Subversion Repositories shark

/*

 * Discrete Cosine Tansform (DCT) for subband synthesis

 *

 * -funroll-loops (for gcc) will remove the loops for better performance

 * using loops in the source-code enhances readabillity

 */

/*

 * TODO: write an optimized version for the down-sampling modes

 *       (in these modes the bands 16-31 (2:1) or 8-31 (4:1) are zero

 */

#include "mpg123.h"

void dct64(real *out0,real *out1,real *samples)

{

  real bufs[64];

 {

  register int i,j;

  register real *b1,*b2,*bs,*costab;

  b1 = samples;

  bs = bufs;

  costab = pnts[0]+16;

  b2 = b1 + 32;

  for(i=15;i>=0;i--)

    *bs++ = (*b1++ + *--b2);

  for(i=15;i>=0;i--)

    *bs++ = (*--b2 - *b1++) * *--costab;

  b1 = bufs;

  costab = pnts[1]+8;

  b2 = b1 + 16;

  {

    for(i=7;i>=0;i--)

      *bs++ = (*b1++ + *--b2);

    for(i=7;i>=0;i--)

      *bs++ = (*--b2 - *b1++) * *--costab;

    b2 += 32;

    costab += 8;

    for(i=7;i>=0;i--)

      *bs++ = (*b1++ + *--b2);

    for(i=7;i>=0;i--)

      *bs++ = (*b1++ - *--b2) * *--costab;

    b2 += 32;

  }

  bs = bufs;

  costab = pnts[2];

  b2 = b1 + 8;

  for(j=2;j;j--)

  {

    for(i=3;i>=0;i--)

      *bs++ = (*b1++ + *--b2);

    for(i=3;i>=0;i--)

      *bs++ = (*--b2 - *b1++) * costab[i];

    b2 += 16;

    for(i=3;i>=0;i--)

      *bs++ = (*b1++ + *--b2);

    for(i=3;i>=0;i--)

      *bs++ = (*b1++ - *--b2) * costab[i];

    b2 += 16;

  }

  b1 = bufs;

  costab = pnts[3];

  b2 = b1 + 4;

  for(j=4;j;j--)

  {

    *bs++ = (*b1++ + *--b2);

    *bs++ = (*b1++ + *--b2);

    *bs++ = (*--b2 - *b1++) * costab[1];

    *bs++ = (*--b2 - *b1++) * costab[0];

    b2 += 8;

    *bs++ = (*b1++ + *--b2);

    *bs++ = (*b1++ + *--b2);

    *bs++ = (*b1++ - *--b2) * costab[1];

    *bs++ = (*b1++ - *--b2) * costab[0];

    b2 += 8;

  }

  bs = bufs;

  costab = pnts[4];

  for(j=8;j;j--)

  {

    real v0,v1;

    v0=*b1++; v1 = *b1++;

    *bs++ = (v0 + v1);

    *bs++ = (v0 - v1) * (*costab);

    v0=*b1++; v1 = *b1++;

    *bs++ = (v0 + v1);

    *bs++ = (v1 - v0) * (*costab);

  }

 }

 {

  register real *b1;

  register int i;

  for(b1=bufs,i=8;i;i--,b1+=4)

    b1[2] += b1[3];

  for(b1=bufs,i=4;i;i--,b1+=8)

  {

    b1[4] += b1[6];

    b1[6] += b1[5];

    b1[5] += b1[7];

  }

  for(b1=bufs,i=2;i;i--,b1+=16)

  {

    b1[8]  += b1[12];

    b1[12] += b1[10];

    b1[10] += b1[14];

    b1[14] += b1[9];

    b1[9]  += b1[13];

    b1[13] += b1[11];

    b1[11] += b1[15];

  }

 }

  out0[0x10*16] = bufs[0];

  out0[0x10*15] = bufs[16+0]  + bufs[16+8];

  out0[0x10*14] = bufs[8];

  out0[0x10*13] = bufs[16+8]  + bufs[16+4];

  out0[0x10*12] = bufs[4];

  out0[0x10*11] = bufs[16+4]  + bufs[16+12];

  out0[0x10*10] = bufs[12];

  out0[0x10* 9] = bufs[16+12] + bufs[16+2];

  out0[0x10* 8] = bufs[2];

  out0[0x10* 7] = bufs[16+2]  + bufs[16+10];

  out0[0x10* 6] = bufs[10];

  out0[0x10* 5] = bufs[16+10] + bufs[16+6];

  out0[0x10* 4] = bufs[6];

  out0[0x10* 3] = bufs[16+6]  + bufs[16+14];

  out0[0x10* 2] = bufs[14];

  out0[0x10* 1] = bufs[16+14] + bufs[16+1];

  out0[0x10* 0] = bufs[1];

  out1[0x10* 0] = bufs[1];

  out1[0x10* 1] = bufs[16+1]  + bufs[16+9];

  out1[0x10* 2] = bufs[9];

  out1[0x10* 3] = bufs[16+9]  + bufs[16+5];

  out1[0x10* 4] = bufs[5];

  out1[0x10* 5] = bufs[16+5]  + bufs[16+13];

  out1[0x10* 6] = bufs[13];

  out1[0x10* 7] = bufs[16+13] + bufs[16+3];

  out1[0x10* 8] = bufs[3];

  out1[0x10* 9] = bufs[16+3]  + bufs[16+11];

  out1[0x10*10] = bufs[11];

  out1[0x10*11] = bufs[16+11] + bufs[16+7];

  out1[0x10*12] = bufs[7];

  out1[0x10*13] = bufs[16+7]  + bufs[16+15];

  out1[0x10*14] = bufs[15];

  out1[0x10*15] = bufs[16+15];

}