diff --git a/README.html b/README.html index 71ef801..d5d9903 100644 --- a/README.html +++ b/README.html @@ -1,767 +1,797 @@ - - - - - - - - SoundTouch library README - - - -
-

SoundTouch audio processing library v1.5.1pre -

-

SoundTouch library Copyright © Olli -Parviainen 2001-2009

-
-

1. Introduction

-

SoundTouch is an open-source audio -processing library that allows changing the sound tempo, pitch -and playback rate parameters independently from each other, i.e.:

- -

1.1 Contact information

-

Author email: oparviai 'at' iki.fi

-

SoundTouch WWW page: http://www.surina.net/soundtouch

-
-

2. Compiling SoundTouch

-

Before compiling, notice that you can choose the sample data format -if it's desirable to use floating point sample -data instead of 16bit integers. See section "sample data format" -for more information.

-

2.1. Building in Microsoft Windows

-

Project files for Microsoft Visual C++ 6.0 and Visual C++ .NET are -supplied with the source code package. 

-

Please notice that SoundTouch -library uses processor-specific optimizations for Pentium III and AMD -processors. Visual Studio .NET and later versions supports the required -instructions by default, but Visual Studio 6.0 requires a processor pack upgrade -to be installed in order to support these optimizations. The processor pack upgrade can be downloaded from -Microsoft site at this URL:

-

http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx

-

If the above URL is unavailable or removed, go -to http://msdn.microsoft.com -and perform a search with keywords "processor pack".

-

To build the binaries with Visual C++ -compiler, either run "make-win.bat" script, or open the -appropriate project files in source code directories with Visual -Studio. The final executable will appear under the "SoundTouch\bin" -directory. If using the Visual Studio IDE instead of the make-win.bat script, directories bin and -lib may need to be created manually to the SoundTouch -package root for the final executables. The make-win.bat script -creates these directories automatically. -

-

2.2. Building in Gnu platforms

-

The SoundTouch library can be compiled in -practically any platform supporting GNU compiler (GCC) tools. -SoundTouch have been tested with gcc version 3.3.4., but it -shouldn't be very specific about the gcc version. Assembler-level -performance optimizations for GNU platform are currently available in -x86 platforms only, they are automatically disabled and replaced with -standard C routines in other processor platforms.

-

To build and install the binaries, run the -following commands in the SoundTouch/ directory:

- - - - - - - - - - - - - - - -
- 
./configure  -
- 		-

Configures the SoundTouch package for the local -environment.

-
- 
make         -
- 		-

Builds the SoundTouch library & -SoundStretch utility.

-
- 
make install -
- 		-

Installs the SoundTouch & BPM libraries -to /usr/local/lib and SoundStretch utility to /usr/local/bin. -Please notice that 'root' privileges may be required to install the -binaries to the destination locations.

-
-

2.2.1 Required GNU tools 

-

Bash shell, GNU C++ compiler, libtool, autoconf and automake tools are required -for compiling -the SoundTouch library. These are usually included with the GNU/Linux distribution, but if -not, install these packages first. For example, in Ubuntu Linux these can be acquired and -installed with the following command:

-
sudo apt-get install automake autoconf libtool build-essential
-

2.2.2 Problems with GCC compiler compatibility

-

At the release time the SoundTouch package has been tested to compile in -GNU/Linux platform. However, in past it's happened that new gcc versions aren't -necessarily compatible with the assembler settings used in the optimized -routines. If you have problems getting the -SoundTouch library compiled, try the workaround of disabling the optimizations -by editing the file "include/STTypes.h" and removing the following -definition there:

-
- 
#define ALLOW_OPTIMIZATIONS 1
-
-

2.2.3 Problems with configure script or build process 

-

Incompatibilities between various GNU toolchain versions may cause errors when running the "configure" script or building the source -codes, if your GNU tool versions are not compatible with the versions used for -preparing the SoundTouch kit. 

-

To resolve the issue, regenerate the configure scripts with your local tool -set by running -the "./bootstrap" script included in the SoundTouch source code -kit. After that, run the configure script and make as usually.

-

2.2.4 Compiler issues with non-x86 processors

-

SoundTouch library works also on non-x86 processors.

-

However, in case that you get compiler errors when trying to compile for non-Intel processor, edit the file -"source\SoundTouch\Makefile.am" and remove the "-msse2" -flag on the AM_CXXFLAGS line:

-
AM_CXXFLAGS=-O3 -fcheck-new -I../../include    # Note: -msse2 flag removed!
-

After that, run "./bootstrap" script, and then run configure -and make again.

-
-

3. About implementation & Usage tips

-

3.1. Supported sample data formats

-

The sample data format can be chosen -between 16bit signed integer and 32bit floating point values, the -default is 32bit floating point.

