rpgxef/code/client/snd_codec_mp3.c

/*
===========================================================================
Copyright (C) 1999-2005 Id Software, Inc.
Copyright (C) 2005 Stuart Dalton (badcdev@gmail.com)
Copyright (C) 2005-2006 Joerg Dietrich <dietrich_joerg@gmx.de>
Copyright (C) 2006 Thilo Schulz <arny@ats.s.bawue.de>

This file is part of Quake III Arena source code.

Quake III Arena source code is free software; you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the License,
or (at your option) any later version.

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

You should have received a copy of the GNU General Public License
along with Quake III Arena source code; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
===========================================================================
*/

// MP3 support is enabled by this define
#if USE_CODEC_MP3

// includes for the Q3 sound system
#include "client.h"
#include "snd_codec.h"

// includes for the MP3 codec
#include "../libmad/mad.h"

#define MP3_SAMPLE_WIDTH		2
#define MP3_PCMSAMPLES_PERSLICE		32

// buffer size used when reading through the mp3
#define MP3_DATA_BUFSIZ			128*1024

// undefine this if you don't want any dithering.
#define MP3_DITHERING

// Q3 MP3 codec
snd_codec_t mp3_codec =
{
	"mp3",
	S_MP3_CodecLoad,
	S_MP3_CodecOpenStream,
	S_MP3_CodecReadStream,
	S_MP3_CodecCloseStream,
	NULL
};

// structure used for info purposes
struct snd_codec_mp3_info
{
	byte encbuf[MP3_DATA_BUFSIZ];	// left over bytes not consumed
					// by the decoder.
	struct mad_stream madstream;	// uses encbuf as buffer.
	struct mad_frame madframe;	// control structures for libmad.
	struct mad_synth madsynth;

	byte *pcmbuf;			// buffer for not-used samples.
	int buflen;			// length of buffer data.
	int pcmbufsize;			// amount of allocated memory for
					// pcmbuf. This should have at least
					// the size of a decoded mp3 frame.	

	byte *dest;			// copy decoded data here.
	int destlen;			// amount of already copied data.
	int destsize;			// amount of bytes we must decode.
};

/*************** MP3 utility functions ***************/

/*
=================
S_MP3_ReadData
=================
*/

// feed libmad with data
int S_MP3_ReadData(snd_stream_t *stream, struct mad_stream *madstream, byte *encbuf, int encbufsize)
{
	int retval;
	int leftover;
	
	if(!stream)
		return -1;
	
	leftover =  madstream->bufend - madstream->next_frame;
	if(leftover > 0)
		memmove(encbuf, madstream->this_frame, leftover);


	// Fill the buffer right to the end
	
	retval = FS_Read(&encbuf[leftover], encbufsize - leftover, stream->file);

	if(retval <= 0)
	{
		// EOF reached, that's ok.
		return 0;
	}
	
	mad_stream_buffer(madstream, encbuf, retval + leftover);
	
	return retval;
}


/*
=================
S_MP3_Scanfile

to determine the samplecount, we apparently must get *all* headers :(
I basically used the xmms-mad plugin source to see how this stuff works.

returns a value < 0 on error.
=================
*/

int S_MP3_Scanfile(snd_stream_t *stream)
{
	struct mad_stream madstream;
	struct mad_header madheader;
	int retval;
	int samplecount;
	byte encbuf[MP3_DATA_BUFSIZ];

	// error out on invalid input.
	if(!stream)
		return -1;

	mad_stream_init(&madstream);
	mad_header_init(&madheader);
	
	while(1)
	{
		retval = S_MP3_ReadData(stream, &madstream, encbuf, sizeof(encbuf));
		if(retval < 0)
			return -1;
		else if(retval == 0)
			break;
		
		// Start decoding the headers.
		while(1)
		{
			if((retval = mad_header_decode(&madheader, &madstream)) < 0)
			{
				if(madstream.error == MAD_ERROR_BUFLEN)
				{
					// We need to read more data
					break;
				}

				if(!MAD_RECOVERABLE (madstream.error))
				{
					// unrecoverable error... we must bail out.
					return retval;
				}

				mad_stream_skip(&madstream, madstream.skiplen);
				continue;
			}
			
			// we got a valid header.
			
