diff options
author | Christoph Egger <Christoph.Egger@gmx.de> | 2010-04-07 22:00:13 +0200 |
---|---|---|
committer | Christoph Egger <Christoph.Egger@gmx.de> | 2010-04-07 22:00:13 +0200 |
commit | 46bf1cf82d855660c212ee3421dff3050ec8ffea (patch) | |
tree | f0fadfc004a86242ad096dee015ed7c20cc7f554 /src/SFML | |
parent | 25177f23f35afa0e6ebdd63d581003c298d3a7df (diff) |
Imported Upstream version 1.6+repack1
Diffstat (limited to 'src/SFML')
44 files changed, 6650 insertions, 6122 deletions
diff --git a/src/SFML/Audio/Makefile b/src/SFML/Audio/Makefile index 109a00a..0215e80 100755 --- a/src/SFML/Audio/Makefile +++ b/src/SFML/Audio/Makefile @@ -10,7 +10,7 @@ ifeq ($(STATIC), yes) else
LIB = libsfml-audio.so
LIBNAME = $(LIBPATH)/$(LIB).$(VERSION)
- INSTALL = && $(LN) $(LNFLAGS) $(DESTLIBDIR)/$(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
+ INSTALL = && $(LN) $(LNFLAGS) $(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
endif
all: $(LIB)
diff --git a/src/SFML/Audio/Sound.cpp b/src/SFML/Audio/Sound.cpp index 51a3bd9..84136db 100755 --- a/src/SFML/Audio/Sound.cpp +++ b/src/SFML/Audio/Sound.cpp @@ -46,15 +46,15 @@ Sound::Sound() /// Construct the sound from its parameters
////////////////////////////////////////////////////////////
Sound::Sound(const SoundBuffer& Buffer, bool Loop, float Pitch, float Volume, const Vector3f& Position) :
-myBuffer(&Buffer)
+myBuffer(NULL)
{
ALCheck(alGenSources(1, &mySource));
- ALCheck(alSourcei (mySource, AL_BUFFER, Buffer.myBuffer));
- ALCheck(alSourcei (mySource, AL_LOOPING, Loop));
- ALCheck(alSourcef (mySource, AL_PITCH, Pitch));
- ALCheck(alSourcef (mySource, AL_GAIN, Volume * 0.01f));
- ALCheck(alSource3f(mySource, AL_POSITION, Position.x, Position.y, Position.z));
+ SetBuffer(Buffer);
+ SetLoop(Loop);
+ SetPitch(Pitch);
+ SetVolume(Volume);
+ SetPosition(Position);
}
@@ -63,15 +63,19 @@ myBuffer(&Buffer) ////////////////////////////////////////////////////////////
Sound::Sound(const Sound& Copy) :
AudioResource(Copy),
-myBuffer (Copy.myBuffer)
+myBuffer(NULL)
{
ALCheck(alGenSources(1, &mySource));
- ALCheck(alSourcei (mySource, AL_BUFFER, myBuffer ? myBuffer->myBuffer : 0));
- ALCheck(alSourcei (mySource, AL_LOOPING, Copy.GetLoop()));
- ALCheck(alSourcef (mySource, AL_PITCH, Copy.GetPitch()));
- ALCheck(alSourcef (mySource, AL_GAIN, Copy.GetVolume() * 0.01f));
- ALCheck(alSource3f(mySource, AL_POSITION, Copy.GetPosition().x, Copy.GetPosition().y, Copy.GetPosition().z));
+ if (Copy.myBuffer)
+ SetBuffer(*Copy.myBuffer);
+ SetLoop(Copy.GetLoop());
+ SetPitch(Copy.GetPitch());
+ SetVolume(Copy.GetVolume());
+ SetPosition(Copy.GetPosition());
+ SetRelativeToListener(Copy.IsRelativeToListener());
+ SetMinDistance(Copy.GetMinDistance());
+ SetAttenuation(Copy.GetAttenuation());
}
@@ -86,6 +90,7 @@ Sound::~Sound() {
Stop();
ALCheck(alSourcei(mySource, AL_BUFFER, 0));
+ myBuffer->DetachSound(this);
}
ALCheck(alDeleteSources(1, &mySource));
}
@@ -124,8 +129,17 @@ void Sound::Stop() ////////////////////////////////////////////////////////////
void Sound::SetBuffer(const SoundBuffer& Buffer)
{
+ // First detach from the previous buffer
+ if (myBuffer)
+ {
+ Stop();
+ myBuffer->DetachSound(this);
+ }
+
+ // Assign and use the new buffer
myBuffer = &Buffer;
- ALCheck(alSourcei(mySource, AL_BUFFER, myBuffer ? myBuffer->myBuffer : 0));
+ myBuffer->AttachSound(this);
+ ALCheck(alSourcei(mySource, AL_BUFFER, myBuffer->myBuffer));
}
@@ -348,12 +362,43 @@ Sound::Status Sound::GetStatus() const ////////////////////////////////////////////////////////////
Sound& Sound::operator =(const Sound& Other)
{
- Sound Temp(Other);
+ // Here we don't use the copy-and-swap idiom, because it would mess up
+ // the list of sound instances contained in the buffers
- std::swap(mySource, Temp.mySource);
- std::swap(myBuffer, Temp.myBuffer);
+ // Detach the sound instance from the previous buffer (if any)
+ if (myBuffer)
+ {
+ Stop();
+ myBuffer->DetachSound(this);
+ myBuffer = NULL;
+ }
+
+ // Copy the sound attributes
+ if (Other.myBuffer)
+ SetBuffer(*Other.myBuffer);
+ SetLoop(Other.GetLoop());
+ SetPitch(Other.GetPitch());
+ SetVolume(Other.GetVolume());
+ SetPosition(Other.GetPosition());
+ SetRelativeToListener(Other.IsRelativeToListener());
+ SetMinDistance(Other.GetMinDistance());
+ SetAttenuation(Other.GetAttenuation());
return *this;
}
+
+////////////////////////////////////////////////////////////
+/// Reset the internal buffer
+////////////////////////////////////////////////////////////
+void Sound::ResetBuffer()
+{
+ // First stop the sound in case it is playing
+ Stop();
+
+ // Detach the buffer
+ ALCheck(alSourcei(mySource, AL_BUFFER, 0));
+ myBuffer = NULL;
+}
+
} // namespace sf
diff --git a/src/SFML/Audio/SoundBuffer.cpp b/src/SFML/Audio/SoundBuffer.cpp index 7636a01..e726b13 100755 --- a/src/SFML/Audio/SoundBuffer.cpp +++ b/src/SFML/Audio/SoundBuffer.cpp @@ -27,6 +27,7 @@ ////////////////////////////////////////////////////////////
#include <SFML/Audio/SoundBuffer.hpp>
#include <SFML/Audio/SoundFile.hpp>
+#include <SFML/Audio/Sound.hpp>
#include <SFML/Audio/AudioDevice.hpp>
#include <SFML/Audio/OpenAL.hpp>
#include <iostream>
@@ -55,7 +56,8 @@ AudioResource (Copy), Resource<SoundBuffer>(Copy),
myBuffer (0),
mySamples (Copy.mySamples),
-myDuration (Copy.myDuration)
+myDuration (Copy.myDuration),
+mySounds () // don't copy the attached sounds
{
// Create the buffer
ALCheck(alGenBuffers(1, &myBuffer));
@@ -70,6 +72,11 @@ myDuration (Copy.myDuration) ////////////////////////////////////////////////////////////
SoundBuffer::~SoundBuffer()
{
+ // First detach the buffer from the sounds that use it (to avoid OpenAL errors)
+ for (SoundList::const_iterator it = mySounds.begin(); it != mySounds.end(); ++it)
+ (*it)->ResetBuffer();
+
+ // Destroy the buffer
if (myBuffer)
ALCheck(alDeleteBuffers(1, &myBuffer));
}
@@ -268,9 +275,10 @@ SoundBuffer& SoundBuffer::operator =(const SoundBuffer& Other) {
SoundBuffer Temp(Other);
- mySamples.swap(Temp.mySamples);
+ std::swap(mySamples, Temp.mySamples);
std::swap(myBuffer, Temp.myBuffer);
std::swap(myDuration, Temp.myDuration);
+ std::swap(mySounds, Temp.mySounds); // swap sounds too, so that they are detached when Temp is destroyed
return *this;
}
@@ -305,4 +313,22 @@ bool SoundBuffer::Update(unsigned int ChannelsCount, unsigned int SampleRate) return true;
}
+
+////////////////////////////////////////////////////////////
+/// Add a sound to the list of sounds that use this buffer
+////////////////////////////////////////////////////////////
+void SoundBuffer::AttachSound(Sound* Instance) const
+{
+ mySounds.insert(Instance);
+}
+
+
+////////////////////////////////////////////////////////////
+/// Remove a sound from the list of sounds that use this buffer
+////////////////////////////////////////////////////////////
+void SoundBuffer::DetachSound(Sound* Instance) const
+{
+ mySounds.erase(Instance);
+}
+
} // namespace sf
diff --git a/src/SFML/Audio/SoundRecorder.cpp b/src/SFML/Audio/SoundRecorder.cpp index ac787b2..c12a49d 100755 --- a/src/SFML/Audio/SoundRecorder.cpp +++ b/src/SFML/Audio/SoundRecorder.cpp @@ -137,7 +137,8 @@ bool SoundRecorder::CanCapture() {
ALCdevice* Device = priv::AudioDevice::GetInstance().GetDevice();
- return alcIsExtensionPresent(Device, "ALC_EXT_CAPTURE") != AL_FALSE;
+ return (alcIsExtensionPresent(Device, "ALC_EXT_CAPTURE") != AL_FALSE) ||
+ (alcIsExtensionPresent(Device, "ALC_EXT_capture") != AL_FALSE); // "bug" in Mac OS X 10.5 and 10.6
}
diff --git a/src/SFML/Audio/SoundStream.cpp b/src/SFML/Audio/SoundStream.cpp index 4fb2ff2..9530d78 100755 --- a/src/SFML/Audio/SoundStream.cpp +++ b/src/SFML/Audio/SoundStream.cpp @@ -188,9 +188,10 @@ bool SoundStream::GetLoop() const ////////////////////////////////////////////////////////////
void SoundStream::Run()
{
- // Create buffers
+ // Create the buffers
ALCheck(alGenBuffers(BuffersCount, myBuffers));
- unsigned int EndBuffer = 0xFFFF;
+ for (int i = 0; i < BuffersCount; ++i)
+ myEndBuffers[i] = false;
// Fill the queue
bool RequestStop = FillQueue();
@@ -225,12 +226,21 @@ void SoundStream::Run() ALuint Buffer;
ALCheck(alSourceUnqueueBuffers(Sound::mySource, 1, &Buffer));
+ // Find its number
+ unsigned int BufferNum = 0;
+ for (int i = 0; i < BuffersCount; ++i)
+ if (myBuffers[i] == Buffer)
+ {
+ BufferNum = i;
+ break;
+ }
+
// Retrieve its size and add it to the samples count
- if (Buffer == EndBuffer)
+ if (myEndBuffers[BufferNum])
{
// This was the last buffer: reset the sample count
mySamplesProcessed = 0;
- EndBuffer = 0xFFFF;
+ myEndBuffers[BufferNum] = false;
}
else
{
@@ -242,21 +252,8 @@ void SoundStream::Run() // Fill it and push it back into the playing queue
if (!RequestStop)
{
- if (FillAndPushBuffer(Buffer))
- {
- // User requested to stop: check if we must loop or really stop
- if (myLoop && OnStart())
- {
- // Looping: mark the current buffer as the last one
- // (to know when to reset the sample count)
- EndBuffer = Buffer;
- }
- else
- {
- // Not looping or restart failed: request stop
- RequestStop = true;
- }
- }
+ if (FillAndPushBuffer(BufferNum))
+ RequestStop = true;
}
}
@@ -281,18 +278,38 @@ void SoundStream::Run() /// Fill a new buffer with audio data, and push it to the
/// playing queue
////////////////////////////////////////////////////////////
-bool SoundStream::FillAndPushBuffer(unsigned int Buffer)
+bool SoundStream::FillAndPushBuffer(unsigned int BufferNum)
{
bool RequestStop = false;
// Acquire audio data
Chunk Data = {NULL, 0};
if (!OnGetData(Data))
- RequestStop = true;
+ {
+ // Mark the buffer as the last one (so that we know when to reset the playing position)
+ myEndBuffers[BufferNum] = true;
+
+ // Check if the stream must loop or stop
+ if (myLoop && OnStart())
+ {
+ // If we succeeded to restart and we previously had no data, try to fill the buffer once again
+ if (!Data.Samples || (Data.NbSamples == 0))
+ {
+ return FillAndPushBuffer(BufferNum);
+ }
+ }
+ else
+ {
+ // Not looping or restart failed: request stop
+ RequestStop = true;
+ }
+ }
- // Create and fill the buffer, and push it to the queue
+ // Fill the buffer if some data was returned
if (Data.Samples && Data.NbSamples)
{
+ unsigned int Buffer = myBuffers[BufferNum];
+
// Fill the buffer
ALsizei Size = static_cast<ALsizei>(Data.NbSamples) * sizeof(Int16);
ALCheck(alBufferData(Buffer, myFormat, Data.Samples, Size, mySampleRate));
@@ -314,7 +331,7 @@ bool SoundStream::FillQueue() bool RequestStop = false;
for (int i = 0; (i < BuffersCount) && !RequestStop; ++i)
{
- if (FillAndPushBuffer(myBuffers[i]))
+ if (FillAndPushBuffer(i))
RequestStop = true;
}
diff --git a/src/SFML/Audio/stb_vorbis/stb_vorbis.c b/src/SFML/Audio/stb_vorbis/stb_vorbis.c index e34405f..902761a 100755 --- a/src/SFML/Audio/stb_vorbis/stb_vorbis.c +++ b/src/SFML/Audio/stb_vorbis/stb_vorbis.c @@ -1,5039 +1,5039 @@ -// Ogg Vorbis I audio decoder -- version 0.99994 -// -// Written in April 2007 by Sean Barrett, sponsored by RAD Game Tools. -// -// Placed in the public domain April 2007 by the author: no copyright is -// claimed, and you may use it for any purpose you like. -// -// No warranty for any purpose is expressed or implied by the author (nor -// by RAD Game Tools). Report bugs and send enhancements to the author. -// -// Get the latest version and other information at: -// http://nothings.org/stb_vorbis/ - - -// Todo: -// -// - seeking (note you can seek yourself using the pushdata API) -// -// Limitations: -// -// - floor 0 not supported (used in old ogg vorbis files) -// - lossless sample-truncation at beginning ignored -// - cannot concatenate multiple vorbis streams -// - sample positions are 32-bit, limiting seekable 192Khz -// files to around 6 hours (Ogg supports 64-bit) -// -// All of these limitations may be removed in future versions. - -#include "stb_vorbis.h" - -#ifndef STB_VORBIS_HEADER_ONLY - -// global configuration settings (e.g. set these in the project/makefile), -// or just set them in this file at the top (although ideally the first few -// should be visible when the header file is compiled too, although it's not -// crucial) - -// STB_VORBIS_NO_PUSHDATA_API -// does not compile the code for the various stb_vorbis_*_pushdata() -// functions -// #define STB_VORBIS_NO_PUSHDATA_API - -// STB_VORBIS_NO_PULLDATA_API -// does not compile the code for the non-pushdata APIs -// #define STB_VORBIS_NO_PULLDATA_API - -// STB_VORBIS_NO_STDIO -// does not compile the code for the APIs that use FILE *s internally -// or externally (implied by STB_VORBIS_NO_PULLDATA_API) -// #define STB_VORBIS_NO_STDIO - -// STB_VORBIS_NO_INTEGER_CONVERSION -// does not compile the code for converting audio sample data from -// float to integer (implied by STB_VORBIS_NO_PULLDATA_API) -// #define STB_VORBIS_NO_INTEGER_CONVERSION - -// STB_VORBIS_NO_FAST_SCALED_FLOAT -// does not use a fast float-to-int trick to accelerate float-to-int on -// most platforms which requires endianness be defined correctly. -#define STB_VORBIS_NO_FAST_SCALED_FLOAT - - -// STB_VORBIS_MAX_CHANNELS [number] -// globally define this to the maximum number of channels you need. -// The spec does not put a restriction on channels except that -// the count is stored in a byte, so 255 is the hard limit. -// Reducing this saves about 16 bytes per value, so using 16 saves -// (255-16)*16 or around 4KB. Plus anything other memory usage -// I forgot to account for. Can probably go as low as 8 (7.1 audio), -// 6 (5.1 audio), or 2 (stereo only). -#ifndef STB_VORBIS_MAX_CHANNELS -#define STB_VORBIS_MAX_CHANNELS 16 // enough for anyone? -#endif - -// STB_VORBIS_PUSHDATA_CRC_COUNT [number] -// after a flush_pushdata(), stb_vorbis begins scanning for the -// next valid page, without backtracking. when it finds something -// that looks like a page, it streams through it and verifies its -// CRC32. Should that validation fail, it keeps scanning. But it's -// possible that _while_ streaming through to check the CRC32 of -// one candidate page, it sees another candidate page. This #define -// determines how many "overlapping" candidate pages it can search -// at once. Note that "real" pages are typically ~4KB to ~8KB, whereas -// garbage pages could be as big as 64KB, but probably average ~16KB. -// So don't hose ourselves by scanning an apparent 64KB page and -// missing a ton of real ones in the interim; so minimum of 2 -#ifndef STB_VORBIS_PUSHDATA_CRC_COUNT -#define STB_VORBIS_PUSHDATA_CRC_COUNT 4 -#endif - -// STB_VORBIS_FAST_HUFFMAN_LENGTH [number] -// sets the log size of the huffman-acceleration table. Maximum -// supported value is 24. with larger numbers, more decodings are O(1), -// but the table size is larger so worse cache missing, so you'll have -// to probe (and try multiple ogg vorbis files) to find the sweet spot. -#ifndef STB_VORBIS_FAST_HUFFMAN_LENGTH -#define STB_VORBIS_FAST_HUFFMAN_LENGTH 10 -#endif - -// STB_VORBIS_FAST_BINARY_LENGTH [number] -// sets the log size of the binary-search acceleration table. this -// is used in similar fashion to the fast-huffman size to set initial -// parameters for the binary search - -// STB_VORBIS_FAST_HUFFMAN_INT -// The fast huffman tables are much more efficient if they can be -// stored as 16-bit results instead of 32-bit results. This restricts -// the codebooks to having only 65535 possible outcomes, though. -// (At least, accelerated by the huffman table.) -#ifndef STB_VORBIS_FAST_HUFFMAN_INT -#define STB_VORBIS_FAST_HUFFMAN_SHORT -#endif - -// STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH -// If the 'fast huffman' search doesn't succeed, then stb_vorbis falls -// back on binary searching for the correct one. This requires storing -// extra tables with the huffman codes in sorted order. Defining this -// symbol trades off space for speed by forcing a linear search in the -// non-fast case, except for "sparse" codebooks. -// #define STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH - -// STB_VORBIS_DIVIDES_IN_RESIDUE -// stb_vorbis precomputes the result of the scalar residue decoding -// that would otherwise require a divide per chunk. you can trade off -// space for time by defining this symbol. -// #define STB_VORBIS_DIVIDES_IN_RESIDUE - -// STB_VORBIS_DIVIDES_IN_CODEBOOK -// vorbis VQ codebooks can be encoded two ways: with every case explicitly -// stored, or with all elements being chosen from a small range of values, -// and all values possible in all elements. By default, stb_vorbis expands -// this latter kind out to look like the former kind for ease of decoding, -// because otherwise an integer divide-per-vector-element is required to -// unpack the index. If you define STB_VORBIS_DIVIDES_IN_CODEBOOK, you can -// trade off storage for speed. -//#define STB_VORBIS_DIVIDES_IN_CODEBOOK - -// STB_VORBIS_CODEBOOK_SHORTS -// The vorbis file format encodes VQ codebook floats as ax+b where a and -// b are floating point per-codebook constants, and x is a 16-bit int. -// Normally, stb_vorbis decodes them to floats rather than leaving them -// as 16-bit ints and computing ax+b while decoding. This is a speed/space -// tradeoff; you can save space by defining this flag. -#ifndef STB_VORBIS_CODEBOOK_SHORTS -#define STB_VORBIS_CODEBOOK_FLOATS -#endif - -// STB_VORBIS_DIVIDE_TABLE -// this replaces small integer divides in the floor decode loop with -// table lookups. made less than 1% difference, so disabled by default. - -// STB_VORBIS_NO_INLINE_DECODE -// disables the inlining of the scalar codebook fast-huffman decode. -// might save a little codespace; useful for debugging -// #define STB_VORBIS_NO_INLINE_DECODE - -// STB_VORBIS_NO_DEFER_FLOOR -// Normally we only decode the floor without synthesizing the actual -// full curve. We can instead synthesize the curve immediately. This -// requires more memory and is very likely slower, so I don't think -// you'd ever want to do it except for debugging. -// #define STB_VORBIS_NO_DEFER_FLOOR - - - - -////////////////////////////////////////////////////////////////////////////// - -#ifdef STB_VORBIS_NO_PULLDATA_API - #define STB_VORBIS_NO_INTEGER_CONVERSION - #define STB_VORBIS_NO_STDIO -#endif - -#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) - #define STB_VORBIS_NO_STDIO 1 -#endif - -#ifndef STB_VORBIS_NO_INTEGER_CONVERSION -#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT - - // only need endianness for fast-float-to-int, which we don't - // use for pushdata - - #ifndef STB_VORBIS_BIG_ENDIAN - #define STB_VORBIS_ENDIAN 0 - #else - #define STB_VORBIS_ENDIAN 1 - #endif - -#endif -#endif - - -#ifndef STB_VORBIS_NO_STDIO -#include <stdio.h> -#endif - -#ifndef STB_VORBIS_NO_CRT -#include <stdlib.h> -#include <string.h> -#include <assert.h> -#include <math.h> - -#if !defined(__APPLE__) && !defined(MACOSX) && !defined(macintosh) && !defined(Macintosh) &&!defined(__FreeBSD__) -#include <malloc.h> -#endif - -#else -#define NULL 0 -#endif - -#ifndef _MSC_VER - #if __GNUC__ - #define __forceinline inline - #else - #define __forceinline - #endif -#endif - -#if STB_VORBIS_MAX_CHANNELS > 256 -#error "Value of STB_VORBIS_MAX_CHANNELS outside of allowed range" -#endif - -#if STB_VORBIS_FAST_HUFFMAN_LENGTH > 24 -#error "Value of STB_VORBIS_FAST_HUFFMAN_LENGTH outside of allowed range" -#endif - - -#define MAX_BLOCKSIZE_LOG 13 // from specification -#define MAX_BLOCKSIZE (1 << MAX_BLOCKSIZE_LOG) - - -typedef unsigned char uint8; -typedef signed char int8; -typedef unsigned short uint16; -typedef signed short int16; -typedef unsigned int uint32; -typedef signed int int32; - -#ifndef TRUE -#define TRUE 1 -#define FALSE 0 -#endif - -#ifdef STB_VORBIS_CODEBOOK_FLOATS -typedef float codetype; -#else -typedef uint16 codetype; -#endif - -// @NOTE -// -// Some arrays below are tagged "//varies", which means it's actually -// a variable-sized piece of data, but rather than malloc I assume it's -// small enough it's better to just allocate it all together with the -// main thing -// -// Most of the variables are specified with the smallest size I could pack -// them into. It might give better performance to make them all full-sized -// integers. It should be safe to freely rearrange the structures or change -// the sizes larger--nothing relies on silently truncating etc., nor the -// order of variables. - -#define FAST_HUFFMAN_TABLE_SIZE (1 << STB_VORBIS_FAST_HUFFMAN_LENGTH) -#define FAST_HUFFMAN_TABLE_MASK (FAST_HUFFMAN_TABLE_SIZE - 1) - -typedef struct -{ - int dimensions, entries; - uint8 *codeword_lengths; - float minimum_value; - float delta_value; - uint8 value_bits; - uint8 lookup_type; - uint8 sequence_p; - uint8 sparse; - uint32 lookup_values; - codetype *multiplicands; - uint32 *codewords; - #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT - int16 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; - #else - int32 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; - #endif - uint32 *sorted_codewords; - int *sorted_values; - int sorted_entries; -} Codebook; - -typedef struct -{ - uint8 order; - uint16 rate; - uint16 bark_map_size; - uint8 amplitude_bits; - uint8 amplitude_offset; - uint8 number_of_books; - uint8 book_list[16]; // varies -} Floor0; - -typedef struct -{ - uint8 partitions; - uint8 partition_class_list[32]; // varies - uint8 class_dimensions[16]; // varies - uint8 class_subclasses[16]; // varies - uint8 class_masterbooks[16]; // varies - int16 subclass_books[16][8]; // varies - uint16 Xlist[31*8+2]; // varies - uint8 sorted_order[31*8+2]; - uint8 neighbors[31*8+2][2]; - uint8 floor1_multiplier; - uint8 rangebits; - int values; -} Floor1; - -typedef union -{ - Floor0 floor0; - Floor1 floor1; -} Floor; - -typedef struct -{ - uint32 begin, end; - uint32 part_size; - uint8 classifications; - uint8 classbook; - uint8 **classdata; - int16 (*residue_books)[8]; -} Residue; - -typedef struct -{ - uint8 magnitude; - uint8 angle; - uint8 mux; -} MappingChannel; - -typedef struct -{ - uint16 coupling_steps; - MappingChannel *chan; - uint8 submaps; - uint8 submap_floor[15]; // varies - uint8 submap_residue[15]; // varies -} Mapping; - -typedef struct -{ - uint8 blockflag; - uint8 mapping; - uint16 windowtype; - uint16 transformtype; -} Mode; - -typedef struct -{ - uint32 goal_crc; // expected crc if match - int bytes_left; // bytes left in packet - uint32 crc_so_far; // running crc - int bytes_done; // bytes processed in _current_ chunk - uint32 sample_loc; // granule pos encoded in page -} CRCscan; - -typedef struct -{ - uint32 page_start, page_end; - uint32 after_previous_page_start; - uint32 first_decoded_sample; - uint32 last_decoded_sample; -} ProbedPage; - -struct stb_vorbis -{ - // user-accessible info - unsigned int sample_rate; - int channels; - - unsigned int setup_memory_required; - unsigned int temp_memory_required; - unsigned int setup_temp_memory_required; - - // input config -#ifndef STB_VORBIS_NO_STDIO - FILE *f; - uint32 f_start; - int close_on_free; -#endif - - uint8 *stream; - uint8 *stream_start; - uint8 *stream_end; - - uint32 stream_len; - - uint8 push_mode; - - uint32 first_audio_page_offset; - - ProbedPage p_first, p_last; - - // memory management - stb_vorbis_alloc alloc; - int setup_offset; - int temp_offset; - - // run-time results - int eof; - enum STBVorbisError error; - - // user-useful data - - // header info - int blocksize[2]; - int blocksize_0, blocksize_1; - int codebook_count; - Codebook *codebooks; - int floor_count; - uint16 floor_types[64]; // varies - Floor *floor_config; - int residue_count; - uint16 residue_types[64]; // varies - Residue *residue_config; - int mapping_count; - Mapping *mapping; - int mode_count; - Mode mode_config[64]; // varies - - uint32 total_samples; - - // decode buffer - float *channel_buffers[STB_VORBIS_MAX_CHANNELS]; - float *outputs [STB_VORBIS_MAX_CHANNELS]; - - float *previous_window[STB_VORBIS_MAX_CHANNELS]; - int previous_length; - - #ifndef STB_VORBIS_NO_DEFER_FLOOR - int16 *finalY[STB_VORBIS_MAX_CHANNELS]; - #else - float *floor_buffers[STB_VORBIS_MAX_CHANNELS]; - #endif - - uint32 current_loc; // sample location of next frame to decode - int current_loc_valid; - - // per-blocksize precomputed data - - // twiddle factors - float *A[2],*B[2],*C[2]; - float *window[2]; - uint16 *bit_reverse[2]; - - // current page/packet/segment streaming info - uint32 serial; // stream serial number for verification - int last_page; - int segment_count; - uint8 segments[255]; - uint8 page_flag; - uint8 bytes_in_seg; - uint8 first_decode; - int next_seg; - int last_seg; // flag that we're on the last segment - int last_seg_which; // what was the segment number of the last seg? - uint32 acc; - int valid_bits; - int packet_bytes; - int end_seg_with_known_loc; - uint32 known_loc_for_packet; - int discard_samples_deferred; - uint32 samples_output; - - // push mode scanning - int page_crc_tests; // only in push_mode: number of tests active; -1 if not searching -#ifndef STB_VORBIS_NO_PUSHDATA_API - CRCscan scan[STB_VORBIS_PUSHDATA_CRC_COUNT]; -#endif - - // sample-access - int channel_buffer_start; - int channel_buffer_end; -}; - -extern int my_prof(int slot); -//#define stb_prof my_prof - -#ifndef stb_prof -#define stb_prof(x) 0 -#endif - -#if defined(STB_VORBIS_NO_PUSHDATA_API) - #define IS_PUSH_MODE(f) FALSE -#elif defined(STB_VORBIS_NO_PULLDATA_API) - #define IS_PUSH_MODE(f) TRUE -#else - #define IS_PUSH_MODE(f) ((f)->push_mode) -#endif - -typedef struct stb_vorbis vorb; - -static int error(vorb *f, enum STBVorbisError e) -{ - f->error = e; - if (!f->eof && e != VORBIS_need_more_data) { - f->error=e; // breakpoint for debugging - } - return 0; -} - - -// these functions are used for allocating temporary memory -// while decoding. if you can afford the stack space, use -// alloca(); otherwise, provide a temp buffer and it will -// allocate out of those. - -#define array_size_required(count,size) (count*(sizeof(void *)+(size))) - -#define temp_alloc(f,size) (f->alloc.alloc_buffer ? setup_temp_malloc(f,size) : alloca(size)) -#ifdef dealloca -#define temp_free(f,p) (f->alloc.alloc_buffer ? 0 : dealloca(size)) -#else -#define temp_free(f,p) 0 -#endif -#define temp_alloc_save(f) ((f)->temp_offset) -#define temp_alloc_restore(f,p) ((f)->temp_offset = (p)) - -#define temp_block_array(f,count,size) make_block_array(temp_alloc(f,array_size_required(count,size)), count, size) - -// given a sufficiently large block of memory, make an array of pointers to subblocks of it -static void *make_block_array(void *mem, int count, int size) -{ - int i; - void ** p = (void **) mem; - char *q = (char *) (p + count); - for (i=0; i < count; ++i) { - p[i] = q; - q += size; - } - return p; -} - -static void *setup_malloc(vorb *f, int sz) -{ - sz = (sz+3) & ~3; - f->setup_memory_required += sz; - if (f->alloc.alloc_buffer) { - void *p = (char *) f->alloc.alloc_buffer + f->setup_offset; - if (f->setup_offset + sz > f->temp_offset) return NULL; - f->setup_offset += sz; - return p; - } - return sz ? malloc(sz) : NULL; -} - -static void setup_free(vorb *f, void *p) -{ - if (f->alloc.alloc_buffer) return; // do nothing; setup mem is not a stack - free(p); -} - -static void *setup_temp_malloc(vorb *f, int sz) -{ - sz = (sz+3) & ~3; - if (f->alloc.alloc_buffer) { - if (f->temp_offset - sz < f->setup_offset) return NULL; - f->temp_offset -= sz; - return (char *) f->alloc.alloc_buffer + f->temp_offset; - } - return malloc(sz); -} - -static void setup_temp_free(vorb *f, void *p, size_t sz) -{ - if (f->alloc.alloc_buffer) { - f->temp_offset += (sz+3)&~3; - return; - } - free(p); -} - -#define CRC32_POLY 0x04c11db7 // from spec - -static uint32 crc_table[256]; -static void crc32_init(void) -{ - int i,j; - uint32 s; - for(i=0; i < 256; i++) { - for (s=i<<24, j=0; j < 8; ++j) - s = (s << 1) ^ (s >= (1<<31) ? CRC32_POLY : 0); - crc_table[i] = s; - } -} - -static __forceinline uint32 crc32_update(uint32 crc, uint8 byte) -{ - return (crc << 8) ^ crc_table[byte ^ (crc >> 24)]; -} - - -// used in setup, and for huffman that doesn't go fast path -static unsigned int bit_reverse(unsigned int n) -{ - n = ((n & 0xAAAAAAAA) >> 1) | ((n & 0x55555555) << 1); - n = ((n & 0xCCCCCCCC) >> 2) | ((n & 0x33333333) << 2); - n = ((n & 0xF0F0F0F0) >> 4) | ((n & 0x0F0F0F0F) << 4); - n = ((n & 0xFF00FF00) >> 8) | ((n & 0x00FF00FF) << 8); - return (n >> 16) | (n << 16); -} - -static float square(float x) -{ - return x*x; -} - -// this is a weird definition of log2() for which log2(1) = 1, log2(2) = 2, log2(4) = 3 -// as required by the specification. fast(?) implementation from stb.h -// @OPTIMIZE: called multiple times per-packet with "constants"; move to setup -static int ilog(int32 n) -{ - static signed char log2_4[16] = { 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4 }; - - // 2 compares if n < 16, 3 compares otherwise (4 if signed or n > 1<<29) - if (n < (1U << 14)) - if (n < (1U << 4)) return 0 + log2_4[n ]; - else if (n < (1U << 9)) return 5 + log2_4[n >> 5]; - else return 10 + log2_4[n >> 10]; - else if (n < (1U << 24)) - if (n < (1U << 19)) return 15 + log2_4[n >> 15]; - else return 20 + log2_4[n >> 20]; - else if (n < (1U << 29)) return 25 + log2_4[n >> 25]; - else if (n < (1U << 31)) return 30 + log2_4[n >> 30]; - else return 0; // signed n returns 0 -} - -#ifndef M_PI - #define M_PI 3.14159265358979323846264f // from CRC -#endif - -// code length assigned to a value with no huffman encoding -#define NO_CODE 255 - -/////////////////////// LEAF SETUP FUNCTIONS ////////////////////////// -// -// these functions are only called at setup, and only a few times -// per file - -static float float32_unpack(uint32 x) -{ - // from the specification - uint32 mantissa = x & 0x1fffff; - uint32 sign = x & 0x80000000; - uint32 exp = (x & 0x7fe00000) >> 21; - double res = sign ? -(double)mantissa : (double)mantissa; - return (float) ldexp((float)res, exp-788); -} - - -// zlib & jpeg huffman tables assume that the output symbols -// can either be arbitrarily arranged, or have monotonically -// increasing frequencies--they rely on the lengths being sorted; -// this makes for a very simple generation algorithm. -// vorbis allows a huffman table with non-sorted lengths. This -// requires a more sophisticated construction, since symbols in -// order do not map to huffman codes "in order". -static void add_entry(Codebook *c, uint32 huff_code, int symbol, int count, int len, uint32 *values) -{ - if (!c->sparse) { - c->codewords [symbol] = huff_code; - } else { - c->codewords [count] = huff_code; - c->codeword_lengths[count] = len; - values [count] = symbol; - } -} - -static int compute_codewords(Codebook *c, uint8 *len, int n, uint32 *values) -{ - int i,k,m=0; - uint32 available[32]; - - memset(available, 0, sizeof(available)); - // find the first entry - for (k=0; k < n; ++k) if (len[k] < NO_CODE) break; - if (k == n) { assert(c->sorted_entries == 0); return TRUE; } - // add to the list - add_entry(c, 0, k, m++, len[k], values); - // add all available leaves - for (i=1; i <= len[k]; ++i) - available[i] = 1 << (32-i); - // note that the above code treats the first case specially, - // but it's really the same as the following code, so they - // could probably be combined (except the initial code is 0, - // and I use 0 in available[] to mean 'empty') - for (i=k+1; i < n; ++i) { - uint32 res; - int z = len[i], y; - if (z == NO_CODE) continue; - // find lowest available leaf (should always be earliest, - // which is what the specification calls for) - // note that this property, and the fact we can never have - // more than one free leaf at a given level, isn't totally - // trivial to prove, but it seems true and the assert never - // fires, so! - while (z > 0 && !available[z]) --z; - if (z == 0) { assert(0); return FALSE; } - res = available[z]; - available[z] = 0; - add_entry(c, bit_reverse(res), i, m++, len[i], values); - // propogate availability up the tree - if (z != len[i]) { - for (y=len[i]; y > z; --y) { - assert(available[y] == 0); - available[y] = res + (1 << (32-y)); - } - } - } - return TRUE; -} - -// accelerated huffman table allows fast O(1) match of all symbols -// of length <= STB_VORBIS_FAST_HUFFMAN_LENGTH -static void compute_accelerated_huffman(Codebook *c) -{ - int i, len; - for (i=0; i < FAST_HUFFMAN_TABLE_SIZE; ++i) - c->fast_huffman[i] = -1; - - len = c->sparse ? c->sorted_entries : c->entries; - #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT - if (len > 32767) len = 32767; // largest possible value we can encode! - #endif - for (i=0; i < len; ++i) { - if (c->codeword_lengths[i] <= STB_VORBIS_FAST_HUFFMAN_LENGTH) { - uint32 z = c->sparse ? bit_reverse(c->sorted_codewords[i]) : c->codewords[i]; - // set table entries for all bit combinations in the higher bits - while (z < FAST_HUFFMAN_TABLE_SIZE) { - c->fast_huffman[z] = i; - z += 1 << c->codeword_lengths[i]; - } - } - } -} - -static int uint32_compare(const void *p, const void *q) -{ - uint32 x = * (uint32 *) p; - uint32 y = * (uint32 *) q; - return x < y ? -1 : x > y; -} - -static int include_in_sort(Codebook *c, uint8 len) -{ - if (c->sparse) { assert(len != NO_CODE); return TRUE; } - if (len == NO_CODE) return FALSE; - if (len > STB_VORBIS_FAST_HUFFMAN_LENGTH) return TRUE; - return FALSE; -} - -// if the fast table above doesn't work, we want to binary -// search them... need to reverse the bits -static void compute_sorted_huffman(Codebook *c, uint8 *lengths, uint32 *values) -{ - int i, len; - // build a list of all the entries - // OPTIMIZATION: don't include the short ones, since they'll be caught by FAST_HUFFMAN. - // this is kind of a frivolous optimization--I don't see any performance improvement, - // but it's like 4 extra lines of code, so. - if (!c->sparse) { - int k = 0; - for (i=0; i < c->entries; ++i) - if (include_in_sort(c, lengths[i])) - c->sorted_codewords[k++] = bit_reverse(c->codewords[i]); - assert(k == c->sorted_entries); - } else { - for (i=0; i < c->sorted_entries; ++i) - c->sorted_codewords[i] = bit_reverse(c->codewords[i]); - } - - qsort(c->sorted_codewords, c->sorted_entries, sizeof(c->sorted_codewords[0]), uint32_compare); - c->sorted_codewords[c->sorted_entries] = 0xffffffff; - - len = c->sparse ? c->sorted_entries : c->entries; - // now we need to indicate how they correspond; we could either - // #1: sort a different data structure that says who they correspond to - // #2: for each sorted entry, search the original list to find who corresponds - // #3: for each original entry, find the sorted entry - // #1 requires extra storage, #2 is slow, #3 can use binary search! - for (i=0; i < len; ++i) { - int huff_len = c->sparse ? lengths[values[i]] : lengths[i]; - if (include_in_sort(c,huff_len)) { - uint32 code = bit_reverse(c->codewords[i]); - int x=0, n=c->sorted_entries; - while (n > 1) { - // invariant: sc[x] <= code < sc[x+n] - int m = x + (n >> 1); - if (c->sorted_codewords[m] <= code) { - x = m; - n -= (n>>1); - } else { - n >>= 1; - } - } - assert(c->sorted_codewords[x] == code); - if (c->sparse) { - c->sorted_values[x] = values[i]; - c->codeword_lengths[x] = huff_len; - } else { - c->sorted_values[x] = i; - } - } - } -} - -// only run while parsing the header (3 times) -static int vorbis_validate(uint8 *data) -{ - static uint8 vorbis[6] = { 'v', 'o', 'r', 'b', 'i', 's' }; - return memcmp(data, vorbis, 6) == 0; -} - -// called from setup only, once per code book -// (formula implied by specification) -static int lookup1_values(int entries, int dim) -{ - int r = (int) floor(exp((float) log((float) entries) / dim)); - if ((int) floor(pow((float) r+1, dim)) <= entries) // (int) cast for MinGW warning; - ++r; // floor() to avoid _ftol() when non-CRT - assert(pow((float) r+1, dim) > entries); - assert((int) floor(pow((float) r, dim)) <= entries); // (int),floor() as above - return r; -} - -// called twice per file -static void compute_twiddle_factors(int n, float *A, float *B, float *C) -{ - int n4 = n >> 2, n8 = n >> 3; - int k,k2; - - for (k=k2=0; k < n4; ++k,k2+=2) { - A[k2 ] = (float) cos(4*k*M_PI/n); - A[k2+1] = (float) -sin(4*k*M_PI/n); - B[k2 ] = (float) cos((k2+1)*M_PI/n/2) * 0.5f; - B[k2+1] = (float) sin((k2+1)*M_PI/n/2) * 0.5f; - } - for (k=k2=0; k < n8; ++k,k2+=2) { - C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); - C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); - } -} - -static void compute_window(int n, float *window) -{ - int n2 = n >> 1, i; - for (i=0; i < n2; ++i) - window[i] = (float) sin(0.5 * M_PI * square((float) sin((i - 0 + 0.5) / n2 * 0.5 * M_PI))); -} - -static void compute_bitreverse(int n, uint16 *rev) -{ - int ld = ilog(n) - 1; // ilog is off-by-one from normal definitions - int i, n8 = n >> 3; - for (i=0; i < n8; ++i) - rev[i] = (bit_reverse(i) >> (32-ld+3)) << 2; -} - -static int init_blocksize(vorb *f, int b, int n) -{ - int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3; - f->A[b] = (float *) setup_malloc(f, sizeof(float) * n2); - f->B[b] = (float *) setup_malloc(f, sizeof(float) * n2); - f->C[b] = (float *) setup_malloc(f, sizeof(float) * n4); - if (!f->A[b] || !f->B[b] || !f->C[b]) return error(f, VORBIS_outofmem); - compute_twiddle_factors(n, f->A[b], f->B[b], f->C[b]); - f->window[b] = (float *) setup_malloc(f, sizeof(float) * n2); - if (!f->window[b]) return error(f, VORBIS_outofmem); - compute_window(n, f->window[b]); - f->bit_reverse[b] = (uint16 *) setup_malloc(f, sizeof(uint16) * n8); - if (!f->bit_reverse[b]) return error(f, VORBIS_outofmem); - compute_bitreverse(n, f->bit_reverse[b]); - return TRUE; -} - -static void neighbors(uint16 *x, int n, int *plow, int *phigh) -{ - int low = -1; - int high = 65536; - int i; - for (i=0; i < n; ++i) { - if (x[i] > low && x[i] < x[n]) { *plow = i; low = x[i]; } - if (x[i] < high && x[i] > x[n]) { *phigh = i; high = x[i]; } - } -} - -// this has been repurposed so y is now the original index instead of y -typedef struct -{ - uint16 x,y; -} Point; - -int point_compare(const void *p, const void *q) -{ - Point *a = (Point *) p; - Point *b = (Point *) q; - return a->x < b->x ? -1 : a->x > b->x; -} - -// -/////////////////////// END LEAF SETUP FUNCTIONS ////////////////////////// - - -#if defined(STB_VORBIS_NO_STDIO) - #define USE_MEMORY(z) TRUE -#else - #define USE_MEMORY(z) ((z)->stream) -#endif - -static uint8 get8(vorb *z) -{ - if (USE_MEMORY(z)) { - if (z->stream >= z->stream_end) { z->eof = TRUE; return 0; } - return *z->stream++; - } - - #ifndef STB_VORBIS_NO_STDIO - { - int c = fgetc(z->f); - if (c == EOF) { z->eof = TRUE; return 0; } - return c; - } - #endif -} - -static uint32 get32(vorb *f) -{ - uint32 x; - x = get8(f); - x += get8(f) << 8; - x += get8(f) << 16; - x += get8(f) << 24; - return x; -} - -static int getn(vorb *z, uint8 *data, int n) -{ - if (USE_MEMORY(z)) { - if (z->stream+n > z->stream_end) { z->eof = 1; return 0; } - memcpy(data, z->stream, n); - z->stream += n; - return 1; - } - - #ifndef STB_VORBIS_NO_STDIO - if (fread(data, n, 1, z->f) == 1) - return 1; - else { - z->eof = 1; - return 0; - } - #endif -} - -static void skip(vorb *z, int n) -{ - if (USE_MEMORY(z)) { - z->stream += n; - if (z->stream >= z->stream_end) z->eof = 1; - return; - } - #ifndef STB_VORBIS_NO_STDIO - { - long x = ftell(z->f); - fseek(z->f, x+n, SEEK_SET); - } - #endif -} - -static int set_file_offset(stb_vorbis *f, unsigned int loc) -{ - #ifndef STB_VORBIS_NO_PUSHDATA_API - if (f->push_mode) return 0; - #endif - f->eof = 0; - if (USE_MEMORY(f)) { - if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) { - f->stream = f->stream_end; - f->eof = 1; - return 0; - } else { - f->stream = f->stream_start + loc; - return 1; - } - } - #ifndef STB_VORBIS_NO_STDIO - if (loc + f->f_start < loc || loc >= 0x80000000) { - loc = 0x7fffffff; - f->eof = 1; - } else { - loc += f->f_start; - } - if (!fseek(f->f, loc, SEEK_SET)) - return 1; - f->eof = 1; - fseek(f->f, f->f_start, SEEK_END); - return 0; - #endif -} - - -static uint8 ogg_page_header[4] = { 0x4f, 0x67, 0x67, 0x53 }; - -static int capture_pattern(vorb *f) -{ - if (0x4f != get8(f)) return FALSE; - if (0x67 != get8(f)) return FALSE; - if (0x67 != get8(f)) return FALSE; - if (0x53 != get8(f)) return FALSE; - return TRUE; -} - -#define PAGEFLAG_continued_packet 1 -#define PAGEFLAG_first_page 2 -#define PAGEFLAG_last_page 4 - -static int start_page_no_capturepattern(vorb *f) -{ - uint32 loc0,loc1,n,i; - // stream structure version - if (0 != get8(f)) return error(f, VORBIS_invalid_stream_structure_version); - // header flag - f->page_flag = get8(f); - // absolute granule position - loc0 = get32(f); - loc1 = get32(f); - // @TODO: validate loc0,loc1 as valid positions? - // stream serial number -- vorbis doesn't interleave, so discard - get32(f); - //if (f->serial != get32(f)) return error(f, VORBIS_incorrect_stream_serial_number); - // page sequence number - n = get32(f); - f->last_page = n; - // CRC32 - get32(f); - // page_segments - f->segment_count = get8(f); - if (!getn(f, f->segments, f->segment_count)) - return error(f, VORBIS_unexpected_eof); - // assume we _don't_ know any the sample position of any segments - f->end_seg_with_known_loc = -2; - if (loc0 != ~0 || loc1 != ~0) { - // determine which packet is the last one that will complete - for (i=f->segment_count-1; i >= 0; --i) - if (f->segments[i] < 255) - break; - // 'i' is now the index of the _last_ segment of a packet that ends - if (i >= 0) { - f->end_seg_with_known_loc = i; - f->known_loc_for_packet = loc0; - } - } - if (f->first_decode) { - int i,len; - ProbedPage p; - len = 0; - for (i=0; i < f->segment_count; ++i) - len += f->segments[i]; - len += 27 + f->segment_count; - p.page_start = f->first_audio_page_offset; - p.page_end = p.page_start + len; - p.after_previous_page_start = p.page_start; - p.first_decoded_sample = 0; - p.last_decoded_sample = loc0; - f->p_first = p; - } - f->next_seg = 0; - return TRUE; -} - -static int start_page(vorb *f) -{ - if (!capture_pattern(f)) return error(f, VORBIS_missing_capture_pattern); - return start_page_no_capturepattern(f); -} - -static int start_packet(vorb *f) -{ - while (f->next_seg == -1) { - if (!start_page(f)) return FALSE; - if (f->page_flag & PAGEFLAG_continued_packet) - return error(f, VORBIS_continued_packet_flag_invalid); - } - f->last_seg = FALSE; - f->valid_bits = 0; - f->packet_bytes = 0; - f->bytes_in_seg = 0; - // f->next_seg is now valid - return TRUE; -} - -static int maybe_start_packet(vorb *f) -{ - if (f->next_seg == -1) { - int x = get8(f); - if (f->eof) return FALSE; // EOF at page boundary is not an error! - if (0x4f != x ) return error(f, VORBIS_missing_capture_pattern); - if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); - if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern); - if (0x53 != get8(f)) return error(f, VORBIS_missing_capture_pattern); - if (!start_page_no_capturepattern(f)) return FALSE; - if (f->page_flag & PAGEFLAG_continued_packet) { - // set up enough state that we can read this packet if we want, - // e.g. during recovery - f->last_seg = FALSE; - f->bytes_in_seg = 0; - return error(f, VORBIS_continued_packet_flag_invalid); - } - } - return start_packet(f); -} - -static int next_segment(vorb *f) -{ - int len; - if (f->last_seg) return 0; - if (f->next_seg == -1) { - f->last_seg_which = f->segment_count-1; // in case start_page fails - if (!start_page(f)) { f->last_seg = 1; return 0; } - if (!(f->page_flag & PAGEFLAG_continued_packet)) return error(f, VORBIS_continued_packet_flag_invalid); - } - len = f->segments[f->next_seg++]; - if (len < 255) { - f->last_seg = TRUE; - f->last_seg_which = f->next_seg-1; - } - if (f->next_seg >= f->segment_count) - f->next_seg = -1; - assert(f->bytes_in_seg == 0); - f->bytes_in_seg = len; - return len; -} - -#define EOP (-1) -#define INVALID_BITS (-1) - -static int get8_packet_raw(vorb *f) -{ - if (!f->bytes_in_seg) - if (f->last_seg) return EOP; - else if (!next_segment(f)) return EOP; - assert(f->bytes_in_seg > 0); - --f->bytes_in_seg; - ++f->packet_bytes; - return get8(f); -} - -static int get8_packet(vorb *f) -{ - int x = get8_packet_raw(f); - f->valid_bits = 0; - return x; -} - -static void flush_packet(vorb *f) -{ - while (get8_packet_raw(f) != EOP); -} - -// @OPTIMIZE: this is the secondary bit decoder, so it's probably not as important -// as the huffman decoder? -static uint32 get_bits(vorb *f, int n) -{ - uint32 z; - - if (f->valid_bits < 0) return 0; - if (f->valid_bits < n) { - if (n > 24) { - // the accumulator technique below would not work correctly in this case - z = get_bits(f, 24); - z += get_bits(f, n-24) << 24; - return z; - } - if (f->valid_bits == 0) f->acc = 0; - while (f->valid_bits < n) { - int z = get8_packet_raw(f); - if (z == EOP) { - f->valid_bits = INVALID_BITS; - return 0; - } - f->acc += z << f->valid_bits; - f->valid_bits += 8; - } - } - if (f->valid_bits < 0) return 0; - z = f->acc & ((1 << n)-1); - f->acc >>= n; - f->valid_bits -= n; - return z; -} - -static int32 get_bits_signed(vorb *f, int n) -{ - uint32 z = get_bits(f, n); - if (z & (1 << (n-1))) - z += ~((1 << n) - 1); - return (int32) z; -} - -// @OPTIMIZE: primary accumulator for huffman -// expand the buffer to as many bits as possible without reading off end of packet -// it might be nice to allow f->valid_bits and f->acc to be stored in registers, -// e.g. cache them locally and decode locally -static __forceinline void prep_huffman(vorb *f) -{ - if (f->valid_bits <= 24) { - if (f->valid_bits == 0) f->acc = 0; - do { - int z; - if (f->last_seg && !f->bytes_in_seg) return; - z = get8_packet_raw(f); - if (z == EOP) return; - f->acc += z << f->valid_bits; - f->valid_bits += 8; - } while (f->valid_bits <= 24); - } -} - -enum -{ - VORBIS_packet_id = 1, - VORBIS_packet_comment = 3, - VORBIS_packet_setup = 5, -}; - -static int codebook_decode_scalar_raw(vorb *f, Codebook *c) -{ - int i; - prep_huffman(f); - - assert(c->sorted_codewords || c->codewords); - // cases to use binary search: sorted_codewords && !c->codewords - // sorted_codewords && c->entries > 8 - if (c->entries > 8 ? c->sorted_codewords!=NULL : !c->codewords) { - // binary search - uint32 code = bit_reverse(f->acc); - int x=0, n=c->sorted_entries, len; - - while (n > 1) { - // invariant: sc[x] <= code < sc[x+n] - int m = x + (n >> 1); - if (c->sorted_codewords[m] <= code) { - x = m; - n -= (n>>1); - } else { - n >>= 1; - } - } - // x is now the sorted index - if (!c->sparse) x = c->sorted_values[x]; - // x is now sorted index if sparse, or symbol otherwise - len = c->codeword_lengths[x]; - if (f->valid_bits >= len) { - f->acc >>= len; - f->valid_bits -= len; - return x; - } - - f->valid_bits = 0; - return -1; - } - - // if small, linear search - assert(!c->sparse); - for (i=0; i < c->entries; ++i) { - if (c->codeword_lengths[i] == NO_CODE) continue; - if (c->codewords[i] == (f->acc & ((1 << c->codeword_lengths[i])-1))) { - if (f->valid_bits >= c->codeword_lengths[i]) { - f->acc >>= c->codeword_lengths[i]; - f->valid_bits -= c->codeword_lengths[i]; - return i; - } - f->valid_bits = 0; - return -1; - } - } - - error(f, VORBIS_invalid_stream); - f->valid_bits = 0; - return -1; -} - -static int codebook_decode_scalar(vorb *f, Codebook *c) -{ - int i; - if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) - prep_huffman(f); - // fast huffman table lookup - i = f->acc & FAST_HUFFMAN_TABLE_MASK; - i = c->fast_huffman[i]; - if (i >= 0) { - f->acc >>= c->codeword_lengths[i]; - f->valid_bits -= c->codeword_lengths[i]; - if (f->valid_bits < 0) { f->valid_bits = 0; return -1; } - return i; - } - return codebook_decode_scalar_raw(f,c); -} - -#ifndef STB_VORBIS_NO_INLINE_DECODE - -#define DECODE_RAW(var, f,c) \ - if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) \ - prep_huffman(f); \ - var = f->acc & FAST_HUFFMAN_TABLE_MASK; \ - var = c->fast_huffman[var]; \ - if (var >= 0) { \ - int n = c->codeword_lengths[var]; \ - f->acc >>= n; \ - f->valid_bits -= n; \ - if (f->valid_bits < 0) { f->valid_bits = 0; var = -1; } \ - } else { \ - var = codebook_decode_scalar_raw(f,c); \ - } - -#else - -#define DECODE_RAW(var,f,c) var = codebook_decode_scalar(f,c); - -#endif - -#define DECODE(var,f,c) \ - DECODE_RAW(var,f,c) \ - if (c->sparse) var = c->sorted_values[var]; - -#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK - #define DECODE_VQ(var,f,c) DECODE_RAW(var,f,c) -#else - #define DECODE_VQ(var,f,c) DECODE(var,f,c) -#endif - - - - - - -// CODEBOOK_ELEMENT_FAST is an optimization for the CODEBOOK_FLOATS case -// where we avoid one addition -#ifndef STB_VORBIS_CODEBOOK_FLOATS - #define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off] * c->delta_value + c->minimum_value) - #define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off] * c->delta_value) - #define CODEBOOK_ELEMENT_BASE(c) (c->minimum_value) -#else - #define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off]) - #define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off]) - #define CODEBOOK_ELEMENT_BASE(c) (0) -#endif - -static int codebook_decode_start(vorb *f, Codebook *c, int len) -{ - int z = -1; - - // type 0 is only legal in a scalar context - if (c->lookup_type == 0) - error(f, VORBIS_invalid_stream); - else { - DECODE_VQ(z,f,c); - if (c->sparse) assert(z < c->sorted_entries); - if (z < 0) { // check for EOP - if (!f->bytes_in_seg) - if (f->last_seg) - return z; - error(f, VORBIS_invalid_stream); - } - } - return z; -} - -static int codebook_decode(vorb *f, Codebook *c, float *output, int len) -{ - int i,z = codebook_decode_start(f,c,len); - if (z < 0) return FALSE; - if (len > c->dimensions) len = c->dimensions; - -#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK - if (c->lookup_type == 1) { - float last = CODEBOOK_ELEMENT_BASE(c); - int div = 1; - for (i=0; i < len; ++i) { - int off = (z / div) % c->lookup_values; - float val = CODEBOOK_ELEMENT_FAST(c,off) + last; - output[i] += val; - if (c->sequence_p) last = val + c->minimum_value; - div *= c->lookup_values; - } - return TRUE; - } -#endif - - z *= c->dimensions; - if (c->sequence_p) { - float last = CODEBOOK_ELEMENT_BASE(c); - for (i=0; i < len; ++i) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - output[i] += val; - last = val + c->minimum_value; - } - } else { - float last = CODEBOOK_ELEMENT_BASE(c); - for (i=0; i < len; ++i) { - output[i] += CODEBOOK_ELEMENT_FAST(c,z+i) + last; - } - } - - return TRUE; -} - -static int codebook_decode_step(vorb *f, Codebook *c, float *output, int len, int step) -{ - int i,z = codebook_decode_start(f,c,len); - float last = CODEBOOK_ELEMENT_BASE(c); - if (z < 0) return FALSE; - if (len > c->dimensions) len = c->dimensions; - -#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK - if (c->lookup_type == 1) { - int div = 1; - for (i=0; i < len; ++i) { - int off = (z / div) % c->lookup_values; - float val = CODEBOOK_ELEMENT_FAST(c,off) + last; - output[i*step] += val; - if (c->sequence_p) last = val; - div *= c->lookup_values; - } - return TRUE; - } -#endif - - z *= c->dimensions; - for (i=0; i < len; ++i) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - output[i*step] += val; - if (c->sequence_p) last = val; - } - - return TRUE; -} - -static int codebook_decode_deinterleave_repeat(vorb *f, Codebook *c, float **outputs, int ch, int *c_inter_p, int *p_inter_p, int len, int total_decode) -{ - int c_inter = *c_inter_p; - int p_inter = *p_inter_p; - int i,z, effective = c->dimensions; - - // type 0 is only legal in a scalar context - if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream); - - while (total_decode > 0) { - float last = CODEBOOK_ELEMENT_BASE(c); - DECODE_VQ(z,f,c); - #ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK - assert(!c->sparse || z < c->sorted_entries); - #endif - if (z < 0) { - if (!f->bytes_in_seg) - if (f->last_seg) return FALSE; - return error(f, VORBIS_invalid_stream); - } - - // if this will take us off the end of the buffers, stop short! - // we check by computing the length of the virtual interleaved - // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), - // and the length we'll be using (effective) - if (c_inter + p_inter*ch + effective > len * ch) { - effective = len*ch - (p_inter*ch - c_inter); - } - - #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK - if (c->lookup_type == 1) { - int div = 1; - for (i=0; i < effective; ++i) { - int off = (z / div) % c->lookup_values; - float val = CODEBOOK_ELEMENT_FAST(c,off) + last; - outputs[c_inter][p_inter] += val; - if (++c_inter == ch) { c_inter = 0; ++p_inter; } - if (c->sequence_p) last = val; - div *= c->lookup_values; - } - } else - #endif - { - z *= c->dimensions; - if (c->sequence_p) { - for (i=0; i < effective; ++i) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - outputs[c_inter][p_inter] += val; - if (++c_inter == ch) { c_inter = 0; ++p_inter; } - last = val; - } - } else { - for (i=0; i < effective; ++i) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - outputs[c_inter][p_inter] += val; - if (++c_inter == ch) { c_inter = 0; ++p_inter; } - } - } - } - - total_decode -= effective; - } - *c_inter_p = c_inter; - *p_inter_p = p_inter; - return TRUE; -} - -#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK -static int codebook_decode_deinterleave_repeat_2(vorb *f, Codebook *c, float **outputs, int *c_inter_p, int *p_inter_p, int len, int total_decode) -{ - int c_inter = *c_inter_p; - int p_inter = *p_inter_p; - int i,z, effective = c->dimensions; - - // type 0 is only legal in a scalar context - if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream); - - while (total_decode > 0) { - float last = CODEBOOK_ELEMENT_BASE(c); - DECODE_VQ(z,f,c); - - if (z < 0) { - if (!f->bytes_in_seg) - if (f->last_seg) return FALSE; - return error(f, VORBIS_invalid_stream); - } - - // if this will take us off the end of the buffers, stop short! - // we check by computing the length of the virtual interleaved - // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), - // and the length we'll be using (effective) - if (c_inter + p_inter*2 + effective > len * 2) { - effective = len*2 - (p_inter*2 - c_inter); - } - - { - z *= c->dimensions; - stb_prof(11); - if (c->sequence_p) { - // haven't optimized this case because I don't have any examples - for (i=0; i < effective; ++i) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - outputs[c_inter][p_inter] += val; - if (++c_inter == 2) { c_inter = 0; ++p_inter; } - last = val; - } - } else { - i=0; - if (c_inter == 1) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - outputs[c_inter][p_inter] += val; - c_inter = 0; ++p_inter; - ++i; - } - { - float *z0 = outputs[0]; - float *z1 = outputs[1]; - for (; i+1 < effective;) { - z0[p_inter] += CODEBOOK_ELEMENT_FAST(c,z+i) + last; - z1[p_inter] += CODEBOOK_ELEMENT_FAST(c,z+i+1) + last; - ++p_inter; - i += 2; - } - } - if (i < effective) { - float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last; - outputs[c_inter][p_inter] += val; - if (++c_inter == 2) { c_inter = 0; ++p_inter; } - } - } - } - - total_decode -= effective; - } - *c_inter_p = c_inter; - *p_inter_p = p_inter; - return TRUE; -} -#endif - -static int predict_point(int x, int x0, int x1, int y0, int y1) -{ - int dy = y1 - y0; - int adx = x1 - x0; - // @OPTIMIZE: force int division to round in the right direction... is this necessary on x86? - int err = abs(dy) * (x - x0); - int off = err / adx; - return dy < 0 ? y0 - off : y0 + off; -} - -// the following table is block-copied from the specification -static float inverse_db_table[256] = -{ - 1.0649863e-07f, 1.1341951e-07f, 1.2079015e-07f, 1.2863978e-07f, - 1.3699951e-07f, 1.4590251e-07f, 1.5538408e-07f, 1.6548181e-07f, - 1.7623575e-07f, 1.8768855e-07f, 1.9988561e-07f, 2.1287530e-07f, - 2.2670913e-07f, 2.4144197e-07f, 2.5713223e-07f, 2.7384213e-07f, - 2.9163793e-07f, 3.1059021e-07f, 3.3077411e-07f, 3.5226968e-07f, - 3.7516214e-07f, 3.9954229e-07f, 4.2550680e-07f, 4.5315863e-07f, - 4.8260743e-07f, 5.1396998e-07f, 5.4737065e-07f, 5.8294187e-07f, - 6.2082472e-07f, 6.6116941e-07f, 7.0413592e-07f, 7.4989464e-07f, - 7.9862701e-07f, 8.5052630e-07f, 9.0579828e-07f, 9.6466216e-07f, - 1.0273513e-06f, 1.0941144e-06f, 1.1652161e-06f, 1.2409384e-06f, - 1.3215816e-06f, 1.4074654e-06f, 1.4989305e-06f, 1.5963394e-06f, - 1.7000785e-06f, 1.8105592e-06f, 1.9282195e-06f, 2.0535261e-06f, - 2.1869758e-06f, 2.3290978e-06f, 2.4804557e-06f, 2.6416497e-06f, - 2.8133190e-06f, 2.9961443e-06f, 3.1908506e-06f, 3.3982101e-06f, - 3.6190449e-06f, 3.8542308e-06f, 4.1047004e-06f, 4.3714470e-06f, - 4.6555282e-06f, 4.9580707e-06f, 5.2802740e-06f, 5.6234160e-06f, - 5.9888572e-06f, 6.3780469e-06f, 6.7925283e-06f, 7.2339451e-06f, - 7.7040476e-06f, 8.2047000e-06f, 8.7378876e-06f, 9.3057248e-06f, - 9.9104632e-06f, 1.0554501e-05f, 1.1240392e-05f, 1.1970856e-05f, - 1.2748789e-05f, 1.3577278e-05f, 1.4459606e-05f, 1.5399272e-05f, - 1.6400004e-05f, 1.7465768e-05f, 1.8600792e-05f, 1.9809576e-05f, - 2.1096914e-05f, 2.2467911e-05f, 2.3928002e-05f, 2.5482978e-05f, - 2.7139006e-05f, 2.8902651e-05f, 3.0780908e-05f, 3.2781225e-05f, - 3.4911534e-05f, 3.7180282e-05f, 3.9596466e-05f, 4.2169667e-05f, - 4.4910090e-05f, 4.7828601e-05f, 5.0936773e-05f, 5.4246931e-05f, - 5.7772202e-05f, 6.1526565e-05f, 6.5524908e-05f, 6.9783085e-05f, - 7.4317983e-05f, 7.9147585e-05f, 8.4291040e-05f, 8.9768747e-05f, - 9.5602426e-05f, 0.00010181521f, 0.00010843174f, 0.00011547824f, - 0.00012298267f, 0.00013097477f, 0.00013948625f, 0.00014855085f, - 0.00015820453f, 0.00016848555f, 0.00017943469f, 0.00019109536f, - 0.00020351382f, 0.00021673929f, 0.00023082423f, 0.00024582449f, - 0.00026179955f, 0.00027881276f, 0.00029693158f, 0.00031622787f, - 0.00033677814f, 0.00035866388f, 0.00038197188f, 0.00040679456f, - 0.00043323036f, 0.00046138411f, 0.00049136745f, 0.00052329927f, - 0.00055730621f, 0.00059352311f, 0.00063209358f, 0.00067317058f, - 0.00071691700f, 0.00076350630f, 0.00081312324f, 0.00086596457f, - 0.00092223983f, 0.00098217216f, 0.0010459992f, 0.0011139742f, - 0.0011863665f, 0.0012634633f, 0.0013455702f, 0.0014330129f, - 0.0015261382f, 0.0016253153f, 0.0017309374f, 0.0018434235f, - 0.0019632195f, 0.0020908006f, 0.0022266726f, 0.0023713743f, - 0.0025254795f, 0.0026895994f, 0.0028643847f, 0.0030505286f, - 0.0032487691f, 0.0034598925f, 0.0036847358f, 0.0039241906f, - 0.0041792066f, 0.0044507950f, 0.0047400328f, 0.0050480668f, - 0.0053761186f, 0.0057254891f, 0.0060975636f, 0.0064938176f, - 0.0069158225f, 0.0073652516f, 0.0078438871f, 0.0083536271f, - 0.0088964928f, 0.009474637f, 0.010090352f, 0.010746080f, - 0.011444421f, 0.012188144f, 0.012980198f, 0.013823725f, - 0.014722068f, 0.015678791f, 0.016697687f, 0.017782797f, - 0.018938423f, 0.020169149f, 0.021479854f, 0.022875735f, - 0.024362330f, 0.025945531f, 0.027631618f, 0.029427276f, - 0.031339626f, 0.033376252f, 0.035545228f, 0.037855157f, - 0.040315199f, 0.042935108f, 0.045725273f, 0.048696758f, - 0.051861348f, 0.055231591f, 0.058820850f, 0.062643361f, - 0.066714279f, 0.071049749f, 0.075666962f, 0.080584227f, - 0.085821044f, 0.091398179f, 0.097337747f, 0.10366330f, - 0.11039993f, 0.11757434f, 0.12521498f, 0.13335215f, - 0.14201813f, 0.15124727f, 0.16107617f, 0.17154380f, - 0.18269168f, 0.19456402f, 0.20720788f, 0.22067342f, - 0.23501402f, 0.25028656f, 0.26655159f, 0.28387361f, - 0.30232132f, 0.32196786f, 0.34289114f, 0.36517414f, - 0.38890521f, 0.41417847f, 0.44109412f, 0.46975890f, - 0.50028648f, 0.53279791f, 0.56742212f, 0.60429640f, - 0.64356699f, 0.68538959f, 0.72993007f, 0.77736504f, - 0.82788260f, 0.88168307f, 0.9389798f, 1.0f -}; - - -// @OPTIMIZE: if you want to replace this bresenham line-drawing routine, -// note that you must produce bit-identical output to decode correctly; -// this specific sequence of operations is specified in the spec (it's -// drawing integer-quantized frequency-space lines that the encoder -// expects to be exactly the same) -// ... also, isn't the whole point of Bresenham's algorithm to NOT -// have to divide in the setup? sigh. -#ifndef STB_VORBIS_NO_DEFER_FLOOR -#define LINE_OP(a,b) a *= b -#else -#define LINE_OP(a,b) a = b -#endif - -#ifdef STB_VORBIS_DIVIDE_TABLE -#define DIVTAB_NUMER 32 -#define DIVTAB_DENOM 64 -int8 integer_divide_table[DIVTAB_NUMER][DIVTAB_DENOM]; // 2KB -#endif - -static __forceinline void draw_line(float *output, int x0, int y0, int x1, int y1, int n) -{ - int dy = y1 - y0; - int adx = x1 - x0; - int ady = abs(dy); - int base; - int x=x0,y=y0; - int err = 0; - int sy; - -#ifdef STB_VORBIS_DIVIDE_TABLE - if (adx < DIVTAB_DENOM && ady < DIVTAB_NUMER) { - if (dy < 0) { - base = -integer_divide_table[ady][adx]; - sy = base-1; - } else { - base = integer_divide_table[ady][adx]; - sy = base+1; - } - } else { - base = dy / adx; - if (dy < 0) - sy = base - 1; - else - sy = base+1; - } -#else - base = dy / adx; - if (dy < 0) - sy = base - 1; - else - sy = base+1; -#endif - ady -= abs(base) * adx; - if (x1 > n) x1 = n; - LINE_OP(output[x], inverse_db_table[y]); - for (++x; x < x1; ++x) { - err += ady; - if (err >= adx) { - err -= adx; - y += sy; - } else - y += base; - LINE_OP(output[x], inverse_db_table[y]); - } -} - -static int