- -

-In Windows environment, the sample data format is chosen -in file "STTypes.h" by choosing one of the following -defines:

- -

-In GNU environment, the floating sample format is used by default, but -integer sample format can be chosen by giving the -following switch to the configure script: -

-
./configure --enable-integer-samples
-
- -

The sample data can have either single (mono) -or double (stereo) audio channel. Stereo data is interleaved so -that every other data value is for left channel and every second -for right channel. Notice that while it'd be possible in theory -to process stereo sound as two separate mono channels, this isn't -recommended because processing the channels separately would -result in losing the phase coherency between the channels, which -consequently would ruin the stereo effect.

-

Sample rates between 8000-48000H are -supported.

-

3.2. Processing latency

-

The processing and latency constraints of -the SoundTouch library are:

- -

3.3. About algorithms

-

SoundTouch provides three seemingly -independent effects: tempo, pitch and playback rate control. -These three controls are implemented as combination of two primary -effects, sample rate transposing and time-stretching.

-

Sample rate transposing affects -both the audio stream duration and pitch. It's implemented simply -by converting the original audio sample stream to the  desired -duration by interpolating from the original audio samples. In SoundTouch, linear interpolation with anti-alias filtering is -used. Theoretically a higher-order interpolation provide better -result than 1st order linear interpolation, but in audio -application linear interpolation together with anti-alias -filtering performs subjectively about as well as higher-order -filtering would.

-

Time-stretching means changing -the audio stream duration without affecting it's pitch. SoundTouch -uses WSOLA-like time-stretching routines that operate in the time -domain. Compared to sample rate transposing, time-stretching is a -much heavier operation and also requires a longer processing -"window" of sound samples used by the -processing algorithm, thus increasing the algorithm input/output -latency. Typical i/o latency for the SoundTouch -time-stretch algorithm is around 100 ms.

-

Sample rate transposing and time-stretching -are then used together to produce the tempo, pitch and rate -controls:

- -

3.4 Tuning the algorithm parameters

-

The time-stretch algorithm has few -parameters that can be tuned to optimize sound quality for -certain application. The current default parameters have been -chosen by iterative if-then analysis (read: "trial and error") -to obtain best subjective sound quality in pop/rock music -processing, but in applications processing different kind of -sound the default parameter set may result into a sub-optimal -result.

-

The time-stretch algorithm default -parameter values are set by the following #defines in file "TDStretch.h":

-
- 
#define DEFAULT_SEQUENCE_MS     AUTOMATIC
+    
+        SoundTouch library README
+        
+        
+        
+        
+        
+        
+        
+    
+    
+        

+        
SoundTouch audio processing library v1.5.1pre
+        

+        
SoundTouch library Copyright © Olli Parviainen 2001-2009
+        

+        

+        
1. Introduction
+        

+        
SoundTouch is an open-source audio processing library that allows changing the 
+            sound tempo, pitch and playback rate parameters independently from each other, 
+            i.e.:

+        
+        
1.1 Contact information
+        

+        
Author email: oparviai 'at' iki.fi
+        

+        
SoundTouch WWW page: http://www.surina.net/soundtouch

+        

+        
2. Compiling SoundTouch

+        
Before compiling, notice that you can choose the sample data format if it's 
+            desirable to use floating point sample data instead of 16bit integers. See 
+            section "sample data format" for more information.

+        
2.1. Building in Microsoft Windows

+        
Project files for Microsoft Visual C++ 6.0 and Visual C++ .NET are supplied with 
+            the source code package. 

+        

+            Please notice that SoundTouch library uses processor-specific optimizations for 
+            Pentium III and AMD processors. Visual Studio .NET and later versions supports 
+            the required instructions by default, but Visual Studio 6.0 requires a 
+            processor pack upgrade to be installed in order to support these optimizations. 
+            The processor pack upgrade can be downloaded from Microsoft site at this URL:

+        
http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx

+        
If the above URL is unavailable or removed, go to 
+                http://msdn.microsoft.com and perform a search with keywords "processor 
+            pack".
+        

+        
To build the binaries with Visual C++ compiler, either run "make-win.bat" 
+            script, or open the appropriate project files in source code directories with 
+            Visual Studio. The final executable will appear under the "SoundTouch\bin" 
+            directory. If using the Visual Studio IDE instead of the make-win.bat script, 
+            directories bin and lib may need to be created manually to the SoundTouch 
+            package root for the final executables. The make-win.bat script creates these 
+            directories automatically.
+        

+        
2.2. Building in Gnu platforms

+        
The SoundTouch library can be compiled in practically any platform supporting 
+            GNU compiler (GCC) tools. SoundTouch have been tested with gcc version 3.3.4., 
+            but it shouldn't be very specific about the gcc version. Assembler-level 
+            performance optimizations for GNU platform are currently available in x86 
+            platforms only, they are automatically disabled and replaced with standard C 
+            routines in other processor platforms.

