pulling vendor branch for rubberband

git-svn-id: svn://localhost/ardour2/branches/2.0-ongoing@2768 d708f5d6-7413-0410-9779-e7cbd77b26cf

libs/rubberband/1.0/COPYING

@@ -0,0 +1,280 @@
+		    GNU GENERAL PUBLIC LICENSE
+		       Version 2, June 1991
+
+ Copyright (C) 1989, 1991 Free Software Foundation, Inc.
+                          675 Mass Ave, Cambridge, MA 02139, USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+			    Preamble
+
+  The licenses for most software are designed to take away your
+freedom to share and change it.  By contrast, the GNU General Public
+License is intended to guarantee your freedom to share and change free
+software--to make sure the software is free for all its users.  This
+General Public License applies to most of the Free Software
+Foundation's software and to any other program whose authors commit to
+using it.  (Some other Free Software Foundation software is covered by
+the GNU Library General Public License instead.)  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+this service if you wish), that you receive source code or can get it
+if you want it, that you can change the software or use pieces of it
+in new free programs; and that you know you can do these things.
+
+  To protect your rights, we need to make restrictions that forbid
+anyone to deny you these rights or to ask you to surrender the rights.
+These restrictions translate to certain responsibilities for you if you
+distribute copies of the software, or if you modify it.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must show them these terms so they know their
+rights.
+
+  We protect your rights with two steps: (1) copyright the software, and
+(2) offer you this license which gives you legal permission to copy,
+distribute and/or modify the software.
+
+  Also, for each author's protection and ours, we want to make certain
+that everyone understands that there is no warranty for this free
+software.  If the software is modified by someone else and passed on, we
+want its recipients to know that what they have is not the original, so
+that any problems introduced by others will not reflect on the original
+authors' reputations.
+
+  Finally, any free program is threatened constantly by software
+patents.  We wish to avoid the danger that redistributors of a free
+program will individually obtain patent licenses, in effect making the
+program proprietary.  To prevent this, we have made it clear that any
+patent must be licensed for everyone's free use or not licensed at all.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+		    GNU GENERAL PUBLIC LICENSE
+   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+  0. This License applies to any program or other work which contains
+a notice placed by the copyright holder saying it may be distributed
+under the terms of this General Public License.  The "Program", below,
+refers to any such program or work, and a "work based on the Program"
+means either the Program or any derivative work under copyright law:
+that is to say, a work containing the Program or a portion of it,
+either verbatim or with modifications and/or translated into another
+language.  (Hereinafter, translation is included without limitation in
+the term "modification".)  Each licensee is addressed as "you".
+
+Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope.  The act of
+running the Program is not restricted, and the output from the Program
+is covered only if its contents constitute a work based on the
+Program (independent of having been made by running the Program).
+Whether that is true depends on what the Program does.
+
+  1. You may copy and distribute verbatim copies of the Program's
+source code as you receive it, in any medium, provided that you
+conspicuously and appropriately publish on each copy an appropriate
+copyright notice and disclaimer of warranty; keep intact all the
+notices that refer to this License and to the absence of any warranty;
+and give any other recipients of the Program a copy of this License
+along with the Program.
+
+You may charge a fee for the physical act of transferring a copy, and
+you may at your option offer warranty protection in exchange for a fee.
+
+  2. You may modify your copy or copies of the Program or any portion
+of it, thus forming a work based on the Program, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+    a) You must cause the modified files to carry prominent notices
+    stating that you changed the files and the date of any change.
+
+    b) You must cause any work that you distribute or publish, that in
+    whole or in part contains or is derived from the Program or any
+    part thereof, to be licensed as a whole at no charge to all third
+    parties under the terms of this License.
+
+    c) If the modified program normally reads commands interactively
+    when run, you must cause it, when started running for such
+    interactive use in the most ordinary way, to print or display an
+    announcement including an appropriate copyright notice and a
+    notice that there is no warranty (or else, saying that you provide
+    a warranty) and that users may redistribute the program under
+    these conditions, and telling the user how to view a copy of this
+    License.  (Exception: if the Program itself is interactive but
+    does not normally print such an announcement, your work based on
+    the Program is not required to print an announcement.)
+
+These requirements apply to the modified work as a whole.  If
+identifiable sections of that work are not derived from the Program,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works.  But when you
+distribute the same sections as part of a whole which is a work based
+on the Program, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Program.
+
+In addition, mere aggregation of another work not based on the Program
+with the Program (or with a work based on the Program) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+  3. You may copy and distribute the Program (or a work based on it,
+under Section 2) in object code or executable form under the terms of
+Sections 1 and 2 above provided that you also do one of the following:
+
+    a) Accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of Sections
+    1 and 2 above on a medium customarily used for software interchange; or,
+
+    b) Accompany it with a written offer, valid for at least three
+    years, to give any third party, for a charge no more than your
+    cost of physically performing source distribution, a complete
+    machine-readable copy of the corresponding source code, to be
+    distributed under the terms of Sections 1 and 2 above on a medium
+    customarily used for software interchange; or,
+
+    c) Accompany it with the information you received as to the offer
+    to distribute corresponding source code.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form with such
+    an offer, in accord with Subsection b above.)
+
+The source code for a work means the preferred form of the work for
+making modifications to it.  For an executable work, complete source
+code means all the source code for all modules it contains, plus any
+associated interface definition files, plus the scripts used to
+control compilation and installation of the executable.  However, as a
+special exception, the source code distributed need not include
+anything that is normally distributed (in either source or binary
+form) with the major components (compiler, kernel, and so on) of the
+operating system on which the executable runs, unless that component
+itself accompanies the executable.
+
+If distribution of executable or object code is made by offering
+access to copy from a designated place, then offering equivalent
+access to copy the source code from the same place counts as
+distribution of the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+  4. You may not copy, modify, sublicense, or distribute the Program
+except as expressly provided under this License.  Any attempt
+otherwise to copy, modify, sublicense or distribute the Program is
+void, and will automatically terminate your rights under this License.
+However, parties who have received copies, or rights, from you under
+this License will not have their licenses terminated so long as such
+parties remain in full compliance.
+
+  5. You are not required to accept this License, since you have not
+signed it.  However, nothing else grants you permission to modify or
+distribute the Program or its derivative works.  These actions are
+prohibited by law if you do not accept this License.  Therefore, by
+modifying or distributing the Program (or any work based on the
+Program), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Program or works based on it.
+
+  6. Each time you redistribute the Program (or any work based on the
+Program), the recipient automatically receives a license from the
+original licensor to copy, distribute or modify the Program subject to
+these terms and conditions.  You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties to
+this License.
+
+  7. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Program at all.  For example, if a patent
+license would not permit royalty-free redistribution of the Program by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Program.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system, which is
+implemented by public license practices.  Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+  8. If the distribution and/or use of the Program is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Program under this License
+may add an explicit geographical distribution limitation excluding
+those countries, so that distribution is permitted only in or among
+countries not thus excluded.  In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+  9. The Free Software Foundation may publish revised and/or new versions
+of the General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+Each version is given a distinguishing version number.  If the Program
+specifies a version number of this License which applies to it and "any
+later version", you have the option of following the terms and conditions
+either of that version or of any later version published by the Free
+Software Foundation.  If the Program does not specify a version number of
+this License, you may choose any version ever published by the Free Software
+Foundation.
+
+  10. If you wish to incorporate parts of the Program into other free
+programs whose distribution conditions are different, write to the author
+to ask for permission.  For software which is copyrighted by the Free
+Software Foundation, write to the Free Software Foundation; we sometimes
+make exceptions for this.  Our decision will be guided by the two goals
+of preserving the free status of all derivatives of our free software and
+of promoting the sharing and reuse of software generally.
+
+			    NO WARRANTY
+
+  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
+FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
+OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
+PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
+TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
+PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
+REPAIR OR CORRECTION.
+
+  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
+REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
+INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
+OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
+TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
+YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
+PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGES.
+
+		     END OF TERMS AND CONDITIONS

libs/rubberband/1.0/Makefile.in

@@ -0,0 +1,189 @@
+
+CXX		:= @CXX@
+CXXFLAGS	:= -DUSE_PTHREADS -DHAVE_LIBSAMPLERATE -DHAVE_FFTW3 -DFFTW_DOUBLE_ONLY @CXXFLAGS@ @SRC_CFLAGS@ @SNDFILE_CFLAGS@ @FFTW_CFLAGS@ @Vamp_CFLAGS@ -Irubberband -Isrc $(OPTFLAGS)
+LDFLAGS		:= @LDFLAGS@ -lpthread $(LDFLAGS)
+
+LIBRARY_LIBS		:= @SRC_LIBS@ @FFTW_LIBS@ 
+PROGRAM_LIBS		:= @SNDFILE_LIBS@ $(LIBRARY_LIBS)
+VAMP_PLUGIN_LIBS	:= @Vamp_LIBS@ $(LIBRARY_LIBS)
+LADSPA_PLUGIN_LIBS	:= $(LIBRARY_LIBS)
+
+MKDIR			:= mkdir
+AR			:= ar
+
+DYNAMIC_LDFLAGS		:= -shared -Wl,-Bsymbolic
+DYNAMIC_EXTENSION	:= .so
+
+PROGRAM_TARGET 		:= bin/rubberband
+STATIC_TARGET  		:= lib/librubberband.a
+DYNAMIC_TARGET 		:= lib/librubberband$(DYNAMIC_EXTENSION)
+VAMP_TARGET    		:= lib/vamp-rubberband$(DYNAMIC_EXTENSION)
+LADSPA_TARGET  		:= lib/ladspa-rubberband$(DYNAMIC_EXTENSION)
+
+INSTALL_BINDIR		:= @prefix@/bin
+INSTALL_INCDIR		:= @prefix@/include/rubberband
+INSTALL_LIBDIR		:= @prefix@/lib
+INSTALL_VAMPDIR		:= @prefix@/lib/vamp
+INSTALL_LADSPADIR	:= @prefix@/lib/ladspa
+INSTALL_PKGDIR		:= @prefix@/lib/pkgconfig
+
+all:	bin lib $(PROGRAM_TARGET) $(STATIC_TARGET) $(DYNAMIC_TARGET) $(VAMP_TARGET) $(LADSPA_TARGET)
+
+PUBLIC_INCLUDES := \
+	rubberband/TimeStretcher.h \
+	rubberband/RubberBandStretcher.h
+
+LIBRARY_INCLUDES := \
+	src/AudioCurve.h \
+	src/ConstantAudioCurve.h \
+	src/FFT.h \
+	src/HighFrequencyAudioCurve.h \
+	src/PercussiveAudioCurve.h \
+	src/Resampler.h \
+	src/RingBuffer.h \
+	src/Scavenger.h \
+	src/SpectralDifferenceAudioCurve.h \
+	src/StretchCalculator.h \
+	src/StretcherImpl.h \
+	src/StretcherChannelData.h \
+	src/Thread.h \
+	src/Window.h \
+	src/sysutils.h
+
+LIBRARY_SOURCES := \
+	src/RubberBandStretcher.cpp \
+	src/ConstantAudioCurve.cpp \
+	src/HighFrequencyAudioCurve.cpp \
+	src/PercussiveAudioCurve.cpp \
+	src/AudioCurve.cpp \
+	src/Resampler.cpp \
+	src/SpectralDifferenceAudioCurve.cpp \
+	src/StretchCalculator.cpp \
+	src/StretcherImpl.cpp \
+	src/StretcherProcess.cpp \
+	src/StretcherChannelData.cpp \
+	src/FFT.cpp \
+	src/Thread.cpp \
+	src/sysutils.cpp
+
+PROGRAM_SOURCES := \
+	src/main.cpp
+
+VAMP_HEADERS := \
+	src/vamp/RubberBandVampPlugin.h
+
+VAMP_SOURCES := \
+	src/vamp/RubberBandVampPlugin.cpp \
+	src/vamp/libmain.cpp
+
+LADSPA_HEADERS := \
+	src/ladspa/RubberBandPitchShifter.h
+
+LADSPA_SOURCES := \
+	src/ladspa/RubberBandPitchShifter.cpp \
+	src/ladspa/libmain.cpp
+
+LIBRARY_OBJECTS := $(LIBRARY_SOURCES:.cpp=.o)
+PROGRAM_OBJECTS := $(PROGRAM_SOURCES:.cpp=.o)
+VAMP_OBJECTS    := $(VAMP_SOURCES:.cpp=.o)
+LADSPA_OBJECTS  := $(LADSPA_SOURCES:.cpp=.o)
+
+$(PROGRAM_TARGET):	$(LIBRARY_OBJECTS) $(PROGRAM_OBJECTS)
+	$(CXX) -o $@ $^ $(PROGRAM_LIBS) $(PROGRAM_LIBS) $(LDFLAGS)
+
+$(STATIC_TARGET):	$(LIBRARY_OBJECTS)
+	$(AR) rsc $@ $^
+
+$(DYNAMIC_TARGET):	$(LIBRARY_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) $^ -o $@ $(LIBRARY_LIBS) $(LDFLAGS)
+
+$(VAMP_TARGET):		$(LIBRARY_OBJECTS) $(VAMP_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) -o $@ $^ $(VAMP_PLUGIN_LIBS) $(LDFLAGS)
+
+$(LADSPA_TARGET):	$(LIBRARY_OBJECTS) $(LADSPA_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) -o $@ $^ $(LADSPA_PLUGIN_LIBS) $(LDFLAGS)
+
+bin:
+	$(MKDIR) $@
+lib:
+	$(MKDIR) $@
+
+install:	all
+	$(MKDIR) -p $(INSTALL_BINDIR)
+	$(MKDIR) -p $(INSTALL_INCDIR)
+	$(MKDIR) -p $(INSTALL_LIBDIR)
+	$(MKDIR) -p $(INSTALL_VAMPDIR)
+	$(MKDIR) -p $(INSTALL_LADSPADIR)
+	cp $(PROGRAM_TARGET) $(INSTALL_BINDIR)
+	cp $(PUBLIC_INCLUDES) $(INSTALL_INCDIR)
+	cp $(STATIC_TARGET) $(INSTALL_LIBDIR)
+	cp $(DYNAMIC_TARGET) $(INSTALL_LIBDIR)
+	cp $(VAMP_TARGET) $(INSTALL_VAMPDIR)
+	cp src/vamp/vamp-rubberband.cat $(INSTALL_VAMPDIR)
+	cp $(LADSPA_TARGET) $(INSTALL_LADSPADIR)
+	cp src/ladspa/ladspa-rubberband.cat $(INSTALL_LADSPADIR)
+	sed "s,%PREFIX%,@prefix@," rubberband.pc.in \
+	  > $(INSTALL_PKGDIR)/rubberband.pc
+
+clean:
+	rm -f $(LIBRARY_OBJECTS) $(PROGRAM_OBJECTS) $(LADSPA_OBJECTS) $(VAMP_OBJECTS)
+
+distclean:	clean
+	rm -f $(PROGRAM_TARGET) $(STATIC_TARGET) $(DYNAMIC_TARGET) $(VAMP_TARGET) $(LADSPA_TARGET)
+
+# DO NOT DELETE
+
+src/AudioCurve.o: src/AudioCurve.h
+src/ConstantAudioCurve.o: src/ConstantAudioCurve.h src/AudioCurve.h
+src/FFT.o: src/FFT.h src/Thread.h
+src/HighFrequencyAudioCurve.o: src/HighFrequencyAudioCurve.h src/AudioCurve.h
+src/HighFrequencyAudioCurve.o: src/Window.h
+src/main.o: src/sysutils.h
+src/PercussiveAudioCurve.o: src/PercussiveAudioCurve.h src/AudioCurve.h
+src/Resampler.o: src/Resampler.h
+src/RubberBandStretcher.o: src/StretcherImpl.h src/Window.h src/Thread.h
+src/RubberBandStretcher.o: src/RingBuffer.h src/Scavenger.h src/sysutils.h
+src/RubberBandStretcher.o: src/FFT.h
+src/SpectralDifferenceAudioCurve.o: src/SpectralDifferenceAudioCurve.h
+src/SpectralDifferenceAudioCurve.o: src/AudioCurve.h src/Window.h
+src/StretchCalculator.o: src/StretchCalculator.h
+src/StretcherChannelData.o: src/StretcherChannelData.h src/StretcherImpl.h
+src/StretcherChannelData.o: src/Window.h src/Thread.h src/RingBuffer.h
+src/StretcherChannelData.o: src/Scavenger.h src/sysutils.h src/FFT.h
+src/StretcherChannelData.o: src/Resampler.h
+src/StretcherImpl.o: src/StretcherImpl.h src/Window.h src/Thread.h
+src/StretcherImpl.o: src/RingBuffer.h src/Scavenger.h src/sysutils.h
+src/StretcherImpl.o: src/FFT.h src/PercussiveAudioCurve.h src/AudioCurve.h
+src/StretcherImpl.o: src/HighFrequencyAudioCurve.h
+src/StretcherImpl.o: src/SpectralDifferenceAudioCurve.h
+src/StretcherImpl.o: src/ConstantAudioCurve.h src/StretchCalculator.h
+src/StretcherImpl.o: src/StretcherChannelData.h src/Resampler.h
+src/StretcherProcess.o: src/StretcherImpl.h src/Window.h src/Thread.h
+src/StretcherProcess.o: src/RingBuffer.h src/Scavenger.h src/sysutils.h
+src/StretcherProcess.o: src/FFT.h src/PercussiveAudioCurve.h src/AudioCurve.h
+src/StretcherProcess.o: src/HighFrequencyAudioCurve.h
+src/StretcherProcess.o: src/ConstantAudioCurve.h src/StretchCalculator.h
+src/StretcherProcess.o: src/StretcherChannelData.h src/Resampler.h
+src/sysutils.o: src/sysutils.h
+src/Thread.o: src/Thread.h
+src/ConstantAudioCurve.o: src/AudioCurve.h
+src/HighFrequencyAudioCurve.o: src/AudioCurve.h src/Window.h
+src/PercussiveAudioCurve.o: src/AudioCurve.h
+src/RingBuffer.o: src/Scavenger.h src/Thread.h src/sysutils.h
+src/Scavenger.o: src/Thread.h src/sysutils.h
+src/SpectralDifferenceAudioCurve.o: src/AudioCurve.h src/Window.h
+src/StretcherChannelData.o: src/StretcherImpl.h src/Window.h src/Thread.h
+src/StretcherChannelData.o: src/RingBuffer.h src/Scavenger.h src/sysutils.h
+src/StretcherChannelData.o: src/FFT.h
+src/StretcherImpl.o: src/Window.h src/Thread.h src/RingBuffer.h
+src/StretcherImpl.o: src/Scavenger.h src/sysutils.h src/FFT.h
+src/vamp/libmain.o: src/vamp/RubberBandVampPlugin.h
+src/vamp/RubberBandVampPlugin.o: src/vamp/RubberBandVampPlugin.h
+src/vamp/RubberBandVampPlugin.o: src/StretchCalculator.h
+src/ladspa/libmain.o: src/ladspa/RubberBandPitchShifter.h src/RingBuffer.h
+src/ladspa/libmain.o: src/Scavenger.h src/Thread.h src/sysutils.h
+src/ladspa/RubberBandPitchShifter.o: src/ladspa/RubberBandPitchShifter.h
+src/ladspa/RubberBandPitchShifter.o: src/RingBuffer.h src/Scavenger.h
+src/ladspa/RubberBandPitchShifter.o: src/Thread.h src/sysutils.h
+src/ladspa/RubberBandPitchShifter.o: src/RingBuffer.h src/Scavenger.h
+src/ladspa/RubberBandPitchShifter.o: src/Thread.h src/sysutils.h

libs/rubberband/1.0/README

@@ -0,0 +1,158 @@
+
+Rubber Band
+===========
+
+An audio time-stretching and pitch-shifting library and utility program.
+
+Copyright 2007 Chris Cannam, cannam@all-day-breakfast.com.
+
+Distributed under the GNU General Public License.
+
+Rubber Band is a library and utility program that permits you to
+change the tempo and pitch of an audio recording independently of one
+another.
+
+
+Attractive features
+~~~~~~~~~~~~~~~~~~~
+
+  * High quality results suitable for musical use
+
+    Rubber Band is a phase-vocoder-based frequency domain time
+    stretcher with partial phase locking to peak frequencies and phase
+    resynchronisation at noisy transients.  It is suitable for most
+    musical uses with its default settings, and has a range of options
+    for fine tuning.
+
+  * Real-time capable
+
+    In addition to the offline mode (for use in situations where all
+    audio data is available beforehand), Rubber Band supports a true
+    real-time, lock-free streaming mode, in which the time and pitch
+    scaling ratios may be dynamically adjusted during use.
+
+  * Sample-accurate duration adjustment
+
+    In offline mode, Rubber Band ensures that the output has exactly
+    the right number of samples for the given stretch ratio.  (In
+    real-time mode Rubber Band aims to keep as closely as possible to
+    the exact ratio, although this depends on the audio material
+    itself.)
+
+  * Multiprocessor/multi-core support
+
+    Rubber Band's offline mode can take advantage of more than one
+    processor core if available, when processing data with two or more
+    audio channels.
+
+  * No job too big, or too small
+
+    Rubber Band is tuned so as to work well with the default settings
+    for any stretch ratio, from tiny deviations from the original
+    speed to very extreme stretches.
+
+  * Handy utilities included
+
+    The Rubber Band code includes a useful command-line time-stretch
+    and pitch shift utility (called simply rubberband), two LADSPA
+    pitch shifter plugins (Rubber Band Mono Pitch Shifter and Rubber
+    Band Stereo Pitch Shifter), and a Vamp audio analysis plugin which
+    may be used to inspect the stretch profile decisions Rubber Band
+    is taking.
+
+  * Free Software
+
+    Rubber Band is Free Software published under the GNU General
+    Public License.
+
+
+Limitations
+~~~~~~~~~~~
+
+  * Not especially fast
+
+    The algorithm used by Rubber Band is very processor intensive, and
+    Rubber Band is not the fastest implementation on earth.
+
+  * Not especially state of the art
+
+    Rubber Band employs well known algorithms which work well in many
+    situations, but it isn't "cutting edge" in any interesting sense.
+
+  * Relatively complex
+
+    While the fundamental algorithms in Rubber Band are not especially
+    complex, the implementation is complicated by the support for
+    multiple processing modes, exact sample precision, threading, and
+    other features that add to the flexibility of the API.
+
+
+Compiling Rubber Band
+---------------------
+
+Rubber Band is supplied with build scripts that have been tested on
+Linux platforms.  It is also possible to build Rubber Band on other
+platforms, including both POSIX platforms such as OS/X and non-POSIX
+platforms such as Win32.  There are some example Makefiles in the misc
+directory, but if you're using a proprietary platform and you get
+stuck I'm afraid you're on your own, unless you want to pay us...
+
+To build Rubber Band you will also need libsndfile, libsamplerate,
+FFTW3, the Vamp plugin SDK, the LADSPA plugin header, the pthread
+library (except on Win32), and a C++ compiler.  The code has been
+tested with GCC 4.x and with the Intel C++ compiler.
+
+Rubber Band comes with a simple autoconf script.  Run 
+
+  $ ./configure
+  $ make
+
+to compile, and optionally
+
+  # make install
+
+to install.
+
+
+Using the Rubber Band utility
+-----------------------------
+
+The Rubber Band command-line utility builds as bin/rubberband.  The
+basic incantation is
+
+  $ rubberband -t <timeratio> -p <pitchratio> <infile.wav> <outfile.wav>
+
+For example,
+
+  $ rubberband -t 1.5 -p 2.0 test.wav output.wav
+
+stretches the file test.wav to 50% longer than its original duration,
+shifts it up in pitch by one octave, and writes the output to output.wav.
+
+Several further options are available: run "rubberband -h" for help.
+In particular, different types of music may benefit from different
+"crispness" options (-c <n> where <n> is from 0 to 5).
+
+
+Using the Rubber Band library
+-----------------------------
+
+The Rubber Band library has a public API that consists of one C++
+class, called RubberBandStretcher in the RubberBand namespace.  You
+should #include <rubberband/RubberBandStretcher.h> to use this class.
+There is extensive documentation in the class header.
+
+The source code for the command-line utility (src/main.cpp) provides a
+good example of how to use Rubber Band in offline mode; the LADSPA
+pitch shifter plugin (src/ladspa/RubberBandPitchShifter.cpp) may be
+used as an example of Rubber Band in real-time mode.
+
+IMPORTANT: Please ensure you have read and understood the licensing
+terms for Rubber Band before using it in another application.  This
+library is provided under the GNU General Public License, which means
+that any application that uses it must also be published under the GPL
+or a compatible license (i.e. with its full source code also available
+for modification and redistribution).  See the file COPYING for more
+details.  Alternative commercial and proprietary licensing terms are
+available; please contact the author if you are interested.
+

libs/rubberband/1.0/configure.ac

@@ -0,0 +1,38 @@
+
+AC_INIT(RubberBand, 0.1, cannam@all-day-breakfast.com)
+
+AC_CONFIG_SRCDIR(src/StretcherImpl.h)
+AC_PROG_CXX
+AC_HEADER_STDC
+AC_C_BIGENDIAN
+
+PKG_CHECK_MODULES([SRC],[samplerate])
+AC_SUBST(SRC_CFLAGS)
+AC_SUBST(SRC_LIBS)
+
+PKG_CHECK_MODULES([SNDFILE],[sndfile])
+AC_SUBST(SNDFILE_CFLAGS)
+AC_SUBST(SNDFILE_LIBS)
+
+PKG_CHECK_MODULES([FFTW],[fftw3])
+AC_SUBST(FFTW_CFLAGS)
+AC_SUBST(FFTW_LIBS)
+
+AC_CHECK_HEADERS(ladspa.h)
+AC_CHECK_HEADERS(pthread.h)
+
+PKG_CHECK_MODULES([Vamp],[vamp-sdk])
+AC_SUBST(Vamp_CFLAGS)
+AC_SUBST(Vamp_LIBS)
+
+changequote(,)dnl
+if test "x$GCC" = "xyes"; then
+  case " $CXXFLAGS " in
+    *[\ \	]-fPIC\ -Wall[\ \	]*) ;;
+    *) CXXFLAGS="$CXXFLAGS -fPIC -Wall" ;;
+  esac
+fi
+changequote([,])dnl
+
+AC_OUTPUT([Makefile])
+

libs/rubberband/1.0/misc/Makefile.osx

@@ -0,0 +1,144 @@
+
+CXX		= g++
+CXXFLAGS	= -isysroot /Developer/SDKs/MacOSX10.4u.sdk -O3 -arch i386 -arch ppc -msse -msse2 -I../include -I../vamp-plugin-sdk -Irubberband -Isrc
+LDFLAGS		= -L../lib -L../vamp-plugin-sdk/vamp-sdk
+
+LIBRARY_LIBS		= -lsamplerate -lfftw3 -lfftw3f
+PROGRAM_LIBS		= -lsndfile $(LIBRARY_LIBS)
+VAMP_PLUGIN_LIBS	= -lvamp-sdk $(LIBRARY_LIBS)
+LADSPA_PLUGIN_LIBS	= $(LIBRARY_LIBS)
+
+MKDIR		= mkdir
+AR		= ar
+
+PROGRAM_TARGET := bin/rubberband
+STATIC_TARGET  := lib/librubberband.a
+DYNAMIC_TARGET := lib/librubberband.dylib
+VAMP_TARGET    := lib/vamp-rubberband.dylib
+LADSPA_TARGET  := lib/ladspa-rubberband.dylib
+
+#DYNAMIC_LDFLAGS	:= -shared -Wl,-Bsymbolic
+DYNAMIC_LDFLAGS := -dynamiclib
+
+all: bin lib $(PROGRAM_TARGET) $(STATIC_TARGET) $(DYNAMIC_TARGET) $(VAMP_TARGET) $(LADSPA_TARGET)
+
+PUBLIC_INCLUDES := \
+	rubberband/TimeStretcher.h \
+	rubberband/RubberBandStretcher.h
+
+LIBRARY_INCLUDES := \
+	src/AudioCurve.h \
+	src/ConstantAudioCurve.h \
+	src/FFT.h \
+	src/HighFrequencyAudioCurve.h \
+	src/PercussiveAudioCurve.h \
+	src/Resampler.h \
+	src/RingBuffer.h \
+	src/Scavenger.h \
+	src/StretchCalculator.h \
+	src/StretcherImpl.h \
+	src/StretcherChannelData.h \
+	src/Thread.h \
+	src/Window.h \
+	src/sysutils.h
+
+LIBRARY_SOURCES := \
+	src/RubberBandStretcher.cpp \
+	src/ConstantAudioCurve.cpp \
+	src/HighFrequencyAudioCurve.cpp \
+	src/PercussiveAudioCurve.cpp \
+	src/AudioCurve.cpp \
+	src/Resampler.cpp \
+	src/StretchCalculator.cpp \
+	src/StretcherImpl.cpp \
+	src/StretcherProcess.cpp \
+	src/StretcherChannelData.cpp \
+	src/FFT.cpp \
+	src/Thread.cpp \
+	src/sysutils.cpp
+
+PROGRAM_SOURCES := \
+	src/main.cpp
+
+VAMP_HEADERS := \
+	src/vamp/RubberBandVampPlugin.h
+
+VAMP_SOURCES := \
+	src/vamp/RubberBandVampPlugin.cpp \
+	src/vamp/libmain.cpp
+
+LADSPA_HEADERS := \
+	src/ladspa/RubberBandPitchShifter.h
+
+LADSPA_SOURCES := \
+	src/ladspa/RubberBandPitchShifter.cpp \
+	src/ladspa/libmain.cpp
+
+LIBRARY_OBJECTS := $(LIBRARY_SOURCES:.cpp=.o)
+PROGRAM_OBJECTS := $(PROGRAM_SOURCES:.cpp=.o)
+VAMP_OBJECTS    := $(VAMP_SOURCES:.cpp=.o)
+LADSPA_OBJECTS  := $(LADSPA_SOURCES:.cpp=.o)
+
+$(PROGRAM_TARGET):	$(LIBRARY_OBJECTS) $(PROGRAM_OBJECTS)
+	$(CXX) -o $@ $^ $(PROGRAM_LIBS) $(PROGRAM_LIBS) $(LDFLAGS)
+
+$(STATIC_TARGET):	$(LIBRARY_OBJECTS)
+	$(AR) rsc $@ $^
+
+$(DYNAMIC_TARGET):	$(LIBRARY_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) $^ -o $@ $(LIBRARY_LIBS) $(LDFLAGS)
+
+$(VAMP_TARGET):		$(LIBRARY_OBJECTS) $(VAMP_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) -o $@ $^ $(VAMP_PLUGIN_LIBS) $(LDFLAGS)
+
+$(LADSPA_TARGET):	$(LIBRARY_OBJECTS) $(LADSPA_OBJECTS)
+	$(CXX) $(DYNAMIC_LDFLAGS) -o $@ $^ $(LADSPA_PLUGIN_LIBS) $(LDFLAGS)
+
+bin:
+	$(MKDIR) $@
+lib:
+	$(MKDIR) $@
+
+clean:
+	rm -f $(LIBRARY_OBJECTS) $(PROGRAM_OBJECTS) $(LADSPA_OBJECTS) $(VAMP_OBJECTS)
+
+distclean:	clean
+	rm -f $(PROGRAM_TARGET) $(STATIC_TARGET) $(DYNAMIC_TARGET) $(VAMP_TARGET) $(LADSPA_TARGET)
+
+# DO NOT DELETE
+
+src/AudioCurve.o: src/AudioCurve.h
+src/ConstantAudioCurve.o: src/ConstantAudioCurve.h src/AudioCurve.h
+src/FFT.o: src/FFT.h src/Thread.h
+src/HighFrequencyAudioCurve.o: src/HighFrequencyAudioCurve.h src/AudioCurve.h
+src/HighFrequencyAudioCurve.o: src/Window.h
+src/main.o: rubberband/RubberBandStretcher.h rubberband/TimeStretcher.h
+src/PercussiveAudioCurve.o: src/PercussiveAudioCurve.h src/AudioCurve.h
+src/Resampler.o: src/Resampler.h
+src/RubberBandStretcher.o: src/StretcherImpl.h
+src/RubberBandStretcher.o: rubberband/RubberBandStretcher.h
+src/RubberBandStretcher.o: rubberband/TimeStretcher.h src/Window.h
+src/RubberBandStretcher.o: src/Thread.h src/RingBuffer.h src/Scavenger.h
+src/RubberBandStretcher.o: src/FFT.h src/sysutils.h
+src/StretchCalculator.o: src/StretchCalculator.h
+src/StretcherChannelData.o: src/StretcherChannelData.h src/StretcherImpl.h
+src/StretcherChannelData.o: rubberband/RubberBandStretcher.h
+src/StretcherChannelData.o: rubberband/TimeStretcher.h src/Window.h
+src/StretcherChannelData.o: src/Thread.h src/RingBuffer.h src/Scavenger.h
+src/StretcherChannelData.o: src/FFT.h src/sysutils.h src/Resampler.h
+src/StretcherImpl.o: src/StretcherImpl.h rubberband/RubberBandStretcher.h
+src/StretcherImpl.o: rubberband/TimeStretcher.h src/Window.h src/Thread.h
+src/StretcherImpl.o: src/RingBuffer.h src/Scavenger.h src/FFT.h
+src/StretcherImpl.o: src/sysutils.h src/PercussiveAudioCurve.h
+src/StretcherImpl.o: src/AudioCurve.h src/HighFrequencyAudioCurve.h
+src/StretcherImpl.o: src/ConstantAudioCurve.h src/StretchCalculator.h
+src/StretcherImpl.o: src/StretcherChannelData.h src/Resampler.h
+src/StretcherProcess.o: src/StretcherImpl.h rubberband/RubberBandStretcher.h
+src/StretcherProcess.o: rubberband/TimeStretcher.h src/Window.h src/Thread.h
+src/StretcherProcess.o: src/RingBuffer.h src/Scavenger.h src/FFT.h
+src/StretcherProcess.o: src/sysutils.h src/PercussiveAudioCurve.h
+src/StretcherProcess.o: src/AudioCurve.h src/HighFrequencyAudioCurve.h
+src/StretcherProcess.o: src/ConstantAudioCurve.h src/StretchCalculator.h
+src/StretcherProcess.o: src/StretcherChannelData.h src/Resampler.h
+src/sysutils.o: src/sysutils.h
+src/Thread.o: src/Thread.h