			if(madheader.layer != MAD_LAYER_III)
			{
				// we don't support non-mp3s
				return -1;
			}

			if(!stream->info.samples)
			{
				// This here is the very first frame. Set initial values now,
				// that we expect to stay constant throughout the whole mp3.
				
				stream->info.rate = madheader.samplerate;
				stream->info.width = MP3_SAMPLE_WIDTH;
				stream->info.channels = MAD_NCHANNELS(&madheader);
				stream->info.samples = 0;
				stream->info.size = 0;				// same here.
				stream->info.dataofs = 0;
			}
			else
			{
				// Check whether something changed that shouldn't.
				
				if(stream->info.rate != madheader.samplerate ||
				   stream->info.channels != MAD_NCHANNELS(&madheader))
					return -1;
			}

			// Update the counters
			samplecount = MAD_NSBSAMPLES(&madheader) * MP3_PCMSAMPLES_PERSLICE;
			stream->info.samples += samplecount;
			stream->info.size += samplecount * stream->info.channels * stream->info.width;			
		}
	}
	
	// Reset the file pointer so we can do the real decoding.
	FS_Seek(stream->file, 0, FS_SEEK_SET);
	
	return 0;
}

/************************ dithering functions ***************************/

#ifdef MP3_DITHERING

// All dithering done here is taken from the GPL'ed xmms-mad plugin.

/* Copyright (C) 1997 Makoto Matsumoto and Takuji Nishimura.       */
/* Any feedback is very welcome. For any question, comments,       */
/* see http://www.math.keio.ac.jp/matumoto/emt.html or email       */
/* matumoto@math.keio.ac.jp                                        */

/* Period parameters */
#define MP3_DITH_N 624
#define MP3_DITH_M 397
#define MATRIX_A 0x9908b0df   /* constant vector a */
#define UPPER_MASK 0x80000000 /* most significant w-r bits */
#define LOWER_MASK 0x7fffffff /* least significant r bits */

/* Tempering parameters */
#define TEMPERING_MASK_B 0x9d2c5680
#define TEMPERING_MASK_C 0xefc60000
#define TEMPERING_SHIFT_U(y)  (y >> 11)
#define TEMPERING_SHIFT_S(y)  (y << 7)
#define TEMPERING_SHIFT_T(y)  (y << 15)
#define TEMPERING_SHIFT_L(y)  (y >> 18)

static unsigned long mt[MP3_DITH_N]; /* the array for the state vector  */
static int mti=MP3_DITH_N+1; /* mti==MP3_DITH_N+1 means mt[MP3_DITH_N] is not initialized */

/* initializing the array with a NONZERO seed */
void sgenrand(unsigned long seed)
{
    /* setting initial seeds to mt[MP3_DITH_N] using         */
    /* the generator Line 25 of Table 1 in          */
    /* [KNUTH 1981, The Art of Computer Programming */
    /*    Vol. 2 (2nd Ed.), pp102]                  */
    mt[0]= seed & 0xffffffff;
    for (mti=1; mti<MP3_DITH_N; mti++)
        mt[mti] = (69069 * mt[mti-1]) & 0xffffffff;
}

unsigned long genrand(void)
{
    unsigned long y;
    static unsigned long mag01[2]={0x0, MATRIX_A};
    /* mag01[x] = x * MATRIX_A  for x=0,1 */

    if (mti >= MP3_DITH_N) { /* generate MP3_DITH_N words at one time */
        int kk;

        if (mti == MP3_DITH_N+1)   /* if sgenrand() has not been called, */
            sgenrand(4357); /* a default initial seed is used   */

        for (kk=0;kk<MP3_DITH_N-MP3_DITH_M;kk++) {
            y = (mt[kk]&UPPER_MASK)|(mt[kk+1]&LOWER_MASK);
            mt[kk] = mt[kk+MP3_DITH_M] ^ (y >> 1) ^ mag01[y & 0x1];
        }
        for (;kk<MP3_DITH_N-1;kk++) {
            y = (mt[kk]&UPPER_MASK)|(mt[kk+1]&LOWER_MASK);
            mt[kk] = mt[kk+(MP3_DITH_M-MP3_DITH_N)] ^ (y >> 1) ^ mag01[y & 0x1];
        }
        y = (mt[MP3_DITH_N-1]&UPPER_MASK)|(mt[0]&LOWER_MASK);
        mt[MP3_DITH_N-1] = mt[MP3_DITH_M-1] ^ (y >> 1) ^ mag01[y & 0x1];