+        
To build and install the binaries, run the following commands in the SoundTouch/ 
+            directory:

+        
+            
+                
+                    
+                    
+                
+                
+                    
+                    
+                
+                
+                    
+                    
+                
+            
+        

+                        
./configure  -

+                    		
+                        
Configures the SoundTouch package for the local environment.

+                    

+                        
make         -

+                    		
+                        
Builds the SoundTouch library & SoundStretch utility.

+                    

+                        
make install -

+                    		
+                        
Installs the SoundTouch & BPM libraries to /usr/local/lib and 
+                            SoundStretch utility to /usr/local/bin. Please notice that 'root' 
+                            privileges may be required to install the binaries to the destination 
+                            locations.

+                    

+        
2.2.1 Required GNU tools 

+        

+            Bash shell, GNU C++ compiler, libtool, autoconf and automake tools are required 
+            for compiling the SoundTouch library. These are usually included with the 
+            GNU/Linux distribution, but if not, install these packages first. For example, 
+            in Ubuntu Linux these can be acquired and installed with the following command:

+        
sudo apt-get install automake autoconf libtool build-essential

+        
2.2.2 Problems with GCC compiler compatibility

+        
At the release time the SoundTouch package has been tested to compile in 
+            GNU/Linux platform. However, in past it's happened that new gcc versions aren't 
+            necessarily compatible with the assembler settings used in the optimized 
+            routines. If you have problems getting the SoundTouch library compiled, try the 
+                workaround of disabling the optimizations by editing the file 
+            "include/STTypes.h" and removing the following definition there:

+        

+            
#define ALLOW_OPTIMIZATIONS 1

+        

+        
2.2.3 Problems with configure script or build process 

+        
Incompatibilities between various GNU toolchain versions may cause errors when 
+            running the "configure" script or building the source codes, if your GNU tool 
+            versions are not compatible with the versions used for preparing the SoundTouch 
+            kit. 

+        
To resolve the issue, regenerate the configure scripts with your local tool set 
+            by running the "./bootstrap" script included in the SoundTouch source 
+            code kit. After that, run the configure script and make as 
+            usually.

+        
2.2.4 Compiler issues with non-x86 processors

+        
SoundTouch library works also on non-x86 processors.

+        
However, in case that you get compiler errors when trying to compile for 
+            non-Intel processor, edit the file "source\SoundTouch\Makefile.am" and 
+            remove the "-msse2" flag on the AM_CXXFLAGS line:

+        
AM_CXXFLAGS=-O3 -fcheck-new -I../../include    # Note: -msse2 flag removed!

+        
After that, run "./bootstrap" script, and then run configure and make
+            again.

+        

+        
3. About implementation & Usage tips

+        
3.1. Supported sample data formats

+        
The sample data format can be chosen between 16bit signed integer and 32bit 
+            floating point values, the default is 32bit floating point.
+        

+        

+            In Windows environment, the sample data format is chosen in file "STTypes.h" by 
+            choosing one of the following defines:

+        
+        

+            In GNU environment, the floating sample format is used by default, but integer 
+            sample format can be chosen by giving the following switch to the configure 
+            script: 

+                
./configure --enable-integer-samples

+            

+        
The sample data can have either single (mono) or double (stereo) audio channel. 
+            Stereo data is interleaved so that every other data value is for left channel 
+            and every second for right channel. Notice that while it'd be possible in 
+            theory to process stereo sound as two separate mono channels, this isn't 
+            recommended because processing the channels separately would result in losing 
+            the phase coherency between the channels, which consequently would ruin the 
+            stereo effect.

+        
Sample rates between 8000-48000H are supported.

+        
3.2. Processing latency

+        
The processing and latency constraints of the SoundTouch library are:

+        
+        
3.3. About algorithms

+        
SoundTouch provides three seemingly independent effects: tempo, pitch and 
+            playback rate control. These three controls are implemented as combination of 
+            two primary effects, sample rate transposing and time-stretching.

+        
Sample rate transposing affects both the audio stream duration and 
+            pitch. It's implemented simply by converting the original audio sample stream 
+            to the  desired duration by interpolating from the original audio samples. 
+            In SoundTouch, linear interpolation with anti-alias filtering is used. 
+            Theoretically a higher-order interpolation provide better result than 1st order 
+            linear interpolation, but in audio application linear interpolation together 
+            with anti-alias filtering performs subjectively about as well as higher-order 
+            filtering would.

+        
Time-stretching means changing the audio stream duration without 
+            affecting it's pitch. SoundTouch uses WSOLA-like time-stretching routines that 
+            operate in the time domain. Compared to sample rate transposing, 
+            time-stretching is a much heavier operation and also requires a longer 
+            processing "window" of sound samples used by the processing algorithm, thus 
+            increasing the algorithm input/output latency. Typical i/o latency for the 
+            SoundTouch time-stretch algorithm is around 100 ms.

