2 ============================================================================
3 LZO -- a real-time data compression library
4 ============================================================================
6 Author : Markus Franz Xaver Johannes Oberhumer
8 http://www.oberhumer.com/opensource/lzo/
15 LZO is a portable lossless data compression library written in ANSI C.
16 It offers pretty fast compression and very fast decompression.
17 Decompression requires no memory.
19 In addition there are slower compression levels achieving a quite
20 competitive compression ratio while still decompressing at
23 The LZO algorithms and implementations are copyrighted OpenSource
24 distributed under the GNU General Public License.
29 LZO is a data compression library which is suitable for data
30 de-/compression in real-time. This means it favours speed
31 over compression ratio.
33 The acronym LZO is standing for Lempel-Ziv-Oberhumer.
35 LZO is written in ANSI C. Both the source code and the compressed
36 data format are designed to be portable across platforms.
38 LZO implements a number of algorithms with the following features:
40 - Decompression is simple and *very* fast.
41 - Requires no memory for decompression.
42 - Compression is pretty fast.
43 - Requires 64 kB of memory for compression.
44 - Allows you to dial up extra compression at a speed cost in the
45 compressor. The speed of the decompressor is not reduced.
46 - Includes compression levels for generating pre-compressed
47 data which achieve a quite competitive compression ratio.
48 - There is also a compression level which needs only 8 kB for compression.
49 - Algorithm is thread safe.
50 - Algorithm is lossless.
52 LZO supports overlapping compression and in-place decompression.
57 LZO was designed with speed in mind. Decompressor speed has been
58 favoured over compressor speed. Real-time decompression should be
59 possible for virtually any application. The implementation of the
60 LZO1X decompressor in optimized i386 assembler code runs about at
61 the third of the speed of a memcpy() - and even faster for many files.
63 In fact I first wrote the decompressor of each algorithm thereby
64 defining the compressed data format, verified it with manually
65 created test data and at last added the compressor.
70 To keep you interested, here is an overview of the average results
71 when compressing the Calgary Corpus test suite with a blocksize
72 of 256 kB, originally done on an ancient Intel Pentium 133.
74 The naming convention of the various algorithms goes LZOxx-N, where N is
75 the compression level. Range 1-9 indicates the fast standard levels using
76 64 kB memory for compression. Level 99 offers better compression at the
77 cost of more memory (256 kB), and is still reasonably fast.
78 Level 999 achieves nearly optimal compression - but it is slow
79 and uses much memory, and is mainly intended for generating
82 The C version of LZO1X-1 is about 4-5 times faster than the fastest
83 zlib compression level, and it also outperforms other algorithms
84 like LZRW1-A and LZV in both compression ratio and compression speed
85 and decompression speed.
87 +------------------------------------------------------------------------+
88 | Algorithm Length CxB ComLen %Remn Bits Com K/s Dec K/s |
89 | --------- ------ --- ------ ----- ---- ------- ------- |
91 | memcpy() 224401 1 224401 100.0 8.00 60956.83 59124.58 |
93 | LZO1-1 224401 1 117362 53.1 4.25 4665.24 13341.98 |
94 | LZO1-99 224401 1 101560 46.7 3.73 1373.29 13823.40 |
