extern/ffmpeg/libavcodec/fdctref.c

   1 /**
   2  * @file fdctref.c
   3  * forward discrete cosine transform, double precision.
   4  */
   5
   6 /* Copyright (C) 1996, MPEG Software Simulation Group. All Rights Reserved. */
   7
   8 /*
   9  * Disclaimer of Warranty
  10  *
  11  * These software programs are available to the user without any license fee or
  12  * royalty on an "as is" basis.  The MPEG Software Simulation Group disclaims
  13  * any and all warranties, whether express, implied, or statuary, including any
  14  * implied warranties or merchantability or of fitness for a particular
  15  * purpose.  In no event shall the copyright-holder be liable for any
  16  * incidental, punitive, or consequential damages of any kind whatsoever
  17  * arising from the use of these programs.
  18  *
  19  * This disclaimer of warranty extends to the user of these programs and user's
  20  * customers, employees, agents, transferees, successors, and assigns.
  21  *
  22  * The MPEG Software Simulation Group does not represent or warrant that the
  23  * programs furnished hereunder are free of infringement of any third-party
  24  * patents.
  25  *
  26  * Commercial implementations of MPEG-1 and MPEG-2 video, including shareware,
  27  * are subject to royalty fees to patent holders.  Many of these patents are
  28  * general enough such that they are unavoidable regardless of implementation
  29  * design.
  30  *
  31  */
  32
  33 #include <math.h>
  34
  35 #ifndef PI
  36 # ifdef M_PI
  37 #  define PI M_PI
  38 # else
  39 #  define PI 3.14159265358979323846
  40 # endif
  41 #endif
  42
  43 /* global declarations */
  44 void init_fdct (void);
  45 void fdct (short *block);
  46
  47 /* private data */
  48 static double c[8][8]; /* transform coefficients */
  49
  50 void init_fdct()
  51 {
  52   int i, j;
  53   double s;
  54
  55   for (i=0; i<8; i++)
  56   {
  57     s = (i==0) ? sqrt(0.125) : 0.5;
  58
  59     for (j=0; j<8; j++)
  60       c[i][j] = s * cos((PI/8.0)*i*(j+0.5));
  61   }
  62 }
  63
  64 void fdct(block)
  65 short *block;
  66 {
  67         register int i, j;
  68         double s;
  69         double tmp[64];
  70
  71         for(i = 0; i < 8; i++)
  72             for(j = 0; j < 8; j++)
  73             {
  74                     s = 0.0;
  75
  76 /*
  77  *                     for(k = 0; k < 8; k++)
  78  *                         s += c[j][k] * block[8 * i + k];
  79  */
  80                 s += c[j][0] * block[8 * i + 0];
  81                 s += c[j][1] * block[8 * i + 1];
  82                 s += c[j][2] * block[8 * i + 2];
  83                 s += c[j][3] * block[8 * i + 3];
  84                 s += c[j][4] * block[8 * i + 4];
  85                 s += c[j][5] * block[8 * i + 5];
  86                 s += c[j][6] * block[8 * i + 6];
  87                 s += c[j][7] * block[8 * i + 7];
  88
  89                     tmp[8 * i + j] = s;
  90             }
  91
  92         for(j = 0; j < 8; j++)
  93             for(i = 0; i < 8; i++)
  94             {
  95                     s = 0.0;
  96
  97 /*
  98  *                       for(k = 0; k < 8; k++)
  99  *                    s += c[i][k] * tmp[8 * k + j];
 100  */
 101                 s += c[i][0] * tmp[8 * 0 + j];
 102                 s += c[i][1] * tmp[8 * 1 + j];
 103                 s += c[i][2] * tmp[8 * 2 + j];
 104                 s += c[i][3] * tmp[8 * 3 + j];
 105                 s += c[i][4] * tmp[8 * 4 + j];
 106                 s += c[i][5] * tmp[8 * 5 + j];
 107                 s += c[i][6] * tmp[8 * 6 + j];
 108                 s += c[i][7] * tmp[8 * 7 + j];
 109                 s*=8.0;
 110
 111                     block[8 * i + j] = (short)floor(s + 0.499999);
 112 /*
 113  * reason for adding 0.499999 instead of 0.5:
 114  * s is quite often x.5 (at least for i and/or j = 0 or 4)
 115  * and setting the rounding threshold exactly to 0.5 leads to an
 116  * extremely high arithmetic implementation dependency of the result;
 117  * s being between x.5 and x.500001 (which is now incorrectly rounded
 118  * downwards instead of upwards) is assumed to occur less often
 119  * (if at all)
 120  */
 121       }
 122 }
 123
 124 /* perform IDCT matrix multiply for 8x8 coefficient block */
 125
 126 void idct(block)
 127 short *block;
 128 {
 129   int i, j, k, v;
 130   double partial_product;
 131   double tmp[64];
 132
 133   for (i=0; i<8; i++)
 134     for (j=0; j<8; j++)
 135     {
 136       partial_product = 0.0;
 137
 138       for (k=0; k<8; k++)
 139         partial_product+= c[k][j]*block[8*i+k];
 140
 141       tmp[8*i+j] = partial_product;
 142     }
 143
 144   /* Transpose operation is integrated into address mapping by switching
 145      loop order of i and j */
 146
 147   for (j=0; j<8; j++)
 148     for (i=0; i<8; i++)
 149     {
 150       partial_product = 0.0;
 151
 152       for (k=0; k<8; k++)
 153         partial_product+= c[k][i]*tmp[8*k+j];
 154
 155       v = (int) floor(partial_product+0.5);
 156       block[8*i+j] = v;
 157     }
 158 }