SDL-mirror/src/audio/sdlgenaudiocvt.pl

761 lines
21 KiB
Perl
Raw Normal View History

First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
#!/usr/bin/perl -w
use warnings;
use strict;
my @audiotypes = qw(
U8
S8
U16LSB
S16LSB
U16MSB
S16MSB
S32LSB
S32MSB
F32LSB
F32MSB
);
my @channels = ( 1, 2, 4, 6, 8 );
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
my %funcs;
my $custom_converters = 0;
sub getTypeConvertHashId {
my ($from, $to) = @_;
return "TYPECONVERTER $from/$to";
}
sub getResamplerHashId {
my ($from, $channels, $upsample, $multiple) = @_;
return "RESAMPLER $from/$channels/$upsample/$multiple";
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
sub outputHeader {
print <<EOF;
/* DO NOT EDIT! This file is generated by sdlgenaudiocvt.pl */
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
/*
2011-04-08 13:03:26 -07:00
Simple DirectMedia Layer
2013-02-15 08:47:44 -08:00
Copyright (C) 1997-2013 Sam Lantinga <slouken\@libsdl.org>
2011-04-08 13:03:26 -07:00
This software is provided 'as-is', without any express or implied
warranty. In no event will the authors be held liable for any damages
arising from the use of this software.
Permission is granted to anyone to use this software for any purpose,
including commercial applications, and to alter it and redistribute it
freely, subject to the following restrictions:
1. The origin of this software must not be misrepresented; you must not
claim that you wrote the original software. If you use this software
in a product, an acknowledgment in the product documentation would be
appreciated but is not required.
2. Altered source versions must be plainly marked as such, and must not be
misrepresented as being the original software.
3. This notice may not be removed or altered from any source distribution.
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
*/
#include "SDL_config.h"
#include "SDL_audio.h"
#include "SDL_audio_c.h"
#ifndef DEBUG_CONVERT
#define DEBUG_CONVERT 0
#endif
/* If you can guarantee your data and need space, you can eliminate code... */
/* Just build the arbitrary resamplers if you're saving code space. */
#ifndef LESS_RESAMPLERS
#define LESS_RESAMPLERS 0
#endif
/* Don't build any resamplers if you're REALLY saving code space. */
#ifndef NO_RESAMPLERS
#define NO_RESAMPLERS 0
#endif
/* Don't build any type converters if you're saving code space. */
#ifndef NO_CONVERTERS
#define NO_CONVERTERS 0
#endif
/* *INDENT-OFF* */
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
EOF
my @vals = ( 127, 32767, 2147483647 );
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
foreach (@vals) {
my $val = $_;
my $fval = 1.0 / $val;
print("#define DIVBY${val} ${fval}f\n");
}
print("\n");
}
sub outputFooter {
print <<EOF;
/* $custom_converters converters generated. */
/* *INDENT-ON* */
/* vi: set ts=4 sw=4 expandtab: */
EOF
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
sub splittype {
my $t = shift;
my ($signed, $size, $endian) = $t =~ /([USF])(\d+)([LM]SB|)/;
my $float = ($signed eq 'F') ? 1 : 0;
$signed = (($float) or ($signed eq 'S')) ? 1 : 0;
$endian = 'NONE' if ($endian eq '');
my $ctype = '';
if ($float) {
$ctype = (($size == 32) ? 'float' : 'double');
} else {
$ctype = (($signed) ? 'S' : 'U') . "int${size}";
}