        mti = 0;
    }

    y = mt[mti++];
    y ^= TEMPERING_SHIFT_U(y);
    y ^= TEMPERING_SHIFT_S(y) & TEMPERING_MASK_B;
    y ^= TEMPERING_SHIFT_T(y) & TEMPERING_MASK_C;
    y ^= TEMPERING_SHIFT_L(y);

    return y;
}

long triangular_dither_noise(int nbits) {
    // parameter nbits : the peak-to-peak amplitude desired (in bits)
    //  use with nbits set to    2 + nber of bits to be trimmed.
    // (because triangular is made from two uniformly distributed processes,
    // it starts at 2 bits peak-to-peak amplitude)
    // see The Theory of Dithered Quantization by Robert Alexander Wannamaker
    // for complete proof of why that's optimal

    long v = (genrand()/2 - genrand()/2); // in ]-2^31, 2^31[
    //int signe = (v>0) ? 1 : -1;
    long P = 1 << (32 - nbits); // the power of 2
    v /= P;
    // now v in ]-2^(nbits-1), 2^(nbits-1) [

    return v;
}

#endif // MP3_DITHERING

/************************ decoder functions ***************************/

/*
=================
S_MP3_Scale

Converts the signal to 16 bit LE-PCM data and does dithering.

- borrowed from xmms-mad plugin source.
=================
*/

/*
 * xmms-mad - mp3 plugin for xmms
 * Copyright (C) 2001-2002 Sam Clegg
 */

signed int S_MP3_Scale(mad_fixed_t sample)
{
	int n_bits_to_loose = MAD_F_FRACBITS + 1 - 16;
#ifdef MP3_DITHERING
	int dither;
#endif
	
	// round
	sample += (1L << (n_bits_to_loose - 1));

#ifdef MP3_DITHERING
	dither = triangular_dither_noise(n_bits_to_loose + 1);
	sample += dither;
#endif

	/* clip */
	if (sample >= MAD_F_ONE)
		sample = MAD_F_ONE - 1;
	else if (sample < -MAD_F_ONE)
		sample = -MAD_F_ONE;

	/* quantize */
	return sample >> n_bits_to_loose;
}



/*
=================
S_MP3_PCMCopy

Copy and convert pcm data until bytecount bytes have been written.
return the position in pcm->samples.
indicate the amount of actually written bytes in wrotecnt.
=================
*/

int S_MP3_PCMCopy(byte *buf, struct mad_pcm *pcm, int bufofs,
			 int sampleofs, int bytecount, int *wrotecnt)
{
	int written = 0;
	signed int sample;
	int framesize = pcm->channels * MP3_SAMPLE_WIDTH;

	// add new pcm data.
	while(written < bytecount && sampleofs < pcm->length)
	{
		sample = S_MP3_Scale(pcm->samples[0][sampleofs]);

#ifdef Q3_BIG_ENDIAN
		// output to 16 bit big endian PCM
		buf[bufofs++] = (sample >> 8) & 0xff;
		buf[bufofs++] = sample & 0xff;
#else
		// output to 16 bit little endian PCM
		buf[bufofs++] = sample & 0xff;
		buf[bufofs++] = (sample >> 8) & 0xff;
#endif
		
		if(pcm->channels == 2)
		{
			sample = S_MP3_Scale(pcm->samples[1][sampleofs]);