libs/rubberband/1.0/rubberband.pc.in

@@ -0,0 +1,10 @@
+prefix=%PREFIX%
+exec_prefix=${prefix}
+libdir=${exec_prefix}/lib
+includedir=${prefix}/include
+
+Name: rubberband
+Version: 1.0
+Description: 
+Libs: -L${libdir} -lrubberband
+Cflags: -I${includedir} 

libs/rubberband/1.0/rubberband/RubberBandStretcher.h

@@ -0,0 +1,410 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBANDSTRETCHER_H_
+#define _RUBBERBANDSTRETCHER_H_
+
+#include "TimeStretcher.h"
+
+#include <vector>
+
+namespace RubberBand
+{
+
+class RubberBandStretcher : public TimeStretcher
+{
+public:
+
+    /**
+     * Processing options for the timestretcher.  The preferred
+     * options should normally be set in the constructor, as a bitwise
+     * OR of the option flags.  The default value (DefaultOptions) is
+     * intended to give good results in most situations.
+     *
+     * 1. Flags prefixed \c OptionProcess determine how the timestretcher
+     * will be invoked.  These options may not be changed after
+     * construction.
+     * 
+     *   \li \c OptionProcessOffline - Run the stretcher in offline
+     *   mode.  In this mode the input data needs to be provided
+     *   twice, once to study(), which calculates a stretch profile
+     *   for the audio, and once to process(), which stretches it.
+     *
+     *   \li \c OptionProcessRealTime - Run the stretcher in real-time
+     *   mode.  In this mode only process() should be called, and the
+     *   stretcher adjusts dynamically in response to the input audio.
+     * 
+     * The Process setting is likely to depend on your architecture:
+     * non-real-time operation on seekable files: Offline; real-time
+     * or streaming operation: RealTime.
+     *
+     * 2. Flags prefixed \c OptionStretch control the profile used for
+     * variable timestretching.  Rubber Band always adjusts the
+     * stretch profile to minimise stretching of busy broadband
+     * transient sounds, but the degree to which it does so is
+     * adjustable.  These options may not be changed after
+     * construction.
+     *
+     *   \li \c OptionStretchElastic - Only meaningful in offline
+     *   mode, and the default in that mode.  The audio will be
+     *   stretched at a variable rate, aimed at preserving the quality
+     *   of transient sounds as much as possible.  The timings of low
+     *   activity regions between transients may be less exact than
+     *   when the precise flag is set.
+     * 
+     *   \li \c OptionStretchPrecise - Although still using a variable
+     *   stretch rate, the audio will be stretched so as to maintain
+     *   as close as possible to a linear stretch ratio throughout.
+     *   Timing may be better than when using \c OptionStretchElastic, at
+     *   slight cost to the sound quality of transients.  This setting
+     *   is always used when running in real-time mode.
+     *
+     * 3. Flags prefixed \c OptionTransients control the component
+     * frequency phase-reset mechanism that may be used at transient
+     * points to provide clarity and realism to percussion and other
+     * significant transient sounds.  These options may be changed
+     * after construction when running in real-time mode, but not when
+     * running in offline mode.
+     * 
+     *   \li \c OptionTransientsCrisp - Reset component phases at the
+     *   peak of each transient (the start of a significant note or
+     *   percussive event).  This, the default setting, usually
+     *   results in a clear-sounding output; but it is not always
+     *   consistent, and may cause interruptions in stable sounds
+     *   present at the same time as transient events.
+     *
+     *   \li \c OptionTransientsMixed - Reset component phases at the
+     *   peak of each transient, outside a frequency range typical of
+     *   musical fundamental frequencies.  The results may be more
+     *   regular for mixed stable and percussive notes than
+     *   \c OptionTransientsCrisp, but with a "phasier" sound.  The
+     *   balance may sound very good for certain types of music and
+     *   fairly bad for others.
+     *
+     *   \li \c OptionTransientsSmooth - Do not reset component phases
+     *   at any point.  The results will be smoother and more regular
+     *   but may be less clear than with either of the other
+     *   transients flags.
+     *
+     * 4. Flags prefixed \c OptionPhase control the adjustment of
+     * component frequency phases from one analysis window to the next
+     * during non-transient segments.  These options may be changed at
+     * any time.
+     *
+     *   \li \c OptionPhaseAdaptive - Lock the adjustments of phase
+     *   for frequencies close to peak frequencies to those of the
+     *   peak, but reduce the degree of locking as the stretch ratio
+     *   gets longer.  This, the default setting, should give a good
+     *   balance between clarity and smoothness in most situations.
+     *
+     *   \li \c OptionPhasePeakLocked - Lock the adjustments of phase
+     *   for frequencies close to peak frequencies to those of the
+     *   peak.  This should give a clear result in situations with
+     *   relatively low stretch ratios, but a relatively metallic
+     *   sound at longer stretches.
+     *
+     *   \li \c OptionPhaseIndependent - Do not lock phase adjustments
+     *   to peak frequencies.  This usually results in a softer,
+     *   phasier sound.
+     *
+     * 5. Flags prefixed \c OptionThreading control the threading
+     * model of the stretcher.  These options may not be changed after
+     * construction.
+     *
+     *   \li \c OptionThreadingAuto - Permit the stretcher to
+     *   determine its own threading model.  Usually this means using
+     *   one processing thread per audio channel in offline mode if
+     *   the stretcher is able to determine that more than one CPU is
+     *   available, and one thread only in realtime mode.
+     *
+     *   \li \c OptionThreadingNever - Never use more than one thread.
+     *  
+     *   \li \c OptionThreadingAlways - Use multiple threads in any
+     *   situation where \c OptionThreadingAuto would do so, except omit
+     *   the check for multiple CPUs and instead assume it to be true.
+     *
+     * 6. Flags prefixed \c OptionWindow control the window size for
+     * FFT processing.  The window size actually used will depend on
+     * many factors, but it can be influenced.  These options may not
+     * be changed after construction.
+     *
+     *   \li \c OptionWindowStandard - Use the default window size.
+     *   The actual size will vary depending on other parameters.
+     *   This option is expected to produce better results than the
+     *   other window options in most situations.
+     *
+     *   \li \c OptionWindowShort - Use a shorter window.  This may
+     *   result in crisper sound for audio that depends strongly on
+     *   its timing qualities.
+     *
+     *   \li \c OptionWindowLong - Use a longer window.  This is
+     *   likely to result in a smoother sound at the expense of
+     *   clarity and timing.
+     */
+    typedef int Options;
+    
+    static const int OptionProcessOffline   = 0x00000000;
+    static const int OptionProcessRealTime  = 0x00000001;
+
+    static const int OptionStretchElastic   = 0x00000000;
+    static const int OptionStretchPrecise   = 0x00000010;
+    
+    static const int OptionTransientsCrisp  = 0x00000000;
+    static const int OptionTransientsMixed  = 0x00000100;
+    static const int OptionTransientsSmooth = 0x00000200;
+
+    static const int OptionPhaseAdaptive    = 0x00000000;
+    static const int OptionPhasePeakLocked  = 0x00001000;
+    static const int OptionPhaseIndependent = 0x00002000;
+    
+    static const int OptionThreadingAuto    = 0x00000000;
+    static const int OptionThreadingNever   = 0x00010000;
+    static const int OptionThreadingAlways  = 0x00020000;
+
+    static const int OptionWindowStandard   = 0x00000000;
+    static const int OptionWindowShort      = 0x00100000;
+    static const int OptionWindowLong       = 0x00200000;
+
+    static const int DefaultOptions         = 0x00000000;
+    static const int PercussiveOptions      = OptionWindowShort | \
+                                              OptionPhaseIndependent;
+
+    /**
+     * Construct a time and pitch stretcher object to run at the given
+     * sample rate, with the given number of channels.  Processing
+     * options and the time and pitch scaling ratios may be provided.
+     * The time and pitch ratios may be changed after construction,
+     * but most of the options may not.  See the option documentation
+     * above for more details.
+     */
+    RubberBandStretcher(size_t sampleRate,
+                        size_t channels,
+                        Options options = DefaultOptions,
+                        double initialTimeRatio = 1.0,
+                        double initialPitchScale = 1.0);
+    virtual ~RubberBandStretcher();
+
+    /**
+     * Reset the stretcher's internal buffers.  The stretcher should
+     * subsequently behave as if it had just been constructed
+     * (although retaining the current time and pitch ratio).
+     */
+    virtual void reset();
+
+    /**
+     * Set the time ratio for the stretcher.  This is the ratio of
+     * stretched to unstretched duration -- not tempo.  For example, a
+     * ratio of 2.0 would make the audio twice as long (i.e. halve the
+     * tempo); 0.5 would make it half as long (i.e. double the tempo);
+     * 1.0 would leave the duration unaffected.
+     *
+     * If the stretcher was constructed in Offline mode, the time
+     * ratio is fixed throughout operation; this function may be
+     * called any number of times between construction (or a call to
+     * reset()) and the first call to study() or process(), but may
+     * not be called after study() or process() has been called.
+     *
+     * If the stretcher was constructed in RealTime mode, the time
+     * ratio may be varied during operation; this function may be
+     * called at any time, so long as it is not called concurrently
+     * with process().  You should either call this function from the
+     * same thread as process(), or provide your own mutex or similar
+     * mechanism to ensure that setTimeRatio and process() cannot be
+     * run at once (there is no internal mutex for this purpose).
+     */
+    virtual void setTimeRatio(double ratio);
+
+    /**
+     * Set the pitch scaling ratio for the stretcher.  This is the
+     * ratio of target frequency to source frequency.  For example, a
+     * ratio of 2.0 would shift up by one octave; 0.5 down by one
+     * octave; or 1.0 leave the pitch unaffected.
+     *
+     * To put this in musical terms, a pitch scaling ratio
+     * corresponding to a shift of S equal-tempered semitones (where S
+     * is positive for an upwards shift and negative for downwards) is
+     * pow(2.0, S / 12.0).
+     *
+     * If the stretcher was constructed in Offline mode, the pitch
+     * scaling ratio is fixed throughout operation; this function may
+     * be called any number of times between construction (or a call
+     * to reset()) and the first call to study() or process(), but may
+     * not be called after study() or process() has been called.
+     *
+     * If the stretcher was constructed in RealTime mode, the pitch
+     * scaling ratio may be varied during operation; this function may
+     * be called at any time, so long as it is not called concurrently
+     * with process().  You should either call this function from the
+     * same thread as process(), or provide your own mutex or similar
+     * mechanism to ensure that setPitchScale and process() cannot be
+     * run at once (there is no internal mutex for this purpose).
+     */
+    virtual void setPitchScale(double scale);
+
+    /**
+     * Return the last time ratio value that was set (either on
+     * construction or with setTimeRatio()).
+     */
+    virtual double getTimeRatio() const;
+
+    /**
+     * Return the last pitch scaling ratio value that was set (either
+     * on construction or with setPitchScale()).
+     */
+    virtual double getPitchScale() const;
+
+    /**
+     * Return the processing latency of the stretcher.  This is the
+     * number of audio samples that one would have to discard at the
+     * start of the output in order to ensure that the resulting audio
+     * aligned with the input audio at the start.  In Offline mode,
+     * latency is automatically adjusted for and the result is zero.
+     * In RealTime mode, the latency may depend on the time and pitch
+     * ratio and other options.
+     */
+    virtual size_t getLatency() const;
+
+    /**
+     * Change an OptionTransients configuration setting.  This may be
+     * called at any time in RealTime mode.  It may not be called in
+     * Offline mode (for which the transients option is fixed on
+     * construction).
+     */
+    virtual void setTransientsOption(Options options);
+
+    /**
+     * Change an OptionPhase configuration setting.  This may be
+     * called at any time in any mode.
+     *
+     * Note that if running multi-threaded in Offline mode, the change
+     * may not take effect immediately if processing is already under
+     * way when this function is called.
+     */
+    virtual void setPhaseOption(Options options);
+
+    /**
+     * Tell the stretcher exactly how many input samples it will
+     * receive.  This is only useful in Offline mode, when it allows
+     * the stretcher to ensure that the number of output samples is
+     * exactly correct.  In RealTime mode no such guarantee is
+     * possible and this value is ignored.
+     */
+    virtual void setExpectedInputDuration(size_t samples);
+
+    /**
+     * Ask the stretcher how many audio sample frames should be
+     * provided as input in order to ensure that some more output
+     * becomes available.  Normal usage consists of querying this
+     * function, providing that number of samples to process(),
+     * reading the output using available() and retrieve(), and then
+     * repeating.
+     *
+     * Note that this value is only relevant to process(), not to
+     * study() (to which you may pass any number of samples at a time,
+     * and from which there is no output).
+     */
+     virtual size_t getSamplesRequired() const;
+
+    /**
+     * Tell the stretcher the maximum number of sample frames that you
+     * will ever be passing in to a single process() call. If you
+     * don't call this function, the stretcher will assume that you
+     * never pass in more samples than getSamplesRequired() suggested
+     * you should.  You should not pass in more samples than that
+     * unless you have called setMaxProcessSize first.
+     *
+     * This function may not be called after the first call to study()
+     * or process().
+     *
+     * Note that this value is only relevant to process(), not to
+     * study() (to which you may pass any number of samples at a time,
+     * and from which there is no output).
+     */
+    virtual void setMaxProcessSize(size_t samples);
+
+    /**
+     * Provide a block of "samples" sample frames for the stretcher to
+     * study and calculate a stretch profile from.
+     *
+     * This is only meaningful in Offline mode, and is required if
+     * running in that mode.  You should pass the entire input through
+     * study() before any process() calls are made, as a sequence of
+     * blocks in individual study() calls, or as a single large block.
+     *
+     * "input" should point to de-interleaved audio data with one
+     * float array per channel.  "samples" supplies the number of
+     * audio sample frames available in "input".  If "samples" is
+     * zero, "input" may be NULL.
+     * 
+     * Set "final" to true if this is the last block of data that will
+     * be provided to study() before the first process() call.
+     */
+    virtual void study(const float *const *input, size_t samples, bool final);
+
+    /**
+     * Provide a block of "samples" sample frames for processing.
+     * See also getSamplesRequired() and setMaxProcessSize().
+     *
+     * Set "final" to true if this is the last block of input data.
+     */
+    virtual void process(const float *const *input, size_t samples, bool final);
+
+    /**
+     * Ask the stretcher how many audio sample frames of output data
+     * are available for reading (via retrieve()).
+     * 
+     * This function returns 0 if no frames are available: this
+     * usually means more input data needs to be provided, but if the
+     * stretcher is running in threaded mode it may just mean that not
+     * enough data has yet been processed.  Call getSamplesRequired()
+     * to discover whether more input is needed.
+     *
+     * This function returns -1 if all data has been fully processed
+     * and all output read, and the stretch process is now finished.
+     */
+    virtual int available() const;
+
+    /**
+     * Obtain some processed output data from the stretcher.  Up to
+     * "samples" samples will be stored in the output arrays (one per
+     * channel for de-interleaved audio data) pointed to by "output".
+     * The return value is the actual number of sample frames
+     * retrieved.
+     */
+    virtual size_t retrieve(float *const *output, size_t samples) const;
+
+    virtual float getFrequencyCutoff(int n) const;
+    virtual void setFrequencyCutoff(int n, float f);
+    
+    virtual size_t getInputIncrement() const;
+    virtual std::vector<int> getOutputIncrements() const; //!!! document particular meaning in RT mode
+    virtual std::vector<float> getPhaseResetCurve() const; //!!! document particular meaning in RT mode
+    virtual std::vector<int> getExactTimePoints() const; //!!! meaningless in RT mode
+
+    virtual size_t getChannelCount() const;
+    
+    virtual void calculateStretch();
+
+    virtual void setDebugLevel(int level);
+
+    static void setDefaultDebugLevel(int level);
+
+protected:
+    class Impl;
+    Impl *m_d;
+};
+
+}
+
+#endif

libs/rubberband/1.0/rubberband/TimeStretcher.h

@@ -0,0 +1,58 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_TIMESTRETCHER_H_
+#define _RUBBERBAND_TIMESTRETCHER_H_
+
+#include <sys/types.h>
+
+namespace RubberBand
+{
+
+/**
+ * Base class for time stretchers.  RubberBand currently provides only
+ * a single subclass implementation.
+ *
+ * @see RubberBandStretcher
+ */
+class TimeStretcher
+{
+public:
+    TimeStretcher(size_t sampleRate, size_t channels) :
+        m_sampleRate(sampleRate),
+        m_channels(channels)
+    { }
+    virtual ~TimeStretcher()
+    { }
+
+    virtual void reset() = 0;
+    virtual void setTimeRatio(double ratio) = 0;
+    virtual void setPitchScale(double scale) = 0;
+    virtual size_t getLatency() const = 0;
+
+    virtual void study(const float *const *input, size_t samples, bool final) = 0;
+    virtual size_t getSamplesRequired() const = 0;
+    virtual void process(const float *const *input, size_t samples, bool final) = 0;
+    virtual int available() const = 0;
+    virtual size_t retrieve(float *const *output, size_t samples) const = 0;
+
+protected:
+    size_t m_sampleRate;
+    size_t m_channels;
+};
+
+}
+
+#endif
+    

libs/rubberband/1.0/src/AudioCurve.cpp

@@ -0,0 +1,31 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "AudioCurve.h"
+
+namespace RubberBand
+{
+
+AudioCurve::AudioCurve(size_t sampleRate, size_t windowSize) :
+    m_sampleRate(sampleRate),
+    m_windowSize(windowSize)
+{
+}
+
+AudioCurve::~AudioCurve()
+{
+}
+
+
+}

libs/rubberband/1.0/src/AudioCurve.h

@@ -0,0 +1,42 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _AUDIO_CURVE_H_
+#define _AUDIO_CURVE_H_
+
+#include <sys/types.h>
+
+namespace RubberBand 
+{
+
+class AudioCurve
+{
+public:
+    AudioCurve(size_t sampleRate, size_t windowSize);
+    virtual ~AudioCurve();
+
+    virtual void setWindowSize(size_t newSize) = 0;
+    
+    virtual float process(float *mag, size_t increment) = 0;
+    virtual void reset() = 0;
+
+protected:
+    size_t m_sampleRate;
+    size_t m_windowSize;
+};
+
+}
+
+#endif
+

libs/rubberband/1.0/src/ConstantAudioCurve.cpp

@@ -0,0 +1,47 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "ConstantAudioCurve.h"
+
+namespace RubberBand
+{
+
+ConstantAudioCurve::ConstantAudioCurve(size_t sampleRate, size_t windowSize) :
+    AudioCurve(sampleRate, windowSize)
+{
+}
+
+ConstantAudioCurve::~ConstantAudioCurve()
+{
+}
+
+void
+ConstantAudioCurve::reset()
+{
+}
+
+void
+ConstantAudioCurve::setWindowSize(size_t newSize)
+{
+    m_windowSize = newSize;
+}
+
+float
+ConstantAudioCurve::process(float *, size_t)
+{
+    return 1.f;
+}
+
+}
+

libs/rubberband/1.0/src/ConstantAudioCurve.h

@@ -0,0 +1,37 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _CONSTANT_AUDIO_CURVE_H_
+#define _CONSTANT_AUDIO_CURVE_H_
+
+#include "AudioCurve.h"
+
+namespace RubberBand
+{
+
+class ConstantAudioCurve : public AudioCurve
+{
+public:
+    ConstantAudioCurve(size_t sampleRate, size_t windowSize);
+    virtual ~ConstantAudioCurve();
+
+    virtual void setWindowSize(size_t newSize);
+
+    virtual float process(float *mag, size_t increment);
+    virtual void reset();
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/FFT.cpp

@@ -0,0 +1,867 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "FFT.h"
+#include "Thread.h"
+
+
+#include <fftw3.h>
+
+#include <cmath>
+#include <iostream>
+#include <map>
+#include <cstdio>
+#include <vector>
+
+namespace RubberBand {
+
+class FFTImpl
+{
+public:
+    virtual ~FFTImpl() { }
+
+    virtual void initFloat() = 0;
+    virtual void initDouble() = 0;
+
+    virtual void forward(double *realIn, double *realOut, double *imagOut) = 0;
+    virtual void forwardPolar(double *realIn, double *magOut, double *phaseOut) = 0;
+    virtual void forwardMagnitude(double *realIn, double *magOut) = 0;
+
+    virtual void forward(float *realIn, float *realOut, float *imagOut) = 0;
+    virtual void forwardPolar(float *realIn, float *magOut, float *phaseOut) = 0;
+    virtual void forwardMagnitude(float *realIn, float *magOut) = 0;
+
+    virtual void inverse(double *realIn, double *imagIn, double *realOut) = 0;
+    virtual void inversePolar(double *magIn, double *phaseIn, double *realOut) = 0;
+
+    virtual void inverse(float *realIn, float *imagIn, float *realOut) = 0;
+    virtual void inversePolar(float *magIn, float *phaseIn, float *realOut) = 0;
+
+    virtual float *getFloatTimeBuffer() = 0;
+    virtual double *getDoubleTimeBuffer() = 0;
+};    
+
+
+
+
+// Define FFTW_DOUBLE_ONLY to make all uses of FFTW functions be
+// double-precision (so "float" FFTs are calculated by casting to
+// doubles and using the double-precision FFTW function).
+//
+// Define FFTW_FLOAT_ONLY to make all uses of FFTW functions be
+// single-precision (so "double" FFTs are calculated by casting to
+// floats and using the single-precision FFTW function).
+//
+// Neither of these flags is terribly desirable -- FFTW_FLOAT_ONLY
+// obviously loses you precision, and neither is handled in the most
+// efficient way so any performance improvement will be small at best.
+// The only real reason to define either flag would be to avoid
+// linking against both fftw3 and fftw3f libraries.
+
+//#define FFTW_DOUBLE_ONLY 1
+//#define FFTW_FLOAT_ONLY 1
+
+#ifdef FFTW_DOUBLE_ONLY
+#ifdef FFTW_FLOAT_ONLY
+// Can't meaningfully define both
+#undef FFTW_DOUBLE_ONLY
+#undef FFTW_FLOAT_ONLY
+#else /* !FFTW_FLOAT_ONLY */
+#define fftwf_complex fftw_complex
+#define fftwf_plan fftw_plan
+#define fftwf_plan_dft_r2c_1d fftw_plan_dft_r2c_1d
+#define fftwf_plan_dft_c2r_1d fftw_plan_dft_c2r_1d
+#define fftwf_destroy_plan fftw_destroy_plan
+#define fftwf_malloc fftw_malloc
+#define fftwf_free fftw_free
+#define fftwf_execute fftw_execute
+#define atan2f atan2
+#define sqrtf sqrt
+#define cosf cos
+#define sinf sin
+#endif /* !FFTW_FLOAT_ONLY */
+
+#ifdef FFTW_FLOAT_ONLY
+#define fftw_complex fftwf_complex
+#define fftw_plan fftwf_plan
+#define fftw_plan_dft_r2c_1d fftwf_plan_dft_r2c_1d
+#define fftw_plan_dft_c2r_1d fftwf_plan_dft_c2r_1d
+#define fftw_destroy_plan fftwf_destroy_plan
+#define fftw_malloc fftwf_malloc
+#define fftw_free fftwf_free
+#define fftw_execute fftwf_execute
+#define atan2 atan2f
+#define sqrt sqrtf
+#define cos cosf
+#define sif sinf
+#endif /* FFTW_FLOAT_ONLY */
+
+class D_FFTW : public FFTImpl
+{
+public:
+    D_FFTW(unsigned int size) : m_fplanf(0)
+#ifdef FFTW_DOUBLE_ONLY
+                              , m_frb(0)
+#endif
+                              , m_dplanf(0)
+#ifdef FFTW_FLOAT_ONLY
+                              , m_drb(0)
+#endif
+                              , m_size(size)
+    {
+    }
+
+    ~D_FFTW() {
+        if (m_fplanf) {
+            bool save = false;
+            m_extantMutex.lock();
+            if (m_extantf > 0 && --m_extantf == 0) save = true;
+            m_extantMutex.unlock();
+            if (save) saveWisdom('f');
+            fftwf_destroy_plan(m_fplanf);
+            fftwf_destroy_plan(m_fplani);
+            fftwf_free(m_fbuf);
+            fftwf_free(m_fpacked);
+#ifdef FFTW_DOUBLE_ONLY
+            if (m_frb) fftw_free(m_frb);
+#endif
+        }
+        if (m_dplanf) {
+            bool save = false;
+            m_extantMutex.lock();
+            if (m_extantd > 0 && --m_extantd == 0) save = true;
+            m_extantMutex.unlock();
+            if (save) saveWisdom('d');
+            fftw_destroy_plan(m_dplanf);
+            fftw_destroy_plan(m_dplani);
+            fftw_free(m_dbuf);
+            fftw_free(m_dpacked);
+#ifdef FFTW_FLOAT_ONLY
+            if (m_drb) fftwf_free(m_drb);
+#endif
+        }
+    }
+
+    void initFloat() {
+        if (m_fplanf) return;
+        bool load = false;
+        m_extantMutex.lock();
+        if (m_extantf++ == 0) load = true;
+        m_extantMutex.unlock();
+#ifdef FFTW_DOUBLE_ONLY
+        if (load) loadWisdom('d');
+        m_fbuf = (double *)fftw_malloc(m_size * sizeof(double));
+#else
+        if (load) loadWisdom('f');
+        m_fbuf = (float *)fftwf_malloc(m_size * sizeof(float));
+#endif
+        m_fpacked = (fftwf_complex *)fftw_malloc
+            ((m_size/2 + 1) * sizeof(fftwf_complex));
+        m_fplanf = fftwf_plan_dft_r2c_1d
+            (m_size, m_fbuf, m_fpacked, FFTW_MEASURE);
+        m_fplani = fftwf_plan_dft_c2r_1d
+            (m_size, m_fpacked, m_fbuf, FFTW_MEASURE);
+    }
+
+    void initDouble() {
+        if (m_dplanf) return;
+        bool load = false;
+        m_extantMutex.lock();
+        if (m_extantd++ == 0) load = true;
+        m_extantMutex.unlock();
+#ifdef FFTW_FLOAT_ONLY
+        if (load) loadWisdom('f');
+        m_dbuf = (float *)fftwf_malloc(m_size * sizeof(float));
+#else
+        if (load) loadWisdom('d');
+        m_dbuf = (double *)fftw_malloc(m_size * sizeof(double));
+#endif
+        m_dpacked = (fftw_complex *)fftw_malloc
+            ((m_size/2 + 1) * sizeof(fftw_complex));
+        m_dplanf = fftw_plan_dft_r2c_1d
+            (m_size, m_dbuf, m_dpacked, FFTW_MEASURE);
+        m_dplani = fftw_plan_dft_c2r_1d
+            (m_size, m_dpacked, m_dbuf, FFTW_MEASURE);
+    }
+
+    void loadWisdom(char type) { wisdom(false, type); }
+    void saveWisdom(char type) { wisdom(true, type); }
+
+    void wisdom(bool save, char type) {
+
+#ifdef FFTW_DOUBLE_ONLY
+        if (type == 'f') return;
+#endif
+#ifdef FFTW_FLOAT_ONLY
+        if (type == 'd') return;
+#endif
+
+        const char *home = getenv("HOME");
+        if (!home) return;
+
+        char fn[256];
+        snprintf(fn, 256, "%s/%s.%c", home, ".rubberband.wisdom", type);
+
+        FILE *f = fopen(fn, save ? "wb" : "rb");
+        if (!f) return;
+
+        if (save) {
+            switch (type) {
+#ifdef FFTW_DOUBLE_ONLY
+            case 'f': break;
+#else
+            case 'f': fftwf_export_wisdom_to_file(f); break;
+#endif
+#ifdef FFTW_FLOAT_ONLY
+            case 'd': break;
+#else
+            case 'd': fftw_export_wisdom_to_file(f); break;
+#endif
+            default: break;
+            }
+        } else {
+            switch (type) {
+#ifdef FFTW_DOUBLE_ONLY
+            case 'f': break;
+#else
+            case 'f': fftwf_import_wisdom_from_file(f); break;
+#endif
+#ifdef FFTW_FLOAT_ONLY
+            case 'd': break;
+#else
+            case 'd': fftw_import_wisdom_from_file(f); break;
+#endif
+            default: break;
+            }
+        }
+
+        fclose(f);
+    }
+
+    void packFloat(float *re, float *im) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            m_fpacked[i][0] = re[i];
+            m_fpacked[i][1] = im[i];
+        }
+    }
+
+    void packDouble(double *re, double *im) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            m_dpacked[i][0] = re[i];
+            m_dpacked[i][1] = im[i];
+        }
+    }
+
+    void unpackFloat(float *re, float *im) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            re[i] = m_fpacked[i][0];
+            im[i] = m_fpacked[i][1];
+        }
+    }        
+
+    void unpackDouble(double *re, double *im) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            re[i] = m_dpacked[i][0];
+            im[i] = m_dpacked[i][1];
+        }
+    }        
+
+    void forward(double *realIn, double *realOut, double *imagOut) {
+        if (!m_dplanf) initDouble();
+#ifndef FFTW_FLOAT_ONLY
+        if (realIn != m_dbuf) 
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_dbuf[i] = realIn[i];
+            }
+        fftw_execute(m_dplanf);
+        unpackDouble(realOut, imagOut);
+    }
+
+    void forwardPolar(double *realIn, double *magOut, double *phaseOut) {
+        if (!m_dplanf) initDouble();
+#ifndef FFTW_FLOAT_ONLY
+        if (realIn != m_dbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_dbuf[i] = realIn[i];
+            }
+        fftw_execute(m_dplanf);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_dpacked[i][0] * m_dpacked[i][0] +
+                             m_dpacked[i][1] * m_dpacked[i][1]);
+        }
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            phaseOut[i] = atan2(m_dpacked[i][1], m_dpacked[i][0]);
+        }
+    }
+
+    void forwardMagnitude(double *realIn, double *magOut) {
+        if (!m_dplanf) initDouble();
+#ifndef FFTW_FLOAT_ONLY
+        if (realIn != m_dbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_dbuf[i] = realIn[i];
+            }
+        fftw_execute(m_dplanf);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_dpacked[i][0] * m_dpacked[i][0] +
+                             m_dpacked[i][1] * m_dpacked[i][1]);
+        }
+    }
+
+    void forward(float *realIn, float *realOut, float *imagOut) {
+        if (!m_fplanf) initFloat();
+#ifndef FFTW_DOUBLE_ONLY
+        if (realIn != m_fbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_fbuf[i] = realIn[i];
+            }
+        fftwf_execute(m_fplanf);
+        unpackFloat(realOut, imagOut);
+    }
+
+    void forwardPolar(float *realIn, float *magOut, float *phaseOut) {
+        if (!m_fplanf) initFloat();
+#ifndef FFTW_DOUBLE_ONLY
+        if (realIn != m_fbuf) 
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_fbuf[i] = realIn[i];
+            }
+        fftwf_execute(m_fplanf);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrtf(m_fpacked[i][0] * m_fpacked[i][0] +
+                              m_fpacked[i][1] * m_fpacked[i][1]);
+        }
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+          phaseOut[i] = atan2f(m_fpacked[i][1], m_fpacked[i][0]) ;
+        }
+    }
+
+    void forwardMagnitude(float *realIn, float *magOut) {
+        if (!m_fplanf) initFloat();
+#ifndef FFTW_DOUBLE_ONLY
+        if (realIn != m_fbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                m_fbuf[i] = realIn[i];
+            }
+        fftwf_execute(m_fplanf);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrtf(m_fpacked[i][0] * m_fpacked[i][0] +
+                              m_fpacked[i][1] * m_fpacked[i][1]);
+        }
+    }
+
+    void inverse(double *realIn, double *imagIn, double *realOut) {
+        if (!m_dplanf) initDouble();
+        packDouble(realIn, imagIn);
+        fftw_execute(m_dplani);
+#ifndef FFTW_FLOAT_ONLY
+        if (realOut != m_dbuf) 
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                realOut[i] = m_dbuf[i];
+            }
+    }
+
+    void inversePolar(double *magIn, double *phaseIn, double *realOut) {
+        if (!m_dplanf) initDouble();
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            m_dpacked[i][0] = magIn[i] * cos(phaseIn[i]);
+            m_dpacked[i][1] = magIn[i] * sin(phaseIn[i]);
+        }
+        fftw_execute(m_dplani);
+#ifndef FFTW_FLOAT_ONLY
+        if (realOut != m_dbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                realOut[i] = m_dbuf[i];
+            }
+    }
+
+    void inverse(float *realIn, float *imagIn, float *realOut) {
+        if (!m_fplanf) initFloat();
+        packFloat(realIn, imagIn);
+        fftwf_execute(m_fplani);
+#ifndef FFTW_DOUBLE_ONLY
+        if (realOut != m_fbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                realOut[i] = m_fbuf[i];
+            }
+    }
+
+    void inversePolar(float *magIn, float *phaseIn, float *realOut) {
+        if (!m_fplanf) initFloat();
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            m_fpacked[i][0] = magIn[i] * cosf(phaseIn[i]);
+            m_fpacked[i][1] = magIn[i] * sinf(phaseIn[i]);
+        }
+        fftwf_execute(m_fplani);
+#ifndef FFTW_DOUBLE_ONLY
+        if (realOut != m_fbuf)
+#endif
+            for (unsigned int i = 0; i < m_size; ++i) {
+                realOut[i] = m_fbuf[i];
+            }
+    }
+
+    float *getFloatTimeBuffer() {
+        initFloat();
+#ifdef FFTW_DOUBLE_ONLY
+        if (!m_frb) m_frb = (float *)fftw_malloc(m_size * sizeof(float));
+        return m_frb;
+#else
+        return m_fbuf;
+#endif
+    }
+
+    double *getDoubleTimeBuffer() {
+        initDouble();
+#ifdef FFTW_FLOAT_ONLY
+        if (!m_drb) m_drb = (double *)fftwf_malloc(m_size * sizeof(double));
+        return m_drb;
+#else
+        return m_dbuf;
+#endif
+    }
+
+private:
+    fftwf_plan m_fplanf;
+    fftwf_plan m_fplani;
+#ifdef FFTW_DOUBLE_ONLY
+    float *m_frb;
+    double *m_fbuf;
+#else
+    float *m_fbuf;
+#endif
+    fftwf_complex *m_fpacked;
+    fftw_plan m_dplanf;
+    fftw_plan m_dplani;
+#ifdef FFTW_FLOAT_ONLY
+    float *m_dbuf;
+    double *m_drb;
+#else
+    double *m_dbuf;
+#endif
+    fftw_complex *m_dpacked;
+    unsigned int m_size;
+    static unsigned int m_extantf;
+    static unsigned int m_extantd;
+    static Mutex m_extantMutex;
+};
+
+unsigned int
+D_FFTW::m_extantf = 0;
+
+unsigned int
+D_FFTW::m_extantd = 0;
+
+Mutex
+D_FFTW::m_extantMutex;
+
+#endif
+
+class D_Cross : public FFTImpl
+{
+public:
+    D_Cross(unsigned int size) : m_size(size), m_table(0), m_frb(0), m_drb(0) {
+        
+        m_a = new double[size];
+        m_b = new double[size];
+        m_c = new double[size];
+        m_d = new double[size];
+
+        m_table = new int[m_size];
+    
+        unsigned int bits;
+        unsigned int i, j, k, m;
+
+        for (i = 0; ; ++i) {
+            if (m_size & (1 << i)) {
+                bits = i;
+                break;
+            }
+        }
+        
+        for (i = 0; i < m_size; ++i) {
+            
+            m = i;
+            
+            for (j = k = 0; j < bits; ++j) {
+                k = (k << 1) | (m & 1);
+                m >>= 1;
+            }
+            
+            m_table[i] = k;
+        }
+    }
+
+    ~D_Cross() {
+        delete[] m_table;
+        delete[] m_a;
+        delete[] m_b;
+        delete[] m_c;
+        delete[] m_d;
+        delete[] m_frb;
+        delete[] m_drb;
+    }
+
+    void initFloat() { }
+    void initDouble() { }
+
+    void forward(double *realIn, double *realOut, double *imagOut) {
+        basefft(false, realIn, 0, m_c, m_d);
+        for (size_t i = 0; i <= m_size/2; ++i) realOut[i] = m_c[i];
+        for (size_t i = 0; i <= m_size/2; ++i) imagOut[i] = m_d[i];
+    }
+
+    void forwardPolar(double *realIn, double *magOut, double *phaseOut) {
+        basefft(false, realIn, 0, m_c, m_d);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]);
+            phaseOut[i] = atan2(m_d[i], m_c[i]) ;
+        }
+    }
+
+    void forwardMagnitude(double *realIn, double *magOut) {
+        basefft(false, realIn, 0, m_c, m_d);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]);
+        }
+    }
+
+    void forward(float *realIn, float *realOut, float *imagOut) {
+        for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i];
+        basefft(false, m_a, 0, m_c, m_d);
+        for (size_t i = 0; i <= m_size/2; ++i) realOut[i] = m_c[i];
+        for (size_t i = 0; i <= m_size/2; ++i) imagOut[i] = m_d[i];
+    }
+
+    void forwardPolar(float *realIn, float *magOut, float *phaseOut) {
+        for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i];
+        basefft(false, m_a, 0, m_c, m_d);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]);
+            phaseOut[i] = atan2(m_d[i], m_c[i]) ;
+        }
+    }
+
+    void forwardMagnitude(float *realIn, float *magOut) {
+        for (size_t i = 0; i < m_size; ++i) m_a[i] = realIn[i];
+        basefft(false, m_a, 0, m_c, m_d);
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            magOut[i] = sqrt(m_c[i] * m_c[i] + m_d[i] * m_d[i]);
+        }
+    }
+
+    void inverse(double *realIn, double *imagIn, double *realOut) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            double real = realIn[i];
+            double imag = imagIn[i];
+            m_a[i] = real;
+            m_b[i] = imag;
+            if (i > 0) {
+                m_a[m_size-i] = real;
+                m_b[m_size-i] = -imag;
+            }
+        }
+        basefft(true, m_a, m_b, realOut, m_d);
+    }
+
+    void inversePolar(double *magIn, double *phaseIn, double *realOut) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            double real = magIn[i] * cos(phaseIn[i]);
+            double imag = magIn[i] * sin(phaseIn[i]);
+            m_a[i] = real;
+            m_b[i] = imag;
+            if (i > 0) {
+                m_a[m_size-i] = real;
+                m_b[m_size-i] = -imag;
+            }
+        }
+        basefft(true, m_a, m_b, realOut, m_d);
+    }
+
+    void inverse(float *realIn, float *imagIn, float *realOut) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            float real = realIn[i];
+            float imag = imagIn[i];
+            m_a[i] = real;
+            m_b[i] = imag;
+            if (i > 0) {
+                m_a[m_size-i] = real;
+                m_b[m_size-i] = -imag;
+            }
+        }
+        basefft(true, m_a, m_b, m_c, m_d);
+        for (unsigned int i = 0; i < m_size; ++i) realOut[i] = m_c[i];
+    }
+
+    void inversePolar(float *magIn, float *phaseIn, float *realOut) {
+        for (unsigned int i = 0; i <= m_size/2; ++i) {
+            float real = magIn[i] * cosf(phaseIn[i]);
+            float imag = magIn[i] * sinf(phaseIn[i]);
+            m_a[i] = real;
+            m_b[i] = imag;
+            if (i > 0) {
+                m_a[m_size-i] = real;
+                m_b[m_size-i] = -imag;
+            }
+        }
+        basefft(true, m_a, m_b, m_c, m_d);
+        for (unsigned int i = 0; i < m_size; ++i) realOut[i] = m_c[i];
+    }
+
+    float *getFloatTimeBuffer() {
+        if (!m_frb) m_frb = new float[m_size];
+        return m_frb;
+    }
+
+    double *getDoubleTimeBuffer() {
+        if (!m_drb) m_drb = new double[m_size];
+        return m_drb;
+    }
+
+private:
+    unsigned int m_size;
+    int *m_table;
+    float *m_frb;
+    double *m_drb;
+    double *m_a;
+    double *m_b;
+    double *m_c;
+    double *m_d;
+    void basefft(bool inverse, double *ri, double *ii, double *ro, double *io);
+};
+
+void
+D_Cross::basefft(bool inverse, double *ri, double *ii, double *ro, double *io)
+{
+    if (!ri || !ro || !io) return;
+
+    unsigned int i, j, k, m;
+    unsigned int blockSize, blockEnd;
+
+    double tr, ti;
+
+    double angle = 2.0 * M_PI;
+    if (inverse) angle = -angle;
+
+    const unsigned int n = m_size;
+
+    if (ii) {
+	for (i = 0; i < n; ++i) {
+	    ro[m_table[i]] = ri[i];
+	    io[m_table[i]] = ii[i];
+	}
+    } else {
+	for (i = 0; i < n; ++i) {
+	    ro[m_table[i]] = ri[i];
+	    io[m_table[i]] = 0.0;
+	}
+    }
+
+    blockEnd = 1;
+
+    for (blockSize = 2; blockSize <= n; blockSize <<= 1) {
+
+	double delta = angle / (double)blockSize;
+	double sm2 = -sin(-2 * delta);
+	double sm1 = -sin(-delta);
+	double cm2 = cos(-2 * delta);
+	double cm1 = cos(-delta);
+	double w = 2 * cm1;
+	double ar[3], ai[3];
+
+	for (i = 0; i < n; i += blockSize) {
+
+	    ar[2] = cm2;
+	    ar[1] = cm1;
+
+	    ai[2] = sm2;
+	    ai[1] = sm1;
+
+	    for (j = i, m = 0; m < blockEnd; j++, m++) {
+
+		ar[0] = w * ar[1] - ar[2];
+		ar[2] = ar[1];
+		ar[1] = ar[0];
+
+		ai[0] = w * ai[1] - ai[2];
+		ai[2] = ai[1];
+		ai[1] = ai[0];
+
+		k = j + blockEnd;
+		tr = ar[0] * ro[k] - ai[0] * io[k];
+		ti = ar[0] * io[k] + ai[0] * ro[k];
+
+		ro[k] = ro[j] - tr;
+		io[k] = io[j] - ti;
+
+		ro[j] += tr;
+		io[j] += ti;
+	    }
+	}
+
+	blockEnd = blockSize;
+    }
+
+/* fftw doesn't rescale, so nor will we
+
+    if (inverse) {
+
+	double denom = (double)n;
+
+	for (i = 0; i < n; i++) {
+	    ro[i] /= denom;
+	    io[i] /= denom;
+	}
+    }
+*/
+}
+
+int
+FFT::m_method = -1;
+
+FFT::FFT(unsigned int size)
+{
+    if (size < 2) throw InvalidSize;
+    if (size & (size-1)) throw InvalidSize;
+
+    if (m_method == -1) {
+        m_method = 1;
+    }
+
+    switch (m_method) {
+
+    case 0:
+        d = new D_Cross(size);
+        break;
+
+    case 1:
+//        std::cerr << "FFT::FFT(" << size << "): using FFTW3 implementation"
+//                  << std::endl;
+        d = new D_FFTW(size);
+        break;
+
+    default:
+        std::cerr << "FFT::FFT(" << size << "): WARNING: using slow built-in implementation"
+                  << std::endl;
+        d = new D_Cross(size);
+        break;
+    }
+}
+
+FFT::~FFT()
+{
+    delete d;
+}
+
+void
+FFT::forward(double *realIn, double *realOut, double *imagOut)
+{
+    d->forward(realIn, realOut, imagOut);
+}
+
+void
+FFT::forwardPolar(double *realIn, double *magOut, double *phaseOut)
+{
+    d->forwardPolar(realIn, magOut, phaseOut);
+}
+
+void
+FFT::forwardMagnitude(double *realIn, double *magOut)
+{
+    d->forwardMagnitude(realIn, magOut);
+}
+
+void
+FFT::forward(float *realIn, float *realOut, float *imagOut)
+{
+    d->forward(realIn, realOut, imagOut);
+}
+
+void
+FFT::forwardPolar(float *realIn, float *magOut, float *phaseOut)
+{
+    d->forwardPolar(realIn, magOut, phaseOut);
+}
+
+void
+FFT::forwardMagnitude(float *realIn, float *magOut)
+{
+    d->forwardMagnitude(realIn, magOut);
+}
+
+void
+FFT::inverse(double *realIn, double *imagIn, double *realOut)
+{
+    d->inverse(realIn, imagIn, realOut);
+}
+
+void
+FFT::inversePolar(double *magIn, double *phaseIn, double *realOut)
+{
+    d->inversePolar(magIn, phaseIn, realOut);
+}
+
+void
+FFT::inverse(float *realIn, float *imagIn, float *realOut)
+{
+    d->inverse(realIn, imagIn, realOut);
+}
+
+void
+FFT::inversePolar(float *magIn, float *phaseIn, float *realOut)
+{
+    d->inversePolar(magIn, phaseIn, realOut);
+}
+
+void
+FFT::initFloat() 
+{
+    d->initFloat();
+}
+
+void
+FFT::initDouble() 
+{
+    d->initDouble();
+}
+
+float *
+FFT::getFloatTimeBuffer()
+{
+    return d->getFloatTimeBuffer();
+}
+
+double *
+FFT::getDoubleTimeBuffer()
+{
+    return d->getDoubleTimeBuffer();
+}
+
+
+void
+FFT::tune()
+{
+}
+
+
+}