+        
Sample rate transposing and time-stretching are then used together to produce 
+            the tempo, pitch and rate controls:

+        
+        
3.4 Tuning the algorithm parameters

+        
The time-stretch algorithm has few parameters that can be tuned to optimize 
+            sound quality for certain application. The current default parameters have been 
+            chosen by iterative if-then analysis (read: "trial and error") to obtain best 
+            subjective sound quality in pop/rock music processing, but in applications 
+            processing different kind of sound the default parameter set may result into a 
+            sub-optimal result.

+        
The time-stretch algorithm default parameter values are set by the following 
+            #defines in file "TDStretch.h":

+        

+            
#define DEFAULT_SEQUENCE_MS     AUTOMATIC
 #define DEFAULT_SEEKWINDOW_MS   AUTOMATIC
 #define DEFAULT_OVERLAP_MS      8

-

-
These parameters affect to the time-stretch
-algorithm as follows:

-
-
Notice that these parameters can also be
-set during execution time with functions "TDStretch::setParameters()"
-and "SoundTouch::setSetting()".

-
The table below summaries how the
-parameters can be adjusted for different applications:

-
-  
-    
-      
-      
-      
-      
-      
-    
-    
-      
-      
-      
-      
-      
-    
-    
-      
-      
-      
-      
-      
-    
-    
-      
-      
-      
-      
-      
-    
-  
-
Parameter nameDefault value
-magnitudeLarger value
-affects...Smaller value
-affects...Effect to CPU burden

-      
SEQUENCE_MS

-      		Default value is relatively
-large, chosen for slowing down music tempoLarger value is usually
-better for slowing down tempo. Growing the value decelerates the
-"echoing" artifact when slowing down the tempo.Smaller value might be better
-for speeding up tempo. Reducing the value accelerates the "echoing"
-artifact when slowing down the tempo Increasing the parameter
-value reduces computation burden

-      
SEEKWINDOW_MS

-      		Default value is relatively
-large, chosen for slowing down music tempoLarger value eases finding a
-good mixing position, but may cause a "drifting" artifactSmaller reduce possibility to
-find a good mixing position, but reduce the "drifting" artifact.Increasing the parameter
-value increases computation burden

-      
OVERLAP_MS

-      		Default value is relatively
-large, chosen to suit with above parameters. If you reduce the "sequence
-ms" setting, you might wish to try a smaller value.Increasing the parameter
-value increases computation burden

-
3.5 Performance Optimizations 

-
General optimizations:

-
The time-stretch routine has a 'quick' mode
-that substantially speeds up the algorithm but may degrade the
-sound quality by a small amount. This mode is activated by
-calling SoundTouch::setSetting() function with parameter  id
-of SETTING_USE_QUICKSEEK and value "1", i.e. 

-

-  
setSetting(SETTING_USE_QUICKSEEK, 1);

-

-
CPU-specific optimizations:

-
-

-
4. SoundStretch audio processing utility
-

-
SoundStretch audio processing utility

-Copyright (c) Olli Parviainen 2002-2009

-
SoundStretch is a simple command-line
-application that can change tempo, pitch and playback rates of
-WAV sound files. This program is intended primarily to
-demonstrate how the "SoundTouch" library can be used to
-process sound in your own program, but it can as well be used for
-processing sound files.

-
4.1. SoundStretch Usage Instructions

-
SoundStretch Usage syntax:

-

-  
soundstretch infilename outfilename [switches]

-

-
Where: 

-
-  
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-  
-

-      
"infilename"

-      		Name of the input sound
-data file (in .WAV audio file format). Give "stdin" as filename to use
-        standard input pipe. 

-      
"outfilename"

-      		Name of the output sound
-file where the resulting sound is saved (in .WAV audio file format).
-This parameter may be omitted if you  don't want to save the
-output
-(e.g. when only calculating BPM rate with '-bpm' switch). Give "stdout"
-        as filename to use standard output pipe.

-      
 [switches]

-      		Are one or more control
-switches.

-
Available control switches are:

-
-  
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-    
-      
-      
-    
-  
-

-      
-tempo=n 

-      		Change the sound tempo by n
-percents (n = -95.0 .. +5000.0 %) 

-      
-pitch=n

-      		Change the sound pitch by n
-semitones (n = -60.0 .. + 60.0 semitones) 

-      
-rate=n

-      		Change the sound playback rate by
-n percents (n = -95.0 .. +5000.0 %) 

-      
-bpm=n

-      		Detect the Beats-Per-Minute (BPM) rate of the sound and adjust the tempo to meet 'n'
-        BPMs. When this switch is
-        applied, the "-tempo" switch is ignored. If "=n" is
-omitted, i.e. switch "-bpm" is used alone, then the BPM rate is
-        estimated and displayed, but tempo not adjusted according to the BPM
-value. 