#ifdef Q3_BIG_ENDIAN
			buf[bufofs++] = (sample >> 8) & 0xff;
			buf[bufofs++] = sample & 0xff;
#else
			buf[bufofs++] = sample & 0xff;
			buf[bufofs++] = (sample >> 8) & 0xff;
#endif
		}
		
		sampleofs++;
		written += framesize;
	}	

	if(wrotecnt)
		*wrotecnt = written;

	return sampleofs;
}


/*
=================
S_MP3_Decode
=================
*/

// gets executed for every decoded frame.
int S_MP3_Decode(snd_stream_t *stream)
{
	struct snd_codec_mp3_info *mp3info;
	struct mad_stream *madstream;
	struct mad_frame *madframe;
	struct mad_synth *madsynth;
	struct mad_pcm *pcm;
	int cursize;
	int samplecount;
	int needcount;
	int wrote;
	int retval;

	if(!stream)
		return -1;

	mp3info = stream->ptr;
	madstream = &mp3info->madstream;
	madframe = &mp3info->madframe;

	if(mad_frame_decode(madframe, madstream))
	{
		if(madstream->error == MAD_ERROR_BUFLEN)
		{
			// we need more data. Read another chunk.
			retval = S_MP3_ReadData(stream, madstream, mp3info->encbuf, sizeof(mp3info->encbuf));

			// call myself again now that buffer is full.
			if(retval > 0)
				retval = S_MP3_Decode(stream);
		}
		else if(MAD_RECOVERABLE(madstream->error))
		{
			mad_stream_skip(madstream, madstream->skiplen);
			return S_MP3_Decode(stream);
		}
		else
			retval = -1;

		return retval;
	}

	// check whether this really is an mp3
	if(madframe->header.layer != MAD_LAYER_III)
		return -1;

	// generate pcm data
	madsynth = &mp3info->madsynth;
	mad_synth_frame(madsynth, madframe);

	pcm = &madsynth->pcm;

	// perform a few checks to see whether something changed that shouldn't.
		
	if(stream->info.rate != pcm->samplerate ||
	   stream->info.channels != pcm->channels)
	{
		return -1;
	}
	// see whether we have got enough data now.
	cursize = pcm->length * pcm->channels * stream->info.width;
	needcount = mp3info->destsize - mp3info->destlen;

	// Copy exactly as many samples as required.
	samplecount = S_MP3_PCMCopy(mp3info->dest, pcm,
				    mp3info->destlen, 0, needcount, &wrote);
	mp3info->destlen += wrote;
	
	if(samplecount < pcm->length)
	{
		// Not all samples got copied. Copy the rest into the pcm buffer.
		samplecount = S_MP3_PCMCopy(mp3info->pcmbuf, pcm,
					    mp3info->buflen,
					    samplecount,
					    mp3info->pcmbufsize - mp3info->buflen,
					    &wrote);
		mp3info->buflen += wrote;
		

		if(samplecount < pcm->length)
		{
			// The pcm buffer was not large enough. Make it bigger.
			byte *newbuf = Z_Malloc(cursize);
			
			if(mp3info->pcmbuf)
			{
				memcpy(newbuf, mp3info->pcmbuf, mp3info->buflen);
				Z_Free(mp3info->pcmbuf);
			}
			
			mp3info->pcmbuf = newbuf;
			mp3info->pcmbufsize = cursize;
			
			samplecount = S_MP3_PCMCopy(mp3info->pcmbuf, pcm,
						    mp3info->buflen,
						    samplecount,
						    mp3info->pcmbufsize - mp3info->buflen,
						    &wrote);
			mp3info->buflen += wrote;		
		}		
		
		// we're definitely done.
		retval = 0;
	}
	else if(mp3info->destlen >= mp3info->destsize)
		retval = 0;
	else
		retval = 1;

	return retval;
}

/*************** Callback functions for quake3 ***************/

/*
=================
S_MP3_CodecOpenStream
=================
*/

snd_stream_t *S_MP3_CodecOpenStream(const char *filename)
{
	snd_stream_t *stream;
	struct snd_codec_mp3_info *mp3info;

	// Open the stream
	stream = S_CodecUtilOpen(filename, &mp3_codec);
	if(!stream || stream->length <= 0)
		return NULL;

	// We have to scan through the MP3 to determine the important mp3 info.
	if(S_MP3_Scanfile(stream) < 0)
	{
		// scanning didn't work out...
		S_CodecUtilClose(&stream);
		return NULL;
	}