libs/rubberband/1.0/src/FFT.h

@@ -0,0 +1,76 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_FFT_H_
+#define _RUBBERBAND_FFT_H_
+
+namespace RubberBand {
+
+class FFTImpl;
+
+/**
+ * Provide the basic FFT computations we need, using one of a set of
+ * candidate FFT implementations (depending on compile flags).
+ *
+ * Implements real->complex FFTs of power-of-two sizes only.  Note
+ * that only the first half of the output signal is returned (the
+ * complex conjugates half is omitted), so the "complex" arrays need
+ * room for size/2+1 elements.
+ *
+ * Not thread safe: use a separate instance per thread.
+ */
+
+class FFT
+{
+public:
+    enum Exception { InvalidSize };
+
+    FFT(unsigned int size); // may throw InvalidSize
+    ~FFT();
+
+    void forward(double *realIn, double *realOut, double *imagOut);
+    void forwardPolar(double *realIn, double *magOut, double *phaseOut);
+    void forwardMagnitude(double *realIn, double *magOut);
+
+    void forward(float *realIn, float *realOut, float *imagOut);
+    void forwardPolar(float *realIn, float *magOut, float *phaseOut);
+    void forwardMagnitude(float *realIn, float *magOut);
+
+    void inverse(double *realIn, double *imagIn, double *realOut);
+    void inversePolar(double *magIn, double *phaseIn, double *realOut);
+
+    void inverse(float *realIn, float *imagIn, float *realOut);
+    void inversePolar(float *magIn, float *phaseIn, float *realOut);
+
+    // Calling one or both of these is optional -- if neither is
+    // called, the first call to a forward or inverse method will call
+    // init().  You only need call these if you don't want to risk
+    // expensive allocations etc happening in forward or inverse.
+    void initFloat();
+    void initDouble();
+
+    float *getFloatTimeBuffer();
+    double *getDoubleTimeBuffer();
+
+    static void tune();
+
+protected:
+    FFTImpl *d;
+    static int m_method;
+};
+
+}
+
+#endif
+

libs/rubberband/1.0/src/HighFrequencyAudioCurve.cpp

@@ -0,0 +1,53 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "HighFrequencyAudioCurve.h"
+
+namespace RubberBand
+{
+
+HighFrequencyAudioCurve::HighFrequencyAudioCurve(size_t sampleRate, size_t windowSize) :
+    AudioCurve(sampleRate, windowSize)
+{
+}
+
+HighFrequencyAudioCurve::~HighFrequencyAudioCurve()
+{
+}
+
+void
+HighFrequencyAudioCurve::reset()
+{
+}
+
+void
+HighFrequencyAudioCurve::setWindowSize(size_t newSize)
+{
+    m_windowSize = newSize;
+}
+
+float
+HighFrequencyAudioCurve::process(float *mag, size_t increment)
+{
+    float result = 0.0;
+
+    for (size_t n = 0; n <= m_windowSize / 2; ++n) {
+        result += mag[n] * n;
+    }
+
+    return result;
+}
+
+}
+

libs/rubberband/1.0/src/HighFrequencyAudioCurve.h

@@ -0,0 +1,39 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _HIGHFREQUENCY_AUDIO_CURVE_H_
+#define _HIGHFREQUENCY_AUDIO_CURVE_H_
+
+#include "AudioCurve.h"
+#include "Window.h"
+
+namespace RubberBand
+{
+
+class HighFrequencyAudioCurve : public AudioCurve
+{
+public:
+    HighFrequencyAudioCurve(size_t sampleRate, size_t windowSize);
+
+    virtual ~HighFrequencyAudioCurve();
+
+    virtual void setWindowSize(size_t newSize);
+
+    virtual float process(float *mag, size_t increment);
+    virtual void reset();
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/PercussiveAudioCurve.cpp

@@ -0,0 +1,78 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "PercussiveAudioCurve.h"
+
+#include <cmath>
+
+namespace RubberBand
+{
+
+PercussiveAudioCurve::PercussiveAudioCurve(size_t sampleRate, size_t windowSize) :
+    AudioCurve(sampleRate, windowSize)
+{
+    m_prevMag = new double[m_windowSize/2 + 1];
+
+    for (size_t i = 0; i <= m_windowSize/2; ++i) {
+        m_prevMag[i] = 0.f;
+    }
+}
+
+PercussiveAudioCurve::~PercussiveAudioCurve()
+{
+    delete[] m_prevMag;
+}
+
+void
+PercussiveAudioCurve::reset()
+{
+    for (size_t i = 0; i <= m_windowSize/2; ++i) {
+        m_prevMag[i] = 0;
+    }
+}
+
+void
+PercussiveAudioCurve::setWindowSize(size_t newSize)
+{
+    delete[] m_prevMag;
+    m_windowSize = newSize;
+    
+    m_prevMag = new double[m_windowSize/2 + 1];
+
+    reset();
+}
+
+float
+PercussiveAudioCurve::process(float *mag, size_t increment)
+{
+    static float threshold = pow(10, 0.3);
+    static float zeroThresh = pow(10, -16);
+
+    size_t count = 0;
+    size_t nonZeroCount = 0;
+
+    for (size_t n = 1; n <= m_windowSize / 2; ++n) {
+	float sqrmag = mag[n] * mag[n];
+        bool above = ((sqrmag / m_prevMag[n]) >= threshold);
+        if (above) ++count;
+        if (sqrmag > zeroThresh) ++nonZeroCount;
+	m_prevMag[n] = sqrmag;
+    }
+
+    if (nonZeroCount == 0) return 0;
+    else return float(count) / float(nonZeroCount);
+}
+
+}
+

libs/rubberband/1.0/src/PercussiveAudioCurve.h

@@ -0,0 +1,41 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _PERCUSSIVE_AUDIO_CURVE_H_
+#define _PERCUSSIVE_AUDIO_CURVE_H_
+
+#include "AudioCurve.h"
+
+namespace RubberBand
+{
+
+class PercussiveAudioCurve : public AudioCurve
+{
+public:
+    PercussiveAudioCurve(size_t sampleRate, size_t windowSize);
+
+    virtual ~PercussiveAudioCurve();
+
+    virtual void setWindowSize(size_t newSize);
+
+    virtual float process(float *mag, size_t increment);
+    virtual void reset();
+
+protected:
+    double *m_prevMag;
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/Resampler.cpp

@@ -0,0 +1,169 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "Resampler.h"
+
+#include <cstdlib>
+#include <cmath>
+
+#include <iostream>
+
+
+#include <samplerate.h>
+
+namespace RubberBand {
+
+class Resampler::D
+{
+public:
+    D(Quality quality, size_t channels, size_t maxBufferSize);
+    ~D();
+
+    size_t resample(float **in, float **out,
+                    size_t incount, float ratio, bool final);
+
+    void reset();
+
+protected:
+    SRC_STATE *m_src;
+    float *m_iin;
+    float *m_iout;
+    size_t m_channels;
+    size_t m_iinsize;
+    size_t m_ioutsize;
+};
+
+Resampler::D::D(Quality quality, size_t channels, size_t maxBufferSize) :
+    m_src(0),
+    m_iin(0),
+    m_iout(0),
+    m_channels(channels),
+    m_iinsize(0),
+    m_ioutsize(0)
+{
+//    std::cerr << "Resampler::Resampler: using libsamplerate implementation"
+//              << std::endl;
+
+    int err = 0;
+    m_src = src_new(quality == Best ? SRC_SINC_BEST_QUALITY :
+                    quality == Fastest ? SRC_LINEAR :
+                    SRC_SINC_FASTEST,
+                    channels, &err);
+
+    //!!! check err, throw
+
+    if (maxBufferSize > 0 && m_channels > 1) {
+        //!!! alignment?
+        m_iinsize = maxBufferSize * m_channels;
+        m_ioutsize = maxBufferSize * m_channels * 2;
+        m_iin = (float *)malloc(m_iinsize * sizeof(float));
+        m_iout = (float *)malloc(m_ioutsize * sizeof(float));
+    }
+}
+
+Resampler::D::~D()
+{
+    src_delete(m_src);
+    if (m_iinsize > 0) {
+        free(m_iin);
+    }
+    if (m_ioutsize > 0) {
+        free(m_iout);
+    }
+}
+
+size_t
+Resampler::D::resample(float **in, float **out, size_t incount, float ratio,
+                       bool final)
+{
+    SRC_DATA data;
+
+    size_t outcount = lrintf(ceilf(incount * ratio));
+
+    if (m_channels == 1) {
+        data.data_in = *in;
+        data.data_out = *out;
+    } else {
+        if (incount * m_channels > m_iinsize) {
+            m_iinsize = incount * m_channels;
+            m_iin = (float *)realloc(m_iin, m_iinsize * sizeof(float));
+        }
+        if (outcount * m_channels > m_ioutsize) {
+            m_ioutsize = outcount * m_channels;
+            m_iout = (float *)realloc(m_iout, m_ioutsize * sizeof(float));
+        }
+        for (size_t i = 0; i < incount; ++i) {
+            for (size_t c = 0; c < m_channels; ++c) {
+                m_iin[i * m_channels + c] = in[c][i];
+            }
+        }
+        data.data_in = m_iin;
+        data.data_out = m_iout;
+    }
+
+    data.input_frames = incount;
+    data.output_frames = outcount;
+    data.src_ratio = ratio;
+    data.end_of_input = (final ? 1 : 0);
+
+    int err = src_process(m_src, &data);
+
+    //!!! check err, respond appropriately
+
+    if (m_channels > 1) {
+        for (int i = 0; i < data.output_frames_gen; ++i) {
+            for (size_t c = 0; c < m_channels; ++c) {
+                out[c][i] = m_iout[i * m_channels + c];
+            }
+        }
+    }
+
+    return data.output_frames_gen;
+}
+
+void
+Resampler::D::reset()
+{
+    src_reset(m_src);
+}
+
+} // end namespace
+
+
+namespace RubberBand {
+
+Resampler::Resampler(Quality quality, size_t channels, size_t maxBufferSize)
+{
+    m_d = new D(quality, channels, maxBufferSize);
+}
+
+Resampler::~Resampler()
+{
+    delete m_d;
+}
+
+size_t 
+Resampler::resample(float **in, float **out,
+                    size_t incount, float ratio, bool final)
+{
+    return m_d->resample(in, out, incount, ratio, final);
+}
+
+void
+Resampler::reset()
+{
+    m_d->reset();
+}
+
+}

libs/rubberband/1.0/src/Resampler.h

@@ -0,0 +1,48 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_RESAMPLER_H_
+#define _RUBBERBAND_RESAMPLER_H_
+
+#include <sys/types.h>
+
+namespace RubberBand {
+
+class Resampler
+{
+public:
+    enum Quality { Best, FastestTolerable, Fastest };
+
+    /**
+     * Construct a resampler with the given quality level and channel
+     * count.  maxBufferSize gives a bound on the maximum incount size
+     * that may be passed to the resample function before the
+     * resampler needs to reallocate its internal buffers.
+     */
+    Resampler(Quality quality, size_t channels, size_t maxBufferSize = 0);
+    ~Resampler();
+
+    size_t resample(float **in, float **out,
+                    size_t incount, float ratio, bool final = false);
+
+    void reset();
+
+protected:
+    class D;
+    D *m_d;
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/RingBuffer.h

@@ -0,0 +1,636 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_RINGBUFFER_H_
+#define _RUBBERBAND_RINGBUFFER_H_
+
+#include <sys/types.h>
+
+#ifndef _WIN32
+#include <sys/mman.h>
+#endif
+
+#include "Scavenger.h"
+
+//#define DEBUG_RINGBUFFER 1
+
+#ifdef _WIN32
+#define MLOCK(a,b) 1
+#define MUNLOCK(a,b) 1
+#else
+#define MLOCK(a,b) ::mlock(a,b)
+#define MUNLOCK(a,b) ::munlock(a,b)
+#endif
+
+#ifdef DEBUG_RINGBUFFER
+#include <iostream>
+#endif
+
+namespace RubberBand {
+
+/**
+ * RingBuffer implements a lock-free ring buffer for one writer and N
+ * readers, that is to be used to store a sample type T.
+ */
+
+template <typename T, int N = 1>
+class RingBuffer
+{
+public:
+    /**
+     * Create a ring buffer with room to write n samples.
+     *
+     * Note that the internal storage size will actually be n+1
+     * samples, as one element is unavailable for administrative
+     * reasons.  Since the ring buffer performs best if its size is a
+     * power of two, this means n should ideally be some power of two
+     * minus one.
+     */
+    RingBuffer(size_t n);
+
+    virtual ~RingBuffer();
+
+    /**
+     * Return the total capacity of the ring buffer in samples.
+     * (This is the argument n passed to the constructor.)
+     */
+    size_t getSize() const;
+
+    /**
+     * Resize the ring buffer.  This also empties it; use resized()
+     * below if you do not want this to happen.  Actually swaps in a
+     * new, larger buffer; the old buffer is scavenged after a seemly
+     * delay.  Should be called from the write thread.
+     */
+    void resize(size_t newSize);
+
+    /**
+     * Return a new ring buffer (allocated with "new" -- called must
+     * delete when no longer needed) of the given size, containing the
+     * same data as this one.  If another thread reads from or writes
+     * to this buffer during the call, the results may be incomplete
+     * or inconsistent.  If this buffer's data will not fit in the new
+     * size, the contents are undefined.
+     */
+    RingBuffer<T, N> *resized(size_t newSize, int R = 0) const;
+
+    /**
+     * Lock the ring buffer into physical memory.  Returns true
+     * for success.
+     */
+    bool mlock();
+
+    /**
+     * Reset read and write pointers, thus emptying the buffer.
+     * Should be called from the write thread.
+     */
+    void reset();
+
+    /**
+     * Return the amount of data available for reading by reader R, in
+     * samples.
+     */
+    size_t getReadSpace(int R = 0) const;
+
+    /**
+     * Return the amount of space available for writing, in samples.
+     */
+    size_t getWriteSpace() const;
+
+    /**
+     * Read n samples from the buffer, for reader R.  If fewer than n
+     * are available, the remainder will be zeroed out.  Returns the
+     * number of samples actually read.
+     */
+    size_t read(T *destination, size_t n, int R = 0);
+
+    /**
+     * Read n samples from the buffer, for reader R, adding them to
+     * the destination.  If fewer than n are available, the remainder
+     * will be left alone.  Returns the number of samples actually
+     * read.
+     */
+    size_t readAdding(T *destination, size_t n, int R = 0);
+
+    /**
+     * Read one sample from the buffer, for reader R.  If no sample is
+     * available, this will silently return zero.  Calling this
+     * repeatedly is obviously slower than calling read once, but it
+     * may be good enough if you don't want to allocate a buffer to
+     * read into.
+     */
+    T readOne(int R = 0);
+
+    /**
+     * Read n samples from the buffer, if available, for reader R,
+     * without advancing the read pointer -- i.e. a subsequent read()
+     * or skip() will be necessary to empty the buffer.  If fewer than
+     * n are available, the remainder will be zeroed out.  Returns the
+     * number of samples actually read.
+     */
+    size_t peek(T *destination, size_t n, int R = 0) const;
+
+    /**
+     * Read one sample from the buffer, if available, without
+     * advancing the read pointer -- i.e. a subsequent read() or
+     * skip() will be necessary to empty the buffer.  Returns zero if
+     * no sample was available.
+     */
+    T peekOne(int R = 0) const;
+
+    /**
+     * Pretend to read n samples from the buffer, for reader R,
+     * without actually returning them (i.e. discard the next n
+     * samples).  Returns the number of samples actually available for
+     * discarding.
+     */
+    size_t skip(size_t n, int R = 0);
+
+    /**
+     * Write n samples to the buffer.  If insufficient space is
+     * available, not all samples may actually be written.  Returns
+     * the number of samples actually written.
+     */
+    size_t write(const T *source, size_t n);
+
+    /**
+     * Write n zero-value samples to the buffer.  If insufficient
+     * space is available, not all zeros may actually be written.
+     * Returns the number of zeroes actually written.
+     */
+    size_t zero(size_t n);
+
+protected:
+    T               *m_buffer;
+    volatile size_t  m_writer;
+    volatile size_t  m_readers[N];
+    size_t           m_size;
+    bool             m_mlocked;
+
+    static Scavenger<ScavengerArrayWrapper<T> > m_scavenger;
+
+private:
+    RingBuffer(const RingBuffer &); // not provided
+    RingBuffer &operator=(const RingBuffer &); // not provided
+};
+
+template <typename T, int N>
+Scavenger<ScavengerArrayWrapper<T> > RingBuffer<T, N>::m_scavenger;
+
+template <typename T, int N>
+RingBuffer<T, N>::RingBuffer(size_t n) :
+    m_buffer(new T[n + 1]),
+    m_writer(0),
+    m_size(n + 1),
+    m_mlocked(false)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::RingBuffer(" << n << ")" << std::endl;
+#endif
+
+    for (int i = 0; i < N; ++i) m_readers[i] = 0;
+
+    m_scavenger.scavenge();
+}
+
+template <typename T, int N>
+RingBuffer<T, N>::~RingBuffer()
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::~RingBuffer" << std::endl;
+#endif
+
+    if (m_mlocked) {
+	MUNLOCK((void *)m_buffer, m_size * sizeof(T));
+    }
+    delete[] m_buffer;
+
+    m_scavenger.scavenge();
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::getSize() const
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::getSize(): " << m_size-1 << std::endl;
+#endif
+
+    return m_size - 1;
+}
+
+template <typename T, int N>
+void
+RingBuffer<T, N>::resize(size_t newSize)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::resize(" << newSize << ")" << std::endl;
+#endif
+
+    m_scavenger.scavenge();
+
+    if (m_mlocked) {
+	MUNLOCK((void *)m_buffer, m_size * sizeof(T));
+    }
+
+    m_scavenger.claim(new ScavengerArrayWrapper<T>(m_buffer));
+
+    reset();
+    m_buffer = new T[newSize + 1];
+    m_size = newSize + 1;
+
+    if (m_mlocked) {
+	if (MLOCK((void *)m_buffer, m_size * sizeof(T))) {
+	    m_mlocked = false;
+	}
+    }
+}
+
+template <typename T, int N>
+RingBuffer<T, N> *
+RingBuffer<T, N>::resized(size_t newSize, int R) const
+{
+    RingBuffer<T, N> *newBuffer = new RingBuffer<T, N>(newSize);
+
+    size_t w = m_writer;
+    size_t r = m_readers[R];
+
+    while (r != w) {
+        T value = m_buffer[r];
+        newBuffer->write(&value, 1);
+        if (++r == m_size) r = 0;
+    }
+
+    return newBuffer;
+}
+
+template <typename T, int N>
+bool
+RingBuffer<T, N>::mlock()
+{
+    if (MLOCK((void *)m_buffer, m_size * sizeof(T))) return false;
+    m_mlocked = true;
+    return true;
+}
+
+template <typename T, int N>
+void
+RingBuffer<T, N>::reset()
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::reset" << std::endl;
+#endif
+
+    m_writer = 0;
+    for (int i = 0; i < N; ++i) m_readers[i] = 0;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::getReadSpace(int R) const
+{
+    size_t writer = m_writer;
+    size_t reader = m_readers[R];
+    size_t space;
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::getReadSpace(" << R << "): reader " << reader << ", writer " << writer << std::endl;
+#endif
+
+    if (writer > reader) space = writer - reader;
+    else if (writer < reader) space = (writer + m_size) - reader;
+    else space = 0;
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::getReadSpace(" << R << "): " << space << std::endl;
+#endif
+
+    return space;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::getWriteSpace() const
+{
+    size_t space = 0;
+    for (int i = 0; i < N; ++i) {
+        size_t writer = m_writer;
+        size_t reader = m_readers[i];
+	size_t here = (reader + m_size - writer - 1);
+        if (here >= m_size) here -= m_size;
+	if (i == 0 || here < space) space = here;
+    }
+
+#ifdef DEBUG_RINGBUFFER
+    size_t rs(getReadSpace()), rp(m_readers[0]);
+
+    std::cerr << "RingBuffer: write space " << space << ", read space "
+	      << rs << ", total " << (space + rs) << ", m_size " << m_size << std::endl;
+    std::cerr << "RingBuffer: reader " << rp << ", writer " << m_writer << std::endl;
+#endif
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::getWriteSpace(): " << space << std::endl;
+#endif
+
+    return space;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::read(T *destination, size_t n, int R)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::read(dest, " << n << ", " << R << ")" << std::endl;
+#endif
+
+    size_t available = getReadSpace(R);
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only " << available << " samples available"
+		  << std::endl;
+#endif
+        for (size_t i = available; i < n; ++i) {
+            destination[i] = 0;
+        }
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t reader = m_readers[R];
+    size_t here = m_size - reader;
+
+    if (here >= n) {
+        for (size_t i = 0; i < n; ++i) {
+            destination[i] = (m_buffer + reader)[i];
+        }
+    } else {
+        for (size_t i = 0; i < here; ++i) {
+            destination[i] = (m_buffer + reader)[i];
+        }
+        for (size_t i = 0; i < (n - here); ++i) {
+            destination[i + here] = m_buffer[i];
+        }
+    }
+
+    reader += n;
+    while (reader >= m_size) reader -= m_size;
+    m_readers[R] = reader;
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::read: read " << n << ", reader now " << m_readers[R] << std::endl;
+#endif
+
+    return n;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::readAdding(T *destination, size_t n, int R)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::readAdding(dest, " << n << ", " << R << ")" << std::endl;
+#endif
+
+    size_t available = getReadSpace(R);
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only " << available << " samples available"
+		  << std::endl;
+#endif
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t reader = m_readers[R];
+    size_t here = m_size - reader;
+
+    if (here >= n) {
+	for (size_t i = 0; i < n; ++i) {
+	    destination[i] += (m_buffer + reader)[i];
+	}
+    } else {
+	for (size_t i = 0; i < here; ++i) {
+	    destination[i] += (m_buffer + reader)[i];
+	}
+	for (size_t i = 0; i < (n - here); ++i) {
+	    destination[i + here] += m_buffer[i];
+	}
+    }
+
+    reader += n;
+    while (reader >= m_size) reader -= m_size;
+    m_readers[R] = reader;
+    return n;
+}
+
+template <typename T, int N>
+T
+RingBuffer<T, N>::readOne(int R)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::readOne(" << R << ")" << std::endl;
+#endif
+
+    if (m_writer == m_readers[R]) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: No sample available"
+		  << std::endl;
+#endif
+	return 0;
+    }
+    size_t reader = m_readers[R];
+    T value = m_buffer[reader];
+    if (++reader == m_size) reader = 0;
+    m_readers[R] = reader;
+    return value;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::peek(T *destination, size_t n, int R) const
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::peek(dest, " << n << ", " << R << ")" << std::endl;
+#endif
+
+    size_t available = getReadSpace(R);
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only " << available << " samples available"
+		  << std::endl;
+#endif
+	memset(destination + available, 0, (n - available) * sizeof(T));
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t reader = m_readers[R];
+    size_t here = m_size - reader;
+
+    if (here >= n) {
+        for (size_t i = 0; i < n; ++i) {
+            destination[i] = (m_buffer + reader)[i];
+        }
+    } else {
+        for (size_t i = 0; i < here; ++i) {
+            destination[i] = (m_buffer + reader)[i];
+        }
+        for (size_t i = 0; i < (n - here); ++i) {
+            destination[i + here] = m_buffer[i];
+        }
+    }
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::peek: read " << n << std::endl;
+#endif
+
+    return n;
+}
+
+template <typename T, int N>
+T
+RingBuffer<T, N>::peekOne(int R) const
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::peek(" << R << ")" << std::endl;
+#endif
+
+    if (m_writer == m_readers[R]) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: No sample available"
+		  << std::endl;
+#endif
+	return 0;
+    }
+    T value = m_buffer[m_readers[R]];
+    return value;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::skip(size_t n, int R)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::skip(" << n << ", " << R << ")" << std::endl;
+#endif
+
+    size_t available = getReadSpace(R);
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only " << available << " samples available"
+		  << std::endl;
+#endif
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t reader = m_readers[R];
+    reader += n;
+    while (reader >= m_size) reader -= m_size;
+    m_readers[R] = reader;
+    return n;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::write(const T *source, size_t n)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::write(" << n << ")" << std::endl;
+#endif
+
+    size_t available = getWriteSpace();
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only room for " << available << " samples"
+		  << std::endl;
+#endif
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t writer = m_writer;
+    size_t here = m_size - writer;
+    if (here >= n) {
+        for (size_t i = 0; i < n; ++i) {
+            (m_buffer + writer)[i] = source[i];
+        }
+    } else {
+        for (size_t i = 0; i < here; ++i) {
+            (m_buffer + writer)[i] = source[i];
+        }
+        for (size_t i = 0; i < (n - here); ++i) {
+            m_buffer[i] = (source + here)[i];
+        }
+    }
+
+    writer += n;
+    while (writer >= m_size) writer -= m_size;
+    m_writer = writer;
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::write: wrote " << n << ", writer now " << m_writer << std::endl;
+#endif
+
+    return n;
+}
+
+template <typename T, int N>
+size_t
+RingBuffer<T, N>::zero(size_t n)
+{
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "RingBuffer<T," << N << ">[" << this << "]::zero(" << n << ")" << std::endl;
+#endif
+
+    size_t available = getWriteSpace();
+    if (n > available) {
+#ifdef DEBUG_RINGBUFFER
+	std::cerr << "WARNING: Only room for " << available << " samples"
+		  << std::endl;
+#endif
+	n = available;
+    }
+    if (n == 0) return n;
+
+    size_t writer = m_writer;
+    size_t here = m_size - writer;
+    if (here >= n) {
+        for (size_t i = 0; i < n; ++i) {
+            (m_buffer + writer)[i] = 0;
+        }
+    } else {
+        for (size_t i = 0; i < here; ++i) {
+            (m_buffer + writer)[i] = 0;
+        }
+        for (size_t i = 0; i < (n - here); ++i) {
+            m_buffer[i] = 0;
+        }
+    }
+
+    writer += n;
+    while (writer >= m_size) writer -= m_size;
+    m_writer = writer;
+
+#ifdef DEBUG_RINGBUFFER
+    std::cerr << "writer -> " << m_writer << std::endl;
+#endif
+
+    return n;
+}
+
+}
+
+#endif // _RINGBUFFER_H_