return ($signed, $float, $size, $endian, $ctype);
}
sub getSwapFunc {
my ($size, $signed, $float, $endian, $val) = @_;
my $BEorLE = (($endian eq 'MSB') ? 'BE' : 'LE');
my $code = '';
if ($float) {
$code = "SDL_SwapFloat${BEorLE}($val)";
} else {
if ($size > 8) {
$code = "SDL_Swap${BEorLE}${size}($val)";
} else {
$code = $val;
}
if (($signed) and (!$float)) {
$code = "((Sint${size}) $code)";
}
}
return "${code}";
}
sub maxIntVal {
my $size = shift;
if ($size == 8) {
return 0x7F;
} elsif ($size == 16) {
return 0x7FFF;
} elsif ($size == 32) {
return 0x7FFFFFFF;
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
die("bug in script.\n");
}
sub getFloatToIntMult {
my $size = shift;
my $val = maxIntVal($size) . '.0';
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
$val .= 'f' if ($size < 32);
return $val;
}
sub getIntToFloatDivBy {
my $size = shift;
return 'DIVBY' . maxIntVal($size);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
sub getSignFlipVal {
my $size = shift;
if ($size == 8) {
return '0x80';
} elsif ($size == 16) {
return '0x8000';
} elsif ($size == 32) {
return '0x80000000';
}
die("bug in script.\n");
}
sub buildCvtFunc {
my ($from, $to) = @_;
my ($fsigned, $ffloat, $fsize, $fendian, $fctype) = splittype($from);
my ($tsigned, $tfloat, $tsize, $tendian, $tctype) = splittype($to);
my $diffs = 0;
$diffs++ if ($fsize != $tsize);
$diffs++ if ($fsigned != $tsigned);
$diffs++ if ($ffloat != $tfloat);
$diffs++ if ($fendian ne $tendian);
return if ($diffs == 0);
my $hashid = getTypeConvertHashId($from, $to);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
if (1) { # !!! FIXME: if ($diffs > 1) {
my $sym = "SDL_Convert_${from}_to_${to}";
$funcs{$hashid} = $sym;
$custom_converters++;
# Always unsigned for ints, for possible byteswaps.
my $srctype = (($ffloat) ? 'float' : "Uint${fsize}");
print <<EOF;
static void SDLCALL
${sym}(SDL_AudioCVT * cvt, SDL_AudioFormat format)
{
int i;
const $srctype *src;
$tctype *dst;
#if DEBUG_CONVERT
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
fprintf(stderr, "Converting AUDIO_${from} to AUDIO_${to}.\\n");
#endif
EOF
if ($fsize < $tsize) {
my $mult = $tsize / $fsize;
print <<EOF;
src = ((const $srctype *) (cvt->buf + cvt->len_cvt)) - 1;
dst = (($tctype *) (cvt->buf + cvt->len_cvt * $mult)) - 1;
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
for (i = cvt->len_cvt / sizeof ($srctype); i; --i, --src, --dst) {
EOF
} else {
print <<EOF;
src = (const $srctype *) cvt->buf;
dst = ($tctype *) cvt->buf;
for (i = cvt->len_cvt / sizeof ($srctype); i; --i, ++src, ++dst) {
EOF
}
# Have to convert to/from float/int.
# !!! FIXME: cast through double for int32<->float?
my $code = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, '*src');
if ($ffloat != $tfloat) {
if ($ffloat) {
my $mult = getFloatToIntMult($tsize);
if (!$tsigned) { # bump from -1.0f/1.0f to 0.0f/2.0f
$code = "($code + 1.0f)";
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
$code = "(($tctype) ($code * $mult))";
} else {
# $divby will be the reciprocal, to avoid pipeline stalls
# from floating point division...so multiply it.
my $divby = getIntToFloatDivBy($fsize);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
$code = "(((float) $code) * $divby)";
if (!$fsigned) { # bump from 0.0f/2.0f to -1.0f/1.0f.