residue_decode(vorb *f, Codebook *book, float *target, int offset, int n, int rtype) -{ - int k; - if (rtype == 0) { - int step = n / book->dimensions; - for (k=0; k < step; ++k) - if (!codebook_decode_step(f, book, target+offset+k, n-offset-k, step)) - return FALSE; - } else { - for (k=0; k < n; ) { - if (!codebook_decode(f, book, target+offset, n-k)) - return FALSE; - k += book->dimensions; - offset += book->dimensions; - } - } - return TRUE; -} - -static void decode_residue(vorb *f, float *residue_buffers[], int ch, int n, int rn, uint8 *do_not_decode) -{ - int i,j,pass; - Residue *r = f->residue_config + rn; - int rtype = f->residue_types[rn]; - int c = r->classbook; - int classwords = f->codebooks[c].dimensions; - int n_read = r->end - r->begin; - int part_read = n_read / r->part_size; - int temp_alloc_point = temp_alloc_save(f); - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - uint8 ***part_classdata = (uint8 ***) temp_block_array(f,f->channels, part_read * sizeof(**part_classdata)); - #else - int **classifications = (int **) temp_block_array(f,f->channels, part_read * sizeof(**classifications)); - #endif - - stb_prof(2); - for (i=0; i < ch; ++i) - if (!do_not_decode[i]) - memset(residue_buffers[i], 0, sizeof(float) * n); - - if (rtype == 2 && ch != 1) { - int len = ch * n; - for (j=0; j < ch; ++j) - if (!do_not_decode[j]) - break; - if (j == ch) - goto done; - - stb_prof(3); - for (pass=0; pass < 8; ++pass) { - int pcount = 0, class_set = 0; - if (ch == 2) { - stb_prof(13); - while (pcount < part_read) { - int z = r->begin + pcount*r->part_size; - int c_inter = (z & 1), p_inter = z>>1; - if (pass == 0) { - Codebook *c = f->codebooks+r->classbook; - int q; - DECODE(q,f,c); - if (q == EOP) goto done; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - part_classdata[0][class_set] = r->classdata[q]; - #else - for (i=classwords-1; i >= 0; --i) { - classifications[0][i+pcount] = q % r->classifications; - q /= r->classifications; - } - #endif - } - stb_prof(5); - for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { - int z = r->begin + pcount*r->part_size; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - int c = part_classdata[0][class_set][i]; - #else - int c = classifications[0][pcount]; - #endif - int b = r->residue_books[c][pass]; - if (b >= 0) { - Codebook *book = f->codebooks + b; - stb_prof(20); // accounts for X time - #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK - if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) - goto done; - #else - // saves 1% - if (!codebook_decode_deinterleave_repeat_2(f, book, residue_buffers, &c_inter, &p_inter, n, r->part_size)) - goto done; - #endif - stb_prof(7); - } else { - z += r->part_size; - c_inter = z & 1; - p_inter = z >> 1; - } - } - stb_prof(8); - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - ++class_set; - #endif - } - } else if (ch == 1) { - while (pcount < part_read) { - int z = r->begin + pcount*r->part_size; - int c_inter = 0, p_inter = z; - if (pass == 0) { - Codebook *c = f->codebooks+r->classbook; - int q; - DECODE(q,f,c); - if (q == EOP) goto done; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - part_classdata[0][class_set] = r->classdata[q]; - #else - for (i=classwords-1; i >= 0; --i) { - classifications[0][i+pcount] = q % r->classifications; - q /= r->classifications; - } - #endif - } - for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { - int z = r->begin + pcount*r->part_size; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - int c = part_classdata[0][class_set][i]; - #else - int c = classifications[0][pcount]; - #endif - int b = r->residue_books[c][pass]; - if (b >= 0) { - Codebook *book = f->codebooks + b; - stb_prof(22); - if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) - goto done; - stb_prof(3); - } else { - z += r->part_size; - c_inter = 0; - p_inter = z; - } - } - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - ++class_set; - #endif - } - } else { - while (pcount < part_read) { - int z = r->begin + pcount*r->part_size; - int c_inter = z % ch, p_inter = z/ch; - if (pass == 0) { - Codebook *c = f->codebooks+r->classbook; - int q; - DECODE(q,f,c); - if (q == EOP) goto done; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - part_classdata[0][class_set] = r->classdata[q]; - #else - for (i=classwords-1; i >= 0; --i) { - classifications[0][i+pcount] = q % r->classifications; - q /= r->classifications; - } - #endif - } - for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { - int z = r->begin + pcount*r->part_size; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - int c = part_classdata[0][class_set][i]; - #else - int c = classifications[0][pcount]; - #endif - int b = r->residue_books[c][pass]; - if (b >= 0) { - Codebook *book = f->codebooks + b; - stb_prof(22); - if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) - goto done; - stb_prof(3); - } else { - z += r->part_size; - c_inter = z % ch; - p_inter = z / ch; - } - } - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - ++class_set; - #endif - } - } - } - goto done; - } - stb_prof(9); - - for (pass=0; pass < 8; ++pass) { - int pcount = 0, class_set=0; - while (pcount < part_read) { - if (pass == 0) { - for (j=0; j < ch; ++j) { - if (!do_not_decode[j]) { - Codebook *c = f->codebooks+r->classbook; - int temp; - DECODE(temp,f,c); - if (temp == EOP) goto done; - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - part_classdata[j][class_set] = r->classdata[temp]; - #else - for (i=classwords-1; i >= 0; --i) { - classifications[j][i+pcount] = temp % r->classifications; - temp /= r->classifications; - } - #endif - } - } - } - for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) { - for (j=0; j < ch; ++j) { - if (!do_not_decode[j]) { - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - int c = part_classdata[j][class_set][i]; - #else - int c = classifications[j][pcount]; - #endif - int b = r->residue_books[c][pass]; - if (b >= 0) { - float *target = residue_buffers[j]; - int offset = r->begin + pcount * r->part_size; - int n = r->part_size; - Codebook *book = f->codebooks + b; - if (!residue_decode(f, book, target, offset, n, rtype)) - goto done; - } - } - } - } - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - ++class_set; - #endif - } - } - done: - stb_prof(0); - temp_alloc_restore(f,temp_alloc_point); -} - - -#if 0 -// slow way for debugging -void inverse_mdct_slow(float *buffer, int n) -{ - int i,j; - int n2 = n >> 1; - float *x = (float *) malloc(sizeof(*x) * n2); - memcpy(x, buffer, sizeof(*x) * n2); - for (i=0; i < n; ++i) { - float acc = 0; - for (j=0; j < n2; ++j) - // formula from paper: - //acc += n/4.0f * x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); - // formula from wikipedia - //acc += 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); - // these are equivalent, except the formula from the paper inverts the multiplier! - // however, what actually works is NO MULTIPLIER!?! - //acc += 64 * 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); - acc += x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); - buffer[i] = acc; - } - free(x); -} -#elif 0 -// same as above, but just barely able to run in real time on modern machines -void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) -{ - float mcos[16384]; - int i,j; - int n2 = n >> 1, nmask = (n << 2) -1; - float *x = (float *) malloc(sizeof(*x) * n2); - memcpy(x, buffer, sizeof(*x) * n2); - for (i=0; i < 4*n; ++i) - mcos[i] = (float) cos(M_PI / 2 * i / n); - - for (i=0; i < n; ++i) { - float acc = 0; - for (j=0; j < n2; ++j) - acc += x[j] * mcos[(2 * i + 1 + n2)*(2*j+1) & nmask]; - buffer[i] = acc; - } - free(x); -} -#else -// transform to use a slow dct-iv; this is STILL basically trivial, -// but only requires half as many ops -void dct_iv_slow(float *buffer, int n) -{ - float mcos[16384]; - float x[2048]; - int i,j; - int n2 = n >> 1, nmask = (n << 3) - 1; - memcpy(x, buffer, sizeof(*x) * n); - for (i=0; i < 8*n; ++i) - mcos[i] = (float) cos(M_PI / 4 * i / n); - for (i=0; i < n; ++i) { - float acc = 0; - for (j=0; j < n; ++j) - acc += x[j] * mcos[((2 * i + 1)*(2*j+1)) & nmask]; - //acc += x[j] * cos(M_PI / n * (i + 0.5) * (j + 0.5)); - buffer[i] = acc; - } -} - -void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) -{ - int i, n4 = n >> 2, n2 = n >> 1, n3_4 = n - n4; - float temp[4096]; - - memcpy(temp, buffer, n2 * sizeof(float)); - dct_iv_slow(temp, n2); // returns -c'-d, a-b' - - for (i=0; i < n4 ; ++i) buffer[i] = temp[i+n4]; // a-b' - for ( ; i < n3_4; ++i) buffer[i] = -temp[n3_4 - i - 1]; // b-a', c+d' - for ( ; i < n ; ++i) buffer[i] = -temp[i - n3_4]; // c'+d -} -#endif - -#ifndef LIBVORBIS_MDCT -#define LIBVORBIS_MDCT 0 -#endif - -#if LIBVORBIS_MDCT -// directly call the vorbis MDCT using an interface documented -// by Jeff Roberts... useful for performance comparison -typedef struct -{ - int n; - int log2n; - - float *trig; - int *bitrev; - - float scale; -} mdct_lookup; - -extern void mdct_init(mdct_lookup *lookup, int n); -extern void mdct_clear(mdct_lookup *l); -extern void mdct_backward(mdct_lookup *init, float *in, float *out); - -mdct_lookup M1,M2; - -void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) -{ - mdct_lookup *M; - if (M1.n == n) M = &M1; - else if (M2.n == n) M = &M2; - else if (M1.n == 0) { mdct_init(&M1, n); M = &M1; } - else { - if (M2.n) __asm int 3; - mdct_init(&M2, n); - M = &M2; - } - - mdct_backward(M, buffer, buffer); -} -#endif - - -// the following were split out into separate functions while optimizing; -// they could be pushed back up but eh. __forceinline showed no change; -// they're probably already being inlined. -static void imdct_step3_iter0_loop(int n, float *e, int i_off, int k_off, float *A) -{ - float *ee0 = e + i_off; - float *ee2 = ee0 + k_off; - int i; - - assert((n & 3) == 0); - for (i=(n>>2); i > 0; --i) { - float k00_20, k01_21; - k00_20 = ee0[ 0] - ee2[ 0]; - k01_21 = ee0[-1] - ee2[-1]; - ee0[ 0] += ee2[ 0];//ee0[ 0] = ee0[ 0] + ee2[ 0]; - ee0[-1] += ee2[-1];//ee0[-1] = ee0[-1] + ee2[-1]; - ee2[ 0] = k00_20 * A[0] - k01_21 * A[1]; - ee2[-1] = k01_21 * A[0] + k00_20 * A[1]; - A += 8; - - k00_20 = ee0[-2] - ee2[-2]; - k01_21 = ee0[-3] - ee2[-3]; - ee0[-2] += ee2[-2];//ee0[-2] = ee0[-2] + ee2[-2]; - ee0[-3] += ee2[-3];//ee0[-3] = ee0[-3] + ee2[-3]; - ee2[-2] = k00_20 * A[0] - k01_21 * A[1]; - ee2[-3] = k01_21 * A[0] + k00_20 * A[1]; - A += 8; - - k00_20 = ee0[-4] - ee2[-4]; - k01_21 = ee0[-5] - ee2[-5]; - ee0[-4] += ee2[-4];//ee0[-4] = ee0[-4] + ee2[-4]; - ee0[-5] += ee2[-5];//ee0[-5] = ee0[-5] + ee2[-5]; - ee2[-4] = k00_20 * A[0] - k01_21 * A[1]; - ee2[-5] = k01_21 * A[0] + k00_20 * A[1]; - A += 8; - - k00_20 = ee0[-6] - ee2[-6]; - k01_21 = ee0[-7] - ee2[-7]; - ee0[-6] += ee2[-6];//ee0[-6] = ee0[-6] + ee2[-6]; - ee0[-7] += ee2[-7];//ee0[-7] = ee0[-7] + ee2[-7]; - ee2[-6] = k00_20 * A[0] - k01_21 * A[1]; - ee2[-7] = k01_21 * A[0] + k00_20 * A[1]; - A += 8; - ee0 -= 8; - ee2 -= 8; - } -} - -static void imdct_step3_inner_r_loop(int lim, float *e, int d0, int k_off, float *A, int k1) -{ - int i; - float k00_20, k01_21; - - float *e0 = e + d0; - float *e2 = e0 + k_off; - - for (i=lim >> 2; i > 0; --i) { - k00_20 = e0[-0] - e2[-0]; - k01_21 = e0[-1] - e2[-1]; - e0[-0] += e2[-0];//e0[-0] = e0[-0] + e2[-0]; - e0[-1] += e2[-1];//e0[-1] = e0[-1] + e2[-1]; - e2[-0] = (k00_20)*A[0] - (k01_21) * A[1]; - e2[-1] = (k01_21)*A[0] + (k00_20) * A[1]; - - A += k1; - - k00_20 = e0[-2] - e2[-2]; - k01_21 = e0[-3] - e2[-3]; - e0[-2] += e2[-2];//e0[-2] = e0[-2] + e2[-2]; - e0[-3] += e2[-3];//e0[-3] = e0[-3] + e2[-3]; - e2[-2] = (k00_20)*A[0] - (k01_21) * A[1]; - e2[-3] = (k01_21)*A[0] + (k00_20) * A[1]; - - A += k1; - - k00_20 = e0[-4] - e2[-4]; - k01_21 = e0[-5] - e2[-5]; - e0[-4] += e2[-4];//e0[-4] = e0[-4] + e2[-4]; - e0[-5] += e2[-5];//e0[-5] = e0[-5] + e2[-5]; - e2[-4] = (k00_20)*A[0] - (k01_21) * A[1]; - e2[-5] = (k01_21)*A[0] + (k00_20) * A[1]; - - A += k1; - - k00_20 = e0[-6] - e2[-6]; - k01_21 = e0[-7] - e2[-7]; - e0[-6] += e2[-6];//e0[-6] = e0[-6] + e2[-6]; - e0[-7] += e2[-7];//e0[-7] = e0[-7] + e2[-7]; - e2[-6] = (k00_20)*A[0] - (k01_21) * A[1]; - e2[-7] = (k01_21)*A[0] + (k00_20) * A[1]; - - e0 -= 8; - e2 -= 8; - - A += k1; - } -} - -static void imdct_step3_inner_s_loop(int n, float *e, int i_off, int k_off, float *A, int a_off, int k0) -{ - int i; - float A0 = A[0]; - float A1 = A[0+1]; - float A2 = A[0+a_off]; - float A3 = A[0+a_off+1]; - float A4 = A[0+a_off*2+0]; - float A5 = A[0+a_off*2+1]; - float A6 = A[0+a_off*3+0]; - float A7 = A[0+a_off*3+1]; - - float k00,k11; - - float *ee0 = e +i_off; - float *ee2 = ee0+k_off; - - for (i=n; i > 0; --i) { - k00 = ee0[ 0] - ee2[ 0]; - k11 = ee0[-1] - ee2[-1]; - ee0[ 0] = ee0[ 0] + ee2[ 0]; - ee0[-1] = ee0[-1] + ee2[-1]; - ee2[ 0] = (k00) * A0 - (k11) * A1; - ee2[-1] = (k11) * A0 + (k00) * A1; - - k00 = ee0[-2] - ee2[-2]; - k11 = ee0[-3] - ee2[-3]; - ee0[-2] = ee0[-2] + ee2[-2]; - ee0[-3] = ee0[-3] + ee2[-3]; - ee2[-2] = (k00) * A2 - (k11) * A3; - ee2[-3] = (k11) * A2 + (k00) * A3; - - k00 = ee0[-4] - ee2[-4]; - k11 = ee0[-5] - ee2[-5]; - ee0[-4] = ee0[-4] + ee2[-4]; - ee0[-5] = ee0[-5] + ee2[-5]; - ee2[-4] = (k00) * A4 - (k11) * A5; - ee2[-5] = (k11) * A4 + (k00) * A5; - - k00 = ee0[-6] - ee2[-6]; - k11 = ee0[-7] - ee2[-7]; - ee0[-6] = ee0[-6] + ee2[-6]; - ee0[-7] = ee0[-7] + ee2[-7]; - ee2[-6] = (k00) * A6 - (k11) * A7; - ee2[-7] = (k11) * A6 + (k00) * A7; - - ee0 -= k0; - ee2 -= k0; - } -} - -static __forceinline void iter_54(float *z) -{ - float k00,k11,k22,k33; - float y0,y1,y2,y3; - - k00 = z[ 0] - z[-4]; - y0 = z[ 0] + z[-4]; - y2 = z[-2] + z[-6]; - k22 = z[-2] - z[-6]; - - z[-0] = y0 + y2; // z0 + z4 + z2 + z6 - z[-2] = y0 - y2; // z0 + z4 - z2 - z6 - - // done with y0,y2 - - k33 = z[-3] - z[-7]; - - z[-4] = k00 + k33; // z0 - z4 + z3 - z7 - z[-6] = k00 - k33; // z0 - z4 - z3 + z7 - - // done with k33 - - k11 = z[-1] - z[-5]; - y1 = z[-1] + z[-5]; - y3 = z[-3] + z[-7]; - - z[-1] = y1 + y3; // z1 + z5 + z3 + z7 - z[-3] = y1 - y3; // z1 + z5 - z3 - z7 - z[-5] = k11 - k22; // z1 - z5 + z2 - z6 - z[-7] = k11 + k22; // z1 - z5 - z2 + z6 -} - -static void imdct_step3_inner_s_loop_ld654(int n, float *e, int i_off, float *A, int base_n) -{ - int k_off = -8; - int a_off = base_n >> 3; - float A2 = A[0+a_off]; - float *z = e + i_off; - float *base = z - 16 * n; - - while (z > base) { - float k00,k11; - - k00 = z[-0] - z[-8]; - k11 = z[-1] - z[-9]; - z[-0] = z[-0] + z[-8]; - z[-1] = z[-1] + z[-9]; - z[-8] = k00; - z[-9] = k11 ; - - k00 = z[ -2] - z[-10]; - k11 = z[ -3] - z[-11]; - z[ -2] = z[ -2] + z[-10]; - z[ -3] = z[ -3] + z[-11]; - z[-10] = (k00+k11) * A2; - z[-11] = (k11-k00) * A2; - - k00 = z[-12] - z[ -4]; // reverse to avoid a unary negation - k11 = z[ -5] - z[-13]; - z[ -4] = z[ -4] + z[-12]; - z[ -5] = z[ -5] + z[-13]; - z[-12] = k11; - z[-13] = k00; - - k00 = z[-14] - z[ -6]; // reverse to avoid a unary negation - k11 = z[ -7] - z[-15]; - z[ -6] = z[ -6] + z[-14]; - z[ -7] = z[ -7] + z[-15]; - z[-14] = (k00+k11) * A2; - z[-15] = (k00-k11) * A2; - - iter_54(z); - iter_54(z-8); - z -= 16; - } -} - -static void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) -{ - int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; - int n3_4 = n - n4, ld; - // @OPTIMIZE: reduce register pressure by using fewer variables? - int save_point = temp_alloc_save(f); - float *buf2 = (float *) temp_alloc(f, n2 * sizeof(*buf2)); - float *u=NULL,*v=NULL; - // twiddle factors - float *A = f->A[blocktype]; - - // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" - // See notes about bugs in that paper in less-optimal implementation 'inverse_mdct_old' after this function. - - // kernel from paper - - - // merged: - // copy and reflect spectral data - // step 0 - - // note that it turns out that the items added together during - // this step are, in fact, being added to themselves (as reflected - // by step 0). inexplicable inefficiency! this became obvious - // once I combined the passes. - - // so there's a missing 'times 2' here (for adding X to itself). - // this propogates through linearly to the end, where the numbers - // are 1/2 too small, and need to be compensated for. - - { - float *d,*e, *AA, *e_stop; - d = &buf2[n2-2]; - AA = A; - e = &buffer[0]; - e_stop = &buffer[n2]; - while (e != e_stop) { - d[1] = (e[0] * AA[0] - e[2]*AA[1]); - d[0] = (e[0] * AA[1] + e[2]*AA[0]); - d -= 2; - AA += 2; - e += 4; - } - - e = &buffer[n2-3]; - while (d >= buf2) { - d[1] = (-e[2] * AA[0] - -e[0]*AA[1]); - d[0] = (-e[2] * AA[1] + -e[0]*AA[0]); - d -= 2; - AA += 2; - e -= 4; - } - } - - // now we use symbolic names for these, so that we can - // possibly swap their meaning as we change which operations - // are in place - - u = buffer; - v = buf2; - - // step 2 (paper output is w, now u) - // this could be in place, but the data ends up in the wrong - // place... _somebody_'s got to swap it, so this is nominated - { - float *AA = &A[n2-8]; - float *d0,*d1, *e0, *e1; - - e0 = &v[n4]; - e1 = &v[0]; - - d0 = &u[n4]; - d1 = &u[0]; - - while (AA >= A) { - float v40_20, v41_21; - - v41_21 = e0[1] - e1[1]; - v40_20 = e0[0] - e1[0]; - d0[1] = e0[1] + e1[1]; - d0[0] = e0[0] + e1[0]; - d1[1] = v41_21*AA[4] - v40_20*AA[5]; - d1[0] = v40_20*AA[4] + v41_21*AA[5]; - - v41_21 = e0[3] - e1[3]; - v40_20 = e0[2] - e1[2]; - d0[3] = e0[3] + e1[3]; - d0[2] = e0[2] + e1[2]; - d1[3] = v41_21*AA[0] - v40_20*AA[1]; - d1[2] = v40_20*AA[0] + v41_21*AA[1]; - - AA -= 8; - - d0 += 4; - d1 += 4; - e0 += 4; - e1 += 4; - } - } - - // step 3 - ld = ilog(n) - 1; // ilog is off-by-one from normal definitions - - // optimized step 3: - - // the original step3 loop can be nested r inside s or s inside r; - // it's written originally as s inside r, but this is dumb when r - // iterates many times, and s few. So I have two copies of it and - // switch between them halfway. - - // this is iteration 0 of step 3 - imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*0, -(n >> 3), A); - imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*1, -(n >> 3), A); - - // this is iteration 1 of step 3 - imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*0, -(n >> 4), A, 16); - imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*1, -(n >> 4), A, 16); - imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*2, -(n >> 4), A, 16); - imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*3, -(n >> 4), A, 16); - - l=2; - for (; l < (ld-3)>>1; ++l) { - int k0 = n >> (l+2), k0_2 = k0>>1; - int lim = 1 << (l+1); - int i; - for (i=0; i < lim; ++i) - imdct_step3_inner_r_loop(n >> (l+4), u, n2-1 - k0*i, -k0_2, A, 1 << (l+3)); - } - - for (; l < ld-6; ++l) { - int k0 = n >> (l+2), k1 = 1 << (l+3), k0_2 = k0>>1; - int rlim = n >> (l+6), r; - int lim = 1 << (l+1); - int i_off; - float *A0 = A; - i_off = n2-1; - for (r=rlim; r > 0; --r) { - imdct_step3_inner_s_loop(lim, u, i_off, -k0_2, A0, k1, k0); - A0 += k1*4; - i_off -= 8; - } - } - - // iterations with count: - // ld-6,-5,-4 all interleaved together - // the big win comes from getting rid of needless flops - // due to the constants on pass 5 & 4 being all 1 and 0; - // combining them to be simultaneous to improve cache made little difference - imdct_step3_inner_s_loop_ld654(n >> 5, u, n2-1, A, n); - - // output is u - - // step 4, 5, and 6 - // cannot be in-place because of step 5 - { - uint16 *bitrev = f->bit_reverse[blocktype]; - // weirdly, I'd have thought reading sequentially and writing - // erratically would have been better than vice-versa, but in - // fact that's not what my testing showed. (That is, with - // j = bitreverse(i), do you read i and write j, or read j and write i.) - - float *d0 = &v[n4-4]; - float *d1 = &v[n2-4]; - while (d0 >= v) { - int k4; - - k4 = bitrev[0]; - d1[3] = u[k4+0]; - d1[2] = u[k4+1]; - d0[3] = u[k4+2]; - d0[2] = u[k4+3]; - - k4 = bitrev[1]; - d1[1] = u[k4+0]; - d1[0] = u[k4+1]; - d0[1] = u[k4+2]; - d0[0] = u[k4+3]; - - d0 -= 4; - d1 -= 4; - bitrev += 2; - } - } - // (paper output is u, now v) - - - // data must be in buf2 - assert(v == buf2); - - // step 7 (paper output is v, now v) - // this is now in place - { - float *C = f->C[blocktype]; - float *d, *e; - - d = v; - e = v + n2 - 4; - - while (d < e) { - float a02,a11,b0,b1,b2,b3; - - a02 = d[0] - e[2]; - a11 = d[1] + e[3]; - - b0 = C[1]*a02 + C[0]*a11; - b1 = C[1]*a11 - C[0]*a02; - - b2 = d[0] + e[ 2]; - b3 = d[1] - e[ 3]; - - d[0] = b2 + b0; - d[1] = b3 + b1; - e[2] = b2 - b0; - e[3] = b1 - b3; - - a02 = d[2] - e[0]; - a11 = d[3] + e[1]; - - b0 = C[3]*a02 + C[2]*a11; - b1 = C[3]*a11 - C[2]*a02; - - b2 = d[2] + e[ 0]; - b3 = d[3] - e[ 1]; - - d[2] = b2 + b0; - d[3] = b3 + b1; - e[0] = b2 - b0; - e[1] = b1 - b3; - - C += 4; - d += 4; - e -= 4; - } - } - - // data must be in buf2 - - - // step 8+decode (paper output is X, now buffer) - // this generates pairs of data a la 8 and pushes them directly through - // the decode kernel (pushing rather than pulling) to avoid having - // to make another pass later - - // this cannot POSSIBLY be in place, so we refer to the buffers directly - - { - float *d0,*d1,*d2,*d3; - - float *B = f->B[blocktype] + n2 - 8; - float *e = buf2 + n2 - 8; - d0 = &buffer[0]; - d1 = &buffer[n2-4]; - d2 = &buffer[n2]; - d3 = &buffer[n-4]; - while (e >= v) { - float p0,p1,p2,p3; - - p3 = e[6]*B[7] - e[7]*B[6]; - p2 = -e[6]*B[6] - e[7]*B[7]; - - d0[0] = p3; - d1[3] = - p3; - d2[0] = p2; - d3[3] = p2; - - p1 = e[4]*B[5] - e[5]*B[4]; - p0 = -e[4]*B[4] - e[5]*B[5]; - - d0[1] = p1; - d1[2] = - p1; - d2[1] = p0; - d3[2] = p0; - - p3 = e[2]*B[3] - e[3]*B[2]; - p2 = -e[2]*B[2] - e[3]*B[3]; - - d0[2] = p3; - d1[1] = - p3; - d2[2] = p2; - d3[1] = p2; - - p1 = e[0]*B[1] - e[1]*B[0]; - p0 = -e[0]*B[0] - e[1]*B[1]; - - d0[3] = p1; - d1[0] = - p1; - d2[3] = p0; - d3[0] = p0; - - B -= 8; - e -= 8; - d0 += 4; - d2 += 4; - d1 -= 4; - d3 -= 4; - } - } - - temp_alloc_restore(f,save_point); -} - -#if 0 -// this is the original version of the above code, if you want to optimize it from scratch -void inverse_mdct_naive(float *buffer, int n) -{ - float s; - float A[1 << 12], B[1 << 12], C[1 << 11]; - int i,k,k2,k4, n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; - int n3_4 = n - n4, ld; - // how can they claim this only uses N words?! - // oh, because they're only used sparsely, whoops - float u[1 << 13], X[1 << 13], v[1 << 13], w[1 << 13]; - // set up twiddle factors - - for (k=k2=0; k < n4; ++k,k2+=2) { - A[k2 ] = (float) cos(4*k*M_PI/n); - A[k2+1] = (float) -sin(4*k*M_PI/n); - B[k2 ] = (float) cos((k2+1)*M_PI/n/2); - B[k2+1] = (float) sin((k2+1)*M_PI/n/2); - } - for (k=k2=0; k < n8; ++k,k2+=2) { - C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); - C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); - } - - // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" - // Note there are bugs in that pseudocode, presumably due to them attempting - // to rename the arrays nicely rather than representing the way their actual - // implementation bounces buffers back and forth. As a result, even in the - // "some formulars corrected" version, a direct implementation fails. These - // are noted below as "paper bug". - - // copy and reflect spectral data - for (k=0; k < n2; ++k) u[k] = buffer[k]; - for ( ; k < n ; ++k) u[k] = -buffer[n - k - 1]; - // kernel from paper - // step 1 - for (k=k2=k4=0; k < n4; k+=1, k2+=2, k4+=4) { - v[n-k4-1] = (u[k4] - u[n-k4-1]) * A[k2] - (u[k4+2] - u[n-k4-3])*A[k2+1]; - v[n-k4-3] = (u[k4] - u[n-k4-1]) * A[k2+1] + (u[k4+2] - u[n-k4-3])*A[k2]; - } - // step 2 - for (k=k4=0; k < n8; k+=1, k4+=4) { - w[n2+3+k4] = v[n2+3+k4] + v[k4+3]; - w[n2+1+k4] = v[n2+1+k4] + v[k4+1]; - w[k4+3] = (v[n2+3+k4] - v[k4+3])*A[n2-4-k4] - (v[n2+1+k4]-v[k4+1])*A[n2-3-k4]; - w[k4+1] = (v[n2+1+k4] - v[k4+1])*A[n2-4-k4] + (v[n2+3+k4]-v[k4+3])*A[n2-3-k4]; - } - // step 3 - ld = ilog(n) - 1; // ilog is off-by-one from normal definitions - for (l=0; l < ld-3; ++l) { - int k0 = n >> (l+2), k1 = 1 << (l+3); - int rlim = n >> (l+4), r4, r; - int s2lim = 1 << (l+2), s2; - for (r=r4=0; r < rlim; r4+=4,++r) { - for (s2=0; s2 < s2lim; s2+=2) { - u[n-1-k0*s2-r4] = w[n-1-k0*s2-r4] + w[n-1-k0*(s2+1)-r4]; - u[n-3-k0*s2-r4] = w[n-3-k0*s2-r4] + w[n-3-k0*(s2+1)-r4]; - u[n-1-k0*(s2+1)-r4] = (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1] - - (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1+1]; - u[n-3-k0*(s2+1)-r4] = (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1] - + (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1+1]; - } - } - if (l+1 < ld-3) { - // paper bug: ping-ponging of u&w here is omitted - memcpy(w, u, sizeof(u)); - } - } - - // step 4 - for (i=0; i < n8; ++i) { - int j = bit_reverse(i) >> (32-ld+3); - assert(j < n8); - if (i == j) { - // paper bug: original code probably swapped in place; if copying, - // need to directly copy in this case - int i8 = i << 3; - v[i8+1] = u[i8+1]; - v[i8+3] = u[i8+3]; - v[i8+5] = u[i8+5]; - v[i8+7] = u[i8+7]; - } else if (i < j) { - int i8 = i << 3, j8 = j << 3; - v[j8+1] = u[i8+1], v[i8+1] = u[j8 + 1]; - v[j8+3] = u[i8+3], v[i8+3] = u[j8 + 3]; - v[j8+5] = u[i8+5], v[i8+5] = u[j8 + 5]; - v[j8+7] = u[i8+7], v[i8+7] = u[j8 + 7]; - } - } - // step 5 - for (k=0; k < n2; ++k) { - w[k] = v[k*2+1]; - } - // step 6 - for (k=k2=k4=0; k < n8; ++k, k2 += 2, k4 += 4) { - u[n-1-k2] = w[k4]; - u[n-2-k2] = w[k4+1]; - u[n3_4 - 1 - k2] = w[k4+2]; - u[n3_4 - 2 - k2] = w[k4+3]; - } - // step 7 - for (k=k2=0; k < n8; ++k, k2 += 2) { - v[n2 + k2 ] = ( u[n2 + k2] + u[n-2-k2] + C[k2+1]*(u[n2+k2]-u[n-2-k2]) + C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; - v[n-2 - k2] = ( u[n2 + k2] + u[n-2-k2] - C[k2+1]*(u[n2+k2]-u[n-2-k2]) - C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; - v[n2+1+ k2] = ( u[n2+1+k2] - u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; - v[n-1 - k2] = (-u[n2+1+k2] + u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; - } - // step 8 - for (k=k2=0; k < n4; ++k,k2 += 2) { - X[k] = v[k2+n2]*B[k2 ] + v[k2+1+n2]*B[k2+1]; - X[n2-1-k] = v[k2+n2]*B[k2+1] - v[k2+1+n2]*B[k2 ]; - } - - // decode kernel to output - // determined the following value experimentally - // (by first figuring out what made inverse_mdct_slow work); then matching that here - // (probably vorbis encoder premultiplies by n or n/2, to save it on the decoder?) - s = 0.5; // theoretically would be n4 - - // [[[ note! the s value of 0.5 is compensated for by the B[] in the current code, - // so it needs to use the "old" B values to behave correctly, or else - // set s to 1.0 ]]] - for (i=0; i < n4 ; ++i) buffer[i] = s * X[i+n4]; - for ( ; i < n3_4; ++i) buffer[i] = -s * X[n3_4 - i - 1]; - for ( ; i < n ; ++i) buffer[i] = -s * X[i - n3_4]; -} -#endif - -static float *get_window(vorb *f, int len) -{ - len <<= 1; - if (len == f->blocksize_0) return f->window[0]; - if (len == f->blocksize_1) return f->window[1]; - assert(0); - return NULL; -} - -#ifndef STB_VORBIS_NO_DEFER_FLOOR -typedef int16 YTYPE; -#else -typedef int YTYPE; -#endif -static int do_floor(vorb *f, Mapping *map, int i, int n, float *target, YTYPE *finalY, uint8 *step2_flag) -{ - int n2 = n >> 1; - int s = map->chan[i].mux, floor; - floor = map->submap_floor[s]; - if (f->floor_types[floor] == 0) { - return error(f, VORBIS_invalid_stream); - } else { - Floor1 *g = &f->floor_config[floor].floor1; - int j,q; - int lx = 0, ly = finalY[0] * g->floor1_multiplier; - for (q=1; q < g->values; ++q) { - j = g->sorted_order[q]; - #ifndef STB_VORBIS_NO_DEFER_FLOOR - if (finalY[j] >= 0) - #else - if (step2_flag[j]) - #endif - { - int hy = finalY[j] * g->floor1_multiplier; - int hx = g->Xlist[j]; - draw_line(target, lx,ly, hx,hy, n2); - lx = hx, ly = hy; - } - } - if (lx < n2) - // optimization of: draw_line(target, lx,ly, n,ly, n2); - for (j=lx; j < n2; ++j) - LINE_OP(target[j], inverse_db_table[ly]); - } - return TRUE; -} - -static int vorbis_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) -{ - Mode *m; - int i, n, prev, next, window_center; - f->channel_buffer_start = f->channel_buffer_end = 0; - - retry: - if (f->eof) return FALSE; - if (!maybe_start_packet(f)) - return FALSE; - // check packet type - if (get_bits(f,1) != 0) { - if (IS_PUSH_MODE(f)) - return error(f,VORBIS_bad_packet_type); - while (EOP != get8_packet(f)); - goto retry; - } - - if (f->alloc.alloc_buffer) - assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); - - i = get_bits(f, ilog(f->mode_count-1)); - if (i == EOP) return FALSE; - if (i >= f->mode_count) return FALSE; - *mode = i; - m = f->mode_config + i; - if (m->blockflag) { - n = f->blocksize_1; - prev = get_bits(f,1); - next = get_bits(f,1); - } else { - prev = next = 0; - n = f->blocksize_0; - } - -// WINDOWING - - window_center = n >> 1; - if (m->blockflag && !prev) { - *p_left_start = (n - f->blocksize_0) >> 2; - *p_left_end = (n + f->blocksize_0) >> 2; - } else { - *p_left_start = 0; - *p_left_end = window_center; - } - if (m->blockflag && !next) { - *p_right_start = (n*3 - f->blocksize_0) >> 2; - *p_right_end = (n*3 + f->blocksize_0) >> 2; - } else { - *p_right_start = window_center; - *p_right_end = n; - } - return TRUE; -} - -static int vorbis_decode_packet_rest(vorb *f, int *len, Mode *m, int left_start, int left_end, int right_start, int right_end, int *p_left) -{ - Mapping *map; - int i,j,k,n,n2; - int zero_channel[256]; - int really_zero_channel[256]; - int window_center; - -// WINDOWING - - n = f->blocksize[m->blockflag]; - window_center = n >> 1; - - map = &f->mapping[m->mapping]; - -// FLOORS - n2 = n >> 1; - - stb_prof(1); - for (i=0; i < f->channels; ++i) { - int s = map->chan[i].mux, floor; - zero_channel[i] = FALSE; - floor = map->submap_floor[s]; - if (f->floor_types[floor] == 0) { - return error(f, VORBIS_invalid_stream); - } else { - Floor1 *g = &f->floor_config[floor].floor1; - if (get_bits(f, 1)) { - short *finalY; - uint8 step2_flag[256]; - static int range_list[4] = { 256, 128, 86, 64 }; - int range = range_list[g->floor1_multiplier-1]; - int offset = 2; - finalY = f->finalY[i]; - finalY[0] = get_bits(f, ilog(range)-1); - finalY[1] = get_bits(f, ilog(range)-1); - for (j=0; j < g->partitions; ++j) { - int pclass = g->partition_class_list[j]; - int cdim = g->class_dimensions[pclass]; - int cbits = g->class_subclasses[pclass]; - int csub = (1 << cbits)-1; - int cval = 0; - if (cbits) { - Codebook *c = f->codebooks + g->class_masterbooks[pclass]; - DECODE(cval,f,c); - } - for (k=0; k < cdim; ++k) { - int book = g->subclass_books[pclass][cval & csub]; - cval = cval >> cbits; - if (book >= 0) { - int temp; - Codebook *c = f->codebooks + book; - DECODE(temp,f,c); - finalY[offset++] = temp; - } else - finalY[offset++] = 0; - } - } - if (f->valid_bits == INVALID_BITS) goto error; // behavior according to spec - step2_flag[0] = step2_flag[1] = 1; - for (j=2; j < g->values; ++j) { - int low, high, pred, highroom, lowroom, room, val; - low = g->neighbors[j][0]; - high = g->neighbors[j][1]; - //neighbors(g->Xlist, j, &low, &high); - pred = predict_point(g->Xlist[j], g->Xlist[low], g->Xlist[high], finalY[low], finalY[high]); - val = finalY[j]; - highroom = range - pred; - lowroom = pred; - if (highroom < lowroom) - room = highroom * 2; - else - room = lowroom * 2; - if (val) { - step2_flag[low] = step2_flag[high] = 1; - step2_flag[j] = 1; - if (val >= room) - if (highroom > lowroom) - finalY[j] = val - lowroom + pred; - else - finalY[j] = pred - val + highroom - 1; - else - if (val & 1) - finalY[j] = pred - ((val+1)>>1); - else - finalY[j] = pred + (val>>1); - } else { - step2_flag[j] = 0; - finalY[j] = pred; - } - } - -#ifdef STB_VORBIS_NO_DEFER_FLOOR - do_floor(f, map, i, n, f->floor_buffers[i], finalY, step2_flag); -#else - // defer final floor computation until _after_ residue - for (j=0; j < g->values; ++j) { - if (!step2_flag[j]) - finalY[j] = -1; - } -#endif - } else { - error: - zero_channel[i] = TRUE; - } - // So we just defer everything else to later - - // at this point we've decoded the floor into buffer - } - } - stb_prof(0); - // at this point we've decoded all floors - - if (f->alloc.alloc_buffer) - assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); - - // re-enable coupled channels if necessary - memcpy(really_zero_channel, zero_channel, sizeof(really_zero_channel[0]) * f->channels); - for (i=0; i < map->coupling_steps; ++i) - if (!zero_channel[map->chan[i].magnitude] || !zero_channel[map->chan[i].angle]) { - zero_channel[map->chan[i].magnitude] = zero_channel[map->chan[i].angle] = FALSE; - } - -// RESIDUE DECODE - for (i=0; i < map->submaps; ++i) { - float *residue_buffers[STB_VORBIS_MAX_CHANNELS]; - int r,t; - uint8 do_not_decode[256]; - int ch = 0; - for (j=0; j < f->channels; ++j) { - if (map->chan[j].mux == i) { - if (zero_channel[j]) { - do_not_decode[ch] = TRUE; - residue_buffers[ch] = NULL; - } else { - do_not_decode[ch] = FALSE; - residue_buffers[ch] = f->channel_buffers[j]; - } - ++ch; - } - } - r = map->submap_residue[i]; - t = f->residue_types[r]; - decode_residue(f, residue_buffers, ch, n2, r, do_not_decode); - } - - if (f->alloc.alloc_buffer) - assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); - -// INVERSE COUPLING - stb_prof(14); - for (i = map->coupling_steps-1; i >= 0; --i) { - int n2 = n >> 1; - float *m = f->channel_buffers[map->chan[i].magnitude]; - float *a = f->channel_buffers[map->chan[i].angle ]; - for (j=0; j < n2; ++j) { - float a2,m2; - if (m[j] > 0) - if (a[j] > 0) - m2 = m[j], a2 = m[j] - a[j]; - else - a2 = m[j], m2 = m[j] + a[j]; - else - if (a[j] > 0) - m2 = m[j], a2 = m[j] + a[j]; - else - a2 = m[j], m2 = m[j] - a[j]; - m[j] = m2; - a[j] = a2; - } - } - - // finish decoding the floors -#ifndef STB_VORBIS_NO_DEFER_FLOOR - stb_prof(15); - for (i=0; i < f->channels; ++i) { - if (really_zero_channel[i]) { - memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); - } else { - do_floor(f, map, i, n, f->channel_buffers[i], f->finalY[i], NULL); - } - } -#else - for (i=0; i < f->channels; ++i) { - if (really_zero_channel[i]) { - memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); - } else { - for (j=0; j < n2; ++j) - f->channel_buffers[i][j] *= f->floor_buffers[i][j]; - } - } -#endif - -// INVERSE MDCT - stb_prof(16); - for (i=0; i < f->channels; ++i) - inverse_mdct(f->channel_buffers[i], n, f, m->blockflag); - stb_prof(0); - - // this shouldn't be necessary, unless we exited on an error - // and want to flush to get to the next packet - flush_packet(f); - - if (f->first_decode) { - // assume we start so first non-discarded sample is sample 0 - // this isn't to spec, but spec would require us to read ahead - // and decode the size of all current frames--could be done, - // but presumably it's not a commonly used feature - f->current_loc = -n2; // start of first frame is positioned for discard - // we might have to discard samples "from" the next frame too, - // if we're lapping a large block then a small at the start? - f->discard_samples_deferred = n - right_end; - f->current_loc_valid = TRUE; - f->first_decode = FALSE; - } else if (f->discard_samples_deferred) { - left_start += f->discard_samples_deferred; - *p_left = left_start; - f->discard_samples_deferred = 0; - } else if (f->previous_length == 0 && f->current_loc_valid) { - // we're recovering from a seek... that means we're going to discard - // the samples from this packet even though we know our position from - // the last page header, so we need to update the position based on - // the discarded samples here - // but wait, the code below is going to add this in itself even - // on a discard, so we don't need to do it here... - } - - // check if we have ogg information about the sample # for this packet - if (f->last_seg_which == f->end_seg_with_known_loc) { - // if we have a valid current loc, and this is final: - if (f->current_loc_valid && (f->page_flag & PAGEFLAG_last_page)) { - uint32 current_end = f->known_loc_for_packet - (n-right_end); - // then let's infer the size of the (probably) short final frame - if (current_end < f->current_loc + right_end) { - if (current_end < f->current_loc) { - // negative truncation, that's impossible! - *len = 0; - } else { - *len = current_end - f->current_loc; - } - *len += left_start; - f->current_loc += *len; - return TRUE; - } - } - // otherwise, just set our sample loc - // guess that the ogg granule pos refers to the _middle_ of the - // last frame? - // set f->current_loc to the position of left_start - f->current_loc = f->known_loc_for_packet - (n2-left_start); - f->current_loc_valid = TRUE; - } - if (f->current_loc_valid) - f->current_loc += (right_start - left_start); - - if (f->alloc.alloc_buffer) - assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); - *len = right_end; // ignore samples after the window goes to 0 - return TRUE; -} - -static int vorbis_decode_packet(vorb *f, int *len, int *p_left, int *p_right) -{ - int mode, left_end, right_end; - if (!vorbis_decode_initial(f, p_left, &left_end, p_right, &right_end, &mode)) return 0; - return vorbis_decode_packet_rest(f, len, f->mode_config + mode, *p_left, left_end, *p_right, right_end, p_left); -} - -static int vorbis_finish_frame(stb_vorbis *f, int len, int left, int right) -{ - int prev,i,j; - // we use right&left (the start of the right- and left-window sin()-regions) - // to determine how much to return, rather than inferring from the rules - // (same result, clearer code); 'left' indicates where our sin() window - // starts, therefore where the previous window's right edge starts, and - // therefore where to start mixing from the previous buffer. 'right' - // indicates where our sin() ending-window starts, therefore that's where - // we start saving, and where our returned-data ends. - - // mixin from previous window - if (f->previous_length) { - int i,j, n = f->previous_length; - float *w = get_window(f, n); - for (i=0; i < f->channels; ++i) { - for (j=0; j < n; ++j) - f->channel_buffers[i][left+j] = - f->channel_buffers[i][left+j]*w[ j] + - f->previous_window[i][ j]*w[n-1-j]; - } - } - - prev = f->previous_length; - - // last half of this data becomes previous window - f->previous_length = len - right; - - // @OPTIMIZE: could avoid this copy by double-buffering the - // output (flipping previous_window with channel_buffers), but - // then previous_window would have to be 2x as large, and - // channel_buffers couldn't be temp mem (although they're NOT - // currently temp mem, they could be (unless we want to level - // performance by spreading out the computation)) - for (i=0; i < f->channels; ++i) - for (j=0; right+j < len; ++j) - f->previous_window[i][j] = f->channel_buffers[i][right+j]; - - if (!prev) - // there was no previous packet, so this data isn't valid... - // this isn't entirely true, only the would-have-overlapped data - // isn't valid, but this seems to be what the spec requires - return 0; - - // truncate a short frame - if (len < right) right = len; - - f->samples_output += right-left; - - return right - left; -} - -static void vorbis_pump_first_frame(stb_vorbis *f) -{ - int len, right, left; - if (vorbis_decode_packet(f, &len, &left, &right)) - vorbis_finish_frame(f, len, left, right); -} - -#ifndef STB_VORBIS_NO_PUSHDATA_API -static int is_whole_packet_present(stb_vorbis *f, int end_page) -{ - // make sure that we have the packet available before continuing... - // this requires a full ogg parse, but we know we can fetch from f->stream - - // instead of coding this out explicitly, we could save the current read state, - // read the next packet with get8() until end-of-packet, check f->eof, then - // reset the state? but that would be slower, esp. since we'd have over 256 bytes - // of state to restore (primarily the page segment table) - - int s = f->next_seg, first = TRUE; - uint8 *p = f->stream; - - if (s != -1) { // if we're not starting the packet with a 'continue on next page' flag - for (; s < f->segment_count; ++s) { - p += f->segments[s]; - if (f->segments[s] < 255) // stop at first short segment - break; - } - // either this continues, or it ends it... - if (end_page) - if (s < f->segment_count-1) return error(f, VORBIS_invalid_stream); - if (s == f->segment_count) - s = -1; // set 'crosses page' flag - if (p > f->stream_end) return error(f, VORBIS_need_more_data); - first = FALSE; - } - for (; s == -1;) { - uint8 *q; - int n; - - // check that we have the page header ready - if (p + 26 >= f->stream_end) return error(f, VORBIS_need_more_data); - // validate the page - if (memcmp(p, ogg_page_header, 4)) return error(f, VORBIS_invalid_stream); - if (p[4] != 0) return error(f, VORBIS_invalid_stream); - if (first) { // the first segment must NOT have 'continued_packet', later ones MUST - if (f->previous_length) - if ((p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); - // if no previous length, we're resynching, so we can come in on a continued-packet, - // which we'll just drop - } else { - if (!(p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream); - } - n = p[26]; // segment counts - q = p+27; // q points to segment table - p = q + n; // advance past header - // make sure we've read the segment table - if (p > f->stream_end) return error(f, VORBIS_need_more_data); - for (s=0; s < n; ++s) { - p += q[s]; - if (q[s] < 255) - break; - } - if (end_page) - if (s < n-1) return error(f, VORBIS_invalid_stream); - if (s == f->segment_count) - s = -1; // set 'crosses page' flag - if (p > f->stream_end) return error(f, VORBIS_need_more_data); - first = FALSE; - } - return TRUE; -} -#endif // !STB_VORBIS_NO_PUSHDATA_API - -static int start_decoder(vorb *f) -{ - uint8 header[6], x,y; - int len,i,j,k, max_submaps = 0; - int longest_floorlist=0; - - // first page, first packet - - if (!start_page(f)) return FALSE; - // validate page flag - if (!(f->page_flag & PAGEFLAG_first_page)) return error(f, VORBIS_invalid_first_page); - if (f->page_flag & PAGEFLAG_last_page) return error(f, VORBIS_invalid_first_page); - if (f->page_flag & PAGEFLAG_continued_packet) return error(f, VORBIS_invalid_first_page); - // check for expected packet length - if (f->segment_count != 1) return error(f, VORBIS_invalid_first_page); - if (f->segments[0] != 30) return error(f, VORBIS_invalid_first_page); - // read packet - // check packet header - if (get8(f) != VORBIS_packet_id) return error(f, VORBIS_invalid_first_page); - if (!getn(f, header, 6)) return error(f, VORBIS_unexpected_eof); - if (!vorbis_validate(header)) return error(f, VORBIS_invalid_first_page); - // vorbis_version - if (get32(f) != 0) return error(f, VORBIS_invalid_first_page); - f->channels = get8(f); if (!f->channels) return error(f, VORBIS_invalid_first_page); - if (f->channels > STB_VORBIS_MAX_CHANNELS) return error(f, VORBIS_too_many_channels); - f->sample_rate = get32(f); if (!f->sample_rate) return error(f, VORBIS_invalid_first_page); - get32(f); // bitrate_maximum - get32(f); // bitrate_nominal - get32(f); // bitrate_minimum - x = get8(f); - { int log0,log1; - log0 = x & 15; - log1 = x >> 4; - f->blocksize_0 = 1 << log0; - f->blocksize_1 = 1 << log1; - if (log0 < 6 || log0 > 13) return error(f, VORBIS_invalid_setup); - if (log1 < 6 || log1 > 13) return error(f, VORBIS_invalid_setup); - if (log0 > log1) return error(f, VORBIS_invalid_setup); - } - - // framing_flag - x = get8(f); - if (!(x & 1)) return error(f, VORBIS_invalid_first_page); - - // second packet! - if (!start_page(f)) return FALSE; - - if (!start_packet(f)) return FALSE; - do { - len = next_segment(f); - skip(f, len); - f->bytes_in_seg = 0; - } while (len); - - // third packet! - if (!start_packet(f)) return FALSE; - - #ifndef STB_VORBIS_NO_PUSHDATA_API - if (IS_PUSH_MODE(f)) { - if (!is_whole_packet_present(f, TRUE)) { - // convert error in ogg header to write type - if (f->error == VORBIS_invalid_stream) - f->error = VORBIS_invalid_setup; - return FALSE; - } - } - #endif - - crc32_init(); // always init it, to avoid multithread race conditions - - if (get8_packet(f) != VORBIS_packet_setup) return error(f, VORBIS_invalid_setup); - for (i=0; i < 6; ++i) header[i] = get8_packet(f); - if (!vorbis_validate(header)) return error(f, VORBIS_invalid_setup); - - // codebooks - - f->codebook_count = get_bits(f,8) + 1; - f->codebooks = (Codebook *) setup_malloc(f, sizeof(*f->codebooks) * f->codebook_count); - if (f->codebooks == NULL) return error(f, VORBIS_outofmem); - memset(f->codebooks, 0, sizeof(*f->codebooks) * f->codebook_count); - for (i=0; i < f->codebook_count; ++i) { - uint32 *values; - int ordered, sorted_count; - int total=0; - uint8 *lengths; - Codebook *c = f->codebooks+i; - x = get_bits(f, 8); if (x != 0x42) return error(f, VORBIS_invalid_setup); - x = get_bits(f, 8); if (x != 0x43) return error(f, VORBIS_invalid_setup); - x = get_bits(f, 8); if (x != 0x56) return error(f, VORBIS_invalid_setup); - x = get_bits(f, 8); - c->dimensions = (get_bits(f, 8)<<8) + x; - x = get_bits(f, 8); - y = get_bits(f, 8); - c->entries = (get_bits(f, 8)<<16) + (y<<8) + x; - ordered = get_bits(f,1); - c->sparse = ordered ? 0 : get_bits(f,1); - - if (c->sparse) - lengths = (uint8 *) setup_temp_malloc(f, c->entries); - else - lengths = c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); - - if (!lengths) return error(f, VORBIS_outofmem); - - if (ordered) { - int current_entry = 0; - int current_length = get_bits(f,5) + 1; - while (current_entry < c->entries) { - int limit = c->entries - current_entry; - int n = get_bits(f, ilog(limit)); - if (current_entry + n > (int) c->entries) { return error(f, VORBIS_invalid_setup); } - memset(lengths + current_entry, current_length, n); - current_entry += n; - ++current_length; - } - } else { - for (j=0; j < c->entries; ++j) { - int present = c->sparse ? get_bits(f,1) : 1; - if (present) { - lengths[j] = get_bits(f, 5) + 1; - ++total; - } else { - lengths[j] = NO_CODE; - } - } - } - - if (c->sparse && total >= c->entries >> 2) { - // convert sparse items to non-sparse! - if (c->entries > (int) f->setup_temp_memory_required) - f->setup_temp_memory_required = c->entries; - - c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries); - memcpy(c->codeword_lengths, lengths, c->entries); - setup_temp_free(f, lengths, c->entries); // note this is only safe if there have been no intervening temp mallocs! - lengths = c->codeword_lengths; - c->sparse = 0; - } - - // compute the size of the sorted tables - if (c->sparse) { - sorted_count = total; - //assert(total != 0); - } else { - sorted_count = 0; - #ifndef STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH - for (j=0; j < c->entries; ++j) - if (lengths[j] > STB_VORBIS_FAST_HUFFMAN_LENGTH && lengths[j] != NO_CODE) - ++sorted_count; - #endif - } - - c->sorted_entries = sorted_count; - values = NULL; - - if (!c->sparse) { - c->codewords = (uint32 *) setup_malloc(f, sizeof(c->codewords[0]) * c->entries); - if (!c->codewords) return error(f, VORBIS_outofmem); - } else { - unsigned int size; - if (c->sorted_entries) { - c->codeword_lengths = (uint8 *) setup_malloc(f, c->sorted_entries); - if (!c->codeword_lengths) return error(f, VORBIS_outofmem); - c->codewords = (uint32 *) setup_temp_malloc(f, sizeof(*c->codewords) * c->sorted_entries); - if (!c->codewords) return error(f, VORBIS_outofmem); - values = (uint32 *) setup_temp_malloc(f, sizeof(*values) * c->sorted_entries); - if (!values) return error(f, VORBIS_outofmem); - } - size = c->entries + (sizeof(*c->codewords) + sizeof(*values)) * c->sorted_entries; - if (size > f->setup_temp_memory_required) - f->setup_temp_memory_required = size; - } - - if (!compute_codewords(c, lengths, c->entries, values)) { - if (c->sparse) setup_temp_free(f, values, 0); - return error(f, VORBIS_invalid_setup); - } - - if (c->sorted_entries) { - // allocate an extra slot for sentinels - c->sorted_codewords = (uint32 *) setup_malloc(f, sizeof(*c->sorted_codewords) * (c->sorted_entries+1)); - // allocate an extra slot at the front so that c->sorted_values[-1] is defined - // so that we can catch that case without an extra if - c->sorted_values = ( int *) setup_malloc(f, sizeof(*c->sorted_values ) * (c->sorted_entries+1)); - if (c->sorted_values) { ++c->sorted_values; c->sorted_values[-1] = -1; } - compute_sorted_huffman(c, lengths, values); - } - - if (c->sparse) { - setup_temp_free(f, values, sizeof(*values)*c->sorted_entries); - setup_temp_free(f, c->codewords, sizeof(*c->codewords)*c->sorted_entries); - setup_temp_free(f, lengths, c->entries); - c->codewords = NULL; - } - - compute_accelerated_huffman(c); - - c->lookup_type = get_bits(f, 4); - if (c->lookup_type > 2) return error(f, VORBIS_invalid_setup); - if (c->lookup_type > 0) { - uint16 *mults; - c->minimum_value = float32_unpack(get_bits(f, 32)); - c->delta_value = float32_unpack(get_bits(f, 32)); - c->value_bits = get_bits(f, 4)+1; - c->sequence_p = get_bits(f,1); - if (c->lookup_type == 1) { - c->lookup_values = lookup1_values(c->entries, c->dimensions); - } else { - c->lookup_values = c->entries * c->dimensions; - } - mults = (uint16 *) setup_temp_malloc(f, sizeof(mults[0]) * c->lookup_values); - if (mults == NULL) return error(f, VORBIS_outofmem); - for (j=0; j < (int) c->lookup_values; ++j) { - int q = get_bits(f, c->value_bits); - if (q == EOP) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_invalid_setup); } - mults[j] = q; - } - -#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK - if (c->lookup_type == 1) { - int len, sparse = c->sparse; - // pre-expand the lookup1-style multiplicands, to avoid a divide in the inner loop - if (sparse) { - if (c->sorted_entries == 0) goto skip; - c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->sorted_entries * c->dimensions); - } else - c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->entries * c->dimensions); - if (c->multiplicands == NULL) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); } - len = sparse ? c->sorted_entries : c->entries; - for (j=0; j < len; ++j) { - int z = sparse ? c->sorted_values[j] : j, div=1; - for (k=0; k < c->dimensions; ++k) { - int off = (z / div) % c->lookup_values; - c->multiplicands[j*c->dimensions + k] = - #ifndef STB_VORBIS_CODEBOOK_FLOATS - mults[off]; - #else - mults[off]*c->delta_value + c->minimum_value; - // in this case (and this case only) we could pre-expand c->sequence_p, - // and throw away the decode logic for it; have to ALSO do - // it in the case below, but it can only be done if - // STB_VORBIS_CODEBOOK_FLOATS - // !STB_VORBIS_DIVIDES_IN_CODEBOOK - #endif - div *= c->lookup_values; - } - } - setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); - c->lookup_type = 2; - } - else -#endif - { - c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->lookup_values); - #ifndef STB_VORBIS_CODEBOOK_FLOATS - memcpy(c->multiplicands, mults, sizeof(c->multiplicands[0]) * c->lookup_values); - #else - for (j=0; j < (int) c->lookup_values; ++j) - c->multiplicands[j] = mults[j] * c->delta_value + c->minimum_value; - #endif - setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values); - } - skip:; - - #ifdef STB_VORBIS_CODEBOOK_FLOATS - if (c->lookup_type == 2 && c->sequence_p) { - for (j=1; j < (int) c->lookup_values; ++j) - c->multiplicands[j] = c->multiplicands[j-1]; - c->sequence_p = 0; - } - #endif - } - } - - // time domain transfers (notused) - - x = get_bits(f, 6) + 1; - for (i=0; i < x; ++i) { - uint32 z = get_bits(f, 16); - if (z != 0) return error(f, VORBIS_invalid_setup); - } - - // Floors - f->floor_count = get_bits(f, 6)+1; - f->floor_config = (Floor *) setup_malloc(f, f->floor_count * sizeof(*f->floor_config)); - for (i=0; i < f->floor_count; ++i) { - f->floor_types[i] = get_bits(f, 16); - if (f->floor_types[i] > 1) return error(f, VORBIS_invalid_setup); - if (f->floor_types[i] == 0) { - Floor0 *g = &f->floor_config[i].floor0; - g->order = get_bits(f,8); - g->rate = get_bits(f,16); - g->bark_map_size = get_bits(f,16); - g->amplitude_bits = get_bits(f,6); - g->amplitude_offset = get_bits(f,8); - g->number_of_books = get_bits(f,4) + 1; - for (j=0; j < g->number_of_books; ++j) - g->book_list[j] = get_bits(f,8); - return error(f, VORBIS_feature_not_supported); - } else { - Point p[31*8+2]; - Floor1 *g = &f->floor_config[i].floor1; - int max_class = -1; - g->partitions = get_bits(f, 5); - for (j=0; j < g->partitions; ++j) { - g->partition_class_list[j] = get_bits(f, 4); - if (g->partition_class_list[j] > max_class) - max_class = g->partition_class_list[j]; - } - for (j=0; j <= max_class; ++j) { - g->class_dimensions[j] = get_bits(f, 3)+1; - g->class_subclasses[j] = get_bits(f, 2); - if (g->class_subclasses[j]) { - g->class_masterbooks[j] = get_bits(f, 8); - if (g->class_masterbooks[j] >= f->codebook_count) return error(f, VORBIS_invalid_setup); - } - for (k=0; k < 1 << g->class_subclasses[j]; ++k) { - g->subclass_books[j][k] = get_bits(f,8)-1; - if (g->subclass_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); - } - } - g->floor1_multiplier = get_bits(f,2)+1; - g->rangebits = get_bits(f,4); - g->Xlist[0] = 0; - g->Xlist[1] = 1 << g->rangebits; - g->values = 2; - for (j=0; j < g->partitions; ++j) { - int c = g->partition_class_list[j]; - for (k=0; k < g->class_dimensions[c]; ++k) { - g->Xlist[g->values] = get_bits(f, g->rangebits); - ++g->values; - } - } - // precompute the sorting - for (j=0; j < g->values; ++j) { - p[j].x = g->Xlist[j]; - p[j].y = j; - } - qsort(p, g->values, sizeof(p[0]), point_compare); - for (j=0; j < g->values; ++j) - g->sorted_order[j] = (uint8) p[j].y; - // precompute the neighbors - for (j=2; j < g->values; ++j) { - int low,hi; - neighbors(g->Xlist, j, &low,&hi); - g->neighbors[j][0] = low; - g->neighbors[j][1] = hi; - } - - if (g->values > longest_floorlist) - longest_floorlist = g->values; - } - } - - // Residue - f->residue_count = get_bits(f, 6)+1; - f->residue_config = (Residue *) setup_malloc(f, f->residue_count * sizeof(*f->residue_config)); - for (i=0; i < f->residue_count; ++i) { - uint8 residue_cascade[64]; - Residue *r = f->residue_config+i; - f->residue_types[i] = get_bits(f, 16); - if (f->residue_types[i] > 2) return error(f, VORBIS_invalid_setup); - r->begin = get_bits(f, 24); - r->end = get_bits(f, 24); - r->part_size = get_bits(f,24)+1; - r->classifications = get_bits(f,6)+1; - r->classbook = get_bits(f,8); - for (j=0; j < r->classifications; ++j) { - uint8 high_bits=0; - uint8 low_bits=get_bits(f,3); - if (get_bits(f,1)) - high_bits = get_bits(f,5); - residue_cascade[j] = high_bits*8 + low_bits; - } - r->residue_books = (short (*)[8]) setup_malloc(f, sizeof(r->residue_books[0]) * r->classifications); - for (j=0; j < r->classifications; ++j) { - for (k=0; k < 8; ++k) { - if (residue_cascade[j] & (1 << k)) { - r->residue_books[j][k] = get_bits(f, 8); - if (r->residue_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup); - } else { - r->residue_books[j][k] = -1; - } - } - } - // precompute the classifications[] array to avoid inner-loop mod/divide - // call it 'classdata' since we already have r->classifications - r->classdata = (uint8 **) setup_malloc(f, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); - if (!r->classdata) return error(f, VORBIS_outofmem); - memset(r->classdata, 0, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); - for (j=0; j < f->codebooks[r->classbook].entries; ++j) { - int classwords = f->codebooks[r->classbook].dimensions; - int temp = j; - r->classdata[j] = (uint8 *) setup_malloc(f, sizeof(r->classdata[j][0]) * classwords); - for (k=classwords-1; k >= 0; --k) { - r->classdata[j][k] = temp % r->classifications; - temp /= r->classifications; - } - } - } - - f->mapping_count = get_bits(f,6)+1; - f->mapping = (Mapping *) setup_malloc(f, f->mapping_count * sizeof(*f->mapping)); - for (i=0; i < f->mapping_count; ++i) { - Mapping *m = f->mapping + i; - int mapping_type = get_bits(f,16); - if (mapping_type != 0) return error(f, VORBIS_invalid_setup); - m->chan = (MappingChannel *) setup_malloc(f, f->channels * sizeof(*m->chan)); - if (get_bits(f,1)) - m->submaps = get_bits(f,4); - else - m->submaps = 1; - if (m->submaps > max_submaps) - max_submaps = m->submaps; - if (get_bits(f,1)) { - m->coupling_steps = get_bits(f,8)+1; - for (k=0; k < m->coupling_steps; ++k) { - m->chan[k].magnitude = get_bits(f, ilog(f->channels)-1); - m->chan[k].angle = get_bits(f, ilog(f->channels)-1); - if (m->chan[k].magnitude >= f->channels) return error(f, VORBIS_invalid_setup); - if (m->chan[k].angle >= f->channels) return error(f, VORBIS_invalid_setup); - if (m->chan[k].magnitude == m->chan[k].angle) return error(f, VORBIS_invalid_setup); - } - } else - m->coupling_steps = 0; - - // reserved field - if (get_bits(f,2)) return error(f, VORBIS_invalid_setup); - if (m->submaps > 1) { - for (j=0; j < f->channels; ++j) { - m->chan[j].mux = get_bits(f, 4); - if (m->chan[j].mux >= m->submaps) return error(f, VORBIS_invalid_setup); - } - } else - // @SPECIFICATION: this case is missing from the spec - for (j=0; j < f->channels; ++j) - m->chan[j].mux = 0; - - for (j=0; j < m->submaps; ++j) { - get_bits(f,8); // discard - m->submap_floor[j] = get_bits(f,8); - m->submap_residue[j] = get_bits(f,8); - if (m->submap_floor[j] >= f->floor_count) return error(f, VORBIS_invalid_setup); - if (m->submap_residue[j] >= f->residue_count) return error(f, VORBIS_invalid_setup); - } - } - - // Modes - f->mode_count = get_bits(f, 6)+1; - for (i=0; i < f->mode_count; ++i) { - Mode *m = f->mode_config+i; - m->blockflag = get_bits(f,1); - m->windowtype = get_bits(f,16); - m->transformtype = get_bits(f,16); - m->mapping = get_bits(f,8); - if (m->windowtype != 0) return error(f, VORBIS_invalid_setup); - if (m->transformtype != 0) return error(f, VORBIS_invalid_setup); - if (m->mapping >= f->mapping_count) return error(f, VORBIS_invalid_setup); - } - - flush_packet(f); - - f->previous_length = 0; - - for (i=0; i < f->channels; ++i) { - f->channel_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1); - f->previous_window[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); - f->finalY[i] = (int16 *) setup_malloc(f, sizeof(int16) * longest_floorlist); - #ifdef STB_VORBIS_NO_DEFER_FLOOR - f->floor_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2); - #endif - } - - if (!init_blocksize(f, 0, f->blocksize_0)) return FALSE; - if (!init_blocksize(f, 1, f->blocksize_1)) return FALSE; - f->blocksize[0] = f->blocksize_0; - f->blocksize[1] = f->blocksize_1; - -#ifdef STB_VORBIS_DIVIDE_TABLE - if (integer_divide_table[1][1]==0) - for (i=0; i < DIVTAB_NUMER; ++i) - for (j=1; j < DIVTAB_DENOM; ++j) - integer_divide_table[i][j] = i / j; -#endif - - // compute how much temporary memory is needed - - // 1. - { - uint32 imdct_mem = (f->blocksize_1 * sizeof(float) >> 1); - uint32 classify_mem; - int i,max_part_read=0; - for (i=0; i < f->residue_count; ++i) { - Residue *r = f->residue_config + i; - int n_read = r->end - r->begin; - int part_read = n_read / r->part_size; - if (part_read > max_part_read) - max_part_read = part_read; - } - #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE - classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(uint8 *)); - #else - classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(int *)); - #endif - - f->temp_memory_required = classify_mem; - if (imdct_mem > f->temp_memory_required) - f->temp_memory_required = imdct_mem; - } - - f->first_decode = TRUE; - - if (f->alloc.alloc_buffer) { - assert(f->temp_offset == f->alloc.alloc_buffer_length_in_bytes); - // check if there's enough temp memory so we don't error later - if (f->setup_offset + sizeof(*f) + f->temp_memory_required > (unsigned) f->temp_offset) - return error(f, VORBIS_outofmem); - } - - f->first_audio_page_offset = stb_vorbis_get_file_offset(f); - - return TRUE; -} - -static void vorbis_deinit(stb_vorbis *p) -{ - int i,j; - for (i=0; i < p->residue_count; ++i) { - Residue *r = p->residue_config+i; - if (r->classdata) { - for (j=0; j < p->codebooks[r->classbook].entries; ++j) - setup_free(p, r->classdata[j]); - setup_free(p, r->classdata); - } - setup_free(p, r->residue_books); - } - - if (p->codebooks) { - for (i=0; i < p->codebook_count; ++i) { - Codebook *c = p->codebooks + i; - setup_free(p, c->codeword_lengths); - setup_free(p, c->multiplicands); - setup_free(p, c->codewords); - setup_free(p, c->sorted_codewords); - // c->sorted_values[-1] is the first entry in the array - setup_free(p, c->sorted_values ? c->sorted_values-1 : NULL); - } - setup_free(p, p->codebooks); - } - setup_free(p, p->floor_config); - setup_free(p, p->residue_config); - for (i=0; i < p->mapping_count; ++i) - setup_free(p, p->mapping[i].chan); - setup_free(p, p->mapping); - for (i=0; i < p->channels; ++i) { - setup_free(p, p->channel_buffers[i]); - setup_free(p, p->previous_window[i]); - #ifdef STB_VORBIS_NO_DEFER_FLOOR - setup_free(p, p->floor_buffers[i]); - #endif - setup_free(p, p->finalY[i]); - } - for (i=0; i < 2; ++i) { - setup_free(p, p->A[i]); - setup_free(p, p->B[i]); - setup_free(p, p->C[i]); - setup_free(p, p->window[i]); +// Ogg Vorbis I audio decoder -- version 0.99994
+//
+// Written in April 2007 by Sean Barrett, sponsored by RAD Game Tools.