libs/rubberband/1.0/src/RubberBandStretcher.cpp

@@ -0,0 +1,189 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "StretcherImpl.h"
+
+namespace RubberBand {
+
+
+RubberBandStretcher::RubberBandStretcher(size_t sampleRate,
+                                         size_t channels,
+                                         Options options,
+                                         double initialTimeRatio,
+                                         double initialPitchScale) :
+    TimeStretcher(sampleRate, channels),
+    m_d(new Impl(this, sampleRate, channels, options,
+                 initialTimeRatio, initialPitchScale))
+{
+}
+
+RubberBandStretcher::~RubberBandStretcher()
+{
+    delete m_d;
+}
+
+void
+RubberBandStretcher::reset()
+{
+    m_d->reset();
+}
+
+void
+RubberBandStretcher::setTimeRatio(double ratio)
+{
+    m_d->setTimeRatio(ratio);
+}
+
+void
+RubberBandStretcher::setPitchScale(double scale)
+{
+    m_d->setPitchScale(scale);
+}
+
+double
+RubberBandStretcher::getTimeRatio() const
+{
+    return m_d->getTimeRatio();
+}
+
+double
+RubberBandStretcher::getPitchScale() const
+{
+    return m_d->getPitchScale();
+}
+
+size_t
+RubberBandStretcher::getLatency() const
+{
+    return m_d->getLatency();
+}
+
+void
+RubberBandStretcher::setTransientsOption(Options options) 
+{
+    m_d->setTransientsOption(options);
+}
+
+void
+RubberBandStretcher::setPhaseOption(Options options) 
+{
+    m_d->setPhaseOption(options);
+}
+
+void
+RubberBandStretcher::setExpectedInputDuration(size_t samples) 
+{
+    m_d->setExpectedInputDuration(samples);
+}
+
+void
+RubberBandStretcher::setMaxProcessSize(size_t samples)
+{
+    m_d->setMaxProcessSize(samples);
+}
+
+size_t
+RubberBandStretcher::getSamplesRequired() const
+{
+    return m_d->getSamplesRequired();
+}
+
+void
+RubberBandStretcher::study(const float *const *input, size_t samples,
+                           bool final)
+{
+    m_d->study(input, samples, final);
+}
+
+void
+RubberBandStretcher::process(const float *const *input, size_t samples,
+                             bool final)
+{
+    m_d->process(input, samples, final);
+}
+
+int
+RubberBandStretcher::available() const
+{
+    return m_d->available();
+}
+
+size_t
+RubberBandStretcher::retrieve(float *const *output, size_t samples) const
+{
+    return m_d->retrieve(output, samples);
+}
+
+float
+RubberBandStretcher::getFrequencyCutoff(int n) const
+{
+    return m_d->getFrequencyCutoff(n);
+}
+
+void
+RubberBandStretcher::setFrequencyCutoff(int n, float f) 
+{
+    m_d->setFrequencyCutoff(n, f);
+}
+
+size_t
+RubberBandStretcher::getInputIncrement() const
+{
+    return m_d->getInputIncrement();
+}
+
+std::vector<int>
+RubberBandStretcher::getOutputIncrements() const
+{
+    return m_d->getOutputIncrements();
+}
+
+std::vector<float>
+RubberBandStretcher::getPhaseResetCurve() const
+{
+    return m_d->getPhaseResetCurve();
+}
+
+std::vector<int>
+RubberBandStretcher::getExactTimePoints() const
+{
+    return m_d->getExactTimePoints();
+}
+
+size_t
+RubberBandStretcher::getChannelCount() const
+{
+    return m_d->getChannelCount();
+}
+
+void
+RubberBandStretcher::calculateStretch()
+{
+    m_d->calculateStretch();
+}
+
+void
+RubberBandStretcher::setDebugLevel(int level)
+{
+    m_d->setDebugLevel(level);
+}
+
+void
+RubberBandStretcher::setDefaultDebugLevel(int level)
+{
+    Impl::setDefaultDebugLevel(level);
+}
+
+}
+

libs/rubberband/1.0/src/Scavenger.h

@@ -0,0 +1,199 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_SCAVENGER_H_
+#define _RUBBERBAND_SCAVENGER_H_
+
+#include <vector>
+#include <list>
+#include <sys/time.h>
+#include <iostream>
+
+#include "Thread.h"
+#include "sysutils.h"
+
+namespace RubberBand {
+
+/**
+ * A very simple class that facilitates running things like plugins
+ * without locking, by collecting unwanted objects and deleting them
+ * after a delay so as to be sure nobody's in the middle of using
+ * them.  Requires scavenge() to be called regularly from a non-RT
+ * thread.
+ *
+ * This is currently not at all suitable for large numbers of objects
+ * -- it's just a quick hack for use with things like plugins.
+ */
+
+template <typename T>
+class Scavenger
+{
+public:
+    Scavenger(int sec = 2, int defaultObjectListSize = 200);
+    ~Scavenger();
+
+    /**
+     * Call from an RT thread etc., to pass ownership of t to us.
+     * Only one thread should be calling this on any given scavenger.
+     */
+    void claim(T *t);
+
+    /**
+     * Call from a non-RT thread.
+     * Only one thread should be calling this on any given scavenger.
+     */
+    void scavenge(bool clearNow = false);
+
+protected:
+    typedef std::pair<T *, int> ObjectTimePair;
+    typedef std::vector<ObjectTimePair> ObjectTimeList;
+    ObjectTimeList m_objects;
+    int m_sec;
+
+    typedef std::list<T *> ObjectList;
+    ObjectList m_excess;
+    int m_lastExcess;
+    Mutex m_excessMutex;
+    void pushExcess(T *);
+    void clearExcess(int);
+
+    unsigned int m_claimed;
+    unsigned int m_scavenged;
+};
+
+/**
+ * A wrapper to permit arrays to be scavenged.
+ */
+
+template <typename T>
+class ScavengerArrayWrapper
+{
+public:
+    ScavengerArrayWrapper(T *array) : m_array(array) { }
+    ~ScavengerArrayWrapper() { delete[] m_array; }
+
+private:
+    T *m_array;
+};
+
+
+template <typename T>
+Scavenger<T>::Scavenger(int sec, int defaultObjectListSize) :
+    m_objects(ObjectTimeList(defaultObjectListSize)),
+    m_sec(sec),
+    m_claimed(0),
+    m_scavenged(0)
+{
+}
+
+template <typename T>
+Scavenger<T>::~Scavenger()
+{
+    if (m_scavenged < m_claimed) {
+	for (size_t i = 0; i < m_objects.size(); ++i) {
+	    ObjectTimePair &pair = m_objects[i];
+	    if (pair.first != 0) {
+		T *ot = pair.first;
+		pair.first = 0;
+		delete ot;
+		++m_scavenged;
+	    }
+	}
+    }
+
+    clearExcess(0);
+}
+
+template <typename T>
+void
+Scavenger<T>::claim(T *t)
+{
+//    std::cerr << "Scavenger::claim(" << t << ")" << std::endl;
+
+    struct timeval tv;
+    (void)gettimeofday(&tv, 0);
+    int sec = tv.tv_sec;
+
+    for (size_t i = 0; i < m_objects.size(); ++i) {
+	ObjectTimePair &pair = m_objects[i];
+	if (pair.first == 0) {
+	    pair.second = sec;
+	    pair.first = t;
+	    ++m_claimed;
+	    return;
+	}
+    }
+
+    std::cerr << "WARNING: Scavenger::claim(" << t << "): run out of slots, "
+	      << "using non-RT-safe method" << std::endl;
+    pushExcess(t);
+}
+
+template <typename T>
+void
+Scavenger<T>::scavenge(bool clearNow)
+{
+//    std::cerr << "Scavenger::scavenge: scavenged " << m_scavenged << ", claimed " << m_claimed << std::endl;
+
+    if (m_scavenged >= m_claimed) return;
+    
+    struct timeval tv;
+    (void)gettimeofday(&tv, 0);
+    int sec = tv.tv_sec;
+
+    for (size_t i = 0; i < m_objects.size(); ++i) {
+	ObjectTimePair &pair = m_objects[i];
+	if (clearNow ||
+	    (pair.first != 0 && pair.second + m_sec < sec)) {
+	    T *ot = pair.first;
+	    pair.first = 0;
+	    delete ot;
+	    ++m_scavenged;
+	}
+    }
+
+    if (sec > m_lastExcess + m_sec) {
+        clearExcess(sec);
+    }
+}
+
+template <typename T>
+void
+Scavenger<T>::pushExcess(T *t)
+{
+    m_excessMutex.lock();
+    m_excess.push_back(t);
+    struct timeval tv;
+    (void)gettimeofday(&tv, 0);
+    m_lastExcess = tv.tv_sec;
+    m_excessMutex.unlock();
+}
+
+template <typename T>
+void
+Scavenger<T>::clearExcess(int sec)
+{
+    m_excessMutex.lock();
+    for (typename ObjectList::iterator i = m_excess.begin();
+	 i != m_excess.end(); ++i) {
+	delete *i;
+    }
+    m_excess.clear();
+    m_lastExcess = sec;
+    m_excessMutex.unlock();
+}
+
+}
+
+#endif

libs/rubberband/1.0/src/SpectralDifferenceAudioCurve.cpp

@@ -0,0 +1,64 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "SpectralDifferenceAudioCurve.h"
+
+namespace RubberBand
+{
+
+SpectralDifferenceAudioCurve::SpectralDifferenceAudioCurve(size_t sampleRate, size_t windowSize) :
+    AudioCurve(sampleRate, windowSize)
+{
+    m_prevMag = new double[m_windowSize/2 + 1];
+
+    for (size_t i = 0; i <= m_windowSize/2; ++i) {
+        m_prevMag[i] = 0.f;
+    }
+}
+
+SpectralDifferenceAudioCurve::~SpectralDifferenceAudioCurve()
+{
+    delete[] m_prevMag;
+}
+
+void
+SpectralDifferenceAudioCurve::reset()
+{
+    for (size_t i = 0; i <= m_windowSize/2; ++i) {
+        m_prevMag[i] = 0;
+    }
+}
+
+void
+SpectralDifferenceAudioCurve::setWindowSize(size_t newSize)
+{
+    m_windowSize = newSize;
+}
+
+float
+SpectralDifferenceAudioCurve::process(float *mag, size_t increment)
+{
+    float result = 0.0;
+
+    for (size_t n = 0; n <= m_windowSize / 2; ++n) {
+        result += sqrtf(fabsf((mag[n] * mag[n]) -
+                              (m_prevMag[n] * m_prevMag[n])));
+        m_prevMag[n] = mag[n];
+    }
+
+    return result;
+}
+
+}
+

libs/rubberband/1.0/src/SpectralDifferenceAudioCurve.h

@@ -0,0 +1,42 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _SPECTRALDIFFERENCE_AUDIO_CURVE_H_
+#define _SPECTRALDIFFERENCE_AUDIO_CURVE_H_
+
+#include "AudioCurve.h"
+#include "Window.h"
+
+namespace RubberBand
+{
+
+class SpectralDifferenceAudioCurve : public AudioCurve
+{
+public:
+    SpectralDifferenceAudioCurve(size_t sampleRate, size_t windowSize);
+
+    virtual ~SpectralDifferenceAudioCurve();
+
+    virtual void setWindowSize(size_t newSize);
+
+    virtual float process(float *mag, size_t increment);
+    virtual void reset();
+
+protected:
+    double *m_prevMag;
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/StretchCalculator.cpp

@@ -0,0 +1,790 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "StretchCalculator.h"
+
+#include <math.h>
+#include <iostream>
+#include <deque>
+#include <set>
+#include <cassert>
+
+namespace RubberBand
+{
+	
+StretchCalculator::StretchCalculator(size_t sampleRate,
+                                     size_t inputIncrement,
+                                     bool useHardPeaks) :
+    m_sampleRate(sampleRate),
+    m_increment(inputIncrement),
+    m_prevDf(0),
+    m_divergence(0),
+    m_recovery(0),
+    m_prevRatio(1.0),
+    m_transientAmnesty(0),
+    m_useHardPeaks(useHardPeaks)
+{
+//    std::cerr << "StretchCalculator::StretchCalculator: useHardPeaks = " << useHardPeaks << std::endl;
+}    
+
+StretchCalculator::~StretchCalculator()
+{
+}
+
+std::vector<int>
+StretchCalculator::calculate(double ratio, size_t inputDuration,
+                             const std::vector<float> &phaseResetDf,
+                             const std::vector<float> &stretchDf)
+{
+    assert(phaseResetDf.size() == stretchDf.size());
+    
+    m_lastPeaks = findPeaks(phaseResetDf);
+    std::vector<Peak> &peaks = m_lastPeaks;
+    size_t totalCount = phaseResetDf.size();
+
+    std::vector<int> increments;
+
+    size_t outputDuration = lrint(inputDuration * ratio);
+
+    if (m_debugLevel > 0) {
+        std::cerr << "StretchCalculator::calculate(): inputDuration " << inputDuration << ", ratio " << ratio << ", outputDuration " << outputDuration;
+    }
+
+    outputDuration = lrint((phaseResetDf.size() * m_increment) * ratio);
+
+    if (m_debugLevel > 0) {
+        std::cerr << " (rounded up to " << outputDuration << ")";
+        std::cerr << ", df size " << phaseResetDf.size() << std::endl;
+    }
+
+    std::vector<size_t> fixedAudioChunks;
+    for (size_t i = 0; i < peaks.size(); ++i) {
+        fixedAudioChunks.push_back
+            (lrint((double(peaks[i].chunk) * outputDuration) / totalCount));
+    }
+
+    if (m_debugLevel > 1) {
+        std::cerr << "have " << peaks.size() << " fixed positions" << std::endl;
+    }
+
+    size_t totalInput = 0, totalOutput = 0;
+
+    // For each region between two consecutive time sync points, we
+    // want to take the number of output chunks to be allocated and
+    // the detection function values within the range, and produce a
+    // series of increments that sum to the number of output chunks,
+    // such that each increment is displaced from the input increment
+    // by an amount inversely proportional to the magnitude of the
+    // stretch detection function at that input step.
+
+    size_t regionTotalChunks = 0;
+
+    for (size_t i = 0; i <= peaks.size(); ++i) {
+        
+        size_t regionStart, regionStartChunk, regionEnd, regionEndChunk;
+        bool phaseReset = false;
+
+        if (i == 0) {
+            regionStartChunk = 0;
+            regionStart = 0;
+        } else {
+            regionStartChunk = peaks[i-1].chunk;
+            regionStart = fixedAudioChunks[i-1];
+            phaseReset = peaks[i-1].hard;
+        }
+
+        if (i == peaks.size()) {
+            regionEndChunk = totalCount;
+            regionEnd = outputDuration;
+        } else {
+            regionEndChunk = peaks[i].chunk;
+            regionEnd = fixedAudioChunks[i];
+        }
+        
+        size_t regionDuration = regionEnd - regionStart;
+        regionTotalChunks += regionDuration;
+
+        std::vector<float> dfRegion;
+
+        for (size_t j = regionStartChunk; j != regionEndChunk; ++j) {
+            dfRegion.push_back(stretchDf[j]);
+        }
+
+        if (m_debugLevel > 1) {
+            std::cerr << "distributeRegion from " << regionStartChunk << " to " << regionEndChunk << " (chunks " << regionStart << " to " << regionEnd << ")" << std::endl;
+        }
+
+        dfRegion = smoothDF(dfRegion);
+        
+        std::vector<int> regionIncrements = distributeRegion
+            (dfRegion, regionDuration, ratio, phaseReset);
+
+        size_t totalForRegion = 0;
+
+        for (size_t j = 0; j < regionIncrements.size(); ++j) {
+
+            int incr = regionIncrements[j];
+
+            if (j == 0 && phaseReset) increments.push_back(-incr);
+            else increments.push_back(incr);
+
+            if (incr > 0) totalForRegion += incr;
+            else totalForRegion += -incr;
+
+            totalInput += m_increment;
+        }
+
+        if (totalForRegion != regionDuration) {
+            std::cerr << "*** WARNING: distributeRegion returned wrong duration " << totalForRegion << ", expected " << regionDuration << std::endl;
+        }
+
+        totalOutput += totalForRegion;
+    }
+
+    if (m_debugLevel > 0) {
+        std::cerr << "total input increment = " << totalInput << " (= " << totalInput / m_increment << " chunks), output = " << totalOutput << ", ratio = " << double(totalOutput)/double(totalInput) << ", ideal output " << size_t(ceil(totalInput * ratio)) << std::endl;
+        std::cerr << "(region total = " << regionTotalChunks << ")" << std::endl;
+    }
+
+    return increments;
+}
+
+int
+StretchCalculator::calculateSingle(double ratio,
+                                   size_t inputDurationSoFar,
+                                   float df)
+{
+    bool isTransient = false;
+
+    // We want to ensure, as close as possible, that the phase reset
+    // points appear at _exactly_ the right audio frame numbers.
+
+    // In principle, the threshold depends on chunk size: larger chunk
+    // sizes need higher thresholds.  Since chunk size depends on
+    // ratio, I suppose we could in theory calculate the threshold
+    // from the ratio directly.  For the moment we're happy if it
+    // works well in common situations.
+
+    float transientThreshold = 0.35;
+    if (ratio > 1) transientThreshold = 0.25;
+
+    if (m_useHardPeaks && df > m_prevDf * 1.1 && df > transientThreshold) {
+        isTransient = true;
+    }
+
+    if (m_debugLevel > 2) {
+        std::cerr << "df = " << df << ", prevDf = " << m_prevDf
+                  << ", thresh = " << transientThreshold << std::endl;
+    }
+
+    m_prevDf = df;
+
+    if (isTransient && m_transientAmnesty == 0) {
+        if (m_debugLevel > 1) {
+            std::cerr << "StretchCalculator::calculateSingle: transient found at "
+                      << inputDurationSoFar << std::endl;
+        }
+        m_divergence += m_increment - (m_increment * ratio);
+
+        // as in offline mode, 0.05 sec approx min between transients
+        m_transientAmnesty =
+            lrint(ceil(double(m_sampleRate) / (20 * double(m_increment))));
+
+        m_recovery = m_divergence / ((m_sampleRate / 10.0) / m_increment);
+        return -m_increment;
+    }
+
+    if (m_prevRatio != ratio) {
+        m_recovery = m_divergence / ((m_sampleRate / 10.0) / m_increment);
+        m_prevRatio = ratio;
+    }
+
+    if (m_transientAmnesty > 0) --m_transientAmnesty;
+
+    int incr = lrint(m_increment * ratio - m_recovery);
+    if (m_debugLevel > 2 || (m_debugLevel > 1 && m_divergence != 0)) {
+        std::cerr << "divergence = " << m_divergence << ", recovery = " << m_recovery << ", incr = " << incr << ", ";
+    }
+    if (incr < lrint((m_increment * ratio) / 2)) {
+        incr = lrint((m_increment * ratio) / 2);
+    } else if (incr > lrint(m_increment * ratio * 2)) {
+        incr = lrint(m_increment * ratio * 2);
+    }
+
+    double divdiff = (m_increment * ratio) - incr;
+
+    if (m_debugLevel > 2 || (m_debugLevel > 1 && m_divergence != 0)) {
+        std::cerr << "divdiff = " << divdiff << std::endl;
+    }
+
+    double prevDivergence = m_divergence;
+    m_divergence -= divdiff;
+    if ((prevDivergence < 0 && m_divergence > 0) ||
+        (prevDivergence > 0 && m_divergence < 0)) {
+        m_recovery = m_divergence / ((m_sampleRate / 10.0) / m_increment);
+    }
+
+    return incr;
+}
+
+void
+StretchCalculator::reset()
+{
+    m_prevDf = 0;
+    m_divergence = 0;
+}
+
+std::vector<StretchCalculator::Peak>
+StretchCalculator::findPeaks(const std::vector<float> &rawDf)
+{
+    std::vector<float> df = smoothDF(rawDf);
+
+    // We distinguish between "soft" and "hard" peaks.  A soft peak is
+    // simply the result of peak-picking on the smoothed onset
+    // detection function, and it represents any (strong-ish) onset.
+    // We aim to ensure always that soft peaks are placed at the
+    // correct position in time.  A hard peak is where there is a very
+    // rapid rise in detection function, and it presumably represents
+    // a more broadband, noisy transient.  For these we perform a
+    // phase reset (if in the appropriate mode), and we locate the
+    // reset at the first point where we notice enough of a rapid
+    // rise, rather than necessarily at the peak itself, in order to
+    // preserve the shape of the transient.
+            
+    std::set<size_t> hardPeakCandidates;
+    std::set<size_t> softPeakCandidates;
+
+    if (m_useHardPeaks) {
+
+        // 0.05 sec approx min between hard peaks
+        size_t hardPeakAmnesty = lrint(ceil(double(m_sampleRate) /
+                                            (20 * double(m_increment))));
+        size_t prevHardPeak = 0;
+
+        if (m_debugLevel > 1) {
+            std::cerr << "hardPeakAmnesty = " << hardPeakAmnesty << std::endl;
+        }
+
+        for (size_t i = 1; i + 1 < df.size(); ++i) {
+
+            if (df[i] < 0.1) continue;
+            if (df[i] <= df[i-1] * 1.1) continue;
+            if (df[i] < 0.22) continue;
+
+            if (!hardPeakCandidates.empty() &&
+                i < prevHardPeak + hardPeakAmnesty) {
+                continue;
+            }
+
+            bool hard = (df[i] > 0.4);
+            
+            if (hard && (m_debugLevel > 1)) {
+                std::cerr << "hard peak at " << i << ": " << df[i] 
+                          << " > absolute " << 0.4
+                          << std::endl;
+            }
+
+            if (!hard) {
+                hard = (df[i] > df[i-1] * 1.4);
+
+                if (hard && (m_debugLevel > 1)) {
+                    std::cerr << "hard peak at " << i << ": " << df[i] 
+                              << " > prev " << df[i-1] << " * 1.4"
+                              << std::endl;
+                }
+            }
+
+            if (!hard && i > 1) {
+                hard = (df[i]   > df[i-1] * 1.2 &&
+                        df[i-1] > df[i-2] * 1.2);
+
+                if (hard && (m_debugLevel > 1)) {
+                    std::cerr << "hard peak at " << i << ": " << df[i] 
+                              << " > prev " << df[i-1] << " * 1.2 and "
+                              << df[i-1] << " > prev " << df[i-2] << " * 1.2"
+                              << std::endl;
+                }
+            }
+
+            if (!hard && i > 2) {
+                // have already established that df[i] > df[i-1] * 1.1
+                hard = (df[i] > 0.3 &&
+                        df[i-1] > df[i-2] * 1.1 &&
+                        df[i-2] > df[i-3] * 1.1);
+
+                if (hard && (m_debugLevel > 1)) {
+                    std::cerr << "hard peak at " << i << ": " << df[i] 
+                              << " > prev " << df[i-1] << " * 1.1 and "
+                              << df[i-1] << " > prev " << df[i-2] << " * 1.1 and "
+                              << df[i-2] << " > prev " << df[i-3] << " * 1.1"
+                              << std::endl;
+                }
+            }
+
+            if (!hard) continue;
+
+//            (df[i+1] > df[i] && df[i+1] > df[i-1] * 1.8) ||
+//                df[i] > 0.4) {
+
+            size_t peakLocation = i;
+
+            if (i + 1 < rawDf.size() &&
+                rawDf[i + 1] > rawDf[i] * 1.4) {
+
+                ++peakLocation;
+
+                if (m_debugLevel > 1) {
+                    std::cerr << "pushing hard peak forward to " << peakLocation << ": " << df[peakLocation] << " > " << df[peakLocation-1] << " * " << 1.4 << std::endl;
+                }
+            }
+
+            hardPeakCandidates.insert(peakLocation);
+            prevHardPeak = peakLocation;
+        }
+    }
+
+    size_t medianmaxsize = lrint(ceil(double(m_sampleRate) /
+                                 double(m_increment))); // 1 sec ish
+
+    if (m_debugLevel > 1) {
+        std::cerr << "mediansize = " << medianmaxsize << std::endl;
+    }
+    if (medianmaxsize < 7) {
+        medianmaxsize = 7;
+        if (m_debugLevel > 1) {
+            std::cerr << "adjusted mediansize = " << medianmaxsize << std::endl;
+        }
+    }
+
+    int minspacing = lrint(ceil(double(m_sampleRate) /
+                                (20 * double(m_increment)))); // 0.05 sec ish
+    
+    std::deque<float> medianwin;
+    std::vector<float> sorted;
+    int softPeakAmnesty = 0;
+
+    for (size_t i = 0; i < medianmaxsize/2; ++i) {
+        medianwin.push_back(0);
+    }
+    for (size_t i = 0; i < medianmaxsize/2 && i < df.size(); ++i) {
+        medianwin.push_back(df[i]);
+    }
+
+    size_t lastSoftPeak = 0;
+
+    for (size_t i = 0; i < df.size(); ++i) {
+        
+        size_t mediansize = medianmaxsize;
+
+        if (medianwin.size() < mediansize) {
+            mediansize = medianwin.size();
+        }
+
+        size_t middle = medianmaxsize / 2;
+        if (middle >= mediansize) middle = mediansize-1;
+
+        size_t nextDf = i + mediansize - middle;
+
+        if (mediansize < 2) {
+            if (mediansize > medianmaxsize) { // absurd, but never mind that
+                medianwin.pop_front();
+            }
+            if (nextDf < df.size()) {
+                medianwin.push_back(df[nextDf]);
+            } else {
+                medianwin.push_back(0);
+            }
+            continue;
+        }
+
+        if (m_debugLevel > 2) {
+//            std::cerr << "have " << mediansize << " in median buffer" << std::endl;
+        }
+
+        sorted.clear();
+        for (size_t j = 0; j < mediansize; ++j) {
+            sorted.push_back(medianwin[j]);
+        }
+        std::sort(sorted.begin(), sorted.end());
+
+        size_t n = 90; // percentile above which we pick peaks
+        size_t index = (sorted.size() * n) / 100;
+        if (index >= sorted.size()) index = sorted.size()-1;
+        if (index == sorted.size()-1 && index > 0) --index;
+        float thresh = sorted[index];
+
+//        if (m_debugLevel > 2) {
+//            std::cerr << "medianwin[" << middle << "] = " << medianwin[middle] << ", thresh = " << thresh << std::endl;
+//            if (medianwin[middle] == 0.f) {
+//                std::cerr << "contents: ";
+//                for (size_t j = 0; j < medianwin.size(); ++j) {
+//                    std::cerr << medianwin[j] << " ";
+//                }
+//                std::cerr << std::endl;
+//            }
+//        }
+
+        if (medianwin[middle] > thresh &&
+            medianwin[middle] > medianwin[middle-1] &&
+            medianwin[middle] > medianwin[middle+1] &&
+            softPeakAmnesty == 0) {
+
+            size_t maxindex = middle;
+            float maxval = medianwin[middle];
+
+            for (size_t j = middle+1; j < mediansize; ++j) {
+                if (medianwin[j] > maxval) {
+                    maxval = medianwin[j];
+                    maxindex = j;
+                } else if (medianwin[j] < medianwin[middle]) {
+                    break;
+                }
+            }
+
+            size_t peak = i + maxindex - middle;
+
+//            std::cerr << "i = " << i << ", maxindex = " << maxindex << ", middle = " << middle << ", so peak at " << peak << std::endl;
+
+            if (softPeakCandidates.empty() || lastSoftPeak != peak) {
+
+                if (m_debugLevel > 1) {
+                    std::cerr << "soft peak at " << peak << " ("
+                              << peak * m_increment << "): "
+                              << medianwin[middle] << " > "
+                              << thresh << " and "
+                              << medianwin[middle]
+                              << " > " << medianwin[middle-1] << " and "
+                              << medianwin[middle]
+                              << " > " << medianwin[middle+1]
+                              << std::endl;
+                }
+
+                if (peak >= df.size()) {
+                    if (m_debugLevel > 2) {
+                        std::cerr << "peak is beyond end"  << std::endl;
+                    }
+                } else {
+                    softPeakCandidates.insert(peak);
+                    lastSoftPeak = peak;
+                }
+            }
+
+            softPeakAmnesty = minspacing + maxindex - middle;
+            if (m_debugLevel > 2) {
+                std::cerr << "amnesty = " << softPeakAmnesty << std::endl;
+            }
+
+        } else if (softPeakAmnesty > 0) --softPeakAmnesty;
+
+        if (mediansize >= medianmaxsize) {
+            medianwin.pop_front();
+        }
+        if (nextDf < df.size()) {
+            medianwin.push_back(df[nextDf]);
+        } else {
+            medianwin.push_back(0);
+        }
+    }
+
+    std::vector<Peak> peaks;
+
+    while (!hardPeakCandidates.empty() || !softPeakCandidates.empty()) {
+
+        bool haveHardPeak = !hardPeakCandidates.empty();
+        bool haveSoftPeak = !softPeakCandidates.empty();
+
+        size_t hardPeak = (haveHardPeak ? *hardPeakCandidates.begin() : 0);
+        size_t softPeak = (haveSoftPeak ? *softPeakCandidates.begin() : 0);
+
+        Peak peak;
+        peak.hard = false;
+        peak.chunk = softPeak;
+
+        bool ignore = false;
+
+        if (haveHardPeak &&
+            (!haveSoftPeak || hardPeak <= softPeak)) {
+
+            if (m_debugLevel > 2) {
+                std::cerr << "Hard peak: " << hardPeak << std::endl;
+            }
+
+            peak.hard = true;
+            peak.chunk = hardPeak;
+            hardPeakCandidates.erase(hardPeakCandidates.begin());
+
+        } else {
+            if (m_debugLevel > 2) {
+                std::cerr << "Soft peak: " << softPeak << std::endl;
+            }
+            if (!peaks.empty() &&
+                peaks[peaks.size()-1].hard &&
+                peaks[peaks.size()-1].chunk + 3 >= softPeak) {
+                if (m_debugLevel > 2) {
+                    std::cerr << "(ignoring, as we just had a hard peak)"
+                              << std::endl;
+                }
+                ignore = true;
+            }
+        }            
+
+        if (haveSoftPeak && peak.chunk == softPeak) {
+            softPeakCandidates.erase(softPeakCandidates.begin());
+        }
+
+        if (!ignore) {
+            peaks.push_back(peak);
+        }
+    }                
+
+    return peaks;
+}
+
+std::vector<float>
+StretchCalculator::smoothDF(const std::vector<float> &df)
+{
+    std::vector<float> smoothedDF;
+    
+    for (size_t i = 0; i < df.size(); ++i) {
+        // three-value moving mean window for simple smoothing
+        float total = 0.f, count = 0;
+        if (i > 0) { total += df[i-1]; ++count; }
+        total += df[i]; ++count;
+        if (i+1 < df.size()) { total += df[i+1]; ++count; }
+        float mean = total / count;
+        smoothedDF.push_back(mean);
+    }
+
+    return smoothedDF;
+}
+
+std::vector<int>
+StretchCalculator::distributeRegion(const std::vector<float> &dfIn,
+                                    size_t duration, float ratio, bool phaseReset)
+{
+    std::vector<float> df(dfIn);
+    std::vector<int> increments;
+
+    // The peak for the stretch detection function may appear after
+    // the peak that we're using to calculate the start of the region.
+    // We don't want that.  If we find a peak in the first half of
+    // the region, we should set all the values up to that point to
+    // the same value as the peak.
+
+    // (This might not be subtle enough, especially if the region is
+    // long -- we want a bound that corresponds to acoustic perception
+    // of the audible bounce.)
+
+    for (size_t i = 1; i < df.size()/2; ++i) {
+        if (df[i] < df[i-1]) {
+            if (m_debugLevel > 1) {
+                std::cerr << "stretch peak offset: " << i-1 << " (peak " << df[i-1] << ")" << std::endl;
+            }
+            for (size_t j = 0; j < i-1; ++j) {
+                df[j] = df[i-1];
+            }
+            break;
+        }
+    }
+
+    float maxDf = 0;
+
+    for (size_t i = 0; i < df.size(); ++i) {
+        if (i == 0 || df[i] > maxDf) maxDf = df[i];
+    }
+
+    // We want to try to ensure the last 100ms or so (if possible) are
+    // tending back towards the maximum df, so that the stretchiness
+    // reduces at the end of the stretched region.
+    
+    int reducedRegion = lrint((0.1 * m_sampleRate) / m_increment);
+    if (reducedRegion > int(df.size()/5)) reducedRegion = df.size()/5;
+
+    for (int i = 0; i < reducedRegion; ++i) {
+        size_t index = df.size() - reducedRegion + i;
+        df[index] = df[index] + ((maxDf - df[index]) * i) / reducedRegion;
+    }
+
+    long toAllot = long(duration) - long(m_increment * df.size());
+    
+    if (m_debugLevel > 1) {
+        std::cerr << "region of " << df.size() << " chunks, output duration " << duration << ", toAllot " << toAllot << std::endl;
+    }
+
+    size_t totalIncrement = 0;
+
+    // We place limits on the amount of displacement per chunk.  if
+    // ratio < 0, no increment should be larger than increment*ratio
+    // or smaller than increment*ratio/2; if ratio > 0, none should be
+    // smaller than increment*ratio or larger than increment*ratio*2.
+    // We need to enforce this in the assignment of displacements to
+    // allotments, not by trying to respond if something turns out
+    // wrong.
+
+    // Note that the ratio is only provided to this function for the
+    // purposes of establishing this bound to the displacement.
+    
+    // so if
+    // maxDisplacement / totalDisplacement > increment * ratio*2 - increment
+    // (for ratio > 1)
+    // or
+    // maxDisplacement / totalDisplacement < increment * ratio/2
+    // (for ratio < 1)
+
+    // then we need to adjust and accommodate
+    
+    bool acceptableSquashRange = false;
+
+    double totalDisplacement = 0;
+    double maxDisplacement = 0; // min displacement will be 0 by definition
+
+    maxDf = 0;
+    float adj = 0;
+
+    while (!acceptableSquashRange) {
+
+        acceptableSquashRange = true;
+        calculateDisplacements(df, maxDf, totalDisplacement, maxDisplacement,
+                               adj);
+
+        if (m_debugLevel > 1) {
+            std::cerr << "totalDisplacement " << totalDisplacement << ", max " << maxDisplacement << " (maxDf " << maxDf << ", df count " << df.size() << ")" << std::endl;
+        }
+
+        if (totalDisplacement == 0) {
+// Not usually a problem, in fact
+//            std::cerr << "WARNING: totalDisplacement == 0 (duration " << duration << ", " << df.size() << " values in df)" << std::endl;
+            if (!df.empty() && adj == 0) {
+                acceptableSquashRange = false;
+                adj = 1;
+            }
+            continue;
+        }
+
+        int extremeIncrement = m_increment + lrint((toAllot * maxDisplacement) / totalDisplacement);
+        if (ratio < 1.0) {
+            if (extremeIncrement > lrint(ceil(m_increment * ratio))) {
+                std::cerr << "ERROR: extreme increment " << extremeIncrement << " > " << m_increment * ratio << " (this should not happen)" << std::endl;
+            } else if (extremeIncrement < (m_increment * ratio) / 2) {
+                if (m_debugLevel > 0) {
+                    std::cerr << "WARNING: extreme increment " << extremeIncrement << " < " << (m_increment * ratio) / 2 << std::endl;
+                }
+                acceptableSquashRange = false;
+            }
+        } else {
+            if (extremeIncrement > m_increment * ratio * 2) {
+                if (m_debugLevel > 0) {
+                    std::cerr << "WARNING: extreme increment " << extremeIncrement << " > " << m_increment * ratio * 2 << std::endl;
+                }
+                acceptableSquashRange = false;
+            } else if (extremeIncrement < lrint(floor(m_increment * ratio))) {
+                std::cerr << "ERROR: extreme increment " << extremeIncrement << " < " << m_increment * ratio << " (I thought this couldn't happen?)" << std::endl;
+            }
+        }
+
+        if (!acceptableSquashRange) {
+            // Need to make maxDisplacement smaller as a proportion of
+            // the total displacement, yet ensure that the
+            // displacements still sum to the total.
+            adj += maxDf/10;
+        }
+    }
+
+    for (size_t i = 0; i < df.size(); ++i) {
+
+        double displacement = maxDf - df[i];
+        if (displacement < 0) displacement -= adj;
+        else displacement += adj;
+
+        if (i == 0 && phaseReset) {
+            if (df.size() == 1) {
+                increments.push_back(duration);
+                totalIncrement += duration;
+            } else {
+                increments.push_back(m_increment);
+                totalIncrement += m_increment;
+            }
+            totalDisplacement -= displacement;
+            continue;
+        }
+
+        double theoreticalAllotment = 0;
+
+        if (totalDisplacement != 0) {
+            theoreticalAllotment = (toAllot * displacement) / totalDisplacement;
+        }
+        int allotment = lrint(theoreticalAllotment);
+        if (i + 1 == df.size()) allotment = toAllot;
+
+        int increment = m_increment + allotment;
+
+        if (increment <= 0) {
+            // this is a serious problem, the allocation is quite
+            // wrong if it allows increment to diverge so far from the
+            // input increment
+            std::cerr << "*** WARNING: increment " << increment << " <= 0, rounding to zero" << std::endl;
+            increment = 0;
+            allotment = increment - m_increment;
+        }
+
+        increments.push_back(increment);
+        totalIncrement += increment;
+
+        toAllot -= allotment;
+        totalDisplacement -= displacement;
+
+        if (m_debugLevel > 2) {
+            std::cerr << "df " << df[i] << ", smoothed " << df[i] << ", disp " << displacement << ", allot " << theoreticalAllotment << ", incr " << increment << ", remain " << toAllot << std::endl;
+        }
+    }
+    
+    if (m_debugLevel > 2) {
+        std::cerr << "total increment: " << totalIncrement << ", left over: " << toAllot << " to allot, displacement " << totalDisplacement << std::endl;
+    }
+
+    if (totalIncrement != duration) {
+        std::cerr << "*** WARNING: calculated output duration " << totalIncrement << " != expected " << duration << std::endl;
+    }
+
+    return increments;
+}
+
+void
+StretchCalculator::calculateDisplacements(const std::vector<float> &df,
+                                          float &maxDf,
+                                          double &totalDisplacement,
+                                          double &maxDisplacement,
+                                          float adj) const
+{
+    totalDisplacement = maxDisplacement = 0;
+
+    maxDf = 0;
+
+    for (size_t i = 0; i < df.size(); ++i) {
+        if (i == 0 || df[i] > maxDf) maxDf = df[i];
+    }
+
+    for (size_t i = 0; i < df.size(); ++i) {
+        double displacement = maxDf - df[i];
+        if (displacement < 0) displacement -= adj;
+        else displacement += adj;
+        totalDisplacement += displacement;
+        if (i == 0 || displacement > maxDisplacement) {
+            maxDisplacement = displacement;
+        }
+    }
+}
+
+}
+

