idct1.c

xvid解码的精简版本.非常好的版本,节省了分离xvid源代码的过程

#include "portab.h"

#define IdctAdjustBeforeShift 2
#define xC1S7 64277
#define xC2S6 60547
#define xC3S5 54491
#define xC4S4 46341
#define xC5S3 36410
#define xC6S2 25080
#define xC7S1 12785


void IDctSlow(  int16_t  * IntermediateData,
                int16_t  * InputData) {
  //int32_t IntermediateData[64];
  int16_t * ip = IntermediateData;
  int16_t * op = InputData;

  int32_t _A, _B, _C, _D, _Ad, _Bd, _Cd, _Dd, _E, _F, _G, _H;
  int32_t _Ed, _Gd, _Add, _Bdd, _Fd, _Hd;
  int32_t t1, t2;

  int loop;

  //dequant_slow( QuantMatrix, InputData, IntermediateData);

  /* Inverse DCT on the rows now */
  for ( loop = 0; loop < 8; loop++){
    /* Check for non-zero values */
    if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
      t1 = (xC1S7 * ip[1]);
      t2 = (xC7S1 * ip[7]);
      t1 >>= 16;
      t2 >>= 16;
      _A = t1 + t2;

      t1 = (xC7S1 * ip[1]);
      t2 = (xC1S7 * ip[7]);
      t1 >>= 16;
      t2 >>= 16;
      _B = t1 - t2;

      t1 = (xC3S5 * ip[3]);
      t2 = (xC5S3 * ip[5]);
      t1 >>= 16;
      t2 >>= 16;
      _C = t1 + t2;

      t1 = (xC3S5 * ip[5]);
      t2 = (xC5S3 * ip[3]);
      t1 >>= 16;
      t2 >>= 16;
      _D = t1 - t2;

      t1 = (xC4S4 * (_A - _C));
      t1 >>= 16;
      _Ad = t1;

      t1 = (xC4S4 * (_B - _D));
      t1 >>= 16;
      _Bd = t1;


      _Cd = _A + _C;
      _Dd = _B + _D;

      t1 = (xC4S4 * (ip[0] + ip[4]));
      t1 >>= 16;
      _E = t1;

      t1 = (xC4S4 * (ip[0] - ip[4]));
      t1 >>= 16;
      _F = t1;

      t1 = (xC2S6 * ip[2]);
      t2 = (xC6S2 * ip[6]);
      t1 >>= 16;
      t2 >>= 16;
      _G = t1 + t2;

      t1 = (xC6S2 * ip[2]);
      t2 = (xC2S6 * ip[6]);
      t1 >>= 16;
      t2 >>= 16;
      _H = t1 - t2;


      _Ed = _E - _G;
      _Gd = _E + _G;

      _Add = _F + _Ad;
      _Bdd = _Bd - _H;

      _Fd = _F - _Ad;
      _Hd = _Bd + _H;

      /* Final sequence of operations over-write original inputs. */
      ip[0] = (int16_t)((_Gd + _Cd )   >> 0);
      ip[7] = (int16_t)((_Gd - _Cd )   >> 0);

      ip[1] = (int16_t)((_Add + _Hd )  >> 0);
      ip[2] = (int16_t)((_Add - _Hd )  >> 0);

      ip[3] = (int16_t)((_Ed + _Dd )   >> 0);
      ip[4] = (int16_t)((_Ed - _Dd )   >> 0);

      ip[5] = (int16_t)((_Fd + _Bdd )  >> 0);
      ip[6] = (int16_t)((_Fd - _Bdd )  >> 0);

    }

    ip += 8;                    /* next row */
  }

  ip = IntermediateData;

  for ( loop = 0; loop < 8; loop++){
    /* Check for non-zero values (bitwise or faster than ||) */
    if ( ip[0 * 8] | ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
         ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {

      t1 = (xC1S7 * ip[1*8]);
      t2 = (xC7S1 * ip[7*8]);
      t1 >>= 16;
      t2 >>= 16;
      _A = t1 + t2;

      t1 = (xC7S1 * ip[1*8]);
      t2 = (xC1S7 * ip[7*8]);
      t1 >>= 16;
      t2 >>= 16;
      _B = t1 - t2;

      t1 = (xC3S5 * ip[3*8]);
      t2 = (xC5S3 * ip[5*8]);
      t1 >>= 16;
      t2 >>= 16;
      _C = t1 + t2;

      t1 = (xC3S5 * ip[5*8]);
      t2 = (xC5S3 * ip[3*8]);
      t1 >>= 16;
      t2 >>= 16;
      _D = t1 - t2;

      t1 = (xC4S4 * (_A - _C));
      t1 >>= 16;
      _Ad = t1;

      t1 = (xC4S4 * (_B - _D));
      t1 >>= 16;
      _Bd = t1;


      _Cd = _A + _C;
      _Dd = _B + _D;

      t1 = (xC4S4 * (ip[0*8] + ip[4*8]));
      t1 >>= 16;
      _E = t1;

      t1 = (xC4S4 * (ip[0*8] - ip[4*8]));
      t1 >>= 16;
      _F = t1;

      t1 = (xC2S6 * ip[2*8]);
      t2 = (xC6S2 * ip[6*8]);
      t1 >>= 16;
      t2 >>= 16;
      _G = t1 + t2;

      t1 = (xC6S2 * ip[2*8]);
      t2 = (xC2S6 * ip[6*8]);
      t1 >>= 16;
      t2 >>= 16;
      _H = t1 - t2;

      _Ed = _E - _G;
      _Gd = _E + _G;

      _Add = _F + _Ad;
      _Bdd = _Bd - _H;

      _Fd = _F - _Ad;
      _Hd = _Bd + _H;

      _Gd += IdctAdjustBeforeShift;
      _Add += IdctAdjustBeforeShift;
      _Ed += IdctAdjustBeforeShift;
      _Fd += IdctAdjustBeforeShift;

      /* Final sequence of operations over-write original inputs. */
      op[0*8] = (int16_t)((_Gd + _Cd )   >> 2);
      op[7*8] = (int16_t)((_Gd - _Cd )   >> 2);

      op[1*8] = (int16_t)((_Add + _Hd )  >> 2);
      op[2*8] = (int16_t)((_Add - _Hd )  >> 2);

      op[3*8] = (int16_t)((_Ed + _Dd )   >> 2);
      op[4*8] = (int16_t)((_Ed - _Dd )   >> 2);

      op[5*8] = (int16_t)((_Fd + _Bdd )  >> 2);
      op[6*8] = (int16_t)((_Fd - _Bdd )  >> 2);
    }else{
      op[0*8] = 0;
      op[7*8] = 0;
      op[1*8] = 0;
      op[2*8] = 0;
      op[3*8] = 0;
      op[4*8] = 0;
      op[5*8] = 0;
      op[6*8] = 0;
    }

    ip++;                       /* next column */
    op++;
  }
}