+//
+// Placed in the public domain April 2007 by the author: no copyright is
+// claimed, and you may use it for any purpose you like.
+//
+// No warranty for any purpose is expressed or implied by the author (nor
+// by RAD Game Tools). Report bugs and send enhancements to the author.
+//
+// Get the latest version and other information at:
+// http://nothings.org/stb_vorbis/
+
+
+// Todo:
+//
+// - seeking (note you can seek yourself using the pushdata API)
+//
+// Limitations:
+//
+// - floor 0 not supported (used in old ogg vorbis files)
+// - lossless sample-truncation at beginning ignored
+// - cannot concatenate multiple vorbis streams
+// - sample positions are 32-bit, limiting seekable 192Khz
+// files to around 6 hours (Ogg supports 64-bit)
+//
+// All of these limitations may be removed in future versions.
+
+#include "stb_vorbis.h"
+
+#ifndef STB_VORBIS_HEADER_ONLY
+
+// global configuration settings (e.g. set these in the project/makefile),
+// or just set them in this file at the top (although ideally the first few
+// should be visible when the header file is compiled too, although it's not
+// crucial)
+
+// STB_VORBIS_NO_PUSHDATA_API
+// does not compile the code for the various stb_vorbis_*_pushdata()
+// functions
+// #define STB_VORBIS_NO_PUSHDATA_API
+
+// STB_VORBIS_NO_PULLDATA_API
+// does not compile the code for the non-pushdata APIs
+// #define STB_VORBIS_NO_PULLDATA_API
+
+// STB_VORBIS_NO_STDIO
+// does not compile the code for the APIs that use FILE *s internally
+// or externally (implied by STB_VORBIS_NO_PULLDATA_API)
+// #define STB_VORBIS_NO_STDIO
+
+// STB_VORBIS_NO_INTEGER_CONVERSION
+// does not compile the code for converting audio sample data from
+// float to integer (implied by STB_VORBIS_NO_PULLDATA_API)
+// #define STB_VORBIS_NO_INTEGER_CONVERSION
+
+// STB_VORBIS_NO_FAST_SCALED_FLOAT
+// does not use a fast float-to-int trick to accelerate float-to-int on
+// most platforms which requires endianness be defined correctly.
+#define STB_VORBIS_NO_FAST_SCALED_FLOAT
+
+
+// STB_VORBIS_MAX_CHANNELS [number]
+// globally define this to the maximum number of channels you need.
+// The spec does not put a restriction on channels except that
+// the count is stored in a byte, so 255 is the hard limit.
+// Reducing this saves about 16 bytes per value, so using 16 saves
+// (255-16)*16 or around 4KB. Plus anything other memory usage
+// I forgot to account for. Can probably go as low as 8 (7.1 audio),
+// 6 (5.1 audio), or 2 (stereo only).
+#ifndef STB_VORBIS_MAX_CHANNELS
+#define STB_VORBIS_MAX_CHANNELS 16 // enough for anyone?
+#endif
+
+// STB_VORBIS_PUSHDATA_CRC_COUNT [number]
+// after a flush_pushdata(), stb_vorbis begins scanning for the
+// next valid page, without backtracking. when it finds something
+// that looks like a page, it streams through it and verifies its
+// CRC32. Should that validation fail, it keeps scanning. But it's
+// possible that _while_ streaming through to check the CRC32 of
+// one candidate page, it sees another candidate page. This #define
+// determines how many "overlapping" candidate pages it can search
+// at once. Note that "real" pages are typically ~4KB to ~8KB, whereas
+// garbage pages could be as big as 64KB, but probably average ~16KB.
+// So don't hose ourselves by scanning an apparent 64KB page and
+// missing a ton of real ones in the interim; so minimum of 2
+#ifndef STB_VORBIS_PUSHDATA_CRC_COUNT
+#define STB_VORBIS_PUSHDATA_CRC_COUNT 4
+#endif
+
+// STB_VORBIS_FAST_HUFFMAN_LENGTH [number]
+// sets the log size of the huffman-acceleration table. Maximum
+// supported value is 24. with larger numbers, more decodings are O(1),
+// but the table size is larger so worse cache missing, so you'll have
+// to probe (and try multiple ogg vorbis files) to find the sweet spot.
+#ifndef STB_VORBIS_FAST_HUFFMAN_LENGTH
+#define STB_VORBIS_FAST_HUFFMAN_LENGTH 10
+#endif
+
+// STB_VORBIS_FAST_BINARY_LENGTH [number]
+// sets the log size of the binary-search acceleration table. this
+// is used in similar fashion to the fast-huffman size to set initial
+// parameters for the binary search
+
+// STB_VORBIS_FAST_HUFFMAN_INT
+// The fast huffman tables are much more efficient if they can be
+// stored as 16-bit results instead of 32-bit results. This restricts
+// the codebooks to having only 65535 possible outcomes, though.
+// (At least, accelerated by the huffman table.)
+#ifndef STB_VORBIS_FAST_HUFFMAN_INT
+#define STB_VORBIS_FAST_HUFFMAN_SHORT
+#endif
+
+// STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH
+// If the 'fast huffman' search doesn't succeed, then stb_vorbis falls
+// back on binary searching for the correct one. This requires storing
+// extra tables with the huffman codes in sorted order. Defining this
+// symbol trades off space for speed by forcing a linear search in the
+// non-fast case, except for "sparse" codebooks.
+// #define STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH
+
+// STB_VORBIS_DIVIDES_IN_RESIDUE
+// stb_vorbis precomputes the result of the scalar residue decoding
+// that would otherwise require a divide per chunk. you can trade off
+// space for time by defining this symbol.
+// #define STB_VORBIS_DIVIDES_IN_RESIDUE
+
+// STB_VORBIS_DIVIDES_IN_CODEBOOK
+// vorbis VQ codebooks can be encoded two ways: with every case explicitly
+// stored, or with all elements being chosen from a small range of values,
+// and all values possible in all elements. By default, stb_vorbis expands
+// this latter kind out to look like the former kind for ease of decoding,
+// because otherwise an integer divide-per-vector-element is required to
+// unpack the index. If you define STB_VORBIS_DIVIDES_IN_CODEBOOK, you can
+// trade off storage for speed.
+//#define STB_VORBIS_DIVIDES_IN_CODEBOOK
+
+// STB_VORBIS_CODEBOOK_SHORTS
+// The vorbis file format encodes VQ codebook floats as ax+b where a and
+// b are floating point per-codebook constants, and x is a 16-bit int.
+// Normally, stb_vorbis decodes them to floats rather than leaving them
+// as 16-bit ints and computing ax+b while decoding. This is a speed/space
+// tradeoff; you can save space by defining this flag.
+#ifndef STB_VORBIS_CODEBOOK_SHORTS
+#define STB_VORBIS_CODEBOOK_FLOATS
+#endif
+
+// STB_VORBIS_DIVIDE_TABLE
+// this replaces small integer divides in the floor decode loop with
+// table lookups. made less than 1% difference, so disabled by default.
+
+// STB_VORBIS_NO_INLINE_DECODE
+// disables the inlining of the scalar codebook fast-huffman decode.
+// might save a little codespace; useful for debugging
+// #define STB_VORBIS_NO_INLINE_DECODE
+
+// STB_VORBIS_NO_DEFER_FLOOR
+// Normally we only decode the floor without synthesizing the actual
+// full curve. We can instead synthesize the curve immediately. This
+// requires more memory and is very likely slower, so I don't think
+// you'd ever want to do it except for debugging.
+// #define STB_VORBIS_NO_DEFER_FLOOR
+
+
+
+
+//////////////////////////////////////////////////////////////////////////////
+
+#ifdef STB_VORBIS_NO_PULLDATA_API
+ #define STB_VORBIS_NO_INTEGER_CONVERSION
+ #define STB_VORBIS_NO_STDIO
+#endif
+
+#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO)
+ #define STB_VORBIS_NO_STDIO 1
+#endif
+
+#ifndef STB_VORBIS_NO_INTEGER_CONVERSION
+#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT
+
+ // only need endianness for fast-float-to-int, which we don't
+ // use for pushdata
+
+ #ifndef STB_VORBIS_BIG_ENDIAN
+ #define STB_VORBIS_ENDIAN 0
+ #else
+ #define STB_VORBIS_ENDIAN 1
+ #endif
+
+#endif
+#endif
+
+
+#ifndef STB_VORBIS_NO_STDIO
+#include <stdio.h>
+#endif
+
+#ifndef STB_VORBIS_NO_CRT
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+#include <math.h>
+
+#if !defined(__APPLE__) && !defined(MACOSX) && !defined(macintosh) && !defined(Macintosh) &&!defined(__FreeBSD__)
+#include <malloc.h>
+#endif
+
+#else
+#define NULL 0
+#endif
+
+#ifndef _MSC_VER
+ #if __GNUC__
+ #define __forceinline inline
+ #else
+ #define __forceinline
+ #endif
+#endif
+
+#if STB_VORBIS_MAX_CHANNELS > 256
+#error "Value of STB_VORBIS_MAX_CHANNELS outside of allowed range"
+#endif
+
+#if STB_VORBIS_FAST_HUFFMAN_LENGTH > 24
+#error "Value of STB_VORBIS_FAST_HUFFMAN_LENGTH outside of allowed range"
+#endif
+
+
+#define MAX_BLOCKSIZE_LOG 13 // from specification
+#define MAX_BLOCKSIZE (1 << MAX_BLOCKSIZE_LOG)
+
+
+typedef unsigned char uint8;
+typedef signed char int8;
+typedef unsigned short uint16;
+typedef signed short int16;
+typedef unsigned int uint32;
+typedef signed int int32;
+
+#ifndef TRUE
+#define TRUE 1
+#define FALSE 0
+#endif
+
+#ifdef STB_VORBIS_CODEBOOK_FLOATS
+typedef float codetype;
+#else
+typedef uint16 codetype;
+#endif
+
+// @NOTE
+//
+// Some arrays below are tagged "//varies", which means it's actually
+// a variable-sized piece of data, but rather than malloc I assume it's
+// small enough it's better to just allocate it all together with the
+// main thing
+//
+// Most of the variables are specified with the smallest size I could pack
+// them into. It might give better performance to make them all full-sized
+// integers. It should be safe to freely rearrange the structures or change
+// the sizes larger--nothing relies on silently truncating etc., nor the
+// order of variables.
+
+#define FAST_HUFFMAN_TABLE_SIZE (1 << STB_VORBIS_FAST_HUFFMAN_LENGTH)
+#define FAST_HUFFMAN_TABLE_MASK (FAST_HUFFMAN_TABLE_SIZE - 1)
+
+typedef struct
+{
+ int dimensions, entries;
+ uint8 *codeword_lengths;
+ float minimum_value;
+ float delta_value;
+ uint8 value_bits;
+ uint8 lookup_type;
+ uint8 sequence_p;
+ uint8 sparse;
+ uint32 lookup_values;
+ codetype *multiplicands;
+ uint32 *codewords;
+ #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT
+ int16 fast_huffman[FAST_HUFFMAN_TABLE_SIZE];
+ #else
+ int32 fast_huffman[FAST_HUFFMAN_TABLE_SIZE];
+ #endif
+ uint32 *sorted_codewords;
+ int *sorted_values;
+ int sorted_entries;
+} Codebook;
+
+typedef struct
+{
+ uint8 order;
+ uint16 rate;
+ uint16 bark_map_size;
+ uint8 amplitude_bits;
+ uint8 amplitude_offset;
+ uint8 number_of_books;
+ uint8 book_list[16]; // varies
+} Floor0;
+
+typedef struct
+{
+ uint8 partitions;
+ uint8 partition_class_list[32]; // varies
+ uint8 class_dimensions[16]; // varies
+ uint8 class_subclasses[16]; // varies
+ uint8 class_masterbooks[16]; // varies
+ int16 subclass_books[16][8]; // varies
+ uint16 Xlist[31*8+2]; // varies
+ uint8 sorted_order[31*8+2];
+ uint8 neighbors[31*8+2][2];
+ uint8 floor1_multiplier;
+ uint8 rangebits;
+ int values;
+} Floor1;
+
+typedef union
+{
+ Floor0 floor0;
+ Floor1 floor1;
+} Floor;
+
+typedef struct
+{
+ uint32 begin, end;
+ uint32 part_size;
+ uint8 classifications;
+ uint8 classbook;
+ uint8 **classdata;
+ int16 (*residue_books)[8];
+} Residue;
+
+typedef struct
+{
+ uint8 magnitude;
+ uint8 angle;
+ uint8 mux;
+} MappingChannel;
+
+typedef struct
+{
+ uint16 coupling_steps;
+ MappingChannel *chan;
+ uint8 submaps;
+ uint8 submap_floor[15]; // varies
+ uint8 submap_residue[15]; // varies
+} Mapping;
+
+typedef struct
+{
+ uint8 blockflag;
+ uint8 mapping;
+ uint16 windowtype;
+ uint16 transformtype;
+} Mode;
+
+typedef struct
+{
+ uint32 goal_crc; // expected crc if match
+ int bytes_left; // bytes left in packet
+ uint32 crc_so_far; // running crc
+ int bytes_done; // bytes processed in _current_ chunk
+ uint32 sample_loc; // granule pos encoded in page
+} CRCscan;
+
+typedef struct
+{
+ uint32 page_start, page_end;
+ uint32 after_previous_page_start;
+ uint32 first_decoded_sample;
+ uint32 last_decoded_sample;
+} ProbedPage;
+
+struct stb_vorbis
+{
+ // user-accessible info
+ unsigned int sample_rate;
+ int channels;
+
+ unsigned int setup_memory_required;
+ unsigned int temp_memory_required;
+ unsigned int setup_temp_memory_required;
+
+ // input config
+#ifndef STB_VORBIS_NO_STDIO
+ FILE *f;
+ uint32 f_start;
+ int close_on_free;
+#endif
+
+ uint8 *stream;
+ uint8 *stream_start;
+ uint8 *stream_end;
+
+ uint32 stream_len;
+
+ uint8 push_mode;
+
+ uint32 first_audio_page_offset;
+
+ ProbedPage p_first, p_last;
+
+ // memory management
+ stb_vorbis_alloc alloc;
+ int setup_offset;
+ int temp_offset;
+
+ // run-time results
+ int eof;
+ enum STBVorbisError error;
+
+ // user-useful data
+
+ // header info
+ int blocksize[2];
+ int blocksize_0, blocksize_1;
+ int codebook_count;
+ Codebook *codebooks;
+ int floor_count;
+ uint16 floor_types[64]; // varies
+ Floor *floor_config;
+ int residue_count;
+ uint16 residue_types[64]; // varies
+ Residue *residue_config;
+ int mapping_count;
+ Mapping *mapping;
+ int mode_count;
+ Mode mode_config[64]; // varies
+
+ uint32 total_samples;
+
+ // decode buffer
+ float *channel_buffers[STB_VORBIS_MAX_CHANNELS];
+ float *outputs [STB_VORBIS_MAX_CHANNELS];
+
+ float *previous_window[STB_VORBIS_MAX_CHANNELS];
+ int previous_length;
+
+ #ifndef STB_VORBIS_NO_DEFER_FLOOR
+ int16 *finalY[STB_VORBIS_MAX_CHANNELS];
+ #else
+ float *floor_buffers[STB_VORBIS_MAX_CHANNELS];
+ #endif
+
+ uint32 current_loc; // sample location of next frame to decode
+ int current_loc_valid;
+
+ // per-blocksize precomputed data
+
+ // twiddle factors
+ float *A[2],*B[2],*C[2];
+ float *window[2];
+ uint16 *bit_reverse[2];
+
+ // current page/packet/segment streaming info
+ uint32 serial; // stream serial number for verification
+ int last_page;
+ int segment_count;
+ uint8 segments[255];
+ uint8 page_flag;
+ uint8 bytes_in_seg;
+ uint8 first_decode;
+ int next_seg;
+ int last_seg; // flag that we're on the last segment
+ int last_seg_which; // what was the segment number of the last seg?
+ uint32 acc;
+ int valid_bits;
+ int packet_bytes;
+ int end_seg_with_known_loc;
+ uint32 known_loc_for_packet;
+ int discard_samples_deferred;
+ uint32 samples_output;
+
+ // push mode scanning
+ int page_crc_tests; // only in push_mode: number of tests active; -1 if not searching
+#ifndef STB_VORBIS_NO_PUSHDATA_API
+ CRCscan scan[STB_VORBIS_PUSHDATA_CRC_COUNT];
+#endif
+
+ // sample-access
+ int channel_buffer_start;
+ int channel_buffer_end;
+};
+
+extern int my_prof(int slot);
+//#define stb_prof my_prof
+
+#ifndef stb_prof
+#define stb_prof(x) 0
+#endif
+
+#if defined(STB_VORBIS_NO_PUSHDATA_API)
+ #define IS_PUSH_MODE(f) FALSE
+#elif defined(STB_VORBIS_NO_PULLDATA_API)
+ #define IS_PUSH_MODE(f) TRUE
+#else
+ #define IS_PUSH_MODE(f) ((f)->push_mode)
+#endif
+
+typedef struct stb_vorbis vorb;
+
+static int error(vorb *f, enum STBVorbisError e)
+{
+ f->error = e;
+ if (!f->eof && e != VORBIS_need_more_data) {
+ f->error=e; // breakpoint for debugging
+ }
+ return 0;
+}
+
+
+// these functions are used for allocating temporary memory
+// while decoding. if you can afford the stack space, use
+// alloca(); otherwise, provide a temp buffer and it will
+// allocate out of those.
+
+#define array_size_required(count,size) (count*(sizeof(void *)+(size)))
+
+#define temp_alloc(f,size) (f->alloc.alloc_buffer ? setup_temp_malloc(f,size) : alloca(size))
+#ifdef dealloca
+#define temp_free(f,p) (f->alloc.alloc_buffer ? 0 : dealloca(size))
+#else
+#define temp_free(f,p) 0
+#endif
+#define temp_alloc_save(f) ((f)->temp_offset)
+#define temp_alloc_restore(f,p) ((f)->temp_offset = (p))
+
+#define temp_block_array(f,count,size) make_block_array(temp_alloc(f,array_size_required(count,size)), count, size)
+
+// given a sufficiently large block of memory, make an array of pointers to subblocks of it
+static void *make_block_array(void *mem, int count, int size)
+{
+ int i;
+ void ** p = (void **) mem;
+ char *q = (char *) (p + count);
+ for (i=0; i < count; ++i) {
+ p[i] = q;
+ q += size;
+ }
+ return p;
+}
+
+static void *setup_malloc(vorb *f, int sz)
+{
+ sz = (sz+3) & ~3;
+ f->setup_memory_required += sz;
+ if (f->alloc.alloc_buffer) {
+ void *p = (char *) f->alloc.alloc_buffer + f->setup_offset;
+ if (f->setup_offset + sz > f->temp_offset) return NULL;
+ f->setup_offset += sz;
+ return p;
+ }
+ return sz ? malloc(sz) : NULL;
+}
+
+static void setup_free(vorb *f, void *p)
+{
+ if (f->alloc.alloc_buffer) return; // do nothing; setup mem is not a stack
+ free(p);
+}
+
+static void *setup_temp_malloc(vorb *f, int sz)
+{
+ sz = (sz+3) & ~3;
+ if (f->alloc.alloc_buffer) {
+ if (f->temp_offset - sz < f->setup_offset) return NULL;
+ f->temp_offset -= sz;
+ return (char *) f->alloc.alloc_buffer + f->temp_offset;
+ }
+ return malloc(sz);
+}
+
+static void setup_temp_free(vorb *f, void *p, size_t sz)
+{
+ if (f->alloc.alloc_buffer) {
+ f->temp_offset += (sz+3)&~3;
+ return;
+ }
+ free(p);
+}
+
+#define CRC32_POLY 0x04c11db7 // from spec
+
+static uint32 crc_table[256];
+static void crc32_init(void)
+{
+ int i,j;
+ uint32 s;
+ for(i=0; i < 256; i++) {
+ for (s=i<<24, j=0; j < 8; ++j)
+ s = (s << 1) ^ (s >= (1<<31) ? CRC32_POLY : 0);
+ crc_table[i] = s;
+ }
+}
+
+static __forceinline uint32 crc32_update(uint32 crc, uint8 byte)
+{
+ return (crc << 8) ^ crc_table[byte ^ (crc >> 24)];
+}
+
+
+// used in setup, and for huffman that doesn't go fast path
+static unsigned int bit_reverse(unsigned int n)
+{
+ n = ((n & 0xAAAAAAAA) >> 1) | ((n & 0x55555555) << 1);
+ n = ((n & 0xCCCCCCCC) >> 2) | ((n & 0x33333333) << 2);
+ n = ((n & 0xF0F0F0F0) >> 4) | ((n & 0x0F0F0F0F) << 4);
+ n = ((n & 0xFF00FF00) >> 8) | ((n & 0x00FF00FF) << 8);
+ return (n >> 16) | (n << 16);
+}
+
+static float square(float x)
+{
+ return x*x;
+}
+
+// this is a weird definition of log2() for which log2(1) = 1, log2(2) = 2, log2(4) = 3
+// as required by the specification. fast(?) implementation from stb.h
+// @OPTIMIZE: called multiple times per-packet with "constants"; move to setup
+static int ilog(int32 n)
+{
+ static signed char log2_4[16] = { 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4 };
+
+ // 2 compares if n < 16, 3 compares otherwise (4 if signed or n > 1<<29)
+ if (n < (1U << 14))
+ if (n < (1U << 4)) return 0 + log2_4[n ];
+ else if (n < (1U << 9)) return 5 + log2_4[n >> 5];
+ else return 10 + log2_4[n >> 10];
+ else if (n < (1U << 24))
+ if (n < (1U << 19)) return 15 + log2_4[n >> 15];
+ else return 20 + log2_4[n >> 20];
+ else if (n < (1U << 29)) return 25 + log2_4[n >> 25];
+ else if (n < (1U << 31)) return 30 + log2_4[n >> 30];
+ else return 0; // signed n returns 0
+}
+
+#ifndef M_PI
+ #define M_PI 3.14159265358979323846264f // from CRC
+#endif
+
+// code length assigned to a value with no huffman encoding
+#define NO_CODE 255
+
+/////////////////////// LEAF SETUP FUNCTIONS //////////////////////////
+//
+// these functions are only called at setup, and only a few times
+// per file
+
+static float float32_unpack(uint32 x)
+{
+ // from the specification
+ uint32 mantissa = x & 0x1fffff;
+ uint32 sign = x & 0x80000000;
+ uint32 exp = (x & 0x7fe00000) >> 21;
+ double res = sign ? -(double)mantissa : (double)mantissa;
+ return (float) ldexp((float)res, exp-788);
+}
+
+
+// zlib & jpeg huffman tables assume that the output symbols
+// can either be arbitrarily arranged, or have monotonically
+// increasing frequencies--they rely on the lengths being sorted;
+// this makes for a very simple generation algorithm.
+// vorbis allows a huffman table with non-sorted lengths. This
+// requires a more sophisticated construction, since symbols in
+// order do not map to huffman codes "in order".
+static void add_entry(Codebook *c, uint32 huff_code, int symbol, int count, int len, uint32 *values)
+{
+ if (!c->sparse) {
+ c->codewords [symbol] = huff_code;
+ } else {
+ c->codewords [count] = huff_code;
+ c->codeword_lengths[count] = len;
+ values [count] = symbol;
+ }
+}
+
+static int compute_codewords(Codebook *c, uint8 *len, int n, uint32 *values)
+{
+ int i,k,m=0;
+ uint32 available[32];
+
+ memset(available, 0, sizeof(available));
+ // find the first entry
+ for (k=0; k < n; ++k) if (len[k] < NO_CODE) break;
+ if (k == n) { assert(c->sorted_entries == 0); return TRUE; }
+ // add to the list
+ add_entry(c, 0, k, m++, len[k], values);
+ // add all available leaves
+ for (i=1; i <= len[k]; ++i)
+ available[i] = 1 << (32-i);
+ // note that the above code treats the first case specially,
+ // but it's really the same as the following code, so they
+ // could probably be combined (except the initial code is 0,
+ // and I use 0 in available[] to mean 'empty')
+ for (i=k+1; i < n; ++i) {
+ uint32 res;
+ int z = len[i], y;
+ if (z == NO_CODE) continue;
+ // find lowest available leaf (should always be earliest,
+ // which is what the specification calls for)
+ // note that this property, and the fact we can never have
+ // more than one free leaf at a given level, isn't totally
+ // trivial to prove, but it seems true and the assert never
+ // fires, so!
+ while (z > 0 && !available[z]) --z;
+ if (z == 0) { assert(0); return FALSE; }
+ res = available[z];
+ available[z] = 0;
+ add_entry(c, bit_reverse(res), i, m++, len[i], values);
+ // propogate availability up the tree
+ if (z != len[i]) {
+ for (y=len[i]; y > z; --y) {
+ assert(available[y] == 0);
+ available[y] = res + (1 << (32-y));
+ }
+ }
+ }
+ return TRUE;
+}
+
+// accelerated huffman table allows fast O(1) match of all symbols
+// of length <= STB_VORBIS_FAST_HUFFMAN_LENGTH
+static void compute_accelerated_huffman(Codebook *c)
+{
+ int i, len;
+ for (i=0; i < FAST_HUFFMAN_TABLE_SIZE; ++i)
+ c->fast_huffman[i] = -1;
+
+ len = c->sparse ? c->sorted_entries : c->entries;
+ #ifdef STB_VORBIS_FAST_HUFFMAN_SHORT
+ if (len > 32767) len = 32767; // largest possible value we can encode!
+ #endif
+ for (i=0; i < len; ++i) {
+ if (c->codeword_lengths[i] <= STB_VORBIS_FAST_HUFFMAN_LENGTH) {
+ uint32 z = c->sparse ? bit_reverse(c->sorted_codewords[i]) : c->codewords[i];
+ // set table entries for all bit combinations in the higher bits
+ while (z < FAST_HUFFMAN_TABLE_SIZE) {
+ c->fast_huffman[z] = i;
+ z += 1 << c->codeword_lengths[i];
+ }
+ }
+ }
+}
+
+static int uint32_compare(const void *p, const void *q)
+{
+ uint32 x = * (uint32 *) p;
+ uint32 y = * (uint32 *) q;
+ return x < y ? -1 : x > y;
+}
+
+static int include_in_sort(Codebook *c, uint8 len)
+{
+ if (c->sparse) { assert(len != NO_CODE); return TRUE; }
+ if (len == NO_CODE) return FALSE;
+ if (len > STB_VORBIS_FAST_HUFFMAN_LENGTH) return TRUE;
+ return FALSE;
+}
+
+// if the fast table above doesn't work, we want to binary
+// search them... need to reverse the bits
+static void compute_sorted_huffman(Codebook *c, uint8 *lengths, uint32 *values)
+{
+ int i, len;
+ // build a list of all the entries
+ // OPTIMIZATION: don't include the short ones, since they'll be caught by FAST_HUFFMAN.
+ // this is kind of a frivolous optimization--I don't see any performance improvement,
+ // but it's like 4 extra lines of code, so.
+ if (!c->sparse) {
+ int k = 0;
+ for (i=0; i < c->entries; ++i)
+ if (include_in_sort(c, lengths[i]))
+ c->sorted_codewords[k++] = bit_reverse(c->codewords[i]);
+ assert(k == c->sorted_entries);
+ } else {
+ for (i=0; i < c->sorted_entries; ++i)
+ c->sorted_codewords[i] = bit_reverse(c->codewords[i]);
+ }
+
+ qsort(c->sorted_codewords, c->sorted_entries, sizeof(c->sorted_codewords[0]), uint32_compare);
+ c->sorted_codewords[c->sorted_entries] = 0xffffffff;
+
+ len = c->sparse ? c->sorted_entries : c->entries;
+ // now we need to indicate how they correspond; we could either
+ // #1: sort a different data structure that says who they correspond to
+ // #2: for each sorted entry, search the original list to find who corresponds
+ // #3: for each original entry, find the sorted entry
+ // #1 requires extra storage, #2 is slow, #3 can use binary search!
+ for (i=0; i < len; ++i) {
+ int huff_len = c->sparse ? lengths[values[i]] : lengths[i];
+ if (include_in_sort(c,huff_len)) {
+ uint32 code = bit_reverse(c->codewords[i]);
+ int x=0, n=c->sorted_entries;
+ while (n > 1) {
+ // invariant: sc[x] <= code < sc[x+n]
+ int m = x + (n >> 1);
+ if (c->sorted_codewords[m] <= code) {
+ x = m;
+ n -= (n>>1);
+ } else {
+ n >>= 1;
+ }
+ }
+ assert(c->sorted_codewords[x] == code);
+ if (c->sparse) {
+ c->sorted_values[x] = values[i];
+ c->codeword_lengths[x] = huff_len;
+ } else {
+ c->sorted_values[x] = i;
+ }
+ }
+ }
+}
+
+// only run while parsing the header (3 times)
+static int vorbis_validate(uint8 *data)
+{
+ static uint8 vorbis[6] = { 'v', 'o', 'r', 'b', 'i', 's' };
+ return memcmp(data, vorbis, 6) == 0;
+}
+
+// called from setup only, once per code book
+// (formula implied by specification)
+static int lookup1_values(int entries, int dim)
+{
+ int r = (int) floor(exp((float) log((float) entries) / dim));
+ if ((int) floor(pow((float) r+1, dim)) <= entries) // (int) cast for MinGW warning;
+ ++r; // floor() to avoid _ftol() when non-CRT
+ assert(pow((float) r+1, dim) > entries);
+ assert((int) floor(pow((float) r, dim)) <= entries); // (int),floor() as above
+ return r;
+}
+
+// called twice per file
+static void compute_twiddle_factors(int n, float *A, float *B, float *C)
+{
+ int n4 = n >> 2, n8 = n >> 3;
+ int k,k2;
+
+ for (k=k2=0; k < n4; ++k,k2+=2) {
+ A[k2 ] = (float) cos(4*k*M_PI/n);
+ A[k2+1] = (float) -sin(4*k*M_PI/n);
+ B[k2 ] = (float) cos((k2+1)*M_PI/n/2) * 0.5f;
+ B[k2+1] = (float) sin((k2+1)*M_PI/n/2) * 0.5f;
+ }
+ for (k=k2=0; k < n8; ++k,k2+=2) {
+ C[k2 ] = (float) cos(2*(k2+1)*M_PI/n);
+ C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n);
+ }
+}
+
+static void compute_window(int n, float *window)
+{
+ int n2 = n >> 1, i;
+ for (i=0; i < n2; ++i)
+ window[i] = (float) sin(0.5 * M_PI * square((float) sin((i - 0 + 0.5) / n2 * 0.5 * M_PI)));
+}
+
+static void compute_bitreverse(int n, uint16 *rev)
+{
+ int ld = ilog(n) - 1; // ilog is off-by-one from normal definitions
+ int i, n8 = n >> 3;
+ for (i=0; i < n8; ++i)
+ rev[i] = (bit_reverse(i) >> (32-ld+3)) << 2;
+}
+
+static int init_blocksize(vorb *f, int b, int n)
+{
+ int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3;
+ f->A[b] = (float *) setup_malloc(f, sizeof(float) * n2);
+ f->B[b] = (float *) setup_malloc(f, sizeof(float) * n2);
+ f->C[b] = (float *) setup_malloc(f, sizeof(float) * n4);
+ if (!f->A[b] || !f->B[b] || !f->C[b]) return error(f, VORBIS_outofmem);
+ compute_twiddle_factors(n, f->A[b], f->B[b], f->C[b]);
+ f->window[b] = (float *) setup_malloc(f, sizeof(float) * n2);
+ if (!f->window[b]) return error(f, VORBIS_outofmem);
+ compute_window(n, f->window[b]);
+ f->bit_reverse[b] = (uint16 *) setup_malloc(f, sizeof(uint16) * n8);
+ if (!f->bit_reverse[b]) return error(f, VORBIS_outofmem);
+ compute_bitreverse(n, f->bit_reverse[b]);
+ return TRUE;
+}
+
+static void neighbors(uint16 *x, int n, int *plow, int *phigh)
+{
+ int low = -1;
+ int high = 65536;
+ int i;
+ for (i=0; i < n; ++i) {
+ if (x[i] > low && x[i] < x[n]) { *plow = i; low = x[i]; }
+ if (x[i] < high && x[i] > x[n]) { *phigh = i; high = x[i]; }
+ }
+}
+
+// this has been repurposed so y is now the original index instead of y
+typedef struct
+{
+ uint16 x,y;
+} Point;
+
+int point_compare(const void *p, const void *q)
+{
+ Point *a = (Point *) p;
+ Point *b = (Point *) q;
+ return a->x < b->x ? -1 : a->x > b->x;
+}
+
+//
+/////////////////////// END LEAF SETUP FUNCTIONS //////////////////////////
+
+
+#if defined(STB_VORBIS_NO_STDIO)
+ #define USE_MEMORY(z) TRUE
+#else
+ #define USE_MEMORY(z) ((z)->stream)
+#endif
+
+static uint8 get8(vorb *z)
+{
+ if (USE_MEMORY(z)) {
+ if (z->stream >= z->stream_end) { z->eof = TRUE; return 0; }
+ return *z->stream++;
+ }
+
+ #ifndef STB_VORBIS_NO_STDIO
+ {
+ int c = fgetc(z->f);
+ if (c == EOF) { z->eof = TRUE; return 0; }
+ return c;
+ }
+ #endif
+}
+
+static uint32 get32(vorb *f)
+{
+ uint32 x;
+ x = get8(f);
+ x += get8(f) << 8;
+ x += get8(f) << 16;
+ x += get8(f) << 24;
+ return x;
+}
+
+static int getn(vorb *z, uint8 *data, int n)
+{
+ if (USE_MEMORY(z)) {
+ if (z->stream+n > z->stream_end) { z->eof = 1; return 0; }
+ memcpy(data, z->stream, n);
+ z->stream += n;
+ return 1;
+ }
+
+ #ifndef STB_VORBIS_NO_STDIO
+ if (fread(data, n, 1, z->f) == 1)
+ return 1;
+ else {
+ z->eof = 1;
+ return 0;
+ }
+ #endif
+}
+
+static void skip(vorb *z, int n)
+{
+ if (USE_MEMORY(z)) {
+ z->stream += n;
+ if (z->stream >= z->stream_end) z->eof = 1;
+ return;
+ }
+ #ifndef STB_VORBIS_NO_STDIO
+ {
+ long x = ftell(z->f);
+ fseek(z->f, x+n, SEEK_SET);
+ }
+ #endif
+}
+
+static int set_file_offset(stb_vorbis *f, unsigned int loc)
+{
+ #ifndef STB_VORBIS_NO_PUSHDATA_API
+ if (f->push_mode) return 0;
+ #endif
+ f->eof = 0;
+ if (USE_MEMORY(f)) {
+ if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) {
+ f->stream = f->stream_end;
+ f->eof = 1;
+ return 0;
+ } else {
+ f->stream = f->stream_start + loc;
+ return 1;
+ }
+ }
+ #ifndef STB_VORBIS_NO_STDIO
+ if (loc + f->f_start < loc || loc >= 0x80000000) {
+ loc = 0x7fffffff;
+ f->eof = 1;
+ } else {
+ loc += f->f_start;
+ }
+ if (!fseek(f->f, loc, SEEK_SET))
+ return 1;
+ f->eof = 1;
+ fseek(f->f, f->f_start, SEEK_END);
+ return 0;
+ #endif
+}
+
+
+static uint8 ogg_page_header[4] = { 0x4f, 0x67, 0x67, 0x53 };
+
+static int capture_pattern(vorb *f)
+{
+ if (0x4f != get8(f)) return FALSE;
+ if (0x67 != get8(f)) return FALSE;
+ if (0x67 != get8(f)) return FALSE;
+ if (0x53 != get8(f)) return FALSE;
+ return TRUE;
+}
+
+#define PAGEFLAG_continued_packet 1
+#define PAGEFLAG_first_page 2
+#define PAGEFLAG_last_page 4
+
+static int start_page_no_capturepattern(vorb *f)
+{
+ uint32 loc0,loc1,n,i;
+ // stream structure version
+ if (0 != get8(f)) return error(f, VORBIS_invalid_stream_structure_version);
+ // header flag
+ f->page_flag = get8(f);
+ // absolute granule position
+ loc0 = get32(f);
+ loc1 = get32(f);
+ // @TODO: validate loc0,loc1 as valid positions?
+ // stream serial number -- vorbis doesn't interleave, so discard
+ get32(f);
+ //if (f->serial != get32(f)) return error(f, VORBIS_incorrect_stream_serial_number);
+ // page sequence number
+ n = get32(f);
+ f->last_page = n;
+ // CRC32
+ get32(f);
+ // page_segments
+ f->segment_count = get8(f);
+ if (!getn(f, f->segments, f->segment_count))
+ return error(f, VORBIS_unexpected_eof);
+ // assume we _don't_ know any the sample position of any segments
+ f->end_seg_with_known_loc = -2;
+ if (loc0 != ~0 || loc1 != ~0) {
+ // determine which packet is the last one that will complete
+ for (i=f->segment_count-1; i >= 0; --i)
+ if (f->segments[i] < 255)
+ break;
+ // 'i' is now the index of the _last_ segment of a packet that ends
+ if (i >= 0) {
+ f->end_seg_with_known_loc = i;
+ f->known_loc_for_packet = loc0;
+ }
+ }
+ if (f->first_decode) {
+ int i,len;
+ ProbedPage p;
+ len = 0;
+ for (i=0; i < f->segment_count; ++i)
+ len += f->segments[i];
+ len += 27 + f->segment_count;
+ p.page_start = f->first_audio_page_offset;
+ p.page_end = p.page_start + len;
+ p.after_previous_page_start = p.page_start;
+ p.first_decoded_sample = 0;
+ p.last_decoded_sample = loc0;
+ f->p_first = p;
+ }
+ f->next_seg = 0;
+ return TRUE;
+}
+
+static int start_page(vorb *f)
+{
+ if (!capture_pattern(f)) return error(f, VORBIS_missing_capture_pattern);
+ return start_page_no_capturepattern(f);
+}
+
+static int start_packet(vorb *f)
+{
+ while (f->next_seg == -1) {
+ if (!start_page(f)) return FALSE;
+ if (f->page_flag & PAGEFLAG_continued_packet)
+ return error(f, VORBIS_continued_packet_flag_invalid);
+ }
+ f->last_seg = FALSE;
+ f->valid_bits = 0;
+ f->packet_bytes = 0;
+ f->bytes_in_seg = 0;
+ // f->next_seg is now valid
+ return TRUE;
+}
+
+static int maybe_start_packet(vorb *f)
+{
+ if (f->next_seg == -1) {
+ int x = get8(f);
+ if (f->eof) return FALSE; // EOF at page boundary is not an error!
+ if (0x4f != x ) return error(f, VORBIS_missing_capture_pattern);
+ if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern);
+ if (0x67 != get8(f)) return error(f, VORBIS_missing_capture_pattern);
+ if (0x53 != get8(f)) return error(f, VORBIS_missing_capture_pattern);
+ if (!start_page_no_capturepattern(f)) return FALSE;
+ if (f->page_flag & PAGEFLAG_continued_packet) {
+ // set up enough state that we can read this packet if we want,
+ // e.g. during recovery
+ f->last_seg = FALSE;
+ f->bytes_in_seg = 0;
+ return error(f, VORBIS_continued_packet_flag_invalid);
+ }
+ }
+ return start_packet(f);
+}
+
+static int next_segment(vorb *f)
+{
+ int len;
+ if (f->last_seg) return 0;
+ if (f->next_seg == -1) {
+ f->last_seg_which = f->segment_count-1; // in case start_page fails
+ if (!start_page(f)) { f->last_seg = 1; return 0; }
+ if (!(f->page_flag & PAGEFLAG_continued_packet)) return error(f, VORBIS_continued_packet_flag_invalid);
+ }
+ len = f->segments[f->next_seg++];
+ if (len < 255) {
+ f->last_seg = TRUE;
+ f->last_seg_which = f->next_seg-1;
+ }
+ if (f->next_seg >= f->segment_count)
+ f->next_seg = -1;
+ assert(f->bytes_in_seg == 0);
+ f->bytes_in_seg = len;
+ return len;
+}
+
+#define EOP (-1)
+#define INVALID_BITS (-1)
+
+static int get8_packet_raw(vorb *f)
+{
+ if (!f->bytes_in_seg)
+ if (f->last_seg) return EOP;
+ else if (!next_segment(f)) return EOP;
+ assert(f->bytes_in_seg > 0);
+ --f->bytes_in_seg;
+ ++f->packet_bytes;
+ return get8(f);
+}
+
+static int get8_packet(vorb *f)
+{
+ int x = get8_packet_raw(f);
+ f->valid_bits = 0;
+ return x;
+}
+
+static void flush_packet(vorb *f)
+{
+ while (get8_packet_raw(f) != EOP);
+}
+
+// @OPTIMIZE: this is the secondary bit decoder, so it's probably not as important
+// as the huffman decoder?
+static uint32 get_bits(vorb *f, int n)
+{
+ uint32 z;
+
+ if (f->valid_bits < 0) return 0;
+ if (f->valid_bits < n) {
+ if (n > 24) {
+ // the accumulator technique below would not work correctly in this case
+ z = get_bits(f, 24);
+ z += get_bits(f, n-24) << 24;
+ return z;
+ }
+ if (f->valid_bits == 0) f->acc = 0;
+ while (f->valid_bits < n) {
+ int z = get8_packet_raw(f);
+ if (z == EOP) {
+ f->valid_bits = INVALID_BITS;
+ return 0;
+ }
+ f->acc += z << f->valid_bits;
+ f->valid_bits += 8;
+ }
+ }
+ if (f->valid_bits < 0) return 0;
+ z = f->acc & ((1 << n)-1);
+ f->acc >>= n;
+ f->valid_bits -= n;
+ return z;
+}
+
+static int32 get_bits_signed(vorb *f, int n)
+{
+ uint32 z = get_bits(f, n);
+ if (z & (1 << (n-1)))
+ z += ~((1 << n) - 1);
+ return (int32) z;
+}
+
+// @OPTIMIZE: primary accumulator for huffman
+// expand the buffer to as many bits as possible without reading off end of packet
+// it might be nice to allow f->valid_bits and f->acc to be stored in registers,
+// e.g. cache them locally and decode locally
+static __forceinline void prep_huffman(vorb *f)
+{
+ if (f->valid_bits <= 24) {
+ if (f->valid_bits == 0) f->acc = 0;
+ do {
+ int z;
+ if (f->last_seg && !f->bytes_in_seg) return;
+ z = get8_packet_raw(f);
+ if (z == EOP) return;
+ f->acc += z << f->valid_bits;
+ f->valid_bits += 8;
+ } while (f->valid_bits <= 24);
+ }
+}
+
+enum
+{
+ VORBIS_packet_id = 1,
+ VORBIS_packet_comment = 3,
+ VORBIS_packet_setup = 5,
+};
+
+static int codebook_decode_scalar_raw(vorb *f, Codebook *c)
+{
+ int i;
+ prep_huffman(f);
+
+ assert(c->sorted_codewords || c->codewords);
+ // cases to use binary search: sorted_codewords && !c->codewords
+ // sorted_codewords && c->entries > 8
+ if (c->entries > 8 ? c->sorted_codewords!=NULL : !c->codewords) {
+ // binary search
+ uint32 code = bit_reverse(f->acc);
+ int x=0, n=c->sorted_entries, len;
+
+ while (n > 1) {
+ // invariant: sc[x] <= code < sc[x+n]
+ int m = x + (n >> 1);
+ if (c->sorted_codewords[m] <= code) {
+ x = m;
+ n -= (n>>1);
+ } else {
+ n >>= 1;
+ }
+ }
+ // x is now the sorted index
+ if (!c->sparse) x = c->sorted_values[x];
+ // x is now sorted index if sparse, or symbol otherwise
+ len = c->codeword_lengths[x];
+ if (f->valid_bits >= len) {
+ f->acc >>= len;
+ f->valid_bits -= len;
+ return x;
+ }
+
+ f->valid_bits = 0;
+ return -1;
+ }
+
+ // if small, linear search
+ assert(!c->sparse);
+ for (i=0; i < c->entries; ++i) {
+ if (c->codeword_lengths[i] == NO_CODE) continue;
+ if (c->codewords[i] == (f->acc & ((1 << c->codeword_lengths[i])-1))) {
+ if (f->valid_bits >= c->codeword_lengths[i]) {
+ f->acc >>= c->codeword_lengths[i];
+ f->valid_bits -= c->codeword_lengths[i];
+ return i;
+ }
+ f->valid_bits = 0;
+ return -1;
+ }
+ }
+
+ error(f, VORBIS_invalid_stream);
+ f->valid_bits = 0;
+ return -1;
+}
+
+static int codebook_decode_scalar(vorb *f, Codebook *c)
+{
+ int i;
+ if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH)
+ prep_huffman(f);
+ // fast huffman table lookup
+ i = f->acc & FAST_HUFFMAN_TABLE_MASK;
+ i = c->fast_huffman[i];
+ if (i >= 0) {
+ f->acc >>= c->codeword_lengths[i];
+ f->valid_bits -= c->codeword_lengths[i];
+ if (f->valid_bits < 0) { f->valid_bits = 0; return -1; }
+ return i;
+ }
+ return codebook_decode_scalar_raw(f,c);
+}
+
+#ifndef STB_VORBIS_NO_INLINE_DECODE
+
+#define DECODE_RAW(var, f,c) \
+ if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) \
+ prep_huffman(f); \
+ var = f->acc & FAST_HUFFMAN_TABLE_MASK; \
+ var = c->fast_huffman[var]; \
+ if (var >= 0) { \
+ int n = c->codeword_lengths[var]; \
+ f->acc >>= n; \
+ f->valid_bits -= n; \
+ if (f->valid_bits < 0) { f->valid_bits = 0; var = -1; } \
+ } else { \
+ var = codebook_decode_scalar_raw(f,c); \
+ }
+
+#else
+
+#define DECODE_RAW(var,f,c) var = codebook_decode_scalar(f,c);
+
+#endif
+
+#define DECODE(var,f,c) \
+ DECODE_RAW(var,f,c) \
+ if (c->sparse) var = c->sorted_values[var];
+
+#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ #define DECODE_VQ(var,f,c) DECODE_RAW(var,f,c)
+#else
+ #define DECODE_VQ(var,f,c) DECODE(var,f,c)
+#endif
+
+
+
+
+
+
+// CODEBOOK_ELEMENT_FAST is an optimization for the CODEBOOK_FLOATS case
+// where we avoid one addition
+#ifndef STB_VORBIS_CODEBOOK_FLOATS
+ #define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off] * c->delta_value + c->minimum_value)
+ #define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off] * c->delta_value)
+ #define CODEBOOK_ELEMENT_BASE(c) (c->minimum_value)
+#else
+ #define CODEBOOK_ELEMENT(c,off) (c->multiplicands[off])
+ #define CODEBOOK_ELEMENT_FAST(c,off) (c->multiplicands[off])
+ #define CODEBOOK_ELEMENT_BASE(c) (0)
+#endif
+
+static int codebook_decode_start(vorb *f, Codebook *c, int len)
+{
+ int z = -1;
+
+ // type 0 is only legal in a scalar context
+ if (c->lookup_type == 0)
+ error(f, VORBIS_invalid_stream);
+ else {
+ DECODE_VQ(z,f,c);
+ if (c->sparse) assert(z < c->sorted_entries);
+ if (z < 0) { // check for EOP
+ if (!f->bytes_in_seg)
+ if (f->last_seg)
+ return z;
+ error(f, VORBIS_invalid_stream);
+ }
+ }
+ return z;
+}
+
+static int codebook_decode(vorb *f, Codebook *c, float *output, int len)
+{
+ int i,z = codebook_decode_start(f,c,len);
+ if (z < 0) return FALSE;
+ if (len > c->dimensions) len = c->dimensions;
+
+#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ if (c->lookup_type == 1) {
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ int div = 1;
+ for (i=0; i < len; ++i) {
+ int off = (z / div) % c->lookup_values;
+ float val = CODEBOOK_ELEMENT_FAST(c,off) + last;
+ output[i] += val;
+ if (c->sequence_p) last = val + c->minimum_value;
+ div *= c->lookup_values;
+ }
+ return TRUE;
+ }
+#endif
+
+ z *= c->dimensions;
+ if (c->sequence_p) {
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ for (i=0; i < len; ++i) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ output[i] += val;
+ last = val + c->minimum_value;
+ }
+ } else {
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ for (i=0; i < len; ++i) {
+ output[i] += CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ }
+ }
+
+ return TRUE;
+}
+
+static int codebook_decode_step(vorb *f, Codebook *c, float *output, int len, int step)
+{
+ int i,z = codebook_decode_start(f,c,len);
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ if (z < 0) return FALSE;
+ if (len > c->dimensions) len = c->dimensions;
+
+#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ if (c->lookup_type == 1) {
+ int div = 1;
+ for (i=0; i < len; ++i) {
+ int off = (z / div) % c->lookup_values;
+ float val = CODEBOOK_ELEMENT_FAST(c,off) + last;
+ output[i*step] += val;
+ if (c->sequence_p) last = val;
+ div *= c->lookup_values;
+ }
+ return TRUE;
+ }
+#endif
+
+ z *= c->dimensions;
+ for (i=0; i < len; ++i) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ output[i*step] += val;
+ if (c->sequence_p) last = val;
+ }
+
+ return TRUE;
+}
+
+static int codebook_decode_deinterleave_repeat(vorb *f, Codebook *c, float **outputs, int ch, int *c_inter_p, int *p_inter_p, int len, int total_decode)
+{
+ int c_inter = *c_inter_p;
+ int p_inter = *p_inter_p;
+ int i,z, effective = c->dimensions;
+
+ // type 0 is only legal in a scalar context
+ if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream);
+
+ while (total_decode > 0) {
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ DECODE_VQ(z,f,c);
+ #ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ assert(!c->sparse || z < c->sorted_entries);
+ #endif
+ if (z < 0) {
+ if (!f->bytes_in_seg)
+ if (f->last_seg) return FALSE;
+ return error(f, VORBIS_invalid_stream);
+ }
+
+ // if this will take us off the end of the buffers, stop short!
+ // we check by computing the length of the virtual interleaved
+ // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter),
+ // and the length we'll be using (effective)
+ if (c_inter + p_inter*ch + effective > len * ch) {
+ effective = len*ch - (p_inter*ch - c_inter);
+ }
+
+ #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ if (c->lookup_type == 1) {
+ int div = 1;
+ for (i=0; i < effective; ++i) {
+ int off = (z / div) % c->lookup_values;
+ float val = CODEBOOK_ELEMENT_FAST(c,off) + last;
+ outputs[c_inter][p_inter] += val;
+ if (++c_inter == ch) { c_inter = 0; ++p_inter; }
+ if (c->sequence_p) last = val;
+ div *= c->lookup_values;
+ }
+ } else
+ #endif
+ {
+ z *= c->dimensions;
+ if (c->sequence_p) {
+ for (i=0; i < effective; ++i) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ outputs[c_inter][p_inter] += val;
+ if (++c_inter == ch) { c_inter = 0; ++p_inter; }
+ last = val;
+ }
+ } else {
+ for (i=0; i < effective; ++i) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ outputs[c_inter][p_inter] += val;
+ if (++c_inter == ch) { c_inter = 0; ++p_inter; }
+ }
+ }
+ }
+
+ total_decode -= effective;
+ }
+ *c_inter_p = c_inter;
+ *p_inter_p = p_inter;
+ return TRUE;
+}
+
+#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK
+static int codebook_decode_deinterleave_repeat_2(vorb *f, Codebook *c, float **outputs, int *c_inter_p, int *p_inter_p, int len, int total_decode)
+{
+ int c_inter = *c_inter_p;
+ int p_inter = *p_inter_p;
+ int i,z, effective = c->dimensions;
+
+ // type 0 is only legal in a scalar context
+ if (c->lookup_type == 0) return error(f, VORBIS_invalid_stream);
+
+ while (total_decode > 0) {
+ float last = CODEBOOK_ELEMENT_BASE(c);
+ DECODE_VQ(z,f,c);
+
+ if (z < 0) {
+ if (!f->bytes_in_seg)
+ if (f->last_seg) return FALSE;
+ return error(f, VORBIS_invalid_stream);
+ }
+
+ // if this will take us off the end of the buffers, stop short!