libs/rubberband/1.0/src/StretchCalculator.h

@@ -0,0 +1,95 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_STRETCH_CALCULATOR_H_
+#define _RUBBERBAND_STRETCH_CALCULATOR_H_
+
+#include <sys/types.h>
+
+#include <vector>
+
+namespace RubberBand
+{
+
+class StretchCalculator
+{
+public:
+    StretchCalculator(size_t sampleRate, size_t inputIncrement, bool useHardPeaks);
+    virtual ~StretchCalculator();
+
+    /**
+     * Calculate phase increments for a region of audio, given the
+     * overall target stretch ratio, input duration in audio samples,
+     * and the audio curves to use for identifying phase lock points
+     * (lockAudioCurve) and for allocating stretches to relatively
+     * less prominent points (stretchAudioCurve).
+     */
+    virtual std::vector<int> calculate(double ratio, size_t inputDuration,
+                                       const std::vector<float> &lockAudioCurve,
+                                       const std::vector<float> &stretchAudioCurve);
+
+    /**
+     * Calculate the phase increment for a single audio block, given
+     * the overall target stretch ratio and the block's value on the
+     * phase-lock audio curve.  State is retained between calls in the
+     * StretchCalculator object; call reset() to reset it.  This uses
+     * a less sophisticated method than the offline calculate().
+     */
+    virtual int calculateSingle(double ratio, size_t inputDurationSoFar,
+                                float curveValue);
+
+    void setUseHardPeaks(bool use) { m_useHardPeaks = use; }
+
+    void reset();
+  
+    void setDebugLevel(int level) { m_debugLevel = level; }
+
+    struct Peak {
+        size_t chunk;
+        bool hard;
+    };
+    std::vector<Peak> getLastCalculatedPeaks() const { return m_lastPeaks; }
+
+    std::vector<float> smoothDF(const std::vector<float> &df);
+
+protected:
+    std::vector<Peak> findPeaks(const std::vector<float> &audioCurve);
+
+    std::vector<int> distributeRegion(const std::vector<float> &regionCurve,
+                                      size_t outputDuration, float ratio,
+                                      bool phaseReset);
+
+    void calculateDisplacements(const std::vector<float> &df,
+                                float &maxDf,
+                                double &totalDisplacement,
+                                double &maxDisplacement,
+                                float adj) const;
+
+    size_t m_sampleRate;
+    size_t m_blockSize;
+    size_t m_increment;
+    float m_prevDf;
+    double m_divergence;
+    float m_recovery;
+    float m_prevRatio;
+    int m_transientAmnesty; // only in RT mode; handled differently offline
+    int m_debugLevel;
+    bool m_useHardPeaks;
+    
+    std::vector<Peak> m_lastPeaks;
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/StretcherChannelData.cpp

@@ -0,0 +1,271 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+#include "StretcherChannelData.h"
+
+#include "Resampler.h"
+
+namespace RubberBand 
+{
+
+RubberBandStretcher::Impl::ChannelData::ChannelData(size_t windowSize,
+                                                    size_t outbufSize)
+{
+    std::set<size_t> s;
+    construct(s, windowSize, outbufSize);
+}
+
+RubberBandStretcher::Impl::ChannelData::ChannelData(const std::set<size_t> &windowSizes,
+                                                    size_t initialWindowSize,
+                                                    size_t outbufSize)
+{
+    construct(windowSizes, initialWindowSize, outbufSize);
+}
+
+void
+RubberBandStretcher::Impl::ChannelData::construct(const std::set<size_t> &windowSizes,
+                                                  size_t initialWindowSize,
+                                                  size_t outbufSize)
+{
+    size_t maxSize = initialWindowSize;
+
+    if (!windowSizes.empty()) {
+        // std::set is ordered by value
+        std::set<size_t>::const_iterator i = windowSizes.end();
+        maxSize = *--i;
+    }
+    if (windowSizes.find(initialWindowSize) == windowSizes.end()) {
+        if (initialWindowSize > maxSize) maxSize = initialWindowSize;
+    }
+
+    size_t realSize = maxSize/2 + 1; // size of the real "half" of freq data
+
+//    std::cerr << "ChannelData::construct([" << windowSizes.size() << "], " << maxSize << ", " << outbufSize << ")" << std::endl;
+    
+    if (outbufSize < maxSize) outbufSize = maxSize;
+
+    inbuf = new RingBuffer<float>(maxSize);
+    outbuf = new RingBuffer<float>(outbufSize);
+
+    mag = new double[realSize];
+    phase = new double[realSize];
+    prevPhase = new double[realSize];
+    unwrappedPhase = new double[realSize];
+    freqPeak = new size_t[realSize];
+
+    accumulator = new float[maxSize];
+    windowAccumulator = new float[maxSize];
+
+    fltbuf = new float[maxSize];
+
+    for (std::set<size_t>::const_iterator i = windowSizes.begin();
+         i != windowSizes.end(); ++i) {
+        ffts[*i] = new FFT(*i);
+        ffts[*i]->initDouble();
+    }
+    if (windowSizes.find(initialWindowSize) == windowSizes.end()) {
+        ffts[initialWindowSize] = new FFT(initialWindowSize);
+        ffts[initialWindowSize]->initDouble();
+    }
+    fft = ffts[initialWindowSize];
+
+    dblbuf = fft->getDoubleTimeBuffer();
+
+    resampler = 0;
+    resamplebuf = 0;
+    resamplebufSize = 0;
+
+    reset();
+
+    for (size_t i = 0; i < realSize; ++i) {
+        mag[i] = 0.0;
+        phase[i] = 0.0;
+        prevPhase[i] = 0.0;
+        unwrappedPhase[i] = 0.0;
+        freqPeak[i] = 0;
+    }
+
+    for (size_t i = 0; i < initialWindowSize; ++i) {
+        dblbuf[i] = 0.0;
+    }
+
+    for (size_t i = 0; i < maxSize; ++i) {
+        accumulator[i] = 0.f;
+        windowAccumulator[i] = 0.f;
+        fltbuf[i] = 0.0;
+    }
+}
+
+void
+RubberBandStretcher::Impl::ChannelData::setWindowSize(size_t windowSize)
+{
+    size_t oldSize = inbuf->getSize();
+    size_t realSize = windowSize/2 + 1;
+
+//    std::cerr << "ChannelData::setWindowSize(" << windowSize << ") [from " << oldSize << "]" << std::endl;
+
+    if (oldSize >= windowSize) {
+
+        // no need to reallocate buffers, just reselect fft
+
+        //!!! we can't actually do this without locking against the
+        //process thread, can we?  we need to zero the mag/phase
+        //buffers without interference
+
+        if (ffts.find(windowSize) == ffts.end()) {
+            //!!! this also requires a lock, but it shouldn't occur in
+            //RT mode with proper initialisation
+            ffts[windowSize] = new FFT(windowSize);
+            ffts[windowSize]->initDouble();
+        }
+        
+        fft = ffts[windowSize];
+
+        dblbuf = fft->getDoubleTimeBuffer();
+
+        for (size_t i = 0; i < windowSize; ++i) {
+            dblbuf[i] = 0.0;
+        }
+
+        for (size_t i = 0; i < realSize; ++i) {
+            mag[i] = 0.0;
+            phase[i] = 0.0;
+            prevPhase[i] = 0.0;
+            unwrappedPhase[i] = 0.0;
+            freqPeak[i] = 0;
+        }
+
+        return;
+    }
+
+    //!!! at this point we need a lock in case a different client
+    //thread is calling process() -- we need this lock even if we
+    //aren't running in threaded mode ourselves -- if we're in RT
+    //mode, then the process call should trylock and fail if the lock
+    //is unavailable (since this should never normally be the case in
+    //general use in RT mode)
+
+    RingBuffer<float> *newbuf = inbuf->resized(windowSize);
+    delete inbuf;
+    inbuf = newbuf;
+
+    // We don't want to preserve data in these arrays
+
+    delete[] mag;
+    delete[] phase;
+    delete[] prevPhase;
+    delete[] unwrappedPhase;
+    delete[] freqPeak;
+
+    mag = new double[realSize];
+    phase = new double[realSize];
+    prevPhase = new double[realSize];
+    unwrappedPhase = new double[realSize];
+    freqPeak = new size_t[realSize];
+
+    delete[] fltbuf;
+    fltbuf = new float[windowSize];
+
+    // But we do want to preserve data in these
+
+    float *newAcc = new float[windowSize];
+    for (size_t i = 0; i < oldSize; ++i) newAcc[i] = accumulator[i];
+    delete[] accumulator;
+    accumulator = newAcc;
+
+    newAcc = new float[windowSize];
+    for (size_t i = 0; i < oldSize; ++i) newAcc[i] = windowAccumulator[i];
+    delete[] windowAccumulator;
+    windowAccumulator = newAcc;
+    
+    //!!! and resampler?
+
+    for (size_t i = 0; i < realSize; ++i) {
+        mag[i] = 0.0;
+        phase[i] = 0.0;
+        prevPhase[i] = 0.0;
+        unwrappedPhase[i] = 0.0;
+        freqPeak[i] = 0;
+    }
+
+    for (size_t i = 0; i < windowSize; ++i) {
+        fltbuf[i] = 0.0;
+    }
+
+    for (size_t i = oldSize; i < windowSize; ++i) {
+        accumulator[i] = 0.f;
+        windowAccumulator[i] = 0.f;
+    }
+
+    if (ffts.find(windowSize) == ffts.end()) {
+        ffts[windowSize] = new FFT(windowSize);
+        ffts[windowSize]->initDouble();
+    }
+    
+    fft = ffts[windowSize];
+
+    dblbuf = fft->getDoubleTimeBuffer();
+
+    for (size_t i = 0; i < windowSize; ++i) {
+        dblbuf[i] = 0.0;
+    }
+}
+
+void
+RubberBandStretcher::Impl::ChannelData::setOutbufSize(size_t outbufSize)
+{
+    size_t oldSize = outbuf->getSize();
+
+//    std::cerr << "ChannelData::setOutbufSize(" << outbufSize << ") [from " << oldSize << "]" << std::endl;
+
+    if (oldSize < outbufSize) {
+
+        //!!! at this point we need a lock in case a different client
+        //thread is calling process()
+
+        RingBuffer<float> *newbuf = outbuf->resized(outbufSize);
+        delete outbuf;
+        outbuf = newbuf;
+    }
+}
+
+RubberBandStretcher::Impl::ChannelData::~ChannelData()
+{
+    delete resampler;
+    delete[] resamplebuf;
+
+    delete inbuf;
+    delete outbuf;
+    delete[] mag;
+    delete[] phase;
+    delete[] prevPhase;
+    delete[] unwrappedPhase;
+    delete[] freqPeak;
+    delete[] accumulator;
+    delete[] windowAccumulator;
+    delete[] fltbuf;
+
+    for (std::map<size_t, FFT *>::iterator i = ffts.begin();
+         i != ffts.end(); ++i) {
+        delete i->second;
+    }
+}
+
+void
+RubberBandStretcher::Impl::ChannelData::reset()
+{
+    inbuf->reset();
+    outbuf->reset();
+
+    if (resampler) resampler->reset();
+
+    accumulatorFill = 0;
+    prevIncrement = 0;
+    chunkCount = 0;
+    inCount = 0;
+    inputSize = -1;
+    outCount = 0;
+    draining = false;
+    outputComplete = false;
+}
+
+}

libs/rubberband/1.0/src/StretcherChannelData.h

@@ -0,0 +1,121 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_STRETCHERCHANNELDATA_H_
+#define _RUBBERBAND_STRETCHERCHANNELDATA_H_
+
+#include "StretcherImpl.h"
+
+#include <set>
+
+namespace RubberBand
+{
+
+class Resampler;
+	
+class RubberBandStretcher::Impl::ChannelData
+{
+public:        
+    /**
+     * Construct a ChannelData structure.
+     *
+     * The window size passed in here is the size for the FFT
+     * calculation, and most of the buffer sizes also depend on
+     * it.  In practice it is always a power of two and except for
+     * very extreme stretches is always either 1024, 2048 or 4096.
+     *
+     * The outbuf size depends on other factors as well, including
+     * the pitch scale factor and any maximum processing block
+     * size specified by the user of the code.
+     */
+    ChannelData(size_t windowSize, size_t outbufSize);
+
+    /**
+     * Construct a ChannelData structure that can process at
+     * different FFT sizes without requiring reallocation when the
+     * size changes.  The size can subsequently be changed with a
+     * call to setWindowSize.  Reallocation will only be necessary
+     * if setWindowSize is called with a value not equal to one of
+     * those passed in to the constructor.
+     *
+     * The outbufSize should be the maximum possible outbufSize to
+     * avoid reallocation, which will happen if setOutbufSize is
+     * called subsequently.
+     */
+    ChannelData(const std::set<size_t> &windowSizes,
+                size_t initialWindowSize, size_t outbufSize);
+    ~ChannelData();
+
+    /**
+     * Reset buffers
+     */
+    void reset();
+
+    /**
+     * Set the FFT and buffer sizes from the given processing
+     * window size.  If this ChannelData was constructed with a set
+     * of window sizes and the given window size here was among
+     * them, no reallocation will be required.
+     */
+    void setWindowSize(size_t windowSize);
+
+    /**
+     * Set the outbufSize for the channel data.  Reallocation will
+     * occur.
+     */
+    void setOutbufSize(size_t outbufSize);
+
+    RingBuffer<float> *inbuf;
+    RingBuffer<float> *outbuf;
+
+    double *mag;
+    double *phase;
+
+    double *prevPhase;
+    double *unwrappedPhase;
+
+    size_t *freqPeak;
+
+    float *accumulator;
+    size_t accumulatorFill;
+    float *windowAccumulator;
+
+    float *fltbuf;
+    double *dblbuf; // owned by FFT object, only used for time domain FFT i/o
+
+    size_t prevIncrement; // only used in RT mode
+
+    size_t chunkCount;
+    size_t inCount;
+    long inputSize; // set only after known (when data ended); -1 previously
+    size_t outCount;
+
+    bool draining;
+    bool outputComplete;
+
+    FFT *fft;
+    std::map<size_t, FFT *> ffts;
+
+    Resampler *resampler;
+    float *resamplebuf;
+    size_t resamplebufSize;
+
+private:
+    void construct(const std::set<size_t> &windowSizes,
+                   size_t initialWindowSize, size_t outbufSize);
+};        
+
+}
+
+#endif

libs/rubberband/1.0/src/StretcherImpl.h

@@ -0,0 +1,194 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_STRETCHERIMPL_H_
+#define _RUBBERBAND_STRETCHERIMPL_H_
+
+#include "RubberBandStretcher.h"
+
+#include "Window.h"
+#include "Thread.h"
+#include "RingBuffer.h"
+#include "FFT.h"
+#include "sysutils.h"
+
+#include <set>
+
+namespace RubberBand
+{
+
+class AudioCurve;
+class StretchCalculator;
+
+class RubberBandStretcher::Impl
+{
+public:
+    Impl(RubberBandStretcher *stretcher,
+         size_t sampleRate, size_t channels, Options options,
+         double initialTimeRatio, double initialPitchScale);
+    ~Impl();
+    
+    void reset();
+    void setTimeRatio(double ratio);
+    void setPitchScale(double scale);
+
+    double getTimeRatio() const;
+    double getPitchScale() const;
+
+    size_t getLatency() const;
+
+    void setTransientsOption(Options);
+    void setPhaseOption(Options);
+
+    void setExpectedInputDuration(size_t samples);
+    void setMaxProcessSize(size_t samples);
+
+    size_t getSamplesRequired() const;
+
+    void study(const float *const *input, size_t samples, bool final);
+    void process(const float *const *input, size_t samples, bool final);
+
+    int available() const;
+    size_t retrieve(float *const *output, size_t samples) const;
+
+    float getFrequencyCutoff(int n) const;
+    void setFrequencyCutoff(int n, float f);
+
+    size_t getInputIncrement() const {
+        return m_increment;
+    }
+
+    std::vector<int> getOutputIncrements() const;
+    std::vector<float> getPhaseResetCurve() const;
+    std::vector<int> getExactTimePoints() const;
+
+    size_t getChannelCount() const {
+        return m_channels;
+    }
+    
+    void calculateStretch();
+
+    void setDebugLevel(int level);
+    static void setDefaultDebugLevel(int level) { m_defaultDebugLevel = level; }
+
+protected:
+    RubberBandStretcher *m_stretcher;
+    size_t m_channels;
+
+    size_t consumeChannel(size_t channel, const float *input, size_t samples);
+    void processChunks(size_t channel, bool &any, bool &last);
+    bool processOneChunk(); // across all channels, for real time use
+    bool processChunkForChannel(size_t channel, size_t phaseIncrement,
+                                size_t shiftIncrement, bool phaseReset);
+    bool testInbufReadSpace(size_t channel);
+    void calculateIncrements(size_t &phaseIncrement,
+                             size_t &shiftIncrement, bool &phaseReset);
+    bool getIncrements(size_t channel, size_t &phaseIncrement,
+                       size_t &shiftIncrement, bool &phaseReset);
+    void analyseChunk(size_t channel);
+    void modifyChunk(size_t channel, size_t outputIncrement, bool phaseReset);
+    void synthesiseChunk(size_t channel);
+    void writeChunk(size_t channel, size_t shiftIncrement, bool last);
+
+    void calculateSizes();
+    void configure();
+    void reconfigure();
+
+    double getEffectiveRatio() const;
+    
+    size_t roundUp(size_t value); // to next power of two
+
+    double m_timeRatio;
+    double m_pitchScale;
+
+    size_t m_windowSize;
+    size_t m_increment;
+    size_t m_outbufSize;
+
+    size_t m_maxProcessSize;
+    size_t m_expectedInputDuration;
+    
+    bool m_threaded;
+    bool m_realtime;
+    Options m_options;
+    int m_debugLevel;
+
+    enum ProcessMode {
+        JustCreated,
+        Studying,
+        Processing,
+        Finished
+    };
+
+    ProcessMode m_mode;
+
+    std::map<size_t, Window<float> *> m_windows;
+    Window<float> *m_window;
+    FFT *m_studyFFT;
+
+    Condition m_spaceAvailable;
+    
+    class ProcessThread : public Thread
+    {
+    public:
+        ProcessThread(Impl *s, size_t c);
+        void run();
+        void signalDataAvailable();
+        void abandon();
+    private:
+        Impl *m_s;
+        size_t m_channel;
+        Condition m_dataAvailable;
+        bool m_abandoning;
+    };
+
+    mutable Mutex m_threadSetMutex;
+    typedef std::set<ProcessThread *> ThreadSet;
+    ThreadSet m_threadSet;
+    
+
+    size_t m_inputDuration;
+    std::vector<float> m_phaseResetDf;
+    std::vector<float> m_stretchDf;
+
+    class ChannelData; 
+    std::vector<ChannelData *> m_channelData;
+
+    std::vector<int> m_outputIncrements;
+
+    mutable RingBuffer<int> m_lastProcessOutputIncrements;
+    mutable RingBuffer<float> m_lastProcessPhaseResetDf;
+
+    AudioCurve *m_phaseResetAudioCurve;
+    AudioCurve *m_stretchAudioCurve;
+    StretchCalculator *m_stretchCalculator;
+
+    float m_freq0;
+    float m_freq1;
+    float m_freq2;
+
+    size_t m_baseWindowSize;
+    float m_rateMultiple;
+
+    void writeOutput(RingBuffer<float> &to, float *from,
+                     size_t qty, size_t &outCount, size_t theoreticalOut);
+
+    static int m_defaultDebugLevel;
+    static const size_t m_defaultIncrement;
+    static const size_t m_defaultWindowSize;
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/StretcherProcess.cpp

