view decoders/timidity/resample.c @ 474:c66080364dff

Most decoders now report total sample play time, now. Technically, this breaks binary compatibility with the 1.0 branch, since it extends the Sound_Sample struct, but most (all?) programs are just passing pointers allocated by SDL_sound around, and might be okay. Source-level compatibility is not broken...yet! :) --ryan. -------- Original Message -------- Subject: SDL_sound patch: Finding total length of time of sound file. Date: Sun, 26 Jan 2003 09:31:17 -0800 (PST) Hi Ryan, I am working with Eric Wing and helping him modify SDL_sound. AS part of our efforts in improving and enhancing SDL_sound, we like to submit this patch. We modified the codecs to find the total time of a sound file. Below is the explanation of the patch. The patch is appended as an attachment to this email. * MOTIVATION: We needed the ability to get the total play time of a sample (And we noticed that we're not the only ones). Since SDL_sound blocks direct access to the specific decoders, there is no way for a user to know this information short of decoding the whole thing. Because of this, we believe this will be a useful addition, even though the accuracy may not be perfect (subject to each decoder) or the information may not always be available. * CONTRIBUTORS: Wesley Leong (modified the majority of the codecs and verified the results) Eric Wing (showed everyone how to do modify codec, modified mikmod) Wang Lam (modified a handful of codecs, researched into specs and int overflow) Ahilan Anantha (modified a few codecs and helped with integer math) * GENERAL ISSUES: We chose the value to be milliseconds as an Sint32. Milliseconds because that's what Sound_Seek takes as a parameter and -1 to allow for instances/codecs where the value could not be determined. We are not sure if this is the final convention you want, so we are willing to work with you on this. We also expect the total_time field to be set on open and never again modified by SDL_sound. Users may access it directly much like the sample buffer and buffer_size. We thought about recomputing the time on DecodeAll, but since users may seek or decode small chunks first, not all the data may be there. So this is better done by the user. This may be good information to document. Currently, all the main codecs are implemented except for QuickTime.
author Ryan C. Gordon <icculus@icculus.org>
date Sat, 08 May 2004 08:19:50 +0000
parents a73c51c12452
children f33471c47efe
line wrap: on
line source

/*

    TiMidity -- Experimental MIDI to WAVE converter
    Copyright (C) 1995 Tuukka Toivonen <toivonen@clinet.fi>

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

    This program 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 this program; if not, write to the Free Software
    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

    resample.c
*/

#if HAVE_CONFIG_H
#  include <config.h>
#endif

#include <math.h>
#include <stdio.h>
#include <stdlib.h>

#include "SDL_sound.h"

#define __SDL_SOUND_INTERNAL__
#include "SDL_sound_internal.h"

#include "timidity.h"
#include "options.h"
#include "common.h"
#include "instrum.h"
#include "playmidi.h"
#include "tables.h"
#include "resample.h"

/*************** resampling with fixed increment *****************/

static sample_t *rs_plain(MidiSong *song, int v, Sint32 *countptr)
{

  /* Play sample until end, then free the voice. */

  sample_t v1, v2;
  Voice 
    *vp=&(song->voice[v]);
  sample_t 
    *dest=song->resample_buffer,
    *src=vp->sample->data;
  Sint32 
    ofs=vp->sample_offset,
    incr=vp->sample_increment,
    le=vp->sample->data_length,
    count=*countptr;
  Sint32 i;

  if (incr<0) incr = -incr; /* In case we're coming out of a bidir loop */

  /* Precalc how many times we should go through the loop.
     NOTE: Assumes that incr > 0 and that ofs <= le */
  i = (le - ofs) / incr + 1;

  if (i > count)
    {
      i = count;
      count = 0;
    } 
  else count -= i;

  while (i--) 
    {
      v1 = src[ofs >> FRACTION_BITS];
      v2 = src[(ofs >> FRACTION_BITS)+1];
      *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
      ofs += incr;
    }

  if (ofs >= le) 
    {
      if (ofs == le)
	*dest++ = src[ofs >> FRACTION_BITS];
      vp->status=VOICE_FREE;
      *countptr-=count+1;
    }
  
  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

static sample_t *rs_loop(MidiSong *song, Voice *vp, Sint32 count)
{

  /* Play sample until end-of-loop, skip back and continue. */

  sample_t v1, v2;
  Sint32 
    ofs=vp->sample_offset, 
    incr=vp->sample_increment,
    le=vp->sample->loop_end, 
    ll=le - vp->sample->loop_start;
  sample_t
    *dest=song->resample_buffer,
    *src=vp->sample->data;
  Sint32 i;
  
  while (count) 
    {
      if (ofs >= le)
	/* NOTE: Assumes that ll > incr and that incr > 0. */
	ofs -= ll;
      /* Precalc how many times we should go through the loop */
      i = (le - ofs) / incr + 1;
      if (i > count) 
	{
	  i = count;
	  count = 0;
	} 
      else count -= i;
      while (i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
    }