+ // we check by computing the length of the virtual interleaved
+ // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter),
+ // and the length we'll be using (effective)
+ if (c_inter + p_inter*2 + effective > len * 2) {
+ effective = len*2 - (p_inter*2 - c_inter);
+ }
+
+ {
+ z *= c->dimensions;
+ stb_prof(11);
+ if (c->sequence_p) {
+ // haven't optimized this case because I don't have any examples
+ for (i=0; i < effective; ++i) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ outputs[c_inter][p_inter] += val;
+ if (++c_inter == 2) { c_inter = 0; ++p_inter; }
+ last = val;
+ }
+ } else {
+ i=0;
+ if (c_inter == 1) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ outputs[c_inter][p_inter] += val;
+ c_inter = 0; ++p_inter;
+ ++i;
+ }
+ {
+ float *z0 = outputs[0];
+ float *z1 = outputs[1];
+ for (; i+1 < effective;) {
+ z0[p_inter] += CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ z1[p_inter] += CODEBOOK_ELEMENT_FAST(c,z+i+1) + last;
+ ++p_inter;
+ i += 2;
+ }
+ }
+ if (i < effective) {
+ float val = CODEBOOK_ELEMENT_FAST(c,z+i) + last;
+ outputs[c_inter][p_inter] += val;
+ if (++c_inter == 2) { c_inter = 0; ++p_inter; }
+ }
+ }
+ }
+
+ total_decode -= effective;
+ }
+ *c_inter_p = c_inter;
+ *p_inter_p = p_inter;
+ return TRUE;
+}
+#endif
+
+static int predict_point(int x, int x0, int x1, int y0, int y1)
+{
+ int dy = y1 - y0;
+ int adx = x1 - x0;
+ // @OPTIMIZE: force int division to round in the right direction... is this necessary on x86?
+ int err = abs(dy) * (x - x0);
+ int off = err / adx;
+ return dy < 0 ? y0 - off : y0 + off;
+}
+
+// the following table is block-copied from the specification
+static float inverse_db_table[256] =
+{
+ 1.0649863e-07f, 1.1341951e-07f, 1.2079015e-07f, 1.2863978e-07f,
+ 1.3699951e-07f, 1.4590251e-07f, 1.5538408e-07f, 1.6548181e-07f,
+ 1.7623575e-07f, 1.8768855e-07f, 1.9988561e-07f, 2.1287530e-07f,
+ 2.2670913e-07f, 2.4144197e-07f, 2.5713223e-07f, 2.7384213e-07f,
+ 2.9163793e-07f, 3.1059021e-07f, 3.3077411e-07f, 3.5226968e-07f,
+ 3.7516214e-07f, 3.9954229e-07f, 4.2550680e-07f, 4.5315863e-07f,
+ 4.8260743e-07f, 5.1396998e-07f, 5.4737065e-07f, 5.8294187e-07f,
+ 6.2082472e-07f, 6.6116941e-07f, 7.0413592e-07f, 7.4989464e-07f,
+ 7.9862701e-07f, 8.5052630e-07f, 9.0579828e-07f, 9.6466216e-07f,
+ 1.0273513e-06f, 1.0941144e-06f, 1.1652161e-06f, 1.2409384e-06f,
+ 1.3215816e-06f, 1.4074654e-06f, 1.4989305e-06f, 1.5963394e-06f,
+ 1.7000785e-06f, 1.8105592e-06f, 1.9282195e-06f, 2.0535261e-06f,
+ 2.1869758e-06f, 2.3290978e-06f, 2.4804557e-06f, 2.6416497e-06f,
+ 2.8133190e-06f, 2.9961443e-06f, 3.1908506e-06f, 3.3982101e-06f,
+ 3.6190449e-06f, 3.8542308e-06f, 4.1047004e-06f, 4.3714470e-06f,
+ 4.6555282e-06f, 4.9580707e-06f, 5.2802740e-06f, 5.6234160e-06f,
+ 5.9888572e-06f, 6.3780469e-06f, 6.7925283e-06f, 7.2339451e-06f,
+ 7.7040476e-06f, 8.2047000e-06f, 8.7378876e-06f, 9.3057248e-06f,
+ 9.9104632e-06f, 1.0554501e-05f, 1.1240392e-05f, 1.1970856e-05f,
+ 1.2748789e-05f, 1.3577278e-05f, 1.4459606e-05f, 1.5399272e-05f,
+ 1.6400004e-05f, 1.7465768e-05f, 1.8600792e-05f, 1.9809576e-05f,
+ 2.1096914e-05f, 2.2467911e-05f, 2.3928002e-05f, 2.5482978e-05f,
+ 2.7139006e-05f, 2.8902651e-05f, 3.0780908e-05f, 3.2781225e-05f,
+ 3.4911534e-05f, 3.7180282e-05f, 3.9596466e-05f, 4.2169667e-05f,
+ 4.4910090e-05f, 4.7828601e-05f, 5.0936773e-05f, 5.4246931e-05f,
+ 5.7772202e-05f, 6.1526565e-05f, 6.5524908e-05f, 6.9783085e-05f,
+ 7.4317983e-05f, 7.9147585e-05f, 8.4291040e-05f, 8.9768747e-05f,
+ 9.5602426e-05f, 0.00010181521f, 0.00010843174f, 0.00011547824f,
+ 0.00012298267f, 0.00013097477f, 0.00013948625f, 0.00014855085f,
+ 0.00015820453f, 0.00016848555f, 0.00017943469f, 0.00019109536f,
+ 0.00020351382f, 0.00021673929f, 0.00023082423f, 0.00024582449f,
+ 0.00026179955f, 0.00027881276f, 0.00029693158f, 0.00031622787f,
+ 0.00033677814f, 0.00035866388f, 0.00038197188f, 0.00040679456f,
+ 0.00043323036f, 0.00046138411f, 0.00049136745f, 0.00052329927f,
+ 0.00055730621f, 0.00059352311f, 0.00063209358f, 0.00067317058f,
+ 0.00071691700f, 0.00076350630f, 0.00081312324f, 0.00086596457f,
+ 0.00092223983f, 0.00098217216f, 0.0010459992f, 0.0011139742f,
+ 0.0011863665f, 0.0012634633f, 0.0013455702f, 0.0014330129f,
+ 0.0015261382f, 0.0016253153f, 0.0017309374f, 0.0018434235f,
+ 0.0019632195f, 0.0020908006f, 0.0022266726f, 0.0023713743f,
+ 0.0025254795f, 0.0026895994f, 0.0028643847f, 0.0030505286f,
+ 0.0032487691f, 0.0034598925f, 0.0036847358f, 0.0039241906f,
+ 0.0041792066f, 0.0044507950f, 0.0047400328f, 0.0050480668f,
+ 0.0053761186f, 0.0057254891f, 0.0060975636f, 0.0064938176f,
+ 0.0069158225f, 0.0073652516f, 0.0078438871f, 0.0083536271f,
+ 0.0088964928f, 0.009474637f, 0.010090352f, 0.010746080f,
+ 0.011444421f, 0.012188144f, 0.012980198f, 0.013823725f,
+ 0.014722068f, 0.015678791f, 0.016697687f, 0.017782797f,
+ 0.018938423f, 0.020169149f, 0.021479854f, 0.022875735f,
+ 0.024362330f, 0.025945531f, 0.027631618f, 0.029427276f,
+ 0.031339626f, 0.033376252f, 0.035545228f, 0.037855157f,
+ 0.040315199f, 0.042935108f, 0.045725273f, 0.048696758f,
+ 0.051861348f, 0.055231591f, 0.058820850f, 0.062643361f,
+ 0.066714279f, 0.071049749f, 0.075666962f, 0.080584227f,
+ 0.085821044f, 0.091398179f, 0.097337747f, 0.10366330f,
+ 0.11039993f, 0.11757434f, 0.12521498f, 0.13335215f,
+ 0.14201813f, 0.15124727f, 0.16107617f, 0.17154380f,
+ 0.18269168f, 0.19456402f, 0.20720788f, 0.22067342f,
+ 0.23501402f, 0.25028656f, 0.26655159f, 0.28387361f,
+ 0.30232132f, 0.32196786f, 0.34289114f, 0.36517414f,
+ 0.38890521f, 0.41417847f, 0.44109412f, 0.46975890f,
+ 0.50028648f, 0.53279791f, 0.56742212f, 0.60429640f,
+ 0.64356699f, 0.68538959f, 0.72993007f, 0.77736504f,
+ 0.82788260f, 0.88168307f, 0.9389798f, 1.0f
+};
+
+
+// @OPTIMIZE: if you want to replace this bresenham line-drawing routine,
+// note that you must produce bit-identical output to decode correctly;
+// this specific sequence of operations is specified in the spec (it's
+// drawing integer-quantized frequency-space lines that the encoder
+// expects to be exactly the same)
+// ... also, isn't the whole point of Bresenham's algorithm to NOT
+// have to divide in the setup? sigh.
+#ifndef STB_VORBIS_NO_DEFER_FLOOR
+#define LINE_OP(a,b) a *= b
+#else
+#define LINE_OP(a,b) a = b
+#endif
+
+#ifdef STB_VORBIS_DIVIDE_TABLE
+#define DIVTAB_NUMER 32
+#define DIVTAB_DENOM 64
+int8 integer_divide_table[DIVTAB_NUMER][DIVTAB_DENOM]; // 2KB
+#endif
+
+static __forceinline void draw_line(float *output, int x0, int y0, int x1, int y1, int n)
+{
+ int dy = y1 - y0;
+ int adx = x1 - x0;
+ int ady = abs(dy);
+ int base;
+ int x=x0,y=y0;
+ int err = 0;
+ int sy;
+
+#ifdef STB_VORBIS_DIVIDE_TABLE
+ if (adx < DIVTAB_DENOM && ady < DIVTAB_NUMER) {
+ if (dy < 0) {
+ base = -integer_divide_table[ady][adx];
+ sy = base-1;
+ } else {
+ base = integer_divide_table[ady][adx];
+ sy = base+1;
+ }
+ } else {
+ base = dy / adx;
+ if (dy < 0)
+ sy = base - 1;
+ else
+ sy = base+1;
+ }
+#else
+ base = dy / adx;
+ if (dy < 0)
+ sy = base - 1;
+ else
+ sy = base+1;
+#endif
+ ady -= abs(base) * adx;
+ if (x1 > n) x1 = n;
+ LINE_OP(output[x], inverse_db_table[y]);
+ for (++x; x < x1; ++x) {
+ err += ady;
+ if (err >= adx) {
+ err -= adx;
+ y += sy;
+ } else
+ y += base;
+ LINE_OP(output[x], inverse_db_table[y]);
+ }
+}
+
+static int residue_decode(vorb *f, Codebook *book, float *target, int offset, int n, int rtype)
+{
+ int k;
+ if (rtype == 0) {
+ int step = n / book->dimensions;
+ for (k=0; k < step; ++k)
+ if (!codebook_decode_step(f, book, target+offset+k, n-offset-k, step))
+ return FALSE;
+ } else {
+ for (k=0; k < n; ) {
+ if (!codebook_decode(f, book, target+offset, n-k))
+ return FALSE;
+ k += book->dimensions;
+ offset += book->dimensions;
+ }
+ }
+ return TRUE;
+}
+
+static void decode_residue(vorb *f, float *residue_buffers[], int ch, int n, int rn, uint8 *do_not_decode)
+{
+ int i,j,pass;
+ Residue *r = f->residue_config + rn;
+ int rtype = f->residue_types[rn];
+ int c = r->classbook;
+ int classwords = f->codebooks[c].dimensions;
+ int n_read = r->end - r->begin;
+ int part_read = n_read / r->part_size;
+ int temp_alloc_point = temp_alloc_save(f);
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ uint8 ***part_classdata = (uint8 ***) temp_block_array(f,f->channels, part_read * sizeof(**part_classdata));
+ #else
+ int **classifications = (int **) temp_block_array(f,f->channels, part_read * sizeof(**classifications));
+ #endif
+
+ stb_prof(2);
+ for (i=0; i < ch; ++i)
+ if (!do_not_decode[i])
+ memset(residue_buffers[i], 0, sizeof(float) * n);
+
+ if (rtype == 2 && ch != 1) {
+ int len = ch * n;
+ for (j=0; j < ch; ++j)
+ if (!do_not_decode[j])
+ break;
+ if (j == ch)
+ goto done;
+
+ stb_prof(3);
+ for (pass=0; pass < 8; ++pass) {
+ int pcount = 0, class_set = 0;
+ if (ch == 2) {
+ stb_prof(13);
+ while (pcount < part_read) {
+ int z = r->begin + pcount*r->part_size;
+ int c_inter = (z & 1), p_inter = z>>1;
+ if (pass == 0) {
+ Codebook *c = f->codebooks+r->classbook;
+ int q;
+ DECODE(q,f,c);
+ if (q == EOP) goto done;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ part_classdata[0][class_set] = r->classdata[q];
+ #else
+ for (i=classwords-1; i >= 0; --i) {
+ classifications[0][i+pcount] = q % r->classifications;
+ q /= r->classifications;
+ }
+ #endif
+ }
+ stb_prof(5);
+ for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) {
+ int z = r->begin + pcount*r->part_size;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ int c = part_classdata[0][class_set][i];
+ #else
+ int c = classifications[0][pcount];
+ #endif
+ int b = r->residue_books[c][pass];
+ if (b >= 0) {
+ Codebook *book = f->codebooks + b;
+ stb_prof(20); // accounts for X time
+ #ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size))
+ goto done;
+ #else
+ // saves 1%
+ if (!codebook_decode_deinterleave_repeat_2(f, book, residue_buffers, &c_inter, &p_inter, n, r->part_size))
+ goto done;
+ #endif
+ stb_prof(7);
+ } else {
+ z += r->part_size;
+ c_inter = z & 1;
+ p_inter = z >> 1;
+ }
+ }
+ stb_prof(8);
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ ++class_set;
+ #endif
+ }
+ } else if (ch == 1) {
+ while (pcount < part_read) {
+ int z = r->begin + pcount*r->part_size;
+ int c_inter = 0, p_inter = z;
+ if (pass == 0) {
+ Codebook *c = f->codebooks+r->classbook;
+ int q;
+ DECODE(q,f,c);
+ if (q == EOP) goto done;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ part_classdata[0][class_set] = r->classdata[q];
+ #else
+ for (i=classwords-1; i >= 0; --i) {
+ classifications[0][i+pcount] = q % r->classifications;
+ q /= r->classifications;
+ }
+ #endif
+ }
+ for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) {
+ int z = r->begin + pcount*r->part_size;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ int c = part_classdata[0][class_set][i];
+ #else
+ int c = classifications[0][pcount];
+ #endif
+ int b = r->residue_books[c][pass];
+ if (b >= 0) {
+ Codebook *book = f->codebooks + b;
+ stb_prof(22);
+ if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size))
+ goto done;
+ stb_prof(3);
+ } else {
+ z += r->part_size;
+ c_inter = 0;
+ p_inter = z;
+ }
+ }
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ ++class_set;
+ #endif
+ }
+ } else {
+ while (pcount < part_read) {
+ int z = r->begin + pcount*r->part_size;
+ int c_inter = z % ch, p_inter = z/ch;
+ if (pass == 0) {
+ Codebook *c = f->codebooks+r->classbook;
+ int q;
+ DECODE(q,f,c);
+ if (q == EOP) goto done;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ part_classdata[0][class_set] = r->classdata[q];
+ #else
+ for (i=classwords-1; i >= 0; --i) {
+ classifications[0][i+pcount] = q % r->classifications;
+ q /= r->classifications;
+ }
+ #endif
+ }
+ for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) {
+ int z = r->begin + pcount*r->part_size;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ int c = part_classdata[0][class_set][i];
+ #else
+ int c = classifications[0][pcount];
+ #endif
+ int b = r->residue_books[c][pass];
+ if (b >= 0) {
+ Codebook *book = f->codebooks + b;
+ stb_prof(22);
+ if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size))
+ goto done;
+ stb_prof(3);
+ } else {
+ z += r->part_size;
+ c_inter = z % ch;
+ p_inter = z / ch;
+ }
+ }
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ ++class_set;
+ #endif
+ }
+ }
+ }
+ goto done;
+ }
+ stb_prof(9);
+
+ for (pass=0; pass < 8; ++pass) {
+ int pcount = 0, class_set=0;
+ while (pcount < part_read) {
+ if (pass == 0) {
+ for (j=0; j < ch; ++j) {
+ if (!do_not_decode[j]) {
+ Codebook *c = f->codebooks+r->classbook;
+ int temp;
+ DECODE(temp,f,c);
+ if (temp == EOP) goto done;
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ part_classdata[j][class_set] = r->classdata[temp];
+ #else
+ for (i=classwords-1; i >= 0; --i) {
+ classifications[j][i+pcount] = temp % r->classifications;
+ temp /= r->classifications;
+ }
+ #endif
+ }
+ }
+ }
+ for (i=0; i < classwords && pcount < part_read; ++i, ++pcount) {
+ for (j=0; j < ch; ++j) {
+ if (!do_not_decode[j]) {
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ int c = part_classdata[j][class_set][i];
+ #else
+ int c = classifications[j][pcount];
+ #endif
+ int b = r->residue_books[c][pass];
+ if (b >= 0) {
+ float *target = residue_buffers[j];
+ int offset = r->begin + pcount * r->part_size;
+ int n = r->part_size;
+ Codebook *book = f->codebooks + b;
+ if (!residue_decode(f, book, target, offset, n, rtype))
+ goto done;
+ }
+ }
+ }
+ }
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ ++class_set;
+ #endif
+ }
+ }
+ done:
+ stb_prof(0);
+ temp_alloc_restore(f,temp_alloc_point);
+}
+
+
+#if 0
+// slow way for debugging
+void inverse_mdct_slow(float *buffer, int n)
+{
+ int i,j;
+ int n2 = n >> 1;
+ float *x = (float *) malloc(sizeof(*x) * n2);
+ memcpy(x, buffer, sizeof(*x) * n2);
+ for (i=0; i < n; ++i) {
+ float acc = 0;
+ for (j=0; j < n2; ++j)
+ // formula from paper:
+ //acc += n/4.0f * x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1));
+ // formula from wikipedia
+ //acc += 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5));
+ // these are equivalent, except the formula from the paper inverts the multiplier!
+ // however, what actually works is NO MULTIPLIER!?!
+ //acc += 64 * 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5));
+ acc += x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1));
+ buffer[i] = acc;
+ }
+ free(x);
+}
+#elif 0
+// same as above, but just barely able to run in real time on modern machines
+void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype)
+{
+ float mcos[16384];
+ int i,j;
+ int n2 = n >> 1, nmask = (n << 2) -1;
+ float *x = (float *) malloc(sizeof(*x) * n2);
+ memcpy(x, buffer, sizeof(*x) * n2);
+ for (i=0; i < 4*n; ++i)
+ mcos[i] = (float) cos(M_PI / 2 * i / n);
+
+ for (i=0; i < n; ++i) {
+ float acc = 0;
+ for (j=0; j < n2; ++j)
+ acc += x[j] * mcos[(2 * i + 1 + n2)*(2*j+1) & nmask];
+ buffer[i] = acc;
+ }
+ free(x);
+}
+#else
+// transform to use a slow dct-iv; this is STILL basically trivial,
+// but only requires half as many ops
+void dct_iv_slow(float *buffer, int n)
+{
+ float mcos[16384];
+ float x[2048];
+ int i,j;
+ int n2 = n >> 1, nmask = (n << 3) - 1;
+ memcpy(x, buffer, sizeof(*x) * n);
+ for (i=0; i < 8*n; ++i)
+ mcos[i] = (float) cos(M_PI / 4 * i / n);
+ for (i=0; i < n; ++i) {
+ float acc = 0;
+ for (j=0; j < n; ++j)
+ acc += x[j] * mcos[((2 * i + 1)*(2*j+1)) & nmask];
+ //acc += x[j] * cos(M_PI / n * (i + 0.5) * (j + 0.5));
+ buffer[i] = acc;
+ }
+}
+
+void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype)
+{
+ int i, n4 = n >> 2, n2 = n >> 1, n3_4 = n - n4;
+ float temp[4096];
+
+ memcpy(temp, buffer, n2 * sizeof(float));
+ dct_iv_slow(temp, n2); // returns -c'-d, a-b'
+
+ for (i=0; i < n4 ; ++i) buffer[i] = temp[i+n4]; // a-b'
+ for ( ; i < n3_4; ++i) buffer[i] = -temp[n3_4 - i - 1]; // b-a', c+d'
+ for ( ; i < n ; ++i) buffer[i] = -temp[i - n3_4]; // c'+d
+}
+#endif
+
+#ifndef LIBVORBIS_MDCT
+#define LIBVORBIS_MDCT 0
+#endif
+
+#if LIBVORBIS_MDCT
+// directly call the vorbis MDCT using an interface documented
+// by Jeff Roberts... useful for performance comparison
+typedef struct
+{
+ int n;
+ int log2n;
+
+ float *trig;
+ int *bitrev;
+
+ float scale;
+} mdct_lookup;
+
+extern void mdct_init(mdct_lookup *lookup, int n);
+extern void mdct_clear(mdct_lookup *l);
+extern void mdct_backward(mdct_lookup *init, float *in, float *out);
+
+mdct_lookup M1,M2;
+
+void inverse_mdct(float *buffer, int n, vorb *f, int blocktype)
+{
+ mdct_lookup *M;
+ if (M1.n == n) M = &M1;
+ else if (M2.n == n) M = &M2;
+ else if (M1.n == 0) { mdct_init(&M1, n); M = &M1; }
+ else {
+ if (M2.n) __asm int 3;
+ mdct_init(&M2, n);
+ M = &M2;
+ }
+
+ mdct_backward(M, buffer, buffer);
+}
+#endif
+
+
+// the following were split out into separate functions while optimizing;
+// they could be pushed back up but eh. __forceinline showed no change;
+// they're probably already being inlined.
+static void imdct_step3_iter0_loop(int n, float *e, int i_off, int k_off, float *A)
+{
+ float *ee0 = e + i_off;
+ float *ee2 = ee0 + k_off;
+ int i;
+
+ assert((n & 3) == 0);
+ for (i=(n>>2); i > 0; --i) {
+ float k00_20, k01_21;
+ k00_20 = ee0[ 0] - ee2[ 0];
+ k01_21 = ee0[-1] - ee2[-1];
+ ee0[ 0] += ee2[ 0];//ee0[ 0] = ee0[ 0] + ee2[ 0];
+ ee0[-1] += ee2[-1];//ee0[-1] = ee0[-1] + ee2[-1];
+ ee2[ 0] = k00_20 * A[0] - k01_21 * A[1];
+ ee2[-1] = k01_21 * A[0] + k00_20 * A[1];
+ A += 8;
+
+ k00_20 = ee0[-2] - ee2[-2];
+ k01_21 = ee0[-3] - ee2[-3];
+ ee0[-2] += ee2[-2];//ee0[-2] = ee0[-2] + ee2[-2];
+ ee0[-3] += ee2[-3];//ee0[-3] = ee0[-3] + ee2[-3];
+ ee2[-2] = k00_20 * A[0] - k01_21 * A[1];
+ ee2[-3] = k01_21 * A[0] + k00_20 * A[1];
+ A += 8;
+
+ k00_20 = ee0[-4] - ee2[-4];
+ k01_21 = ee0[-5] - ee2[-5];
+ ee0[-4] += ee2[-4];//ee0[-4] = ee0[-4] + ee2[-4];
+ ee0[-5] += ee2[-5];//ee0[-5] = ee0[-5] + ee2[-5];
+ ee2[-4] = k00_20 * A[0] - k01_21 * A[1];
+ ee2[-5] = k01_21 * A[0] + k00_20 * A[1];
+ A += 8;
+
+ k00_20 = ee0[-6] - ee2[-6];
+ k01_21 = ee0[-7] - ee2[-7];
+ ee0[-6] += ee2[-6];//ee0[-6] = ee0[-6] + ee2[-6];
+ ee0[-7] += ee2[-7];//ee0[-7] = ee0[-7] + ee2[-7];
+ ee2[-6] = k00_20 * A[0] - k01_21 * A[1];
+ ee2[-7] = k01_21 * A[0] + k00_20 * A[1];
+ A += 8;
+ ee0 -= 8;
+ ee2 -= 8;
+ }
+}
+
+static void imdct_step3_inner_r_loop(int lim, float *e, int d0, int k_off, float *A, int k1)
+{
+ int i;
+ float k00_20, k01_21;
+
+ float *e0 = e + d0;
+ float *e2 = e0 + k_off;
+
+ for (i=lim >> 2; i > 0; --i) {
+ k00_20 = e0[-0] - e2[-0];
+ k01_21 = e0[-1] - e2[-1];
+ e0[-0] += e2[-0];//e0[-0] = e0[-0] + e2[-0];
+ e0[-1] += e2[-1];//e0[-1] = e0[-1] + e2[-1];
+ e2[-0] = (k00_20)*A[0] - (k01_21) * A[1];
+ e2[-1] = (k01_21)*A[0] + (k00_20) * A[1];
+
+ A += k1;
+
+ k00_20 = e0[-2] - e2[-2];
+ k01_21 = e0[-3] - e2[-3];
+ e0[-2] += e2[-2];//e0[-2] = e0[-2] + e2[-2];
+ e0[-3] += e2[-3];//e0[-3] = e0[-3] + e2[-3];
+ e2[-2] = (k00_20)*A[0] - (k01_21) * A[1];
+ e2[-3] = (k01_21)*A[0] + (k00_20) * A[1];
+
+ A += k1;
+
+ k00_20 = e0[-4] - e2[-4];
+ k01_21 = e0[-5] - e2[-5];
+ e0[-4] += e2[-4];//e0[-4] = e0[-4] + e2[-4];
+ e0[-5] += e2[-5];//e0[-5] = e0[-5] + e2[-5];
+ e2[-4] = (k00_20)*A[0] - (k01_21) * A[1];
+ e2[-5] = (k01_21)*A[0] + (k00_20) * A[1];
+
+ A += k1;
+
+ k00_20 = e0[-6] - e2[-6];
+ k01_21 = e0[-7] - e2[-7];
+ e0[-6] += e2[-6];//e0[-6] = e0[-6] + e2[-6];
+ e0[-7] += e2[-7];//e0[-7] = e0[-7] + e2[-7];
+ e2[-6] = (k00_20)*A[0] - (k01_21) * A[1];
+ e2[-7] = (k01_21)*A[0] + (k00_20) * A[1];
+
+ e0 -= 8;
+ e2 -= 8;
+
+ A += k1;
+ }
+}
+
+static void imdct_step3_inner_s_loop(int n, float *e, int i_off, int k_off, float *A, int a_off, int k0)
+{
+ int i;
+ float A0 = A[0];
+ float A1 = A[0+1];
+ float A2 = A[0+a_off];
+ float A3 = A[0+a_off+1];
+ float A4 = A[0+a_off*2+0];
+ float A5 = A[0+a_off*2+1];
+ float A6 = A[0+a_off*3+0];
+ float A7 = A[0+a_off*3+1];
+
+ float k00,k11;
+
+ float *ee0 = e +i_off;
+ float *ee2 = ee0+k_off;
+
+ for (i=n; i > 0; --i) {
+ k00 = ee0[ 0] - ee2[ 0];
+ k11 = ee0[-1] - ee2[-1];
+ ee0[ 0] = ee0[ 0] + ee2[ 0];
+ ee0[-1] = ee0[-1] + ee2[-1];
+ ee2[ 0] = (k00) * A0 - (k11) * A1;
+ ee2[-1] = (k11) * A0 + (k00) * A1;
+
+ k00 = ee0[-2] - ee2[-2];
+ k11 = ee0[-3] - ee2[-3];
+ ee0[-2] = ee0[-2] + ee2[-2];
+ ee0[-3] = ee0[-3] + ee2[-3];
+ ee2[-2] = (k00) * A2 - (k11) * A3;
+ ee2[-3] = (k11) * A2 + (k00) * A3;
+
+ k00 = ee0[-4] - ee2[-4];
+ k11 = ee0[-5] - ee2[-5];
+ ee0[-4] = ee0[-4] + ee2[-4];
+ ee0[-5] = ee0[-5] + ee2[-5];
+ ee2[-4] = (k00) * A4 - (k11) * A5;
+ ee2[-5] = (k11) * A4 + (k00) * A5;
+
+ k00 = ee0[-6] - ee2[-6];
+ k11 = ee0[-7] - ee2[-7];
+ ee0[-6] = ee0[-6] + ee2[-6];
+ ee0[-7] = ee0[-7] + ee2[-7];
+ ee2[-6] = (k00) * A6 - (k11) * A7;
+ ee2[-7] = (k11) * A6 + (k00) * A7;
+
+ ee0 -= k0;
+ ee2 -= k0;
+ }
+}
+
+static __forceinline void iter_54(float *z)
+{
+ float k00,k11,k22,k33;
+ float y0,y1,y2,y3;
+
+ k00 = z[ 0] - z[-4];
+ y0 = z[ 0] + z[-4];
+ y2 = z[-2] + z[-6];
+ k22 = z[-2] - z[-6];
+
+ z[-0] = y0 + y2; // z0 + z4 + z2 + z6
+ z[-2] = y0 - y2; // z0 + z4 - z2 - z6
+
+ // done with y0,y2
+
+ k33 = z[-3] - z[-7];
+
+ z[-4] = k00 + k33; // z0 - z4 + z3 - z7
+ z[-6] = k00 - k33; // z0 - z4 - z3 + z7
+
+ // done with k33
+
+ k11 = z[-1] - z[-5];
+ y1 = z[-1] + z[-5];
+ y3 = z[-3] + z[-7];
+
+ z[-1] = y1 + y3; // z1 + z5 + z3 + z7
+ z[-3] = y1 - y3; // z1 + z5 - z3 - z7
+ z[-5] = k11 - k22; // z1 - z5 + z2 - z6
+ z[-7] = k11 + k22; // z1 - z5 - z2 + z6
+}
+
+static void imdct_step3_inner_s_loop_ld654(int n, float *e, int i_off, float *A, int base_n)
+{
+ int k_off = -8;
+ int a_off = base_n >> 3;
+ float A2 = A[0+a_off];
+ float *z = e + i_off;
+ float *base = z - 16 * n;
+
+ while (z > base) {
+ float k00,k11;
+
+ k00 = z[-0] - z[-8];
+ k11 = z[-1] - z[-9];
+ z[-0] = z[-0] + z[-8];
+ z[-1] = z[-1] + z[-9];
+ z[-8] = k00;
+ z[-9] = k11 ;
+
+ k00 = z[ -2] - z[-10];
+ k11 = z[ -3] - z[-11];
+ z[ -2] = z[ -2] + z[-10];
+ z[ -3] = z[ -3] + z[-11];
+ z[-10] = (k00+k11) * A2;
+ z[-11] = (k11-k00) * A2;
+
+ k00 = z[-12] - z[ -4]; // reverse to avoid a unary negation
+ k11 = z[ -5] - z[-13];
+ z[ -4] = z[ -4] + z[-12];
+ z[ -5] = z[ -5] + z[-13];
+ z[-12] = k11;
+ z[-13] = k00;
+
+ k00 = z[-14] - z[ -6]; // reverse to avoid a unary negation
+ k11 = z[ -7] - z[-15];
+ z[ -6] = z[ -6] + z[-14];
+ z[ -7] = z[ -7] + z[-15];
+ z[-14] = (k00+k11) * A2;
+ z[-15] = (k00-k11) * A2;
+
+ iter_54(z);
+ iter_54(z-8);
+ z -= 16;
+ }
+}
+
+static void inverse_mdct(float *buffer, int n, vorb *f, int blocktype)
+{
+ int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l;
+ int n3_4 = n - n4, ld;
+ // @OPTIMIZE: reduce register pressure by using fewer variables?
+ int save_point = temp_alloc_save(f);
+ float *buf2 = (float *) temp_alloc(f, n2 * sizeof(*buf2));
+ float *u=NULL,*v=NULL;
+ // twiddle factors
+ float *A = f->A[blocktype];
+
+ // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio"
+ // See notes about bugs in that paper in less-optimal implementation 'inverse_mdct_old' after this function.
+
+ // kernel from paper
+
+
+ // merged:
+ // copy and reflect spectral data
+ // step 0
+
+ // note that it turns out that the items added together during
+ // this step are, in fact, being added to themselves (as reflected
+ // by step 0). inexplicable inefficiency! this became obvious
+ // once I combined the passes.
+
+ // so there's a missing 'times 2' here (for adding X to itself).
+ // this propogates through linearly to the end, where the numbers
+ // are 1/2 too small, and need to be compensated for.
+
+ {
+ float *d,*e, *AA, *e_stop;
+ d = &buf2[n2-2];
+ AA = A;
+ e = &buffer[0];
+ e_stop = &buffer[n2];
+ while (e != e_stop) {
+ d[1] = (e[0] * AA[0] - e[2]*AA[1]);
+ d[0] = (e[0] * AA[1] + e[2]*AA[0]);
+ d -= 2;
+ AA += 2;
+ e += 4;
+ }
+
+ e = &buffer[n2-3];
+ while (d >= buf2) {
+ d[1] = (-e[2] * AA[0] - -e[0]*AA[1]);
+ d[0] = (-e[2] * AA[1] + -e[0]*AA[0]);
+ d -= 2;
+ AA += 2;
+ e -= 4;
+ }
+ }
+
+ // now we use symbolic names for these, so that we can
+ // possibly swap their meaning as we change which operations
+ // are in place
+
+ u = buffer;
+ v = buf2;
+
+ // step 2 (paper output is w, now u)
+ // this could be in place, but the data ends up in the wrong
+ // place... _somebody_'s got to swap it, so this is nominated
+ {
+ float *AA = &A[n2-8];
+ float *d0,*d1, *e0, *e1;
+
+ e0 = &v[n4];
+ e1 = &v[0];
+
+ d0 = &u[n4];
+ d1 = &u[0];
+
+ while (AA >= A) {
+ float v40_20, v41_21;
+
+ v41_21 = e0[1] - e1[1];
+ v40_20 = e0[0] - e1[0];
+ d0[1] = e0[1] + e1[1];
+ d0[0] = e0[0] + e1[0];
+ d1[1] = v41_21*AA[4] - v40_20*AA[5];
+ d1[0] = v40_20*AA[4] + v41_21*AA[5];
+
+ v41_21 = e0[3] - e1[3];
+ v40_20 = e0[2] - e1[2];
+ d0[3] = e0[3] + e1[3];
+ d0[2] = e0[2] + e1[2];
+ d1[3] = v41_21*AA[0] - v40_20*AA[1];
+ d1[2] = v40_20*AA[0] + v41_21*AA[1];
+
+ AA -= 8;
+
+ d0 += 4;
+ d1 += 4;
+ e0 += 4;
+ e1 += 4;
+ }
+ }
+
+ // step 3
+ ld = ilog(n) - 1; // ilog is off-by-one from normal definitions
+
+ // optimized step 3:
+
+ // the original step3 loop can be nested r inside s or s inside r;
+ // it's written originally as s inside r, but this is dumb when r
+ // iterates many times, and s few. So I have two copies of it and
+ // switch between them halfway.
+
+ // this is iteration 0 of step 3
+ imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*0, -(n >> 3), A);
+ imdct_step3_iter0_loop(n >> 4, u, n2-1-n4*1, -(n >> 3), A);
+
+ // this is iteration 1 of step 3
+ imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*0, -(n >> 4), A, 16);
+ imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*1, -(n >> 4), A, 16);
+ imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*2, -(n >> 4), A, 16);
+ imdct_step3_inner_r_loop(n >> 5, u, n2-1 - n8*3, -(n >> 4), A, 16);
+
+ l=2;
+ for (; l < (ld-3)>>1; ++l) {
+ int k0 = n >> (l+2), k0_2 = k0>>1;
+ int lim = 1 << (l+1);
+ int i;
+ for (i=0; i < lim; ++i)
+ imdct_step3_inner_r_loop(n >> (l+4), u, n2-1 - k0*i, -k0_2, A, 1 << (l+3));
+ }
+
+ for (; l < ld-6; ++l) {
+ int k0 = n >> (l+2), k1 = 1 << (l+3), k0_2 = k0>>1;
+ int rlim = n >> (l+6), r;
+ int lim = 1 << (l+1);
+ int i_off;
+ float *A0 = A;
+ i_off = n2-1;
+ for (r=rlim; r > 0; --r) {
+ imdct_step3_inner_s_loop(lim, u, i_off, -k0_2, A0, k1, k0);
+ A0 += k1*4;
+ i_off -= 8;
+ }
+ }
+
+ // iterations with count:
+ // ld-6,-5,-4 all interleaved together
+ // the big win comes from getting rid of needless flops
+ // due to the constants on pass 5 & 4 being all 1 and 0;
+ // combining them to be simultaneous to improve cache made little difference
+ imdct_step3_inner_s_loop_ld654(n >> 5, u, n2-1, A, n);
+
+ // output is u
+
+ // step 4, 5, and 6
+ // cannot be in-place because of step 5
+ {
+ uint16 *bitrev = f->bit_reverse[blocktype];
+ // weirdly, I'd have thought reading sequentially and writing
+ // erratically would have been better than vice-versa, but in
+ // fact that's not what my testing showed. (That is, with
+ // j = bitreverse(i), do you read i and write j, or read j and write i.)
+
+ float *d0 = &v[n4-4];
+ float *d1 = &v[n2-4];
+ while (d0 >= v) {
+ int k4;
+
+ k4 = bitrev[0];
+ d1[3] = u[k4+0];
+ d1[2] = u[k4+1];
+ d0[3] = u[k4+2];
+ d0[2] = u[k4+3];
+
+ k4 = bitrev[1];
+ d1[1] = u[k4+0];
+ d1[0] = u[k4+1];
+ d0[1] = u[k4+2];
+ d0[0] = u[k4+3];
+
+ d0 -= 4;
+ d1 -= 4;
+ bitrev += 2;
+ }
+ }
+ // (paper output is u, now v)
+
+
+ // data must be in buf2
+ assert(v == buf2);
+
+ // step 7 (paper output is v, now v)
+ // this is now in place
+ {
+ float *C = f->C[blocktype];
+ float *d, *e;
+
+ d = v;
+ e = v + n2 - 4;
+
+ while (d < e) {
+ float a02,a11,b0,b1,b2,b3;
+
+ a02 = d[0] - e[2];
+ a11 = d[1] + e[3];
+
+ b0 = C[1]*a02 + C[0]*a11;
+ b1 = C[1]*a11 - C[0]*a02;
+
+ b2 = d[0] + e[ 2];
+ b3 = d[1] - e[ 3];
+
+ d[0] = b2 + b0;
+ d[1] = b3 + b1;
+ e[2] = b2 - b0;
+ e[3] = b1 - b3;
+
+ a02 = d[2] - e[0];
+ a11 = d[3] + e[1];
+
+ b0 = C[3]*a02 + C[2]*a11;
+ b1 = C[3]*a11 - C[2]*a02;
+
+ b2 = d[2] + e[ 0];
+ b3 = d[3] - e[ 1];
+
+ d[2] = b2 + b0;
+ d[3] = b3 + b1;
+ e[0] = b2 - b0;
+ e[1] = b1 - b3;
+
+ C += 4;
+ d += 4;
+ e -= 4;
+ }
+ }
+
+ // data must be in buf2
+
+
+ // step 8+decode (paper output is X, now buffer)
+ // this generates pairs of data a la 8 and pushes them directly through
+ // the decode kernel (pushing rather than pulling) to avoid having
+ // to make another pass later
+
+ // this cannot POSSIBLY be in place, so we refer to the buffers directly
+
+ {
+ float *d0,*d1,*d2,*d3;
+
+ float *B = f->B[blocktype] + n2 - 8;
+ float *e = buf2 + n2 - 8;
+ d0 = &buffer[0];
+ d1 = &buffer[n2-4];
+ d2 = &buffer[n2];
+ d3 = &buffer[n-4];
+ while (e >= v) {
+ float p0,p1,p2,p3;
+
+ p3 = e[6]*B[7] - e[7]*B[6];
+ p2 = -e[6]*B[6] - e[7]*B[7];
+
+ d0[0] = p3;
+ d1[3] = - p3;
+ d2[0] = p2;
+ d3[3] = p2;
+
+ p1 = e[4]*B[5] - e[5]*B[4];
+ p0 = -e[4]*B[4] - e[5]*B[5];
+
+ d0[1] = p1;
+ d1[2] = - p1;
+ d2[1] = p0;
+ d3[2] = p0;
+
+ p3 = e[2]*B[3] - e[3]*B[2];
+ p2 = -e[2]*B[2] - e[3]*B[3];
+
+ d0[2] = p3;
+ d1[1] = - p3;
+ d2[2] = p2;
+ d3[1] = p2;
+
+ p1 = e[0]*B[1] - e[1]*B[0];
+ p0 = -e[0]*B[0] - e[1]*B[1];
+
+ d0[3] = p1;
+ d1[0] = - p1;
+ d2[3] = p0;
+ d3[0] = p0;
+
+ B -= 8;
+ e -= 8;
+ d0 += 4;
+ d2 += 4;
+ d1 -= 4;
+ d3 -= 4;
+ }
+ }
+
+ temp_alloc_restore(f,save_point);
+}
+
+#if 0
+// this is the original version of the above code, if you want to optimize it from scratch
+void inverse_mdct_naive(float *buffer, int n)
+{
+ float s;
+ float A[1 << 12], B[1 << 12], C[1 << 11];
+ int i,k,k2,k4, n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l;
+ int n3_4 = n - n4, ld;
+ // how can they claim this only uses N words?!
+ // oh, because they're only used sparsely, whoops
+ float u[1 << 13], X[1 << 13], v[1 << 13], w[1 << 13];
+ // set up twiddle factors
+
+ for (k=k2=0; k < n4; ++k,k2+=2) {
+ A[k2 ] = (float) cos(4*k*M_PI/n);
+ A[k2+1] = (float) -sin(4*k*M_PI/n);
+ B[k2 ] = (float) cos((k2+1)*M_PI/n/2);
+ B[k2+1] = (float) sin((k2+1)*M_PI/n/2);
+ }
+ for (k=k2=0; k < n8; ++k,k2+=2) {
+ C[k2 ] = (float) cos(2*(k2+1)*M_PI/n);
+ C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n);
+ }
+
+ // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio"
+ // Note there are bugs in that pseudocode, presumably due to them attempting
+ // to rename the arrays nicely rather than representing the way their actual
+ // implementation bounces buffers back and forth. As a result, even in the
+ // "some formulars corrected" version, a direct implementation fails. These
+ // are noted below as "paper bug".
+
+ // copy and reflect spectral data
+ for (k=0; k < n2; ++k) u[k] = buffer[k];
+ for ( ; k < n ; ++k) u[k] = -buffer[n - k - 1];
+ // kernel from paper
+ // step 1
+ for (k=k2=k4=0; k < n4; k+=1, k2+=2, k4+=4) {
+ v[n-k4-1] = (u[k4] - u[n-k4-1]) * A[k2] - (u[k4+2] - u[n-k4-3])*A[k2+1];
+ v[n-k4-3] = (u[k4] - u[n-k4-1]) * A[k2+1] + (u[k4+2] - u[n-k4-3])*A[k2];
+ }
+ // step 2
+ for (k=k4=0; k < n8; k+=1, k4+=4) {
+ w[n2+3+k4] = v[n2+3+k4] + v[k4+3];
+ w[n2+1+k4] = v[n2+1+k4] + v[k4+1];
+ w[k4+3] = (v[n2+3+k4] - v[k4+3])*A[n2-4-k4] - (v[n2+1+k4]-v[k4+1])*A[n2-3-k4];
+ w[k4+1] = (v[n2+1+k4] - v[k4+1])*A[n2-4-k4] + (v[n2+3+k4]-v[k4+3])*A[n2-3-k4];
+ }
+ // step 3
+ ld = ilog(n) - 1; // ilog is off-by-one from normal definitions
+ for (l=0; l < ld-3; ++l) {
+ int k0 = n >> (l+2), k1 = 1 << (l+3);
+ int rlim = n >> (l+4), r4, r;
+ int s2lim = 1 << (l+2), s2;
+ for (r=r4=0; r < rlim; r4+=4,++r) {
+ for (s2=0; s2 < s2lim; s2+=2) {
+ u[n-1-k0*s2-r4] = w[n-1-k0*s2-r4] + w[n-1-k0*(s2+1)-r4];
+ u[n-3-k0*s2-r4] = w[n-3-k0*s2-r4] + w[n-3-k0*(s2+1)-r4];
+ u[n-1-k0*(s2+1)-r4] = (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1]
+ - (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1+1];
+ u[n-3-k0*(s2+1)-r4] = (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1]
+ + (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1+1];
+ }
+ }
+ if (l+1 < ld-3) {
+ // paper bug: ping-ponging of u&w here is omitted
+ memcpy(w, u, sizeof(u));
+ }
+ }
+
+ // step 4
+ for (i=0; i < n8; ++i) {
+ int j = bit_reverse(i) >> (32-ld+3);
+ assert(j < n8);
+ if (i == j) {
+ // paper bug: original code probably swapped in place; if copying,
+ // need to directly copy in this case
+ int i8 = i << 3;
+ v[i8+1] = u[i8+1];
+ v[i8+3] = u[i8+3];
+ v[i8+5] = u[i8+5];
+ v[i8+7] = u[i8+7];
+ } else if (i < j) {
+ int i8 = i << 3, j8 = j << 3;
+ v[j8+1] = u[i8+1], v[i8+1] = u[j8 + 1];
+ v[j8+3] = u[i8+3], v[i8+3] = u[j8 + 3];
+ v[j8+5] = u[i8+5], v[i8+5] = u[j8 + 5];
+ v[j8+7] = u[i8+7], v[i8+7] = u[j8 + 7];
+ }
+ }
+ // step 5
+ for (k=0; k < n2; ++k) {
+ w[k] = v[k*2+1];
+ }
+ // step 6
+ for (k=k2=k4=0; k < n8; ++k, k2 += 2, k4 += 4) {
+ u[n-1-k2] = w[k4];
+ u[n-2-k2] = w[k4+1];
+ u[n3_4 - 1 - k2] = w[k4+2];
+ u[n3_4 - 2 - k2] = w[k4+3];
+ }
+ // step 7
+ for (k=k2=0; k < n8; ++k, k2 += 2) {
+ v[n2 + k2 ] = ( u[n2 + k2] + u[n-2-k2] + C[k2+1]*(u[n2+k2]-u[n-2-k2]) + C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2;
+ v[n-2 - k2] = ( u[n2 + k2] + u[n-2-k2] - C[k2+1]*(u[n2+k2]-u[n-2-k2]) - C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2;
+ v[n2+1+ k2] = ( u[n2+1+k2] - u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2;
+ v[n-1 - k2] = (-u[n2+1+k2] + u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2;
+ }
+ // step 8
+ for (k=k2=0; k < n4; ++k,k2 += 2) {
+ X[k] = v[k2+n2]*B[k2 ] + v[k2+1+n2]*B[k2+1];
+ X[n2-1-k] = v[k2+n2]*B[k2+1] - v[k2+1+n2]*B[k2 ];
+ }
+
+ // decode kernel to output
+ // determined the following value experimentally
+ // (by first figuring out what made inverse_mdct_slow work); then matching that here
+ // (probably vorbis encoder premultiplies by n or n/2, to save it on the decoder?)
+ s = 0.5; // theoretically would be n4
+
+ // [[[ note! the s value of 0.5 is compensated for by the B[] in the current code,
+ // so it needs to use the "old" B values to behave correctly, or else
+ // set s to 1.0 ]]]
+ for (i=0; i < n4 ; ++i) buffer[i] = s * X[i+n4];
+ for ( ; i < n3_4; ++i) buffer[i] = -s * X[n3_4 - i - 1];
+ for ( ; i < n ; ++i) buffer[i] = -s * X[i - n3_4];
+}
+#endif
+
+static float *get_window(vorb *f, int len)
+{
+ len <<= 1;
+ if (len == f->blocksize_0) return f->window[0];
+ if (len == f->blocksize_1) return f->window[1];
+ assert(0);
+ return NULL;
+}
+
+#ifndef STB_VORBIS_NO_DEFER_FLOOR
+typedef int16 YTYPE;
+#else
+typedef int YTYPE;
+#endif
+static int do_floor(vorb *f, Mapping *map, int i, int n, float *target, YTYPE *finalY, uint8 *step2_flag)
+{
+ int n2 = n >> 1;
+ int s = map->chan[i].mux, floor;
+ floor = map->submap_floor[s];
+ if (f->floor_types[floor] == 0) {
+ return error(f, VORBIS_invalid_stream);
+ } else {
+ Floor1 *g = &f->floor_config[floor].floor1;
+ int j,q;
+ int lx = 0, ly = finalY[0] * g->floor1_multiplier;
+ for (q=1; q < g->values; ++q) {
+ j = g->sorted_order[q];
+ #ifndef STB_VORBIS_NO_DEFER_FLOOR
+ if (finalY[j] >= 0)
+ #else
+ if (step2_flag[j])
+ #endif
+ {
+ int hy = finalY[j] * g->floor1_multiplier;
+ int hx = g->Xlist[j];
+ draw_line(target, lx,ly, hx,hy, n2);
+ lx = hx, ly = hy;
+ }
+ }
+ if (lx < n2)
+ // optimization of: draw_line(target, lx,ly, n,ly, n2);
+ for (j=lx; j < n2; ++j)
+ LINE_OP(target[j], inverse_db_table[ly]);
+ }
+ return TRUE;
+}
+
+static int vorbis_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode)
+{
+ Mode *m;
+ int i, n, prev, next, window_center;
+ f->channel_buffer_start = f->channel_buffer_end = 0;
+
+ retry:
+ if (f->eof) return FALSE;
+ if (!maybe_start_packet(f))
+ return FALSE;
+ // check packet type
+ if (get_bits(f,1) != 0) {
+ if (IS_PUSH_MODE(f))
+ return error(f,VORBIS_bad_packet_type);
+ while (EOP != get8_packet(f));
+ goto retry;
+ }
+
+ if (f->alloc.alloc_buffer)
+ assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset);
+
+ i = get_bits(f, ilog(f->mode_count-1));
+ if (i == EOP) return FALSE;
+ if (i >= f->mode_count) return FALSE;
+ *mode = i;
+ m = f->mode_config + i;
+ if (m->blockflag) {
+ n = f->blocksize_1;
+ prev = get_bits(f,1);
+ next = get_bits(f,1);
+ } else {
+ prev = next = 0;
+ n = f->blocksize_0;
+ }
+
+// WINDOWING
+
+ window_center = n >> 1;
+ if (m->blockflag && !prev) {
+ *p_left_start = (n - f->blocksize_0) >> 2;
+ *p_left_end = (n + f->blocksize_0) >> 2;
+ } else {
+ *p_left_start = 0;
+ *p_left_end = window_center;
+ }
+ if (m->blockflag && !next) {
+ *p_right_start = (n*3 - f->blocksize_0) >> 2;
+ *p_right_end = (n*3 + f->blocksize_0) >> 2;
+ } else {
+ *p_right_start = window_center;
+ *p_right_end = n;
+ }
+ return TRUE;
+}
+
+static int vorbis_decode_packet_rest(vorb *f, int *len, Mode *m, int left_start, int left_end, int right_start, int right_end, int *p_left)
+{
+ Mapping *map;
+ int i,j,k,n,n2;
+ int zero_channel[256];
+ int really_zero_channel[256];
+ int window_center;
+
+// WINDOWING
+
+ n = f->blocksize[m->blockflag];
+ window_center = n >> 1;
+
+ map = &f->mapping[m->mapping];
+
+// FLOORS
+ n2 = n >> 1;
+
+ stb_prof(1);
+ for (i=0; i < f->channels; ++i) {
+ int s = map->chan[i].mux, floor;
+ zero_channel[i] = FALSE;
+ floor = map->submap_floor[s];
+ if (f->floor_types[floor] == 0) {
+ return error(f, VORBIS_invalid_stream);
+ } else {
+ Floor1 *g = &f->floor_config[floor].floor1;
+ if (get_bits(f, 1)) {
+ short *finalY;
+ uint8 step2_flag[256];
+ static int range_list[4] = { 256, 128, 86, 64 };
+ int range = range_list[g->floor1_multiplier-1];
+ int offset = 2;
+ finalY = f->finalY[i];
+ finalY[0] = get_bits(f, ilog(range)-1);
+ finalY[1] = get_bits(f, ilog(range)-1);
+ for (j=0; j < g->partitions; ++j) {
+ int pclass = g->partition_class_list[j];
+ int cdim = g->class_dimensions[pclass];
+ int cbits = g->class_subclasses[pclass];
+ int csub = (1 << cbits)-1;
+ int cval = 0;
+ if (cbits) {
+ Codebook *c = f->codebooks + g->class_masterbooks[pclass];
+ DECODE(cval,f,c);
+ }
+ for (k=0; k < cdim; ++k) {
+ int book = g->subclass_books[pclass][cval & csub];
+ cval = cval >> cbits;
+ if (book >= 0) {
+ int temp;
+ Codebook *c = f->codebooks + book;
+ DECODE(temp,f,c);
+ finalY[offset++] = temp;
+ } else
+ finalY[offset++] = 0;
+ }
+ }
+ if (f->valid_bits == INVALID_BITS) goto error; // behavior according to spec
+ step2_flag[0] = step2_flag[1] = 1;
+ for (j=2; j < g->values; ++j) {
+ int low, high, pred, highroom, lowroom, room, val;
+ low = g->neighbors[j][0];
+ high = g->neighbors[j][1];
+ //neighbors(g->Xlist, j, &low, &high);
+ pred = predict_point(g->Xlist[j], g->Xlist[low], g->Xlist[high], finalY[low], finalY[high]);
+ val = finalY[j];
+ highroom = range - pred;
+ lowroom = pred;
+ if (highroom < lowroom)
+ room = highroom * 2;
+ else
+ room = lowroom * 2;
+ if (val) {
+ step2_flag[low] = step2_flag[high] = 1;
+ step2_flag[j] = 1;
+ if (val >= room)
+ if (highroom > lowroom)
+ finalY[j] = val - lowroom + pred;
+ else
+ finalY[j] = pred - val + highroom - 1;
+ else
+ if (val & 1)
+ finalY[j] = pred - ((val+1)>>1);
+ else
+ finalY[j] = pred + (val>>1);
+ } else {
+ step2_flag[j] = 0;
+ finalY[j] = pred;
+ }
+ }
+
+#ifdef STB_VORBIS_NO_DEFER_FLOOR
+ do_floor(f, map, i, n, f->floor_buffers[i], finalY, step2_flag);
+#else
+ // defer final floor computation until _after_ residue
+ for (j=0; j < g->values; ++j) {
+ if (!step2_flag[j])
+ finalY[j] = -1;
+ }
+#endif
+ } else {
+ error:
+ zero_channel[i] = TRUE;
+ }
+ // So we just defer everything else to later
+
+ // at this point we've decoded the floor into buffer
+ }
+ }
+ stb_prof(0);
+ // at this point we've decoded all floors
+
+ if (f->alloc.alloc_buffer)
+ assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset);
+
+ // re-enable coupled channels if necessary
+ memcpy(really_zero_channel, zero_channel, sizeof(really_zero_channel[0]) * f->channels);
+ for (i=0; i < map->coupling_steps; ++i)
+ if (!zero_channel[map->chan[i].magnitude] || !zero_channel[map->chan[i].angle]) {
+ zero_channel[map->chan[i].magnitude] = zero_channel[map->chan[i].angle] = FALSE;
+ }
+
+// RESIDUE DECODE
+ for (i=0; i < map->submaps; ++i) {
+ float *residue_buffers[STB_VORBIS_MAX_CHANNELS];
+ int r,t;
+ uint8 do_not_decode[256];
+ int ch = 0;
+ for (j=0; j < f->channels; ++j) {
+ if (map->chan[j].mux == i) {
+ if (zero_channel[j]) {
+ do_not_decode[ch] = TRUE;
+ residue_buffers[ch] = NULL;
+ } else {
+ do_not_decode[ch] = FALSE;
+ residue_buffers[ch] = f->channel_buffers[j];
+ }
+ ++ch;
+ }
+ }
+ r = map->submap_residue[i];
+ t = f->residue_types[r];
+ decode_residue(f, residue_buffers, ch, n2, r, do_not_decode);
+ }
+
+ if (f->alloc.alloc_buffer)
+ assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset);
+
+// INVERSE COUPLING
+ stb_prof(14);
+ for (i = map->coupling_steps-1; i >= 0; --i) {
+ int n2 = n >> 1;
+ float *m = f->channel_buffers[map->chan[i].magnitude];
+ float *a = f->channel_buffers[map->chan[i].angle ];
+ for (j=0; j < n2; ++j) {
+ float a2,m2;
+ if (m[j] > 0)
+ if (a[j] > 0)
+ m2 = m[j], a2 = m[j] - a[j];
+ else
+ a2 = m[j], m2 = m[j] + a[j];
+ else
+ if (a[j] > 0)
+ m2 = m[j], a2 = m[j] + a[j];
+ else
+ a2 = m[j], m2 = m[j] - a[j];
+ m[j] = m2;
+ a[j] = a2;
+ }
+ }
+
+ // finish decoding the floors
+#ifndef STB_VORBIS_NO_DEFER_FLOOR
+ stb_prof(15);
+ for (i=0; i < f->channels; ++i) {
+ if (really_zero_channel[i]) {
+ memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2);
+ } else {
+ do_floor(f, map, i, n, f->channel_buffers[i], f->finalY[i], NULL);
+ }
+ }
+#else
+ for (i=0; i < f->channels; ++i) {
+ if (really_zero_channel[i]) {
+ memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2);
+ } else {
+ for (j=0; j < n2; ++j)
+ f->channel_buffers[i][j] *= f->floor_buffers[i][j];
+ }
+ }
+#endif
+
+// INVERSE MDCT
+ stb_prof(16);
+ for (i=0; i < f->channels; ++i)
+ inverse_mdct(f->channel_buffers[i], n, f, m->blockflag);
+ stb_prof(0);
+
+ // this shouldn't be necessary, unless we exited on an error
+ // and want to flush to get to the next packet
+ flush_packet(f);
+
+ if (f->first_decode) {
+ // assume we start so first non-discarded sample is sample 0
+ // this isn't to spec, but spec would require us to read ahead
+ // and decode the size of all current frames--could be done,
+ // but presumably it's not a commonly used feature
+ f->current_loc = -n2; // start of first frame is positioned for discard
+ // we might have to discard samples "from" the next frame too,
+ // if we're lapping a large block then a small at the start?
+ f->discard_samples_deferred = n - right_end;
+ f->current_loc_valid = TRUE;
+ f->first_decode = FALSE;
+ } else if (f->discard_samples_deferred) {
+ left_start += f->discard_samples_deferred;
+ *p_left = left_start;
+ f->discard_samples_deferred = 0;
+ } else if (f->previous_length == 0 && f->current_loc_valid) {
+ // we're recovering from a seek... that means we're going to discard
+ // the samples from this packet even though we know our position from
+ // the last page header, so we need to update the position based on
+ // the discarded samples here
+ // but wait, the code below is going to add this in itself even
+ // on a discard, so we don't need to do it here...
+ }
+
+ // check if we have ogg information about the sample # for this packet
+ if (f->last_seg_which == f->end_seg_with_known_loc) {
+ // if we have a valid current loc, and this is final:
+ if (f->current_loc_valid && (f->page_flag & PAGEFLAG_last_page)) {
+ uint32 current_end = f->known_loc_for_packet - (n-right_end);
+ // then let's infer the size of the (probably) short final frame
+ if (current_end < f->current_loc + right_end) {
+ if (current_end < f->current_loc) {
+ // negative truncation, that's impossible!
+ *len = 0;
+ } else {
+ *len = current_end - f->current_loc;
+ }
+ *len += left_start;
+ f->current_loc += *len;
+ return TRUE;
+ }
+ }
+ // otherwise, just set our sample loc
+ // guess that the ogg granule pos refers to the _middle_ of the
+ // last frame?
+ // set f->current_loc to the position of left_start
+ f->current_loc = f->known_loc_for_packet - (n2-left_start);
+ f->current_loc_valid = TRUE;
+ }
+ if (f->current_loc_valid)
+ f->current_loc += (right_start - left_start);
+
+ if (f->alloc.alloc_buffer)
+ assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset);
+ *len = right_end; // ignore samples after the window goes to 0
+ return TRUE;
+}
+
+static int vorbis_decode_packet(vorb *f, int *len, int *p_left, int *p_right)
+{
+ int mode, left_end, right_end;
+ if (!vorbis_decode_initial(f, p_left, &left_end, p_right, &right_end, &mode)) return 0;
+ return vorbis_decode_packet_rest(f, len, f->mode_config + mode, *p_left, left_end, *p_right, right_end, p_left);
+}
+
+static int vorbis_finish_frame(stb_vorbis *f, int len, int left, int right)
+{
+ int prev,i,j;
+ // we use right&left (the start of the right- and left-window sin()-regions)
+ // to determine how much to return, rather than inferring from the rules
+ // (same result, clearer code); 'left' indicates where our sin() window
+ // starts, therefore where the previous window's right edge starts, and
+ // therefore where to start mixing from the previous buffer. 'right'
+ // indicates where our sin() ending-window starts, therefore that's where
+ // we start saving, and where our returned-data ends.
+
+ // mixin from previous window
+ if (f->previous_length) {
+ int i,j, n = f->previous_length;
+ float *w = get_window(f, n);
+ for (i=0; i < f->channels; ++i) {
+ for (j=0; j < n; ++j)
+ f->channel_buffers[i][left+j] =
+ f->channel_buffers[i][left+j]*w[ j] +
+ f->previous_window[i][ j]*w[n-1-j];
+ }
+ }
+
+ prev = f->previous_length;
+
+ // last half of this data becomes previous window
+ f->previous_length = len - right;
+
+ // @OPTIMIZE: could avoid this copy by double-buffering the
+ // output (flipping previous_window with channel_buffers), but
+ // then previous_window would have to be 2x as large, and
+ // channel_buffers couldn't be temp mem (although they're NOT
+ // currently temp mem, they could be (unless we want to level
+ // performance by spreading out the computation))
+ for (i=0; i < f->channels; ++i)
+ for (j=0; right+j < len; ++j)
+ f->previous_window[i][j] = f->channel_buffers[i][right+j];
+
+ if (!prev)
+ // there was no previous packet, so this data isn't valid...
+ // this isn't entirely true, only the would-have-overlapped data
+ // isn't valid, but this seems to be what the spec requires
+ return 0;
+
+ // truncate a short frame
+ if (len < right) right = len;
+
+ f->samples_output += right-left;
+
+ return right - left;
+}
+
+static void vorbis_pump_first_frame(stb_vorbis *f)
+{
+ int len, right, left;
+ if (vorbis_decode_packet(f, &len, &left, &right))
+ vorbis_finish_frame(f, len, left, right);
+}
+
+#ifndef STB_VORBIS_NO_PUSHDATA_API
+static int is_whole_packet_present(stb_vorbis *f, int end_page)
+{
+ // make sure that we have the packet available before continuing...