@@ -0,0 +1,927 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "StretcherImpl.h"
+#include "PercussiveAudioCurve.h"
+#include "HighFrequencyAudioCurve.h"
+#include "ConstantAudioCurve.h"
+#include "StretchCalculator.h"
+#include "StretcherChannelData.h"
+#include "Resampler.h"
+
+#include <cassert>
+#include <cmath>
+#include <set>
+#include <map>
+
+using std::cerr;
+using std::endl;
+
+namespace RubberBand {
+
+RubberBandStretcher::Impl::ProcessThread::ProcessThread(Impl *s, size_t c) :
+    m_s(s),
+    m_channel(c),
+    m_dataAvailable(std::string("data ") + char('A' + c)),
+    m_abandoning(false)
+{ }
+
+void
+RubberBandStretcher::Impl::ProcessThread::run()
+{
+    if (m_s->m_debugLevel > 1) {
+        cerr << "thread " << m_channel << " getting going" << endl;
+    }
+
+    ChannelData &cd = *m_s->m_channelData[m_channel];
+
+    while (cd.inputSize == -1 ||
+           cd.inbuf->getReadSpace() > 0) {
+
+//        if (cd.inputSize != -1) {
+//            cerr << "inputSize == " << cd.inputSize
+//                 << ", readSpace == " << cd.inbuf->getReadSpace() << endl;
+//        }
+        
+        bool any = false, last = false;
+        m_s->processChunks(m_channel, any, last);
+
+        if (last) break;
+
+        if (any) m_s->m_spaceAvailable.signal();
+
+        m_dataAvailable.lock();
+        if (!m_s->testInbufReadSpace(m_channel) && !m_abandoning) {
+            m_dataAvailable.wait();
+        } else {
+            m_dataAvailable.unlock();
+        }
+
+        if (m_abandoning) {
+            if (m_s->m_debugLevel > 1) {
+                cerr << "thread " << m_channel << " abandoning" << endl;
+            }
+            return;
+        }
+    }
+
+    bool any = false, last = false;
+    m_s->processChunks(m_channel, any, last);
+    m_s->m_spaceAvailable.signal();
+    
+    if (m_s->m_debugLevel > 1) {
+        cerr << "thread " << m_channel << " done" << endl;
+    }
+}
+
+void
+RubberBandStretcher::Impl::ProcessThread::signalDataAvailable()
+{
+    m_dataAvailable.signal();
+}
+
+void
+RubberBandStretcher::Impl::ProcessThread::abandon()
+{
+    m_abandoning = true;
+}
+
+void
+RubberBandStretcher::Impl::processChunks(size_t c, bool &any, bool &last)
+{
+    // Process as many chunks as there are available on the input
+    // buffer for channel c.  This requires that the increments have
+    // already been calculated.
+
+    ChannelData &cd = *m_channelData[c];
+
+    last = false;
+    any = false;
+
+    while (!last) {
+
+        if (!testInbufReadSpace(c)) {
+//            cerr << "not enough input" << endl;
+            break;
+        }
+
+        any = true;
+
+        if (!cd.draining) {
+            size_t got = cd.inbuf->peek(cd.fltbuf, m_windowSize);
+            assert(got == m_windowSize || cd.inputSize >= 0);
+            cd.inbuf->skip(m_increment);
+            analyseChunk(c);
+        }
+
+        bool phaseReset = false;
+        size_t phaseIncrement, shiftIncrement;
+        getIncrements(c, phaseIncrement, shiftIncrement, phaseReset);
+
+        last = processChunkForChannel(c, phaseIncrement, shiftIncrement, phaseReset);
+        cd.chunkCount++;
+        if (m_debugLevel > 2) {
+            cerr << "channel " << c << ": last = " << last << ", chunkCount = " << cd.chunkCount << endl;
+        }
+    }
+}
+
+bool
+RubberBandStretcher::Impl::processOneChunk()
+{
+    // Process a single chunk for all channels, provided there is
+    // enough data on each channel for at least one chunk.  This is
+    // able to calculate increments as it goes along.
+
+    for (size_t c = 0; c < m_channels; ++c) {
+        if (!testInbufReadSpace(c)) return false;
+        ChannelData &cd = *m_channelData[c];
+        if (!cd.draining) {
+            size_t got = cd.inbuf->peek(cd.fltbuf, m_windowSize);
+            assert(got == m_windowSize || cd.inputSize >= 0);
+            cd.inbuf->skip(m_increment);
+            analyseChunk(c);
+        }
+    }
+    
+    bool phaseReset = false;
+    size_t phaseIncrement, shiftIncrement;
+    if (!getIncrements(0, phaseIncrement, shiftIncrement, phaseReset)) {
+        calculateIncrements(phaseIncrement, shiftIncrement, phaseReset);
+    }
+
+    bool last = false;
+    for (size_t c = 0; c < m_channels; ++c) {
+        last = processChunkForChannel(c, phaseIncrement, shiftIncrement, phaseReset);
+        m_channelData[c]->chunkCount++;
+    }
+
+    return last;
+}
+
+bool
+RubberBandStretcher::Impl::testInbufReadSpace(size_t c)
+{
+    ChannelData &cd = *m_channelData[c];
+    RingBuffer<float> &inbuf = *cd.inbuf;
+
+    size_t rs = inbuf.getReadSpace();
+
+    if (rs < m_windowSize && !cd.draining) {
+            
+        if (cd.inputSize == -1) {
+
+            // Not all the input data has been written to the inbuf
+            // (that's why the input size is not yet set).  We can't
+            // process, because we don't have a full chunk of data, so
+            // our process chunk would contain some empty padding in
+            // its input -- and that would give incorrect output, as
+            // we know there is more input to come.
+
+            if (!m_threaded) {
+//                cerr << "WARNING: RubberBandStretcher: read space < chunk size ("
+//                          << inbuf.getReadSpace() << " < " << m_windowSize
+//                          << ") when not all input written, on processChunks for channel " << c << endl;
+            }
+            return false;
+        }
+        
+        if (rs == 0) {
+
+            if (m_debugLevel > 1) {
+                cerr << "read space = 0, giving up" << endl;
+            }
+            return false;
+
+        } else if (rs < m_windowSize/2) {
+
+            if (m_debugLevel > 1) {
+                cerr << "read space = " << rs << ", setting draining true" << endl;
+            }
+            
+            cd.draining = true;
+        }
+    }
+
+    return true;
+}
+
+bool 
+RubberBandStretcher::Impl::processChunkForChannel(size_t c,
+                                                  size_t phaseIncrement,
+                                                  size_t shiftIncrement,
+                                                  bool phaseReset)
+{
+    // Process a single chunk on a single channel.  This assumes
+    // enough input data is available; caller must have tested this
+    // using e.g. testInbufReadSpace first.  Return true if this is
+    // the last chunk on the channel.
+
+    if (phaseReset && (m_debugLevel > 1)) {
+        cerr << "processChunkForChannel: phase reset found, incrs "
+                  << phaseIncrement << ":" << shiftIncrement << endl;
+    }
+
+    ChannelData &cd = *m_channelData[c];
+
+    if (!cd.draining) {
+        
+        // This is the normal processing case -- draining is only
+        // set when all the input has been used and we only need
+        // to write from the existing accumulator into the output.
+        
+        // We know we have enough samples available in m_inbuf --
+        // this is usually m_windowSize, but we know that if fewer
+        // are available, it's OK to use zeroes for the rest
+        // (which the ring buffer will provide) because we've
+        // reached the true end of the data.
+        
+        // We need to peek m_windowSize samples for processing, and
+        // then skip m_increment to advance the read pointer.
+        
+        modifyChunk(c, phaseIncrement, phaseReset);
+        synthesiseChunk(c); // reads from cd.mag, cd.phase
+
+        if (m_debugLevel > 2) {
+            if (phaseReset) {
+                for (int i = 0; i < 10; ++i) {
+                    cd.accumulator[i] = 1.2f - (i % 3) * 1.2f;
+                }
+            }
+        }
+    }
+
+    bool last = false;
+
+    if (cd.draining) {
+        if (m_debugLevel > 1) {
+            cerr << "draining: accumulator fill = " << cd.accumulatorFill << " (shiftIncrement = " << shiftIncrement << ")" <<  endl;
+        }
+        if (shiftIncrement == 0) {
+            cerr << "WARNING: draining: shiftIncrement == 0, can't handle that in this context: setting to " << m_increment << endl;
+            shiftIncrement = m_increment;
+        }
+        if (cd.accumulatorFill <= shiftIncrement) {
+            if (m_debugLevel > 1) {
+                cerr << "reducing shift increment from " << shiftIncrement
+                          << " to " << cd.accumulatorFill
+                          << " and marking as last" << endl;
+            }
+            shiftIncrement = cd.accumulatorFill;
+            last = true;
+        }
+    }
+        
+    if (m_threaded) {
+        size_t required = shiftIncrement;
+        if (m_pitchScale != 1.0) {
+            required = int(required / m_pitchScale) + 1;
+        }
+        
+        if (cd.outbuf->getWriteSpace() < required) {
+            if (m_debugLevel > 0) {
+                cerr << "Buffer overrun on output for channel " << c << endl;
+            }
+
+            //!!! The only correct thing we can do here is resize the
+            // buffer.  We can't wait for the client thread to read
+            // some data out from the buffer so as to make more space,
+            // because the client thread is probably stuck in a
+            // process() call waiting for us to stow away enough input
+            // increments to allow the process() call to complete.
+
+        }
+    }
+    
+    writeChunk(c, shiftIncrement, last);
+    return last;
+}
+
+void
+RubberBandStretcher::Impl::calculateIncrements(size_t &phaseIncrementRtn,
+                                               size_t &shiftIncrementRtn,
+                                               bool &phaseReset)
+{
+//    cerr << "calculateIncrements" << endl;
+    
+    // Calculate the next upcoming phase and shift increment, on the
+    // basis that both channels are in sync.  This is in contrast to
+    // getIncrements, which requires that all the increments have been
+    // calculated in advance but can then return increments
+    // corresponding to different chunks in different channels.
+
+    // Requires frequency domain representations of channel data in
+    // the mag and phase buffers in the channel.
+
+    // This function is only used in real-time mode.
+
+    phaseIncrementRtn = m_increment;
+    shiftIncrementRtn = m_increment;
+    phaseReset = false;
+
+    if (m_channels == 0) return;
+
+    ChannelData &cd = *m_channelData[0];
+
+    size_t bc = cd.chunkCount;
+    for (size_t c = 1; c < m_channels; ++c) {
+        if (m_channelData[c]->chunkCount != bc) {
+            cerr << "ERROR: RubberBandStretcher::Impl::calculateIncrements: Channels are not in sync" << endl;
+            return;
+        }
+    }
+
+    // Normally we would mix down the time-domain signal and apply a
+    // single FFT, or else mix down the Cartesian form of the
+    // frequency-domain signal.  Both of those would be inefficient
+    // from this position.  Fortunately, the onset detectors should
+    // work reasonably well (maybe even better?) if we just sum the
+    // magnitudes of the frequency-domain channel signals and forget
+    // about phase entirely.  Normally we don't expect the channel
+    // phases to cancel each other, and broadband effects will still
+    // be apparent.
+
+    for (size_t i = 0; i <= m_windowSize/2; ++i) {
+        cd.fltbuf[i] = 0.0;
+    }
+
+    for (size_t c = 0; c < m_channels; ++c) {
+        for (size_t i = 0; i <= m_windowSize/2; ++i) {
+            cd.fltbuf[i] += m_channelData[c]->mag[i];
+        }
+    }
+    
+    float df = m_phaseResetAudioCurve->process(cd.fltbuf, m_increment);
+
+    int incr = m_stretchCalculator->calculateSingle
+        (getEffectiveRatio(),
+         m_inputDuration, //!!! no, totally wrong... fortunately it doesn't matter atm
+         df);
+
+    m_lastProcessPhaseResetDf.write(&df, 1);
+    m_lastProcessOutputIncrements.write(&incr, 1);
+
+    if (incr < 0) {
+        phaseReset = true;
+        incr = -incr;
+    }
+    
+    // The returned increment is the phase increment.  The shift
+    // increment for one chunk is the same as the phase increment for
+    // the following chunk (see comment below).  This means we don't
+    // actually know the shift increment until we see the following
+    // phase increment... which is a bit of a problem.
+
+    // This implies we should use this increment for the shift
+    // increment, and make the following phase increment the same as
+    // it.  This means in RT mode we'll be one chunk later with our
+    // phase reset than we would be in non-RT mode.  The sensitivity
+    // of the broadband onset detector may mean that this isn't a
+    // problem -- test it and see.
+
+    shiftIncrementRtn = incr;
+
+    if (cd.prevIncrement == 0) {
+        phaseIncrementRtn = shiftIncrementRtn;
+    } else {
+        phaseIncrementRtn = cd.prevIncrement;
+    }
+
+    cd.prevIncrement = shiftIncrementRtn;
+}
+
+bool
+RubberBandStretcher::Impl::getIncrements(size_t channel,
+                                         size_t &phaseIncrementRtn,
+                                         size_t &shiftIncrementRtn,
+                                         bool &phaseReset)
+{
+    if (channel >= m_channels) {
+        phaseIncrementRtn = m_increment;
+        shiftIncrementRtn = m_increment;
+        phaseReset = false;
+        return false;
+    }
+
+    // There are two relevant output increments here.  The first is
+    // the phase increment which we use when recalculating the phases
+    // for the current chunk; the second is the shift increment used
+    // to determine how far to shift the processing buffer after
+    // writing the chunk.  The shift increment for one chunk is the
+    // same as the phase increment for the following chunk.
+    
+    // When an onset occurs for which we need to reset phases, the
+    // increment given will be negative.
+    
+    // When we reset phases, the previous shift increment (and so
+    // current phase increments) must have been m_increment to ensure
+    // consistency.
+    
+    // m_outputIncrements stores phase increments.
+
+    ChannelData &cd = *m_channelData[channel];
+    bool gotData = true;
+
+    if (cd.chunkCount >= m_outputIncrements.size()) {
+//        cerr << "WARNING: RubberBandStretcher::Impl::getIncrements:"
+//             << " chunk count " << cd.chunkCount << " >= "
+//             << m_outputIncrements.size() << endl;
+        if (m_outputIncrements.size() == 0) {
+            phaseIncrementRtn = m_increment;
+            shiftIncrementRtn = m_increment;
+            phaseReset = false;
+            return false;
+        } else {
+            cd.chunkCount = m_outputIncrements.size()-1;
+            gotData = false;
+        }
+    }
+    
+    int phaseIncrement = m_outputIncrements[cd.chunkCount];
+    
+    int shiftIncrement = phaseIncrement;
+    if (cd.chunkCount + 1 < m_outputIncrements.size()) {
+        shiftIncrement = m_outputIncrements[cd.chunkCount + 1];
+    }
+    
+    if (phaseIncrement < 0) {
+        phaseIncrement = -phaseIncrement;
+        phaseReset = true;
+    }
+    
+    if (shiftIncrement < 0) {
+        shiftIncrement = -shiftIncrement;
+    }
+    
+    if (shiftIncrement >= int(m_windowSize)) {
+        cerr << "*** ERROR: RubberBandStretcher::Impl::processChunks: shiftIncrement " << shiftIncrement << " >= windowSize " << m_windowSize << " at " << cd.chunkCount << " (of " << m_outputIncrements.size() << ")" << endl;
+        shiftIncrement = m_windowSize;
+    }
+
+    phaseIncrementRtn = phaseIncrement;
+    shiftIncrementRtn = shiftIncrement;
+    if (cd.chunkCount == 0) phaseReset = true; // don't mess with the first chunk
+    return gotData;
+}
+
+void
+RubberBandStretcher::Impl::analyseChunk(size_t channel)
+{
+    size_t i;
+
+    ChannelData &cd = *m_channelData[channel];
+
+    // cd.fltbuf is known to contain m_windowSize samples
+
+    m_window->cut(cd.fltbuf);
+
+    for (i = 0; i < m_windowSize/2; ++i) {
+	cd.dblbuf[i] = cd.fltbuf[i + m_windowSize/2];
+	cd.dblbuf[i + m_windowSize/2] = cd.fltbuf[i];
+    }
+
+    cd.fft->forwardPolar(cd.dblbuf, cd.mag, cd.phase);
+}
+
+double mod(double x, double y) { return x - (y * floor(x / y)); }
+double princarg(double a) { return mod(a + M_PI, -2 * M_PI) + M_PI; }
+
+void
+RubberBandStretcher::Impl::modifyChunk(size_t channel, size_t outputIncrement,
+                                       bool phaseReset)
+{
+    ChannelData &cd = *m_channelData[channel];
+
+    if (phaseReset && m_debugLevel > 1) {
+        cerr << "phase reset: leaving phases unmodified" << endl;
+    }
+
+    size_t count = m_windowSize/2;
+    size_t pfp = 0;
+    double rate = m_stretcher->m_sampleRate;
+
+    if (!(m_options & OptionPhaseIndependent)) {
+
+        cd.freqPeak[0] = 0;
+
+        float freq0 = m_freq0;
+        float freq1 = m_freq1;
+        float freq2 = m_freq2;
+
+        // As the stretch ratio increases, so the frequency thresholds
+        // for phase lamination should increase.  Beyond a ratio of
+        // about 1.5, the threshold should be about 1200Hz; beyond a
+        // ratio of 2, we probably want no lamination to happen at all
+        // by default.  This calculation aims for more or less that.
+        // We only do this if the phase option is OptionPhaseAdaptive
+        // (the default), i.e. not Independent or PeakLocked.
+
+        if (!(m_options & OptionPhasePeakLocked)) {
+            float r = getEffectiveRatio();
+            if (r > 1) {
+                float rf0 = 600 + (600 * ((r-1)*(r-1)*(r-1)*2));
+                float f1ratio = freq1 / freq0;
+                float f2ratio = freq2 / freq0;
+                freq0 = std::max(freq0, rf0);
+                freq1 = freq0 * f1ratio;
+                freq2 = freq0 * f2ratio;
+            }
+        }
+
+        size_t limit0 = lrint((freq0 * m_windowSize) / rate);
+        size_t limit1 = lrint((freq1 * m_windowSize) / rate);
+        size_t limit2 = lrint((freq2 * m_windowSize) / rate);
+
+        size_t range = 0;
+
+        if (limit1 < limit0) limit1 = limit0;
+        if (limit2 < limit1) limit2 = limit1;
+    
+//        cerr << "limit0 = " << limit0 << " limit1 = " << limit1 << " limit2 = " << limit2 << endl;
+
+        int peakCount = 0;
+
+        for (size_t i = 0; i <= count; ++i) {
+
+            double mag = cd.mag[i];
+            bool isPeak = true;
+
+            for (size_t j = 1; j <= range; ++j) {
+
+                if (mag < cd.mag[i-j]) {
+                    isPeak = false;
+                    break;
+                }
+
+                if (mag < cd.mag[i+j]) {
+                    isPeak = false;
+                    break;
+                }
+            }        
+
+            if (isPeak) {
+
+                // i is a peak bin.
+
+                // The previous peak bin was at pfp; make freqPeak entries
+                // from pfp to half-way between pfp and i point at pfp, and
+                // those from the half-way mark to i point at i.
+            
+                size_t halfway = (pfp + i) / 2;
+                if (halfway == pfp) halfway = pfp + 1;
+
+                for (size_t j = pfp + 1; j < halfway; ++j) {
+                    cd.freqPeak[j] = pfp;
+                }
+                for (size_t j = halfway; j <= i; ++j) {
+                    cd.freqPeak[j] = i;
+                }
+
+                pfp = i;
+
+                ++peakCount;
+            }
+
+            if (i == limit0) range = 1;
+            if (i == limit1) range = 2;
+            if (i >= limit2) {
+                range = 3;
+                if (i + range + 1 > count) range = count - i;
+            }
+        }
+
+//        cerr << "peakCount = " << peakCount << endl;
+        
+        cd.freqPeak[count-1] = count-1;
+        cd.freqPeak[count] = count;
+    }
+
+    double peakInPhase = 0.0;
+    double peakOutPhase = 0.0;
+    size_t p, pp;
+
+    for (size_t i = 0; i <= count; ++i) {
+        
+        if (m_options & OptionPhaseIndependent) {
+            p = i;
+            pp = i-1;
+        } else {
+            p = cd.freqPeak[i];
+            pp = cd.freqPeak[i-1];
+        }
+
+        bool resetThis = phaseReset;
+        
+        if (m_options & OptionTransientsMixed) {
+            size_t low = lrint((150 * m_windowSize) / rate);
+            size_t high = lrint((1000 * m_windowSize) / rate);
+            if (resetThis) {
+                if (i > low && i < high) resetThis = false;
+            }
+        }
+
+        if (!resetThis) {
+
+            if (i == 0 || p != pp) {
+	
+                double omega = (2 * M_PI * m_increment * p) / m_windowSize;
+                double expectedPhase = cd.prevPhase[p] + omega;
+                double phaseError = princarg(cd.phase[p] - expectedPhase);
+                double phaseIncrement = (omega + phaseError) / m_increment;
+            
+                double unwrappedPhase = cd.unwrappedPhase[p] +
+                    outputIncrement * phaseIncrement;
+
+                cd.prevPhase[p] = cd.phase[p];
+                cd.phase[p] = unwrappedPhase;
+                cd.unwrappedPhase[p] = unwrappedPhase;
+
+                peakInPhase = cd.prevPhase[p];
+                peakOutPhase = unwrappedPhase;
+            }
+
+            if (i != p) {
+
+                double diffToPeak = peakInPhase - cd.phase[i];
+                double unwrappedPhase = peakOutPhase - diffToPeak;
+                
+                cd.prevPhase[i] = cd.phase[i];
+                cd.phase[i] = unwrappedPhase;
+                cd.unwrappedPhase[i] = unwrappedPhase;
+            }
+
+        } else {
+            cd.prevPhase[i] = cd.phase[i];
+            cd.unwrappedPhase[i] = cd.phase[i];
+        }
+    }
+}
+
+void
+RubberBandStretcher::Impl::synthesiseChunk(size_t channel)
+{
+    ChannelData &cd = *m_channelData[channel];
+
+    cd.fft->inversePolar(cd.mag, cd.phase, cd.dblbuf);
+
+    for (size_t i = 0; i < m_windowSize/2; ++i) {
+        cd.fltbuf[i] = cd.dblbuf[i + m_windowSize/2];
+        cd.fltbuf[i + m_windowSize/2] = cd.dblbuf[i];
+    }
+
+    // our ffts produced unscaled results
+    for (size_t i = 0; i < m_windowSize; ++i) {
+        cd.fltbuf[i] = cd.fltbuf[i] / m_windowSize;
+    }
+
+    m_window->cut(cd.fltbuf);
+
+    for (size_t i = 0; i < m_windowSize; ++i) {
+        cd.accumulator[i] += cd.fltbuf[i];
+    }
+
+    cd.accumulatorFill = m_windowSize;
+
+    float fixed = m_window->getArea() * 1.5;
+
+    for (size_t i = 0; i < m_windowSize; ++i) {
+        float val = m_window->getValue(i);
+        cd.windowAccumulator[i] += val * fixed;
+    }
+}
+
+void
+RubberBandStretcher::Impl::writeChunk(size_t channel, size_t shiftIncrement, bool last)
+{
+    ChannelData &cd = *m_channelData[channel];
+
+    if (m_debugLevel > 2) {
+        cerr << "writeChunk(" << channel << ", " << shiftIncrement << ", " << last << ")" << endl;
+    }
+
+    for (int i = 0; i < shiftIncrement; ++i) {
+        if (cd.windowAccumulator[i] > 0.f) {
+            cd.accumulator[i] /= cd.windowAccumulator[i];
+        }
+    }
+
+    // for exact sample scaling (probably not meaningful if we
+    // were running in RT mode)
+    size_t theoreticalOut = 0;
+    if (cd.inputSize >= 0) {
+        theoreticalOut = lrint(cd.inputSize * m_timeRatio);
+    }
+
+    if (m_pitchScale != 1.0 && cd.resampler) {
+
+        size_t reqSize = int(ceil(shiftIncrement / m_pitchScale));
+        if (reqSize > cd.resamplebufSize) {
+            // This shouldn't normally happen -- the buffer is
+            // supposed to be initialised with enough space in the
+            // first place.  But we retain this check in case the
+            // pitch scale has changed since then, or the stretch
+            // calculator has gone mad, or something.
+            cerr << "WARNING: RubberBandStretcher::Impl::writeChunk: resizing resampler buffer from "
+                      << cd.resamplebufSize << " to " << reqSize << endl;
+            cd.resamplebufSize = reqSize;
+            if (cd.resamplebuf) delete[] cd.resamplebuf;
+            cd.resamplebuf = new float[cd.resamplebufSize];
+        }
+
+
+        size_t outframes = cd.resampler->resample(&cd.accumulator,
+                                                  &cd.resamplebuf,
+                                                  shiftIncrement,
+                                                  1.0 / m_pitchScale,
+                                                  last);
+
+
+        writeOutput(*cd.outbuf, cd.resamplebuf,
+                    outframes, cd.outCount, theoreticalOut);
+
+    } else {
+        writeOutput(*cd.outbuf, cd.accumulator,
+                    shiftIncrement, cd.outCount, theoreticalOut);
+    }
+    
+    for (size_t i = 0; i < m_windowSize - shiftIncrement; ++i) {
+        cd.accumulator[i] = cd.accumulator[i + shiftIncrement];
+    }
+    
+    for (size_t i = m_windowSize - shiftIncrement; i < m_windowSize; ++i) {
+        cd.accumulator[i] = 0.0f;
+    }
+    
+    for (size_t i = 0; i < m_windowSize - shiftIncrement; ++i) {
+        cd.windowAccumulator[i] = cd.windowAccumulator[i + shiftIncrement];
+    }
+    
+    for (size_t i = m_windowSize - shiftIncrement; i < m_windowSize; ++i) {
+        cd.windowAccumulator[i] = 0.0f;
+    }
+    
+    if (cd.accumulatorFill > shiftIncrement) {
+        cd.accumulatorFill -= shiftIncrement;
+    } else {
+        cd.accumulatorFill = 0;
+        if (cd.draining) {
+            if (m_debugLevel > 1) {
+                cerr << "RubberBandStretcher::Impl::processChunks: setting outputComplete to true" << endl;
+            }
+            cd.outputComplete = true;
+        }
+    }
+}
+
+void
+RubberBandStretcher::Impl::writeOutput(RingBuffer<float> &to, float *from, size_t qty, size_t &outCount, size_t theoreticalOut)
+{
+    // In non-RT mode, we don't want to write the first startSkip
+    // samples, because the first chunk is centred on the start of the
+    // output.  In RT mode we didn't apply any pre-padding in
+    // configure(), so we don't want to remove any here.
+
+    size_t startSkip = 0;
+    if (!m_realtime) {
+        startSkip = lrintf((m_windowSize/2) / m_pitchScale);
+    }
+
+    if (outCount > startSkip) {
+        
+        // this is the normal case
+
+        if (theoreticalOut > 0) {
+            if (m_debugLevel > 1) {
+                cerr << "theoreticalOut = " << theoreticalOut
+                     << ", outCount = " << outCount
+                     << ", startSkip = " << startSkip
+                     << ", qty = " << qty << endl;
+            }
+            if (outCount - startSkip <= theoreticalOut &&
+                outCount - startSkip + qty > theoreticalOut) {
+                qty = theoreticalOut - (outCount - startSkip);
+                if (m_debugLevel > 1) {
+                    cerr << "reduce qty to " << qty << endl;
+                }
+            }
+        }
+
+        if (m_debugLevel > 2) {
+            cerr << "writing " << qty << endl;
+        }
+
+        size_t written = to.write(from, qty);
+
+        if (written < qty) {
+            cerr << "WARNING: RubberBandStretcher::Impl::writeOutput: "
+                 << "Buffer overrun on output: wrote " << written
+                 << " of " << qty << " samples" << endl;
+        }
+
+        outCount += written;
+        return;
+    }
+
+    // the rest of this is only used during the first startSkip samples
+
+    if (outCount + qty <= startSkip) {
+        if (m_debugLevel > 1) {
+            cerr << "qty = " << qty << ", startSkip = "
+                 << startSkip << ", outCount = " << outCount
+                 << ", discarding" << endl;
+        }
+        outCount += qty;
+        return;
+    }
+
+    size_t off = startSkip - outCount;
+    if (m_debugLevel > 1) {
+        cerr << "qty = " << qty << ", startSkip = "
+             << startSkip << ", outCount = " << outCount
+             << ", writing " << qty - off
+             << " from start offset " << off << endl;
+    }
+    to.write(from + off, qty - off);
+    outCount += qty;
+}
+
+int
+RubberBandStretcher::Impl::available() const
+{
+    if (m_threaded) {
+        MutexLocker locker(&m_threadSetMutex);
+        if (m_channelData.empty()) return 0;
+    } else {
+        if (m_channelData.empty()) return 0;
+    }
+
+    if (!m_threaded) {
+        for (size_t c = 0; c < m_channels; ++c) {
+            if (m_channelData[c]->inputSize >= 0) {
+//                cerr << "available: m_done true" << endl;
+                if (m_channelData[c]->inbuf->getReadSpace() > 0) {
+//                    cerr << "calling processChunks(" << c << ") from available" << endl;
+                    //!!! do we ever actually do this? if so, this method should not be const
+                    // ^^^ yes, we do sometimes -- e.g. when fed a very short file
+                    bool any = false, last = false;
+                    ((RubberBandStretcher::Impl *)this)->processChunks(c, any, last);
+                }
+            }
+        }
+    }
+
+    size_t min = 0;
+    bool consumed = true;
+    bool haveResamplers = false;
+
+    for (size_t i = 0; i < m_channels; ++i) {
+        size_t availIn = m_channelData[i]->inbuf->getReadSpace();
+        size_t availOut = m_channelData[i]->outbuf->getReadSpace();
+        if (m_debugLevel > 2) {
+            cerr << "available on channel " << i << ": " << availOut << " (waiting: " << availIn << ")" << endl;
+        }
+        if (i == 0 || availOut < min) min = availOut;
+        if (!m_channelData[i]->outputComplete) consumed = false;
+        if (m_channelData[i]->resampler) haveResamplers = true;
+    }
+
+    if (min == 0 && consumed) return -1;
+    if (m_pitchScale == 1.0) return min;
+
+    if (haveResamplers) return min; // resampling has already happened
+    return int(floor(min / m_pitchScale));
+}
+
+size_t
+RubberBandStretcher::Impl::retrieve(float *const *output, size_t samples) const
+{
+    size_t got = samples;
+
+    for (size_t c = 0; c < m_channels; ++c) {
+        size_t gotHere = m_channelData[c]->outbuf->read(output[c], got);
+        if (gotHere < got) {
+            if (c > 0) {
+                if (m_debugLevel > 0) {
+                    cerr << "RubberBandStretcher::Impl::retrieve: WARNING: channel imbalance detected" << endl;
+                }
+            }
+            got = gotHere;
+        }
+    }
+
+    return got;
+}
+
+}
+

libs/rubberband/1.0/src/Thread.cpp

@@ -0,0 +1,530 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "Thread.h"
+
+#include <iostream>
+
+#include <sys/time.h>
+#include <time.h>
+
+//#define DEBUG_THREAD 1
+//#define DEBUG_MUTEX 1
+//#define DEBUG_CONDITION 1
+
+using std::cerr;
+using std::endl;
+using std::string;
+
+namespace RubberBand
+{
+
+#ifdef _WIN32
+
+Thread::Thread() :
+    m_id(0),
+    m_extant(false)
+{
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Created thread object " << this << endl;
+#endif
+}
+
+Thread::~Thread()
+{
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Destroying thread object " << this << ", id " << m_id << endl;
+#endif
+    if (m_extant) {
+        WaitForSingleObject(m_id, INFINITE);
+    }
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Destroyed thread object " << this << endl;
+#endif
+}
+
+void
+Thread::start()
+{
+    m_id = CreateThread(NULL, 0, staticRun, this, 0, 0);
+    if (!m_id) {
+        cerr << "ERROR: thread creation failed" << endl;
+        exit(1);
+    } else {
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Created thread " << m_id << " for thread object " << this << endl;
+#endif
+        m_extant = true;
+    }
+}    
+
+void 
+Thread::wait()
+{
+    if (m_extant) {
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Waiting on thread " << m_id << " for thread object " << this << endl;
+#endif
+        WaitForSingleObject(m_id, INFINITE);
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Waited on thread " << m_id << " for thread object " << this << endl;
+#endif
+        m_extant = false;
+    }
+}
+
+Thread::Id
+Thread::id()
+{
+    return m_id;
+}
+
+bool
+Thread::threadingAvailable()
+{
+    return true;
+}
+
+DWORD
+Thread::staticRun(LPVOID arg)
+{
+    Thread *thread = static_cast<Thread *>(arg);
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: " << (void *)GetCurrentThreadId() << ": Running thread " << thread->m_id << " for thread object " << thread << endl;
+#endif
+    thread->run();
+    return 0;
+}
+
+Mutex::Mutex() :
+    m_locked(false)
+{
+    m_mutex = CreateMutex(NULL, FALSE, NULL);
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Initialised mutex " << &m_mutex << endl;
+#endif
+}
+
+Mutex::~Mutex()
+{
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Destroying mutex " << &m_mutex << endl;
+#endif
+    CloseHandle(m_mutex);
+}
+
+void
+Mutex::lock()
+{
+    if (m_locked) {
+        cerr << "ERROR: Deadlock on mutex " << &m_mutex << endl;
+    }
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Want to lock mutex " << &m_mutex << endl;
+#endif
+    WaitForSingleObject(m_mutex, INFINITE);
+    m_locked = true;
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Locked mutex " << &m_mutex << endl;
+#endif
+}
+
+void
+Mutex::unlock()
+{
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Unlocking mutex " << &m_mutex << endl;
+#endif
+    m_locked = false;
+    ReleaseMutex(m_mutex);
+}
+
+bool
+Mutex::trylock()
+{
+    DWORD result = WaitForSingleObject(m_mutex, 0);
+    if (result == WAIT_TIMEOUT || result == WAIT_FAILED) {
+#ifdef DEBUG_MUTEX
+        cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Mutex " << &m_mutex << " unavailable" << endl;
+#endif
+        return false;
+    } else {
+        m_locked = true;
+#ifdef DEBUG_MUTEX
+        cerr << "MUTEX DEBUG: " << (void *)GetCurrentThreadId() << ": Locked mutex " << &m_mutex << " (from trylock)" << endl;
+#endif
+        return true;
+    }
+}
+
+Condition::Condition(string name) :
+    m_name(name),
+    m_locked(false)
+{
+    m_mutex = CreateMutex(NULL, FALSE, NULL);
+    m_condition = CreateEvent(NULL, FALSE, FALSE, NULL);
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Initialised condition " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+}
+
+Condition::~Condition()
+{
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Destroying condition " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    if (m_locked) ReleaseMutex(m_mutex);
+    CloseHandle(m_condition);
+    CloseHandle(m_mutex);
+}
+
+void
+Condition::lock()
+{
+    if (m_locked) {
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Already locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        return;
+    }
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Want to lock " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    WaitForSingleObject(m_mutex, INFINITE);
+    m_locked = true;
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+}
+
+void
+Condition::unlock()
+{
+    if (!m_locked) {
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Not locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        return;
+    }
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Unlocking " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    m_locked = false;
+    ReleaseMutex(m_mutex);
+}
+
+void 
+Condition::wait(int us)
+{
+    if (!m_locked) lock();
+
+    if (us == 0) {
+
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Waiting on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        SignalObjectAndWait(m_mutex, m_condition, INFINITE, FALSE);
+        WaitForSingleObject(m_mutex, INFINITE);
+
+    } else {
+
+        DWORD ms = us / 1000;
+        if (us > 0 && ms == 0) ms = 1;
+    
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Timed waiting on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        SignalObjectAndWait(m_mutex, m_condition, ms, FALSE);
+        WaitForSingleObject(m_mutex, INFINITE);
+    }
+
+    ReleaseMutex(m_mutex);
+
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Wait done on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    m_locked = false;
+}
+
+void
+Condition::signal()
+{
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)GetCurrentThreadId() << ": Signalling " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    SetEvent(m_condition);
+}
+
+#else /* !_WIN32 */
+
+
+Thread::Thread() :
+    m_id(0),
+    m_extant(false)
+{
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Created thread object " << this << endl;
+#endif
+}
+
+Thread::~Thread()
+{
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Destroying thread object " << this << ", id " << m_id << endl;
+#endif
+    if (m_extant) {
+        pthread_join(m_id, 0);
+    }
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: Destroyed thread object " << this << endl;
+#endif
+}
+
+void
+Thread::start()
+{
+    if (pthread_create(&m_id, 0, staticRun, this)) {
+        cerr << "ERROR: thread creation failed" << endl;
+        exit(1);
+    } else {
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Created thread " << m_id << " for thread object " << this << endl;
+#endif
+        m_extant = true;
+    }
+}    
+
+void 
+Thread::wait()
+{
+    if (m_extant) {
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Waiting on thread " << m_id << " for thread object " << this << endl;
+#endif
+        pthread_join(m_id, 0);
+#ifdef DEBUG_THREAD
+        cerr << "THREAD DEBUG: Waited on thread " << m_id << " for thread object " << this << endl;
+#endif
+        m_extant = false;
+    }
+}
+
+Thread::Id
+Thread::id()
+{
+    return m_id;
+}
+
+bool
+Thread::threadingAvailable()
+{
+    return true;
+}
+
+void *
+Thread::staticRun(void *arg)
+{
+    Thread *thread = static_cast<Thread *>(arg);
+#ifdef DEBUG_THREAD
+    cerr << "THREAD DEBUG: " << (void *)pthread_self() << ": Running thread " << thread->m_id << " for thread object " << thread << endl;
+#endif
+    thread->run();
+    return 0;
+}
+
+Mutex::Mutex() :
+    m_locked(false)
+{
+    pthread_mutex_init(&m_mutex, 0);
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Initialised mutex " << &m_mutex << endl;
+#endif
+}
+
+Mutex::~Mutex()
+{
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Destroying mutex " << &m_mutex << endl;
+#endif
+    pthread_mutex_destroy(&m_mutex);
+}
+
+void
+Mutex::lock()
+{
+    if (m_locked) {
+        cerr << "ERROR: Deadlock on mutex " << &m_mutex << endl;
+    }
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Want to lock mutex " << &m_mutex << endl;
+#endif
+    pthread_mutex_lock(&m_mutex);
+    m_locked = true;
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Locked mutex " << &m_mutex << endl;
+#endif
+}
+
+void
+Mutex::unlock()
+{
+#ifdef DEBUG_MUTEX
+    cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Unlocking mutex " << &m_mutex << endl;
+#endif
+    m_locked = false;
+    pthread_mutex_unlock(&m_mutex);
+}
+
+bool
+Mutex::trylock()
+{
+    if (pthread_mutex_trylock(&m_mutex)) {
+#ifdef DEBUG_MUTEX
+        cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Mutex " << &m_mutex << " unavailable" << endl;
+#endif
+        return false;
+    } else {
+        m_locked = true;
+#ifdef DEBUG_MUTEX
+        cerr << "MUTEX DEBUG: " << (void *)pthread_self() << ": Locked mutex " << &m_mutex << " (from trylock)" << endl;
+#endif
+        return true;
+    }
+}
+
+Condition::Condition(string name) :
+    m_locked(false),
+    m_name(name)
+{
+    pthread_mutex_init(&m_mutex, 0);
+    pthread_cond_init(&m_condition, 0);
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Initialised condition " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+}
+
+Condition::~Condition()
+{
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Destroying condition " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    if (m_locked) pthread_mutex_unlock(&m_mutex);
+    pthread_cond_destroy(&m_condition);
+    pthread_mutex_destroy(&m_mutex);
+}
+
+void
+Condition::lock()
+{
+    if (m_locked) {
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Already locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        return;
+    }
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Want to lock " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    pthread_mutex_lock(&m_mutex);
+    m_locked = true;
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+}
+
+void
+Condition::unlock()
+{
+    if (!m_locked) {
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Not locked " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        return;
+    }
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Unlocking " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    m_locked = false;
+    pthread_mutex_unlock(&m_mutex);
+}
+
+void 
+Condition::wait(int us)
+{
+    if (!m_locked) lock();
+
+    if (us == 0) {
+
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Waiting on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        pthread_cond_wait(&m_condition, &m_mutex);
+
+    } else {
+
+        struct timeval now;
+        gettimeofday(&now, 0);
+
+        now.tv_usec += us;
+        while (now.tv_usec > 1000000) {
+            now.tv_usec -= 1000000;
+            ++now.tv_sec;
+        }
+
+        struct timespec timeout;
+        timeout.tv_sec = now.tv_sec;
+        timeout.tv_nsec = now.tv_usec * 1000;
+    
+#ifdef DEBUG_CONDITION
+        cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Timed waiting on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+        pthread_cond_timedwait(&m_condition, &m_mutex, &timeout);
+    }
+
+    pthread_mutex_unlock(&m_mutex);
+
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Wait done on " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    m_locked = false;
+}
+
+void
+Condition::signal()
+{
+#ifdef DEBUG_CONDITION
+    cerr << "CONDITION DEBUG: " << (void *)pthread_self() << ": Signalling " << &m_condition << " \"" << m_name << "\"" << endl;
+#endif
+    pthread_cond_signal(&m_condition);
+}
+
+#endif /* !_WIN32 */
+
+MutexLocker::MutexLocker(Mutex *mutex) :
+    m_mutex(mutex)
+{
+    if (m_mutex) {
+        m_mutex->lock();
+    }
+}
+
+MutexLocker::~MutexLocker()
+{
+    if (m_mutex) {
+        m_mutex->unlock();
+    }
+}
+
+}
+