	// Initialize the mp3 info structure we need for streaming
	mp3info = Z_Malloc(sizeof(*mp3info));
	if(!mp3info)
	{
		S_CodecUtilClose(&stream);
		return NULL;
	}

	stream->ptr = mp3info;

	// initialize the libmad control structures.
	mad_stream_init(&mp3info->madstream);
	mad_frame_init(&mp3info->madframe);
	mad_synth_init(&mp3info->madsynth);

	if(S_MP3_ReadData(stream, &mp3info->madstream, mp3info->encbuf, sizeof(mp3info->encbuf)) <= 0)
	{
		// we didnt read anything, that's bad.
		S_MP3_CodecCloseStream(stream);
		return NULL;
	}

	return stream;
}

/*
=================
S_MP3_CodecCloseStream
=================
*/

// free all memory we allocated.
void S_MP3_CodecCloseStream(snd_stream_t *stream)
{
	struct snd_codec_mp3_info *mp3info;
	
	if(!stream)
		return;
		
	// free all data in our mp3info tree

	if(stream->ptr)
	{
		mp3info = stream->ptr;

		if(mp3info->pcmbuf)
			Z_Free(mp3info->pcmbuf);

		mad_synth_finish(&mp3info->madsynth);
		mad_frame_finish(&mp3info->madframe);
		mad_stream_finish(&mp3info->madstream);
	
		Z_Free(stream->ptr);
	}

	S_CodecUtilClose(&stream);
}

/*
=================
S_MP3_CodecReadStream
=================
*/
int S_MP3_CodecReadStream(snd_stream_t *stream, int bytes, void *buffer)
{
	struct snd_codec_mp3_info *mp3info;
	int retval;
	
	if(!stream)
		return -1;
		
	mp3info = stream->ptr;

	// Make sure we get complete frames all the way through.
	bytes -= bytes % (stream->info.channels * stream->info.width);

	if(mp3info->buflen)
	{
		if(bytes < mp3info->buflen)
		{
			// we still have enough bytes in our decoded pcm buffer
			memcpy(buffer, mp3info->pcmbuf, bytes);
		
			// remove the portion from our buffer.
			mp3info->buflen -= bytes;
			memmove(mp3info->pcmbuf, &mp3info->pcmbuf[bytes], mp3info->buflen);
			return bytes;
		}
		else
		{
			// copy over the samples we already have.
			memcpy(buffer, mp3info->pcmbuf, mp3info->buflen);
			mp3info->destlen = mp3info->buflen;
			mp3info->buflen = 0;
		}
	}
	else
		mp3info->destlen = 0;
	
	mp3info->dest = buffer;
	mp3info->destsize = bytes;

	do
	{
		retval = S_MP3_Decode(stream);
	} while(retval > 0);
	
	// if there was an error return nothing.
	if(retval < 0)
		return 0;
	
	return mp3info->destlen;
}

/*
=====================================================================
S_MP3_CodecLoad

We handle S_MP3_CodecLoad as a special case of the streaming functions 
where we read the whole stream at once.
======================================================================
*/
void *S_MP3_CodecLoad(const char *filename, snd_info_t *info)
{
	snd_stream_t *stream;
	byte *pcmbuffer;

	// check if input is valid
	if(!filename)
		return NULL;

	stream = S_MP3_CodecOpenStream(filename);
	
	if(!stream)
		return NULL;
		
        // copy over the info
        info->rate = stream->info.rate;
        info->width = stream->info.width;
        info->channels = stream->info.channels;
        info->samples = stream->info.samples;
        info->dataofs = stream->info.dataofs;
	
	// allocate enough buffer for all pcm data
	pcmbuffer = Hunk_AllocateTempMemory(stream->info.size);
	if(!pcmbuffer)
	{
		S_MP3_CodecCloseStream(stream);
		return NULL;
	}

	info->size = S_MP3_CodecReadStream(stream, stream->info.size, pcmbuffer);

	if(info->size <= 0)
	{
		// we didn't read anything at all. darn.
		Hunk_FreeTempMemory(pcmbuffer);
		pcmbuffer = NULL;
	}

	S_MP3_CodecCloseStream(stream);

	return pcmbuffer;
}

#endif // USE_CODEC_MP3