+ // this requires a full ogg parse, but we know we can fetch from f->stream
+
+ // instead of coding this out explicitly, we could save the current read state,
+ // read the next packet with get8() until end-of-packet, check f->eof, then
+ // reset the state? but that would be slower, esp. since we'd have over 256 bytes
+ // of state to restore (primarily the page segment table)
+
+ int s = f->next_seg, first = TRUE;
+ uint8 *p = f->stream;
+
+ if (s != -1) { // if we're not starting the packet with a 'continue on next page' flag
+ for (; s < f->segment_count; ++s) {
+ p += f->segments[s];
+ if (f->segments[s] < 255) // stop at first short segment
+ break;
+ }
+ // either this continues, or it ends it...
+ if (end_page)
+ if (s < f->segment_count-1) return error(f, VORBIS_invalid_stream);
+ if (s == f->segment_count)
+ s = -1; // set 'crosses page' flag
+ if (p > f->stream_end) return error(f, VORBIS_need_more_data);
+ first = FALSE;
+ }
+ for (; s == -1;) {
+ uint8 *q;
+ int n;
+
+ // check that we have the page header ready
+ if (p + 26 >= f->stream_end) return error(f, VORBIS_need_more_data);
+ // validate the page
+ if (memcmp(p, ogg_page_header, 4)) return error(f, VORBIS_invalid_stream);
+ if (p[4] != 0) return error(f, VORBIS_invalid_stream);
+ if (first) { // the first segment must NOT have 'continued_packet', later ones MUST
+ if (f->previous_length)
+ if ((p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream);
+ // if no previous length, we're resynching, so we can come in on a continued-packet,
+ // which we'll just drop
+ } else {
+ if (!(p[5] & PAGEFLAG_continued_packet)) return error(f, VORBIS_invalid_stream);
+ }
+ n = p[26]; // segment counts
+ q = p+27; // q points to segment table
+ p = q + n; // advance past header
+ // make sure we've read the segment table
+ if (p > f->stream_end) return error(f, VORBIS_need_more_data);
+ for (s=0; s < n; ++s) {
+ p += q[s];
+ if (q[s] < 255)
+ break;
+ }
+ if (end_page)
+ if (s < n-1) return error(f, VORBIS_invalid_stream);
+ if (s == f->segment_count)
+ s = -1; // set 'crosses page' flag
+ if (p > f->stream_end) return error(f, VORBIS_need_more_data);
+ first = FALSE;
+ }
+ return TRUE;
+}
+#endif // !STB_VORBIS_NO_PUSHDATA_API
+
+static int start_decoder(vorb *f)
+{
+ uint8 header[6], x,y;
+ int len,i,j,k, max_submaps = 0;
+ int longest_floorlist=0;
+
+ // first page, first packet
+
+ if (!start_page(f)) return FALSE;
+ // validate page flag
+ if (!(f->page_flag & PAGEFLAG_first_page)) return error(f, VORBIS_invalid_first_page);
+ if (f->page_flag & PAGEFLAG_last_page) return error(f, VORBIS_invalid_first_page);
+ if (f->page_flag & PAGEFLAG_continued_packet) return error(f, VORBIS_invalid_first_page);
+ // check for expected packet length
+ if (f->segment_count != 1) return error(f, VORBIS_invalid_first_page);
+ if (f->segments[0] != 30) return error(f, VORBIS_invalid_first_page);
+ // read packet
+ // check packet header
+ if (get8(f) != VORBIS_packet_id) return error(f, VORBIS_invalid_first_page);
+ if (!getn(f, header, 6)) return error(f, VORBIS_unexpected_eof);
+ if (!vorbis_validate(header)) return error(f, VORBIS_invalid_first_page);
+ // vorbis_version
+ if (get32(f) != 0) return error(f, VORBIS_invalid_first_page);
+ f->channels = get8(f); if (!f->channels) return error(f, VORBIS_invalid_first_page);
+ if (f->channels > STB_VORBIS_MAX_CHANNELS) return error(f, VORBIS_too_many_channels);
+ f->sample_rate = get32(f); if (!f->sample_rate) return error(f, VORBIS_invalid_first_page);
+ get32(f); // bitrate_maximum
+ get32(f); // bitrate_nominal
+ get32(f); // bitrate_minimum
+ x = get8(f);
+ { int log0,log1;
+ log0 = x & 15;
+ log1 = x >> 4;
+ f->blocksize_0 = 1 << log0;
+ f->blocksize_1 = 1 << log1;
+ if (log0 < 6 || log0 > 13) return error(f, VORBIS_invalid_setup);
+ if (log1 < 6 || log1 > 13) return error(f, VORBIS_invalid_setup);
+ if (log0 > log1) return error(f, VORBIS_invalid_setup);
+ }
+
+ // framing_flag
+ x = get8(f);
+ if (!(x & 1)) return error(f, VORBIS_invalid_first_page);
+
+ // second packet!
+ if (!start_page(f)) return FALSE;
+
+ if (!start_packet(f)) return FALSE;
+ do {
+ len = next_segment(f);
+ skip(f, len);
+ f->bytes_in_seg = 0;
+ } while (len);
+
+ // third packet!
+ if (!start_packet(f)) return FALSE;
+
+ #ifndef STB_VORBIS_NO_PUSHDATA_API
+ if (IS_PUSH_MODE(f)) {
+ if (!is_whole_packet_present(f, TRUE)) {
+ // convert error in ogg header to write type
+ if (f->error == VORBIS_invalid_stream)
+ f->error = VORBIS_invalid_setup;
+ return FALSE;
+ }
+ }
+ #endif
+
+ crc32_init(); // always init it, to avoid multithread race conditions
+
+ if (get8_packet(f) != VORBIS_packet_setup) return error(f, VORBIS_invalid_setup);
+ for (i=0; i < 6; ++i) header[i] = get8_packet(f);
+ if (!vorbis_validate(header)) return error(f, VORBIS_invalid_setup);
+
+ // codebooks
+
+ f->codebook_count = get_bits(f,8) + 1;
+ f->codebooks = (Codebook *) setup_malloc(f, sizeof(*f->codebooks) * f->codebook_count);
+ if (f->codebooks == NULL) return error(f, VORBIS_outofmem);
+ memset(f->codebooks, 0, sizeof(*f->codebooks) * f->codebook_count);
+ for (i=0; i < f->codebook_count; ++i) {
+ uint32 *values;
+ int ordered, sorted_count;
+ int total=0;
+ uint8 *lengths;
+ Codebook *c = f->codebooks+i;
+ x = get_bits(f, 8); if (x != 0x42) return error(f, VORBIS_invalid_setup);
+ x = get_bits(f, 8); if (x != 0x43) return error(f, VORBIS_invalid_setup);
+ x = get_bits(f, 8); if (x != 0x56) return error(f, VORBIS_invalid_setup);
+ x = get_bits(f, 8);
+ c->dimensions = (get_bits(f, 8)<<8) + x;
+ x = get_bits(f, 8);
+ y = get_bits(f, 8);
+ c->entries = (get_bits(f, 8)<<16) + (y<<8) + x;
+ ordered = get_bits(f,1);
+ c->sparse = ordered ? 0 : get_bits(f,1);
+
+ if (c->sparse)
+ lengths = (uint8 *) setup_temp_malloc(f, c->entries);
+ else
+ lengths = c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries);
+
+ if (!lengths) return error(f, VORBIS_outofmem);
+
+ if (ordered) {
+ int current_entry = 0;
+ int current_length = get_bits(f,5) + 1;
+ while (current_entry < c->entries) {
+ int limit = c->entries - current_entry;
+ int n = get_bits(f, ilog(limit));
+ if (current_entry + n > (int) c->entries) { return error(f, VORBIS_invalid_setup); }
+ memset(lengths + current_entry, current_length, n);
+ current_entry += n;
+ ++current_length;
+ }
+ } else {
+ for (j=0; j < c->entries; ++j) {
+ int present = c->sparse ? get_bits(f,1) : 1;
+ if (present) {
+ lengths[j] = get_bits(f, 5) + 1;
+ ++total;
+ } else {
+ lengths[j] = NO_CODE;
+ }
+ }
+ }
+
+ if (c->sparse && total >= c->entries >> 2) {
+ // convert sparse items to non-sparse!
+ if (c->entries > (int) f->setup_temp_memory_required)
+ f->setup_temp_memory_required = c->entries;
+
+ c->codeword_lengths = (uint8 *) setup_malloc(f, c->entries);
+ memcpy(c->codeword_lengths, lengths, c->entries);
+ setup_temp_free(f, lengths, c->entries); // note this is only safe if there have been no intervening temp mallocs!
+ lengths = c->codeword_lengths;
+ c->sparse = 0;
+ }
+
+ // compute the size of the sorted tables
+ if (c->sparse) {
+ sorted_count = total;
+ //assert(total != 0);
+ } else {
+ sorted_count = 0;
+ #ifndef STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH
+ for (j=0; j < c->entries; ++j)
+ if (lengths[j] > STB_VORBIS_FAST_HUFFMAN_LENGTH && lengths[j] != NO_CODE)
+ ++sorted_count;
+ #endif
+ }
+
+ c->sorted_entries = sorted_count;
+ values = NULL;
+
+ if (!c->sparse) {
+ c->codewords = (uint32 *) setup_malloc(f, sizeof(c->codewords[0]) * c->entries);
+ if (!c->codewords) return error(f, VORBIS_outofmem);
+ } else {
+ unsigned int size;
+ if (c->sorted_entries) {
+ c->codeword_lengths = (uint8 *) setup_malloc(f, c->sorted_entries);
+ if (!c->codeword_lengths) return error(f, VORBIS_outofmem);
+ c->codewords = (uint32 *) setup_temp_malloc(f, sizeof(*c->codewords) * c->sorted_entries);
+ if (!c->codewords) return error(f, VORBIS_outofmem);
+ values = (uint32 *) setup_temp_malloc(f, sizeof(*values) * c->sorted_entries);
+ if (!values) return error(f, VORBIS_outofmem);
+ }
+ size = c->entries + (sizeof(*c->codewords) + sizeof(*values)) * c->sorted_entries;
+ if (size > f->setup_temp_memory_required)
+ f->setup_temp_memory_required = size;
+ }
+
+ if (!compute_codewords(c, lengths, c->entries, values)) {
+ if (c->sparse) setup_temp_free(f, values, 0);
+ return error(f, VORBIS_invalid_setup);
+ }
+
+ if (c->sorted_entries) {
+ // allocate an extra slot for sentinels
+ c->sorted_codewords = (uint32 *) setup_malloc(f, sizeof(*c->sorted_codewords) * (c->sorted_entries+1));
+ // allocate an extra slot at the front so that c->sorted_values[-1] is defined
+ // so that we can catch that case without an extra if
+ c->sorted_values = ( int *) setup_malloc(f, sizeof(*c->sorted_values ) * (c->sorted_entries+1));
+ if (c->sorted_values) { ++c->sorted_values; c->sorted_values[-1] = -1; }
+ compute_sorted_huffman(c, lengths, values);
+ }
+
+ if (c->sparse) {
+ setup_temp_free(f, values, sizeof(*values)*c->sorted_entries);
+ setup_temp_free(f, c->codewords, sizeof(*c->codewords)*c->sorted_entries);
+ setup_temp_free(f, lengths, c->entries);
+ c->codewords = NULL;
+ }
+
+ compute_accelerated_huffman(c);
+
+ c->lookup_type = get_bits(f, 4);
+ if (c->lookup_type > 2) return error(f, VORBIS_invalid_setup);
+ if (c->lookup_type > 0) {
+ uint16 *mults;
+ c->minimum_value = float32_unpack(get_bits(f, 32));
+ c->delta_value = float32_unpack(get_bits(f, 32));
+ c->value_bits = get_bits(f, 4)+1;
+ c->sequence_p = get_bits(f,1);
+ if (c->lookup_type == 1) {
+ c->lookup_values = lookup1_values(c->entries, c->dimensions);
+ } else {
+ c->lookup_values = c->entries * c->dimensions;
+ }
+ mults = (uint16 *) setup_temp_malloc(f, sizeof(mults[0]) * c->lookup_values);
+ if (mults == NULL) return error(f, VORBIS_outofmem);
+ for (j=0; j < (int) c->lookup_values; ++j) {
+ int q = get_bits(f, c->value_bits);
+ if (q == EOP) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_invalid_setup); }
+ mults[j] = q;
+ }
+
+#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK
+ if (c->lookup_type == 1) {
+ int len, sparse = c->sparse;
+ // pre-expand the lookup1-style multiplicands, to avoid a divide in the inner loop
+ if (sparse) {
+ if (c->sorted_entries == 0) goto skip;
+ c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->sorted_entries * c->dimensions);
+ } else
+ c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->entries * c->dimensions);
+ if (c->multiplicands == NULL) { setup_temp_free(f,mults,sizeof(mults[0])*c->lookup_values); return error(f, VORBIS_outofmem); }
+ len = sparse ? c->sorted_entries : c->entries;
+ for (j=0; j < len; ++j) {
+ int z = sparse ? c->sorted_values[j] : j, div=1;
+ for (k=0; k < c->dimensions; ++k) {
+ int off = (z / div) % c->lookup_values;
+ c->multiplicands[j*c->dimensions + k] =
+ #ifndef STB_VORBIS_CODEBOOK_FLOATS
+ mults[off];
+ #else
+ mults[off]*c->delta_value + c->minimum_value;
+ // in this case (and this case only) we could pre-expand c->sequence_p,
+ // and throw away the decode logic for it; have to ALSO do
+ // it in the case below, but it can only be done if
+ // STB_VORBIS_CODEBOOK_FLOATS
+ // !STB_VORBIS_DIVIDES_IN_CODEBOOK
+ #endif
+ div *= c->lookup_values;
+ }
+ }
+ setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values);
+ c->lookup_type = 2;
+ }
+ else
+#endif
+ {
+ c->multiplicands = (codetype *) setup_malloc(f, sizeof(c->multiplicands[0]) * c->lookup_values);
+ #ifndef STB_VORBIS_CODEBOOK_FLOATS
+ memcpy(c->multiplicands, mults, sizeof(c->multiplicands[0]) * c->lookup_values);
+ #else
+ for (j=0; j < (int) c->lookup_values; ++j)
+ c->multiplicands[j] = mults[j] * c->delta_value + c->minimum_value;
+ #endif
+ setup_temp_free(f, mults,sizeof(mults[0])*c->lookup_values);
+ }
+ skip:;
+
+ #ifdef STB_VORBIS_CODEBOOK_FLOATS
+ if (c->lookup_type == 2 && c->sequence_p) {
+ for (j=1; j < (int) c->lookup_values; ++j)
+ c->multiplicands[j] = c->multiplicands[j-1];
+ c->sequence_p = 0;
+ }
+ #endif
+ }
+ }
+
+ // time domain transfers (notused)
+
+ x = get_bits(f, 6) + 1;
+ for (i=0; i < x; ++i) {
+ uint32 z = get_bits(f, 16);
+ if (z != 0) return error(f, VORBIS_invalid_setup);
+ }
+
+ // Floors
+ f->floor_count = get_bits(f, 6)+1;
+ f->floor_config = (Floor *) setup_malloc(f, f->floor_count * sizeof(*f->floor_config));
+ for (i=0; i < f->floor_count; ++i) {
+ f->floor_types[i] = get_bits(f, 16);
+ if (f->floor_types[i] > 1) return error(f, VORBIS_invalid_setup);
+ if (f->floor_types[i] == 0) {
+ Floor0 *g = &f->floor_config[i].floor0;
+ g->order = get_bits(f,8);
+ g->rate = get_bits(f,16);
+ g->bark_map_size = get_bits(f,16);
+ g->amplitude_bits = get_bits(f,6);
+ g->amplitude_offset = get_bits(f,8);
+ g->number_of_books = get_bits(f,4) + 1;
+ for (j=0; j < g->number_of_books; ++j)
+ g->book_list[j] = get_bits(f,8);
+ return error(f, VORBIS_feature_not_supported);
+ } else {
+ Point p[31*8+2];
+ Floor1 *g = &f->floor_config[i].floor1;
+ int max_class = -1;
+ g->partitions = get_bits(f, 5);
+ for (j=0; j < g->partitions; ++j) {
+ g->partition_class_list[j] = get_bits(f, 4);
+ if (g->partition_class_list[j] > max_class)
+ max_class = g->partition_class_list[j];
+ }
+ for (j=0; j <= max_class; ++j) {
+ g->class_dimensions[j] = get_bits(f, 3)+1;
+ g->class_subclasses[j] = get_bits(f, 2);
+ if (g->class_subclasses[j]) {
+ g->class_masterbooks[j] = get_bits(f, 8);
+ if (g->class_masterbooks[j] >= f->codebook_count) return error(f, VORBIS_invalid_setup);
+ }
+ for (k=0; k < 1 << g->class_subclasses[j]; ++k) {
+ g->subclass_books[j][k] = get_bits(f,8)-1;
+ if (g->subclass_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup);
+ }
+ }
+ g->floor1_multiplier = get_bits(f,2)+1;
+ g->rangebits = get_bits(f,4);
+ g->Xlist[0] = 0;
+ g->Xlist[1] = 1 << g->rangebits;
+ g->values = 2;
+ for (j=0; j < g->partitions; ++j) {
+ int c = g->partition_class_list[j];
+ for (k=0; k < g->class_dimensions[c]; ++k) {
+ g->Xlist[g->values] = get_bits(f, g->rangebits);
+ ++g->values;
+ }
+ }
+ // precompute the sorting
+ for (j=0; j < g->values; ++j) {
+ p[j].x = g->Xlist[j];
+ p[j].y = j;
+ }
+ qsort(p, g->values, sizeof(p[0]), point_compare);
+ for (j=0; j < g->values; ++j)
+ g->sorted_order[j] = (uint8) p[j].y;
+ // precompute the neighbors
+ for (j=2; j < g->values; ++j) {
+ int low,hi;
+ neighbors(g->Xlist, j, &low,&hi);
+ g->neighbors[j][0] = low;
+ g->neighbors[j][1] = hi;
+ }
+
+ if (g->values > longest_floorlist)
+ longest_floorlist = g->values;
+ }
+ }
+
+ // Residue
+ f->residue_count = get_bits(f, 6)+1;
+ f->residue_config = (Residue *) setup_malloc(f, f->residue_count * sizeof(*f->residue_config));
+ for (i=0; i < f->residue_count; ++i) {
+ uint8 residue_cascade[64];
+ Residue *r = f->residue_config+i;
+ f->residue_types[i] = get_bits(f, 16);
+ if (f->residue_types[i] > 2) return error(f, VORBIS_invalid_setup);
+ r->begin = get_bits(f, 24);
+ r->end = get_bits(f, 24);
+ r->part_size = get_bits(f,24)+1;
+ r->classifications = get_bits(f,6)+1;
+ r->classbook = get_bits(f,8);
+ for (j=0; j < r->classifications; ++j) {
+ uint8 high_bits=0;
+ uint8 low_bits=get_bits(f,3);
+ if (get_bits(f,1))
+ high_bits = get_bits(f,5);
+ residue_cascade[j] = high_bits*8 + low_bits;
+ }
+ r->residue_books = (short (*)[8]) setup_malloc(f, sizeof(r->residue_books[0]) * r->classifications);
+ for (j=0; j < r->classifications; ++j) {
+ for (k=0; k < 8; ++k) {
+ if (residue_cascade[j] & (1 << k)) {
+ r->residue_books[j][k] = get_bits(f, 8);
+ if (r->residue_books[j][k] >= f->codebook_count) return error(f, VORBIS_invalid_setup);
+ } else {
+ r->residue_books[j][k] = -1;
+ }
+ }
+ }
+ // precompute the classifications[] array to avoid inner-loop mod/divide
+ // call it 'classdata' since we already have r->classifications
+ r->classdata = (uint8 **) setup_malloc(f, sizeof(*r->classdata) * f->codebooks[r->classbook].entries);
+ if (!r->classdata) return error(f, VORBIS_outofmem);
+ memset(r->classdata, 0, sizeof(*r->classdata) * f->codebooks[r->classbook].entries);
+ for (j=0; j < f->codebooks[r->classbook].entries; ++j) {
+ int classwords = f->codebooks[r->classbook].dimensions;
+ int temp = j;
+ r->classdata[j] = (uint8 *) setup_malloc(f, sizeof(r->classdata[j][0]) * classwords);
+ for (k=classwords-1; k >= 0; --k) {
+ r->classdata[j][k] = temp % r->classifications;
+ temp /= r->classifications;
+ }
+ }
+ }
+
+ f->mapping_count = get_bits(f,6)+1;
+ f->mapping = (Mapping *) setup_malloc(f, f->mapping_count * sizeof(*f->mapping));
+ for (i=0; i < f->mapping_count; ++i) {
+ Mapping *m = f->mapping + i;
+ int mapping_type = get_bits(f,16);
+ if (mapping_type != 0) return error(f, VORBIS_invalid_setup);
+ m->chan = (MappingChannel *) setup_malloc(f, f->channels * sizeof(*m->chan));
+ if (get_bits(f,1))
+ m->submaps = get_bits(f,4);
+ else
+ m->submaps = 1;
+ if (m->submaps > max_submaps)
+ max_submaps = m->submaps;
+ if (get_bits(f,1)) {
+ m->coupling_steps = get_bits(f,8)+1;
+ for (k=0; k < m->coupling_steps; ++k) {
+ m->chan[k].magnitude = get_bits(f, ilog(f->channels)-1);
+ m->chan[k].angle = get_bits(f, ilog(f->channels)-1);
+ if (m->chan[k].magnitude >= f->channels) return error(f, VORBIS_invalid_setup);
+ if (m->chan[k].angle >= f->channels) return error(f, VORBIS_invalid_setup);
+ if (m->chan[k].magnitude == m->chan[k].angle) return error(f, VORBIS_invalid_setup);
+ }
+ } else
+ m->coupling_steps = 0;
+
+ // reserved field
+ if (get_bits(f,2)) return error(f, VORBIS_invalid_setup);
+ if (m->submaps > 1) {
+ for (j=0; j < f->channels; ++j) {
+ m->chan[j].mux = get_bits(f, 4);
+ if (m->chan[j].mux >= m->submaps) return error(f, VORBIS_invalid_setup);
+ }
+ } else
+ // @SPECIFICATION: this case is missing from the spec
+ for (j=0; j < f->channels; ++j)
+ m->chan[j].mux = 0;
+
+ for (j=0; j < m->submaps; ++j) {
+ get_bits(f,8); // discard
+ m->submap_floor[j] = get_bits(f,8);
+ m->submap_residue[j] = get_bits(f,8);
+ if (m->submap_floor[j] >= f->floor_count) return error(f, VORBIS_invalid_setup);
+ if (m->submap_residue[j] >= f->residue_count) return error(f, VORBIS_invalid_setup);
+ }
+ }
+
+ // Modes
+ f->mode_count = get_bits(f, 6)+1;
+ for (i=0; i < f->mode_count; ++i) {
+ Mode *m = f->mode_config+i;
+ m->blockflag = get_bits(f,1);
+ m->windowtype = get_bits(f,16);
+ m->transformtype = get_bits(f,16);
+ m->mapping = get_bits(f,8);
+ if (m->windowtype != 0) return error(f, VORBIS_invalid_setup);
+ if (m->transformtype != 0) return error(f, VORBIS_invalid_setup);
+ if (m->mapping >= f->mapping_count) return error(f, VORBIS_invalid_setup);
+ }
+
+ flush_packet(f);
+
+ f->previous_length = 0;
+
+ for (i=0; i < f->channels; ++i) {
+ f->channel_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1);
+ f->previous_window[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2);
+ f->finalY[i] = (int16 *) setup_malloc(f, sizeof(int16) * longest_floorlist);
+ #ifdef STB_VORBIS_NO_DEFER_FLOOR
+ f->floor_buffers[i] = (float *) setup_malloc(f, sizeof(float) * f->blocksize_1/2);
+ #endif
+ }
+
+ if (!init_blocksize(f, 0, f->blocksize_0)) return FALSE;
+ if (!init_blocksize(f, 1, f->blocksize_1)) return FALSE;
+ f->blocksize[0] = f->blocksize_0;
+ f->blocksize[1] = f->blocksize_1;
+
+#ifdef STB_VORBIS_DIVIDE_TABLE
+ if (integer_divide_table[1][1]==0)
+ for (i=0; i < DIVTAB_NUMER; ++i)
+ for (j=1; j < DIVTAB_DENOM; ++j)
+ integer_divide_table[i][j] = i / j;
+#endif
+
+ // compute how much temporary memory is needed
+
+ // 1.
+ {
+ uint32 imdct_mem = (f->blocksize_1 * sizeof(float) >> 1);
+ uint32 classify_mem;
+ int i,max_part_read=0;
+ for (i=0; i < f->residue_count; ++i) {
+ Residue *r = f->residue_config + i;
+ int n_read = r->end - r->begin;
+ int part_read = n_read / r->part_size;
+ if (part_read > max_part_read)
+ max_part_read = part_read;
+ }
+ #ifndef STB_VORBIS_DIVIDES_IN_RESIDUE
+ classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(uint8 *));
+ #else
+ classify_mem = f->channels * (sizeof(void*) + max_part_read * sizeof(int *));
+ #endif
+
+ f->temp_memory_required = classify_mem;
+ if (imdct_mem > f->temp_memory_required)
+ f->temp_memory_required = imdct_mem;
+ }
+
+ f->first_decode = TRUE;
+
+ if (f->alloc.alloc_buffer) {
+ assert(f->temp_offset == f->alloc.alloc_buffer_length_in_bytes);
+ // check if there's enough temp memory so we don't error later
+ if (f->setup_offset + sizeof(*f) + f->temp_memory_required > (unsigned) f->temp_offset)
+ return error(f, VORBIS_outofmem);
+ }
+
+ f->first_audio_page_offset = stb_vorbis_get_file_offset(f);
+
+ return TRUE;
+}
+
+static void vorbis_deinit(stb_vorbis *p)
+{
+ int i,j;
+ for (i=0; i < p->residue_count; ++i) {
+ Residue *r = p->residue_config+i;
+ if (r->classdata) {
+ for (j=0; j < p->codebooks[r->classbook].entries; ++j)
+ setup_free(p, r->classdata[j]);
+ setup_free(p, r->classdata);
+ }
+ setup_free(p, r->residue_books);
+ }
+
+ if (p->codebooks) {
+ for (i=0; i < p->codebook_count; ++i) {
+ Codebook *c = p->codebooks + i;
+ setup_free(p, c->codeword_lengths);
+ setup_free(p, c->multiplicands);
+ setup_free(p, c->codewords);
+ setup_free(p, c->sorted_codewords);
+ // c->sorted_values[-1] is the first entry in the array
+ setup_free(p, c->sorted_values ? c->sorted_values-1 : NULL);
+ }
+ setup_free(p, p->codebooks);
+ }
+ setup_free(p, p->floor_config);
+ setup_free(p, p->residue_config);
+ for (i=0; i < p->mapping_count; ++i)
+ setup_free(p, p->mapping[i].chan);
+ setup_free(p, p->mapping);
+ for (i=0; i < p->channels; ++i) {
+ setup_free(p, p->channel_buffers[i]);
+ setup_free(p, p->previous_window[i]);
+ #ifdef STB_VORBIS_NO_DEFER_FLOOR
+ setup_free(p, p->floor_buffers[i]);
+ #endif
+ setup_free(p, p->finalY[i]);
+ }
+ for (i=0; i < 2; ++i) {
+ setup_free(p, p->A[i]);
+ setup_free(p, p->B[i]);
+ setup_free(p, p->C[i]);
+ setup_free(p, p->window[i]);
setup_free(p, p->bit_reverse[i]);
- } - #ifndef STB_VORBIS_NO_STDIO - if (p->close_on_free) fclose(p->f); - #endif -} - -void stb_vorbis_close(stb_vorbis *p) -{ - if (p == NULL) return; - vorbis_deinit(p); - setup_free(p,p); -} - -static void vorbis_init(stb_vorbis *p, stb_vorbis_alloc *z) -{ - memset(p, 0, sizeof(*p)); // NULL out all malloc'd pointers to start - if (z) { - p->alloc = *z; - p->alloc.alloc_buffer_length_in_bytes = (p->alloc.alloc_buffer_length_in_bytes+3) & ~3; - p->temp_offset = p->alloc.alloc_buffer_length_in_bytes; - } - p->eof = 0; - p->error = VORBIS__no_error; - p->stream = NULL; - p->codebooks = NULL; - p->page_crc_tests = -1; - #ifndef STB_VORBIS_NO_STDIO - p->close_on_free = FALSE; - p->f = NULL; - #endif -} - -int stb_vorbis_get_sample_offset(stb_vorbis *f) -{ - if (f->current_loc_valid) - return f->current_loc; - else - return -1; -} - -stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f) -{ - stb_vorbis_info d; - d.channels = f->channels; - d.sample_rate = f->sample_rate; - d.setup_memory_required = f->setup_memory_required; - d.setup_temp_memory_required = f->setup_temp_memory_required; - d.temp_memory_required = f->temp_memory_required; - d.max_frame_size = f->blocksize_1 >> 1; - return d; -} - -int stb_vorbis_get_error(stb_vorbis *f) -{ - int e = f->error; - f->error = VORBIS__no_error; - return e; -} - -static stb_vorbis * vorbis_alloc(stb_vorbis *f) -{ - stb_vorbis *p = (stb_vorbis *) setup_malloc(f, sizeof(*p)); - return p; -} - -#ifndef STB_VORBIS_NO_PUSHDATA_API - -void stb_vorbis_flush_pushdata(stb_vorbis *f) -{ - f->previous_length = 0; - f->page_crc_tests = 0; - f->discard_samples_deferred = 0; - f->current_loc_valid = FALSE; - f->first_decode = FALSE; - f->samples_output = 0; - f->channel_buffer_start = 0; - f->channel_buffer_end = 0; -} - -static int vorbis_search_for_page_pushdata(vorb *f, uint8 *data, int data_len) -{ - int i,n; - for (i=0; i < f->page_crc_tests; ++i) - f->scan[i].bytes_done = 0; - - // if we have room for more scans, search for them first, because - // they may cause us to stop early if their header is incomplete - if (f->page_crc_tests < STB_VORBIS_PUSHDATA_CRC_COUNT) { - if (data_len < 4) return 0; - data_len -= 3; // need to look for 4-byte sequence, so don't miss - // one that straddles a boundary - for (i=0; i < data_len; ++i) { - if (data[i] == 0x4f) { - if (0==memcmp(data+i, ogg_page_header, 4)) { - int j,len; - uint32 crc; - // make sure we have the whole page header - if (i+26 >= data_len || i+27+data[i+26] >= data_len) { - // only read up to this page start, so hopefully we'll - // have the whole page header start next time - data_len = i; - break; - } - // ok, we have it all; compute the length of the page - len = 27 + data[i+26]; - for (j=0; j < data[i+26]; ++j) - len += data[i+27+j]; - // scan everything up to the embedded crc (which we must 0) - crc = 0; - for (j=0; j < 22; ++j) - crc = crc32_update(crc, data[i+j]); - // now process 4 0-bytes - for ( ; j < 26; ++j) - crc = crc32_update(crc, 0); - // len is the total number of bytes we need to scan - n = f->page_crc_tests++; - f->scan[n].bytes_left = len-j; - f->scan[n].crc_so_far = crc; - f->scan[n].goal_crc = data[i+22] + (data[i+23] << 8) + (data[i+24]<<16) + (data[i+25]<<24); - // if the last frame on a page is continued to the next, then - // we can't recover the sample_loc immediately - if (data[i+27+data[i+26]-1] == 255) - f->scan[n].sample_loc = ~0; - else - f->scan[n].sample_loc = data[i+6] + (data[i+7] << 8) + (data[i+ 8]<<16) + (data[i+ 9]<<24); - f->scan[n].bytes_done = i+j; - if (f->page_crc_tests == STB_VORBIS_PUSHDATA_CRC_COUNT) - break; - // keep going if we still have room for more - } - } - } - } - - for (i=0; i < f->page_crc_tests;) { - uint32 crc; - int j; - int n = f->scan[i].bytes_done; - int m = f->scan[i].bytes_left; - if (m > data_len - n) m = data_len - n; - // m is the bytes to scan in the current chunk - crc = f->scan[i].crc_so_far; - for (j=0; j < m; ++j) - crc = crc32_update(crc, data[n+j]); - f->scan[i].bytes_left -= m; - f->scan[i].crc_so_far = crc; - if (f->scan[i].bytes_left == 0) { - // does it match? - if (f->scan[i].crc_so_far == f->scan[i].goal_crc) { - // Houston, we have page - data_len = n+m; // consumption amount is wherever that scan ended - f->page_crc_tests = -1; // drop out of page scan mode - f->previous_length = 0; // decode-but-don't-output one frame - f->next_seg = -1; // start a new page - f->current_loc = f->scan[i].sample_loc; // set the current sample location - // to the amount we'd have decoded had we decoded this page - f->current_loc_valid = f->current_loc != ~0; - return data_len; - } - // delete entry - f->scan[i] = f->scan[--f->page_crc_tests]; - } else { - ++i; - } - } - - return data_len; -} - -// return value: number of bytes we used -int stb_vorbis_decode_frame_pushdata( - stb_vorbis *f, // the file we're decoding - uint8 *data, int data_len, // the memory available for decoding - int *channels, // place to write number of float * buffers - float ***output, // place to write float ** array of float * buffers - int *samples // place to write number of output samples - ) -{ - int i; - int len,right,left; - - if (!IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); - - if (f->page_crc_tests >= 0) { - *samples = 0; - return vorbis_search_for_page_pushdata(f, data, data_len); - } - - f->stream = data; - f->stream_end = data + data_len; - f->error = VORBIS__no_error; - - // check that we have the entire packet in memory - if (!is_whole_packet_present(f, FALSE)) { - *samples = 0; - return 0; - } - - if (!vorbis_decode_packet(f, &len, &left, &right)) { - // save the actual error we encountered - enum STBVorbisError error = f->error; - if (error == VORBIS_bad_packet_type) { - // flush and resynch - f->error = VORBIS__no_error; - while (get8_packet(f) != EOP) - if (f->eof) break; - *samples = 0; - return f->stream - data; - } - if (error == VORBIS_continued_packet_flag_invalid) { - if (f->previous_length == 0) { - // we may be resynching, in which case it's ok to hit one - // of these; just discard the packet - f->error = VORBIS__no_error; - while (get8_packet(f) != EOP) - if (f->eof) break; - *samples = 0; - return f->stream - data; - } - } - // if we get an error while parsing, what to do? - // well, it DEFINITELY won't work to continue from where we are! - stb_vorbis_flush_pushdata(f); - // restore the error that actually made us bail - f->error = error; - *samples = 0; - return 1; - } - - // success! - len = vorbis_finish_frame(f, len, left, right); - for (i=0; i < f->channels; ++i) - f->outputs[i] = f->channel_buffers[i] + left; - - if (channels) *channels = f->channels; - *samples = len; - *output = f->outputs; - return f->stream - data; -} - -stb_vorbis *stb_vorbis_open_pushdata( - unsigned char *data, int data_len, // the memory available for decoding - int *data_used, // only defined if result is not NULL - int *error, stb_vorbis_alloc *alloc) -{ - stb_vorbis *f, p; - vorbis_init(&p, alloc); - p.stream = data; - p.stream_end = data + data_len; - p.push_mode = TRUE; - if (!start_decoder(&p)) { - if (p.eof) - *error = VORBIS_need_more_data; - else - *error = p.error; - return NULL; - } - f = vorbis_alloc(&p); - if (f) { - *f = p; - *data_used = f->stream - data; - *error = 0; - return f; - } else { - vorbis_deinit(&p); - return NULL; - } -} -#endif // STB_VORBIS_NO_PUSHDATA_API - -unsigned int stb_vorbis_get_file_offset(stb_vorbis *f) -{ - #ifndef STB_VORBIS_NO_PUSHDATA_API - if (f->push_mode) return 0; - #endif - if (USE_MEMORY(f)) return f->stream - f->stream_start; - #ifndef STB_VORBIS_NO_STDIO - return ftell(f->f) - f->f_start; - #endif -} - -#ifndef STB_VORBIS_NO_PULLDATA_API -// -// DATA-PULLING API -// - -static uint32 vorbis_find_page(stb_vorbis *f, uint32 *end, uint32 *last) -{ - for(;;) { - int n; - if (f->eof) return 0; - n = get8(f); - if (n == 0x4f) { // page header - unsigned int retry_loc = stb_vorbis_get_file_offset(f); - int i; - // check if we're off the end of a file_section stream - if (retry_loc - 25 > f->stream_len) - return 0; - // check the rest of the header - for (i=1; i < 4; ++i) - if (get8(f) != ogg_page_header[i]) - break; - if (f->eof) return 0; - if (i == 4) { - uint8 header[27]; - uint32 i, crc, goal, len; - for (i=0; i < 4; ++i) - header[i] = ogg_page_header[i]; - for (; i < 27; ++i) - header[i] = get8(f); - if (f->eof) return 0; - if (header[4] != 0) goto invalid; - goal = header[22] + (header[23] << 8) + (header[24]<<16) + (header[25]<<24); - for (i=22; i < 26; ++i) - header[i] = 0; - crc = 0; - for (i=0; i < 27; ++i) - crc = crc32_update(crc, header[i]); - len = 0; - for (i=0; i < header[26]; ++i) { - int s = get8(f); - crc = crc32_update(crc, s); - len += s; - } - if (len && f->eof) return 0; - for (i=0; i < len; ++i) - crc = crc32_update(crc, get8(f)); - // finished parsing probable page - if (crc == goal) { - // we could now check that it's either got the last - // page flag set, OR it's followed by the capture - // pattern, but I guess TECHNICALLY you could have - // a file with garbage between each ogg page and recover - // from it automatically? So even though that paranoia - // might decrease the chance of an invalid decode by - // another 2^32, not worth it since it would hose those - // invalid-but-useful files? - if (end) - *end = stb_vorbis_get_file_offset(f); - if (last) - if (header[5] & 0x04) - *last = 1; - else - *last = 0; - set_file_offset(f, retry_loc-1); - return 1; - } - } - invalid: - // not a valid page, so rewind and look for next one - set_file_offset(f, retry_loc); - } - } -} - -// seek is implemented with 'interpolation search'--this is like -// binary search, but we use the data values to estimate the likely -// location of the data item (plus a bit of a bias so when the -// estimation is wrong we don't waste overly much time) - -#define SAMPLE_unknown 0xffffffff - - -// ogg vorbis, in its insane infinite wisdom, only provides -// information about the sample at the END of the page. -// therefore we COULD have the data we need in the current -// page, and not know it. we could just use the end location -// as our only knowledge for bounds, seek back, and eventually -// the binary search finds it. or we can try to be smart and -// not waste time trying to locate more pages. we try to be -// smart, since this data is already in memory anyway, so -// doing needless I/O would be crazy! -static int vorbis_analyze_page(stb_vorbis *f, ProbedPage *z) -{ - uint8 header[27], lacing[255]; - uint8 packet_type[255]; - int num_packet, packet_start, previous =0; - int i,len; - uint32 samples; - - // record where the page starts - z->page_start = stb_vorbis_get_file_offset(f); - - // parse the header - getn(f, header, 27); - assert(header[0] == 'O' && header[1] == 'g' && header[2] == 'g' && header[3] == 'S'); - getn(f, lacing, header[26]); - - // determine the length of the payload - len = 0; - for (i=0; i < header[26]; ++i) - len += lacing[i]; - - // this implies where the page ends - z->page_end = z->page_start + 27 + header[26] + len; - - // read the last-decoded sample out of the data - z->last_decoded_sample = header[6] + (header[7] << 8) + (header[8] << 16) + (header[9] << 16); - - if (header[5] & 4) { - // if this is the last page, it's not possible to work - // backwards to figure out the first sample! whoops! fuck. - z->first_decoded_sample = SAMPLE_unknown; - set_file_offset(f, z->page_start); - return 1; - } - - // scan through the frames to determine the sample-count of each one... - // our goal is the sample # of the first fully-decoded sample on the - // page, which is the first decoded sample of the 2nd page - - num_packet=0; - - packet_start = ((header[5] & 1) == 0); - - for (i=0; i < header[26]; ++i) { - if (packet_start) { - uint8 n,b,m; - if (lacing[i] == 0) goto bail; // trying to read from zero-length packet - n = get8(f); - // if bottom bit is non-zero, we've got corruption - if (n & 1) goto bail; - n >>= 1; - b = ilog(f->mode_count-1); - m = n >> b; - n &= (1 << b)-1; - if (n >= f->mode_count) goto bail; - if (num_packet == 0 && f->mode_config[n].blockflag) - previous = (m & 1); - packet_type[num_packet++] = f->mode_config[n].blockflag; - skip(f, lacing[i]-1); - } else - skip(f, lacing[i]); - packet_start = (lacing[i] < 255); - } - - // now that we know the sizes of all the pages, we can start determining - // how much sample data there is. - - samples = 0; - - // for the last packet, we step by its whole length, because the definition - // is that we encoded the end sample loc of the 'last packet completed', - // where 'completed' refers to packets being split, and we are left to guess - // what 'end sample loc' means. we assume it means ignoring the fact that - // the last half of the data is useless without windowing against the next - // packet... (so it's not REALLY complete in that sense) - if (num_packet > 1) - samples += f->blocksize[packet_type[num_packet-1]]; - - for (i=num_packet-2; i >= 1; --i) { - // now, for this packet, how many samples do we have that - // do not overlap the following packet? - if (packet_type[i] == 1) - if (packet_type[i+1] == 1) - samples += f->blocksize_1 >> 1; - else - samples += ((f->blocksize_1 - f->blocksize_0) >> 2) + (f->blocksize_0 >> 1); - else - samples += f->blocksize_0 >> 1; - } - // now, at this point, we've rewound to the very beginning of the - // _second_ packet. if we entirely discard the first packet after - // a seek, this will be exactly the right sample number. HOWEVER! - // we can't as easily compute this number for the LAST page. The - // only way to get the sample offset of the LAST page is to use - // the end loc from the previous page. But what that returns us - // is _exactly_ the place where we get our first non-overlapped - // sample. (I think. Stupid spec for being ambiguous.) So for - // consistency it's better to do that here, too. However, that - // will then require us to NOT discard all of the first frame we - // decode, in some cases, which means an even weirder frame size - // and extra code. what a fucking pain. - - // we're going to discard the first packet if we - // start the seek here, so we don't care about it. (we could actually - // do better; if the first packet is long, and the previous packet - // is short, there's actually data in the first half of the first - // packet that doesn't need discarding... but not worth paying the - // effort of tracking that of that here and in the seeking logic) - // except crap, if we infer it from the _previous_ packet's end - // location, we DO need to use that definition... and we HAVE to - // infer the start loc of the LAST packet from the previous packet's - // end location. fuck you, ogg vorbis. - - z->first_decoded_sample = z->last_decoded_sample - samples; - - // restore file state to where we were - set_file_offset(f, z->page_start); - return 1; - - // restore file state to where we were - bail: - set_file_offset(f, z->page_start); - return 0; -} - -static int vorbis_seek_frame_from_page(stb_vorbis *f, uint32 page_start, uint32 first_sample, uint32 target_sample, int fine) -{ - int left_start, left_end, right_start, right_end, mode,i; - int frame=0; - uint32 frame_start; - int frames_to_skip, data_to_skip; - - // first_sample is the sample # of the first sample that doesn't - // overlap the previous page... note that this requires us to - // _partially_ discard the first packet! bleh. - set_file_offset(f, page_start); - - f->next_seg = -1; // force page resync - - frame_start = first_sample; - // frame start is where the previous packet's last decoded sample - // was, which corresponds to left_end... EXCEPT if the previous - // packet was long and this packet is short? Probably a bug here. - - - // now, we can start decoding frames... we'll only FAKE decode them, - // until we find the frame that contains our sample; then we'll rewind, - // and try again - for (;;) { - int start; - - if (!vorbis_decode_initial(f, &left_start, &left_end, &right_start, &right_end, &mode)) - return error(f, VORBIS_seek_failed); - - if (frame == 0) - start = left_end; - else - start = left_start; - - // the window starts at left_start; the last valid sample we generate - // before the next frame's window start is right_start-1 - if (target_sample < frame_start + right_start-start) - break; - - flush_packet(f); - if (f->eof) - return error(f, VORBIS_seek_failed); - - frame_start += right_start - start; - - ++frame; - } - - // ok, at this point, the sample we want is contained in frame #'frame' - - // to decode frame #'frame' normally, we have to decode the - // previous frame first... but if it's the FIRST frame of the page - // we can't. if it's the first frame, it means it falls in the part - // of the first frame that doesn't overlap either of the other frames. - // so, if we have to handle that case for the first frame, we might - // as well handle it for all of them, so: - if (target_sample > frame_start + (left_end - left_start)) { - // so what we want to do is go ahead and just immediately decode - // this frame, but then make it so the next get_frame_float() uses - // this already-decoded data? or do we want to go ahead and rewind, - // and leave a flag saying to skip the first N data? let's do that - frames_to_skip = frame; // if this is frame #1, skip 1 frame (#0) - data_to_skip = left_end - left_start; - } else { - // otherwise, we want to skip frames 0, 1, 2, ... frame-2 - // (which means frame-2+1 total frames) then decode frame-1, - // then leave frame pending - frames_to_skip = frame - 1; - assert(frames_to_skip >= 0); - data_to_skip = -1; - } - - set_file_offset(f, page_start); - f->next_seg = - 1; // force page resync - - for (i=0; i < frames_to_skip; ++i) { - maybe_start_packet(f); - flush_packet(f); - } - - if (data_to_skip >= 0) { - int i,j,n = f->blocksize_0 >> 1; - f->discard_samples_deferred = data_to_skip; - for (i=0; i < f->channels; ++i) - for (j=0; j < n; ++j) - f->previous_window[i][j] = 0; - f->previous_length = n; - frame_start += data_to_skip; - } else { - f->previous_length = 0; - vorbis_pump_first_frame(f); - } - - // at this point, the NEXT decoded frame will generate the desired sample - if (fine) { - // so if we're doing sample accurate streaming, we want to go ahead and decode it! - if (target_sample != frame_start) { - int n; - stb_vorbis_get_frame_float(f, &n, NULL); - assert(target_sample > frame_start); - assert(f->channel_buffer_start + (int) (target_sample-frame_start) < f->channel_buffer_end); - f->channel_buffer_start += (target_sample - frame_start); - } - } - - return 0; -} - -static int vorbis_seek_base(stb_vorbis *f, unsigned int sample_number, int fine) -{ - ProbedPage p[2],q; - if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); - - // do we know the location of the last page? - if (f->p_last.page_start == 0) { - uint32 z = stb_vorbis_stream_length_in_samples(f); - if (z == 0) return error(f, VORBIS_cant_find_last_page); - } - - p[0] = f->p_first; - p[1] = f->p_last; - - if (sample_number >= f->p_last.last_decoded_sample) - sample_number = f->p_last.last_decoded_sample-1; - - if (sample_number < f->p_first.last_decoded_sample) { - vorbis_seek_frame_from_page(f, p[0].page_start, 0, sample_number, fine); - return 0; - } else { - int attempts=0; - while (p[0].page_end < p[1].page_start) { - uint32 probe; - uint32 start_offset, end_offset; - uint32 start_sample, end_sample; - - // copy these into local variables so we can tweak them - // if any are unknown - start_offset = p[0].page_end; - end_offset = p[1].after_previous_page_start; // an address known to seek to page p[1] - start_sample = p[0].last_decoded_sample; - end_sample = p[1].last_decoded_sample; - - // currently there is no such tweaking logic needed/possible? - if (start_sample == SAMPLE_unknown || end_sample == SAMPLE_unknown) - return error(f, VORBIS_seek_failed); - - // now we want to lerp between these for the target samples... - - // step 1: we need to bias towards the page start... - if (start_offset + 4000 < end_offset) - end_offset -= 4000; - - // now compute an interpolated search loc - probe = start_offset + (int) floor((float) (end_offset - start_offset) / (end_sample - start_sample) * (sample_number - start_sample)); - - // next we need to bias towards binary search... - // code is a little wonky to allow for full 32-bit unsigned values - if (attempts >= 4) { - uint32 probe2 = start_offset + ((end_offset - start_offset) >> 1); - if (attempts >= 8) - probe = probe2; - else if (probe < probe2) - probe = probe + ((probe2 - probe) >> 1); - else - probe = probe2 + ((probe - probe2) >> 1); - } - ++attempts; - - set_file_offset(f, probe); - if (!vorbis_find_page(f, NULL, NULL)) return error(f, VORBIS_seek_failed); - if (!vorbis_analyze_page(f, &q)) return error(f, VORBIS_seek_failed); - q.after_previous_page_start = probe; - - // it's possible we've just found the last page again - if (q.page_start == p[1].page_start) { - p[1] = q; - continue; - } - - if (sample_number < q.last_decoded_sample) - p[1] = q; - else - p[0] = q; - } - - if (p[0].last_decoded_sample <= sample_number && sample_number < p[1].last_decoded_sample) { - vorbis_seek_frame_from_page(f, p[1].page_start, p[0].last_decoded_sample, sample_number, fine); - return 0; - } - return error(f, VORBIS_seek_failed); - } -} - -int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number) -{ - return vorbis_seek_base(f, sample_number, FALSE); -} - -int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number) -{ - return vorbis_seek_base(f, sample_number, TRUE); -} - -void stb_vorbis_seek_start(stb_vorbis *f) -{ - if (IS_PUSH_MODE(f)) { error(f, VORBIS_invalid_api_mixing); return; } - set_file_offset(f, f->first_audio_page_offset); - f->previous_length = 0; - f->first_decode = TRUE; - f->next_seg = -1; - vorbis_pump_first_frame(f); -} - -unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f) -{ - unsigned int restore_offset, previous_safe; - unsigned int end, last_page_loc; - - if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); - if (!f->total_samples) { - int last; - uint32 lo,hi; - char header[6]; - - // first, store the current decode position so we can restore it - restore_offset = stb_vorbis_get_file_offset(f); - - // now we want to seek back 64K from the end (the last page must - // be at most a little less than 64K, but let's allow a little slop) - if (f->stream_len >= 65536 && f->stream_len-65536 >= f->first_audio_page_offset) - previous_safe = f->stream_len - 65536; - else - previous_safe = f->first_audio_page_offset; - - set_file_offset(f, previous_safe); - // previous_safe is now our candidate 'earliest known place that seeking - // to will lead to the final page' - - if (!vorbis_find_page(f, &end, (int unsigned *)&last)) { - // if we can't find a page, we're hosed! - f->error = VORBIS_cant_find_last_page; - f->total_samples = 0xffffffff; - goto done; - } - - // check if there are more pages - last_page_loc = stb_vorbis_get_file_offset(f); - - // stop when the last_page flag is set, not when we reach eof; - // this allows us to stop short of a 'file_section' end without - // explicitly checking the length of the section - while (!last) { - set_file_offset(f, end); - if (!vorbis_find_page(f, &end, (int unsigned *)&last)) { - // the last page we found didn't have the 'last page' flag - // set. whoops! - break; - } - previous_safe = last_page_loc+1; - last_page_loc = stb_vorbis_get_file_offset(f); - } - - set_file_offset(f, last_page_loc); - - // parse the header - getn(f, (unsigned char *)header, 6); - // extract the absolute granule position - lo = get32(f); - hi = get32(f); - if (lo == 0xffffffff && hi == 0xffffffff) { - f->error = VORBIS_cant_find_last_page; - f->total_samples = SAMPLE_unknown; - goto done; - } - if (hi) - lo = 0xfffffffe; // saturate - f->total_samples = lo; - - f->p_last.page_start = last_page_loc; - f->p_last.page_end = end; - f->p_last.last_decoded_sample = lo; - f->p_last.first_decoded_sample = SAMPLE_unknown; - f->p_last.after_previous_page_start = previous_safe; - - done: - set_file_offset(f, restore_offset); - } - return f->total_samples == SAMPLE_unknown ? 0 : f->total_samples; -} - -float stb_vorbis_stream_length_in_seconds(stb_vorbis *f) -{ - return stb_vorbis_stream_length_in_samples(f) / (float) f->sample_rate; -} - - - -int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output) -{ - int len, right,left,i; - if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing); - - if (!vorbis_decode_packet(f, &len, &left, &right)) { - f->channel_buffer_start = f->channel_buffer_end = 0; - return 0; - } - - len = vorbis_finish_frame(f, len, left, right); - for (i=0; i < f->channels; ++i) - f->outputs[i] = f->channel_buffers[i] + left; - - f->channel_buffer_start = left; - f->channel_buffer_end = left+len; - - if (channels) *channels = f->channels; - if (output) *output = f->outputs; - return len; -} - -#ifndef STB_VORBIS_NO_STDIO - -stb_vorbis * stb_vorbis_open_file_section(FILE *file, int close_on_free, int *error, stb_vorbis_alloc *alloc, unsigned int length) -{ - stb_vorbis *f, p; - vorbis_init(&p, alloc); - p.f = file; - p.f_start = ftell(file); - p.stream_len = length; - p.close_on_free = close_on_free; - if (start_decoder(&p)) { - f = vorbis_alloc(&p); - if (f) { - *f = p; - vorbis_pump_first_frame(f); - return f; - } - } - if (error) *error = p.error; - vorbis_deinit(&p); - return NULL; -} - -stb_vorbis * stb_vorbis_open_file(FILE *file, int close_on_free, int *error, stb_vorbis_alloc *alloc) -{ - unsigned int len, start; - start = ftell(file); - fseek(file, 0, SEEK_END); - len = ftell(file) - start; - fseek(file, start, SEEK_SET); - return stb_vorbis_open_file_section(file, close_on_free, error, alloc, len); -} - -stb_vorbis * stb_vorbis_open_filename(char *filename, int *error, stb_vorbis_alloc *alloc) -{ - FILE *f = fopen(filename, "rb"); - if (f) - return stb_vorbis_open_file(f, TRUE, error, alloc); - if (error) *error = VORBIS_file_open_failure; - return NULL; -} -#endif // STB_VORBIS_NO_STDIO - -stb_vorbis * stb_vorbis_open_memory(unsigned char *data, int len, int *error, stb_vorbis_alloc *alloc) -{ - stb_vorbis *f, p; - if (data == NULL) return NULL; - vorbis_init(&p, alloc); - p.stream = data; - p.stream_end = data + len; - p.stream_start = p.stream; - p.stream_len = len; - p.push_mode = FALSE; - if (start_decoder(&p)) { - f = vorbis_alloc(&p); - if (f) { - *f = p; - vorbis_pump_first_frame(f); - return f; - } - } - if (error) *error = p.error; - vorbis_deinit(&p); - return NULL; -} - -#ifndef STB_VORBIS_NO_INTEGER_CONVERSION -#define PLAYBACK_MONO 1 -#define PLAYBACK_LEFT 2 -#define PLAYBACK_RIGHT 4 - -#define L (PLAYBACK_LEFT | PLAYBACK_MONO) -#define C (PLAYBACK_LEFT | PLAYBACK_RIGHT | PLAYBACK_MONO) -#define R (PLAYBACK_RIGHT | PLAYBACK_MONO) - -static int8 channel_position[7][6] = -{ - { 0 }, - { C }, - { L, R }, - { L, C, R }, - { L, R, L, R }, - { L, C, R, L, R }, - { L, C, R, L, R, C }, -}; - - -#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT - // add (1<<23) to convert to int, then divide by 2^SHIFT, then add 0.5/2^SHIFT to round - #define MAGIC(SHIFT) (1.5f * (1 << (23-SHIFT)) + 0.5f/(1 << SHIFT)) - #define ADDEND(SHIFT) (((150-SHIFT) << 23) + (1 << 22)) - #define FAST_SCALED_FLOAT_TO_INT(x,s) ((temp = (x) + MAGIC(s)), (*(int *)&temp) - ADDEND(s)) - #define check_endianness() - typedef char stb_vorbis_float_size_test[sizeof(float)==4 && sizeof(int) == 4]; - #define FASTDEF(x) x -#else - #define FAST_SCALED_FLOAT_TO_INT(x,s) ((int) ((x) * (1 << (s)))) - #define check_endianness() - #define FASTDEF(x) -#endif - -static void copy_samples(short *dest, float *src, int len) -{ - int i; - FASTDEF(float temp); - check_endianness(); - for (i=0; i < len; ++i) { - int v = FAST_SCALED_FLOAT_TO_INT(src[i],15); - if ((unsigned int) (v + 32768) > 65535) - v = v < 0 ? -32768 : 32767; - dest[i] = v; - } -} - -static void compute_samples(int mask, short *output, int num_c, float **data, int d_offset, int len) -{ - #define BUFFER_SIZE 32 - float buffer[BUFFER_SIZE]; - int i,j,o,n = BUFFER_SIZE; - FASTDEF(float temp); - check_endianness(); - for (o = 0; o < len; o += BUFFER_SIZE) { - memset(buffer, 0, sizeof(buffer)); - if (o + n > len) n = len - o; - for (j=0; j < num_c; ++j) { - if (channel_position[num_c][j] & mask) { - for (i=0; i < n; ++i) - buffer[i] += data[j][d_offset+o+i]; - } - } - for (i=0; i < n; ++i) { - int v = FAST_SCALED_FLOAT_TO_INT(buffer[i],15); - if ((unsigned int) (v + 32768) > 65535) - v = v < 0 ? -32768 : 32767; - output[o+i] = v; - } - } -} - -static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} }; -static void compute_stereo_samples(short *output, int num_c, float **data, int d_offset, int len) -{ - #define BUFFER_SIZE 32 - float buffer[BUFFER_SIZE]; - int i,j,o,n = BUFFER_SIZE >> 1; - FASTDEF(float temp); - // o is the offset in the source data - check_endianness(); - for (o = 0; o < len; o += BUFFER_SIZE >> 1) { - // o2 is the offset in the output data - int o2 = o << 1; - memset(buffer, 0, sizeof(buffer)); - if (o + n > len) n = len - o; - for (j=0; j < num_c; ++j) { - int m = channel_position[num_c][j] & (PLAYBACK_LEFT | PLAYBACK_RIGHT); - if (m == (PLAYBACK_LEFT | PLAYBACK_RIGHT)) { - for (i=0; i < n; ++i) { - buffer[i*2+0] += data[j][d_offset+o+i]; - buffer[i*2+1] += data[j][d_offset+o+i]; - } - } else if (m == PLAYBACK_LEFT) { - for (i=0; i < n; ++i) { - buffer[i*2+0] += data[j][d_offset+o+i]; - } - } else if (m == PLAYBACK_RIGHT) { - for (i=0; i < n; ++i) { - buffer[i*2+1] += data[j][d_offset+o+i]; - } - } - } - for (i=0; i < (n<<1); ++i) { - int v = FAST_SCALED_FLOAT_TO_INT(buffer[i],15); - if ((unsigned int) (v + 32768) > 65535) - v = v < 0 ? -32768 : 32767; - output[o2+i] = v; - } - } -} - -static void convert_samples_short(int buf_c, short **buffer, int b_offset, int data_c, float **data, int d_offset, int samples) -{ - int i; - if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { - static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} }; - for (i=0; i < buf_c; ++i) - compute_samples(channel_selector[buf_c][i], buffer[i]+b_offset, data_c, data, d_offset, samples); - } else { - int limit = buf_c < data_c ? buf_c : data_c; - for (i=0; i < limit; ++i) - copy_samples(buffer[i]+b_offset, data[i], samples); - for ( ; i < buf_c; ++i) - memset(buffer[i]+b_offset, 0, sizeof(short) * samples); - } -} - -int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples) -{ - float **output; - int len = stb_vorbis_get_frame_float(f, NULL, &output); - if (len > num_samples) len = num_samples; - if (len) - convert_samples_short(num_c, buffer, 0, f->channels, output, 0, len); - return len; -} - -static void convert_channels_short_interleaved(int buf_c, short *buffer, int data_c, float **data, int d_offset, int len) -{ - int i; - check_endianness(); - if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { - assert(buf_c == 2); - for (i=0; i < buf_c; ++i) - compute_stereo_samples(buffer, data_c, data, d_offset, len); - } else { - int limit = buf_c < data_c ? buf_c : data_c; - int j; - FASTDEF(float temp); - for (j=0; j < len; ++j) { - for (i=0; i < limit; ++i) { - int v = FAST_SCALED_FLOAT_TO_INT(data[i][d_offset+j],15); - if ((unsigned int) (v + 32768) > 65535) - v = v < 0 ? -32768 : 32767; - *buffer++ = v; - } - for ( ; i < buf_c; ++i) - *buffer++ = 0; - } - } -} - -int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts) -{ - float **output; - int len; - if (num_c == 1) return stb_vorbis_get_frame_short(f,num_c,&buffer, num_shorts); - len = stb_vorbis_get_frame_float(f, NULL, &output); - if (len) { - if (len*num_c > num_shorts) len = num_shorts / num_c; - convert_channels_short_interleaved(num_c, buffer, f->channels, output, 0, len); - } - return len; -} - -int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts) -{ - float **outputs; - int len = num_shorts / channels; - int n=0; - int z = f->channels; - if (z > channels) z = channels; - while (n < len) { - int k = f->channel_buffer_end - f->channel_buffer_start; - if (n+k >= len) k = len - n; - if (k) - convert_channels_short_interleaved(channels, buffer, f->channels, f->channel_buffers, f->channel_buffer_start, k); - buffer += k*channels; - n += k; - f->channel_buffer_start += k; - if (n == len) break; - if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; - } - return n; -} - -int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int len) -{ - float **outputs; - int n=0; - int z = f->channels; - if (z > channels) z = channels; - while (n < len) { - int k = f->channel_buffer_end - f->channel_buffer_start; - if (n+k >= len) k = len - n; - if (k) - convert_samples_short(channels, buffer, n, f->channels, f->channel_buffers, f->channel_buffer_start, k); - n += k; - f->channel_buffer_start += k; - if (n == len) break; - if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; - } - return n; -} - -#ifndef STB_VORBIS_NO_STDIO -int stb_vorbis_decode_filename(char *filename, int *channels, short **output) -{ - int data_len, offset, total, limit, error; - short *data; - stb_vorbis *v = stb_vorbis_open_filename(filename, &error, NULL); - if (v == NULL) return -1; - limit = v->channels * 4096; - *channels = v->channels; - offset = data_len = 0; - total = limit; - data = (short *) malloc(total * sizeof(*data)); - if (data == NULL) { - stb_vorbis_close(v); - return -2; - } - for (;;) { - int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); - if (n == 0) break; - data_len += n; - offset += n * v->channels; - if (offset + limit > total) { - short *data2; - total *= 2; - data2 = (short *) realloc(data, total * sizeof(*data)); - if (data2 == NULL) { - free(data); - stb_vorbis_close(v); - return -2; - } - data = data2; - } - } - *output = data; - return data_len; -} -#endif // NO_STDIO - -int stb_vorbis_decode_memory(uint8 *mem, int len, int *channels, short **output) -{ - int data_len, offset, total, limit, error; - short *data; - stb_vorbis *v = stb_vorbis_open_memory(mem, len, &error, NULL); - if (v == NULL) return -1; - limit = v->channels * 4096; - *channels = v->channels; - offset = data_len = 0; - total = limit; - data = (short *) malloc(total * sizeof(*data)); - if (data == NULL) { - stb_vorbis_close(v); - return -2; - } - for (;;) { - int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset); - if (n == 0) break; - data_len += n; - offset += n * v->channels; - if (offset + limit > total) { - short *data2; - total *= 2; - data2 = (short *) realloc(data, total * sizeof(*data)); - if (data2 == NULL) { - free(data); - stb_vorbis_close(v); - return -2; - } - data = data2; - } - } - *output = data; - return data_len; -} -#endif - -int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats) -{ - float **outputs; - int len = num_floats / channels; - int n=0; - int z = f->channels; - if (z > channels) z = channels; - while (n < len) { - int i,j; - int k = f->channel_buffer_end - f->channel_buffer_start; - if (n+k >= len) k = len - n; - for (j=0; j < k; ++j) { - for (i=0; i < z; ++i) - *buffer++ = f->channel_buffers[i][f->channel_buffer_start+j]; - for ( ; i < channels; ++i) - *buffer++ = 0; - } - n += k; - f->channel_buffer_start += k; - if (n == len) break; - if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; - } - return n; -} - -int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples) -{ - float **outputs; - int n=0; - int z = f->channels; - if (z > channels) z = channels; - while (n < num_samples) { - int i; - int k = f->channel_buffer_end - f->channel_buffer_start; - if (n+k >= num_samples) k = num_samples - n; - if (k) { - for (i=0; i < z; ++i) - memcpy(buffer[i]+n, f->channel_buffers+f->channel_buffer_start, sizeof(float)*k); - for ( ; i < channels; ++i) - memset(buffer[i]+n, 0, sizeof(float) * k); - } - n += k; - f->channel_buffer_start += k; - if (n == num_samples) break; - if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break; - } - return n; -} -#endif // STB_VORBIS_NO_PULLDATA_API - -#endif // STB_VORBIS_HEADER_ONLY + }
+ #ifndef STB_VORBIS_NO_STDIO
+ if (p->close_on_free) fclose(p->f);
+ #endif
+}
+
+void stb_vorbis_close(stb_vorbis *p)
+{
+ if (p == NULL) return;
+ vorbis_deinit(p);
+ setup_free(p,p);
+}
+
+static void vorbis_init(stb_vorbis *p, stb_vorbis_alloc *z)
+{
+ memset(p, 0, sizeof(*p)); // NULL out all malloc'd pointers to start
+ if (z) {
+ p->alloc = *z;
+ p->alloc.alloc_buffer_length_in_bytes = (p->alloc.alloc_buffer_length_in_bytes+3) & ~3;
+ p->temp_offset = p->alloc.alloc_buffer_length_in_bytes;
+ }
+ p->eof = 0;
+ p->error = VORBIS__no_error;
+ p->stream = NULL;
+ p->codebooks = NULL;
+ p->page_crc_tests = -1;
+ #ifndef STB_VORBIS_NO_STDIO
+ p->close_on_free = FALSE;
+ p->f = NULL;
+ #endif
+}
+
+int stb_vorbis_get_sample_offset(stb_vorbis *f)
+{
+ if (f->current_loc_valid)
+ return f->current_loc;
+ else
+ return -1;
+}
+
+stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f)
+{
+ stb_vorbis_info d;
+ d.channels = f->channels;
+ d.sample_rate = f->sample_rate;
+ d.setup_memory_required = f->setup_memory_required;
+ d.setup_temp_memory_required = f->setup_temp_memory_required;
+ d.temp_memory_required = f->temp_memory_required;
+ d.max_frame_size = f->blocksize_1 >> 1;
+ return d;
+}
+
+int stb_vorbis_get_error(stb_vorbis *f)
+{
+ int e = f->error;
+ f->error = VORBIS__no_error;
+ return e;
+}
+
+static stb_vorbis * vorbis_alloc(stb_vorbis *f)
+{
+ stb_vorbis *p = (stb_vorbis *) setup_malloc(f, sizeof(*p));
+ return p;
+}
+
+#ifndef STB_VORBIS_NO_PUSHDATA_API
+
+void stb_vorbis_flush_pushdata(stb_vorbis *f)
+{
+ f->previous_length = 0;
+ f->page_crc_tests = 0;
+ f->discard_samples_deferred = 0;
+ f->current_loc_valid = FALSE;
+ f->first_decode = FALSE;
+ f->samples_output = 0;
+ f->channel_buffer_start = 0;
+ f->channel_buffer_end = 0;
+}
+
+static int vorbis_search_for_page_pushdata(vorb *f, uint8 *data, int data_len)
+{
+ int i,n;
+ for (i=0; i < f->page_crc_tests; ++i)
+ f->scan[i].bytes_done = 0;
+
+ // if we have room for more scans, search for them first, because
+ // they may cause us to stop early if their header is incomplete
+ if (f->page_crc_tests < STB_VORBIS_PUSHDATA_CRC_COUNT) {
+ if (data_len < 4) return 0;
+ data_len -= 3; // need to look for 4-byte sequence, so don't miss
+ // one that straddles a boundary
+ for (i=0; i < data_len; ++i) {
+ if (data[i] == 0x4f) {
+ if (0==memcmp(data+i, ogg_page_header, 4)) {
+ int j,len;
+ uint32 crc;
+ // make sure we have the whole page header
+ if (i+26 >= data_len || i+27+data[i+26] >= data_len) {
+ // only read up to this page start, so hopefully we'll
+ // have the whole page header start next time
+ data_len = i;
+ break;
+ }
+ // ok, we have it all; compute the length of the page
+ len = 27 + data[i+26];
+ for (j=0; j < data[i+26]; ++j)
+ len += data[i+27+j];
+ // scan everything up to the embedded crc (which we must 0)
+ crc = 0;
+ for (j=0; j < 22; ++j)
+ crc = crc32_update(crc, data[i+j]);
+ // now process 4 0-bytes
+ for ( ; j < 26; ++j)
+ crc = crc32_update(crc, 0);
+ // len is the total number of bytes we need to scan
+ n = f->page_crc_tests++;
+ f->scan[n].bytes_left = len-j;
+ f->scan[n].crc_so_far = crc;
+ f->scan[n].goal_crc = data[i+22] + (data[i+23] << 8) + (data[i+24]<<16) + (data[i+25]<<24);
+ // if the last frame on a page is continued to the next, then
+ // we can't recover the sample_loc immediately
+ if (data[i+27+data[i+26]-1] == 255)
+ f->scan[n].sample_loc = ~0;
+ else
+ f->scan[n].sample_loc = data[i+6] + (data[i+7] << 8) + (data[i+ 8]<<16) + (data[i+ 9]<<24);
+ f->scan[n].bytes_done = i+j;
+ if (f->page_crc_tests == STB_VORBIS_PUSHDATA_CRC_COUNT)
+ break;
+ // keep going if we still have room for more
+ }
+ }
+ }
+ }
+
+ for (i=0; i < f->page_crc_tests;) {
+ uint32 crc;
+ int j;
+ int n = f->scan[i].bytes_done;
+ int m = f->scan[i].bytes_left;
+ if (m > data_len - n) m = data_len - n;
+ // m is the bytes to scan in the current chunk
+ crc = f->scan[i].crc_so_far;
+ for (j=0; j < m; ++j)
+ crc = crc32_update(crc, data[n+j]);
+ f->scan[i].bytes_left -= m;
+ f->scan[i].crc_so_far = crc;
+ if (f->scan[i].bytes_left == 0) {
+ // does it match?