96 | LZO1A-1 224401 1 115174 51.7 4.14 4937.83 14410.35 |
97 | LZO1A-99 224401 1 99958 45.5 3.64 1362.72 14734.17 |
99 | LZO1B-1 224401 1 109590 49.6 3.97 4565.53 15438.34 |
100 | LZO1B-2 224401 1 106235 48.4 3.88 4297.33 15492.79 |
101 | LZO1B-3 224401 1 104395 47.8 3.83 4018.21 15373.52 |
102 | LZO1B-4 224401 1 104828 47.4 3.79 3024.48 15100.11 |
103 | LZO1B-5 224401 1 102724 46.7 3.73 2827.82 15427.62 |
104 | LZO1B-6 224401 1 101210 46.0 3.68 2615.96 15325.68 |
105 | LZO1B-7 224401 1 101388 46.0 3.68 2430.89 15361.47 |
106 | LZO1B-8 224401 1 99453 45.2 3.62 2183.87 15402.77 |
107 | LZO1B-9 224401 1 99118 45.0 3.60 1677.06 15069.60 |
108 | LZO1B-99 224401 1 95399 43.6 3.48 1286.87 15656.11 |
109 | LZO1B-999 224401 1 83934 39.1 3.13 232.40 16445.05 |
111 | LZO1C-1 224401 1 111735 50.4 4.03 4883.08 15570.91 |
112 | LZO1C-2 224401 1 108652 49.3 3.94 4424.24 15733.14 |
113 | LZO1C-3 224401 1 106810 48.7 3.89 4127.65 15645.69 |
114 | LZO1C-4 224401 1 105717 47.7 3.82 3007.92 15346.44 |
115 | LZO1C-5 224401 1 103605 47.0 3.76 2829.15 15153.88 |
116 | LZO1C-6 224401 1 102585 46.5 3.72 2631.37 15257.58 |
117 | LZO1C-7 224401 1 101937 46.2 3.70 2378.57 15492.49 |
118 | LZO1C-8 224401 1 100779 45.6 3.65 2171.93 15386.07 |
119 | LZO1C-9 224401 1 100255 45.4 3.63 1691.44 15194.68 |
120 | LZO1C-99 224401 1 97252 44.1 3.53 1462.88 15341.37 |
121 | LZO1C-999 224401 1 87740 40.2 3.21 306.44 16411.94 |
123 | LZO1F-1 224401 1 113412 50.8 4.07 4755.97 16074.12 |
124 | LZO1F-999 224401 1 89599 40.3 3.23 280.68 16553.90 |
126 | LZO1X-1(11) 224401 1 118810 52.6 4.21 4544.42 15879.04 |
127 | LZO1X-1(12) 224401 1 113675 50.6 4.05 4411.15 15721.59 |
128 | LZO1X-1 224401 1 109323 49.4 3.95 4991.76 15584.89 |
129 | LZO1X-1(15) 224401 1 108500 49.1 3.93 5077.50 15744.56 |
130 | LZO1X-999 224401 1 82854 38.0 3.04 135.77 16548.48 |
132 | LZO1Y-1 224401 1 110820 49.8 3.98 4952.52 15638.82 |
133 | LZO1Y-999 224401 1 83614 38.2 3.05 135.07 16385.40 |
135 | LZO1Z-999 224401 1 83034 38.0 3.04 133.31 10553.74 |
137 | LZO2A-999 224401 1 87880 40.0 3.20 301.21 8115.75 |
138 +------------------------------------------------------------------------+
141 - CxB is the number of blocks
142 - K/s is the speed measured in 1000 uncompressed bytes per second
143 - the assembler decompressors are even faster
148 LZO is a block compression algorithm - it compresses and decompresses
149 a block of data. Block size must be the same for compression
152 LZO compresses a block of data into matches (a sliding dictionary)
153 and runs of non-matching literals. LZO takes care about long matches
154 and long literal runs so that it produces good results on highly
155 redundant data and deals acceptably with non-compressible data.
157 When dealing with uncompressible data, LZO expands the input
158 block by a maximum of 16 bytes per 1024 bytes input.
160 I have verified LZO using such tools as valgrind and other memory checkers.
161 And in addition to compressing gigabytes of files when tuning some parameters
162 I have also consulted various `lint' programs to spot potential portability
163 problems. LZO is free of any known bugs.
168 There are too many algorithms implemented. But I want to support
169 unlimited backward compatibility, so I will not reduce the LZO
170 distribution in the future.
172 As the many object files are mostly independent of each other, the
173 size overhead for an executable statically linked with the LZO library
174 is usually pretty low (just a few kB) because the linker will only add
175 the modules that you are actually using.
