 MB/s | 3360 MB/s |
LZ4 default (v1.7.3) | 2.101 | 625 MB/s | 3220 MB/s |
LZO 2.09 | 2.108 | 620 MB/s | 845 MB/s |
QuickLZ 1.5.0 | 2.238 | 510 MB/s | 600 MB/s |
Snappy 1.1.3 | 2.091 | 450 MB/s | 1550 MB/s |
LZF v3.6 | 2.073 | 365 MB/s | 820 MB/s |
Zstandard 1.1.1 -1 | 2.876 | 330 MB/s | 930 MB/s |
Zstandard 1.1.1 -3 | 3.164 | 200 MB/s | 810 MB/s |
zlib deflate 1.2.8 -1 | 2.730 | 100 MB/s | 370 MB/s |
LZ4 HC -9 (v1.7.3) | 2.720 | 34 MB/s | 3240 MB/s |
zlib deflate 1.2.8 -6 | 3.099 | 33 MB/s | 390 MB/s |
LZ4 is also compatible and well optimized for x32 mode, for which it provides an additional +10% speed performance.
make make install # this command may require root access
LZ4's Makefile
supports standard Makefile conventions, including staged installs, redirection, or command redefinition. It is compatible with parallel builds (-j#
).
The raw LZ4 block compression format is detailed within lz4_Block_format.
To compress an arbitrarily long file or data stream, multiple blocks are required. Organizing these blocks and providing a common header format to handle their content is the purpose of the Frame format, defined into lz4_Frame_format. Interoperable versions of LZ4 must respect this frame format.
Beyond the C reference source, many contributors have created versions of lz4 in multiple languages (Java, C#, Python, Perl, Ruby, etc.). A list of known source ports is maintained on the LZ4 Homepage.