annotate mp3lib/decod386.c @ 1108:268c1d28cff2

*** empty log message ***

/*
 * Mpeg Layer-1,2,3 audio decoder
 * ------------------------------
 * copyright (c) 1995,1996,1997 by Michael Hipp, All rights reserved.
 * See also 'README'
 *
 * slighlty optimized for machines without autoincrement/decrement.
 * The performance is highly compiler dependend. Maybe
 * the decode.c version for 'normal' processor may be faster
 * even for Intel processors.
 */



 /* new WRITE_SAMPLE */
   /* sizeof(int) == 4 */
#define WRITE_SAMPLE(samples,sum,clip) { \
  double dtemp; long v;                  \
  dtemp = ((((65536.0 * 65536.0 * 16)+(65536.0 * 0.5))* 65536.0)) + (sum);\
  v = ((*(int *)&dtemp) - 0x80000000); \
  if( v > 32767) { *(samples) = 0x7fff; (clip)++; } \
  else if( v < -32768) { *(samples) = -0x8000; (clip)++; } \
  else { *(samples) = v; } \
}


/*
#define WRITE_SAMPLE(samples,sum,clip) { \
  double dtemp; int v;                    \
  dtemp = ((((65536.0 * 65536.0 * 16)+(65536.0 * 0.5))* 65536.0)) + (sum);\
  v = ((*(int *)&dtemp) - 0x80000000); \
  if( v > 32767) { *(samples) = 0x7fff; (clip)++; } \
  else if( v < -32768) { *(samples) = -0x8000; (clip)++; } \
  else { *(samples) = v; } \
}
*/

static int synth_1to1_mono(real *bandPtr,unsigned char *samples,int *pnt)
{
  short samples_tmp[64];
  short *tmp1 = samples_tmp;
  int i,ret;
  int pnt1 = 0;

  ret = synth_1to1(bandPtr,0,(unsigned char *) samples_tmp,&pnt1);
  samples += *pnt;

  for(i=0;i<32;i++) {
    *( (short *) samples) = *tmp1;
    samples += 2;
    tmp1 += 2;
  }
  *pnt += 64;

  return ret;
}


static int synth_1to1_mono2stereo(real *bandPtr,unsigned char *samples,int *pnt)
{
  int i,ret;

  ret = synth_1to1(bandPtr,0,samples,pnt);
  samples = samples + *pnt - 128;

  for(i=0;i<32;i++) {
    ((short *)samples)[1] = ((short *)samples)[0];
    samples+=4;
  }

  return ret;
}


#ifdef USE_FAKE_MONO
static int synth_1to1_l(real *bandPtr,int channel,unsigned char *out,int *pnt)
{
  int i,ret;

  ret = synth_1to1(bandPtr,channel,out,pnt);
  out = out + *pnt - 128;

  for(i=0;i<32;i++) {
    ((short *)out)[1] = ((short *)out)[0];
    out+=4;
  }

  return ret;
}


static int synth_1to1_r(real *bandPtr,int channel,unsigned char *out,int *pnt)
{
  int i,ret;

  ret = synth_1to1(bandPtr,channel,out,pnt);
  out = out + *pnt - 128;

  for(i=0;i<32;i++) {
    ((short *)out)[0] = ((short *)out)[1];
    out+=4;
  }

  return ret;
}
#endif


static int synth_1to1(real *bandPtr,int channel,unsigned char *out,int *pnt)
{
  static real buffs[2][2][0x110];
  static const int step = 2;
  static int bo = 1;
  short *samples = (short *) (out + *pnt);

  real *b0,(*buf)[0x110];
  int clip = 0;
  int bo1;

  #ifdef HAVE_SSE_MP3
  //if ( _3dnow )
   {
    int ret;
    ret=synth_1to1_sse( bandPtr,channel,out+*pnt );
    *pnt+=128;
    return ret;
   }
  #endif
  #ifdef HAVE_3DNOWEX
  if ( _3dnow > 1 )
   {
    int ret;
    ret=synth_1to1_3dnowex( bandPtr,channel,out+*pnt );
    *pnt+=128;
    return ret;
   }
  #endif
  #ifdef HAVE_3DNOW
  if ( _3dnow )
   {
    int ret;
    ret=synth_1to1_3dnow( bandPtr,channel,out+*pnt );
    *pnt+=128;
    return ret;
   }
  #endif
  if ( _i586 )
   {
     int ret;
     ret=synth_1to1_pent( bandPtr,channel,out+*pnt );
     *pnt+=128;
     return ret;
   }

  if(!channel) {     /* channel=0 */
    bo--;
    bo &= 0xf;
    buf = buffs[0];
  }
  else {
    samples++;
    buf = buffs[1];
  }

  if(bo & 0x1) {
    b0 = buf[0];
    bo1 = bo;
    dct64(buf[1]+((bo+1)&0xf),buf[0]+bo,bandPtr);
  }
  else {
    b0 = buf[1];
    bo1 = bo+1;
    dct64(buf[0]+bo,buf[1]+bo+1,bandPtr);
  }

  {
    register int j;
    real *window = decwin + 16 - bo1;

    for (j=16;j;j--,b0+=0x10,window+=0x20,samples+=step)
    {
      real sum;
      sum  = window[0x0] * b0[0x0];
      sum -= window[0x1] * b0[0x1];
      sum += window[0x2] * b0[0x2];
      sum -= window[0x3] * b0[0x3];
      sum += window[0x4] * b0[0x4];
      sum -= window[0x5] * b0[0x5];
      sum += window[0x6] * b0[0x6];
      sum -= window[0x7] * b0[0x7];
      sum += window[0x8] * b0[0x8];
      sum -= window[0x9] * b0[0x9];
      sum += window[0xA] * b0[0xA];
      sum -= window[0xB] * b0[0xB];
      sum += window[0xC] * b0[0xC];
      sum -= window[0xD] * b0[0xD];
      sum += window[0xE] * b0[0xE];
      sum -= window[0xF] * b0[0xF];

      WRITE_SAMPLE(samples,sum,clip);
    }

    {
      real sum;
      sum  = window[0x0] * b0[0x0];
      sum += window[0x2] * b0[0x2];
      sum += window[0x4] * b0[0x4];
      sum += window[0x6] * b0[0x6];
      sum += window[0x8] * b0[0x8];
      sum += window[0xA] * b0[0xA];
      sum += window[0xC] * b0[0xC];
      sum += window[0xE] * b0[0xE];
      WRITE_SAMPLE(samples,sum,clip);
      b0-=0x10,window-=0x20,samples+=step;
    }
    window += bo1<<1;

    for (j=15;j;j--,b0-=0x10,window-=0x20,samples+=step)
    {
      real sum;
      sum = -window[-0x1] * b0[0x0];
      sum -= window[-0x2] * b0[0x1];
      sum -= window[-0x3] * b0[0x2];
      sum -= window[-0x4] * b0[0x3];
      sum -= window[-0x5] * b0[0x4];
      sum -= window[-0x6] * b0[0x5];
      sum -= window[-0x7] * b0[0x6];
      sum -= window[-0x8] * b0[0x7];
      sum -= window[-0x9] * b0[0x8];
      sum -= window[-0xA] * b0[0x9];
      sum -= window[-0xB] * b0[0xA];
      sum -= window[-0xC] * b0[0xB];
      sum -= window[-0xD] * b0[0xC];
      sum -= window[-0xE] * b0[0xD];
      sum -= window[-0xF] * b0[0xE];
      sum -= window[-0x0] * b0[0xF];

      WRITE_SAMPLE(samples,sum,clip);
    }
  }
  *pnt += 128;

  return clip;

}