libs/rubberband/1.0/src/Thread.h

@@ -0,0 +1,131 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_THREAD_H_
+#define _RUBBERBAND_THREAD_H_
+
+#ifdef _WIN32
+#include <windows.h>
+#else /* !_WIN32 */
+#include <pthread.h>
+#endif /* !_WIN32 */
+
+#include <string>
+
+namespace RubberBand
+{
+
+class Thread
+{
+public:
+#ifdef _WIN32
+    typedef HANDLE Id;
+#else
+    typedef pthread_t Id;
+#endif
+
+    Thread();
+    virtual ~Thread();
+
+    Id id();
+
+    void start();
+    void wait();
+
+    static bool threadingAvailable();
+
+protected:
+    virtual void run() = 0;
+
+private:
+#ifdef _WIN32
+    HANDLE m_id;
+    bool m_extant;
+    static DWORD WINAPI staticRun(LPVOID lpParam);
+#else
+    pthread_t m_id;
+    bool m_extant;
+    static void *staticRun(void *);
+#endif
+};
+
+class Mutex
+{
+public:
+    Mutex();
+    ~Mutex();
+
+    void lock();
+    void unlock();
+    bool trylock();
+
+private:
+#ifdef _WIN32
+    HANDLE m_mutex;
+    bool m_locked;
+#else
+    pthread_mutex_t m_mutex;
+    bool m_locked;
+#endif
+};
+
+class MutexLocker
+{
+public:
+    MutexLocker(Mutex *);
+    ~MutexLocker();
+
+private:
+    Mutex *m_mutex;
+};
+
+class Condition
+{
+public:
+    Condition(std::string name);
+    ~Condition();
+    
+    // To wait on a condition, either simply call wait(), or call
+    // lock() and then wait() (perhaps testing some state in between).
+    // To signal a condition, call signal().
+
+    // Although any thread may signal on a given condition, only one
+    // thread should ever wait on any given condition object --
+    // otherwise there will be a race conditions in the logic that
+    // avoids the thread code having to track whether the condition's
+    // mutex is locked or not.  If that is your requirement, this
+    // Condition wrapper is not for you.
+    void lock();
+    void unlock();
+    void wait(int us = 0);
+
+    void signal();
+    
+private:
+#ifdef _WIN32
+    HANDLE m_mutex;
+    bool m_locked;
+    HANDLE m_condition;
+    std::string m_name;
+#else
+    pthread_mutex_t m_mutex;
+    bool m_locked;
+    pthread_cond_t m_condition;
+    std::string m_name;
+#endif
+};
+
+}
+
+#endif

libs/rubberband/1.0/src/Window.h

@@ -0,0 +1,165 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_WINDOW_H_
+#define _RUBBERBAND_WINDOW_H_
+
+#include <cmath>
+#include <iostream>
+#include <map>
+
+namespace RubberBand {
+
+enum WindowType {
+    RectangularWindow,
+    BartlettWindow,
+    HammingWindow,
+    HanningWindow,
+    BlackmanWindow,
+    GaussianWindow,
+    ParzenWindow,
+    NuttallWindow,
+    BlackmanHarrisWindow
+};
+
+template <typename T>
+class Window
+{
+public:
+    /**
+     * Construct a windower of the given type.
+     */
+    Window(WindowType type, size_t size) : m_type(type), m_size(size) { encache(); }
+    Window(const Window &w) : m_type(w.m_type), m_size(w.m_size) { encache(); }
+    Window &operator=(const Window &w) {
+	if (&w == this) return *this;
+	m_type = w.m_type;
+	m_size = w.m_size;
+	encache();
+	return *this;
+    }
+    virtual ~Window() { delete[] m_cache; }
+    
+    void cut(T *src) const { cut(src, src); }
+    void cut(T *src, T *dst) const {
+	for (size_t i = 0; i < m_size; ++i) dst[i] = src[i] * m_cache[i];
+    }
+
+    T getArea() { return m_area; }
+    T getValue(size_t i) { return m_cache[i]; }
+
+    WindowType getType() const { return m_type; }
+    size_t getSize() const { return m_size; }
+
+protected:
+    WindowType m_type;
+    size_t m_size;
+    T *m_cache;
+    T m_area;
+    
+    void encache();
+    void cosinewin(T *, T, T, T, T);
+};
+
+template <typename T>
+void Window<T>::encache()
+{
+    int n = int(m_size);
+    T *mult = new T[n];
+    int i;
+    for (i = 0; i < n; ++i) mult[i] = 1.0;
+
+    switch (m_type) {
+		
+    case RectangularWindow:
+	for (i = 0; i < n; ++i) {
+	    mult[i] *= 0.5;
+	}
+	break;
+	    
+    case BartlettWindow:
+	for (i = 0; i < n/2; ++i) {
+	    mult[i] *= (i / T(n/2));
+	    mult[i + n/2] *= (1.0 - (i / T(n/2)));
+	}
+	break;
+	    
+    case HammingWindow:
+        cosinewin(mult, 0.54, 0.46, 0.0, 0.0);
+	break;
+	    
+    case HanningWindow:
+        cosinewin(mult, 0.50, 0.50, 0.0, 0.0);
+	break;
+	    
+    case BlackmanWindow:
+        cosinewin(mult, 0.42, 0.50, 0.08, 0.0);
+	break;
+	    
+    case GaussianWindow:
+	for (i = 0; i < n; ++i) {
+            mult[i] *= pow(2, - pow((i - (n-1)/2.0) / ((n-1)/2.0 / 3), 2));
+	}
+	break;
+	    
+    case ParzenWindow:
+    {
+        int N = n-1;
+        for (i = 0; i < N/4; ++i) {
+            T m = 2 * pow(1.0 - (T(N)/2 - i) / (T(N)/2), 3);
+            mult[i] *= m;
+            mult[N-i] *= m;
+        }
+        for (i = N/4; i <= N/2; ++i) {
+            int wn = i - N/2;
+            T m = 1.0 - 6 * pow(wn / (T(N)/2), 2) * (1.0 - abs(wn) / (T(N)/2));
+            mult[i] *= m;
+            mult[N-i] *= m;
+        }            
+        break;
+    }
+
+    case NuttallWindow:
+        cosinewin(mult, 0.3635819, 0.4891775, 0.1365995, 0.0106411);
+	break;
+
+    case BlackmanHarrisWindow:
+        cosinewin(mult, 0.35875, 0.48829, 0.14128, 0.01168);
+        break;
+    }
+	
+    m_cache = mult;
+
+    m_area = 0;
+    for (int i = 0; i < n; ++i) {
+        m_area += m_cache[i];
+    }
+    m_area /= n;
+}
+
+template <typename T>
+void Window<T>::cosinewin(T *mult, T a0, T a1, T a2, T a3)
+{
+    int n = int(m_size);
+    for (int i = 0; i < n; ++i) {
+        mult[i] *= (a0
+                    - a1 * cos(2 * M_PI * i / n)
+                    + a2 * cos(4 * M_PI * i / n)
+                    - a3 * cos(6 * M_PI * i / n));
+    }
+}
+
+}
+
+#endif

libs/rubberband/1.0/src/ladspa/RubberBandPitchShifter.cpp

@@ -0,0 +1,408 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "RubberBandPitchShifter.h"
+
+#include "RubberBandStretcher.h"
+
+#include <iostream>
+#include <cmath>
+
+using namespace RubberBand;
+
+const char *const
+RubberBandPitchShifter::portNamesMono[PortCountMono] =
+{
+    "_latency",
+    "Cents",
+    "Semitones",
+    "Octaves",
+    "Crispness",
+    "Input",
+    "Output"
+};
+
+const char *const
+RubberBandPitchShifter::portNamesStereo[PortCountStereo] =
+{
+    "_latency",
+    "Cents",
+    "Semitones",
+    "Octaves",
+    "Crispness",
+    "Input L",
+    "Output L",
+    "Input R",
+    "Output R"
+};
+
+const LADSPA_PortDescriptor 
+RubberBandPitchShifter::portsMono[PortCountMono] =
+{
+    LADSPA_PORT_OUTPUT | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_AUDIO,
+    LADSPA_PORT_OUTPUT | LADSPA_PORT_AUDIO
+};
+
+const LADSPA_PortDescriptor 
+RubberBandPitchShifter::portsStereo[PortCountStereo] =
+{
+    LADSPA_PORT_OUTPUT | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_CONTROL,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_AUDIO,
+    LADSPA_PORT_OUTPUT | LADSPA_PORT_AUDIO,
+    LADSPA_PORT_INPUT  | LADSPA_PORT_AUDIO,
+    LADSPA_PORT_OUTPUT | LADSPA_PORT_AUDIO
+};
+
+const LADSPA_PortRangeHint 
+RubberBandPitchShifter::hintsMono[PortCountMono] =
+{
+    { 0, 0, 0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE,
+      -100.0, 100.0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+      -12.0, 12.0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+      -4.0, 4.0 },
+    { LADSPA_HINT_DEFAULT_MAXIMUM |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+       0.0, 3.0 },
+    { 0, 0, 0 },
+    { 0, 0, 0 }
+};
+
+const LADSPA_PortRangeHint 
+RubberBandPitchShifter::hintsStereo[PortCountStereo] =
+{
+    { 0, 0, 0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE,
+      -100.0, 100.0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+      -12.0, 12.0 },
+    { LADSPA_HINT_DEFAULT_0 |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+      -4.0, 4.0 },
+    { LADSPA_HINT_DEFAULT_MAXIMUM |
+      LADSPA_HINT_BOUNDED_BELOW |
+      LADSPA_HINT_BOUNDED_ABOVE |
+      LADSPA_HINT_INTEGER,
+       0.0, 3.0 },
+    { 0, 0, 0 },
+    { 0, 0, 0 },
+    { 0, 0, 0 },
+    { 0, 0, 0 }
+};
+
+const LADSPA_Properties
+RubberBandPitchShifter::properties = LADSPA_PROPERTY_HARD_RT_CAPABLE;
+
+const LADSPA_Descriptor 
+RubberBandPitchShifter::ladspaDescriptorMono =
+{
+    2979, // "Unique" ID
+    "rubberband-pitchshifter-mono", // Label
+    properties,
+    "Rubber Band Mono Pitch Shifter", // Name
+    "Chris Cannam",
+    "GPL",
+    PortCountMono,
+    portsMono,
+    portNamesMono,
+    hintsMono,
+    0, // Implementation data
+    instantiate,
+    connectPort,
+    activate,
+    run,
+    0, // Run adding
+    0, // Set run adding gain
+    deactivate,
+    cleanup
+};
+
+const LADSPA_Descriptor 
+RubberBandPitchShifter::ladspaDescriptorStereo =
+{
+    9792, // "Unique" ID
+    "rubberband-pitchshifter-stereo", // Label
+    properties,
+    "Rubber Band Stereo Pitch Shifter", // Name
+    "Chris Cannam",
+    "GPL",
+    PortCountStereo,
+    portsStereo,
+    portNamesStereo,
+    hintsStereo,
+    0, // Implementation data
+    instantiate,
+    connectPort,
+    activate,
+    run,
+    0, // Run adding
+    0, // Set run adding gain
+    deactivate,
+    cleanup
+};
+
+const LADSPA_Descriptor *
+RubberBandPitchShifter::getDescriptor(unsigned long index)
+{
+    if (index == 0) return &ladspaDescriptorMono;
+    if (index == 1) return &ladspaDescriptorStereo;
+    else return 0;
+}
+
+RubberBandPitchShifter::RubberBandPitchShifter(int sampleRate, size_t channels) :
+    m_latency(0),
+    m_cents(0),
+    m_semitones(0),
+    m_octaves(0),
+    m_crispness(0),
+    m_ratio(1.0),
+    m_prevRatio(1.0),
+    m_currentCrispness(-1),
+    m_extraLatency(8192), //!!! this should be at least the maximum possible displacement from linear at input rates, divided by the pitch scale factor.  It could be very large
+    m_stretcher(new RubberBandStretcher
+                (sampleRate, channels,
+                 RubberBandStretcher::OptionProcessRealTime)),
+    m_sampleRate(sampleRate),
+    m_channels(channels)
+{
+    for (size_t c = 0; c < m_channels; ++c) {
+        m_input[c] = 0;
+        m_output[c] = 0;
+        //!!! size must be at least max process size plus m_extraLatency:
+        m_outputBuffer[c] = new RingBuffer<float>(8092); //!!!
+        m_outputBuffer[c]->zero(m_extraLatency);
+        //!!! size must be at least max process size:
+        m_scratch[c] = new float[16384];//!!!
+    }
+}
+
+RubberBandPitchShifter::~RubberBandPitchShifter()
+{
+    delete m_stretcher;
+    for (size_t c = 0; c < m_channels; ++c) {
+        delete m_outputBuffer[c];
+        delete[] m_scratch[c];
+    }
+}
+    
+LADSPA_Handle
+RubberBandPitchShifter::instantiate(const LADSPA_Descriptor *desc, unsigned long rate)
+{
+    if (desc->PortCount == ladspaDescriptorMono.PortCount) {
+        return new RubberBandPitchShifter(rate, 1);
+    } else if (desc->PortCount == ladspaDescriptorStereo.PortCount) {
+        return new RubberBandPitchShifter(rate, 2);
+    }
+    return 0;
+}
+
+void
+RubberBandPitchShifter::connectPort(LADSPA_Handle handle,
+				    unsigned long port, LADSPA_Data *location)
+{
+    RubberBandPitchShifter *shifter = (RubberBandPitchShifter *)handle;
+
+    float **ports[PortCountStereo] = {
+        &shifter->m_latency,
+	&shifter->m_cents,
+	&shifter->m_semitones,
+	&shifter->m_octaves,
+        &shifter->m_crispness,
+	&shifter->m_input[0],
+	&shifter->m_output[0],
+	&shifter->m_input[1],
+	&shifter->m_output[1]
+    };
+
+    *ports[port] = (float *)location;
+}
+
+void
+RubberBandPitchShifter::activate(LADSPA_Handle handle)
+{
+    RubberBandPitchShifter *shifter = (RubberBandPitchShifter *)handle;
+    shifter->updateRatio();
+    shifter->m_prevRatio = shifter->m_ratio;
+    shifter->m_stretcher->reset();
+    shifter->m_stretcher->setPitchScale(shifter->m_ratio);
+}
+
+void
+RubberBandPitchShifter::run(LADSPA_Handle handle, unsigned long samples)
+{
+    RubberBandPitchShifter *shifter = (RubberBandPitchShifter *)handle;
+    shifter->runImpl(samples);
+}
+
+void
+RubberBandPitchShifter::updateRatio()
+{
+    double oct = *m_octaves;
+    oct += *m_semitones / 12;
+    oct += *m_cents / 1200;
+    m_ratio = pow(2.0, oct);
+}
+
+void
+RubberBandPitchShifter::updateCrispness()
+{
+    if (!m_crispness) return;
+    
+    int c = lrintf(*m_crispness);
+    if (c == m_currentCrispness) return;
+    if (c < 0 || c > 3) return;
+    RubberBandStretcher *s = m_stretcher;
+
+    switch (c) {
+    case 0:
+        s->setPhaseOption(RubberBandStretcher::OptionPhaseIndependent);
+        s->setTransientsOption(RubberBandStretcher::OptionTransientsSmooth);
+        break;
+    case 1:
+        s->setPhaseOption(RubberBandStretcher::OptionPhaseAdaptive);
+        s->setTransientsOption(RubberBandStretcher::OptionTransientsSmooth);
+        break;
+    case 2:
+        s->setPhaseOption(RubberBandStretcher::OptionPhaseAdaptive);
+        s->setTransientsOption(RubberBandStretcher::OptionTransientsMixed);
+        break;
+    case 3:
+        s->setPhaseOption(RubberBandStretcher::OptionPhaseAdaptive);
+        s->setTransientsOption(RubberBandStretcher::OptionTransientsCrisp);
+        break;
+    }
+
+    m_currentCrispness = c;
+}
+
+void
+RubberBandPitchShifter::runImpl(unsigned long insamples)
+{
+//    std::cerr << "RubberBandPitchShifter::runImpl(" << insamples << ")" << std::endl;
+
+    updateRatio();
+    if (m_ratio != m_prevRatio) {
+        m_stretcher->setPitchScale(m_ratio);
+        m_prevRatio = m_ratio;
+    }
+
+    if (m_latency) {
+        *m_latency = m_stretcher->getLatency() + m_extraLatency;
+//        std::cerr << "latency = " << *m_latency << std::endl;
+    }
+
+    updateCrispness();
+
+    int samples = insamples;
+    int processed = 0;
+    size_t outTotal = 0;
+
+    float *ptrs[2];
+
+    // We have to break up the input into chunks like this because
+    // insamples could be arbitrarily large 
+
+    while (processed < samples) {
+
+        //!!! size_t:
+        int toCauseProcessing = m_stretcher->getSamplesRequired();
+//        std::cout << "to-cause: " << toCauseProcessing << ", remain = " << samples - processed;
+        int inchunk = std::min(samples - processed, toCauseProcessing);
+        for (size_t c = 0; c < m_channels; ++c) {
+            ptrs[c] = &(m_input[c][processed]);
+        }
+        m_stretcher->process(ptrs, inchunk, false);
+        processed += inchunk;
+
+        int avail = m_stretcher->available();
+        int writable = m_outputBuffer[0]->getWriteSpace();
+        int outchunk = std::min(avail, writable);
+        size_t actual = m_stretcher->retrieve(m_scratch, outchunk);
+        outTotal += actual;
+
+//        std::cout << ", avail: " << avail << ", outchunk = " << outchunk;
+//        if (actual != outchunk) std::cout << " (" << actual << ")";
+//        std::cout << std::endl;
+
+        outchunk = actual;
+
+        for (size_t c = 0; c < m_channels; ++c) {
+            if (int(m_outputBuffer[c]->getWriteSpace()) < outchunk) {
+                std::cerr << "RubberBandPitchShifter::runImpl: buffer overrun: chunk = " << outchunk << ", space = " << m_outputBuffer[c]->getWriteSpace() << std::endl;
+            }                
+            m_outputBuffer[c]->write(m_scratch[c], outchunk);
+        }
+    }
+
+//    std::cout << "processed = " << processed << " in, " << outTotal << " out" << ", fill = " << m_outputBuffer[0]->getReadSpace() << " of " << m_outputBuffer[0]->getSize() << std::endl;
+    
+    for (size_t c = 0; c < m_channels; ++c) {
+        int avail = m_outputBuffer[c]->getReadSpace();
+//        std::cout << "avail: " << avail << std::endl;
+        if (avail < samples && c == 0) {
+            std::cerr << "RubberBandPitchShifter::runImpl: buffer underrun: required = " << samples << ", available = " << avail << std::endl;
+        }
+        int chunk = std::min(avail, samples);
+//        std::cout << "out chunk: " << chunk << std::endl;
+        m_outputBuffer[c]->read(m_output[c], chunk);
+    }
+
+    static int minr = -1;
+    int avail = m_outputBuffer[0]->getReadSpace();
+    if (minr == -1 || (avail >= 0 && avail < minr)) {
+        std::cerr << "RubberBandPitchShifter::runImpl: new min remaining " << avail << " from " << minr << std::endl;
+        minr = avail;
+    }
+}
+
+void
+RubberBandPitchShifter::deactivate(LADSPA_Handle handle)
+{
+    activate(handle); // both functions just reset the plugin
+}
+
+void
+RubberBandPitchShifter::cleanup(LADSPA_Handle handle)
+{
+    delete (RubberBandPitchShifter *)handle;
+}
+

libs/rubberband/1.0/src/ladspa/RubberBandPitchShifter.h

@@ -0,0 +1,95 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_PITCH_SHIFTER_H_
+#define _RUBBERBAND_PITCH_SHIFTER_H_
+
+#include "ladspa.h"
+
+#include "RingBuffer.h"
+
+namespace RubberBand {
+class RubberBandStretcher;
+}
+
+class RubberBandPitchShifter
+{
+public:
+    static const LADSPA_Descriptor *getDescriptor(unsigned long index);
+    
+protected:
+    RubberBandPitchShifter(int sampleRate, size_t channels);
+    ~RubberBandPitchShifter();
+
+    enum {
+        LatencyPort      = 0,
+	OctavesPort      = 1,
+	SemitonesPort    = 2,
+	CentsPort        = 3,
+        CrispnessPort    = 4,
+        InputPort1       = 5,
+        OutputPort1      = 6,
+        PortCountMono    = OutputPort1 + 1,
+        InputPort2       = 7,
+        OutputPort2      = 8,
+        PortCountStereo  = OutputPort2 + 1
+    };
+
+    static const char *const portNamesMono[PortCountMono];
+    static const LADSPA_PortDescriptor portsMono[PortCountMono];
+    static const LADSPA_PortRangeHint hintsMono[PortCountMono];
+
+    static const char *const portNamesStereo[PortCountStereo];
+    static const LADSPA_PortDescriptor portsStereo[PortCountStereo];
+    static const LADSPA_PortRangeHint hintsStereo[PortCountStereo];
+
+    static const LADSPA_Properties properties;
+
+    static const LADSPA_Descriptor ladspaDescriptorMono;
+    static const LADSPA_Descriptor ladspaDescriptorStereo;
+
+    static LADSPA_Handle instantiate(const LADSPA_Descriptor *, unsigned long);
+    static void connectPort(LADSPA_Handle, unsigned long, LADSPA_Data *);
+    static void activate(LADSPA_Handle);
+    static void run(LADSPA_Handle, unsigned long);
+    static void deactivate(LADSPA_Handle);
+    static void cleanup(LADSPA_Handle);
+
+    void runImpl(unsigned long);
+    void updateRatio();
+    void updateCrispness();
+
+    float *m_input[2];
+    float *m_output[2];
+    float *m_latency;
+    float *m_cents;
+    float *m_semitones;
+    float *m_octaves;
+    float *m_crispness;
+    double m_ratio;
+    double m_prevRatio;
+    int m_currentCrispness;
+
+    size_t m_extraLatency;
+
+    RubberBand::RubberBandStretcher *m_stretcher;
+    RubberBand::RingBuffer<float> *m_outputBuffer[2];
+    float *m_scratch[2];
+
+    int m_sampleRate;
+    size_t m_channels;
+};
+
+
+#endif

libs/rubberband/1.0/src/ladspa/ladspa-rubberband.cat

@@ -0,0 +1,2 @@
+ladspa:ladspa-rubberband:rubberband-pitchshifter-mono::Frequency > Pitch shifters
+ladspa:ladspa-rubberband:rubberband-pitchshifter-stereo::Frequency > Pitch shifters

libs/rubberband/1.0/src/ladspa/libmain.cpp

@@ -0,0 +1,26 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "RubberBandPitchShifter.h"
+
+#include <stdio.h>
+
+extern "C" {
+
+const LADSPA_Descriptor *ladspa_descriptor(unsigned long index)
+{
+    return RubberBandPitchShifter::getDescriptor(index);
+}
+
+}

libs/rubberband/1.0/src/main.cpp

@@ -0,0 +1,475 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "RubberBandStretcher.h"
+
+#include <iostream>
+#include <sndfile.h>
+#include <cmath>
+#include <sys/time.h>
+#include <time.h>
+#include "sysutils.h"
+
+#include <getopt.h>
+
+// for import and export of FFTW wisdom
+#include <fftw3.h>
+
+using namespace std;
+using namespace RubberBand;
+
+#ifdef _WIN32
+using RubberBand::gettimeofday;
+using RubberBand::usleep;
+#endif
+
+int main(int argc, char **argv)
+{
+    int c;
+
+    double ratio = 1.0;
+    double pitchshift = 1.0;
+    double frequencyshift = 1.0;
+    int debug = 0;
+    bool realtime = false;
+    bool precise = false;
+    int threading = 0;
+    bool peaklock = true;
+    bool longwin = false;
+    bool shortwin = false;
+    bool softening = true;
+    int crispness = -1;
+    bool help = false;
+    bool quiet = false;
+
+    bool haveRatio = false;
+
+    enum {
+        NoTransients,
+        BandLimitedTransients,
+        Transients
+    } transients = Transients;
+
+    float fthresh0 = -1.f;
+    float fthresh1 = -1.f;
+    float fthresh2 = -1.f;
+
+    while (1) {
+        int optionIndex = 0;
+
+        static struct option longOpts[] = {
+            { "help",          0, 0, 'h' },
+            { "time",          1, 0, 't' },
+            { "tempo",         1, 0, 'T' },
+            { "pitch",         1, 0, 'p' },
+            { "frequency",     1, 0, 'f' },
+            { "crisp",         1, 0, 'c' },
+            { "crispness",     1, 0, 'c' },
+            { "debug",         1, 0, 'd' },
+            { "realtime",      0, 0, 'R' },
+            { "precise",       0, 0, 'P' },
+            { "no-threads",    0, 0, '0' },
+            { "no-transients", 0, 0, '1' },
+            { "no-peaklock",   0, 0, '2' },
+            { "window-long",   0, 0, '3' },
+            { "window-short",  0, 0, '4' },
+            { "thresh0",       1, 0, '5' },
+            { "thresh1",       1, 0, '6' },
+            { "thresh2",       1, 0, '7' },
+            { "bl-transients", 0, 0, '8' },
+            { "no-softening",  0, 0, '9' },
+            { "threads",       0, 0, '@' },
+            { "quiet",         0, 0, 'q' },
+            { 0, 0, 0 }
+        };
+
+        c = getopt_long(argc, argv, "t:p:d:RPc:f:qh", longOpts, &optionIndex);
+        if (c == -1) break;
+
+        switch (c) {
+        case 'h': help = true; break;
+        case 't': ratio *= atof(optarg); haveRatio = true; break;
+        case 'T': { double m = atof(optarg); if (m != 0.0) ratio /= m; }; haveRatio = true; break;
+        case 'p': pitchshift = atof(optarg); haveRatio = true; break;
+        case 'f': frequencyshift = atof(optarg); haveRatio = true; break;
+        case 'd': debug = atoi(optarg); break;
+        case 'R': realtime = true; break;
+        case 'P': precise = true; break;
+        case '0': threading = 1; break;
+        case '@': threading = 2; break;
+        case '1': transients = NoTransients; break;
+        case '2': peaklock = false; break;
+        case '3': longwin = true; break;
+        case '4': shortwin = true; break;
+        case '5': fthresh0 = atof(optarg); break;
+        case '6': fthresh1 = atof(optarg); break;
+        case '7': fthresh2 = atof(optarg); break;
+        case '8': transients = BandLimitedTransients; break;
+        case '9': softening = false; break;
+        case 'c': crispness = atoi(optarg); break;
+        case 'q': quiet = true; break;
+        default:  help = true; break;
+        }
+    }
+
+    if (help || !haveRatio || optind + 2 != argc) {
+        cerr << endl;
+	cerr << "Rubber Band" << endl;
+        cerr << "An audio time-stretching and pitch-shifting library and utility program." << endl;
+	cerr << "Copyright 2007 Chris Cannam.  Distributed under the GNU General Public License." << endl;
+        cerr << endl;
+	cerr << "   Usage: " << argv[0] << " [options] <infile.wav> <outfile.wav>" << endl;
+        cerr << endl;
+        cerr << "You must specify at least one of the following time and pitch ratio options." << endl;
+        cerr << endl;
+        cerr << "  -t<X>, --time <X>       Stretch to X times original duration, or" << endl;
+        cerr << "  -T<X>, --tempo <X>      Change tempo by multiple X (equivalent to --time 1/X)" << endl;
+        cerr << endl;
+        cerr << "  -p<X>, --pitch <X>      Raise pitch by X semitones, or" << endl;
+        cerr << "  -f<X>, --frequency <X>  Change frequency by multiple X" << endl;
+        cerr << endl;
+        cerr << "The following option provides a simple way to adjust the sound.  See below" << endl;
+        cerr << "for more details." << endl;
+        cerr << endl;
+        cerr << "  -c<N>, --crisp <N>      Crispness (N = 0,1,2,3,4,5); default 4 (see below)" << endl;
+        cerr << endl;
+        cerr << "The remaining options fine-tune the processing mode and stretch algorithm." << endl;
+        cerr << "These are mostly included for test purposes; the default settings and standard" << endl;
+        cerr << "crispness parameter are intended to provide the best sounding set of options" << endl;
+        cerr << "for most situations." << endl;
+        cerr << endl;
+        cerr << "  -P,    --precise        Aim for minimal time distortion (implied by -R)" << endl;
+        cerr << "  -R,    --realtime       Select realtime mode (implies -P --no-threads)" << endl;
+        cerr << "         --no-threads     No extra threads regardless of CPU and channel count" << endl;
+        cerr << "         --threads        Assume multi-CPU even if only one CPU is identified" << endl;
+        cerr << "         --no-transients  Disable phase resynchronisation at transients" << endl;
+        cerr << "         --bl-transients  Band-limit phase resync to extreme frequencies" << endl;
+        cerr << "         --no-peaklock    Disable phase locking to peak frequencies" << endl;
+        cerr << "         --no-softening   Disable large-ratio softening of phase locking" << endl;
+        cerr << "         --window-long    Use longer processing window (actual size may vary)" << endl;
+        cerr << "         --window-short   Use shorter processing window" << endl;
+        cerr << "         --thresh<N> <F>  Set internal freq threshold N (N = 0,1,2) to F Hz" << endl;
+        cerr << endl;
+        cerr << "  -d<N>, --debug <N>      Select debug level (N = 0,1,2,3); default 0, full 3" << endl;
+        cerr << "                          (N.B. debug level 3 includes audible ticks in output)" << endl;
+        cerr << "  -q,    --quiet          Suppress progress output" << endl;
+        cerr << endl;
+        cerr << "  -h,    --help           Show this help" << endl;
+        cerr << endl;
+        cerr << "\"Crispness\" levels:" << endl;
+        cerr << "  -c 0   equivalent to --no-transients --no-peaklock --window-long" << endl;
+        cerr << "  -c 1   equivalent to --no-transients --no-peaklock" << endl;
+        cerr << "  -c 2   equivalent to --no-transients" << endl;
+        cerr << "  -c 3   equivalent to --bl-transients" << endl;
+        cerr << "  -c 4   default processing options" << endl;
+        cerr << "  -c 5   equivalent to --no-peaklock --window-short (may be suitable for drums)" << endl;
+        cerr << endl;
+	return 2;
+    }
+
+    switch (crispness) {
+    case -1: crispness = 4; break;
+    case 0: transients = NoTransients; peaklock = false; longwin = true; shortwin = false; break;
+    case 1: transients = NoTransients; peaklock = false; longwin = false; shortwin = false; break;
+    case 2: transients = NoTransients; peaklock = true; longwin = false; shortwin = false; break;
+    case 3: transients = BandLimitedTransients; peaklock = true; longwin = false; shortwin = false; break;
+    case 4: transients = Transients; peaklock = true; longwin = false; shortwin = false; break;
+    case 5: transients = Transients; peaklock = false; longwin = false; shortwin = true; break;
+    };
+
+    if (!quiet) {
+        cerr << "Using crispness level: " << crispness << " (";
+        switch (crispness) {
+        case 0: cerr << "Mushy"; break;
+        case 1: cerr << "Smooth"; break;
+        case 2: cerr << "Balanced multitimbral mixture"; break;
+        case 3: cerr << "Unpitched percussion with stable notes"; break;
+        case 4: cerr << "Crisp monophonic instrumental"; break;
+        case 5: cerr << "Unpitched solo percussion"; break;
+        }
+        cerr << ")" << endl;
+    }
+
+    char *fileName = strdup(argv[optind++]);
+    char *fileNameOut = strdup(argv[optind++]);
+    
+    SNDFILE *sndfile;
+    SNDFILE *sndfileOut;
+    SF_INFO sfinfo;
+    SF_INFO sfinfoOut;
+    memset(&sfinfo, 0, sizeof(SF_INFO));
+
+    sndfile = sf_open(fileName, SFM_READ, &sfinfo);
+    if (!sndfile) {
+	cerr << "ERROR: Failed to open input file \"" << fileName << "\": "
+	     << sf_strerror(sndfile) << endl;
+	return 1;
+    }
+
+    sfinfoOut.channels = sfinfo.channels;
+    sfinfoOut.format = sfinfo.format;
+    sfinfoOut.frames = int(sfinfo.frames * ratio + 0.1);
+    sfinfoOut.samplerate = sfinfo.samplerate;
+    sfinfoOut.sections = sfinfo.sections;
+    sfinfoOut.seekable = sfinfo.seekable;
+
+    sndfileOut = sf_open(fileNameOut, SFM_WRITE, &sfinfoOut) ;
+    if (!sndfileOut) {
+	cerr << "ERROR: Failed to open output file \"" << fileName << "\" for writing: "
+	     << sf_strerror(sndfile) << endl;
+	return 1;
+    }
+    
+    int ibs = 1024;
+    size_t channels = sfinfo.channels;
+
+    RubberBandStretcher::Options options = 0;
+    if (realtime)    options |= RubberBandStretcher::OptionProcessRealTime;
+    if (precise)     options |= RubberBandStretcher::OptionStretchPrecise;
+    if (!peaklock)   options |= RubberBandStretcher::OptionPhaseIndependent;
+    if (!softening)  options |= RubberBandStretcher::OptionPhasePeakLocked;
+    if (longwin)     options |= RubberBandStretcher::OptionWindowLong;
+    if (shortwin)    options |= RubberBandStretcher::OptionWindowShort;
+
+    switch (threading) {
+    case 0:
+        options |= RubberBandStretcher::OptionThreadingAuto;
+        break;
+    case 1:
+        options |= RubberBandStretcher::OptionThreadingNever;
+        break;
+    case 2:
+        options |= RubberBandStretcher::OptionThreadingAlways;
+        break;
+    }
+
+    switch (transients) {
+    case NoTransients:
+        options |= RubberBandStretcher::OptionTransientsSmooth;
+        break;
+    case BandLimitedTransients:
+        options |= RubberBandStretcher::OptionTransientsMixed;
+        break;
+    case Transients:
+        options |= RubberBandStretcher::OptionTransientsCrisp;
+        break;
+    }
+
+    if (pitchshift != 1.0) {
+        frequencyshift *= pow(2.0, pitchshift / 12);
+    }
+
+#ifdef _WIN32
+    RubberBand::
+#endif
+    timeval tv;
+    (void)gettimeofday(&tv, 0);
+
+    RubberBandStretcher::setDefaultDebugLevel(debug);
+
+    RubberBandStretcher ts(sfinfo.samplerate, channels, options,
+                           ratio, frequencyshift);
+
+    ts.setExpectedInputDuration(sfinfo.frames);
+
+    float *fbuf = new float[channels * ibs];
+    float **ibuf = new float *[channels];
+    for (size_t i = 0; i < channels; ++i) ibuf[i] = new float[ibs];
+
+    int frame = 0;
+    int percent = 0;
+
+    sf_seek(sndfile, 0, SEEK_SET);
+
+    if (!realtime) {
+
+        if (!quiet) {
+            cerr << "Pass 1: Studying..." << endl;
+        }
+
+        while (frame < sfinfo.frames) {
+
+            int count = -1;
+
+            if ((count = sf_readf_float(sndfile, fbuf, ibs)) <= 0) break;
+        
+            for (size_t c = 0; c < channels; ++c) {
+                for (int i = 0; i < count; ++i) {
+                    float value = fbuf[i * channels + c];
+                    ibuf[c][i] = value;
+                }
+            }
+
+            bool final = (frame + ibs >= sfinfo.frames);
+
+            ts.study(ibuf, count, final);
+
+            int p = int((double(frame) * 100.0) / sfinfo.frames);
+            if (p > percent || frame == 0) {
+                percent = p;
+                if (!quiet) {
+                    cerr << "\r" << percent << "% ";
+                }
+            }
+
+            frame += ibs;
+        }
+
+        if (!quiet) {
+            cerr << "\rCalculating profile..." << endl;
+        }
+
+        sf_seek(sndfile, 0, SEEK_SET);
+    }
+
+    frame = 0;
+    percent = 0;
+    
+    size_t countIn = 0, countOut = 0;
+
+    while (frame < sfinfo.frames) {
+
+        int count = -1;
+
+	if ((count = sf_readf_float(sndfile, fbuf, ibs)) < 0) break;
+        
+        countIn += count;
+
+        for (size_t c = 0; c < channels; ++c) {
+            for (int i = 0; i < count; ++i) {
+                float value = fbuf[i * channels + c];
+                ibuf[c][i] = value;
+            }
+        }
+
+        bool final = (frame + ibs >= sfinfo.frames);
+
+        ts.process(ibuf, count, final);
+
+        int avail = ts.available();
+        if (debug > 1) cerr << "available = " << avail << endl;
+
+        if (avail > 0) {
+            float **obf = new float *[channels];
+            for (size_t i = 0; i < channels; ++i) {
+                obf[i] = new float[avail];
+            }
+            ts.retrieve(obf, avail);
+            countOut += avail;
+            float *fobf = new float[channels * avail];
+            for (size_t c = 0; c < channels; ++c) {
+                for (int i = 0; i < avail; ++i) {
+                    float value = obf[c][i];
+                    if (value > 1.f) value = 1.f;
+                    if (value < -1.f) value = -1.f;
+                    fobf[i * channels + c] = value;
+                }
+            }
+//            cout << "fobf mean: ";
+//    double d = 0;
+//    for (int i = 0; i < avail; ++i) {
+//        d += fobf[i];
+//    }
+//    d /= avail;
+//    cout << d << endl;
+            sf_writef_float(sndfileOut, fobf, avail);
+            delete[] fobf;
+            for (size_t i = 0; i < channels; ++i) {
+                delete[] obf[i];
+            }
+            delete[] obf;
+        }
+
+        if (frame == 0 && !realtime && !quiet) {
+            cerr << "Pass 2: Processing..." << endl;
+        }
+
+	int p = int((double(frame) * 100.0) / sfinfo.frames);
+	if (p > percent || frame == 0) {
+	    percent = p;
+            if (!quiet) {
+                cerr << "\r" << percent << "% ";
+            }
+	}
+
+        frame += ibs;
+    }
+
+    if (!quiet) {
+        cerr << "\r    " << endl;
+    }
+    int avail;
+
+    while ((avail = ts.available()) >= 0) {
+
+        if (debug > 1) {
+            cerr << "(completing) available = " << avail << endl;
+        }
+
+        if (avail > 0) {
+            float **obf = new float *[channels];
+            for (size_t i = 0; i < channels; ++i) {
+                obf[i] = new float[avail];
+            }
+            ts.retrieve(obf, avail);
+            countOut += avail;
+            float *fobf = new float[channels * avail];
+            for (size_t c = 0; c < channels; ++c) {
+                for (int i = 0; i < avail; ++i) {
+                    float value = obf[c][i];
+                    if (value > 1.f) value = 1.f;
+                    if (value < -1.f) value = -1.f;
+                    fobf[i * channels + c] = value;
+                }
+            }
+
+            sf_writef_float(sndfileOut, fobf, avail);
+            delete[] fobf;
+            for (size_t i = 0; i < channels; ++i) {
+                delete[] obf[i];
+            }
+            delete[] obf;
+        } else {
+            usleep(10000);
+        }
+    }
+
+    sf_close(sndfile);
+    sf_close(sndfileOut);
+
+    if (!quiet) {
+
+        cerr << "in: " << countIn << ", out: " << countOut << ", ratio: " << float(countOut)/float(countIn) << ", ideal output: " << lrint(countIn * ratio) << ", error: " << abs(lrint(countIn * ratio) - int(countOut)) << endl;
+
+#ifdef _WIN32
+        RubberBand::
+#endif
+        timeval etv;
+        (void)gettimeofday(&etv, 0);
+        
+        etv.tv_sec -= tv.tv_sec;
+        if (etv.tv_usec < tv.tv_usec) {
+            etv.tv_usec += 1000000;
+            etv.tv_sec -= 1;
+        }
+        etv.tv_usec -= tv.tv_usec;
+        
+        double sec = double(etv.tv_sec) + (double(etv.tv_usec) / 1000000.0);
+        cerr << "elapsed time: " << sec << " sec, in frames/sec: " << countIn/sec << ", out frames/sec: " << countOut/sec << endl;
+    }
+
+    return 0;
+}
+
+