$code = "($code - 1.0f)";
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
} else {
# All integer conversions here.
if ($fsigned != $tsigned) {
my $signflipval = getSignFlipVal($fsize);
$code = "(($code) ^ $signflipval)";
}
my $shiftval = abs($fsize - $tsize);
if ($fsize < $tsize) {
$code = "((($tctype) $code) << $shiftval)";
} elsif ($fsize > $tsize) {
$code = "(($tctype) ($code >> $shiftval))";
}
}
my $swap = getSwapFunc($tsize, $tsigned, $tfloat, $tendian, 'val');
print <<EOF;
const $tctype val = $code;
*dst = ${swap};
}
EOF
if ($fsize > $tsize) {
my $divby = $fsize / $tsize;
print(" cvt->len_cvt /= $divby;\n");
} elsif ($fsize < $tsize) {
my $mult = $tsize / $fsize;
print(" cvt->len_cvt *= $mult;\n");
}
print <<EOF;
if (cvt->filters[++cvt->filter_index]) {
cvt->filters[cvt->filter_index] (cvt, AUDIO_$to);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
}
EOF
} else {
if ($fsigned != $tsigned) {
$funcs{$hashid} = 'SDL_ConvertSigned';
} elsif ($ffloat != $tfloat) {
$funcs{$hashid} = 'SDL_ConvertFloat';
} elsif ($fsize != $tsize) {
$funcs{$hashid} = 'SDL_ConvertSize';
} elsif ($fendian ne $tendian) {
$funcs{$hashid} = 'SDL_ConvertEndian';
} else {
die("error in script.\n");
}
}
}
sub buildTypeConverters {
print "#if !NO_CONVERTERS\n\n";
foreach (@audiotypes) {
my $from = $_;
foreach (@audiotypes) {
my $to = $_;
buildCvtFunc($from, $to);
}
}
print "#endif /* !NO_CONVERTERS */\n\n\n";
print "const SDL_AudioTypeFilters sdl_audio_type_filters[] =\n{\n";
print "#if !NO_CONVERTERS\n";
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
foreach (@audiotypes) {
my $from = $_;
foreach (@audiotypes) {
my $to = $_;
if ($from ne $to) {
my $hashid = getTypeConvertHashId($from, $to);
my $sym = $funcs{$hashid};
print(" { AUDIO_$from, AUDIO_$to, $sym },\n");
}
}
}
print "#endif /* !NO_CONVERTERS */\n";
print(" { 0, 0, NULL }\n");
print "};\n\n\n";
}
sub getBiggerCtype {
my ($isfloat, $size) = @_;
if ($isfloat) {
if ($size == 32) {
return 'double';
}
die("bug in script.\n");
}
if ($size == 8) {
return 'Sint16';
} elsif ($size == 16) {
return 'Sint32'
} elsif ($size == 32) {
return 'Sint64'
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
die("bug in script.\n");
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
# These handle arbitrary resamples...44100Hz to 48000Hz, for example.
# Man, this code is skanky.
sub buildArbitraryResampleFunc {
# !!! FIXME: we do a lot of unnecessary and ugly casting in here, due to getSwapFunc().
my ($from, $channels, $upsample) = @_;
my ($fsigned, $ffloat, $fsize, $fendian, $fctype) = splittype($from);
my $bigger = getBiggerCtype($ffloat, $fsize);
my $interp = ($ffloat) ? '* 0.5' : '>> 1';
my $resample = ($upsample) ? 'Upsample' : 'Downsample';
my $hashid = getResamplerHashId($from, $channels, $upsample, 0);
my $sym = "SDL_${resample}_${from}_${channels}c";
$funcs{$hashid} = $sym;
$custom_converters++;
my $fudge = $fsize * $channels * 2; # !!! FIXME
my $eps_adjust = ($upsample) ? 'dstsize' : 'srcsize';
my $incr = '';
my $incr2 = '';
# !!! FIXME: DEBUG_CONVERT should report frequencies.
print <<EOF;
static void SDLCALL
${sym}(SDL_AudioCVT * cvt, SDL_AudioFormat format)
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
{
#if DEBUG_CONVERT
fprintf(stderr, "$resample arbitrary (x%f) AUDIO_${from}, ${channels} channels.\\n", cvt->rate_incr);
#endif
const int srcsize = cvt->len_cvt - $fudge;
const int dstsize = (int) (((double)cvt->len_cvt) * cvt->rate_incr);
register int eps = 0;
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
EOF
my $endcomparison = '!=';