+ if (f->scan[i].crc_so_far == f->scan[i].goal_crc) {
+ // Houston, we have page
+ data_len = n+m; // consumption amount is wherever that scan ended
+ f->page_crc_tests = -1; // drop out of page scan mode
+ f->previous_length = 0; // decode-but-don't-output one frame
+ f->next_seg = -1; // start a new page
+ f->current_loc = f->scan[i].sample_loc; // set the current sample location
+ // to the amount we'd have decoded had we decoded this page
+ f->current_loc_valid = f->current_loc != ~0;
+ return data_len;
+ }
+ // delete entry
+ f->scan[i] = f->scan[--f->page_crc_tests];
+ } else {
+ ++i;
+ }
+ }
+
+ return data_len;
+}
+
+// return value: number of bytes we used
+int stb_vorbis_decode_frame_pushdata(
+ stb_vorbis *f, // the file we're decoding
+ uint8 *data, int data_len, // the memory available for decoding
+ int *channels, // place to write number of float * buffers
+ float ***output, // place to write float ** array of float * buffers
+ int *samples // place to write number of output samples
+ )
+{
+ int i;
+ int len,right,left;
+
+ if (!IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing);
+
+ if (f->page_crc_tests >= 0) {
+ *samples = 0;
+ return vorbis_search_for_page_pushdata(f, data, data_len);
+ }
+
+ f->stream = data;
+ f->stream_end = data + data_len;
+ f->error = VORBIS__no_error;
+
+ // check that we have the entire packet in memory
+ if (!is_whole_packet_present(f, FALSE)) {
+ *samples = 0;
+ return 0;
+ }
+
+ if (!vorbis_decode_packet(f, &len, &left, &right)) {
+ // save the actual error we encountered
+ enum STBVorbisError error = f->error;
+ if (error == VORBIS_bad_packet_type) {
+ // flush and resynch
+ f->error = VORBIS__no_error;
+ while (get8_packet(f) != EOP)
+ if (f->eof) break;
+ *samples = 0;
+ return f->stream - data;
+ }
+ if (error == VORBIS_continued_packet_flag_invalid) {
+ if (f->previous_length == 0) {
+ // we may be resynching, in which case it's ok to hit one
+ // of these; just discard the packet
+ f->error = VORBIS__no_error;
+ while (get8_packet(f) != EOP)
+ if (f->eof) break;
+ *samples = 0;
+ return f->stream - data;
+ }
+ }
+ // if we get an error while parsing, what to do?
+ // well, it DEFINITELY won't work to continue from where we are!
+ stb_vorbis_flush_pushdata(f);
+ // restore the error that actually made us bail
+ f->error = error;
+ *samples = 0;
+ return 1;
+ }
+
+ // success!
+ len = vorbis_finish_frame(f, len, left, right);
+ for (i=0; i < f->channels; ++i)
+ f->outputs[i] = f->channel_buffers[i] + left;
+
+ if (channels) *channels = f->channels;
+ *samples = len;
+ *output = f->outputs;
+ return f->stream - data;
+}
+
+stb_vorbis *stb_vorbis_open_pushdata(
+ unsigned char *data, int data_len, // the memory available for decoding
+ int *data_used, // only defined if result is not NULL
+ int *error, stb_vorbis_alloc *alloc)
+{
+ stb_vorbis *f, p;
+ vorbis_init(&p, alloc);
+ p.stream = data;
+ p.stream_end = data + data_len;
+ p.push_mode = TRUE;
+ if (!start_decoder(&p)) {
+ if (p.eof)
+ *error = VORBIS_need_more_data;
+ else
+ *error = p.error;
+ return NULL;
+ }
+ f = vorbis_alloc(&p);
+ if (f) {
+ *f = p;
+ *data_used = f->stream - data;
+ *error = 0;
+ return f;
+ } else {
+ vorbis_deinit(&p);
+ return NULL;
+ }
+}
+#endif // STB_VORBIS_NO_PUSHDATA_API
+
+unsigned int stb_vorbis_get_file_offset(stb_vorbis *f)
+{
+ #ifndef STB_VORBIS_NO_PUSHDATA_API
+ if (f->push_mode) return 0;
+ #endif
+ if (USE_MEMORY(f)) return f->stream - f->stream_start;
+ #ifndef STB_VORBIS_NO_STDIO
+ return ftell(f->f) - f->f_start;
+ #endif
+}
+
+#ifndef STB_VORBIS_NO_PULLDATA_API
+//
+// DATA-PULLING API
+//
+
+static uint32 vorbis_find_page(stb_vorbis *f, uint32 *end, uint32 *last)
+{
+ for(;;) {
+ int n;
+ if (f->eof) return 0;
+ n = get8(f);
+ if (n == 0x4f) { // page header
+ unsigned int retry_loc = stb_vorbis_get_file_offset(f);
+ int i;
+ // check if we're off the end of a file_section stream
+ if (retry_loc - 25 > f->stream_len)
+ return 0;
+ // check the rest of the header
+ for (i=1; i < 4; ++i)
+ if (get8(f) != ogg_page_header[i])
+ break;
+ if (f->eof) return 0;
+ if (i == 4) {
+ uint8 header[27];
+ uint32 i, crc, goal, len;
+ for (i=0; i < 4; ++i)
+ header[i] = ogg_page_header[i];
+ for (; i < 27; ++i)
+ header[i] = get8(f);
+ if (f->eof) return 0;
+ if (header[4] != 0) goto invalid;
+ goal = header[22] + (header[23] << 8) + (header[24]<<16) + (header[25]<<24);
+ for (i=22; i < 26; ++i)
+ header[i] = 0;
+ crc = 0;
+ for (i=0; i < 27; ++i)
+ crc = crc32_update(crc, header[i]);
+ len = 0;
+ for (i=0; i < header[26]; ++i) {
+ int s = get8(f);
+ crc = crc32_update(crc, s);
+ len += s;
+ }
+ if (len && f->eof) return 0;
+ for (i=0; i < len; ++i)
+ crc = crc32_update(crc, get8(f));
+ // finished parsing probable page
+ if (crc == goal) {
+ // we could now check that it's either got the last
+ // page flag set, OR it's followed by the capture
+ // pattern, but I guess TECHNICALLY you could have
+ // a file with garbage between each ogg page and recover
+ // from it automatically? So even though that paranoia
+ // might decrease the chance of an invalid decode by
+ // another 2^32, not worth it since it would hose those
+ // invalid-but-useful files?
+ if (end)
+ *end = stb_vorbis_get_file_offset(f);
+ if (last)
+ if (header[5] & 0x04)
+ *last = 1;
+ else
+ *last = 0;
+ set_file_offset(f, retry_loc-1);
+ return 1;
+ }
+ }
+ invalid:
+ // not a valid page, so rewind and look for next one
+ set_file_offset(f, retry_loc);
+ }
+ }
+}
+
+// seek is implemented with 'interpolation search'--this is like
+// binary search, but we use the data values to estimate the likely
+// location of the data item (plus a bit of a bias so when the
+// estimation is wrong we don't waste overly much time)
+
+#define SAMPLE_unknown 0xffffffff
+
+
+// ogg vorbis, in its insane infinite wisdom, only provides
+// information about the sample at the END of the page.
+// therefore we COULD have the data we need in the current
+// page, and not know it. we could just use the end location
+// as our only knowledge for bounds, seek back, and eventually
+// the binary search finds it. or we can try to be smart and
+// not waste time trying to locate more pages. we try to be
+// smart, since this data is already in memory anyway, so
+// doing needless I/O would be crazy!
+static int vorbis_analyze_page(stb_vorbis *f, ProbedPage *z)
+{
+ uint8 header[27], lacing[255];
+ uint8 packet_type[255];
+ int num_packet, packet_start, previous =0;
+ int i,len;
+ uint32 samples;
+
+ // record where the page starts
+ z->page_start = stb_vorbis_get_file_offset(f);
+
+ // parse the header
+ getn(f, header, 27);
+ assert(header[0] == 'O' && header[1] == 'g' && header[2] == 'g' && header[3] == 'S');
+ getn(f, lacing, header[26]);
+
+ // determine the length of the payload
+ len = 0;
+ for (i=0; i < header[26]; ++i)
+ len += lacing[i];
+
+ // this implies where the page ends
+ z->page_end = z->page_start + 27 + header[26] + len;
+
+ // read the last-decoded sample out of the data
+ z->last_decoded_sample = header[6] + (header[7] << 8) + (header[8] << 16) + (header[9] << 16);
+
+ if (header[5] & 4) {
+ // if this is the last page, it's not possible to work
+ // backwards to figure out the first sample! whoops! fuck.
+ z->first_decoded_sample = SAMPLE_unknown;
+ set_file_offset(f, z->page_start);
+ return 1;
+ }
+
+ // scan through the frames to determine the sample-count of each one...
+ // our goal is the sample # of the first fully-decoded sample on the
+ // page, which is the first decoded sample of the 2nd page
+
+ num_packet=0;
+
+ packet_start = ((header[5] & 1) == 0);
+
+ for (i=0; i < header[26]; ++i) {
+ if (packet_start) {
+ uint8 n,b,m;
+ if (lacing[i] == 0) goto bail; // trying to read from zero-length packet
+ n = get8(f);
+ // if bottom bit is non-zero, we've got corruption
+ if (n & 1) goto bail;
+ n >>= 1;
+ b = ilog(f->mode_count-1);
+ m = n >> b;
+ n &= (1 << b)-1;
+ if (n >= f->mode_count) goto bail;
+ if (num_packet == 0 && f->mode_config[n].blockflag)
+ previous = (m & 1);
+ packet_type[num_packet++] = f->mode_config[n].blockflag;
+ skip(f, lacing[i]-1);
+ } else
+ skip(f, lacing[i]);
+ packet_start = (lacing[i] < 255);
+ }
+
+ // now that we know the sizes of all the pages, we can start determining
+ // how much sample data there is.
+
+ samples = 0;
+
+ // for the last packet, we step by its whole length, because the definition
+ // is that we encoded the end sample loc of the 'last packet completed',
+ // where 'completed' refers to packets being split, and we are left to guess
+ // what 'end sample loc' means. we assume it means ignoring the fact that
+ // the last half of the data is useless without windowing against the next
+ // packet... (so it's not REALLY complete in that sense)
+ if (num_packet > 1)
+ samples += f->blocksize[packet_type[num_packet-1]];
+
+ for (i=num_packet-2; i >= 1; --i) {
+ // now, for this packet, how many samples do we have that
+ // do not overlap the following packet?
+ if (packet_type[i] == 1)
+ if (packet_type[i+1] == 1)
+ samples += f->blocksize_1 >> 1;
+ else
+ samples += ((f->blocksize_1 - f->blocksize_0) >> 2) + (f->blocksize_0 >> 1);
+ else
+ samples += f->blocksize_0 >> 1;
+ }
+ // now, at this point, we've rewound to the very beginning of the
+ // _second_ packet. if we entirely discard the first packet after
+ // a seek, this will be exactly the right sample number. HOWEVER!
+ // we can't as easily compute this number for the LAST page. The
+ // only way to get the sample offset of the LAST page is to use
+ // the end loc from the previous page. But what that returns us
+ // is _exactly_ the place where we get our first non-overlapped
+ // sample. (I think. Stupid spec for being ambiguous.) So for
+ // consistency it's better to do that here, too. However, that
+ // will then require us to NOT discard all of the first frame we
+ // decode, in some cases, which means an even weirder frame size
+ // and extra code. what a fucking pain.
+
+ // we're going to discard the first packet if we
+ // start the seek here, so we don't care about it. (we could actually
+ // do better; if the first packet is long, and the previous packet
+ // is short, there's actually data in the first half of the first
+ // packet that doesn't need discarding... but not worth paying the
+ // effort of tracking that of that here and in the seeking logic)
+ // except crap, if we infer it from the _previous_ packet's end
+ // location, we DO need to use that definition... and we HAVE to
+ // infer the start loc of the LAST packet from the previous packet's
+ // end location. fuck you, ogg vorbis.
+
+ z->first_decoded_sample = z->last_decoded_sample - samples;
+
+ // restore file state to where we were
+ set_file_offset(f, z->page_start);
+ return 1;
+
+ // restore file state to where we were
+ bail:
+ set_file_offset(f, z->page_start);
+ return 0;
+}
+
+static int vorbis_seek_frame_from_page(stb_vorbis *f, uint32 page_start, uint32 first_sample, uint32 target_sample, int fine)
+{
+ int left_start, left_end, right_start, right_end, mode,i;
+ int frame=0;
+ uint32 frame_start;
+ int frames_to_skip, data_to_skip;
+
+ // first_sample is the sample # of the first sample that doesn't
+ // overlap the previous page... note that this requires us to
+ // _partially_ discard the first packet! bleh.
+ set_file_offset(f, page_start);
+
+ f->next_seg = -1; // force page resync
+
+ frame_start = first_sample;
+ // frame start is where the previous packet's last decoded sample
+ // was, which corresponds to left_end... EXCEPT if the previous
+ // packet was long and this packet is short? Probably a bug here.
+
+
+ // now, we can start decoding frames... we'll only FAKE decode them,
+ // until we find the frame that contains our sample; then we'll rewind,
+ // and try again
+ for (;;) {
+ int start;
+
+ if (!vorbis_decode_initial(f, &left_start, &left_end, &right_start, &right_end, &mode))
+ return error(f, VORBIS_seek_failed);
+
+ if (frame == 0)
+ start = left_end;
+ else
+ start = left_start;
+
+ // the window starts at left_start; the last valid sample we generate
+ // before the next frame's window start is right_start-1
+ if (target_sample < frame_start + right_start-start)
+ break;
+
+ flush_packet(f);
+ if (f->eof)
+ return error(f, VORBIS_seek_failed);
+
+ frame_start += right_start - start;
+
+ ++frame;
+ }
+
+ // ok, at this point, the sample we want is contained in frame #'frame'
+
+ // to decode frame #'frame' normally, we have to decode the
+ // previous frame first... but if it's the FIRST frame of the page
+ // we can't. if it's the first frame, it means it falls in the part
+ // of the first frame that doesn't overlap either of the other frames.
+ // so, if we have to handle that case for the first frame, we might
+ // as well handle it for all of them, so:
+ if (target_sample > frame_start + (left_end - left_start)) {
+ // so what we want to do is go ahead and just immediately decode
+ // this frame, but then make it so the next get_frame_float() uses
+ // this already-decoded data? or do we want to go ahead and rewind,
+ // and leave a flag saying to skip the first N data? let's do that
+ frames_to_skip = frame; // if this is frame #1, skip 1 frame (#0)
+ data_to_skip = left_end - left_start;
+ } else {
+ // otherwise, we want to skip frames 0, 1, 2, ... frame-2
+ // (which means frame-2+1 total frames) then decode frame-1,
+ // then leave frame pending
+ frames_to_skip = frame - 1;
+ assert(frames_to_skip >= 0);
+ data_to_skip = -1;
+ }
+
+ set_file_offset(f, page_start);
+ f->next_seg = - 1; // force page resync
+
+ for (i=0; i < frames_to_skip; ++i) {
+ maybe_start_packet(f);
+ flush_packet(f);
+ }
+
+ if (data_to_skip >= 0) {
+ int i,j,n = f->blocksize_0 >> 1;
+ f->discard_samples_deferred = data_to_skip;
+ for (i=0; i < f->channels; ++i)
+ for (j=0; j < n; ++j)
+ f->previous_window[i][j] = 0;
+ f->previous_length = n;
+ frame_start += data_to_skip;
+ } else {
+ f->previous_length = 0;
+ vorbis_pump_first_frame(f);
+ }
+
+ // at this point, the NEXT decoded frame will generate the desired sample
+ if (fine) {
+ // so if we're doing sample accurate streaming, we want to go ahead and decode it!
+ if (target_sample != frame_start) {
+ int n;
+ stb_vorbis_get_frame_float(f, &n, NULL);
+ assert(target_sample > frame_start);
+ assert(f->channel_buffer_start + (int) (target_sample-frame_start) < f->channel_buffer_end);
+ f->channel_buffer_start += (target_sample - frame_start);
+ }
+ }
+
+ return 0;
+}
+
+static int vorbis_seek_base(stb_vorbis *f, unsigned int sample_number, int fine)
+{
+ ProbedPage p[2],q;
+ if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing);
+
+ // do we know the location of the last page?
+ if (f->p_last.page_start == 0) {
+ uint32 z = stb_vorbis_stream_length_in_samples(f);
+ if (z == 0) return error(f, VORBIS_cant_find_last_page);
+ }
+
+ p[0] = f->p_first;
+ p[1] = f->p_last;
+
+ if (sample_number >= f->p_last.last_decoded_sample)
+ sample_number = f->p_last.last_decoded_sample-1;
+
+ if (sample_number < f->p_first.last_decoded_sample) {
+ vorbis_seek_frame_from_page(f, p[0].page_start, 0, sample_number, fine);
+ return 0;
+ } else {
+ int attempts=0;
+ while (p[0].page_end < p[1].page_start) {
+ uint32 probe;
+ uint32 start_offset, end_offset;
+ uint32 start_sample, end_sample;
+
+ // copy these into local variables so we can tweak them
+ // if any are unknown
+ start_offset = p[0].page_end;
+ end_offset = p[1].after_previous_page_start; // an address known to seek to page p[1]
+ start_sample = p[0].last_decoded_sample;
+ end_sample = p[1].last_decoded_sample;
+
+ // currently there is no such tweaking logic needed/possible?
+ if (start_sample == SAMPLE_unknown || end_sample == SAMPLE_unknown)
+ return error(f, VORBIS_seek_failed);
+
+ // now we want to lerp between these for the target samples...
+
+ // step 1: we need to bias towards the page start...
+ if (start_offset + 4000 < end_offset)
+ end_offset -= 4000;
+
+ // now compute an interpolated search loc
+ probe = start_offset + (int) floor((float) (end_offset - start_offset) / (end_sample - start_sample) * (sample_number - start_sample));
+
+ // next we need to bias towards binary search...
+ // code is a little wonky to allow for full 32-bit unsigned values
+ if (attempts >= 4) {
+ uint32 probe2 = start_offset + ((end_offset - start_offset) >> 1);
+ if (attempts >= 8)
+ probe = probe2;
+ else if (probe < probe2)
+ probe = probe + ((probe2 - probe) >> 1);
+ else
+ probe = probe2 + ((probe - probe2) >> 1);
+ }
+ ++attempts;
+
+ set_file_offset(f, probe);
+ if (!vorbis_find_page(f, NULL, NULL)) return error(f, VORBIS_seek_failed);
+ if (!vorbis_analyze_page(f, &q)) return error(f, VORBIS_seek_failed);
+ q.after_previous_page_start = probe;
+
+ // it's possible we've just found the last page again
+ if (q.page_start == p[1].page_start) {
+ p[1] = q;
+ continue;
+ }
+
+ if (sample_number < q.last_decoded_sample)
+ p[1] = q;
+ else
+ p[0] = q;
+ }
+
+ if (p[0].last_decoded_sample <= sample_number && sample_number < p[1].last_decoded_sample) {
+ vorbis_seek_frame_from_page(f, p[1].page_start, p[0].last_decoded_sample, sample_number, fine);
+ return 0;
+ }
+ return error(f, VORBIS_seek_failed);
+ }
+}
+
+int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number)
+{
+ return vorbis_seek_base(f, sample_number, FALSE);
+}
+
+int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number)
+{
+ return vorbis_seek_base(f, sample_number, TRUE);
+}
+
+void stb_vorbis_seek_start(stb_vorbis *f)
+{
+ if (IS_PUSH_MODE(f)) { error(f, VORBIS_invalid_api_mixing); return; }
+ set_file_offset(f, f->first_audio_page_offset);
+ f->previous_length = 0;
+ f->first_decode = TRUE;
+ f->next_seg = -1;
+ vorbis_pump_first_frame(f);
+}
+
+unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f)
+{
+ unsigned int restore_offset, previous_safe;
+ unsigned int end, last_page_loc;
+
+ if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing);
+ if (!f->total_samples) {
+ int last;
+ uint32 lo,hi;
+ char header[6];
+
+ // first, store the current decode position so we can restore it
+ restore_offset = stb_vorbis_get_file_offset(f);
+
+ // now we want to seek back 64K from the end (the last page must
+ // be at most a little less than 64K, but let's allow a little slop)
+ if (f->stream_len >= 65536 && f->stream_len-65536 >= f->first_audio_page_offset)
+ previous_safe = f->stream_len - 65536;
+ else
+ previous_safe = f->first_audio_page_offset;
+
+ set_file_offset(f, previous_safe);
+ // previous_safe is now our candidate 'earliest known place that seeking
+ // to will lead to the final page'
+
+ if (!vorbis_find_page(f, &end, (int unsigned *)&last)) {
+ // if we can't find a page, we're hosed!
+ f->error = VORBIS_cant_find_last_page;
+ f->total_samples = 0xffffffff;
+ goto done;
+ }
+
+ // check if there are more pages
+ last_page_loc = stb_vorbis_get_file_offset(f);
+
+ // stop when the last_page flag is set, not when we reach eof;
+ // this allows us to stop short of a 'file_section' end without
+ // explicitly checking the length of the section
+ while (!last) {
+ set_file_offset(f, end);
+ if (!vorbis_find_page(f, &end, (int unsigned *)&last)) {
+ // the last page we found didn't have the 'last page' flag
+ // set. whoops!
+ break;
+ }
+ previous_safe = last_page_loc+1;
+ last_page_loc = stb_vorbis_get_file_offset(f);
+ }
+
+ set_file_offset(f, last_page_loc);
+
+ // parse the header
+ getn(f, (unsigned char *)header, 6);
+ // extract the absolute granule position
+ lo = get32(f);
+ hi = get32(f);
+ if (lo == 0xffffffff && hi == 0xffffffff) {
+ f->error = VORBIS_cant_find_last_page;
+ f->total_samples = SAMPLE_unknown;
+ goto done;
+ }
+ if (hi)
+ lo = 0xfffffffe; // saturate
+ f->total_samples = lo;
+
+ f->p_last.page_start = last_page_loc;
+ f->p_last.page_end = end;
+ f->p_last.last_decoded_sample = lo;
+ f->p_last.first_decoded_sample = SAMPLE_unknown;
+ f->p_last.after_previous_page_start = previous_safe;
+
+ done:
+ set_file_offset(f, restore_offset);
+ }
+ return f->total_samples == SAMPLE_unknown ? 0 : f->total_samples;
+}
+
+float stb_vorbis_stream_length_in_seconds(stb_vorbis *f)
+{
+ return stb_vorbis_stream_length_in_samples(f) / (float) f->sample_rate;
+}
+
+
+
+int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output)
+{
+ int len, right,left,i;
+ if (IS_PUSH_MODE(f)) return error(f, VORBIS_invalid_api_mixing);
+
+ if (!vorbis_decode_packet(f, &len, &left, &right)) {
+ f->channel_buffer_start = f->channel_buffer_end = 0;
+ return 0;
+ }
+
+ len = vorbis_finish_frame(f, len, left, right);
+ for (i=0; i < f->channels; ++i)
+ f->outputs[i] = f->channel_buffers[i] + left;
+
+ f->channel_buffer_start = left;
+ f->channel_buffer_end = left+len;
+
+ if (channels) *channels = f->channels;
+ if (output) *output = f->outputs;
+ return len;
+}
+
+#ifndef STB_VORBIS_NO_STDIO
+
+stb_vorbis * stb_vorbis_open_file_section(FILE *file, int close_on_free, int *error, stb_vorbis_alloc *alloc, unsigned int length)
+{
+ stb_vorbis *f, p;
+ vorbis_init(&p, alloc);
+ p.f = file;
+ p.f_start = ftell(file);
+ p.stream_len = length;
+ p.close_on_free = close_on_free;
+ if (start_decoder(&p)) {
+ f = vorbis_alloc(&p);
+ if (f) {
+ *f = p;
+ vorbis_pump_first_frame(f);
+ return f;
+ }
+ }
+ if (error) *error = p.error;
+ vorbis_deinit(&p);
+ return NULL;
+}
+
+stb_vorbis * stb_vorbis_open_file(FILE *file, int close_on_free, int *error, stb_vorbis_alloc *alloc)
+{
+ unsigned int len, start;
+ start = ftell(file);
+ fseek(file, 0, SEEK_END);
+ len = ftell(file) - start;
+ fseek(file, start, SEEK_SET);
+ return stb_vorbis_open_file_section(file, close_on_free, error, alloc, len);
+}
+
+stb_vorbis * stb_vorbis_open_filename(char *filename, int *error, stb_vorbis_alloc *alloc)
+{
+ FILE *f = fopen(filename, "rb");
+ if (f)
+ return stb_vorbis_open_file(f, TRUE, error, alloc);
+ if (error) *error = VORBIS_file_open_failure;
+ return NULL;
+}
+#endif // STB_VORBIS_NO_STDIO
+
+stb_vorbis * stb_vorbis_open_memory(unsigned char *data, int len, int *error, stb_vorbis_alloc *alloc)
+{
+ stb_vorbis *f, p;
+ if (data == NULL) return NULL;
+ vorbis_init(&p, alloc);
+ p.stream = data;
+ p.stream_end = data + len;
+ p.stream_start = p.stream;
+ p.stream_len = len;
+ p.push_mode = FALSE;
+ if (start_decoder(&p)) {
+ f = vorbis_alloc(&p);
+ if (f) {
+ *f = p;
+ vorbis_pump_first_frame(f);
+ return f;
+ }
+ }
+ if (error) *error = p.error;
+ vorbis_deinit(&p);
+ return NULL;
+}
+
+#ifndef STB_VORBIS_NO_INTEGER_CONVERSION
+#define PLAYBACK_MONO 1
+#define PLAYBACK_LEFT 2
+#define PLAYBACK_RIGHT 4
+
+#define L (PLAYBACK_LEFT | PLAYBACK_MONO)
+#define C (PLAYBACK_LEFT | PLAYBACK_RIGHT | PLAYBACK_MONO)
+#define R (PLAYBACK_RIGHT | PLAYBACK_MONO)
+
+static int8 channel_position[7][6] =
+{
+ { 0 },
+ { C },
+ { L, R },
+ { L, C, R },
+ { L, R, L, R },
+ { L, C, R, L, R },
+ { L, C, R, L, R, C },
+};
+
+
+#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT
+ // add (1<<23) to convert to int, then divide by 2^SHIFT, then add 0.5/2^SHIFT to round
+ #define MAGIC(SHIFT) (1.5f * (1 << (23-SHIFT)) + 0.5f/(1 << SHIFT))
+ #define ADDEND(SHIFT) (((150-SHIFT) << 23) + (1 << 22))
+ #define FAST_SCALED_FLOAT_TO_INT(x,s) ((temp = (x) + MAGIC(s)), (*(int *)&temp) - ADDEND(s))
+ #define check_endianness()
+ typedef char stb_vorbis_float_size_test[sizeof(float)==4 && sizeof(int) == 4];
+ #define FASTDEF(x) x
+#else
+ #define FAST_SCALED_FLOAT_TO_INT(x,s) ((int) ((x) * (1 << (s))))
+ #define check_endianness()
+ #define FASTDEF(x)
+#endif
+
+static void copy_samples(short *dest, float *src, int len)
+{
+ int i;
+ FASTDEF(float temp);
+ check_endianness();
+ for (i=0; i < len; ++i) {
+ int v = FAST_SCALED_FLOAT_TO_INT(src[i],15);
+ if ((unsigned int) (v + 32768) > 65535)
+ v = v < 0 ? -32768 : 32767;
+ dest[i] = v;
+ }
+}
+
+static void compute_samples(int mask, short *output, int num_c, float **data, int d_offset, int len)
+{
+ #define BUFFER_SIZE 32
+ float buffer[BUFFER_SIZE];
+ int i,j,o,n = BUFFER_SIZE;
+ FASTDEF(float temp);
+ check_endianness();
+ for (o = 0; o < len; o += BUFFER_SIZE) {
+ memset(buffer, 0, sizeof(buffer));
+ if (o + n > len) n = len - o;
+ for (j=0; j < num_c; ++j) {
+ if (channel_position[num_c][j] & mask) {
+ for (i=0; i < n; ++i)
+ buffer[i] += data[j][d_offset+o+i];
+ }
+ }
+ for (i=0; i < n; ++i) {
+ int v = FAST_SCALED_FLOAT_TO_INT(buffer[i],15);
+ if ((unsigned int) (v + 32768) > 65535)
+ v = v < 0 ? -32768 : 32767;
+ output[o+i] = v;
+ }
+ }
+}
+
+static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} };
+static void compute_stereo_samples(short *output, int num_c, float **data, int d_offset, int len)
+{
+ #define BUFFER_SIZE 32
+ float buffer[BUFFER_SIZE];
+ int i,j,o,n = BUFFER_SIZE >> 1;
+ FASTDEF(float temp);
+ // o is the offset in the source data
+ check_endianness();
+ for (o = 0; o < len; o += BUFFER_SIZE >> 1) {
+ // o2 is the offset in the output data
+ int o2 = o << 1;
+ memset(buffer, 0, sizeof(buffer));
+ if (o + n > len) n = len - o;
+ for (j=0; j < num_c; ++j) {
+ int m = channel_position[num_c][j] & (PLAYBACK_LEFT | PLAYBACK_RIGHT);
+ if (m == (PLAYBACK_LEFT | PLAYBACK_RIGHT)) {
+ for (i=0; i < n; ++i) {
+ buffer[i*2+0] += data[j][d_offset+o+i];
+ buffer[i*2+1] += data[j][d_offset+o+i];
+ }
+ } else if (m == PLAYBACK_LEFT) {
+ for (i=0; i < n; ++i) {
+ buffer[i*2+0] += data[j][d_offset+o+i];
+ }
+ } else if (m == PLAYBACK_RIGHT) {
+ for (i=0; i < n; ++i) {
+ buffer[i*2+1] += data[j][d_offset+o+i];
+ }
+ }
+ }
+ for (i=0; i < (n<<1); ++i) {
+ int v = FAST_SCALED_FLOAT_TO_INT(buffer[i],15);
+ if ((unsigned int) (v + 32768) > 65535)
+ v = v < 0 ? -32768 : 32767;
+ output[o2+i] = v;
+ }
+ }
+}
+
+static void convert_samples_short(int buf_c, short **buffer, int b_offset, int data_c, float **data, int d_offset, int samples)
+{
+ int i;
+ if (buf_c != data_c && buf_c <= 2 && data_c <= 6) {
+ static int channel_selector[3][2] = { {0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT} };
+ for (i=0; i < buf_c; ++i)
+ compute_samples(channel_selector[buf_c][i], buffer[i]+b_offset, data_c, data, d_offset, samples);
+ } else {
+ int limit = buf_c < data_c ? buf_c : data_c;
+ for (i=0; i < limit; ++i)
+ copy_samples(buffer[i]+b_offset, data[i], samples);
+ for ( ; i < buf_c; ++i)
+ memset(buffer[i]+b_offset, 0, sizeof(short) * samples);
+ }
+}
+
+int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples)
+{
+ float **output;
+ int len = stb_vorbis_get_frame_float(f, NULL, &output);
+ if (len > num_samples) len = num_samples;
+ if (len)
+ convert_samples_short(num_c, buffer, 0, f->channels, output, 0, len);
+ return len;
+}
+
+static void convert_channels_short_interleaved(int buf_c, short *buffer, int data_c, float **data, int d_offset, int len)
+{
+ int i;
+ check_endianness();
+ if (buf_c != data_c && buf_c <= 2 && data_c <= 6) {
+ assert(buf_c == 2);
+ for (i=0; i < buf_c; ++i)
+ compute_stereo_samples(buffer, data_c, data, d_offset, len);
+ } else {
+ int limit = buf_c < data_c ? buf_c : data_c;
+ int j;
+ FASTDEF(float temp);
+ for (j=0; j < len; ++j) {
+ for (i=0; i < limit; ++i) {
+ int v = FAST_SCALED_FLOAT_TO_INT(data[i][d_offset+j],15);
+ if ((unsigned int) (v + 32768) > 65535)
+ v = v < 0 ? -32768 : 32767;
+ *buffer++ = v;
+ }
+ for ( ; i < buf_c; ++i)
+ *buffer++ = 0;
+ }
+ }
+}
+
+int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts)
+{
+ float **output;
+ int len;
+ if (num_c == 1) return stb_vorbis_get_frame_short(f,num_c,&buffer, num_shorts);
+ len = stb_vorbis_get_frame_float(f, NULL, &output);
+ if (len) {
+ if (len*num_c > num_shorts) len = num_shorts / num_c;
+ convert_channels_short_interleaved(num_c, buffer, f->channels, output, 0, len);
+ }
+ return len;
+}
+
+int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts)
+{
+ float **outputs;
+ int len = num_shorts / channels;
+ int n=0;
+ int z = f->channels;
+ if (z > channels) z = channels;
+ while (n < len) {
+ int k = f->channel_buffer_end - f->channel_buffer_start;
+ if (n+k >= len) k = len - n;
+ if (k)
+ convert_channels_short_interleaved(channels, buffer, f->channels, f->channel_buffers, f->channel_buffer_start, k);
+ buffer += k*channels;
+ n += k;
+ f->channel_buffer_start += k;
+ if (n == len) break;
+ if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break;
+ }
+ return n;
+}
+
+int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int len)
+{
+ float **outputs;
+ int n=0;
+ int z = f->channels;
+ if (z > channels) z = channels;
+ while (n < len) {
+ int k = f->channel_buffer_end - f->channel_buffer_start;
+ if (n+k >= len) k = len - n;
+ if (k)
+ convert_samples_short(channels, buffer, n, f->channels, f->channel_buffers, f->channel_buffer_start, k);
+ n += k;
+ f->channel_buffer_start += k;
+ if (n == len) break;
+ if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break;
+ }
+ return n;
+}
+
+#ifndef STB_VORBIS_NO_STDIO
+int stb_vorbis_decode_filename(char *filename, int *channels, short **output)
+{
+ int data_len, offset, total, limit, error;
+ short *data;
+ stb_vorbis *v = stb_vorbis_open_filename(filename, &error, NULL);
+ if (v == NULL) return -1;
+ limit = v->channels * 4096;
+ *channels = v->channels;
+ offset = data_len = 0;
+ total = limit;
+ data = (short *) malloc(total * sizeof(*data));
+ if (data == NULL) {
+ stb_vorbis_close(v);
+ return -2;
+ }
+ for (;;) {
+ int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset);
+ if (n == 0) break;
+ data_len += n;
+ offset += n * v->channels;
+ if (offset + limit > total) {
+ short *data2;
+ total *= 2;
+ data2 = (short *) realloc(data, total * sizeof(*data));
+ if (data2 == NULL) {
+ free(data);
+ stb_vorbis_close(v);
+ return -2;
+ }
+ data = data2;
+ }
+ }
+ *output = data;
+ return data_len;
+}
+#endif // NO_STDIO
+
+int stb_vorbis_decode_memory(uint8 *mem, int len, int *channels, short **output)
+{
+ int data_len, offset, total, limit, error;
+ short *data;
+ stb_vorbis *v = stb_vorbis_open_memory(mem, len, &error, NULL);
+ if (v == NULL) return -1;
+ limit = v->channels * 4096;
+ *channels = v->channels;
+ offset = data_len = 0;
+ total = limit;
+ data = (short *) malloc(total * sizeof(*data));
+ if (data == NULL) {
+ stb_vorbis_close(v);
+ return -2;
+ }
+ for (;;) {
+ int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data+offset, total-offset);
+ if (n == 0) break;
+ data_len += n;
+ offset += n * v->channels;
+ if (offset + limit > total) {
+ short *data2;
+ total *= 2;
+ data2 = (short *) realloc(data, total * sizeof(*data));
+ if (data2 == NULL) {
+ free(data);
+ stb_vorbis_close(v);
+ return -2;
+ }
+ data = data2;
+ }
+ }
+ *output = data;
+ return data_len;
+}
+#endif
+
+int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats)
+{
+ float **outputs;
+ int len = num_floats / channels;
+ int n=0;
+ int z = f->channels;
+ if (z > channels) z = channels;
+ while (n < len) {
+ int i,j;
+ int k = f->channel_buffer_end - f->channel_buffer_start;
+ if (n+k >= len) k = len - n;
+ for (j=0; j < k; ++j) {
+ for (i=0; i < z; ++i)
+ *buffer++ = f->channel_buffers[i][f->channel_buffer_start+j];
+ for ( ; i < channels; ++i)
+ *buffer++ = 0;
+ }
+ n += k;
+ f->channel_buffer_start += k;
+ if (n == len) break;
+ if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break;
+ }
+ return n;
+}
+
+int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples)
+{
+ float **outputs;
+ int n=0;
+ int z = f->channels;
+ if (z > channels) z = channels;
+ while (n < num_samples) {
+ int i;
+ int k = f->channel_buffer_end - f->channel_buffer_start;
+ if (n+k >= num_samples) k = num_samples - n;
+ if (k) {
+ for (i=0; i < z; ++i)
+ memcpy(buffer[i]+n, f->channel_buffers+f->channel_buffer_start, sizeof(float)*k);
+ for ( ; i < channels; ++i)
+ memset(buffer[i]+n, 0, sizeof(float) * k);
+ }
+ n += k;
+ f->channel_buffer_start += k;
+ if (n == num_samples) break;
+ if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) break;
+ }
+ return n;
+}
+#endif // STB_VORBIS_NO_PULLDATA_API
+
+#endif // STB_VORBIS_HEADER_ONLY
diff --git a/src/SFML/Graphics/Font.cpp b/src/SFML/Graphics/Font.cpp index df2e563..36c4c28 100755 --- a/src/SFML/Graphics/Font.cpp +++ b/src/SFML/Graphics/Font.cpp @@ -158,14 +158,28 @@ const Image& Font::GetImage() const ////////////////////////////////////////////////////////////
const Font& Font::GetDefaultFont()
{
- static Font DefaultFont;
- static bool DefaultFontLoaded = false;
+#if defined(SFML_SYSTEM_WINDOWS) && defined(SFML_DYNAMIC)
+
+ // On Windows dynamic build, the default font causes a crash at global exit.
+ // This is a temporary workaround that turns the crash into a memory leak.
+ // Note that this bug doesn't exist anymore in SFML 2.
+ static Font* DefaultFontPtr = new Font;
+ Font& DefaultFont = *DefaultFontPtr;
+
+#else
+
+ static Font DefaultFont;
+
+#endif
+
+ // Get the raw data of the Arial font file into an array, so that we can load it into the font
static const char DefaultFontData[] =
{
#include <SFML/Graphics/Arial.hpp>
};
// Load the default font on first call
+ static bool DefaultFontLoaded = false;
if (!DefaultFontLoaded)
{
DefaultFont.LoadFromMemory(DefaultFontData, sizeof(DefaultFontData), 30);
diff --git a/src/SFML/Graphics/FontLoader.cpp b/src/SFML/Graphics/FontLoader.cpp index 922cac9..0f6e67b 100755 --- a/src/SFML/Graphics/FontLoader.cpp +++ b/src/SFML/Graphics/FontLoader.cpp @@ -213,10 +213,6 @@ FT_Error FontLoader::CreateBitmapFont(FT_Face FontFace, unsigned int CharSize, c FT_Glyph_To_Bitmap(&Glyph, FT_RENDER_MODE_NORMAL, 0, 1);
FT_BitmapGlyph BitmapGlyph = (FT_BitmapGlyph)Glyph;
- // Should we handle other pixel modes ?
- if (BitmapGlyph->bitmap.pixel_mode != FT_PIXEL_MODE_GRAY)
- return FT_Err_Cannot_Render_Glyph;
-
// Add it to the sorted table of glyphs
Glyphs.insert(std::make_pair(BitmapGlyph, Charset[i]));
}
@@ -293,7 +289,7 @@ FT_Error FontLoader::CreateBitmapFont(FT_Face FontFace, unsigned int CharSize, c for (std::size_t i = 0; i < Charset.size(); ++i)
{
Uint32 CurChar = Charset[i];
- LoadedFont.myGlyphs[CurChar].TexCoords = LoadedFont.myTexture.GetTexCoords(Coords[CurChar], false);
+ LoadedFont.myGlyphs[CurChar].TexCoords = LoadedFont.myTexture.GetTexCoords(Coords[CurChar]);
}
// Update the character size (it may have been changed by the function)
diff --git a/src/SFML/Graphics/Image.cpp b/src/SFML/Graphics/Image.cpp index f90673c..14346fb 100755 --- a/src/SFML/Graphics/Image.cpp +++ b/src/SFML/Graphics/Image.cpp @@ -542,25 +542,15 @@ bool Image::IsSmooth() const /// Convert a subrect expressed in pixels, into float
/// texture coordinates
////////////////////////////////////////////////////////////
-FloatRect Image::GetTexCoords(const IntRect& Rect, bool Adjust) const
+FloatRect Image::GetTexCoords(const IntRect& Rect) const
{
float Width = static_cast<float>(myTextureWidth);
float Height = static_cast<float>(myTextureHeight);
- if (Adjust && myIsSmooth)
- {
- return FloatRect((Rect.Left + 0.5f) / Width,
- (Rect.Top + 0.5f) / Height,
- (Rect.Right - 0.5f) / Width,
- (Rect.Bottom - 0.5f) / Height);
- }
- else
- {
- return FloatRect(Rect.Left / Width,
- Rect.Top / Height,
- Rect.Right / Width,
- Rect.Bottom / Height);
- }
+ return FloatRect(Rect.Left / Width,
+ Rect.Top / Height,
+ Rect.Right / Width,
+ Rect.Bottom / Height);
}
diff --git a/src/SFML/Graphics/Makefile b/src/SFML/Graphics/Makefile index f1cc6ff..4de224d 100755 --- a/src/SFML/Graphics/Makefile +++ b/src/SFML/Graphics/Makefile @@ -18,7 +18,7 @@ ifeq ($(STATIC), yes) else
LIB = libsfml-graphics.so
LIBNAME = $(LIBPATH)/$(LIB).$(VERSION)
- INSTALL = && $(LN) $(LNFLAGS) $(DESTLIBDIR)/$(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
+ INSTALL = && $(LN) $(LNFLAGS) $(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
endif
all: $(LIB)
@@ -27,7 +27,7 @@ libsfml-graphics-s.a: $(OBJ) $(OBJGLEW) $(OBJJPEG) $(OBJPNG) $(OBJSOIL) $(OBJZLI $(AR) $(ARFLAGS) $(LIBNAME) $(OBJ) $(OBJGLEW) $(OBJJPEG) $(OBJPNG) $(OBJSOIL) $(OBJZLIB)
libsfml-graphics.so: $(OBJ) $(OBJGLEW) $(OBJJPEG) $(OBJPNG) $(OBJSOIL) $(OBJZLIB)
- $(CPP) $(LDFLAGS) -Wl,-soname,$(LIB).$(VERSION) -o $(LIBNAME) $(OBJ) $(OBJGLEW) $(OBJJPEG) $(OBJPNG) $(OBJSOIL) $(OBJZLIB) -lfreetype
+ $(CPP) $(LDFLAGS) -Wl,-soname,$(LIB).$(VERSION) -o $(LIBNAME) $(OBJ) $(OBJGLEW) $(OBJJPEG) $(OBJPNG) $(OBJSOIL) $(OBJZLIB) -lfreetype -lX11 -lGL
$(OBJ): %.o: %.cpp
$(CPP) -o $@ -c $< $(CFLAGS) -I/usr/include/freetype2
diff --git a/src/SFML/Graphics/RenderTarget.cpp b/src/SFML/Graphics/RenderTarget.cpp index f9d9096..2dca866 100755 --- a/src/SFML/Graphics/RenderTarget.cpp +++ b/src/SFML/Graphics/RenderTarget.cpp @@ -86,10 +86,10 @@ void RenderTarget::Draw(const Drawable& Object) // Save the current render states and set the SFML ones
if (myPreserveStates)
{
- GLCheck(glMatrixMode(GL_MODELVIEW)); GLCheck(glPushMatrix());
- GLCheck(glMatrixMode(GL_PROJECTION)); GLCheck(glPushMatrix());
GLCheck(glPushAttrib(GL_COLOR_BUFFER_BIT | GL_CURRENT_BIT | GL_ENABLE_BIT |
GL_TEXTURE_BIT | GL_TRANSFORM_BIT | GL_VIEWPORT_BIT));
+ GLCheck(glMatrixMode(GL_MODELVIEW)); GLCheck(glPushMatrix());
+ GLCheck(glMatrixMode(GL_PROJECTION)); GLCheck(glPushMatrix());
SetRenderStates();
}
diff --git a/src/SFML/Graphics/Sprite.cpp b/src/SFML/Graphics/Sprite.cpp index a728d4b..375bd5e 100755 --- a/src/SFML/Graphics/Sprite.cpp +++ b/src/SFML/Graphics/Sprite.cpp @@ -186,6 +186,10 @@ void Sprite::Render(RenderTarget&) const // Check if the image is valid
if (myImage && (myImage->GetWidth() > 0) && (myImage->GetHeight() > 0))
{
+ // Use the "offset trick" to get pixel-perfect rendering
+ // see http://www.opengl.org/resources/faq/technical/transformations.htm#tran0030
+ GLCheck(glTranslatef(0.375f, 0.375f, 0.f));
+
// Bind the texture
myImage->Bind();
diff --git a/src/SFML/Makefile b/src/SFML/Makefile index dbc9d45..d8f633e 100755 --- a/src/SFML/Makefile +++ b/src/SFML/Makefile @@ -21,31 +21,36 @@ export CFLAGS = -W -Wall -pedantic -I../../../include -I../../ $(DEBUGFLAGS) export CFLAGSEXT = -I../../../include -I../.. $(DEBUGFLAGS) $(BUILDFLAGS)
export LDFLAGS = $(LINKFLAGS)
export LIBPATH = ../../../lib
-export VERSION = 1.5
+export VERSION = 1.6
export CP = cp
export LN = ln
export LNFLAGS = -s -f
export AR = ar
export ARFLAGS = rcs
-export DESTDIR = /usr
+export DESTDIR = /usr/local
export DESTLIBDIR = $(DESTDIR)/lib
export DESTINCDIR = $(DESTDIR)/include
all: sfml-system sfml-window sfml-network sfml-graphics sfml-audio
sfml-system:
+ @(mkdir -p ../../lib)
@(cd ./System && $(MAKE))
sfml-window:
+ @(mkdir -p ../../lib)
@(cd ./Window && $(MAKE))
sfml-network:
+ @(mkdir -p ../../lib)
@(cd ./Network && $(MAKE))
sfml-graphics:
+ @(mkdir -p ../../lib)
@(cd ./Graphics && $(MAKE))
sfml-audio:
+ @(mkdir -p ../../lib)
@(cd ./Audio && $(MAKE))
.PHONY: clean mrproper
diff --git a/src/SFML/Network/Ftp.cpp b/src/SFML/Network/Ftp.cpp index 85146d4..c53322c 100755 --- a/src/SFML/Network/Ftp.cpp +++ b/src/SFML/Network/Ftp.cpp @@ -380,7 +380,8 @@ Ftp::Response Ftp::Download(const std::string& DistantFile, const std::string& D std::ofstream File((Path + Filename).c_str(), std::ios_base::binary);
if (!File)
return Response(Response::InvalidFile);
- File.write(&FileData[0], static_cast<std::streamsize>(FileData.size()));
+ if (!FileData.empty())
+ File.write(&FileData[0], static_cast<std::streamsize>(FileData.size()));
}
}
}
@@ -402,7 +403,8 @@ Ftp::Response Ftp::Upload(const std::string& LocalFile, const std::string& DestP std::size_t Length = File.tellg();
File.seekg(0, std::ios::beg);
std::vector<char> FileData(Length);
- File.read(&FileData[0], static_cast<std::streamsize>(Length));
+ if (Length > 0)
+ File.read(&FileData[0], static_cast<std::streamsize>(Length));
// Extract the filename from the file path
std::string Filename = LocalFile;
@@ -700,7 +702,8 @@ void Ftp::DataChannel::Receive(std::vector<char>& Data) void Ftp::DataChannel::Send(const std::vector<char>& Data)
{
// Send data
- myDataSocket.Send(&Data[0], Data.size());
+ if (!Data.empty())
+ myDataSocket.Send(&Data[0], Data.size());
// Close the data socket
myDataSocket.Close();
diff --git a/src/SFML/Network/Http.cpp b/src/SFML/Network/Http.cpp index af34a34..4a9502f 100755 --- a/src/SFML/Network/Http.cpp +++ b/src/SFML/Network/Http.cpp @@ -27,6 +27,8 @@ ////////////////////////////////////////////////////////////
#include <SFML/Network/Http.hpp>
#include <ctype.h>
+#include <algorithm>
+#include <iterator>
#include <sstream>
@@ -51,14 +53,12 @@ namespace sf ////////////////////////////////////////////////////////////
/// Default constructor
////////////////////////////////////////////////////////////
-Http::Request::Request(Method RequestMethod, const std::string& URI, const std::string& Body) :
-myMethod (RequestMethod),
-myURI (URI),
-myMajorVersion(1),
-myMinorVersion(0),
-myBody (Body)
+Http::Request::Request(Method RequestMethod, const std::string& URI, const std::string& Body)
{
-
+ SetMethod(RequestMethod);
+ SetURI(URI);
+ SetHttpVersion(1, 0);
+ SetBody(Body);
}
@@ -180,7 +180,7 @@ myMinorVersion(0) ////////////////////////////////////////////////////////////
const std::string& Http::Response::GetField(const std::string& Field) const
{
- FieldTable::const_iterator It = myFields.find(Field);
+ FieldTable::const_iterator It = myFields.find(ToLower(Field));
if (It != myFields.end())
{
return It->second;
@@ -297,8 +297,7 @@ void Http::Response::FromString(const std::string& Data) // Finally extract the body
myBody.clear();
- while (std::getline(In, Line))
- myBody += Line + "\n";
+ std::copy(std::istreambuf_iterator<char>(In), std::istreambuf_iterator<char>(), std::back_inserter(myBody));
}
@@ -386,6 +385,14 @@ Http::Response Http::SendRequest(const Http::Request& Req, float Timeout) Out << ToSend.myBody.size();
ToSend.SetField("Content-Length", Out.str());
}
+ if ((ToSend.myMethod == Request::Post) && !ToSend.HasField("Content-Type"))
+ {
+ ToSend.SetField("Content-Type", "application/x-www-form-urlencoded");
+ }
+ if ((ToSend.myMajorVersion * 10 + ToSend.myMinorVersion >= 11) && !ToSend.HasField("Connection"))
+ {
+ ToSend.SetField("Connection", "close");
+ }
// Prepare the response
Response Received;
diff --git a/src/SFML/Network/Makefile b/src/SFML/Network/Makefile index d30da62..62fb673 100755 --- a/src/SFML/Network/Makefile +++ b/src/SFML/Network/Makefile @@ -8,7 +8,7 @@ ifeq ($(STATIC), yes) else
LIB = libsfml-network.so
LIBNAME = $(LIBPATH)/$(LIB).$(VERSION)
- INSTALL = && $(LN) $(LNFLAGS) $(DESTLIBDIR)/$(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
+ INSTALL = && $(LN) $(LNFLAGS) $(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
endif
all: $(LIB)
diff --git a/src/SFML/Network/SocketTCP.cpp b/src/SFML/Network/SocketTCP.cpp index 03d96d3..6a200fc 100755 --- a/src/SFML/Network/SocketTCP.cpp +++ b/src/SFML/Network/SocketTCP.cpp @@ -131,11 +131,22 @@ Socket::Status SocketTCP::Connect(unsigned short Port, const IPAddress& HostAddr Time.tv_sec = static_cast<long>(Timeout);
Time.tv_usec = (static_cast<long>(Timeout * 1000) % 1000) * 1000;
- // Wait for something to write on our socket (would mean the connection has been accepted)
+ // Wait for something to write on our socket (which means that the connection request has returned)
if (select(static_cast<int>(mySocket + 1), NULL, &Selector, NULL, &Time) > 0)
{
- // Connection succeeded
- Status = Socket::Done;
+ // At this point the connection may have been either accepted or refused.
+ // To know whether it's a success or a failure, we try to retrieve the name of the connected peer
+ SocketHelper::LengthType Size = sizeof(SockAddr);
+ if (getpeername(mySocket, reinterpret_cast<sockaddr*>(&SockAddr), &Size) != -1)
+ {
+ // Connection accepted
+ Status = Socket::Done;
+ }
+ else
+ {
+ // Connection failed
+ Status = SocketHelper::GetErrorStatus();
+ }
}
else
{
@@ -333,11 +344,20 @@ Socket::Status SocketTCP::Receive(Packet& PacketToReceive) std::size_t Received = 0;
if (myPendingPacketSize < 0)
{
- Socket::Status Status = Receive(reinterpret_cast<char*>(&PacketSize), sizeof(PacketSize), Received);
- if (Status != Socket::Done)
- return Status;
+ // Loop until we've received the entire size of the packet
+ // (even a 4 bytes variable may be received in more than one call)
+ while (myPendingHeaderSize < sizeof(myPendingHeader))
+ {
+ char* Data = reinterpret_cast<char*>(&myPendingHeader) + myPendingHeaderSize;
+ Socket::Status Status = Receive(Data, sizeof(myPendingHeader) - myPendingHeaderSize, Received);
+ myPendingHeaderSize += Received;
+
+ if (Status != Socket::Done)
+ return Status;
+ }
- PacketSize = ntohl(PacketSize);
+ PacketSize = ntohl(myPendingHeader);
+ myPendingHeaderSize = 0;
}
else
{
@@ -461,6 +481,7 @@ void SocketTCP::Create(SocketHelper::SocketType Descriptor) myIsBlocking = true;
// Reset the pending packet
+ myPendingHeaderSize = 0;
myPendingPacket.clear();
myPendingPacketSize = -1;
diff --git a/src/SFML/Network/SocketUDP.cpp b/src/SFML/Network/SocketUDP.cpp index 0c977cd..c1fbc7b 100755 --- a/src/SFML/Network/SocketUDP.cpp +++ b/src/SFML/Network/SocketUDP.cpp @@ -244,20 +244,25 @@ Socket::Status SocketUDP::Send(Packet& PacketToSend, const IPAddress& Address, u ////////////////////////////////////////////////////////////
Socket::Status SocketUDP::Receive(Packet& PacketToReceive, IPAddress& Address, unsigned short& Port)
{
- // This is not safe at all, as data can be lost, duplicated, or arrive in a different order.
- // So if a packet is split into more than one chunk, nobody knows what could happen...
- // Conclusion : we shouldn't use packets with UDP, unless we build a more complex protocol on top of it.
-
// We start by getting the size of the incoming packet
Uint32 PacketSize = 0;
std::size_t Received = 0;
if (myPendingPacketSize < 0)
{
- Socket::Status Status = Receive(reinterpret_cast<char*>(&PacketSize), sizeof(PacketSize), Received, Address, Port);
- if (Status != Socket::Done)
- return Status;
+ // Loop until we've received the entire size of the packet
+ // (even a 4 bytes variable may be received in more than one call)
+ while (myPendingHeaderSize < sizeof(myPendingHeader))
+ {
+ char* Data = reinterpret_cast<char*>(&myPendingHeader) + myPendingHeaderSize;
+ Socket::Status Status = Receive(Data, sizeof(myPendingHeader) - myPendingHeaderSize, Received, Address, Port);
+ myPendingHeaderSize += Received;
- PacketSize = ntohl(PacketSize);
+ if (Status != Socket::Done)
+ return Status;
+ }
+
+ PacketSize = ntohl(myPendingHeader);
+ myPendingHeaderSize = 0;
}
else
{
@@ -265,9 +270,6 @@ Socket::Status SocketUDP::Receive(Packet& PacketToReceive, IPAddress& Address, u PacketSize = myPendingPacketSize;
}
- // Clear the user packet
- PacketToReceive.Clear();
-
// Use another address instance for receiving the packet data ;
// chunks of data coming from a different sender will be discarded (and lost...)