libs/rubberband/1.0/src/sysutils.cpp

@@ -0,0 +1,106 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "sysutils.h"
+
+#ifdef _WIN32
+#include <windows.h>
+#else /* !_WIN32 */
+#ifdef __APPLE__
+#include <sys/sysctl.h>
+#else /* !__APPLE__, !_WIN32 */
+#include <stdio.h>
+#include <string.h>
+#endif /* !__APPLE__, !_WIN32 */
+#endif /* !_WIN32 */
+
+#include <iostream>
+
+namespace RubberBand {
+
+bool
+system_is_multiprocessor()
+{
+    static bool tested = false, mp = false;
+
+    if (tested) return mp;
+    int count = 0;
+
+#ifdef _WIN32
+
+    SYSTEM_INFO sysinfo;
+    GetSystemInfo(&sysinfo);
+    count = sysinfo.dwNumberOfProcessors;
+
+#else /* !_WIN32 */
+#ifdef __APPLE__
+    
+    size_t sz = sizeof(count);
+    if (sysctlbyname("hw.ncpu", &count, &sz, NULL, 0)) {
+        mp = false;
+    } else {
+        mp = (count > 1);
+    }
+
+#else /* !__APPLE__, !_WIN32 */
+    
+    //...
+
+    FILE *cpuinfo = fopen("/proc/cpuinfo", "r");
+    if (!cpuinfo) return false;
+
+    char buf[256];
+    while (!feof(cpuinfo)) {
+        fgets(buf, 256, cpuinfo);
+        if (!strncmp(buf, "processor", 9)) {
+            ++count;
+        }
+        if (count > 1) break;
+    }
+
+    fclose(cpuinfo);
+
+#endif /* !__APPLE__, !_WIN32 */
+#endif /* !_WIN32 */
+
+    mp = (count > 1);
+    tested = true;
+    return mp;
+}
+
+#ifdef _WIN32
+
+void gettimeofday(struct timeval *tv, void *tz)
+{
+    union { 
+	long long ns100;  
+	FILETIME ft; 
+    } now; 
+    
+    ::GetSystemTimeAsFileTime(&now.ft); 
+    tv->tv_usec = (long)((now.ns100 / 10LL) % 1000000LL); 
+    tv->tv_sec = (long)((now.ns100 - 116444736000000000LL) / 10000000LL); 
+}
+
+void usleep(unsigned long usec)
+{
+    ::Sleep(usec == 0 ? 0 : usec < 1000 ? 1 : usec / 1000);
+}
+
+#endif
+
+}
+
+
+

libs/rubberband/1.0/src/sysutils.h

@@ -0,0 +1,30 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_SYSINFO_H_
+#define _RUBBERBAND_SYSINFO_H_
+
+namespace RubberBand {
+
+extern bool system_is_multiprocessor();
+
+#ifdef _WIN32
+struct timeval { long tv_sec; long tv_usec; };
+void gettimeofday(struct timeval *p, void *tz);
+void usleep(unsigned long);
+#endif
+
+}
+
+#endif

libs/rubberband/1.0/src/vamp/RubberBandVampPlugin.cpp

@@ -0,0 +1,647 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include "RubberBandVampPlugin.h"
+
+#include "StretchCalculator.h"
+
+#include <cmath>
+
+using std::string;
+using std::vector;
+using std::cerr;
+using std::endl;
+
+class RubberBandVampPlugin::Impl
+{
+public:
+    size_t m_stepSize;
+    size_t m_blockSize;
+    size_t m_sampleRate;
+
+    float m_timeRatio;
+    float m_pitchRatio;
+
+    bool m_realtime;
+    bool m_elasticTiming;
+    int m_transientMode;
+    bool m_phaseIndependent;
+    int m_windowLength;
+
+    RubberBand::RubberBandStretcher *m_stretcher;
+
+    int m_incrementsOutput;
+    int m_aggregateIncrementsOutput;
+    int m_divergenceOutput;
+    int m_phaseResetDfOutput;
+    int m_smoothedPhaseResetDfOutput;
+    int m_phaseResetPointsOutput;
+    int m_timeSyncPointsOutput;
+
+    size_t m_counter;
+    size_t m_accumulatedIncrement;
+
+    float **m_outputDump;
+
+    FeatureSet processOffline(const float *const *inputBuffers,
+                              Vamp::RealTime timestamp);
+
+    FeatureSet getRemainingFeaturesOffline();
+
+    FeatureSet processRealTime(const float *const *inputBuffers,
+                               Vamp::RealTime timestamp);
+
+    FeatureSet getRemainingFeaturesRealTime();
+
+    FeatureSet createFeatures(size_t inputIncrement,
+                              std::vector<int> &outputIncrements,
+                              std::vector<float> &phaseResetDf,
+                              std::vector<int> &exactPoints,
+                              std::vector<float> &smoothedDf,
+                              size_t baseCount,
+                              bool includeFinal);
+};
+
+
+RubberBandVampPlugin::RubberBandVampPlugin(float inputSampleRate) :
+    Plugin(inputSampleRate)
+{
+    m_d = new Impl();
+    m_d->m_stepSize = 0;
+    m_d->m_timeRatio = 1.f;
+    m_d->m_pitchRatio = 1.f;
+    m_d->m_realtime = false;
+    m_d->m_elasticTiming = true;
+    m_d->m_transientMode = 0;
+    m_d->m_phaseIndependent = false;
+    m_d->m_windowLength = 0;
+    m_d->m_stretcher = 0;
+    m_d->m_sampleRate = lrintf(m_inputSampleRate);
+}
+
+RubberBandVampPlugin::~RubberBandVampPlugin()
+{
+    if (m_d->m_outputDump) {
+        for (size_t i = 0; i < m_d->m_stretcher->getChannelCount(); ++i) {
+            delete[] m_d->m_outputDump[i];
+        }
+        delete[] m_d->m_outputDump;
+    }
+    delete m_d->m_stretcher;
+    delete m_d;
+}
+
+string
+RubberBandVampPlugin::getIdentifier() const
+{
+    return "rubberband";
+}
+
+string
+RubberBandVampPlugin::getName() const
+{
+    return "Rubber Band Timestretch Analysis";
+}
+
+string
+RubberBandVampPlugin::getDescription() const
+{
+    return "Carry out analysis phases of time stretcher process";
+}
+
+string
+RubberBandVampPlugin::getMaker() const
+{
+    return "Rubber Band"; ///!!!
+}
+
+int
+RubberBandVampPlugin::getPluginVersion() const
+{
+    return 1;
+}
+
+string
+RubberBandVampPlugin::getCopyright() const
+{
+    return "";//!!!
+}
+
+RubberBandVampPlugin::OutputList
+RubberBandVampPlugin::getOutputDescriptors() const
+{
+    OutputList list;
+
+    size_t rate = 0;
+    if (m_d->m_stretcher) {
+        rate = lrintf(m_inputSampleRate / m_d->m_stretcher->getInputIncrement());
+    }
+
+    OutputDescriptor d;
+    d.identifier = "increments";
+    d.name = "Output Increments";
+    d.description = "Output time increment for each input step";
+    d.unit = "samples";
+    d.hasFixedBinCount = true;
+    d.binCount = 1;
+    d.hasKnownExtents = false;
+    d.isQuantized = true;
+    d.quantizeStep = 1.0;
+    d.sampleType = OutputDescriptor::VariableSampleRate;
+    d.sampleRate = rate;
+    m_d->m_incrementsOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "aggregate_increments";
+    d.name = "Accumulated Output Increments";
+    d.description = "Accumulated output time increments";
+    d.sampleRate = 0;
+    m_d->m_aggregateIncrementsOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "divergence";
+    d.name = "Divergence from Linear";
+    d.description = "Difference between actual output time and the output time for a theoretical linear stretch";
+    d.isQuantized = false;
+    d.sampleRate = 0;
+    m_d->m_divergenceOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "phaseresetdf";
+    d.name = "Phase Reset Detection Function";
+    d.description = "Curve whose peaks are used to identify transients for phase reset points";
+    d.unit = "";
+    d.sampleRate = rate;
+    m_d->m_phaseResetDfOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "smoothedphaseresetdf";
+    d.name = "Smoothed Phase Reset Detection Function";
+    d.description = "Phase reset curve smoothed for peak picking";
+    d.unit = "";
+    m_d->m_smoothedPhaseResetDfOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "phaseresetpoints";
+    d.name = "Phase Reset Points";
+    d.description = "Points estimated as transients at which phase reset occurs";
+    d.unit = "";
+    d.hasFixedBinCount = true;
+    d.binCount = 0;
+    d.hasKnownExtents = false;
+    d.isQuantized = false;
+    d.sampleRate = 0;
+    m_d->m_phaseResetPointsOutput = list.size();
+    list.push_back(d);
+
+    d.identifier = "timesyncpoints";
+    d.name = "Time Sync Points";
+    d.description = "Salient points which stretcher aims to place with strictly correct timing";
+    d.unit = "";
+    d.hasFixedBinCount = true;
+    d.binCount = 0;
+    d.hasKnownExtents = false;
+    d.isQuantized = false;
+    d.sampleRate = 0;
+    m_d->m_timeSyncPointsOutput = list.size();
+    list.push_back(d);
+
+    return list;
+}
+
+RubberBandVampPlugin::ParameterList
+RubberBandVampPlugin::getParameterDescriptors() const
+{
+    ParameterList list;
+
+    ParameterDescriptor d;
+    d.identifier = "timeratio";
+    d.name = "Time Ratio";
+    d.description = "Ratio to modify overall duration by";
+    d.unit = "%";
+    d.minValue = 1;
+    d.maxValue = 500;
+    d.defaultValue = 100;
+    d.isQuantized = false;
+    list.push_back(d);
+
+    d.identifier = "pitchratio";
+    d.name = "Pitch Scale Ratio";
+    d.description = "Frequency ratio to modify pitch by";
+    d.unit = "%";
+    d.minValue = 1;
+    d.maxValue = 500;
+    d.defaultValue = 100;
+    d.isQuantized = false;
+    list.push_back(d);
+
+    d.identifier = "mode";
+    d.name = "Processing Mode";
+    d.description = ""; //!!!
+    d.unit = "";
+    d.minValue = 0;
+    d.maxValue = 1;
+    d.defaultValue = 0;
+    d.isQuantized = true;
+    d.quantizeStep = 1;
+    d.valueNames.clear();
+    d.valueNames.push_back("Offline");
+    d.valueNames.push_back("Real Time");
+    list.push_back(d);
+
+    d.identifier = "stretchtype";
+    d.name = "Stretch Flexibility";
+    d.description = ""; //!!!
+    d.unit = "";
+    d.minValue = 0;
+    d.maxValue = 1;
+    d.defaultValue = 0;
+    d.isQuantized = true;
+    d.quantizeStep = 1;
+    d.valueNames.clear();
+    d.valueNames.push_back("Elastic");
+    d.valueNames.push_back("Precise");
+    list.push_back(d);
+
+    d.identifier = "transientmode";
+    d.name = "Transient Handling";
+    d.description = ""; //!!!
+    d.unit = "";
+    d.minValue = 0;
+    d.maxValue = 2;
+    d.defaultValue = 0;
+    d.isQuantized = true;
+    d.quantizeStep = 1;
+    d.valueNames.clear();
+    d.valueNames.push_back("Mixed");
+    d.valueNames.push_back("Smooth");
+    d.valueNames.push_back("Crisp");
+    list.push_back(d);
+
+    d.identifier = "phasemode";
+    d.name = "Phase Handling";
+    d.description = ""; //!!!
+    d.unit = "";
+    d.minValue = 0;
+    d.maxValue = 1;
+    d.defaultValue = 0;
+    d.isQuantized = true;
+    d.quantizeStep = 1;
+    d.valueNames.clear();
+    d.valueNames.push_back("Peak Locked");
+    d.valueNames.push_back("Independent");
+    list.push_back(d);
+
+    d.identifier = "windowmode";
+    d.name = "Window Length";
+    d.description = ""; //!!!
+    d.unit = "";
+    d.minValue = 0;
+    d.maxValue = 2;
+    d.defaultValue = 0;
+    d.isQuantized = true;
+    d.quantizeStep = 1;
+    d.valueNames.clear();
+    d.valueNames.push_back("Standard");
+    d.valueNames.push_back("Short");
+    d.valueNames.push_back("Long");
+    list.push_back(d);
+
+    return list;
+}
+
+float
+RubberBandVampPlugin::getParameter(std::string id) const
+{
+    if (id == "timeratio") return m_d->m_timeRatio * 100.f;
+    if (id == "pitchratio") return m_d->m_pitchRatio * 100.f;
+    if (id == "mode") return m_d->m_realtime ? 1 : 0;
+    if (id == "stretchtype") return m_d->m_elasticTiming ? 0 : 1;
+    if (id == "transientmode") return m_d->m_transientMode;
+    if (id == "phasemode") return m_d->m_phaseIndependent ? 1 : 0;
+    if (id == "windowmode") return m_d->m_windowLength;
+    return 0.f;
+}
+
+void
+RubberBandVampPlugin::setParameter(std::string id, float value)
+{
+    if (id == "timeratio") {
+        m_d->m_timeRatio = value / 100;
+    } else if (id == "pitchratio") {
+        m_d->m_pitchRatio = value / 100;
+    } else {
+        bool set = (value > 0.5);
+        if (id == "mode") m_d->m_realtime = set;
+        else if (id == "stretchtype") m_d->m_elasticTiming = !set;
+        else if (id == "transientmode") m_d->m_transientMode = int(value + 0.5);
+        else if (id == "phasemode") m_d->m_phaseIndependent = set;
+        else if (id == "windowmode") m_d->m_windowLength = int(value + 0.5);
+    }
+}
+
+bool
+RubberBandVampPlugin::initialise(size_t channels, size_t stepSize, size_t blockSize)
+{
+    if (channels < getMinChannelCount() ||
+	channels > getMaxChannelCount()) return false;
+
+    m_d->m_stepSize = std::min(stepSize, blockSize);
+    m_d->m_blockSize = stepSize;
+
+    RubberBand::RubberBandStretcher::Options options = 0;
+
+    if (m_d->m_realtime)
+         options |= RubberBand::RubberBandStretcher::OptionProcessRealTime;
+    else options |= RubberBand::RubberBandStretcher::OptionProcessOffline;
+
+    if (m_d->m_elasticTiming)
+         options |= RubberBand::RubberBandStretcher::OptionStretchElastic;
+    else options |= RubberBand::RubberBandStretcher::OptionStretchPrecise;
+ 
+    if (m_d->m_transientMode == 0) 
+         options |= RubberBand::RubberBandStretcher::OptionTransientsMixed;
+    else if (m_d->m_transientMode == 1) 
+         options |= RubberBand::RubberBandStretcher::OptionTransientsSmooth;
+    else options |= RubberBand::RubberBandStretcher::OptionTransientsCrisp;
+
+    if (m_d->m_phaseIndependent) 
+         options |= RubberBand::RubberBandStretcher::OptionPhaseIndependent;
+    else options |= RubberBand::RubberBandStretcher::OptionPhasePeakLocked;
+
+    if (m_d->m_windowLength == 0)
+         options |= RubberBand::RubberBandStretcher::OptionWindowStandard;
+    else if (m_d->m_windowLength == 1)
+         options |= RubberBand::RubberBandStretcher::OptionWindowShort;
+    else options |= RubberBand::RubberBandStretcher::OptionWindowLong;
+
+    delete m_d->m_stretcher;
+    m_d->m_stretcher = new RubberBand::RubberBandStretcher
+        (m_d->m_sampleRate, channels, options);
+    m_d->m_stretcher->setDebugLevel(1);
+    m_d->m_stretcher->setTimeRatio(m_d->m_timeRatio);
+    m_d->m_stretcher->setPitchScale(m_d->m_pitchRatio);
+
+    m_d->m_counter = 0;
+    m_d->m_accumulatedIncrement = 0;
+
+    m_d->m_outputDump = 0;
+
+    return true;
+}
+
+void
+RubberBandVampPlugin::reset()
+{
+//    delete m_stretcher;  //!!! or just if (m_stretcher) m_stretcher->reset();
+//    m_stretcher = new RubberBand::RubberBandStretcher(lrintf(m_inputSampleRate), channels);
+    if (m_d->m_stretcher) m_d->m_stretcher->reset();
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::process(const float *const *inputBuffers,
+                              Vamp::RealTime timestamp)
+{
+    if (m_d->m_realtime) {
+        return m_d->processRealTime(inputBuffers, timestamp);
+    } else {
+        return m_d->processOffline(inputBuffers, timestamp);
+    }        
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::getRemainingFeatures()
+{
+    if (m_d->m_realtime) {
+        return m_d->getRemainingFeaturesRealTime();
+    } else {
+        return m_d->getRemainingFeaturesOffline();
+    }
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::Impl::processOffline(const float *const *inputBuffers,
+                                           Vamp::RealTime timestamp)
+{
+    if (!m_stretcher) {
+	cerr << "ERROR: RubberBandVampPlugin::processOffline: "
+	     << "RubberBandVampPlugin has not been initialised"
+	     << endl;
+	return FeatureSet();
+    }
+
+    m_stretcher->study(inputBuffers, m_blockSize, false);
+    return FeatureSet();
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::Impl::getRemainingFeaturesOffline()
+{
+    m_stretcher->study(0, 0, true);
+
+    m_stretcher->calculateStretch();
+
+    int rate = m_sampleRate;
+
+    RubberBand::StretchCalculator sc(rate,
+                                     m_stretcher->getInputIncrement(),
+                                     true);
+
+    size_t inputIncrement = m_stretcher->getInputIncrement();
+    std::vector<int> outputIncrements = m_stretcher->getOutputIncrements();
+    std::vector<float> phaseResetDf = m_stretcher->getPhaseResetCurve();
+    std::vector<int> peaks = m_stretcher->getExactTimePoints();
+    std::vector<float> smoothedDf = sc.smoothDF(phaseResetDf);
+
+    FeatureSet features = createFeatures
+        (inputIncrement, outputIncrements, phaseResetDf, peaks, smoothedDf,
+         0, true);
+
+    return features;
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::Impl::processRealTime(const float *const *inputBuffers,
+                                            Vamp::RealTime timestamp)
+{
+    // This function is not in any way a real-time function (i.e. it
+    // has no requirement to be RT safe); it simply operates the
+    // stretcher in RT mode.
+
+    if (!m_stretcher) {
+	cerr << "ERROR: RubberBandVampPlugin::processRealTime: "
+	     << "RubberBandVampPlugin has not been initialised"
+	     << endl;
+	return FeatureSet();
+    }
+
+    m_stretcher->process(inputBuffers, m_blockSize, false);
+    
+    size_t inputIncrement = m_stretcher->getInputIncrement();
+    std::vector<int> outputIncrements = m_stretcher->getOutputIncrements();
+    std::vector<float> phaseResetDf = m_stretcher->getPhaseResetCurve();
+    std::vector<float> smoothedDf; // not meaningful in RT mode
+    std::vector<int> dummyPoints;
+    FeatureSet features = createFeatures
+        (inputIncrement, outputIncrements, phaseResetDf, dummyPoints, smoothedDf, 
+         m_counter, false);
+    m_counter += outputIncrements.size();
+
+    int available = 0;
+    while ((available = m_stretcher->available()) > 0) {
+        if (!m_outputDump) {
+            m_outputDump = new float *[m_stretcher->getChannelCount()];
+            for (size_t i = 0; i < m_stretcher->getChannelCount(); ++i) {
+                m_outputDump[i] = new float[m_blockSize];
+            }
+        }
+        m_stretcher->retrieve(m_outputDump,
+                              std::min(int(m_blockSize), available));
+    }
+
+    return features;
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::Impl::getRemainingFeaturesRealTime()
+{
+    return FeatureSet();
+}
+
+RubberBandVampPlugin::FeatureSet
+RubberBandVampPlugin::Impl::createFeatures(size_t inputIncrement,
+                                           std::vector<int> &outputIncrements,
+                                           std::vector<float> &phaseResetDf,
+                                           std::vector<int> &exactPoints,
+                                           std::vector<float> &smoothedDf,
+                                           size_t baseCount,
+                                           bool includeFinal)
+{
+    size_t actual = m_accumulatedIncrement;
+
+    double overallRatio = m_timeRatio * m_pitchRatio;
+
+    char label[200];
+
+    FeatureSet features;
+
+    int rate = m_sampleRate;
+
+    size_t epi = 0;
+
+    for (size_t i = 0; i < outputIncrements.size(); ++i) {
+
+        size_t frame = (baseCount + i) * inputIncrement;
+
+        int oi = outputIncrements[i];
+        bool hard = false;
+        bool soft = false;
+
+        if (oi < 0) {
+            oi = -oi;
+            hard = true;
+        }
+
+        if (epi < exactPoints.size() && int(i) == exactPoints[epi]) {
+            soft = true;
+            ++epi;
+        }
+
+        double linear = (frame * overallRatio);
+
+        Vamp::RealTime t = Vamp::RealTime::frame2RealTime(frame, rate);
+
+        Feature feature;
+        feature.hasTimestamp = true;
+        feature.timestamp = t;
+        feature.values.push_back(oi);
+        feature.label = Vamp::RealTime::frame2RealTime(oi, rate).toText();
+        features[m_incrementsOutput].push_back(feature);
+
+        feature.values.clear();
+        feature.values.push_back(actual);
+        feature.label = Vamp::RealTime::frame2RealTime(actual, rate).toText();
+        features[m_aggregateIncrementsOutput].push_back(feature);
+
+        feature.values.clear();
+        feature.values.push_back(actual - linear);
+
+        sprintf(label, "expected %ld, actual %ld, difference %ld (%s ms)",
+                long(linear), long(actual), long(actual - linear),
+                // frame2RealTime expects an integer frame number,
+                // hence our multiplication factor
+                (Vamp::RealTime::frame2RealTime
+                 (lrintf((actual - linear) * 1000), rate) / 1000)
+                .toText().c_str());
+        feature.label = label;
+
+        features[m_divergenceOutput].push_back(feature);
+        actual += oi;
+        
+        char buf[30];
+
+        if (i < phaseResetDf.size()) {
+            feature.values.clear();
+            feature.values.push_back(phaseResetDf[i]);
+            sprintf(buf, "%d", baseCount + i);
+            feature.label = buf;
+            features[m_phaseResetDfOutput].push_back(feature);
+        }
+
+        if (i < smoothedDf.size()) {
+            feature.values.clear();
+            feature.values.push_back(smoothedDf[i]);
+            features[m_smoothedPhaseResetDfOutput].push_back(feature);
+        }
+
+        if (hard) {
+            feature.values.clear();
+            feature.label = "Phase Reset";
+            features[m_phaseResetPointsOutput].push_back(feature);
+        }
+
+        if (hard || soft) {
+            feature.values.clear();
+            feature.label = "Time Sync";
+            features[m_timeSyncPointsOutput].push_back(feature);
+        }            
+    }
+
+    if (includeFinal) {
+        Vamp::RealTime t = Vamp::RealTime::frame2RealTime
+            (inputIncrement * (baseCount + outputIncrements.size()), rate);
+        Feature feature;
+        feature.hasTimestamp = true;
+        feature.timestamp = t;
+        feature.label = Vamp::RealTime::frame2RealTime(actual, rate).toText();
+        feature.values.clear();
+        feature.values.push_back(actual);
+        features[m_aggregateIncrementsOutput].push_back(feature);
+
+        float linear = ((baseCount + outputIncrements.size())
+                        * inputIncrement * overallRatio);
+        feature.values.clear();
+        feature.values.push_back(actual - linear);
+        feature.label =  // see earlier comment
+            (Vamp::RealTime::frame2RealTime //!!! update this as earlier label
+             (lrintf((actual - linear) * 1000), rate) / 1000)
+            .toText();
+        features[m_divergenceOutput].push_back(feature);
+    }
+
+    m_accumulatedIncrement = actual;
+
+    return features;
+}
+

libs/rubberband/1.0/src/vamp/RubberBandVampPlugin.h

@@ -0,0 +1,56 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#ifndef _RUBBERBAND_VAMP_PLUGIN_H_
+#define _RUBBERBAND_VAMP_PLUGIN_H_
+
+#include <vamp-sdk/Plugin.h>
+
+#include "RubberBandStretcher.h"
+
+class RubberBandVampPlugin : public Vamp::Plugin
+{
+public:
+    RubberBandVampPlugin(float inputSampleRate);
+    virtual ~RubberBandVampPlugin();
+
+    bool initialise(size_t channels, size_t stepSize, size_t blockSize);
+    void reset();
+
+    InputDomain getInputDomain() const { return TimeDomain; }
+
+    std::string getIdentifier() const;
+    std::string getName() const;
+    std::string getDescription() const;
+    std::string getMaker() const;
+    int getPluginVersion() const;
+    std::string getCopyright() const;
+
+    ParameterList getParameterDescriptors() const;
+    float getParameter(std::string id) const;
+    void setParameter(std::string id, float value);
+
+    OutputList getOutputDescriptors() const;
+
+    FeatureSet process(const float *const *inputBuffers,
+                       Vamp::RealTime timestamp);
+
+    FeatureSet getRemainingFeatures();
+
+protected:
+    class Impl;
+    Impl *m_d;
+};
+
+#endif

libs/rubberband/1.0/src/vamp/libmain.cpp

@@ -0,0 +1,32 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+/*
+    Rubber Band
+    An audio time-stretching and pitch-shifting library.
+    Copyright 2007 Chris Cannam.
+    
+    This program is free software; you can redistribute it and/or
+    modify it under the terms of the GNU General Public License as
+    published by the Free Software Foundation; either version 2 of the
+    License, or (at your option) any later version.  See the file
+    COPYING included with this distribution for more information.
+*/
+
+#include <vamp/vamp.h>
+#include <vamp-sdk/PluginAdapter.h>
+
+#include "RubberBandVampPlugin.h"
+
+static Vamp::PluginAdapter<RubberBandVampPlugin> rubberBandAdapter;
+
+const VampPluginDescriptor *vampGetPluginDescriptor(unsigned int version,
+                                                    unsigned int index)
+{
+    if (version < 1) return 0;
+
+    switch (index) {
+    case  0: return rubberBandAdapter.getDescriptor();
+    default: return 0;
+    }
+}
+

libs/rubberband/1.0/src/vamp/vamp-rubberband.cat

@@ -0,0 +1 @@
+vamp:vamp-rubberband:rubberband::Time > Timestretch Analysis