  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

static sample_t *rs_bidir(MidiSong *song, Voice *vp, Sint32 count)
{
  sample_t v1, v2;
  Sint32 
    ofs=vp->sample_offset,
    incr=vp->sample_increment,
    le=vp->sample->loop_end,
    ls=vp->sample->loop_start;
  sample_t 
    *dest=song->resample_buffer, 
    *src=vp->sample->data;
  Sint32
    le2 = le<<1, 
    ls2 = ls<<1,
    i;
  /* Play normally until inside the loop region */

  if (ofs <= ls) 
    {
      /* NOTE: Assumes that incr > 0, which is NOT always the case
	 when doing bidirectional looping.  I have yet to see a case
	 where both ofs <= ls AND incr < 0, however. */
      i = (ls - ofs) / incr + 1;
      if (i > count) 
	{
	  i = count;
	  count = 0;
	} 
      else count -= i;
      while (i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
    }

  /* Then do the bidirectional looping */
  
  while(count) 
    {
      /* Precalc how many times we should go through the loop */
      i = ((incr > 0 ? le : ls) - ofs) / incr + 1;
      if (i > count) 
	{
	  i = count;
	  count = 0;
	} 
      else count -= i;
      while (i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
      if (ofs>=le) 
	{
	  /* fold the overshoot back in */
	  ofs = le2 - ofs;
	  incr *= -1;
	} 
      else if (ofs <= ls) 
	{
	  ofs = ls2 - ofs;
	  incr *= -1;
	}
    }

  vp->sample_increment=incr;
  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

/*********************** vibrato versions ***************************/

/* We only need to compute one half of the vibrato sine cycle */
static int vib_phase_to_inc_ptr(int phase)
{
  if (phase < VIBRATO_SAMPLE_INCREMENTS/2)
    return VIBRATO_SAMPLE_INCREMENTS/2-1-phase;
  else if (phase >= 3*VIBRATO_SAMPLE_INCREMENTS/2)
    return 5*VIBRATO_SAMPLE_INCREMENTS/2-1-phase;
  else
    return phase-VIBRATO_SAMPLE_INCREMENTS/2;
}

static Sint32 update_vibrato(MidiSong *song, Voice *vp, int sign)
{
  Sint32 depth;
  int phase, pb;
  double a;

  if (vp->vibrato_phase++ >= 2*VIBRATO_SAMPLE_INCREMENTS-1)
    vp->vibrato_phase=0;
  phase=vib_phase_to_inc_ptr(vp->vibrato_phase);
  
  if (vp->vibrato_sample_increment[phase])
    {
      if (sign)
	return -vp->vibrato_sample_increment[phase];
      else
	return vp->vibrato_sample_increment[phase];
    }

  /* Need to compute this sample increment. */
    
  depth=vp->sample->vibrato_depth<<7;

  if (vp->vibrato_sweep)
    {
      /* Need to update sweep */
      vp->vibrato_sweep_position += vp->vibrato_sweep;
      if (vp->vibrato_sweep_position >= (1<<SWEEP_SHIFT))
	vp->vibrato_sweep=0;
      else
	{
	  /* Adjust depth */
	  depth *= vp->vibrato_sweep_position;
	  depth >>= SWEEP_SHIFT;
	}
    }

  a = FSCALE(((double)(vp->sample->sample_rate) *
	      (double)(vp->frequency)) /
	     ((double)(vp->sample->root_freq) *
	      (double)(song->rate)),
	     FRACTION_BITS);

  pb=(int)((sine(vp->vibrato_phase * 
		 (SINE_CYCLE_LENGTH/(2*VIBRATO_SAMPLE_INCREMENTS)))
	    * (double)(depth) * VIBRATO_AMPLITUDE_TUNING));

  if (pb<0)
    {
      pb=-pb;
      a /= bend_fine[(pb>>5) & 0xFF] * bend_coarse[pb>>13];
    }
  else
    a *= bend_fine[(pb>>5) & 0xFF] * bend_coarse[pb>>13];
  
  /* If the sweep's over, we can store the newly computed sample_increment */
  if (!vp->vibrato_sweep)
    vp->vibrato_sample_increment[phase]=(Sint32) a;

  if (sign)
    a = -a; /* need to preserve the loop direction */

  return (Sint32) a;
}

static sample_t *rs_vib_plain(MidiSong *song, int v, Sint32 *countptr)
{

  /* Play sample until end, then free the voice. */

  sample_t v1, v2;
  Voice *vp=&(song->voice[v]);
  sample_t 
    *dest=song->resample_buffer, 
    *src=vp->sample->data;
  Sint32 
    le=vp->sample->data_length,
    ofs=vp->sample_offset, 
    incr=vp->sample_increment, 
    count=*countptr;
  int 
    cc=vp->vibrato_control_counter;