IPAddress Sender;
@@ -402,6 +404,7 @@ void SocketUDP::Create(SocketHelper::SocketType Descriptor) myPort = 0;
// Reset the pending packet
+ myPendingHeaderSize = 0;
myPendingPacket.clear();
myPendingPacketSize = -1;
diff --git a/src/SFML/System/Makefile b/src/SFML/System/Makefile index 04514f5..1cf6211 100755 --- a/src/SFML/System/Makefile +++ b/src/SFML/System/Makefile @@ -8,7 +8,7 @@ ifeq ($(STATIC), yes) else
LIB = libsfml-system.so
LIBNAME = $(LIBPATH)/$(LIB).$(VERSION)
- INSTALL = && $(LN) $(LNFLAGS) $(DESTLIBDIR)/$(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
+ INSTALL = && $(LN) $(LNFLAGS) $(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
endif
all: $(LIB)
diff --git a/src/SFML/System/Randomizer.cpp b/src/SFML/System/Randomizer.cpp index 93500bc..18bbe9c 100755 --- a/src/SFML/System/Randomizer.cpp +++ b/src/SFML/System/Randomizer.cpp @@ -26,8 +26,8 @@ // Headers
////////////////////////////////////////////////////////////
#include <SFML/System/Randomizer.hpp>
-#include <SFML/System/Platform.hpp>
-#include <cstdlib>
+#include <stdlib.h>
+#include <time.h>
namespace
@@ -35,7 +35,7 @@ namespace // Set the random numbers sequence seed with the current system time, so that it is always different
unsigned int SetRandomSeed()
{
- unsigned int Seed = static_cast<unsigned int>(sf::priv::Platform::GetSystemTime() * 1000);
+ unsigned int Seed = static_cast<unsigned int>(time(NULL));
srand(Seed);
return Seed;
}
diff --git a/src/SFML/System/Initializer.cpp b/src/SFML/System/Unix/Initializer.cpp index 0d3dbee..0d3dbee 100755 --- a/src/SFML/System/Initializer.cpp +++ b/src/SFML/System/Unix/Initializer.cpp diff --git a/src/SFML/System/Unix/Thread.cpp b/src/SFML/System/Unix/Thread.cpp index 3268f61..da02399 100755 --- a/src/SFML/System/Unix/Thread.cpp +++ b/src/SFML/System/Unix/Thread.cpp @@ -61,8 +61,7 @@ myUserData(UserData) Thread::~Thread()
{
// Wait for the thread to finish before destroying the instance
- if (myIsActive)
- Wait();
+ Wait();
}
@@ -71,6 +70,9 @@ Thread::~Thread() ////////////////////////////////////////////////////////////
void Thread::Launch()
{
+ // Wait for the thread to finish, in case it was already running
+ Wait();
+
// Create the thread
myIsActive = true;
int Error = pthread_create(&myThread, NULL, &Thread::ThreadFunc, this);
diff --git a/src/SFML/System/Win32/Thread.cpp b/src/SFML/System/Win32/Thread.cpp index d2d1d82..3252100 100755 --- a/src/SFML/System/Win32/Thread.cpp +++ b/src/SFML/System/Win32/Thread.cpp @@ -62,8 +62,7 @@ myUserData(UserData) Thread::~Thread()
{
// Wait for the thread to finish before destroying the instance
- if (myHandle)
- Wait();
+ Wait();
}
@@ -72,6 +71,9 @@ Thread::~Thread() ////////////////////////////////////////////////////////////
void Thread::Launch()
{
+ // Wait for the thread to finish, in case it was already running
+ Wait();
+
// Create the thread
myHandle = reinterpret_cast<HANDLE>(_beginthreadex(NULL, 0, &Thread::ThreadFunc, this, 0, NULL));
diff --git a/src/SFML/Window/Cocoa/AppController.h b/src/SFML/Window/Cocoa/AppController.h index f56db4e..de2d055 100755 --- a/src/SFML/Window/Cocoa/AppController.h +++ b/src/SFML/Window/Cocoa/AppController.h @@ -30,18 +30,16 @@ #import <Cocoa/Cocoa.h> -#define SharedAppController [AppController sharedController] - // Fade operations enum { FillScreen, CleanScreen }; -@class WindowWrapper; -@interface AppController : NSObject { +@class sfPrivWindow; +@interface sfPrivAppController : NSObject { BOOL myOwningEventLoop; - WindowWrapper *myFullscreenWrapper; + sfPrivWindow *myFullscreenWrapper; NSAutoreleasePool *myMainPool; sf::VideoMode myDesktopMode; sf::VideoMode myPrevMode; @@ -50,7 +48,17 @@ enum { //////////////////////////////////////////////////////////// /// Return the shared AppController instance. Make one if needed. //////////////////////////////////////////////////////////// -+ (AppController *)sharedController; ++ (sfPrivAppController *)sharedController; + +//////////////////////////////////////////////////////////// +/// Returns the primay computer's screen +//////////////////////////////////////////////////////////// ++ (CGDirectDisplayID)primaryScreen; + +//////////////////////////////////////////////////////////// +/// Reset notifictions about application focus +//////////////////////////////////////////////////////////// +- (void)setNotifications; //////////////////////////////////////////////////////////// /// Make the menu bar @@ -68,7 +76,7 @@ enum { /// Set @window as the current fullscreen window /// Change the screen resolution if needed according to @window and @fullscreenMode //////////////////////////////////////////////////////////// -- (void)setFullscreenWindow:(WindowWrapper *)window mode:(sf::VideoMode *)fullscreenMode; +- (void)setFullscreenWindow:(sfPrivWindow *)window mode:(sf::VideoMode *)fullscreenMode; //////////////////////////////////////////////////////////// /// Perform fade operation where 'operation' is one of { FillScreen, CleanScreen} diff --git a/src/SFML/Window/Cocoa/AppController.mm b/src/SFML/Window/Cocoa/AppController.mm index 8d07eaf..9356090 100755 --- a/src/SFML/Window/Cocoa/AppController.mm +++ b/src/SFML/Window/Cocoa/AppController.mm @@ -33,9 +33,6 @@ #import <iostream> -// AppController singleton object -static AppController *shared = nil; - /* setAppleMenu disappeared from the headers in 10.4 */ #if MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_4 @@ -60,7 +57,7 @@ static AppController *shared = nil; @end -@implementation AppController +@implementation sfPrivAppController //////////////////////////////////////////////////////////// @@ -83,6 +80,11 @@ static AppController *shared = nil; // Make the app autorelease pool myMainPool = [[NSAutoreleasePool alloc] init]; + if (!myMainPool) { + [self release]; + throw std::bad_alloc(); + } + // Don't go on if the user handles the app if (![NSApp isRunning]) { @@ -96,27 +98,12 @@ static AppController *shared = nil; } // Make the app - [NSApplication sharedApplication]; - - NSNotificationCenter *nc = [NSNotificationCenter defaultCenter]; - // I want to go back to the desktop mode - // if we've a fullscreen window when hiding - [nc addObserver:self - selector:@selector(applicationWillHide:) - name:NSApplicationWillHideNotification - object:NSApp]; - - // And restore de fullscreen mode when unhiding - [nc addObserver:self - selector:@selector(applicationWillUnhide:) - name:NSApplicationWillUnhideNotification - object:NSApp]; + if (![NSApplication sharedApplication]) { + [self release]; + throw std::bad_alloc(); + } - // Go back to desktop mode before exit - [nc addObserver:self - selector:@selector(applicationWillTerminate:) - name:NSApplicationWillTerminateNotification - object:NSApp]; + [self setNotifications]; if ([NSApp mainMenu] == nil) { [self makeMenuBar]; @@ -134,7 +121,7 @@ static AppController *shared = nil; - (void)dealloc { [[NSNotificationCenter defaultCenter] removeObserver:self]; - [myFullscreenWrapper release]; + [myMainPool release]; [super dealloc]; } @@ -142,40 +129,80 @@ static AppController *shared = nil; //////////////////////////////////////////////////////////// /// Return the shared AppController instance. Make one if needed. //////////////////////////////////////////////////////////// -+ (AppController *)sharedController ++ (sfPrivAppController *)sharedController { - if (nil == shared) - shared = [[AppController alloc] init]; - + // AppController singleton object + static sfPrivAppController *shared = [[sfPrivAppController alloc] init]; return shared; } //////////////////////////////////////////////////////////// +/// Returns the primay computer's screen +//////////////////////////////////////////////////////////// ++ (CGDirectDisplayID)primaryScreen +{ + static BOOL firstTime = YES; + static CGDirectDisplayID screen = kCGNullDirectDisplay; + + if (firstTime) { + CGDisplayCount numScr; + CGDisplayErr err = CGGetDisplaysWithPoint(CGPointMake(0, 0), 1, &screen, &numScr); + + if (err != kCGErrorSuccess || numScr < 1) { + std::cerr << "Unable to get primary screen (error code " << err + << " ). Using the main screen."; + screen = CGMainDisplayID(); + } + firstTime = NO; + } + + return screen; +} + + +//////////////////////////////////////////////////////////// +/// Reset notifictions about application focus +//////////////////////////////////////////////////////////// +- (void)setNotifications +{ + NSNotificationCenter *nc = [NSNotificationCenter defaultCenter]; + [nc addObserver:self + selector:@selector(applicationWillDeactivate:) + name:NSApplicationWillResignActiveNotification + object:NSApp]; + [nc addObserver:self + selector:@selector(applicationWillActivate:) + name:NSApplicationWillBecomeActiveNotification + object:NSApp]; + + [nc addObserver:self + selector:@selector(applicationWillTerminate:) + name:NSApplicationWillTerminateNotification + object:NSApp]; +} + + +//////////////////////////////////////////////////////////// /// Hide all the fullscreen windows and switch back to the desktop display mode //////////////////////////////////////////////////////////// -- (void)applicationWillHide:(NSNotification *)aNotification +- (void)applicationWillDeactivate:(NSNotification *)aNotification { + // Note: not using fading because it produces reactivation issues if (myFullscreenWrapper) { myPrevMode = sf::VideoMode::GetDesktopMode(); - CFDictionaryRef displayMode = CGDisplayBestModeForParameters (kCGDirectMainDisplay, + CFDictionaryRef displayMode = CGDisplayBestModeForParameters ([sfPrivAppController primaryScreen], myDesktopMode.BitsPerPixel, myDesktopMode.Width, myDesktopMode.Height, NULL); - // Fade to black screen - [self doFadeOperation:FillScreen time:0.2f sync:true]; - // Make the full screen window unvisible [[myFullscreenWrapper window] setAlphaValue:0.0f]; // Switch to the wished display mode - CGDisplaySwitchToMode(kCGDirectMainDisplay, displayMode); - - // Fade to normal screen - [self doFadeOperation:CleanScreen time:0.5f sync:false]; + CGDisplaySwitchToMode([sfPrivAppController primaryScreen], displayMode); } } @@ -183,31 +210,26 @@ static AppController *shared = nil; //////////////////////////////////////////////////////////// /// Unhide all the fullscreen windows and switch to full screen display mode //////////////////////////////////////////////////////////// -- (void)applicationWillUnhide:(NSNotification *)aNotification +- (void)applicationWillActivate:(NSNotification *)aNotification { if (myFullscreenWrapper) { - CFDictionaryRef displayMode = CGDisplayBestModeForParameters (kCGDirectMainDisplay, + CFDictionaryRef displayMode = CGDisplayBestModeForParameters ([sfPrivAppController primaryScreen], myPrevMode.BitsPerPixel, myPrevMode.Width, myPrevMode.Height, NULL); - - // Fade to a black screen - [self doFadeOperation:FillScreen time:0.5f sync:true]; [NSMenu setMenuBarVisible:NO]; // Switch to the wished display mode - CGDisplaySwitchToMode(kCGDirectMainDisplay, displayMode); + CGDisplaySwitchToMode([sfPrivAppController primaryScreen], displayMode); // Show the fullscreen window if existing if (myFullscreenWrapper) { - [[myFullscreenWrapper window] setAlphaValue:1.0f]; - [[myFullscreenWrapper window] center]; + [[myFullscreenWrapper window] setAlphaValue:1.0f]; + [[myFullscreenWrapper window] center]; + [[myFullscreenWrapper window] makeKeyAndOrderFront:self]; } - - // Fade to normal screen - [self doFadeOperation:CleanScreen time:0.5f sync:false]; } } @@ -216,10 +238,6 @@ static AppController *shared = nil; { if (myFullscreenWrapper) [self setFullscreenWindow:nil mode:NULL]; - - // FIXME: should I really do this ? what about the user owned windows ? - // And is this really useful as the application is about to exit ? - [NSApp makeWindowsPerform:@selector(close) inOrder:NO]; } //////////////////////////////////////////////////////////// @@ -228,7 +246,6 @@ static AppController *shared = nil; - (void)makeMenuBar { // Source taken from SDL 1.3 - NSString *appName = nil; NSString *title = nil; NSMenu *appleMenu = nil; @@ -266,10 +283,10 @@ static AppController *shared = nil; keyEquivalent:@"h"]; // + 'Hide other' menu item - menuItem = reinterpret_cast <NSMenuItem *> ([appleMenu addItemWithTitle:@"Hide Others" - action:@selector(hideOtherApplications:) - keyEquivalent:@"h"]); - [menuItem setKeyEquivalentModifierMask:(NSAlternateKeyMask|NSCommandKeyMask)]; + menuItem = (NSMenuItem *)[appleMenu addItemWithTitle:@"Hide Others" + action:@selector(hideOtherApplications:) + keyEquivalent:@"h"]; + [menuItem setKeyEquivalentModifierMask:NSAlternateKeyMask | NSCommandKeyMask]; // + 'Show all' menu item [appleMenu addItemWithTitle:@"Show All" @@ -363,14 +380,14 @@ static AppController *shared = nil; myOwningEventLoop = YES; } - // Clean the autorelease pool - [myMainPool release]; - myMainPool = [[NSAutoreleasePool alloc] init]; - NSEvent *event = nil; if (myOwningEventLoop) { + // Clean the autorelease pool + [myMainPool release]; + myMainPool = [[NSAutoreleasePool alloc] init]; + // Minimal event loop while (nil != (event = [NSApp nextEventMatchingMask:NSAnyEventMask untilDate:nil @@ -387,13 +404,13 @@ static AppController *shared = nil; /// Set @window as the current fullscreen window /// Change the screen resolution if needed according to @window and @fullscreenMode //////////////////////////////////////////////////////////// -- (void)setFullscreenWindow:(WindowWrapper *)aWrapper mode:(sf::VideoMode *)fullscreenMode +- (void)setFullscreenWindow:(sfPrivWindow *)aWindow mode:(sf::VideoMode *)fullscreenMode { // If we have a fullscreen window and want to remove it - if (aWrapper == nil && myFullscreenWrapper) + if (aWindow == nil && myFullscreenWrapper) { // Get the CoreGraphics display mode according to the desktop mode - CFDictionaryRef displayMode = CGDisplayBestModeForParameters (kCGDirectMainDisplay, + CFDictionaryRef displayMode = CGDisplayBestModeForParameters ([sfPrivAppController primaryScreen], myDesktopMode.BitsPerPixel, myDesktopMode.Width, myDesktopMode.Height, @@ -405,7 +422,7 @@ static AppController *shared = nil; #endif // Switch to the desktop display mode - CGDisplaySwitchToMode(kCGDirectMainDisplay, displayMode); + CGDisplaySwitchToMode([sfPrivAppController primaryScreen], displayMode); // Close the window [[myFullscreenWrapper window] close]; @@ -421,12 +438,13 @@ static AppController *shared = nil; // Release the saved window wrapper myFullscreenWrapper = nil; } - else if (aWrapper) + else if (aWindow) { + // else if we want to SET fullscreen assert(fullscreenMode != NULL); // Get the CoreGraphics display mode according to the given sf mode - CFDictionaryRef displayMode = CGDisplayBestModeForParameters (kCGDirectMainDisplay, + CFDictionaryRef displayMode = CGDisplayBestModeForParameters ([sfPrivAppController primaryScreen], fullscreenMode->BitsPerPixel, fullscreenMode->Width, fullscreenMode->Height, @@ -446,7 +464,8 @@ static AppController *shared = nil; if (myPrevMode != *fullscreenMode) { // Switch to the wished display mode - CGDisplaySwitchToMode(kCGDirectMainDisplay, displayMode); + myPrevMode = *fullscreenMode; + CGDisplaySwitchToMode([sfPrivAppController primaryScreen], displayMode); } if (myFullscreenWrapper) @@ -455,8 +474,8 @@ static AppController *shared = nil; } // Open and center the window - [[aWrapper window] makeKeyAndOrderFront:nil]; - [[aWrapper window] center]; + [[aWindow window] makeKeyAndOrderFront:nil]; + [[aWindow window] center]; #if ENABLE_FADE_OPERATIONS // Fade to normal screen @@ -464,11 +483,11 @@ static AppController *shared = nil; #endif // Save the fullscreen wrapper - myFullscreenWrapper = aWrapper; + myFullscreenWrapper = aWindow; } else { - std::cerr << "Inconcistency error for arguments given to -[AppController setFullscreenWindow:mode:]" << std::endl; + std::cerr << "Inconcistency error for arguments given to -[sfPrivAppController setFullscreenWindow:mode:]" << std::endl; } } @@ -495,7 +514,7 @@ static AppController *shared = nil; if (!result) { // Capture display but do not fill the screen with black // so that we can see the fade operation - capture = CGDisplayCaptureWithOptions(kCGDirectMainDisplay, kCGCaptureNoFill); + capture = CGDisplayCaptureWithOptions([sfPrivAppController primaryScreen], kCGCaptureNoFill); if (!capture) { // Do the increasing fade operation @@ -505,11 +524,11 @@ static AppController *shared = nil; 0.0f, 0.0f, 0.0f, sync); // Now, release the non black-filling capture - CGDisplayRelease(kCGDirectMainDisplay); + CGDisplayRelease([sfPrivAppController primaryScreen]); // And capture with filling // so that we don't see the switching in the meantime - CGDisplayCaptureWithOptions(kCGDirectMainDisplay, kCGCaptureNoOptions); + CGDisplayCaptureWithOptions([sfPrivAppController primaryScreen], kCGCaptureNoOptions); } prevToken = token; @@ -522,10 +541,10 @@ static AppController *shared = nil; if (!result) { if (!capture) { // Release the black-filling capture - CGDisplayRelease(kCGDirectMainDisplay); + CGDisplayRelease([sfPrivAppController primaryScreen]); // Capture the display but do not fill with black (still for the fade operation) - CGDisplayCaptureWithOptions(kCGDirectMainDisplay, kCGCaptureNoFill); + CGDisplayCaptureWithOptions([sfPrivAppController primaryScreen], kCGCaptureNoFill); // Do the decreasing fading CGDisplayFade(token, time, @@ -541,7 +560,7 @@ static AppController *shared = nil; } // Release the captured display - CGDisplayRelease(kCGDirectMainDisplay); + CGDisplayRelease([sfPrivAppController primaryScreen]); } } } diff --git a/src/SFML/Window/Cocoa/GLKit.h b/src/SFML/Window/Cocoa/GLKit.h index ac12252..f4cff4a 100755 --- a/src/SFML/Window/Cocoa/GLKit.h +++ b/src/SFML/Window/Cocoa/GLKit.h @@ -32,10 +32,7 @@ //////////////////////////////////////////////////////////// /// Window independant OpenGL context class //////////////////////////////////////////////////////////// -@interface GLContext : NSOpenGLContext -{ - GLContext *mySharedContext; -} +@interface sfPrivGLContext : NSOpenGLContext //////////////////////////////////////////////////////////// /// Return the shared OpenGL context instance (making one if needed) @@ -47,7 +44,8 @@ /// and the shared context @context //////////////////////////////////////////////////////////// - (id)initWithAttributes:(sf::WindowSettings&)attribs - sharedContext:(GLContext *)context; + mode:(const sf::VideoMode&)mode + sharedContext:(sfPrivGLContext *)context; @end @@ -55,22 +53,31 @@ //////////////////////////////////////////////////////////// /// Customized Cocoa OpenGL view //////////////////////////////////////////////////////////// -@interface GLView : NSOpenGLView +@interface sfPrivGLView : NSOpenGLView { sf::priv::WindowImplCocoa *myDelegate; - GLContext *myGLContext; + sfPrivGLContext *myGLContext; } //////////////////////////////////////////////////////////// /// Make a new view according the the rect @frame, /// the video mode @mode, the window settings @settings -/// and the sf window delegate @delegate /// @delegate must not be null //////////////////////////////////////////////////////////// - (id)initWithFrame:(NSRect)frame mode:(const sf::VideoMode&)mode - settings:(sf::WindowSettings&)settings - delegate:(sf::priv::WindowImplCocoa *)delegate; + settings:(sf::WindowSettings&)settings; + + +//////////////////////////////////////////////////////////// +/// Sets @aDelegate as the view delegate +//////////////////////////////////////////////////////////// +- (void)setDelegate:(sf::priv::WindowImplCocoa *)aDelegate; + +//////////////////////////////////////////////////////////// +/// Returns the view delegate +//////////////////////////////////////////////////////////// +- (sf::priv::WindowImplCocoa *)delegate; //////////////////////////////////////////////////////////// /// Finish view setting (after having added it to the window) @@ -95,75 +102,34 @@ @end //////////////////////////////////////////////////////////// -/// Cocoa window implementation to let fullscreen windows -/// catch key events -//////////////////////////////////////////////////////////// -@interface GLWindow : NSWindow - -//////////////////////////////////////////////////////////// -/// Technical note: this class must neither contain new members -/// nor methods. It is used transparently as a NSWindow object -/// by WindowWrapper. Not following this rule could result -/// in a segmentation fault or data corruption. -//////////////////////////////////////////////////////////// - -@end - +/// Parent class for handling general SFML window stuff //////////////////////////////////////////////////////////// -/// WindowWrapper class : handles both imported and self-built windows -//////////////////////////////////////////////////////////// -@interface WindowWrapper : NSObject +@interface sfPrivWindow : NSObject { - GLWindow *myWindow; - GLView *myView; - sf::VideoMode myFullscreenMode; - bool myIsFullscreen; +@private + NSWindow *myWindow; + sfPrivGLView *myView; } //////////////////////////////////////////////////////////// -/// Make a new window wrapper according to the window settings @attribs, -/// the video mode @mode, the window style @style, the window title @title -/// and the sf window implementation delegate @delegate -//////////////////////////////////////////////////////////// -- (id)initWithSettings:(sf::WindowSettings&)attribs - videoMode:(sf::VideoMode&)mode - style:(unsigned long)style - title:(NSString *)title - delegate:(sf::priv::WindowImplCocoa *)delegate; - -//////////////////////////////////////////////////////////// -/// Make a new window wrapper by importing @window and according to -/// the window settings @params and the sf window implementation delegate -/// @delegate -/// @window and @delegate must not be null +/// Return a reference to the internal Cocoa window //////////////////////////////////////////////////////////// -- (id)initWithWindow:(NSWindow *)window - settings:(sf::WindowSettings&)params - delegate:(sf::priv::WindowImplCocoa *)delegate; +- (NSWindow *)window; //////////////////////////////////////////////////////////// -/// Make a new window wrapper by importing @window if it's not null and according to -/// the window settings @params and the sf window implementation delegate -/// @delegate; or by creating a new window if @window is null. In this case @title -/// must therefore not be null and @params must be valid. -/// @delegate must never be null +/// Return a reference to the internal Cocoa OpenGL view //////////////////////////////////////////////////////////// -- (id)initWithWindow:(NSWindow *)window - settings:(sf::WindowSettings&)params - videoMode:(sf::VideoMode&)mode - style:(unsigned long)style - title:(NSString *)title - delegate:(sf::priv::WindowImplCocoa *)delegate; +- (sfPrivGLView *)view; //////////////////////////////////////////////////////////// -/// Return a reference to the internal Cocoa window +/// Sets @aDelegate as the window delegate //////////////////////////////////////////////////////////// -- (NSWindow *)window; +- (void)setDelegate:(sf::priv::WindowImplCocoa *)aDelegate; //////////////////////////////////////////////////////////// -/// Return a reference to the internal Cocoa OpenGL view +/// Returns the window delegate //////////////////////////////////////////////////////////// -- (GLView *)glView; +- (sf::priv::WindowImplCocoa *)delegate; //////////////////////////////////////////////////////////// /// Forward call to set the window position on screen @@ -207,3 +173,61 @@ @end +//////////////////////////////////////////////////////////// +/// Class for creating new SFML windows from informations +//////////////////////////////////////////////////////////// +@interface sfPrivOwnedWindow : sfPrivWindow +{ +@private + sf::VideoMode myFullscreenMode; + bool myIsFullscreen; +} + +//////////////////////////////////////////////////////////// +/// Creates and returns a new SFML window handler with +/// the given parameters +//////////////////////////////////////////////////////////// +- (id)initWithVideoMode:(sf::VideoMode&)aMode + settings:(sf::WindowSettings&)someSettings + style:(unsigned long)aStyle + title:(NSString *)aTitle; + +//////////////////////////////////////////////////////////// +/// Returns the window's fullscreen state +//////////////////////////////////////////////////////////// +- (BOOL)isFullscreen; + +@end + + +//////////////////////////////////////////////////////////// +/// Class for creating SFML windows from Cocoa windows +//////////////////////////////////////////////////////////// +@interface sfPrivImportedWindow : sfPrivWindow + +//////////////////////////////////////////////////////////// +/// Returns a new SFML window handler with the given window +/// and parameters +//////////////////////////////////////////////////////////// +- (id)initWithWindow:(NSWindow *)aWindow + settings:(sf::WindowSettings&)someSettings; + +@end + + +//////////////////////////////////////////////////////////// +/// Class for creating SFML windows from Cocoa views +//////////////////////////////////////////////////////////// +@interface sfPrivImportedView : sfPrivWindow +{ + NSView *parentView; +} + +//////////////////////////////////////////////////////////// +/// Returns a new SFML window handler with the given view +/// and parameters +//////////////////////////////////////////////////////////// +- (id)initWithView:(NSView *)aView + settings:(sf::WindowSettings&)someSettings; + +@end diff --git a/src/SFML/Window/Cocoa/GLKit.mm b/src/SFML/Window/Cocoa/GLKit.mm index a570f44..f19fa77 100755 --- a/src/SFML/Window/Cocoa/GLKit.mm +++ b/src/SFML/Window/Cocoa/GLKit.mm @@ -29,36 +29,62 @@ #import <SFML/Window/Cocoa/AppController.h> #import <SFML/Window/VideoMode.hpp> #import <SFML/Window/WindowStyle.hpp> -#import <SFML/System/Sleep.hpp> #import <OpenGL/gl.h> #import <iostream> +@interface sfPrivWindow (Protected) + +//////////////////////////////////////////////////////////// +/// Registers a reference to the internal Cocoa window +//////////////////////////////////////////////////////////// +- (void)setWindow:(NSWindow *)aWindow; + +//////////////////////////////////////////////////////////// +/// Registers the the OpenGL view and adds it to its parent container +//////////////////////////////////////////////////////////// +- (void)putOpenGLView:(sfPrivGLView *)aView; + +//////////////////////////////////////////////////////////// +/// Registers a reference to the internal Cocoa OpenGL view +//////////////////////////////////////////////////////////// +- (void)setView:(sfPrivGLView *)aView; + +@end + +@interface sfPrivOwnedWindow (Private) + +//////////////////////////////////////////////////////////// +/// Sets the window's fullscreen state +//////////////////////////////////////////////////////////// +- (void)setFullscreen:(BOOL)aFlag; + +@end + + //////////////////////////////////////////////////////////// /// Window independant OpenGL context class //////////////////////////////////////////////////////////// -@implementation GLContext +@implementation sfPrivGLContext -static GLContext *sharedCtx = nil; //////////////////////////////////////////////////////////// /// Return the shared OpenGL context instance (making one if needed) //////////////////////////////////////////////////////////// + (id)sharedContext { - if (sharedCtx == nil) - { - // Make a new context with the default parameters - sf::WindowSettings params(0, 0, 0); - sharedCtx = [[GLContext alloc] initWithAttributes:params sharedContext:nil]; - } - + // Make a new context with the default parameters + sf::WindowSettings params; + sf::VideoMode m; + static sfPrivGLContext *sharedCtx = [[sfPrivGLContext alloc] initWithAttributes:params + mode:m + sharedContext:nil]; return sharedCtx; } - (void)dealloc { - [mySharedContext release]; + [[sfPrivGLContext sharedContext] release]; [super dealloc]; } @@ -66,7 +92,7 @@ static GLContext *sharedCtx = nil; /// Make a new OpenGL context according to the @attribs settings /// and the shared context @context //////////////////////////////////////////////////////////// -- (id)initWithAttributes:(sf::WindowSettings&)attribs sharedContext:(GLContext *)context +- (id)initWithAttributes:(sf::WindowSettings&)attribs mode:(const sf::VideoMode&)mode sharedContext:(sfPrivGLContext *)sharedContext { // Note about antialiasing and other context attributes : // OpenGL context sharing does not allow the shared contexts to use different attributes. @@ -77,8 +103,13 @@ static GLContext *sharedCtx = nil; // // I've no way to fix this for now. - if (attribs.AntialiasingLevel) - std::cerr << "Warning: antialiasing settings are inhibited under Mac OS X for technical reasons" << std::endl; + if (attribs.AntialiasingLevel) { + NSLog(@"Warning: antialiasing settings are inhibited on Mac OS X for technical reasons"); + } + + if (attribs.StencilBits != 8) { + NSLog(@"Warning: stencil bits settings are inhibited on Mac OS X for technical reasons"); + } NSOpenGLPixelFormat *myPixelFormat = nil; unsigned idx = 0; @@ -91,28 +122,44 @@ static GLContext *sharedCtx = nil; ctxtAttribs[idx++] = NSOpenGLPFADoubleBuffer; ctxtAttribs[idx++] = NSOpenGLPFAAccelerated; + // Force use of first screen + //ctxtAttribs[idx++] = NSOpenGLPFAScreenMask; + //ctxtAttribs[idx++] = CGDisplayIDToOpenGLDisplayMask([sfPrivAppController primaryScreen]); + // windowed context (even fullscreen mode uses a window) ctxtAttribs[idx++] = NSOpenGLPFAWindow; - // Color size ; usually 32 bits per pixel + // Color buffer bits ; usually 32 bits per pixel ctxtAttribs[idx++] = NSOpenGLPFAColorSize; - ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) sf::VideoMode::GetDesktopMode().BitsPerPixel; - // Z-buffer size + // well.. whatever I put here, the color size is always 24 bits on my computer.. + ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) mode.BitsPerPixel; + + // Alpha buffer size + // FIXME: I don't really remember whether I can safely use the user video mode settings + // to set this attribute (because of the shared context attributes conflicts) + if (mode.BitsPerPixel > 24) { + ctxtAttribs[idx++] = NSOpenGLPFAAlphaSize; + ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) 8; + } + + // Depth buffer size ctxtAttribs[idx++] = NSOpenGLPFADepthSize; ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) attribs.DepthBits; - // Stencil bits (I don't really know what's that...) + // Stencil buffer bits + // Note: even with "NSOpenGLPFAClosestPolicy" set, the NSOpenGLPFAStencilSize value must be an valid and exact one + // I've been looking for possible diffferent default stencil bits values among different Apple computers + // but I can't again get the website I found oneaday + // (describing all the graphic cards used by Apple, the supported extensions, etc). ctxtAttribs[idx++] = NSOpenGLPFAStencilSize; - ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) attribs.StencilBits; + ctxtAttribs[idx++] = (NSOpenGLPixelFormatAttribute) /* attribs.StencilBits */ 8; myPixelFormat = [[NSOpenGLPixelFormat alloc] initWithAttributes:ctxtAttribs]; if (myPixelFormat) { self = [super initWithFormat:myPixelFormat - shareContext:context]; - - mySharedContext = [context retain]; + shareContext:[sharedContext retain]]; // Get the effective properties from our OpenGL context GLint tmpDepthSize = 0, tmpStencilBits = 0, tmpAntialiasingLevel = 0; @@ -137,6 +184,8 @@ static GLContext *sharedCtx = nil; attribs.AntialiasingLevel = (unsigned) tmpAntialiasingLevel; [myPixelFormat release]; + } else { + NSLog(@"NSOpenGLPixelFormat creation failed! (invalid video settings ?)"); } return self; @@ -148,7 +197,7 @@ static GLContext *sharedCtx = nil; //////////////////////////////////////////////////////////// /// Customized Cocoa OpenGL view //////////////////////////////////////////////////////////// -@implementation GLView +@implementation sfPrivGLView //////////////////////////////////////////////////////////// /// Make a new view according the the rect @frame, @@ -159,10 +208,7 @@ static GLContext *sharedCtx = nil; - (id)initWithFrame:(NSRect)frame mode:(const sf::VideoMode&)mode settings:(sf::WindowSettings&)settings - delegate:(sf::priv::WindowImplCocoa *)delegate { - assert(delegate != NULL); - // make the view self = [super initWithFrame:frame pixelFormat:nil]; @@ -172,7 +218,14 @@ static GLContext *sharedCtx = nil; [self setAutoresizingMask:NSViewWidthSizable | NSViewHeightSizable]; // make the OpenGL context - myGLContext = [[GLContext alloc] initWithAttributes:settings sharedContext:sharedCtx]; + myGLContext = [[sfPrivGLContext alloc] initWithAttributes:settings + mode:mode + sharedContext:[sfPrivGLContext sharedContext]]; + + if (!myGLContext) { + [self release]; + return nil; + } // We need to update the OpenGL view when it's resized NSNotificationCenter *nc = [NSNotificationCenter defaultCenter]; @@ -180,9 +233,6 @@ static GLContext *sharedCtx = nil; selector:@selector(viewFrameDidChange:) name:NSViewFrameDidChangeNotification object:self]; - - // Save the delegate - myDelegate = delegate; } return self; @@ -201,6 +251,22 @@ static GLContext *sharedCtx = nil; [super dealloc]; } +//////////////////////////////////////////////////////////// +/// Sets @aDelegate as the view delegate +//////////////////////////////////////////////////////////// +- (void)setDelegate:(sf::priv::WindowImplCocoa *)aDelegate +{ + myDelegate = aDelegate; +} + +//////////////////////////////////////////////////////////// +/// Returns the view delegate +//////////////////////////////////////////////////////////// +- (sf::priv::WindowImplCocoa *)delegate +{ + return myDelegate; +} + //////////////////////////////////////////////////////////// /// Finish view setting (after having added it to the window) @@ -287,9 +353,12 @@ static GLContext *sharedCtx = nil; { assert(myDelegate != NULL); - NSText *field = [[self window] fieldEditor:YES forObject:nil]; - [field interpretKeyEvents:[NSArray arrayWithObject:theEvent]]; - [field setString:@""]; + if (sf::priv::WindowImplCocoa::IsTextEvent((void *)theEvent)) + { + NSText *field = [[self window] fieldEditor:YES forObject:nil]; + [field interpretKeyEvents:[NSArray arrayWithObject:theEvent]]; + [field setString:@""]; + } myDelegate->HandleKeyDown(theEvent); } @@ -414,10 +483,10 @@ static GLContext *sharedCtx = nil; //////////////////////////////////////////////////////////// -/// Cocoa window implementation to let fullscreen windows +/// Cocoa window category to let fullscreen windows /// catch key events //////////////////////////////////////////////////////////// -@implementation GLWindow +@implementation NSWindow (GLWindow) - (BOOL)canBecomeKeyWindow { @@ -432,309 +501,188 @@ static GLContext *sharedCtx = nil; @end + +@implementation sfPrivWindow + +- (void)dealloc +{ + NSAutoreleasePool *localPool = [[NSAutoreleasePool alloc] init]; + // Remove the notification observer + [[NSNotificationCenter defaultCenter] removeObserver:self]; + + // Close the window + [self show:false]; + + // Release the window and view + [[self view] removeFromSuperviewWithoutNeedingDisplay]; + + [self setWindow:nil]; + [self setView:nil]; + [self setDelegate:nil]; + + [super dealloc]; + [localPool release]; +} + //////////////////////////////////////////////////////////// -/// WindowWrapper class : handles both imported and self-built windows +/// Registers a reference to the internal Cocoa window //////////////////////////////////////////////////////////// -@implementation WindowWrapper +- (void)setWindow:(NSWindow *)aWindow +{ + if (myWindow != aWindow) + { + NSNotificationCenter *nc = [NSNotificationCenter defaultCenter]; + + // Drop the observers on the previously set window + if ([self window]) { + [nc removeObserver:self + name:NSWindowDidBecomeMainNotification + object:[self window]]; + [nc removeObserver:self + name:NSWindowDidResignMainNotification + object:[self window]]; + [nc removeObserver:self + name:NSWindowWillCloseNotification + object:[self window]]; + } + + [myWindow release]; + myWindow = [aWindow retain]; + + // Set the new observers + // We want to know when our window got the focus + [nc addObserver:self + selector:@selector(windowDidBecomeMain:) + name:NSWindowDidBecomeMainNotification + object:[self window]]; + + // We want to know when our window lost the focus + [nc addObserver:self + selector:@selector(windowDidResignMain:) + name:NSWindowDidResignMainNotification + object:[self window]]; + + // We want to know when the user closes the window + [nc addObserver:self + selector:@selector(windowWillClose:) + name:NSWindowWillCloseNotification + object:[self window]]; + + } +} //////////////////////////////////////////////////////////// -/// Make a new window wrapper according to the window settings @attribs, -/// the video mode @mode, the window style @style, the window title @title -/// and the sf window implementation delegate @delegate +/// Return a reference to the internal Cocoa window //////////////////////////////////////////////////////////// -- (id)initWithSettings:(sf::WindowSettings&)params - videoMode:(sf::VideoMode&)mode - style:(unsigned long)style - title:(NSString *)title - delegate:(sf::priv::WindowImplCocoa *)delegate +- (NSWindow *)window { - return [self initWithWindow:nil - settings:params - videoMode:mode - style:style - title:title - delegate:delegate]; + return [[myWindow retain] autorelease]; } //////////////////////////////////////////////////////////// -/// Make a new window wrapper by importing @window and according to -/// the window settings @params and the sf window implementation delegate -/// @delegate -/// @window and @delegate must not be null +/// Registers the the OpenGL view and adds it to its parent container //////////////////////////////////////////////////////////// -- (id)initWithWindow:(NSWindow *)window - settings:(sf::WindowSettings&)params - delegate:(sf::priv::WindowImplCocoa *)delegate +- (void)putOpenGLView:(sfPrivGLView *)aView { - sf::VideoMode mode([[myWindow contentView] frame].size.width, [[myWindow contentView] frame].size.height); - return [self initWithWindow:window - settings:params - videoMode:mode - style:0 - title:nil - delegate:delegate]; + [self setView:aView]; + + // Finish setting up the view and window + // Add the view to our window and tell it to the view + [[[self window] contentView] addSubview:[self view]]; + [[self view] finishInitialization]; } - //////////////////////////////////////////////////////////// -/// Make a new window wrapper by importing @window if it's not null and according to -/// the window settings @params and the sf window implementation delegate -/// @delegate; or by creating a new window if @window is null. In this case @title -/// must therefore not be null and @params must be valid. -/// @delegate must never be null +/// Registers a reference to the internal Cocoa OpenGL view //////////////////////////////////////////////////////////// -- (id)initWithWindow:(NSWindow *)window - settings:(sf::WindowSettings&)params - videoMode:(sf::VideoMode&)mode - style:(unsigned long)style - title:(NSString *)title - delegate:(sf::priv::WindowImplCocoa *)delegate +- (void)setView:(sfPrivGLView *)aView { - assert(delegate != NULL); - - self = [super init]; - - if (self) + if (myView != aView) { - if (window) { - myWindow = (GLWindow *)[window retain]; - } else { - assert(title != nil); - - NSRect frame = NSMakeRect (0.0f, 0.0f, (float) mode.Width, (float) mode.Height); - unsigned int mask = 0; - - if (style & sf::Style::Fullscreen) { - myIsFullscreen = true; - - // Check display mode and put new values in 'mode' if needed - boolean_t exact = true; - - CFDictionaryRef properties = CGDisplayBestModeForParameters(kCGDirectMainDisplay, mode.BitsPerPixel, - mode.Width, mode.Height, &exact); - - if (!properties) { - std::cerr << "Unable to get a display mode with the given parameters" << std::endl; - [self autorelease]; - return nil; - } - - if (exact == false) { - CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, kCGDisplayWidth), - kCFNumberIntType, &mode.Width); - - CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, kCGDisplayHeight), - kCFNumberIntType, &mode.Height); - - CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, kCGDisplayBitsPerPixel), - kCFNumberIntType, &mode.BitsPerPixel); - - } - } - - // We grab options from WindowStyle and add them to our window mask - if (style & sf::Style::None || style & sf::Style::Fullscreen) { - mask |= NSBorderlessWindowMask; - - - - } else { - if (style & sf::Style::Titlebar) { - mask |= NSTitledWindowMask; - mask |= NSMiniaturizableWindowMask; - } - - if (style & sf::Style::Resize) { - mask |= NSTitledWindowMask; - mask |= NSMiniaturizableWindowMask; - mask |= NSResizableWindowMask; - } - - if (style & sf::Style::Close) { - mask |= NSTitledWindowMask; - mask |= NSClosableWindowMask; - mask |= NSMiniaturizableWindowMask; - } - } - - // Now we make the window with the values we got - // Note: defer flag set to NO to be able to use OpenGL in our window - myWindow = [[GLWindow alloc] initWithContentRect:frame - styleMask:mask - backing:NSBackingStoreBuffered - defer:NO]; - - if (myWindow) { - // We set title and window position - [myWindow setTitle:title]; - [myWindow center]; - } else { - std::cerr << "Unable to create the Cocoa window" << std::endl; - [self autorelease]; - return nil; - } - } - - // Make the OpenGL view - myView = [[GLView alloc] initWithFrame:[[myWindow contentView] frame] - mode:mode - settings:params - delegate:delegate]; - - if (myView) { - // Finish setting up the view and window - // Add the view to our window and tell it to the view - [[myWindow contentView] addSubview:myView]; - [myView finishInitialization]; - - NSNotificationCenter *nc = [NSNotificationCenter defaultCenter]; - - // We want to know when our window got the focus - [nc addObserver:self - selector:@selector(windowDidBecomeMain:) - name:NSWindowDidBecomeMainNotification - object:myWindow]; - - // We want to know when our window lost the focus - [nc addObserver:self - selector:@selector(windowDidResignMain:) - name:NSWindowDidResignMainNotification - object:myWindow]; - - // We want to know when the user closes the window - [nc addObserver:self - selector:@selector(windowWillClose:) - name:NSWindowWillCloseNotification - object:myWindow]; - - // I want to re-center the window if it's a full screen one and moved by Spaces - [nc addObserver:self - selector:@selector(windowDidMove:) - name:NSWindowDidMoveNotification - object:myWindow]; - - // Needed not to make application crash when releasing the window in our destructor - // (I prefer to take control of everything :P) - [myWindow setReleasedWhenClosed:NO]; - [myWindow setAcceptsMouseMovedEvents:YES]; - - } else { - std::cerr << "Unable to create the OpenGL view" << std::endl; - [self autorelease]; - return nil; - } - - if (myIsFullscreen) { - myFullscreenMode = mode; - - // Using this because full screen window was not always - // in front of the other application windows when unhiding app - [myWindow setLevel:NSFloatingWindowLevel]; - } + [myView release]; + myView = [aView retain]; } - - return self; } - //////////////////////////////////////////////////////////// -/// Clean the window wrapper +/// Return a reference to the internal Cocoa OpenGL view //////////////////////////////////////////////////////////// -- (void)dealloc +- (sfPrivGLView *)view { - - // Remove the notification observer - [[NSNotificationCenter defaultCenter] removeObserver:self]; - - // Close the window - [self show:false]; - - // Release the window and view - [myView release]; - [myWindow release]; - - [super dealloc]; + return [[myView retain] autorelease]; } - //////////////////////////////////////////////////////////// -/// Return a reference to the internal Cocoa window +/// Sets @aDelegate as the window delegate //////////////////////////////////////////////////////////// -- (NSWindow *)window +- (void)setDelegate:(sf::priv::WindowImplCocoa *)aDelegate { - return myWindow; + [[self view] setDelegate:aDelegate]; } - //////////////////////////////////////////////////////////// -/// Return a reference to the internal Cocoa OpenGL view +/// Returns the window delegate //////////////////////////////////////////////////////////// -- (GLView *)glView +- (sf::priv::WindowImplCocoa *)delegate { - return myView; + return [[self view] delegate]; } - //////////////////////////////////////////////////////////// /// Forward call to set the window position on screen //////////////////////////////////////////////////////////// - (void)setPosition:(NSPoint)pos { - assert(myWindow != nil); + NSAssert([self window] != nil, @"expected valid window"); - if (!myIsFullscreen) { - // Flip Y and set window position - pos.y = [[myWindow screen] frame].size.height - pos.y; - [myWindow setFrameTopLeftPoint:pos]; - } + // Flip Y and set window position + pos.y = [[[self window] screen] frame].size.height - pos.y; + [[self window] setFrameTopLeftPoint:pos]; } - //////////////////////////////////////////////////////////// /// Forward call to set the window size //////////////////////////////////////////////////////////// - (void)setSize:(NSSize)size { - assert(myWindow != nil); + NSAssert([self window] != nil, @"expected valid window"); - if (!myIsFullscreen) { - [myWindow setFrame:NSMakeRect([myWindow frame].origin.x, - [myWindow frame].origin.y, - size.width, size.height) - display:YES]; - } + [[self window] setFrame:NSMakeRect([[self window] frame].origin.x, + [[self window] frame].origin.y, + size.width, size.height) + display:YES]; } - //////////////////////////////////////////////////////////// /// Return the mouse location relative to the internal window //////////////////////////////////////////////////////////// - (NSPoint)mouseLocation { - assert(myWindow != nil); + NSAssert([self view] != nil, @"expected valid window"); + + NSPoint windowPoint = [[self window] convertScreenToBase:[NSEvent mouseLocation]]; + NSPoint viewPoint = [[self view] convertPoint:windowPoint fromView:[[[self window] contentView] superview]]; - NSPoint relativeLocation = [myWindow convertScreenToBase:[NSEvent mouseLocation]]; - relativeLocation.y = [[self glView] frame].size.height - relativeLocation.y; - return relativeLocation; + viewPoint.y = [[self view] frame].size.height - viewPoint.y; + return viewPoint; } - //////////////////////////////////////////////////////////// /// Return whether the mouse is on our window //////////////////////////////////////////////////////////// - (BOOL)mouseInside { - assert(myWindow != nil); - assert(myView != nil); + NSAssert([self window] != nil, @"expected valid window"); + NSAssert([self view] != nil, @"expected valid OpenGL view"); BOOL flag = NO; - if ([myWindow isVisible]) { - NSPoint relativeToWindow = [myWindow mouseLocationOutsideOfEventStream]; - NSPoint relativeToView = [myView convertPoint:relativeToWindow fromView:nil]; + if ([[self window] isVisible]) { + NSPoint relativeToWindow = [[self window] mouseLocationOutsideOfEventStream]; + NSPoint relativeToView = [[self view] convertPoint:relativeToWindow fromView:nil]; - if (NSPointInRect (relativeToView, [myView bounds])) + if (NSPointInRect (relativeToView, [[self view] bounds])) { flag = YES; } @@ -743,32 +691,23 @@ static GLContext *sharedCtx = nil; return flag; } - //////////////////////////////////////////////////////////// /// Close or open the window //////////////////////////////////////////////////////////// - (void)show:(bool)flag { - assert(myWindow != nil); + NSAssert([self window] != nil, @"expected valid window"); - if (flag && ![myWindow isVisible]) { + if (flag && ![[self window] isVisible]) { // Wanna open the closed window - if (myIsFullscreen) { - [SharedAppController setFullscreenWindow:self mode:&myFullscreenMode]; - } else { - // Show the window - [myWindow makeKeyAndOrderFront:nil]; - } - } else if (!flag && [myWindow isVisible]) { + // Show the window + [[self window] makeKeyAndOrderFront:nil]; + } else if (!flag && [[self window] isVisible]) { // Wanna close the opened window - if (myIsFullscreen) { - [SharedAppController setFullscreenWindow:nil mode:NULL]; - } else { - // Close the window - [myWindow close]; - } + // Close the window + [[self window] close]; } } @@ -778,31 +717,28 @@ static GLContext *sharedCtx = nil; //////////////////////////////////////////////////////////// - (void)enableVerticalSync:(bool)flag { - assert(myView != nil); - [myView enableVerticalSync:flag]; + NSAssert([self view] != nil, @"expected valid OpenGL view"); + [[self view] enableVerticalSync:flag]; } - //////////////////////////////////////////////////////////// /// Forward 'setActive' call the the OpenGL view //////////////////////////////////////////////////////////// - (void)setActive:(bool)flag { - assert(myView != nil); - [myView setActive:flag]; + NSAssert([self view] != nil, @"expected valid OpenGL view"); + [[self view] setActive:flag]; } - //////////////////////////////////////////////////////////// /// Forward call to flush the OpenGL view //////////////////////////////////////////////////////////// - (void)flushBuffer { - assert(myView != nil); - [myView flushBuffer]; + NSAssert([self view] != nil, @"expected valid OpenGL view"); + [[self view] flushBuffer]; } - //////////////////////////////////////////////////////////// /// Notification method receiver when the window gains focus //////////////////////////////////////////////////////////// @@ -811,7 +747,7 @@ static GLContext *sharedCtx = nil; sf::Event ev; ev.Type = sf::Event::GainedFocus; - [myView pushEvent:ev]; + [[self view] pushEvent:ev]; } @@ -823,7 +759,7 @@ static GLContext *sharedCtx = nil; sf::Event ev; ev.Type = sf::Event::LostFocus; - [myView pushEvent:ev]; + [[self view] pushEvent:ev]; } @@ -835,9 +771,170 @@ static GLContext *sharedCtx = nil; sf::Event ev; ev.Type = sf::Event::Closed; - [myView pushEvent:ev]; + [[self view] pushEvent:ev]; +} + + +@end + + +@implementation sfPrivOwnedWindow + +- (id)initWithVideoMode:(sf::VideoMode&)aMode + settings:(sf::WindowSettings&)someSettings + style:(unsigned long)aStyle + title:(NSString *)aTitle +{ + self = [super init]; + if (self) + { + if (aStyle & sf::Style::Fullscreen) { + [self setFullscreen:YES]; + } + + NSRect frame = NSMakeRect (0.0f, 0.0f, (float) aMode.Width, (float) aMode.Height); + unsigned int mask = 0; + + if (aStyle & sf::Style::Fullscreen) { + // Check display mode and put new values in 'mode' if needed + boolean_t exact = true; + + CFDictionaryRef properties = + CGDisplayBestModeForParameters([sfPrivAppController primaryScreen], + aMode.BitsPerPixel, + aMode.Width, + aMode.Height, + &exact); + + if (!properties) { + std::cerr + << "Unable to get a display mode with the given parameters" + << std::endl; + + [self autorelease]; + return nil; + } + + if (exact == false) { + CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, + kCGDisplayWidth), + kCFNumberIntType, &aMode.Width); + + CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, + kCGDisplayHeight), + kCFNumberIntType, &aMode.Height); + + CFNumberGetValue((CFNumberRef) CFDictionaryGetValue(properties, + kCGDisplayBitsPerPixel), + kCFNumberIntType, &aMode.BitsPerPixel); + + } + } + + // We grab options from WindowStyle and add them to our window mask + if (aStyle & sf::Style::None || aStyle & sf::Style::Fullscreen) { + mask |= NSBorderlessWindowMask; + } else { + if (aStyle & sf::Style::Titlebar) { + mask |= NSTitledWindowMask; + mask |= NSMiniaturizableWindowMask; + } + + if (aStyle & sf::Style::Resize) { + mask |= NSTitledWindowMask; + mask |= NSMiniaturizableWindowMask; + mask |= NSResizableWindowMask; + } + + if (aStyle & sf::Style::Close) { + mask |= NSTitledWindowMask; + mask |= NSClosableWindowMask; + mask |= NSMiniaturizableWindowMask; + } + } + + // Now we make the window with the values we got + // Note: defer flag set to NO to be able to use OpenGL in our window + NSWindow *newWindow = [[NSWindow alloc] + initWithContentRect:frame + styleMask:mask + backing:NSBackingStoreBuffered + defer:NO]; + + if (newWindow) { + [self setWindow:[newWindow autorelease]]; + } else { + std::cerr << "Unable to create the Cocoa window" << std::endl; + [self autorelease]; + return nil; + } + + // We set title and window position + [[self window] setTitle:aTitle != nil ? aTitle : @""]; + [[self window] center]; + + // Make the OpenGL view + sfPrivGLView *newView = [[sfPrivGLView alloc] + initWithFrame:[[[self window] contentView] frame] + mode:aMode + settings:someSettings]; + + if (!newView) { + std::cerr << "Unable to create the OpenGL view" << std::endl; + [self autorelease]; + return nil; + } + + // Put our view in the window + [self putOpenGLView:[newView autorelease]]; + + // I want to re-center the window if it's a full screen one and moved by Spaces + [[NSNotificationCenter defaultCenter] + addObserver:self + selector:@selector(windowDidMove:) + name:NSWindowDidMoveNotification + object:[self window]]; + + // Needed not to make application crash when releasing the window in our destructor + // (I prefer to take control of everything :P) + [[self window] setReleasedWhenClosed:NO]; + [[self window] setAcceptsMouseMovedEvents:YES]; + + if ([self isFullscreen]) { + myFullscreenMode = aMode; + + // Using this because full screen window was not always + // in front of the other application windows when unhiding app + [[self window] setLevel:NSFloatingWindowLevel]; + } + } + return self; } +- (void)dealloc +{ + [[NSNotificationCenter defaultCenter] + removeObserver:self + name:NSWindowDidMoveNotification + object:[self window]]; + [super dealloc]; +} + +//////////////////////////////////////////////////////////// +/// Sets the window's fullscreen state +//////////////////////////////////////////////////////////// +- (void)setFullscreen:(BOOL)aFlag +{ + myIsFullscreen = aFlag; +} + +//////////////////////////////////////////////////////////// +/// Returns the window's fullscreen state +//////////////////////////////////////////////////////////// +- (BOOL)isFullscreen +{ + return myIsFullscreen; +} //////////////////////////////////////////////////////////// /// Notification method receiver when the window finish moving @@ -846,9 +943,182 @@ static GLContext *sharedCtx = nil; { NSWindow *sender = [notification object]; - if (!