-      
-quick

-      		Use quicker tempo change
-algorithm. Gains speed but loses sound quality. 

-      
-naa

-      		Don't use anti-alias
-filtering in sample rate transposing. Gains speed but loses sound
-quality. 

-      
-license

-      		Displays the program license
-text (LGPL)

-
Notes:

-
-
4.2. SoundStretch usage examples 

-
Example 1

-
The following command increases tempo of
-the sound file "originalfile.wav" by 12.5% and stores result to file "destinationfile.wav":

-

-  
soundstretch originalfile.wav destinationfile.wav -tempo=12.5

-

-
Example 2

-
The following command decreases the sound
-pitch (key) of the sound file "orig.wav" by two
-semitones and stores the result to file "dest.wav":

-

-  
soundstretch orig.wav dest.wav -pitch=-2

-

-
Example 3

-
The following command processes the file "orig.wav" by decreasing the sound tempo by 25.3% and
-increasing the sound pitch (key) by 1.5 semitones. Resulting .wav audio data is
-directed to standard output pipe:

-

-  
soundstretch orig.wav stdout -tempo=-25.3 -pitch=1.5

-

-
Example 4

-
The following command detects the BPM rate
-of the file "orig.wav" and adjusts the tempo to match
-100 beats per minute. Result is stored to file "dest.wav":

-

-  
soundstretch orig.wav dest.wav -bpm=100

-

-
Example 5

-
The following command reads .wav sound data from standard input pipe and
-estimates the BPM rate:

-

-  
soundstretch stdin -bpm

-

-

-
5. Change History

-
5.1. SoundTouch library Change History 

-
-
1.5.1pre:

-
-
-
1.5.0:

-
-
-
1.4.1:

-
-
-
1.4.0:

-
-
-
v1.3.1:
-

-
-
-
v1.3.0:
-

-
-
v1.2.1: 

-
-
v1.2.0: 

-
-
v1.1.1: 

-
-
v1.01: 

-
-
v1.0: 

-
-
 

-
5.2. SoundStretch application Change
-History 

-
-
v1.5.0:

-
-
-
v1.4.0:

-
-
-
v1.3.0:

-
-
v1.2.1: 

-
-
v1.2.0: 

-
-
v1.1.1: 

-
-
v1.1:

-
-
v1.01: 

-
-

-
6. Acknowledgements 

-
Kudos for these people who have contributed to development or submitted
-bugfixes since
-SoundTouch v1.3.1: 

-
-
Moral greetings to all other contributors and users also!

-

-
7. LICENSE 

-
SoundTouch audio processing library

-Copyright (c) Olli Parviainen

-
This library is free software; you can
-redistribute it and/or modify it under the terms of the GNU
-Lesser General Public License version 2.1 as published by the Free Software
-Foundation.

-
This library is distributed in the hope
-that it will be useful, but WITHOUT ANY WARRANTY; without even
-the implied warranty of MERCHANTABILITY or FITNESS FOR A
-PARTICULAR PURPOSE. See the GNU Lesser General Public License for
-more details.

-
You should have received a copy of the GNU
-Lesser General Public License along with this library; if not,
-write to the Free Software Foundation, Inc., 59 Temple Place,
-Suite 330, Boston, MA 02111-1307 USA

-

-
-
+    
 
diff --git a/include/STTypes.h b/include/STTypes.h
index 48c36a6..8aa5b19 100644
--- a/include/STTypes.h
+++ b/include/STTypes.h
@@ -128,11 +128,7 @@ namespace soundtouch
         typedef double LONG_SAMPLETYPE;
 
         #ifdef ALLOW_X86_OPTIMIZATIONS
-                // Allow 3DNow! and SSE optimizations
-            #if WIN32
-                #define ALLOW_3DNOW     1
-            #endif
-
+            // Allow SSE optimizations
             #define ALLOW_SSE       1
         #endif
 