  /* This has never been tested */

  if (incr<0) incr = -incr; /* In case we're coming out of a bidir loop */

  while (count--)
    {
      if (!cc--)
	{
	  cc=vp->vibrato_control_ratio;
	  incr=update_vibrato(song, vp, 0);
	}
      v1 = src[ofs >> FRACTION_BITS];
      v2 = src[(ofs >> FRACTION_BITS)+1];
      *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
      ofs += incr;
      if (ofs >= le)
	{
	  if (ofs == le)
	    *dest++ = src[ofs >> FRACTION_BITS];
	  vp->status=VOICE_FREE;
	  *countptr-=count+1;
	  break;
	}
    }
  
  vp->vibrato_control_counter=cc;
  vp->sample_increment=incr;
  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

static sample_t *rs_vib_loop(MidiSong *song, Voice *vp, Sint32 count)
{

  /* Play sample until end-of-loop, skip back and continue. */
  
  sample_t v1, v2;
  Sint32 
    ofs=vp->sample_offset, 
    incr=vp->sample_increment, 
    le=vp->sample->loop_end,
    ll=le - vp->sample->loop_start;
  sample_t 
    *dest=song->resample_buffer, 
    *src=vp->sample->data;
  int 
    cc=vp->vibrato_control_counter;
  Sint32 i;
  int
    vibflag=0;

  while (count) 
    {
      /* Hopefully the loop is longer than an increment */
      if(ofs >= le)
	ofs -= ll;
      /* Precalc how many times to go through the loop, taking
	 the vibrato control ratio into account this time. */
      i = (le - ofs) / incr + 1;
      if(i > count) i = count;
      if(i > cc)
	{
	  i = cc;
	  vibflag = 1;
	} 
      else cc -= i;
      count -= i;
      while(i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
      if(vibflag) 
	{
	  cc = vp->vibrato_control_ratio;
	  incr = update_vibrato(song, vp, 0);
	  vibflag = 0;
	}
    }

  vp->vibrato_control_counter=cc;
  vp->sample_increment=incr;
  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

static sample_t *rs_vib_bidir(MidiSong *song, Voice *vp, Sint32 count)
{
  sample_t v1, v2;
  Sint32 
    ofs=vp->sample_offset, 
    incr=vp->sample_increment,
    le=vp->sample->loop_end, 
    ls=vp->sample->loop_start;
  sample_t 
    *dest=song->resample_buffer, 
    *src=vp->sample->data;
  int 
    cc=vp->vibrato_control_counter;
  Sint32
    le2=le<<1,
    ls2=ls<<1,
    i;
  int
    vibflag = 0;

  /* Play normally until inside the loop region */
  while (count && (ofs <= ls)) 
    {
      i = (ls - ofs) / incr + 1;
      if (i > count) i = count;
      if (i > cc) 
	{
	  i = cc;
	  vibflag = 1;
	} 
      else cc -= i;
      count -= i;
      while (i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
      if (vibflag) 
	{
	  cc = vp->vibrato_control_ratio;
	  incr = update_vibrato(song, vp, 0);
	  vibflag = 0;
	}
    }
  
  /* Then do the bidirectional looping */

  while (count) 
    {
      /* Precalc how many times we should go through the loop */
      i = ((incr > 0 ? le : ls) - ofs) / incr + 1;
      if(i > count) i = count;
      if(i > cc) 
	{
	  i = cc;
	  vibflag = 1;
	} 
      else cc -= i;
      count -= i;
      while (i--) 
	{
          v1 = src[ofs >> FRACTION_BITS];
          v2 = src[(ofs >> FRACTION_BITS)+1];
          *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
	  ofs += incr;
	}
      if (vibflag) 
	{
	  cc = vp->vibrato_control_ratio;
	  incr = update_vibrato(song, vp, (incr < 0));
	  vibflag = 0;
	}
      if (ofs >= le) 
	{
	  /* fold the overshoot back in */
	  ofs = le2 - ofs;
	  incr *= -1;
	} 
      else if (ofs <= ls) 
	{
	  ofs = ls2 - ofs;
	  incr *= -1;
	}
    }

  vp->vibrato_control_counter=cc;
  vp->sample_increment=incr;
  vp->sample_offset=ofs; /* Update offset */
  return song->resample_buffer;
}

sample_t *resample_voice(MidiSong *song, int v, Sint32 *countptr)
{
  Sint32 ofs;
  Uint8 modes;
  Voice *vp=&(song->voice[v]);
  
  if (!(vp->sample->sample_rate))
    {
      /* Pre-resampled data -- just update the offset and check if
         we're out of data. */
      ofs=vp->sample_offset >> FRACTION_BITS; /* Kind of silly to use
						 FRACTION_BITS here... */
      if (*countptr >= (vp->sample->data_length>>FRACTION_BITS) - ofs)
	{
	  /* Note finished. Free the voice. */
	  vp->status = VOICE_FREE;
	  