([sender styleMask] & NSTitledWindowMask)) + if (myIsFullscreen) [sender center]; } +//////////////////////////////////////////////////////////// +/// Close or open the window +//////////////////////////////////////////////////////////// +- (void)show:(bool)flag +{ + NSAssert([self window] != nil, @"expected valid window"); + + if (flag && ![[self window] isVisible]) { + // Wanna open the closed window + + if ([self isFullscreen]) { + [[sfPrivAppController sharedController] + setFullscreenWindow:self + mode:&myFullscreenMode]; + } else { + // Show the window + [[self window] makeKeyAndOrderFront:nil]; + } + } else if (!flag && [[self window] isVisible]) { + // Wanna close the opened window + + if ([self isFullscreen]) { + [[sfPrivAppController sharedController] + setFullscreenWindow:nil + mode:NULL]; + } else { + // Close the window + [[self window] close]; + } + } +} + +//////////////////////////////////////////////////////////// +/// Forward call to set the window position on screen +//////////////////////////////////////////////////////////// +- (void)setPosition:(NSPoint)aPosition +{ + if (![self isFullscreen]) { + [super setPosition:aPosition]; + } +} + +//////////////////////////////////////////////////////////// +/// Forward call to set the window size +//////////////////////////////////////////////////////////// +- (void)setSize:(NSSize)size +{ + if (![self isFullscreen]) { + [super setSize:size]; + } +} + +@end + + +@implementation sfPrivImportedWindow + +- (id)initWithWindow:(NSWindow *)aWindow + settings:(sf::WindowSettings&)someSettings +{ + self = [super init]; + + if (self) { + [self setWindow:aWindow]; + + // Make the OpenGL view + sf::VideoMode mode([[[self window] contentView] frame].size.width, + [[[self window] contentView] frame].size.height); + sfPrivGLView *newView = [[sfPrivGLView alloc] + initWithFrame:[[[self window] contentView] frame] + mode:mode + settings:someSettings]; + + if (!newView) { + std::cerr << "Unable to create the OpenGL view" << std::endl; + [self autorelease]; + return nil; + } + + [self putOpenGLView:[newView autorelease]]; + + [[self window] setAcceptsMouseMovedEvents:YES]; + } + + return self; +} + +@end + + +@implementation sfPrivImportedView + + +- (id)initWithView:(NSView *)aView + settings:(sf::WindowSettings&)someSettings +{ + self = [super init]; + if (self) + { + parentView = [aView retain]; + [self setWindow:[parentView window]]; + + // Make the OpenGL view + sf::VideoMode mode([parentView bounds].size.width, + [parentView bounds].size.height); + sfPrivGLView *newView = [[sfPrivGLView alloc] + initWithFrame:[parentView bounds] + mode:mode + settings:someSettings]; + + if (!newView) { + std::cerr << "Unable to create the OpenGL view" << std::endl; + [self autorelease]; + return nil; + } + + [self putOpenGLView:[newView autorelease]]; + [[self window] setAcceptsMouseMovedEvents:YES]; + } + return self; +} + +- (void)dealloc +{ + [parentView release]; + [super dealloc]; +} + +//////////////////////////////////////////////////////////// +/// Registers the the OpenGL view and adds it to its parent container +//////////////////////////////////////////////////////////// +- (void)putOpenGLView:(sfPrivGLView *)aView +{ + [self setView:aView]; + + // Finish setting up the view and window + NSRect originalFrame = [[self window] frame]; + NSRect tmpFrame = originalFrame; + originalFrame.origin.x++; + + [[self window] setFrame:tmpFrame display:YES]; + [[self window] setFrame:originalFrame display:YES]; + + + // Add the view to our *parent view* and tell it to the view + [parentView addSubview:[self view]]; + [[self view] finishInitialization]; +} + +//////////////////////////////////////////////////////////// +/// Forward call to set the window position on screen +//////////////////////////////////////////////////////////// +- (void)setPosition:(NSPoint)aPosition +{ + std::cerr + << "Warning: called Window::SetPosition() on a window imported from a widget. " + << "This method has been disabled in this case and has no effect. " + << "Directly use the widget if you want to move it." + << std::endl; +} + +//////////////////////////////////////////////////////////// +/// Forward call to set the window size +//////////////////////////////////////////////////////////// +- (void)setSize:(NSSize)size +{ + std::cerr + << "Warning: called Window::SetSize() on a window imported from a widget. " + << "This method has been disabled in this case and has no effect. " + << "Directly use the widget if you want to resize it." + << std::endl; +} + @end diff --git a/src/SFML/Window/Cocoa/Joystick.cpp b/src/SFML/Window/Cocoa/Joystick.cpp index 4ade09c..f65ac73 100755 --- a/src/SFML/Window/Cocoa/Joystick.cpp +++ b/src/SFML/Window/Cocoa/Joystick.cpp @@ -49,18 +49,17 @@ void Joystick::Initialize(unsigned int Index) JoystickState Joystick::UpdateState() { // Fill a JoystickState instance with the current joystick state - JoystickState s = {0}; + JoystickState s; return s; } //////////////////////////////////////////////////////////// -/// Get the number of axes supported by the joystick +/// Check if the joystick supports the given axis //////////////////////////////////////////////////////////// -unsigned int Joystick::GetAxesCount() const +bool Joystick::HasAxis(Joy::Axis Axis) const { - // Return number of supported axes - return 0; + return false; } diff --git a/src/SFML/Window/Cocoa/Joystick.hpp b/src/SFML/Window/Cocoa/Joystick.hpp index 4259075..60c5db2 100755 --- a/src/SFML/Window/Cocoa/Joystick.hpp +++ b/src/SFML/Window/Cocoa/Joystick.hpp @@ -58,13 +58,15 @@ public : JoystickState UpdateState(); //////////////////////////////////////////////////////////// - /// Get the number of axes supported by the joystick + /// Check if the joystick supports the given axis /// - /// \return Number of axis + /// \param Axis : Axis to check + /// + /// \return True of the axis is supported, false otherwise /// //////////////////////////////////////////////////////////// - unsigned int GetAxesCount() const; - + bool HasAxis(Joy::Axis Axis) const; + //////////////////////////////////////////////////////////// /// Get the number of buttons supported by the joystick /// diff --git a/src/SFML/Window/Cocoa/WindowImplCocoa.hpp b/src/SFML/Window/Cocoa/WindowImplCocoa.hpp index bc93d54..62158f0 100755 --- a/src/SFML/Window/Cocoa/WindowImplCocoa.hpp +++ b/src/SFML/Window/Cocoa/WindowImplCocoa.hpp @@ -33,8 +33,10 @@ #include <string> #ifdef __OBJC__ -#import <Cocoa/Cocoa.h> -@class WindowWrapper; +@class sfPrivWindow; +typedef sfPrivWindow* sfPrivWindowRef; +#else +typedef void* sfPrivWindowRef; #endif namespace sf @@ -108,6 +110,10 @@ public : void HandleMouseMove(void *eventRef); void HandleMouseWheel(void *eventRef); + //////////////////////////////////////////////////////////// + /// Return whether 'ev' must be considered as a TextEntered event + //////////////////////////////////////////////////////////// + static bool IsTextEvent(void *event); private : //////////////////////////////////////////////////////////// @@ -186,13 +192,7 @@ private : //////////////////////////////////////////////////////////// // Member data //////////////////////////////////////////////////////////// - -#ifdef __OBJC__ - WindowWrapper *myWrapper; -#else - void *myWrapper; -#endif - + sfPrivWindowRef myWrapper; bool myUseKeyRepeat; bool myMouseIn; float myWheelStatus; diff --git a/src/SFML/Window/Cocoa/WindowImplCocoa.mm b/src/SFML/Window/Cocoa/WindowImplCocoa.mm index 3b66bbc..13dec83 100755 --- a/src/SFML/Window/Cocoa/WindowImplCocoa.mm +++ b/src/SFML/Window/Cocoa/WindowImplCocoa.mm @@ -31,9 +31,6 @@ #import <SFML/Window/Cocoa/GLKit.h> #import <SFML/Window/WindowStyle.hpp> #import <SFML/System.hpp> -#import <OpenGL/OpenGL.h> -#import <OpenGL/gl.h> -#import <CoreFoundation/CoreFoundation.h> #import <iostream> @@ -54,26 +51,26 @@ __done = 1;\ //////////////////////////////////////////////////////////// /// Private function declarations //////////////////////////////////////////////////////////// -static Key::Code KeyForVirtualCode(unsigned short vCode); -static Key::Code KeyForUnicode(unsigned short uniCode); -static bool IsTextEvent(NSEvent *event); +namespace { + Key::Code KeyForVirtualCode(unsigned short vCode); + Key::Code KeyForUnicode(unsigned short uniCode); +} // anonymous namespace //////////////////////////////////////////////////////////// /// Default constructor /// (creates a dummy window to provide a valid OpenGL context) //////////////////////////////////////////////////////////// - static WindowImplCocoa *globalWin = NULL; WindowImplCocoa::WindowImplCocoa() : myWrapper(nil), myUseKeyRepeat(false), myMouseIn(false), myWheelStatus(0.0f) { - [AppController sharedController]; + [sfPrivAppController sharedController]; // Create the shared OpenGL context - if ([GLContext sharedContext]) { + if ([sfPrivGLContext sharedContext]) { // Then we make it the current active OpenGL context SetActive(); } else { @@ -91,33 +88,46 @@ myUseKeyRepeat(false), myMouseIn(false), myWheelStatus(0.0f) { - if (Handle) - { - if (![(NSWindow *)Handle isKindOfClass:[NSWindow class]]) - std::cerr << "Cannot import this Window Handle because it is not a <NSWindow *> object" - << "(or one of its subclasses). You gave a <" - << [[(NSWindow *)Handle className] UTF8String] - << "> object." << std::endl; - else - { + if (Handle) { + // Classical window import + if ([(id)Handle isKindOfClass:[NSWindow class]]) { + myWrapper = [[sfPrivImportedWindow alloc] + initWithWindow:(NSWindow *)Handle + settings:params]; + } + // Qt "window" import + else if ([(id)Handle isKindOfClass:[NSView class]]) { + myWrapper = [[sfPrivImportedView alloc] + initWithView:(NSView *)Handle + settings:params]; + } else { + std::cerr + << "Cannot import this Window Handle because it is neither" + << "a <NSWindow *> nor <NSView *> object" + << "(or any of its subclasses). You gave a <" + << [[(id)Handle className] UTF8String] + << "> object." + << std::endl; - // We create the window according to the given handle - myWrapper = [[WindowWrapper alloc] initWithWindow:(NSWindow *)Handle - settings:params - delegate:this]; + } + + if (myWrapper) { + [myWrapper setDelegate:this]; - if (myWrapper) - { - // initial mouse state - myMouseIn = [myWrapper mouseInside]; - - // We set the myWidth and myHeight members to the correct values - myWidth = (int) [[myWrapper glView] frame].size.width; - myHeight = (int) [[myWrapper glView] frame].size.height; - } else { - std::cerr << "Failed to make the public window" << std::endl; - } + // initial mouse state + myMouseIn = [myWrapper mouseInside]; + + // We set the myWidth and myHeight members to the correct values + myWidth = (int) [[myWrapper view] frame].size.width; + myHeight = (int) [[myWrapper view] frame].size.height; + } else { + std::cerr << "Failed to make the public window" << std::endl; } + } else { + std::cerr + << "Invalid null handle given to " + << "Window::Window(WindowHandle Handle, const WindowSettings& Params)" + << std::endl; } } @@ -133,17 +143,22 @@ myWheelStatus(0.0f) { // Create a new window with given size, title and style // First we define some objects used for our window - NSString *title = [NSString stringWithUTF8String:(Title.c_str()) ? (Title.c_str()) : ""]; + NSString *title = [NSString stringWithCString:(Title.c_str()) ? (Title.c_str()) : "" + encoding:NSASCIIStringEncoding]; // We create the window - myWrapper = [[WindowWrapper alloc] initWithSettings:params - videoMode:Mode - style:WindowStyle - title:title - delegate:this]; + myWrapper = [[sfPrivOwnedWindow alloc] + initWithVideoMode:Mode + settings:params + style:WindowStyle + title:title]; + + if (myWrapper) { + [myWrapper setDelegate:this]; + // initial mouse state myMouseIn = [myWrapper mouseInside]; @@ -228,18 +243,20 @@ void WindowImplCocoa::HandleKeyDown(void *eventRef) // convert the characters // note: using CFString in order to keep compatibility with Mac OS X 10.4 - // (NSUTF32StringEncoding only defined since Mac OS X 10.5) + // (as NSUTF32StringEncoding is only being defined in Mac OS X 10.5 and later) if (!CFStringGetCString ((CFStringRef)[event characters], (char *)utf32Characters, sizeof(utf32Characters), kCFStringEncodingUTF32)) { - const char *utf8Char = NULL; - if ([[event characters] lengthOfBytesUsingEncoding:NSUTF8StringEncoding]) - utf8Char = [[event characters] UTF8String]; + char asciiChar[3] = {0}; + if ([[event characters] lengthOfBytesUsingEncoding:NSASCIIStringEncoding]) + [[event characters] getCString:asciiChar + maxLength:3 + encoding:NSASCIIStringEncoding]; - std::cerr << "Error while converting character to UTF32 : " - << ((utf8Char) ? utf8Char : "(undefined)") << std::endl; + std::cerr << "Error while converting character to UTF32 : \"" + << asciiChar << "\"" << std::endl; } else { @@ -359,7 +376,9 @@ void WindowImplCocoa::HandleMouseDown(void *eventRef) { NSEvent *event = reinterpret_cast <NSEvent *> (eventRef); Event sfEvent; - NSPoint loc = {0, 0}; + + // Get mouse position relative to the window + NSPoint loc = [myWrapper mouseLocation]; unsigned mods = [event modifierFlags]; switch ([event type]) { @@ -373,9 +392,6 @@ void WindowImplCocoa::HandleMouseDown(void *eventRef) sfEvent.MouseButton.Button = Mouse::Left; } - // Get mouse position relative to the window - loc = [myWrapper mouseLocation]; - sfEvent.MouseButton.X = (int) loc.x; sfEvent.MouseButton.Y = (int) loc.y; @@ -387,9 +403,6 @@ void WindowImplCocoa::HandleMouseDown(void *eventRef) sfEvent.Type = Event::MouseButtonPressed; sfEvent.MouseButton.Button = Mouse::Right; - // Get mouse position relative to the window - loc = [myWrapper mouseLocation]; - sfEvent.MouseButton.X = (int) loc.x; sfEvent.MouseButton.Y = (int) loc.y; @@ -410,7 +423,9 @@ void WindowImplCocoa::HandleMouseUp(void *eventRef) { NSEvent *event = reinterpret_cast <NSEvent *> (eventRef); Event sfEvent; - NSPoint loc = {0, 0}; + + // Get mouse position relative to the window + NSPoint loc = [myWrapper mouseLocation]; unsigned mods = [event modifierFlags]; switch ([event type]) { @@ -424,9 +439,6 @@ void WindowImplCocoa::HandleMouseUp(void *eventRef) sfEvent.MouseButton.Button = Mouse::Left; } - // Get mouse position relative to the window - loc = [myWrapper mouseLocation]; - sfEvent.MouseButton.X = (int) loc.x; sfEvent.MouseButton.Y = (int) loc.y; @@ -438,9 +450,6 @@ void WindowImplCocoa::HandleMouseUp(void *eventRef) sfEvent.Type = Event::MouseButtonReleased; sfEvent.MouseButton.Button = Mouse::Right; - // Get mouse position relative to the window - loc = [myWrapper mouseLocation]; - sfEvent.MouseButton.X = (int) loc.x; sfEvent.MouseButton.Y = (int) loc.y; @@ -460,9 +469,7 @@ void WindowImplCocoa::HandleMouseUp(void *eventRef) void WindowImplCocoa::HandleMouseMove(void *eventRef) { Event sfEvent; - NSPoint loc = {0, 0}; - - loc = [myWrapper mouseLocation]; + NSPoint loc = [myWrapper mouseLocation]; sfEvent.Type = Event::MouseMoved; sfEvent.MouseMove.X = (int) loc.x; @@ -511,6 +518,25 @@ void WindowImplCocoa::HandleMouseWheel(void *eventRef) } } +//////////////////////////////////////////////////////////// +/// Return whether 'ev' must be considered as a TextEntered event +//////////////////////////////////////////////////////////// +bool WindowImplCocoa::IsTextEvent(void *eventRef) +{ + NSEvent *event = (NSEvent *)eventRef; + bool res = false; + + if (event && [event type] == NSKeyDown && [[event characters] length]) { + unichar code = [[event characters] characterAtIndex:0]; + + // Codes from 0xF700 to 0xF8FF are non text keys (see NSEvent.h) + // 0x35 is the Escape key + if ([event keyCode] != 0x35 && (code < 0xF700 || code > 0xF8FF)) + res = true; + } + + return res; +} //////////////////////////////////////////////////////////// /// /see sfWindowImpl::Display @@ -528,7 +554,7 @@ void WindowImplCocoa::Display() void WindowImplCocoa::ProcessEvents() { // Forward event handling call to the application controller - [SharedAppController processEvents]; + [[sfPrivAppController sharedController] processEvents]; } @@ -538,7 +564,18 @@ void WindowImplCocoa::ProcessEvents() void WindowImplCocoa::SetActive(bool Active) const { // Forward the call to the window - [myWrapper setActive:Active]; + if (myWrapper) + [myWrapper setActive:Active]; + else { + // Or directly activate the shared OpenGL context if we're not using a window + if (Active) { + if ([NSOpenGLContext currentContext] != [sfPrivGLContext sharedContext]) + [[sfPrivGLContext sharedContext] makeCurrentContext]; + } else { + if ([NSOpenGLContext currentContext] == [sfPrivGLContext sharedContext]) + [NSOpenGLContext clearCurrentContext]; + } + } } @@ -577,7 +614,7 @@ void WindowImplCocoa::SetCursorPosition(unsigned int Left, unsigned int Top) pos.y = [[myWrapper window] frame].size.height - pos.y; // Adjust for view reference instead of window - pos.y -= [[myWrapper window] frame].size.height - [[myWrapper glView] frame].size.height; + pos.y -= [[myWrapper window] frame].size.height - [[myWrapper view] frame].size.height; // Convert to screen coordinates NSPoint absolute = [[myWrapper window] convertBaseToScreen:pos]; @@ -586,7 +623,8 @@ void WindowImplCocoa::SetCursorPosition(unsigned int Left, unsigned int Top) absolute.y = [[NSScreen mainScreen] frame].size.height - absolute.y; // Move cursor - CGDisplayMoveCursorToPoint(kCGDirectMainDisplay, CGPointMake(absolute.x, absolute.y)); + CGDisplayMoveCursorToPoint([sfPrivAppController primaryScreen], + CGPointMake(absolute.x, absolute.y)); } } @@ -637,183 +675,167 @@ void WindowImplCocoa::SetIcon(unsigned int Width, unsigned int Height, const Uin } -//////////////////////////////////////////////////////////// -/// Return the SFML key corresponding to a key code -//////////////////////////////////////////////////////////// -static Key::Code KeyForVirtualCode(unsigned short vCode) -{ - static struct { - unsigned short code; - Key::Code sfKey; - } virtualTable[] = +namespace { + //////////////////////////////////////////////////////////// + /// Return the SFML key corresponding to a key code + //////////////////////////////////////////////////////////// + Key::Code KeyForVirtualCode(unsigned short vCode) { - {0x35, Key::Escape}, - {0x31, Key::Space}, - {0x24, Key::Return}, // main Return key - {0x4C, Key::Return}, // pav Return key - {0x33, Key::Back}, - {0x30, Key::Tab}, - {0x74, Key::PageUp}, - {0x79, Key::PageDown}, - {0x77, Key::End}, - {0x73, Key::Home}, - {0x72, Key::Insert}, - {0x75, Key::Delete}, - {0x45, Key::Add}, - {0x4E, Key::Subtract}, - {0x43, Key::Multiply}, - {0x4B, Key::Divide}, - - {0x7A, Key::F1}, {0x78, Key::F2}, {0x63, Key::F3}, - {0x76, Key::F4}, {0x60, Key::F5}, {0x61, Key::F6}, - {0x62, Key::F7}, {0x64, Key::F8}, {0x65, Key::F9}, - {0x6D, Key::F10}, {0x67, Key::F11}, {0x6F, Key::F12}, - {0x69, Key::F13}, {0x6B, Key::F14}, {0x71, Key::F15}, - - {0x7B, Key::Left}, - {0x7C, Key::Right}, - {0x7E, Key::Up}, - {0x7D, Key::Down}, - - {0x52, Key::Numpad0}, {0x53, Key::Numpad1}, {0x54, Key::Numpad2}, - {0x55, Key::Numpad3}, {0x56, Key::Numpad4}, {0x57, Key::Numpad5}, - {0x58, Key::Numpad6}, {0x59, Key::Numpad7}, {0x5B, Key::Numpad8}, - {0x5C, Key::Numpad9}, - - {0x1D, Key::Num0}, {0x12, Key::Num1}, {0x13, Key::Num2}, - {0x14, Key::Num3}, {0x15, Key::Num4}, {0x17, Key::Num5}, - {0x16, Key::Num6}, {0x1A, Key::Num7}, {0x1C, Key::Num8}, - {0x19, Key::Num9}, + static struct { + unsigned short code; + Key::Code sfKey; + } virtualTable[] = + { + {0x35, Key::Escape}, + {0x31, Key::Space}, + {0x24, Key::Return}, // main Return key + {0x4C, Key::Return}, // pav Return key + {0x33, Key::Back}, + {0x30, Key::Tab}, + {0x74, Key::PageUp}, + {0x79, Key::PageDown}, + {0x77, Key::End}, + {0x73, Key::Home}, + {0x72, Key::Insert}, + {0x75, Key::Delete}, + {0x45, Key::Add}, + {0x4E, Key::Subtract}, + {0x43, Key::Multiply}, + {0x4B, Key::Divide}, + + {0x7A, Key::F1}, {0x78, Key::F2}, {0x63, Key::F3}, + {0x76, Key::F4}, {0x60, Key::F5}, {0x61, Key::F6}, + {0x62, Key::F7}, {0x64, Key::F8}, {0x65, Key::F9}, + {0x6D, Key::F10}, {0x67, Key::F11}, {0x6F, Key::F12}, + {0x69, Key::F13}, {0x6B, Key::F14}, {0x71, Key::F15}, + + {0x7B, Key::Left}, + {0x7C, Key::Right}, + {0x7E, Key::Up}, + {0x7D, Key::Down}, + + {0x52, Key::Numpad0}, {0x53, Key::Numpad1}, {0x54, Key::Numpad2}, + {0x55, Key::Numpad3}, {0x56, Key::Numpad4}, {0x57, Key::Numpad5}, + {0x58, Key::Numpad6}, {0x59, Key::Numpad7}, {0x5B, Key::Numpad8}, + {0x5C, Key::Numpad9}, + + {0x1D, Key::Num0}, {0x12, Key::Num1}, {0x13, Key::Num2}, + {0x14, Key::Num3}, {0x15, Key::Num4}, {0x17, Key::Num5}, + {0x16, Key::Num6}, {0x1A, Key::Num7}, {0x1C, Key::Num8}, + {0x19, Key::Num9}, + + {0x3B, Key::LControl}, //< Left Ctrl + {0x3A, Key::LAlt}, //< Left Option/Alt + {0x37, Key::LSystem}, //< Left Command + {0x38, Key::LShift}, //< Left Shift + {0x3E, Key::RControl}, //< Right Ctrl + {0x3D, Key::RAlt}, //< Right Option/Alt + {0x36, Key::RSystem}, //< Right Command + {0x3C, Key::RShift}, //< Right Shift + + {0x39, Key::Code(0)} //< Caps Lock (not handled by SFML for now) + }; - {0x3B, Key::LControl}, //< Left Ctrl - {0x3A, Key::LAlt}, //< Left Option/Alt - {0x37, Key::LSystem}, //< Left Command - {0x38, Key::LShift}, //< Left Shift - {0x3E, Key::RControl}, //< Right Ctrl - {0x3D, Key::RAlt}, //< Right Option/Alt - {0x36, Key::RSystem}, //< Right Command - {0x3C, Key::RShift}, //< Right Shift + Key::Code result = Key::Code(0); - {0x39, Key::Code(0)} //< Caps Lock (not handled by SFML for now) - }; - - Key::Code result = Key::Code(0); - - for (unsigned i = 0;virtualTable[i].code;i++) { - if (virtualTable[i].code == vCode) { - result = virtualTable[i].sfKey; - break; + for (unsigned i = 0;virtualTable[i].code;i++) { + if (virtualTable[i].code == vCode) { + result = virtualTable[i].sfKey; + break; + } } + + return result; } - return result; -} - - -//////////////////////////////////////////////////////////// -/// Return the SFML key corresponding to a unicode code -//////////////////////////////////////////////////////////// -static Key::Code KeyForUnicode(unsigned short uniCode) -{ - // TODO: find a better way to get the language independant key - static struct { - unsigned short character; - Key::Code sfKey; - } unicodeTable[] = + + //////////////////////////////////////////////////////////// + /// Return the SFML key corresponding to a unicode code + //////////////////////////////////////////////////////////// + Key::Code KeyForUnicode(unsigned short uniCode) { - {'!', Key::Code(0)}, //< No Key for this code - {'"', Key::Code(0)}, //< No Key for this code - {'#', Key::Code(0)}, //< No Key for this code - {'$', Key::Code(0)}, //< No Key for this code - {'%', Key::Code(0)}, //< No Key for this code - {'&', Key::Code(0)}, //< No Key for this code - {'\'', Key::Quote}, - {'(', Key::Code(0)}, //< No Key for this code - {')', Key::Code(0)}, //< No Key for this code - {'*', Key::Multiply}, - {'+', Key::Add}, - {',', Key::Comma}, - {'-', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'.', Key::Period}, - {'/', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'0', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'1', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'2', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'3', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'4', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'5', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'6', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'7', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'8', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {'9', Key::Code(0)}, //< Handled by KeyForVirtualCode() - {':', Key::Code(0)}, //< No Key for this code - {';', Key::SemiColon}, - {'<', Key::Code(0)}, //< No Key for this code - {'=', Key::Equal}, - {'>', Key::Code(0)}, //< No Key for this code - {'?', Key::Code(0)}, //< No Key for this code - {'@', Key::Code(0)}, //< No Key for this code - {'A', Key::A}, {'B', Key::B}, {'C', Key::C}, - {'D', Key::D}, {'E', Key::E}, {'F', Key::F}, - {'G', Key::G}, {'H', Key::H}, {'I', Key::I}, - {'J', Key::J}, {'K', Key::K}, {'L', Key::L}, - {'M', Key::M}, {'N', Key::N}, {'O', Key::O}, - {'P', Key::P}, {'Q', Key::Q}, {'R', Key::R}, - {'S', Key::S}, {'T', Key::T}, {'U', Key::U}, - {'V', Key::V}, {'W', Key::W}, {'X', Key::X}, - {'Y', Key::Y}, {'Z', Key::Z}, - {'[', Key::LBracket}, - {'\\', Key::BackSlash}, - {']', Key::RBracket}, - {'^', Key::Code(0)}, //< No Key for this code - {'_', Key::Code(0)}, //< No Key for this code - {'`', Key::Code(0)}, //< No Key for this code - {'a', Key::A}, {'b', Key::B}, {'c', Key::C}, - {'d', Key::D}, {'e', Key::E}, {'f', Key::F}, - {'g', Key::G}, {'h', Key::H}, {'i', Key::I}, - {'j', Key::J}, {'k', Key::K}, {'l', Key::L}, - {'m', Key::M}, {'n', Key::N}, {'o', Key::O}, - {'p', Key::P}, {'q', Key::Q}, {'r', Key::R}, - {'s', Key::S}, {'t', Key::T}, {'u', Key::U}, - {'v', Key::V}, {'w', Key::W}, {'x', Key::X}, - {'y', Key::Y}, {'z', Key::Z}, - {'{', Key::Code(0)}, //< No Key for this code - {'|', Key::Code(0)}, //< No Key for this code - {'}', Key::Code(0)}, //< No Key for this code - {'~', Key::Tilde}, - {0, Key::Code(0)} - }; - - Key::Code result = Key::Code(0); - - for (unsigned i = 0;unicodeTable[i].character;i++) { - if (unicodeTable[i].character == uniCode) { - result = unicodeTable[i].sfKey; - break; + // TODO: find a better way to get the language independant key + static struct { + unsigned short character; + Key::Code sfKey; + } unicodeTable[] = + { + {'!', Key::Code(0)}, //< No Key for this code + {'"', Key::Code(0)}, //< No Key for this code + {'#', Key::Code(0)}, //< No Key for this code + {'$', Key::Code(0)}, //< No Key for this code + {'%', Key::Code(0)}, //< No Key for this code + {'&', Key::Code(0)}, //< No Key for this code + {'\'', Key::Quote}, + {'(', Key::Code(0)}, //< No Key for this code + {')', Key::Code(0)}, //< No Key for this code + {'*', Key::Multiply}, + {'+', Key::Add}, + {',', Key::Comma}, + {'-', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'.', Key::Period}, + {'/', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'0', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'1', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'2', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'3', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'4', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'5', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'6', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'7', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'8', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {'9', Key::Code(0)}, //< Handled by KeyForVirtualCode() + {':', Key::Code(0)}, //< No Key for this code + {';', Key::SemiColon}, + {'<', Key::Code(0)}, //< No Key for this code + {'=', Key::Equal}, + {'>', Key::Code(0)}, //< No Key for this code + {'?', Key::Code(0)}, //< No Key for this code + {'@', Key::Code(0)}, //< No Key for this code + {'A', Key::A}, {'B', Key::B}, {'C', Key::C}, + {'D', Key::D}, {'E', Key::E}, {'F', Key::F}, + {'G', Key::G}, {'H', Key::H}, {'I', Key::I}, + {'J', Key::J}, {'K', Key::K}, {'L', Key::L}, + {'M', Key::M}, {'N', Key::N}, {'O', Key::O}, + {'P', Key::P}, {'Q', Key::Q}, {'R', Key::R}, + {'S', Key::S}, {'T', Key::T}, {'U', Key::U}, + {'V', Key::V}, {'W', Key::W}, {'X', Key::X}, + {'Y', Key::Y}, {'Z', Key::Z}, + {'[', Key::LBracket}, + {'\\', Key::BackSlash}, + {']', Key::RBracket}, + {'^', Key::Code(0)}, //< No Key for this code + {'_', Key::Code(0)}, //< No Key for this code + {'`', Key::Code(0)}, //< No Key for this code + {'a', Key::A}, {'b', Key::B}, {'c', Key::C}, + {'d', Key::D}, {'e', Key::E}, {'f', Key::F}, + {'g', Key::G}, {'h', Key::H}, {'i', Key::I}, + {'j', Key::J}, {'k', Key::K}, {'l', Key::L}, + {'m', Key::M}, {'n', Key::N}, {'o', Key::O}, + {'p', Key::P}, {'q', Key::Q}, {'r', Key::R}, + {'s', Key::S}, {'t', Key::T}, {'u', Key::U}, + {'v', Key::V}, {'w', Key::W}, {'x', Key::X}, + {'y', Key::Y}, {'z', Key::Z}, + {'{', Key::Code(0)}, //< No Key for this code + {'|', Key::Code(0)}, //< No Key for this code + {'}', Key::Code(0)}, //< No Key for this code + {'~', Key::Tilde}, + {0, Key::Code(0)} + }; + + Key::Code result = Key::Code(0); + + for (unsigned i = 0;unicodeTable[i].character;i++) { + if (unicodeTable[i].character == uniCode) { + result = unicodeTable[i].sfKey; + break; + } } - } - - return result; -} - - -//////////////////////////////////////////////////////////// -/// Return whether 'ev' must be considered as a TextEntered event -//////////////////////////////////////////////////////////// -static bool IsTextEvent(NSEvent *event) -{ - bool res = false; - - if (event && [event type] == NSKeyDown && [[event characters] length]) { - unichar code = [[event characters] characterAtIndex:0]; - // Codes from 0xF700 to 0xF8FF are non text keys (see NSEvent.h) - if (code < 0xF700 || code > 0xF8FF) - res = true; + return result; } - return res; -} +} // anonymous namespace } // namespace priv diff --git a/src/SFML/Window/Input.cpp b/src/SFML/Window/Input.cpp index 7b10270..01dd54f 100755 --- a/src/SFML/Window/Input.cpp +++ b/src/SFML/Window/Input.cpp @@ -37,23 +37,7 @@ Input::Input() : myMouseX(0),
myMouseY(0)
{
- for (int i = 0; i < Key::Count; ++i)
- myKeys[i] = false;
-
- for (int i = 0; i < Mouse::Count; ++i)
- myMouseButtons[i] = false;
-
- for (int i = 0; i < 16; ++i)
- {
- myJoystickButtons[0][i] = false;
- myJoystickButtons[1][i] = false;
- }
-
- for (int i = 0; i < Joy::Count; ++i)
- {
- myJoystickAxis[0][i] = 0.f;
- myJoystickAxis[1][i] = 0.f;
- }
+ ResetStates();
}
@@ -80,7 +64,7 @@ bool Input::IsMouseButtonDown(Mouse::Button Button) const ////////////////////////////////////////////////////////////
bool Input::IsJoystickButtonDown(unsigned int JoyId, unsigned int Button) const
{
- if ((JoyId < 2) && (Button < 16))
+ if ((JoyId < Joy::Count) && (Button < Joy::ButtonCount))
return myJoystickButtons[JoyId][Button];
else
return false;
@@ -110,7 +94,10 @@ int Input::GetMouseY() const ////////////////////////////////////////////////////////////
float Input::GetJoystickAxis(unsigned int JoyId, Joy::Axis Axis) const
{
- return myJoystickAxis[JoyId][Axis];
+ if (JoyId < Joy::Count)
+ return myJoystickAxis[JoyId][Axis];
+ else
+ return 0.f;
}
@@ -147,17 +134,7 @@ void Input::OnEvent(const Event& EventReceived) // Lost focus event : we must reset all persistent states
case Event::LostFocus :
{
- for (int i = 0; i < Key::Count; ++i)
- myKeys[i] = false;
-
- for (int i = 0; i < Mouse::Count; ++i)
- myMouseButtons[i] = false;
-
- for (int i = 0; i < 16; ++i)
- {
- myJoystickButtons[0][i] = false;
- myJoystickButtons[1][i] = false;
- }
+ ResetStates();
break;
}
@@ -166,4 +143,27 @@ void Input::OnEvent(const Event& EventReceived) }
}
+
+////////////////////////////////////////////////////////////
+/// Reset all the states
+////////////////////////////////////////////////////////////
+void Input::ResetStates()
+{
+ for (int i = 0; i < Key::Count; ++i)
+ myKeys[i] = false;
+
+ for (int i = 0; i < Mouse::ButtonCount; ++i)
+ myMouseButtons[i] = false;
+
+ for (int i = 0; i < Joy::Count; ++i)
+ {
+ for (int j = 0; j < Joy::ButtonCount; ++j)
+ myJoystickButtons[i][j] = false;
+
+ for (int j = 0; j < Joy::AxisCount; ++j)
+ myJoystickAxis[i][j] = 0.f;
+ myJoystickAxis[i][Joy::AxisPOV] = -1.f;
+ }
+}
+
} // namespace sf
diff --git a/src/SFML/Window/Joystick.hpp b/src/SFML/Window/Joystick.hpp index 4a17f4e..6dd1b6e 100755 --- a/src/SFML/Window/Joystick.hpp +++ b/src/SFML/Window/Joystick.hpp @@ -41,10 +41,20 @@ namespace priv //////////////////////////////////////////////////////////// struct JoystickState { - enum {MaxButtons = 32}; - - float Axis[Joy::Count]; ///< Position on each axis in range [-100, 100] (except POV which is [0, 360]) - bool Buttons[MaxButtons]; ///< Status of each button (true = pressed) + JoystickState() + { + // Default value for axes + for (int i = 0; i < Joy::AxisCount; ++i) + Axis[i] = 0.f; + Axis[Joy::AxisPOV] = -1.f; + + // Default value for buttons + for (int i = 0; i < Joy::ButtonCount; ++i) + Buttons[i] = false; + } + + float Axis[Joy::AxisCount]; ///< Position on each axis in range [-100, 100] (except POV which is [0, 360]) + bool Buttons[Joy::ButtonCount]; ///< Status of each button (true = pressed) }; } // namespace priv diff --git a/src/SFML/Window/Linux/Joystick.cpp b/src/SFML/Window/Linux/Joystick.cpp index 509bc53..829a741 100755 --- a/src/SFML/Window/Linux/Joystick.cpp +++ b/src/SFML/Window/Linux/Joystick.cpp @@ -28,13 +28,6 @@ #include <SFML/Window/Joystick.hpp>
#include <sstream>
-#if defined(SFML_SYSTEM_LINUX)
- #include <linux/joystick.h>
- #include <fcntl.h>
-#elif defined(SFML_SYSTEM_FREEBSD)
- // #include <sys/joystick.h> ?
-#endif
-
namespace sf
{
@@ -48,12 +41,18 @@ namespace priv void Joystick::Initialize(unsigned int Index)
{
// Initial state
- myNbAxes = 0;
myNbButtons = 0;
- for (int i = 0; i < Joy::Count; ++i)
- myState.Axis[i] = 0.f;
- for (int i = 0; i < JoystickState::MaxButtons; ++i)
+ myPovX = 0;
+ myPovY = 0;
+ for (int i = 0; i < Joy::ButtonCount; ++i)
+ {
myState.Buttons[i] = false;
+ }
+ for (int i = 0; i < Joy::AxisCount; ++i)
+ {
+ myState.Axis[i] = 0.f;
+ myAxes[i] = false;
+ }
// Open the joystick handle
std::ostringstream oss;
@@ -63,13 +62,32 @@ void Joystick::Initialize(unsigned int Index) {
// Use non-blocking mode
fcntl(myDescriptor, F_SETFL, O_NONBLOCK);
-
- // Get number of axes and buttons
- char NbAxes, NbButtons;
- ioctl(myDescriptor, JSIOCGAXES, &NbAxes);
+
+ // Get number of buttons
+ char NbButtons;
ioctl(myDescriptor, JSIOCGBUTTONS, &NbButtons);
- myNbAxes = NbAxes;
myNbButtons = NbButtons;
+ if (myNbButtons > Joy::ButtonCount)
+ myNbButtons = Joy::ButtonCount;
+
+ // Get the supported axes
+ char NbAxes;
+ ioctl(myDescriptor, JSIOCGAXES, &NbAxes);
+ ioctl(myDescriptor, JSIOCGAXMAP, myAxesMapping);
+ for (int i = 0; i < NbAxes; ++i)
+ {
+ switch (myAxesMapping[i])
+ {
+ case ABS_X : myAxes[Joy::AxisX] = true; break;
+ case ABS_Y : myAxes[Joy::AxisY] = true; break;
+ case ABS_Z : case ABS_THROTTLE : myAxes[Joy::AxisZ] = true; break;
+ case ABS_RZ: case ABS_RUDDER: myAxes[Joy::AxisR] = true; break;
+ case ABS_RX : myAxes[Joy::AxisU] = true; break;
+ case ABS_RY : myAxes[Joy::AxisV] = true; break;
+ case ABS_HAT0X : case ABS_HAT0Y : myAxes[Joy::AxisPOV] = true; break;
+ default : break;
+ }
+ }
}
}
@@ -89,11 +107,41 @@ JoystickState Joystick::UpdateState() // An axis has been moved
case JS_EVENT_AXIS :
{
- if (JoyState.number < Joy::Count)
- myState.Axis[JoyState.number] = JoyState.value * 100.f / 32767.f;
+ switch (myAxesMapping[JoyState.number])
+ {
+ case ABS_X : myState.Axis[Joy::AxisX] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_Y : myState.Axis[Joy::AxisY] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_Z : case ABS_THROTTLE : myState.Axis[Joy::AxisZ] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_RZ: case ABS_RUDDER: myState.Axis[Joy::AxisR] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_RX : myState.Axis[Joy::AxisU] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_RY : myState.Axis[Joy::AxisV] = JoyState.value * 100.f / 32767.f; break;
+ case ABS_HAT0X : myPovX = JoyState.value; break;
+ case ABS_HAT0Y : myPovY = JoyState.value; break;
+ default : break;
+ }
+
+ // Compute the new POV angle
+ if (myPovX > 0)
+ {
+ if (myPovY > 0) myState.Axis[Joy::AxisPOV] = 135.f;
+ else if (myPovY < 0) myState.Axis[Joy::AxisPOV] = 45.f;
+ else myState.Axis[Joy::AxisPOV] = 90.f;
+ }
+ else if (myPovX < 0)
+ {
+ if (myPovY > 0) myState.Axis[Joy::AxisPOV] = 225.f;
+ else if (myPovY < 0) myState.Axis[Joy::AxisPOV] = 315.f;
+ else myState.Axis[Joy::AxisPOV] = 270.f;
+ }
+ else
+ {
+ if (myPovY > 0) myState.Axis[Joy::AxisPOV] = 180.f;
+ else if (myPovY < 0) myState.Axis[Joy::AxisPOV] = 0.f;
+ else myState.Axis[Joy::AxisPOV] = -1.f;
+ }
break;
}
-
+
// A button has been pressed
case JS_EVENT_BUTTON :
{
@@ -110,11 +158,11 @@ JoystickState Joystick::UpdateState() ////////////////////////////////////////////////////////////
-/// Get the number of axes supported by the joystick
+/// Check if the joystick supports the given axis
////////////////////////////////////////////////////////////
-unsigned int Joystick::GetAxesCount() const
+bool Joystick::HasAxis(Joy::Axis Axis) const
{
- return myNbAxes;
+ return myAxes[Axis];
}
@@ -148,11 +196,11 @@ JoystickState Joystick::UpdateState() ////////////////////////////////////////////////////////////
-/// Get the number of axes supported by the joystick
+/// Check if the joystick supports the given axis
////////////////////////////////////////////////////////////
-unsigned int Joystick::GetAxesCount() const
+bool Joystick::HasAxis(Joy::Axis Axis) const
{
- return 0;
+ return false;
}
diff --git a/src/SFML/Window/Linux/Joystick.hpp b/src/SFML/Window/Linux/Joystick.hpp index a035f61..3122264 100755 --- a/src/SFML/Window/Linux/Joystick.hpp +++ b/src/SFML/Window/Linux/Joystick.hpp @@ -28,6 +28,13 @@ ////////////////////////////////////////////////////////////
// Headers
////////////////////////////////////////////////////////////
+#if defined(SFML_SYSTEM_LINUX)
+ #include <linux/joystick.h>
+ #include <fcntl.h>
+#elif defined(SFML_SYSTEM_FREEBSD)
+ // #include <sys/joystick.h> ?
+ #define ABS_MAX 1
+#endif
namespace sf
@@ -58,12 +65,14 @@ public : JoystickState UpdateState();
////////////////////////////////////////////////////////////
- /// Get the number of axes supported by the joystick
+ /// Check if the joystick supports the given axis
///
- /// \return Number of axis
+ /// \param Axis : Axis to check
+ ///
+ /// \return True of the axis is supported, false otherwise
///
////////////////////////////////////////////////////////////
- unsigned int GetAxesCount() const;
+ bool HasAxis(Joy::Axis Axis) const;
////////////////////////////////////////////////////////////
/// Get the number of buttons supported by the joystick
@@ -78,10 +87,13 @@ private : ////////////////////////////////////////////////////////////
// Member data
////////////////////////////////////////////////////////////
- int myDescriptor; ///< Linux descriptor of the joystick
- unsigned int myNbAxes; ///< Number of axis supported by the joystick
- unsigned int myNbButtons; ///< Number of buttons supported by the joystick
- JoystickState myState; ///< Current state of the joystick
+ int myDescriptor; ///< Linux descriptor of the joystick
+ unsigned int myNbButtons; ///< Number of buttons supported by the joystick
+ bool myAxes[Joy::AxisCount]; ///< Supported axes
+ JoystickState myState; ///< Current state of the joystick
+ int myPovX; ///< Last X position of the POV
+ int myPovY; ///< Last Y position of the POV
+ char myAxesMapping[ABS_MAX + 1]; ///< Axes mapping (index --> axis id)
};
} // namespace priv
diff --git a/src/SFML/Window/Linux/WindowImplX11.cpp b/src/SFML/Window/Linux/WindowImplX11.cpp index fa0bd8e..edec19b 100755 --- a/src/SFML/Window/Linux/WindowImplX11.cpp +++ b/src/SFML/Window/Linux/WindowImplX11.cpp @@ -349,20 +349,11 @@ WindowImplX11::~WindowImplX11() // Close the display
CloseDisplay();
}
-
-
+ + ////////////////////////////////////////////////////////////
/// Check if there's an active context on the current thread
-////////////////////
////////////////////////////////////////////////////////////
-/// Filter the received events
-/// (only allow those matching a specific window)
-////////////////////////////////////////////////////////////
-Bool CheckEvent(::Display*, XEvent* Event, XPointer UserData)
-{
- // Just check if the event matches our window
- return Event->xany.window == reinterpret_cast< ::Window >(UserData);
-}////////////////////////////////////////
bool WindowImplX11::IsContextActive()
{
return glXGetCurrentContext() != NULL;
diff --git a/src/SFML/Window/Makefile b/src/SFML/Window/Makefile index 956b34b..0dd689c 100755 --- a/src/SFML/Window/Makefile +++ b/src/SFML/Window/Makefile @@ -9,7 +9,7 @@ ifeq ($(STATIC), yes) else
LIB = libsfml-window.so
LIBNAME = $(LIBPATH)/$(LIB).$(VERSION)
- INSTALL = && $(LN) $(LNFLAGS) $(DESTLIBDIR)/$(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
+ INSTALL = && $(LN) $(LNFLAGS) $(LIB).$(VERSION) $(DESTLIBDIR)/$(LIB)
endif
all: $(LIB)
diff --git a/src/SFML/Window/Win32/Joystick.cpp b/src/SFML/Window/Win32/Joystick.cpp index 10bb0d6..e5e16a9 100755 --- a/src/SFML/Window/Win32/Joystick.cpp +++ b/src/SFML/Window/Win32/Joystick.cpp @@ -42,9 +42,12 @@ namespace priv void Joystick::Initialize(unsigned int Index)
{
// Reset state
- myIndex = JOYSTICKID1;
- myNbAxes = 0;
- myNbButtons = 0;
+ myIndex = JOYSTICKID1;
+ myNbButtons = 0;
+ myIsConnected = false;
+ myHasContinuousPOV = false;
+ for (int i = 0; i < Joy::AxisCount; ++i)
+ myAxes[i] = false;
// Get the Index-th connected joystick
MMRESULT Error;
@@ -60,12 +63,21 @@ void Joystick::Initialize(unsigned int Index) if (NbFound == Index)
{
// Ok : store its parameters and return
+ myIsConnected = true;
JOYCAPS Caps;
joyGetDevCaps(myIndex, &Caps, sizeof(Caps));
- myNbAxes = Caps.wNumAxes;
myNbButtons = Caps.wNumButtons;
- if (myNbButtons > JoystickState::MaxButtons)
- myNbButtons = JoystickState::MaxButtons;
+ if (myNbButtons > Joy::ButtonCount)
+ myNbButtons = Joy::ButtonCount;
+
+ myAxes[Joy::AxisX] = true;
+ myAxes[Joy::AxisY] = true;
+ myAxes[Joy::AxisZ] = (Caps.wCaps & JOYCAPS_HASZ) != 0;
+ myAxes[Joy::AxisR] = (Caps.wCaps & JOYCAPS_HASR) != 0;
+ myAxes[Joy::AxisU] = (Caps.wCaps & JOYCAPS_HASU) != 0;
+ myAxes[Joy::AxisV] = (Caps.wCaps & JOYCAPS_HASV) != 0;
+ myAxes[Joy::AxisPOV] = (Caps.wCaps & JOYCAPS_HASPOV) != 0;
+ myHasContinuousPOV = (Caps.wCaps & JOYCAPS_POVCTS) != 0;
return;
}
@@ -82,32 +94,39 @@ void Joystick::Initialize(unsigned int Index) ////////////////////////////////////////////////////////////
JoystickState Joystick::UpdateState()
{
- JoystickState State = {0};
+ JoystickState State;
- // Get the joystick caps (for range conversions)
- JOYCAPS Caps;
- if (joyGetDevCaps(myIndex, &Caps, sizeof(Caps)) == JOYERR_NOERROR)
+ if (myIsConnected)
{
- // Get the current joystick state
- JOYINFOEX Pos;
- Pos.dwFlags = JOY_RETURNALL;
- Pos.dwSize = sizeof(JOYINFOEX);
- if (joyGetPosEx(myIndex, &Pos) == JOYERR_NOERROR)
+ // Get the joystick caps (for range conversions)
+ JOYCAPS Caps;
+ if (joyGetDevCaps(myIndex, &Caps, sizeof(Caps)) == JOYERR_NOERROR)
{
- // Axes
- State.Axis[Joy::AxisX] = (Pos.dwXpos - (Caps.wXmax + Caps.wXmin) / 2.f) * 200.f / (Caps.wXmax - Caps.wXmin);
- State.Axis[Joy::AxisY] = (Pos.dwYpos - (Caps.wYmax + Caps.wYmin) / 2.f) * 200.f / (Caps.wYmax - Caps.wYmin);
- State.Axis[Joy::AxisZ] = (Pos.dwZpos - (Caps.wZmax + Caps.wZmin) / 2.f) * 200.f / (Caps.wZmax - Caps.wZmin);
- State.Axis[Joy::AxisR] = (Pos.dwRpos - (Caps.wRmax + Caps.wRmin) / 2.f) * 200.f / (Caps.wRmax - Caps.wRmin);
- State.Axis[Joy::AxisU] = (Pos.dwUpos - (Caps.wUmax + Caps.wUmin) / 2.f) * 200.f / (Caps.wUmax - Caps.wUmin);
- State.Axis[Joy::AxisV] = (Pos.dwVpos - (Caps.wVmax + Caps.wVmin) / 2.f) * 200.f / (Caps.wVmax - Caps.wVmin);
-
- // POV
- State.Axis[Joy::AxisPOV] = Pos.dwPOV / 100.f;
-
- // Buttons
- for (unsigned int i = 0; i < GetButtonsCount(); ++i)
- State.Buttons[i] = (Pos.dwButtons & (1 << i)) != 0;
+ // Get the current joystick state
+ JOYINFOEX Pos;
+ Pos.dwFlags = JOY_RETURNX | JOY_RETURNY | JOY_RETURNZ | JOY_RETURNR | JOY_RETURNU | JOY_RETURNV | JOY_RETURNBUTTONS;
+ Pos.dwFlags |= myHasContinuousPOV ? JOY_RETURNPOVCTS : JOY_RETURNPOV;
+ Pos.dwSize = sizeof(JOYINFOEX);
+ if (joyGetPosEx(myIndex, &Pos) == JOYERR_NOERROR)
+ {
+ // Axes
+ State.Axis[Joy::AxisX] = (Pos.dwXpos - (Caps.wXmax + Caps.wXmin) / 2.f) * 200.f / (Caps.wXmax - Caps.wXmin);
+ State.Axis[Joy::AxisY] = (Pos.dwYpos - (Caps.wYmax + Caps.wYmin) / 2.f) * 200.f / (Caps.wYmax - Caps.wYmin);
+ State.Axis[Joy::AxisZ] = (Pos.dwZpos - (Caps.wZmax + Caps.wZmin) / 2.f) * 200.f / (Caps.wZmax - Caps.wZmin);
+ State.Axis[Joy::AxisR] = (Pos.dwRpos - (Caps.wRmax + Caps.wRmin) / 2.f) * 200.f / (Caps.wRmax - Caps.wRmin);
+ State.Axis[Joy::AxisU] = (Pos.dwUpos - (Caps.wUmax + Caps.wUmin) / 2.f) * 200.f / (Caps.wUmax - Caps.wUmin);
+ State.Axis[Joy::AxisV] = (Pos.dwVpos - (Caps.wVmax + Caps.wVmin) / 2.f) * 200.f / (Caps.wVmax - Caps.wVmin);
+
+ // POV
+ if (Pos.dwPOV != 0xFFFF)
+ State.Axis[Joy::AxisPOV] = Pos.dwPOV / 100.f;
+ else
+ State.Axis[Joy::AxisPOV] = -1.f;
+
+ // Buttons
+ for (unsigned int i = 0; i < GetButtonsCount(); ++i)
+ State.Buttons[i] = (Pos.dwButtons & (1 << i)) != 0;
+ }
}
}
@@ -116,11 +135,11 @@ JoystickState Joystick::UpdateState() ////////////////////////////////////////////////////////////
-/// Get the number of axes supported by the joystick
+/// Check if the joystick supports the given axis
////////////////////////////////////////////////////////////
-unsigned int Joystick::GetAxesCount() const
+bool Joystick::HasAxis(Joy::Axis Axis) const
{
- return myNbAxes;
+ return myAxes[Axis];
}
diff --git a/src/SFML/Window/Win32/Joystick.hpp b/src/SFML/Window/Win32/Joystick.hpp index 8303a7e..1538069 100755 --- a/src/SFML/Window/Win32/Joystick.hpp +++ b/src/SFML/Window/Win32/Joystick.hpp @@ -58,12 +58,14 @@ public : JoystickState UpdateState();
////////////////////////////////////////////////////////////
- /// Get the number of axes supported by the joystick
+ /// Check if the joystick supports the given axis
///
- /// \return Number of axis
+ /// \param Axis : Axis to check
+ ///
+ /// \return True of the axis is supported, false otherwise
///
////////////////////////////////////////////////////////////
- unsigned int GetAxesCount() const;
+ bool HasAxis(Joy::Axis Axis) const;
////////////////////////////////////////////////////////////
/// Get the number of buttons supported by the joystick
@@ -78,9 +80,11 @@ private : ////////////////////////////////////////////////////////////
// Member data
////////////////////////////////////////////////////////////
- unsigned int myIndex; ///< Windows ID of the joystick
- unsigned int myNbAxes; ///< Number of axis supported by the joystick
- unsigned int myNbButtons; ///< Number of buttons supported by the joystick
+ bool myIsConnected; ///< Is there a joystick connected?
+ unsigned int myIndex; ///< Windows ID of the joystick
+ unsigned int myNbButtons; ///< Number of buttons supported by the joystick
+ bool myAxes[Joy::AxisCount]; ///< Supported axes
+ bool myHasContinuousPOV; ///< True if the driver supports continuous values for the POV
};
} // namespace priv
diff --git a/src/SFML/Window/Win32/WindowImplWin32.cpp b/src/SFML/Window/Win32/WindowImplWin32.cpp index 9148774..506cb73 100755 --- a/src/SFML/Window/Win32/WindowImplWin32.cpp +++ b/src/SFML/Window/Win32/WindowImplWin32.cpp @@ -36,13 +36,16 @@ #include <iostream>
#include <vector>
-// Old versions of MinGW lack the definition of XBUTTON1 and XBUTTON2
+// MinGW lacks the definition of some Win32 constants
#ifndef XBUTTON1
#define XBUTTON1 0x0001
#endif
#ifndef XBUTTON2
#define XBUTTON2 0x0002
#endif
+#ifndef MAPVK_VK_TO_VSC
+ #define MAPVK_VK_TO_VSC (0)
+#endif
namespace sf
@@ -150,10 +153,12 @@ myIsCursorIn (false) RegisterWindowClass();
// Compute position and size
- int Left = (GetDeviceCaps(GetDC(NULL), HORZRES) - Mode.Width) / 2;
- int Top = (GetDeviceCaps(GetDC(NULL), VERTRES) - Mode.Height) / 2;
+ HDC ScreenDC = GetDC(NULL);
+ int Left = (GetDeviceCaps(ScreenDC, HORZRES) - Mode.Width) / 2;
+ int Top = (GetDeviceCaps(ScreenDC, VERTRES) - Mode.Height) / 2;
int Width = myWidth = Mode.Width;
int Height = myHeight = Mode.Height;
+ ReleaseDC(NULL, ScreenDC);
// Choose the window style according to the Style parameter
DWORD Win32Style = WS_VISIBLE;
@@ -268,7 +273,7 @@ void WindowImplWin32::ProcessEvents() if (!myCallback)
{
MSG Message;
- while (PeekMessage(&Message, myHandle, 0, 0, PM_REMOVE))
+ while (PeekMessage(&Message, NULL, 0, 0, PM_REMOVE))
{
TranslateMessage(&Message);
DispatchMessage(&Message);
@@ -476,20 +481,16 @@ void WindowImplWin32::SwitchToFullscreen(const VideoMode& Mode) return;
}
- // Change window style (no border, no titlebar, ...)
- SetWindowLong(myHandle, GWL_STYLE, WS_POPUP);
+ // Make the window flags compatible with fullscreen mode
+ SetWindowLong(myHandle, GWL_STYLE, WS_POPUP | WS_CLIPCHILDREN | WS_CLIPSIBLINGS);
SetWindowLong(myHandle, GWL_EXSTYLE, WS_EX_APPWINDOW);
- // And resize it so that it fits the entire screen
+ // Resize the window so that it fits the entire screen
SetWindowPos(myHandle, HWND_TOP, 0, 0, Mode.Width, Mode.Height, SWP_FRAMECHANGED);
ShowWindow(myHandle, SW_SHOW);
// Set "this" as the current fullscreen window
ourFullscreenWindow = this;
-
- // SetPixelFormat can fail (really ?) if window style doesn't contain these flags
- long Style = GetWindowLong(myHandle, GWL_STYLE);
- SetWindowLong(myHandle, GWL_STYLE, Style | WS_CLIPCHILDREN | WS_CLIPSIBLINGS);
}
@@ -512,68 +513,76 @@ void WindowImplWin32::CreateContext(const VideoMode& Mode, WindowSettings& Param {
// Get the wglChoosePixelFormatARB function (it is an extension)
PFNWGLCHOOSEPIXELFORMATARBPROC wglChoosePixelFormatARB = reinterpret_cast<PFNWGLCHOOSEPIXELFORMATARBPROC>(wglGetProcAddress("wglChoosePixelFormatARB"));
-
- // Define the basic attributes we want for our window
- int IntAttributes[] =
- {
- WGL_DRAW_TO_WINDOW_ARB, GL_TRUE,
- WGL_SUPPORT_OPENGL_ARB, GL_TRUE,
- WGL_ACCELERATION_ARB, WGL_FULL_ACCELERATION_ARB,
- WGL_DOUBLE_BUFFER_ARB, GL_TRUE,
- WGL_SAMPLE_BUFFERS_ARB, (Params.AntialiasingLevel ? GL_TRUE : GL_FALSE),
- WGL_SAMPLES_ARB, Params.AntialiasingLevel,
- 0, 0
- };
-
- // Let's check how many formats are supporting our requirements
- int Formats[128];
- UINT NbFormats;
- float FloatAttributes[] = {0, 0};
- bool IsValid = wglChoosePixelFormatARB(myDeviceContext, IntAttributes, FloatAttributes, sizeof(Formats) / sizeof(*Formats), Formats, &NbFormats) != 0;
- if (!IsValid || (NbFormats == 0))
+ if (wglChoosePixelFormatARB)
{
- if (Params.AntialiasingLevel > 2)
+ // Define the basic attributes we want for our window
+ int IntAttributes[] =
{
- // No format matching our needs : reduce the multisampling level
- std::cerr << "Failed to find a pixel format supporting "
- << Params.AntialiasingLevel << " antialiasing levels ; trying with 2 levels" << std::endl;
-
- Params.AntialiasingLevel = IntAttributes[1] = 2;
- IsValid = wglChoosePixelFormatARB(myDeviceContext, IntAttributes, FloatAttributes, sizeof(Formats) / sizeof(*Formats), Formats, &NbFormats) != 0;
- }
-
+ WGL_DRAW_TO_WINDOW_ARB, GL_TRUE,
+ WGL_SUPPORT_OPENGL_ARB, GL_TRUE,
+ WGL_ACCELERATION_ARB, WGL_FULL_ACCELERATION_ARB,
+ WGL_DOUBLE_BUFFER_ARB, GL_TRUE,
+ WGL_SAMPLE_BUFFERS_ARB, (Params.AntialiasingLevel ? GL_TRUE : GL_FALSE),
+ WGL_SAMPLES_ARB, Params.AntialiasingLevel,
+ 0, 0
+ };
+
+ // Let's check how many formats are supporting our requirements
+ int Formats[128];
+ UINT NbFormats;
+ float FloatAttributes[] = {0, 0};
+ bool IsValid = wglChoosePixelFormatARB(myDeviceContext, IntAttributes, FloatAttributes, sizeof(Formats) / sizeof(*Formats), Formats, &NbFormats) != 0;
if (!IsValid || (NbFormats == 0))
{
- // Cannot find any pixel format supporting multisampling ; disabling antialiasing
- std::cerr << "Failed to find a pixel format supporting antialiasing ; antialiasing will be disabled" << std::endl;
- Params.AntialiasingLevel = 0;
+ if (Params.AntialiasingLevel > 2)
+ {
+ // No format matching our needs : reduce the multisampling level
+ std::cerr << "Failed to find a pixel format supporting "
+ << Params.AntialiasingLevel << " antialiasing levels ; trying with 2 levels" << std::endl;
+
+ Params.AntialiasingLevel = IntAttributes[11] = 2;
+ IsValid = wglChoosePixelFormatARB(myDeviceContext, IntAttributes, FloatAttributes, sizeof(Formats) / sizeof(*Formats), Formats, &NbFormats) != 0;
+ }
+
+ if (!IsValid || (NbFormats == 0))
+ {
+ // Cannot find any pixel format supporting multisampling ; disabling antialiasing
+ std::cerr << "Failed to find a pixel format supporting antialiasing ; antialiasing will be disabled" << std::endl;
+ Params.AntialiasingLevel = 0;
+ }
}
- }
- // Get the best format among the returned ones
- if (IsValid && (NbFormats > 0))
- {
- int BestScore = 0xFFFF;
- for (UINT i = 0; i < NbFormats; ++i)
+ // Get the best format among the returned ones
+ if (IsValid && (NbFormats > 0))
{
- // Get the current format's attributes
- PIXELFORMATDESCRIPTOR Attribs;
- Attribs.nSize = sizeof(PIXELFORMATDESCRIPTOR);
- Attribs.nVersion = 1;
- DescribePixelFormat(myDeviceContext, Formats[i], sizeof(PIXELFORMATDESCRIPTOR), &Attribs);
-
- // Evaluate the current configuration
- int Color = Attribs.cRedBits + Attribs.cGreenBits + Attribs.cBlueBits + Attribs.cAlphaBits;
- int Score = EvaluateConfig(Mode, Params, Color, Attribs.cDepthBits, Attribs.cStencilBits, Params.AntialiasingLevel);
-
- // Keep it if it's better than the current best
- if (Score < BestScore)
+ int BestScore = 0xFFFF;
+ for (UINT i = 0; i < NbFormats; ++i)
{
- BestScore = Score;
- BestFormat = Formats[i];
+ // Get the current format's attributes
+ PIXELFORMATDESCRIPTOR Attribs;
+ Attribs.nSize = sizeof(PIXELFORMATDESCRIPTOR);
+ Attribs.nVersion = 1;
+ DescribePixelFormat(myDeviceContext, Formats[i], sizeof(PIXELFORMATDESCRIPTOR), &Attribs);
+
+ // Evaluate the current configuration
+ int Color = Attribs.cRedBits + Attribs.cGreenBits + Attribs.cBlueBits + Attribs.cAlphaBits;
+ int Score = EvaluateConfig(Mode, Params, Color, Attribs.cDepthBits, Attribs.cStencilBits, Params.AntialiasingLevel);
+
+ // Keep it if it's better than the current best
+ if (Score < BestScore)
+ {
+ BestScore = Score;
+ BestFormat = Formats[i];
+ }
}
}
}
+ else
+ {
+ // wglChoosePixelFormatARB not supported ; disabling antialiasing
+ std::cerr << "Antialiasing is not supported ; it will be disabled" << std::endl;
+ Params.AntialiasingLevel = 0;
+ }
}
// Find a pixel format with no antialiasing, if not needed or not supported
@@ -590,6 +599,7 @@ void WindowImplWin32::CreateContext(const VideoMode& Mode, WindowSettings& Param PixelDescriptor.cColorBits = static_cast<BYTE>(Mode.BitsPerPixel);
PixelDescriptor.cDepthBits = static_cast<BYTE>(Params.DepthBits);
PixelDescriptor.cStencilBits = static_cast<BYTE>(Params.StencilBits);
+ PixelDescriptor.cAlphaBits = Mode.BitsPerPixel == 32 ? 8 : 0;
// Get the pixel format that best matches our requirements
BestFormat = ChoosePixelFormat(myDeviceContext, &PixelDescriptor);
@@ -745,10 +755,13 @@ void WindowImplWin32::ProcessEvent(UINT Message, WPARAM WParam, LPARAM LParam) // Text event
case WM_CHAR :
{
- Event Evt;
- Evt.Type = Event::TextEntered;
- Evt.Text.Unicode = static_cast<Uint32>(WParam);
- SendEvent(Evt);
+ if (myKeyRepeatEnabled || ((LParam & (1 << 30)) == 0))
+ {
+ Event Evt;
+ Evt.Type = Event::TextEntered;
+ Evt.Text.Unicode = static_cast<Uint32>(WParam);
+ SendEvent(Evt);
+ }
break;
}
@@ -756,26 +769,15 @@ void WindowImplWin32::ProcessEvent(UINT Message, WPARAM WParam, LPARAM LParam) case WM_KEYDOWN :
case WM_SYSKEYDOWN :
{
- if (myKeyRepeatEnabled || ((LParam & (1 << 30)) == 0))
+ if (myKeyRepeatEnabled || ((HIWORD(LParam) & KF_REPEAT) == 0))
{
Event Evt;
Evt.Type = Event::KeyPressed;
Evt.Key.Alt = HIWORD(GetAsyncKeyState(VK_MENU)) != 0;
Evt.Key.Control = HIWORD(GetAsyncKeyState(VK_CONTROL)) != 0;
Evt.Key.Shift = HIWORD(GetAsyncKeyState(VK_SHIFT)) != 0;
-
- if (WParam != VK_SHIFT)
- {
- Evt.Key.Code = VirtualKeyCodeToSF(WParam, LParam);
- SendEvent(Evt);
- }
- else
- {
- // Special case for shift, its state can't be retrieved directly
- Evt.Key.Code = GetShiftState(true);
- if (Evt.Key.Code != 0)
- SendEvent(Evt);
- }
+ Evt.Key.Code = VirtualKeyCodeToSF(WParam, LParam);
+ SendEvent(Evt);
}
break;
}
@@ -789,19 +791,8 @@ void WindowImplWin32::ProcessEvent(UINT Message, WPARAM WParam, LPARAM LParam) Evt.Key.Alt = HIWORD(GetAsyncKeyState(VK_MENU)) != 0;
Evt.Key.Control = HIWORD(GetAsyncKeyState(VK_CONTROL)) != 0;
Evt.Key.Shift = HIWORD(GetAsyncKeyState(VK_SHIFT)) != 0;
-
- if (WParam != VK_SHIFT)
- {
- Evt.Key.Code = VirtualKeyCodeToSF(WParam, LParam);
- SendEvent(Evt);
- }
- else
- {
- // Special case for shift, its state can't be retrieved directly
- Evt.Key.Code = GetShiftState(false);
- if (Evt.Key.Code != 0)
- SendEvent(Evt);
- }
+ Evt.Key.Code = VirtualKeyCodeToSF(WParam, LParam);
+ SendEvent(Evt);
break;
}
@@ -954,46 +945,27 @@ void WindowImplWin32::ProcessEvent(UINT Message, WPARAM WParam, LPARAM LParam) ////////////////////////////////////////////////////////////
-/// Check the state of the shift keys on a key event,
-/// and return the corresponding SF key code
-////////////////////////////////////////////////////////////
-Key::Code WindowImplWin32::GetShiftState(bool KeyDown)
-{
- static bool LShiftPrevDown = false;
- static bool RShiftPrevDown = false;
-
- bool LShiftDown = (HIWORD(GetAsyncKeyState(VK_LSHIFT)) != 0);
- bool RShiftDown = (HIWORD(GetAsyncKeyState(VK_RSHIFT)) != 0);
-
- Key::Code Code = Key::Code(0);
- if (KeyDown)
- {
- if (!LShiftPrevDown && LShiftDown) Code = Key::LShift;
- else if (!RShiftPrevDown && RShiftDown) Code = Key::RShift;
- }
- else
- {
- if (LShiftPrevDown && !LShiftDown) Code = Key::LShift;
- else if (RShiftPrevDown && !RShiftDown) Code = Key::RShift;
- }
-
- LShiftPrevDown = LShiftDown;
- RShiftPrevDown = RShiftDown;
-
- return Code;
-}
-
-
-////////////////////////////////////////////////////////////
/// Convert a Win32 virtual key code to a SFML key code
////////////////////////////////////////////////////////////
Key::Code WindowImplWin32::VirtualKeyCodeToSF(WPARAM VirtualKey, LPARAM Flags)
{
switch (VirtualKey)
{
- // VK_SHIFT is handled by the GetShiftState function
- case VK_MENU : return (Flags & (1 << 24)) ? Key::RAlt : Key::LAlt;
- case VK_CONTROL : return (Flags & (1 << 24)) ? Key::RControl : Key::LControl;
+ // Check the scancode to distinguish between left and right shift
+ case VK_SHIFT :
+ {
+ static UINT LShift = MapVirtualKey(VK_LSHIFT, MAPVK_VK_TO_VSC);
+ UINT scancode = (Flags & (0xFF << 16)) >> 16;
+ return scancode == LShift ? Key::LShift : Key::RShift;
+ }
+
+ // Check the "extended" flag to distinguish between left and right alt
+ case VK_MENU : return (HIWORD(Flags) & KF_EXTENDED) ? Key::RAlt : Key::LAlt;
+
+ // Check the "extended" flag to distinguish between left and right control
+ case VK_CONTROL : return (HIWORD(Flags) & KF_EXTENDED) ? Key::RControl : Key::LControl;
+
+ // Other keys are reported properly
case VK_LWIN : return Key::LSystem;
case VK_RWIN : return Key::RSystem;
case VK_APPS : return Key::Menu;
diff --git a/src/SFML/Window/Win32/WindowImplWin32.hpp b/src/SFML/Window/Win32/WindowImplWin32.hpp index 75c1c59..20557dc 100755 --- a/src/SFML/Window/Win32/WindowImplWin32.hpp +++ b/src/SFML/Window/Win32/WindowImplWin32.hpp @@ -194,17 +194,6 @@ private : void ProcessEvent(UINT Message, WPARAM WParam, LPARAM LParam);
////////////////////////////////////////////////////////////
- /// Check the state of the shift keys on a key event,
- /// and return the corresponding SF key code
- ///
- /// \param KeyDown : True for a keydown event, false for a keyup event
- ///
- /// \return SFML key code corresponding to the shift key
- ///
- ////////////////////////////////////////////////////////////
- static Key::Code GetShiftState(bool KeyDown);
-
- ////////////////////////////////////////////////////////////
/// Convert a Win32 virtual key code to a SFML key code
///
/// \param VirtualKey : Virtual key code to convert
diff --git a/src/SFML/Window/Window.cpp b/src/SFML/Window/Window.cpp index cfaae84..7ea1963 100755 --- a/src/SFML/Window/Window.cpp +++ b/src/SFML/Window/Window.cpp @@ -445,6 +445,10 @@ void Window::Initialize(priv::WindowImpl* Window) myWindow = Window;
myWindow->Initialize();
+ // Clear the event queue
+ while (!myEvents.empty())
+ myEvents.pop();
+
// Listen to events from the new window
myWindow->AddListener(this);
myWindow->AddListener(&myInput);
diff --git a/src/SFML/Window/WindowImpl.cpp b/src/SFML/Window/WindowImpl.cpp index bc60543..dd0584f 100755 --- a/src/SFML/Window/WindowImpl.cpp +++ b/src/SFML/Window/WindowImpl.cpp @@ -125,7 +125,7 @@ void WindowImpl::RemoveListener(WindowListener* Listener) void WindowImpl::Initialize()
{
// Initialize the joysticks
- for (unsigned int i = 0; i < JoysticksCount; ++i)
+ for (unsigned int i = 0; i < Joy::Count; ++i)
{
myJoysticks[i].Initialize(i);
myJoyStates[i] = myJoysticks[i].UpdateState();
@@ -214,25 +214,29 @@ int WindowImpl::EvaluateConfig(const VideoMode& Mode, const WindowSettings& Sett ////////////////////////////////////////////////////////////
void WindowImpl::ProcessJoystickEvents()
{
- for (unsigned int i = 0; i < JoysticksCount; ++i)
+ for (unsigned int i = 0; i < Joy::Count; ++i)
{
// Copy the previous state of the joystick and get the new one
JoystickState PreviousState = myJoyStates[i];
myJoyStates[i] = myJoysticks[i].UpdateState();
// Axis
- for (unsigned int j = 0; j < myJoysticks[i].GetAxesCount(); ++j)
+ for (unsigned int j = 0; j < Joy::AxisCount; ++j)
{
- float PrevPos = PreviousState.Axis[j];
- float CurrPos = myJoyStates[i].Axis[j];
- if (fabs(CurrPos - PrevPos) >= myJoyThreshold)
+ Joy::Axis Axis = static_cast<Joy::Axis>(j);
+ if (myJoysticks[i].HasAxis(Axis))
{
- Event Event;
- Event.Type = Event::JoyMoved;
- Event.JoyMove.JoystickId = i;
- Event.JoyMove.Axis = static_cast<Joy::Axis>(j);
- Event.JoyMove.Position = CurrPos;
- SendEvent(Event);
+ float PrevPos = PreviousState.Axis[j];
+ float CurrPos = myJoyStates[i].Axis[j];
+ if (fabs(CurrPos - PrevPos) >= myJoyThreshold)
+ {
+ Event Event;
+ Event.Type = Event::JoyMoved;
+ Event.JoyMove.JoystickId = i;
+ Event.JoyMove.Axis = Axis;
+ Event.JoyMove.Position = CurrPos;
+ SendEvent(Event);
+ }
}
}
diff --git a/src/SFML/Window/WindowImpl.hpp b/src/SFML/Window/WindowImpl.hpp index 15d29ec..09f8b54 100755 --- a/src/SFML/Window/WindowImpl.hpp +++ b/src/SFML/Window/WindowImpl.hpp @@ -291,17 +291,12 @@ private : virtual void ProcessEvents() = 0;
////////////////////////////////////////////////////////////
- // Total number of joysticks supported
- ////////////////////////////////////////////////////////////
- enum {JoysticksCount = 2};
-
- ////////////////////////////////////////////////////////////
// Member data
////////////////////////////////////////////////////////////
- std::set<WindowListener*> myListeners; ///< Array of listeners connected to the window
- Joystick myJoysticks[JoysticksCount]; ///< Joysticks to observe
- JoystickState myJoyStates[JoysticksCount]; ///< Current states of the joysticks
- float myJoyThreshold; ///< Joystick threshold (minimum motion for MOVE event to be generated)
+ std::set<WindowListener*> myListeners; ///< Array of listeners connected to the window
+ Joystick myJoysticks[Joy::Count]; ///< Joysticks to observe
+ JoystickState myJoyStates[Joy::Count]; ///< Current states of the joysticks
+ float myJoyThreshold; ///< Joystick threshold (minimum motion for MOVE event to be generated)
};
} // namespace priv
|