# Upsampling (growing the buffer) needs to work backwards, since we
# overwrite the buffer as we go.
if ($upsample) {
$endcomparison = '>='; # dst > target
print <<EOF;
$fctype *dst = (($fctype *) (cvt->buf + dstsize)) - $channels;
const $fctype *src = (($fctype *) (cvt->buf + cvt->len_cvt)) - $channels;
const $fctype *target = ((const $fctype *) cvt->buf);
EOF
} else {
$endcomparison = '<'; # dst < target
print <<EOF;
$fctype *dst = ($fctype *) cvt->buf;
const $fctype *src = ($fctype *) cvt->buf;
const $fctype *target = (const $fctype *) (cvt->buf + dstsize);
EOF
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
my $val = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "src[$idx]");
print <<EOF;
$fctype sample${idx} = $val;
EOF
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
print <<EOF;
$fctype last_sample${idx} = sample${idx};
EOF
}
print <<EOF;
while (dst $endcomparison target) {
EOF
if ($upsample) {
for (my $i = 0; $i < $channels; $i++) {
# !!! FIXME: don't do this swap every write, just when the samples change.
my $idx = (($channels - $i) - 1);
my $val = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "sample${idx}");
print <<EOF;
dst[$idx] = $val;
EOF
}
$incr = ($channels == 1) ? 'dst--' : "dst -= $channels";
$incr2 = ($channels == 1) ? 'src--' : "src -= $channels";
print <<EOF;
$incr;
eps += srcsize;
if ((eps << 1) >= dstsize) {
$incr2;
EOF
} else { # downsample.
$incr = ($channels == 1) ? 'src++' : "src += $channels";
print <<EOF;
$incr;
eps += dstsize;
if ((eps << 1) >= srcsize) {
EOF
for (my $i = 0; $i < $channels; $i++) {
my $val = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "sample${i}");
print <<EOF;
dst[$i] = $val;
EOF
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
$incr = ($channels == 1) ? 'dst++' : "dst += $channels";
print <<EOF;
$incr;
EOF
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
my $swapped = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "src[$idx]");
print <<EOF;
sample${idx} = ($fctype) (((($bigger) $swapped) + (($bigger) last_sample${idx})) $interp);
EOF
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
print <<EOF;
last_sample${idx} = sample${idx};
EOF
}
print <<EOF;
eps -= $eps_adjust;
}
}
EOF
print <<EOF;
cvt->len_cvt = dstsize;
if (cvt->filters[++cvt->filter_index]) {
cvt->filters[cvt->filter_index] (cvt, format);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
}
EOF
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
# These handle clean resamples...doubling and quadrupling the sample rate, etc.
sub buildMultipleResampleFunc {
# !!! FIXME: we do a lot of unnecessary and ugly casting in here, due to getSwapFunc().
my ($from, $channels, $upsample, $multiple) = @_;
my ($fsigned, $ffloat, $fsize, $fendian, $fctype) = splittype($from);
my $bigger = getBiggerCtype($ffloat, $fsize);
my $interp = ($ffloat) ? '* 0.5' : '>> 1';
my $interp2 = ($ffloat) ? '* 0.25' : '>> 2';
my $mult3 = ($ffloat) ? '3.0' : '3';
my $lencvtop = ($upsample) ? '*' : '/';
my $resample = ($upsample) ? 'Upsample' : 'Downsample';
my $hashid = getResamplerHashId($from, $channels, $upsample, $multiple);
my $sym = "SDL_${resample}_${from}_${channels}c_x${multiple}";
$funcs{$hashid} = $sym;
$custom_converters++;
