2 libmad - MPEG audio decoder library
3 Copyright (C) 2000-2004 Underbit Technologies, Inc.
7 ===============================================================================
11 MAD (libmad) is a high-quality MPEG audio decoder. It currently supports
12 MPEG-1 and the MPEG-2 extension to Lower Sampling Frequencies, as well as
13 the so-called MPEG 2.5 format. All three audio layers (Layer I, Layer II,
14 and Layer III a.k.a. MP3) are fully implemented.
16 MAD does not yet support MPEG-2 multichannel audio (although it should be
17 backward compatible with such streams) nor does it currently support AAC.
19 MAD has the following special features:
22 - 100% fixed-point (integer) computation
23 - completely new implementation based on the ISO/IEC standards
24 - distributed under the terms of the GNU General Public License (GPL)
26 Because MAD provides full 24-bit PCM output, applications using MAD are
27 able to produce high quality audio. Even when the output device supports
28 only 16-bit PCM, applications can use the extra resolution to increase the
29 audible dynamic range through the use of dithering or noise shaping.
31 Because MAD uses integer computation rather than floating point, it is
32 well suited for architectures without a floating point unit. All
33 calculations are performed with a 32-bit fixed-point integer
36 Because MAD is a new implementation of the ISO/IEC standards, it is
37 unencumbered by the errors of other implementations. MAD is NOT a
38 derivation of the ISO reference source or any other code. Considerable
39 effort has been expended to ensure a correct implementation, even in cases
40 where the standards are ambiguous or misleading.
42 Because MAD is distributed under the terms of the GPL, its redistribution
43 is not generally restricted, so long as the terms of the GPL are followed.
44 This means MAD can be incorporated into other software as long as that
45 software is also distributed under the GPL. (Should this be undesirable,
46 alternate arrangements may be possible by contacting Underbit.)
48 ===============================================================================
52 The code is optimized and performs very well, although specific
53 improvements can still be made. The output from the decoder library
54 consists of 32-bit signed linear fixed-point values that can be easily
55 scaled for any size PCM output, up to 24 bits per sample.
57 The API for libmad can be found in the `mad.h' header file. Note that this
58 file is automatically generated, and will not exist until after you have
61 There are two APIs available, one high-level, and the other low-level.
62 With the low-level API, each step of the decoding process must be handled
63 explicitly, offering the greatest amount of control. With the high-level
64 API, after callbacks are configured, a single routine will decode an
67 The high-level API may either be used synchronously or asynchronously. If
68 used asynchronously, decoding will occur in a separate process.
69 Communication is possible with the decoding process by passing control
72 The file `minimad.c' contains an example usage of the libmad API that
73 shows only the bare minimum required to implement a useful decoder. It
74 expects a regular file to be redirected to standard input, and it sends
75 decoded 16-bit signed little-endian PCM samples to standard output. If a
76 decoding error occurs, it is reported to standard error and decoding
77 continues. Note that the scale() routine in this code is only provided as
78 an example; it rounds MAD's high-resolution samples down to 16 bits, but
79 does not perform any dithering or noise shaping. It is therefore not
80 recommended to use this routine as-is in your own code if sound quality is
85 To get the best possible performance, it is recommended that an assembly
86 version of the fixed-point multiply and related routines be selected.
87 Several such assembly routines have been written for various CPUs.
89 If an assembly version is not available, a fast approximation version will
90 be used. This will result in reduced accuracy of the decoder.
92 Alternatively, if 64-bit integers are supported as a datatype by the
93 compiler, another version can be used that is much more accurate.
94 However, using an assembly version is generally much faster and just as
97 More information can be gathered from the `fixed.h' header file.
99 MAD's CPU-intensive subband synthesis routine can be further optimized at
100 the expense of a slight loss in output accuracy due to a modified method
101 for fixed-point multiplication with a small windowing constant. While this
102 is helpful for performance and the output accuracy loss is generally
103 undetectable, it is disabled by default and must be explicitly enabled.