diff --git a/source/SoundTouch/3dnow_win.cpp b/source/SoundTouch/3dnow_win.cpp
deleted file mode 100644
index 1fbbdab..0000000
--- a/source/SoundTouch/3dnow_win.cpp
+++ /dev/null
@@ -1,349 +0,0 @@
-////////////////////////////////////////////////////////////////////////////////
-///
-/// Win32 version of the AMD 3DNow! optimized routines for AMD K6-2/Athlon 
-/// processors. All 3DNow! optimized functions have been gathered into this
-/// single source code file, regardless to their class or original source code 
-/// file, in order to ease porting the library to other compiler and processor 
-/// platforms.
-///
-/// By the way; the performance gain depends heavily on the CPU generation: On 
-/// K6-2 these routines provided speed-up of even 2.4 times, while on Athlon the 
-/// difference to the original routines stayed at unremarkable 8%! Such a small 
-/// improvement on Athlon is due to 3DNow can perform only two operations in 
-/// parallel, and obviously also the Athlon FPU is doing a very good job with
-/// the standard C floating point routines! Here these routines are anyway, 
-/// although it might not be worth the effort to convert these to GCC platform, 
-/// for Athlon CPU at least. The situation is different regarding the SSE 
-/// optimizations though, thanks to the four parallel operations of SSE that 
-/// already make a difference.
-/// 
-/// This file is to be compiled in Windows platform with Microsoft Visual C++ 
-/// Compiler. Please see '3dnow_gcc.cpp' for the gcc compiler version for all
-/// GNU platforms (if file supplied).
-///
-/// NOTICE: If using Visual Studio 6.0, you'll need to install the "Visual C++ 
-/// 6.0 processor pack" update to support 3DNow! instruction set. The update is 
-/// available for download at Microsoft Developers Network, see here:
-/// http://msdn.microsoft.com/en-us/vstudio/aa718349.aspx
-///
-/// If the above URL is expired or removed, go to "http://msdn.microsoft.com" and 
-/// perform a search with keywords "processor pack".
-///
-/// Author        : Copyright (c) Olli Parviainen
-/// Author e-mail : oparviai 'at' iki.fi
-/// SoundTouch WWW: http://www.surina.net/soundtouch
-///
-////////////////////////////////////////////////////////////////////////////////
-//
-// Last changed  : $Date$
-// File revision : $Revision: 4 $
-//
-// $Id$
-//
-////////////////////////////////////////////////////////////////////////////////
-//
-// License :
-//
-//  SoundTouch audio processing library
-//  Copyright (c) Olli Parviainen
-//
-//  This library is free software; you can redistribute it and/or
-//  modify it under the terms of the GNU Lesser General Public
-//  License as published by the Free Software Foundation; either
-//  version 2.1 of the License, or (at your option) any later version.
-//
-//  This library is distributed in the hope that it will be useful,
-//  but WITHOUT ANY WARRANTY; without even the implied warranty of
-//  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-//  Lesser General Public License for more details.
-//
-//  You should have received a copy of the GNU Lesser General Public
-//  License along with this library; if not, write to the Free Software
-//  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
-//
-////////////////////////////////////////////////////////////////////////////////
-
-#include "cpu_detect.h"
-#include "STTypes.h"
-
-#ifndef WIN32
-#error "wrong platform - this source code file is exclusively for Win32 platform"
-#endif
-
-using namespace soundtouch;
-
-#ifdef ALLOW_3DNOW
-// 3DNow! routines available only with float sample type    
-
-//////////////////////////////////////////////////////////////////////////////
-//
-// implementation of 3DNow! optimized functions of class 'TDStretch3DNow'
-//
-//////////////////////////////////////////////////////////////////////////////
-
-#include "TDStretch.h"
-
-
-// Calculates cross correlation of two buffers
-double TDStretch3DNow::calcCrossCorrStereo(const float *pV1, const float *pV2) const
-{
-    int overlapLengthLocal = overlapLength;
-    float corr = 0;
-
-    // Calculates the cross-correlation value between 'pV1' and 'pV2' vectors
-    /*
-    c-pseudocode:
-
-        corr = 0;
-        for (i = 0; i < overlapLength / 4; i ++)
-        {
-            corr += pV1[0] * pV2[0];
-                    pV1[1] * pV2[1];
-                    pV1[2] * pV2[2];
-                    pV1[3] * pV2[3];
-                    pV1[4] * pV2[4];
-                    pV1[5] * pV2[5];
-                    pV1[6] * pV2[6];
-                    pV1[7] * pV2[7];
-
-            pV1 += 8;
-            pV2 += 8;
-        }
-    */
-
-    _asm 
-    {
-        // give prefetch hints to CPU of what data are to be needed soonish.
-        // give more aggressive hints on pV1 as that changes more between different calls 
-        // while pV2 stays the same.
-        prefetch [pV1]
-        prefetch [pV2]
-        prefetch [pV1 + 32]
-
-        mov     eax, dword ptr pV2
-        mov     ebx, dword ptr pV1
-
-        pxor    mm0, mm0
-
-        mov     ecx, overlapLengthLocal
-        shr     ecx, 2  // div by four
-
-    loop1:
-        movq    mm1, [eax]
-        prefetch [eax + 32]     // give a prefetch hint to CPU what data are to be needed soonish
-        pfmul   mm1, [ebx]
-        prefetch [ebx + 64]     // give a prefetch hint to CPU what data are to be needed soonish
-
-        movq    mm2, [eax + 8]
-        pfadd   mm0, mm1
-        pfmul   mm2, [ebx + 8]
-
-        movq    mm3, [eax + 16]
-        pfadd   mm0, mm2
-        pfmul   mm3, [ebx + 16]
-
-        movq    mm4, [eax + 24]
-        pfadd   mm0, mm3
-        pfmul   mm4, [ebx + 24]
-
-        add     eax, 32
-        pfadd   mm0, mm4
-        add     ebx, 32
-
-        dec     ecx
-        jnz     loop1
-