# !!! FIXME: DEBUG_CONVERT should report frequencies.
print <<EOF;
static void SDLCALL
${sym}(SDL_AudioCVT * cvt, SDL_AudioFormat format)
{
#if DEBUG_CONVERT
fprintf(stderr, "$resample (x${multiple}) AUDIO_${from}, ${channels} channels.\\n");
#endif
const int dstsize = cvt->len_cvt $lencvtop $multiple;
EOF
my $endcomparison = '!=';
# Upsampling (growing the buffer) needs to work backwards, since we
# overwrite the buffer as we go.
if ($upsample) {
$endcomparison = '>='; # dst > target
print <<EOF;
$fctype *dst = (($fctype *) (cvt->buf + dstsize)) - $channels * $multiple;
const $fctype *src = (($fctype *) (cvt->buf + cvt->len_cvt)) - $channels;
const $fctype *target = ((const $fctype *) cvt->buf);
EOF
} else {
$endcomparison = '<'; # dst < target
print <<EOF;
$fctype *dst = ($fctype *) cvt->buf;
const $fctype *src = ($fctype *) cvt->buf;
const $fctype *target = (const $fctype *) (cvt->buf + dstsize);
EOF
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
my $val = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "src[$idx]");
print <<EOF;
$bigger last_sample${idx} = ($bigger) $val;
EOF
}
print <<EOF;
while (dst $endcomparison target) {
EOF
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
my $val = getSwapFunc($fsize, $fsigned, $ffloat, $fendian, "src[$idx]");
print <<EOF;
const $bigger sample${idx} = ($bigger) $val;
EOF
}
my $incr = '';
if ($upsample) {
$incr = ($channels == 1) ? 'src--' : "src -= $channels";
} else {
my $amount = $channels * $multiple;
$incr = "src += $amount"; # can't ever be 1, so no "++" version.
}
print <<EOF;
$incr;
EOF
# !!! FIXME: This really begs for some Altivec or SSE, etc.
if ($upsample) {
if ($multiple == 2) {
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i + $channels;
print <<EOF;
dst[$dsti] = ($fctype) ((sample${i} + last_sample${i}) $interp);
EOF
}
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i;
print <<EOF;
dst[$dsti] = ($fctype) sample${i};
EOF
}
} elsif ($multiple == 4) {
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i + ($channels * 3);
print <<EOF;
dst[$dsti] = ($fctype) ((sample${i} + ($mult3 * last_sample${i})) $interp2);
EOF
}
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i + ($channels * 2);
print <<EOF;
dst[$dsti] = ($fctype) ((sample${i} + last_sample${i}) $interp);
EOF
}
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i + ($channels * 1);
print <<EOF;
dst[$dsti] = ($fctype) ((($mult3 * sample${i}) + last_sample${i}) $interp2);
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
EOF
}
for (my $i = $channels-1; $i >= 0; $i--) {
my $dsti = $i + ($channels * 0);
print <<EOF;
dst[$dsti] = ($fctype) sample${i};
EOF
}
} else {
die('bug in program.'); # we only handle x2 and x4.
}
} else { # downsample.
if ($multiple == 2) {
for (my $i = 0; $i < $channels; $i++) {
print <<EOF;
dst[$i] = ($fctype) ((sample${i} + last_sample${i}) $interp);
EOF
}
} elsif ($multiple == 4) {
# !!! FIXME: interpolate all 4 samples?
for (my $i = 0; $i < $channels; $i++) {
print <<EOF;
dst[$i] = ($fctype) ((sample${i} + last_sample${i}) $interp);
EOF
}
} else {
die('bug in program.'); # we only handle x2 and x4.
}
}
for (my $i = 0; $i < $channels; $i++) {
my $idx = ($upsample) ? (($channels - $i) - 1) : $i;
print <<EOF;
last_sample${idx} = sample${idx};
EOF
}
if ($upsample) {
my $amount = $channels * $multiple;
$incr = "dst -= $amount"; # can't ever be 1, so no "--" version.