177 I first published LZO1 and LZO1A in the Internet newsgroups
178 comp.compression and comp.compression.research in March 1996.
179 They are mainly included for compatibility reasons. The LZO2A
180 decompressor is too slow, and there is no fast compressor anyway.
182 My experiments have shown that LZO1B is good with a large blocksize
183 or with very redundant data, LZO1F is good with a small blocksize or
184 with binary data and that LZO1X is often the best choice of all.
185 LZO1Y and LZO1Z are almost identical to LZO1X - they can achieve a
186 better compression ratio on some files.
187 Beware, your mileage may vary.
192 Despite of its size, the basic usage of LZO is really very simple.
194 Let's assume you want to compress some data with LZO1X-1:
196 * include <lzo/lzo1x.h>
198 compress your data with lzo1x_1_compress()
199 * link your application with the LZO library
201 * include <lzo/lzo1x.h>
203 decompress your data with lzo1x_decompress()
204 * link your application with the LZO library
206 The program examples/simple.c shows a fully working example.
207 See also LZO.FAQ for more information.
212 As LZO uses Autoconf+Automake+Libtool the building process under
213 UNIX systems should be very unproblematic. Shared libraries are
214 supported on many architectures as well.
215 For detailed instructions see the file INSTALL.
217 Please note that due to the design of the ELF executable format
218 the performance of a shared library on i386 systems (e.g. Linux)
219 is a little bit slower, so you may want to link your applications
220 with the static version (liblzo2.a) anyway.
222 For building under DOS, Win16, Win32, OS/2 and other systems
223 take a look at the file B/00readme.txt.
225 In case of troubles (like decompression data errors) try recompiling
226 everything without optimizations - LZO may break the optimizer
227 of your compiler. See the file BUGS.
229 LZO is written in ANSI C. In particular this means:
230 - your compiler must understand prototypes
231 - your compiler must understand prototypes in function pointers
232 - your compiler must correctly promote integrals ("value-preserving")
233 - your preprocessor must implement #elif, #error and stringizing
234 - you must have a conforming and correct <limits.h> header
235 - you must have <stddef.h>, <string.h> and other ANSI C headers
236 - you should have size_t and ptrdiff_t
241 I have built and tested LZO successfully on a variety of platforms
242 including DOS (16 + 32 bit), Windows 3.x (16-bit), Win32, Win64,
243 Linux, *BSD, HP-UX and many more.
245 LZO is also reported to work under AIX, ConvexOS, IRIX, MacOS, PalmOS (Pilot),
246 PSX (Sony Playstation), Solaris, SunOS, TOS (Atari ST) and VxWorks.
247 Furthermore it is said that its performance on a Cray is superior
248 to all other machines...
250 And I think it would be much fun to translate the decompressors
251 to Z-80 or 6502 assembly.
256 Here is what I'm planning for the next months. No promises, though...
258 - interfaces to .NET and Mono
259 - interfaces to Perl, Java, Python, Delphi, Visual Basic, ...
260 - improve documentation and API reference
263 Some comments about the source code
264 -----------------------------------
265 Be warned: the main source code in the `src' directory is a
266 real pain to understand as I've experimented with hundreds of slightly
267 different versions. It contains many #if and some gotos, and
268 is *completely optimized for speed* and not for readability.
269 Code sharing of the different algorithms is implemented by stressing
270 the preprocessor - this can be really confusing. Lots of marcos and
271 assertions don't make things better.
273 Nevertheless LZO compiles very quietly on a variety of
274 compilers with the highest warning levels turned on, even
280 LZO is Copyright (C) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
281 2005, 2006, 2007, 2008 Markus Franz Xaver Johannes Oberhumer
283 LZO is distributed under the terms of the GNU General Public License (GPL).
284 See the file COPYING.
286 Special licenses for commercial and other applications which
287 are not willing to accept the GNU General Public License
288 are available by contacting the author.