-        // add halfs of mm0 together and return the result. 
-        // note: mm1 is used as a dummy parameter only, we actually don't care about it's value
-        pfacc   mm0, mm1
-        movd    corr, mm0
-        femms
-    }
-
-    return corr;
-}
-
-
-
-
-//////////////////////////////////////////////////////////////////////////////
-//
-// implementation of 3DNow! optimized functions of class 'FIRFilter'
-//
-//////////////////////////////////////////////////////////////////////////////
-
-#include "FIRFilter.h"
-
-FIRFilter3DNow::FIRFilter3DNow() : FIRFilter()
-{
-    filterCoeffsUnalign = NULL;
-    filterCoeffsAlign = NULL;
-}
-
-
-FIRFilter3DNow::~FIRFilter3DNow()
-{
-    delete[] filterCoeffsUnalign;
-    filterCoeffsUnalign = NULL;
-    filterCoeffsAlign = NULL;
-}
-
-
-// (overloaded) Calculates filter coefficients for 3DNow! routine
-void FIRFilter3DNow::setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor)
-{
-    uint i;
-    float fDivider;
-
-    FIRFilter::setCoefficients(coeffs, newLength, uResultDivFactor);
-
-    // Scale the filter coefficients so that it won't be necessary to scale the filtering result
-    // also rearrange coefficients suitably for 3DNow!
-    // Ensure that filter coeffs array is aligned to 16-byte boundary
-    delete[] filterCoeffsUnalign;
-    filterCoeffsUnalign = new float[2 * newLength + 4];
-    filterCoeffsAlign = (float *)(((uint)filterCoeffsUnalign + 15) & (uint)-16);
-
-    fDivider = (float)resultDivider;
-
-    // rearrange the filter coefficients for mmx routines 
-    for (i = 0; i < newLength; i ++)
-    {
-        filterCoeffsAlign[2 * i + 0] =
-        filterCoeffsAlign[2 * i + 1] = coeffs[i + 0] / fDivider;
-    }
-}
-
-
-// 3DNow!-optimized version of the filter routine for stereo sound
-uint FIRFilter3DNow::evaluateFilterStereo(float *dest, const float *src, uint numSamples) const
-{
-    float *filterCoeffsLocal = filterCoeffsAlign;
-    uint count = (numSamples - length) & (uint)-2;
-    uint lengthLocal = length / 4;
-
-    assert(length != 0);
-    assert(count % 2 == 0);
-
-    /* original code:
-
-    double suml1, suml2;
-    double sumr1, sumr2;
-    uint i, j;
-
-    for (j = 0; j < count; j += 2)
-    {
-        const float *ptr;
-
-        suml1 = sumr1 = 0.0;
-        suml2 = sumr2 = 0.0;
-        ptr = src;
-        filterCoeffsLocal = filterCoeffs;
-        for (i = 0; i < lengthLocal; i ++) 
-        {
-            // unroll loop for efficiency.
-
-            suml1 += ptr[0] * filterCoeffsLocal[0] + 
-                     ptr[2] * filterCoeffsLocal[2] +
-                     ptr[4] * filterCoeffsLocal[4] +
-                     ptr[6] * filterCoeffsLocal[6];
-
-            sumr1 += ptr[1] * filterCoeffsLocal[1] + 
-                     ptr[3] * filterCoeffsLocal[3] +
-                     ptr[5] * filterCoeffsLocal[5] +
-                     ptr[7] * filterCoeffsLocal[7];
-
-            suml2 += ptr[8] * filterCoeffsLocal[0] + 
-                     ptr[10] * filterCoeffsLocal[2] +
-                     ptr[12] * filterCoeffsLocal[4] +
-                     ptr[14] * filterCoeffsLocal[6];
-
-            sumr2 += ptr[9] * filterCoeffsLocal[1] + 
-                     ptr[11] * filterCoeffsLocal[3] +
-                     ptr[13] * filterCoeffsLocal[5] +
-                     ptr[15] * filterCoeffsLocal[7];
-
-            ptr += 16;
-            filterCoeffsLocal += 8;
-        }
-        dest[0] = (float)suml1;
-        dest[1] = (float)sumr1;
-        dest[2] = (float)suml2;
-        dest[3] = (float)sumr2;
-
-        src += 4;
-        dest += 4;
-    }
-
-    */
-    _asm
-    {
-        mov     eax, dword ptr dest
-        mov     ebx, dword ptr src
-        mov     edx, count
-        shr     edx, 1
-
-    loop1:
-        // "outer loop" : during each round 2*2 output samples are calculated
-        prefetch  [ebx]                 // give a prefetch hint to CPU what data are to be needed soonish
-        prefetch  [filterCoeffsLocal]   // give a prefetch hint to CPU what data are to be needed soonish
-
-        mov     esi, ebx
-        mov     edi, filterCoeffsLocal
-        pxor    mm0, mm0
-        pxor    mm1, mm1
-        mov     ecx, lengthLocal
-
-    loop2:
-        // "inner loop" : during each round four FIR filter taps are evaluated for 2*2 output samples
-        movq    mm2, [edi]
-        movq    mm3, mm2
-        prefetch  [edi + 32]     // give a prefetch hint to CPU what data are to be needed soonish
-        pfmul   mm2, [esi]
-        prefetch  [esi + 32]     // give a prefetch hint to CPU what data are to be needed soonish
-        pfmul   mm3, [esi + 8]
-
-        movq    mm4, [edi + 8]
-        movq    mm5, mm4
-        pfadd   mm0, mm2
-        pfmul   mm4, [esi + 8]
-        pfadd   mm1, mm3
-        pfmul   mm5, [esi + 16]
-
-        movq    mm2, [edi + 16]
-        movq    mm6, mm2
-        pfadd   mm0, mm4
-        pfmul   mm2, [esi + 16]
-        pfadd   mm1, mm5
-        pfmul   mm6, [esi + 24]
-
-        movq    mm3, [edi + 24]
-        movq    mm7, mm3
-        pfadd   mm0, mm2
-        pfmul   mm3, [esi + 24]
-        pfadd   mm1, mm6
-        pfmul   mm7, [esi + 32]
-        add     esi, 32
-        pfadd   mm0, mm3
-        add     edi, 32
-        pfadd   mm1, mm7
-
-        dec     ecx
-        jnz     loop2
-
-        movq    [eax], mm0
-        add     ebx, 16
-        movq    [eax + 8], mm1
-        add     eax, 16
-
-        dec     edx
-        jnz     loop1
-
-        femms
-    }
-
-    return count;
-}
-
-
-#endif  // ALLOW_3DNOW
diff --git a/source/SoundTouch/FIRFilter.cpp b/source/SoundTouch/FIRFilter.cpp
index 72607bd..001f70f 100644
--- a/source/SoundTouch/FIRFilter.cpp
+++ b/source/SoundTouch/FIRFilter.cpp
@@ -233,7 +233,7 @@ FIRFilter * FIRFilter::newInstance()
 