	  /* Let the caller know how much data we had left */
	  *countptr = (vp->sample->data_length>>FRACTION_BITS) - ofs;
	}
      else
	vp->sample_offset += *countptr << FRACTION_BITS;
      
      return vp->sample->data+ofs;
    }

  /* Need to resample. Use the proper function. */
  modes=vp->sample->modes;

  if (vp->vibrato_control_ratio)
    {
      if ((modes & MODES_LOOPING) &&
	  ((modes & MODES_ENVELOPE) ||
	   (vp->status==VOICE_ON || vp->status==VOICE_SUSTAINED)))
	{
	  if (modes & MODES_PINGPONG)
	    return rs_vib_bidir(song, vp, *countptr);
	  else
	    return rs_vib_loop(song, vp, *countptr);
	}
      else
	return rs_vib_plain(song, v, countptr);
    }
  else
    {
      if ((modes & MODES_LOOPING) &&
	  ((modes & MODES_ENVELOPE) ||
	   (vp->status==VOICE_ON || vp->status==VOICE_SUSTAINED)))
	{
	  if (modes & MODES_PINGPONG)
	    return rs_bidir(song, vp, *countptr);
	  else
	    return rs_loop(song, vp, *countptr);
	}
      else
	return rs_plain(song, v, countptr);
    }
}

void pre_resample(MidiSong *song, Sample *sp)
{
  double a, xdiff;
  Sint32 incr, ofs, newlen, count;
  Sint16 *newdata, *dest, *src = (Sint16 *) sp->data;
  Sint16 v1, v2, v3, v4, *vptr;
#ifdef DEBUG_CHATTER
  static const char note_name[12][3] =
  {
    "C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"
  };
#endif

  SNDDBG((" * pre-resampling for note %d (%s%d)\n",
	  sp->note_to_use,
	  note_name[sp->note_to_use % 12], (sp->note_to_use & 0x7F) / 12));

  a = ((double) (sp->sample_rate) * freq_table[(int) (sp->note_to_use)]) /
    ((double) (sp->root_freq) * song->rate);
  newlen = (Sint32)(sp->data_length / a);
  dest = newdata = safe_malloc(newlen >> (FRACTION_BITS - 1));

  count = (newlen >> FRACTION_BITS) - 1;
  ofs = incr = (sp->data_length - (1 << FRACTION_BITS)) / count;

  if (--count)
    *dest++ = src[0];

  /* Since we're pre-processing and this doesn't have to be done in
     real-time, we go ahead and do the full sliding cubic interpolation. */
  while (--count)
    {
      vptr = src + (ofs >> FRACTION_BITS);
          /*
           * Electric Fence to the rescue: Accessing *(vptr - 1) is not a
           * good thing to do when vptr <= src. (TiMidity++ has a similar
           * safe-guard here.)
           */
      v1 = (vptr > src) ? *(vptr - 1) : 0;
      v2 = *vptr;
      v3 = *(vptr + 1);
      v4 = *(vptr + 2);
      xdiff = FSCALENEG(ofs & FRACTION_MASK, FRACTION_BITS);
      *dest++ = (Sint16)(v2 + (xdiff / 6.0) * (-2 * v1 - 3 * v2 + 6 * v3 - v4 +
      xdiff * (3 * (v1 - 2 * v2 + v3) + xdiff * (-v1 + 3 * (v2 - v3) + v4))));
      ofs += incr;
    }

  if (ofs & FRACTION_MASK)
    {
      v1 = src[ofs >> FRACTION_BITS];
      v2 = src[(ofs >> FRACTION_BITS) + 1];
      *dest++ = v1 + (((v2 - v1) * (ofs & FRACTION_MASK)) >> FRACTION_BITS);
    }
  else
    *dest++ = src[ofs >> FRACTION_BITS];

  sp->data_length = newlen;
  sp->loop_start = (Sint32)(sp->loop_start / a);
  sp->loop_end = (Sint32)(sp->loop_end / a);
  free(sp->data);
  sp->data = (sample_t *) newdata;
  sp->sample_rate = 0;
}