105 Under some architectures, other special optimizations may also be
110 The output from MAD has been found to satisfy the ISO/IEC 11172-4
111 computational accuracy requirements for compliance. In most
112 configurations, MAD is a Full Layer III ISO/IEC 11172-3 audio decoder as
113 defined by the standard.
115 When the approximation version of the fixed-point multiply is used, MAD is
116 a limited accuracy ISO/IEC 11172-3 audio decoder as defined by the
119 MAD can alternatively be configured to produce output with less or more
120 accuracy than the default, as a tradeoff with performance.
122 MAD produces output samples with a precision greater than 24 bits. Because
123 most output formats use fewer bits, typically 16, it is recommended that a
124 dithering algorithm be used (rather than rounding or truncating) to obtain
125 the highest quality audio. However, dithering may unfavorably affect an
126 analytic examination of the output (such as compliance testing); you may
127 therefore wish to use rounding in this case instead.
131 GCC is preferred to compile the code, but other compilers may also work.
132 The assembly code in `fixed.h' depends on the inline assembly features of
133 your compiler. If you're not using GCC or MSVC++, you can either write
134 your own assembly macros or use the default (low quality output) version.
136 The union initialization of `huffman.c' may not be portable to all
137 platforms when GCC is not used.
139 The code should not be sensitive to word sizes or byte ordering, however
140 it does assume A % B has the same sign as A.
142 ===============================================================================
144 BUILDING AND INSTALLING
148 MAD can be built under Windows using either MSVC++ or Cygwin. A MSVC++
149 project file can be found under the `msvc++' subdirectory.
151 To build libmad using Cygwin, you will first need to install the Cygwin
154 http://www.cygwin.com/
156 You may then proceed with the following POSIX instructions within the
159 Note that by default Cygwin will build a library that depends on the
160 Cygwin DLL. You can use MinGW to build a library that does not depend on
161 the Cygwin DLL. To do so, give the option --host=mingw32 to `configure'.
163 POSIX Platforms (including Cygwin)
165 The code is distributed with a `configure' script that will generate for
166 you a `Makefile' and a `config.h' for your platform. See the file
167 `INSTALL' for generic instructions.
169 The specific options you may want to give `configure' are:
171 --enable-speed optimize for speed over accuracy
173 --enable-accuracy optimize for accuracy over speed
175 --disable-debugging do not compile with debugging support, and
176 use more optimizations
178 --disable-shared do not build a shared library
180 Note that you need not specify one of --enable-speed or --enable-accuracy;
181 in its default configuration, MAD is optimized for both. You should only
182 use one of these options if you wish to compromise speed or accuracy for
185 By default the package will build a shared library if possible for your
186 platform. If you want only a static library, use --disable-shared.
188 It is not normally necessary to use the following options, but you may
189 fine-tune the configuration with them if desired:
191 --enable-fpm=ARCH use the ARCH-specific version of the
192 fixed-point math assembly routines
193 (current options are: intel, arm, mips,
194 sparc, ppc; also allowed are: 64bit, approx)
196 --enable-sso use the subband synthesis optimization,
197 with reduced accuracy
199 --disable-aso do not use certain architecture-specific
202 By default an appropriate fixed-point assembly routine will be selected
203 for the configured host type, if it can be determined. Thus if you are
204 cross-compiling for another architecture, you should be sure either to
205 give `configure' a host type argument (--host) or to use an explicit
208 If an appropriate assembly routine cannot be determined, the default
209 approximation version will be used. In this case, use of an alternate
210 --enable-fpm is highly recommended.
212 Experimenting and Developing
214 Further options for `configure' that may be useful to developers and
217 --enable-debugging enable diagnostic debugging support and
220 --enable-profiling generate `gprof' profiling code
222 --enable-experimental enable code using the EXPERIMENTAL
225 ===============================================================================
229 Please read the `COPYRIGHT' file for copyright and warranty information.
230 Also, the file `COPYING' contains the full text of the GNU GPL.
232 Send inquiries, comments, bug reports, suggestions, patches, etc. to:
234 Underbit Technologies, Inc. <support@underbit.com>
236 See also the MAD home page on the Web:
238 http://www.underbit.com/products/mad/
240 ===============================================================================