     uExtensions = detectCPUextensions();
 
-    // Check if MMX/SSE/3DNow! instruction set extensions supported by CPU
+    // Check if MMX/SSE instruction set extensions supported by CPU
 
 #ifdef ALLOW_MMX
     // MMX routines available only with integer sample types
@@ -253,15 +253,6 @@ FIRFilter * FIRFilter::newInstance()
     else
 #endif // ALLOW_SSE
 
-#ifdef ALLOW_3DNOW
-    if (uExtensions & SUPPORT_3DNOW)
-    {
-        // 3DNow! support
-        return ::new FIRFilter3DNow;
-    }
-    else
-#endif // ALLOW_3DNOW
-
     {
         // ISA optimizations not supported, use plain C version
         return ::new FIRFilter;
diff --git a/source/SoundTouch/FIRFilter.h b/source/SoundTouch/FIRFilter.h
index 71e88bd..5de98df 100644
--- a/source/SoundTouch/FIRFilter.h
+++ b/source/SoundTouch/FIRFilter.h
@@ -122,25 +122,6 @@ public:
 #endif // ALLOW_MMX
 
 
-#ifdef ALLOW_3DNOW
-
-    /// Class that implements 3DNow! optimized functions exclusive for floating point samples type.
-    class FIRFilter3DNow : public FIRFilter
-    {
-    protected:
-        float *filterCoeffsUnalign;
-        float *filterCoeffsAlign;
-
-        virtual uint evaluateFilterStereo(float *dest, const float *src, uint numSamples) const;
-    public:
-        FIRFilter3DNow();
-        ~FIRFilter3DNow();
-        virtual void setCoefficients(const float *coeffs, uint newLength, uint uResultDivFactor);
-    };
-
-#endif  // ALLOW_3DNOW
-
-
 #ifdef ALLOW_SSE
     /// Class that implements SSE optimized functions exclusive for floating point samples type.
     class FIRFilterSSE : public FIRFilter
diff --git a/source/SoundTouch/SoundTouch.vcproj b/source/SoundTouch/SoundTouch.vcproj
index 62a3d55..a9b3ffe 100644
--- a/source/SoundTouch/SoundTouch.vcproj
+++ b/source/SoundTouch/SoundTouch.vcproj
@@ -125,27 +125,6 @@
 		
-			
-				
-					
-				
-				
-					
-				
-