} else {
$incr = ($channels == 1) ? 'dst++' : "dst += $channels";
}
print <<EOF;
$incr;
}
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
cvt->len_cvt = dstsize;
if (cvt->filters[++cvt->filter_index]) {
cvt->filters[cvt->filter_index] (cvt, format);
}
}
EOF
}
sub buildResamplers {
print "#if !NO_RESAMPLERS\n\n";
foreach (@audiotypes) {
my $from = $_;
foreach (@channels) {
my $channel = $_;
buildArbitraryResampleFunc($from, $channel, 1);
buildArbitraryResampleFunc($from, $channel, 0);
}
}
print "\n#if !LESS_RESAMPLERS\n\n";
foreach (@audiotypes) {
my $from = $_;
foreach (@channels) {
my $channel = $_;
for (my $multiple = 2; $multiple <= 4; $multiple += 2) {
buildMultipleResampleFunc($from, $channel, 1, $multiple);
buildMultipleResampleFunc($from, $channel, 0, $multiple);
}
}
}
print "#endif /* !LESS_RESAMPLERS */\n";
print "#endif /* !NO_RESAMPLERS */\n\n\n";
print "const SDL_AudioRateFilters sdl_audio_rate_filters[] =\n{\n";
print "#if !NO_RESAMPLERS\n";
foreach (@audiotypes) {
my $from = $_;
foreach (@channels) {
my $channel = $_;
for (my $upsample = 0; $upsample <= 1; $upsample++) {
my $hashid = getResamplerHashId($from, $channel, $upsample, 0);
my $sym = $funcs{$hashid};
print(" { AUDIO_$from, $channel, $upsample, 0, $sym },\n");
}
}
}
print "#if !LESS_RESAMPLERS\n";
foreach (@audiotypes) {
my $from = $_;
foreach (@channels) {
my $channel = $_;
for (my $multiple = 2; $multiple <= 4; $multiple += 2) {
for (my $upsample = 0; $upsample <= 1; $upsample++) {
my $hashid = getResamplerHashId($from, $channel, $upsample, $multiple);
my $sym = $funcs{$hashid};
print(" { AUDIO_$from, $channel, $upsample, $multiple, $sym },\n");
}
}
}
}
print "#endif /* !LESS_RESAMPLERS */\n";
print "#endif /* !NO_RESAMPLERS */\n";
print(" { 0, 0, 0, 0, NULL }\n");
print "};\n\n";
}
# mainline ...
outputHeader();
buildTypeConverters();
buildResamplers();
outputFooter();
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00
exit 0;
# end of sdlgenaudiocvt.pl ...
First shot at new audio data types (int32 and float32). Notable changes: - Converters between types are autogenerated. Instead of making multiple passes over the data with seperate filters for endianess, size, signedness, etc, converting between data types is always one specialized filter. This simplifies SDL_BuildAudioCVT(), which otherwise had a million edge cases with the new types, and makes the actually conversions more CPU cache friendly. Left a stub for adding specific optimized versions of these routines (SSE/MMX/Altivec, assembler, etc) - Autogenerated converters are built by SDL/src/audio/sdlgenaudiocvt.pl. This does not need to be run unless tweaking the code, and thus doesn't need integration into the build system. - Went through all the drivers and tried to weed out all the "Uint16" references that are better specified with the new SDL_AudioFormat typedef. - Cleaned out a bunch of hardcoded bitwise magic numbers and replaced them with new SDL_AUDIO_* macros. - Added initial float32 and int32 support code. Theoretically, existing drivers will push these through converters to get the data they want to feed to the hardware. Still TODO: - Optimize and debug new converters. - Update the CoreAudio backend to accept float32 data directly. - Other backends, too? - SDL_LoadWAV() needs to be updated to support int32 and float32 .wav files (both of which exist and can be generated by 'sox' for testing purposes). - Update the mixer to handle new datatypes. - Optionally update SDL_sound and SDL_mixer, etc. --HG-- extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%402029
2006-08-24 12:10:46 +00:00