git.sesse.net Git - x264/blob - common/csp.c

   1 /*****************************************************************************
   2  * csp.c: h264 encoder library
   3  *****************************************************************************
   4  * Copyright (C) 2004 Laurent Aimar
   5  * $Id: csp.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $
   6  *
   7  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
  22  *****************************************************************************/
  23
  24 #include <stdio.h>
  25 #include <string.h>
  26
  27 #include "common.h"
  28
  29 static inline void plane_copy( uint8_t *dst, int i_dst,
  30                                uint8_t *src, int i_src, int w, int h)
  31 {
  32     for( ; h > 0; h-- )
  33     {
  34         memcpy( dst, src, w );
  35         dst += i_dst;
  36         src += i_src;
  37     }
  38 }
  39 static inline void plane_copy_vflip( uint8_t *dst, int i_dst,
  40                                      uint8_t *src, int i_src, int w, int h)
  41 {
  42     plane_copy( dst, i_dst, src + (h -1)*i_src, -i_src, w, h );
  43 }
  44
  45 static inline void plane_subsamplev2( uint8_t *dst, int i_dst,
  46                                       uint8_t *src, int i_src, int w, int h)
  47 {
  48     for( ; h > 0; h-- )
  49     {
  50         uint8_t *d = dst;
  51         uint8_t *s = src;
  52         int     i;
  53         for( i = 0; i < w; i++ )
  54         {
  55             *d++ = ( s[0] + s[i_src] + 1 ) >> 1;
  56             s++;
  57         }
  58         dst += i_dst;
  59         src += 2 * i_src;
  60     }
  61 }
  62
  63 static inline void plane_subsamplev2_vlip( uint8_t *dst, int i_dst,
  64                                            uint8_t *src, int i_src, int w, int h)
  65 {
  66     plane_subsamplev2( dst, i_dst, src + (2*h-1)*i_src, -i_src, w, h );
  67 }
  68
  69 static inline void plane_subsamplehv2( uint8_t *dst, int i_dst,
  70                                        uint8_t *src, int i_src, int w, int h)
  71 {
  72     for( ; h > 0; h-- )
  73     {
  74         uint8_t *d = dst;
  75         uint8_t *s = src;
  76         int     i;
  77         for( i = 0; i < w; i++ )
  78         {
  79             *d++ = ( s[0] + s[1] + s[i_src] + s[i_src+1] + 1 ) >> 2;
  80             s += 2;
  81         }
  82         dst += i_dst;
  83         src += 2 * i_src;
  84     }
  85 }
  86
  87 static inline void plane_subsamplehv2_vlip( uint8_t *dst, int i_dst,
  88                                             uint8_t *src, int i_src, int w, int h)
  89 {
  90     plane_subsamplehv2( dst, i_dst, src + (2*h-1)*i_src, -i_src, w, h );
  91 }
  92
  93 static void i420_to_i420( x264_frame_t *frm, x264_image_t *img,
  94                           int i_width, int i_height )
  95 {
  96     if( img->i_csp & X264_CSP_VFLIP )
  97     {
  98         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
  99                           img->plane[0], img->i_stride[0],
 100                           i_width, i_height );
 101         plane_copy_vflip( frm->plane[1], frm->i_stride[1],
 102                           img->plane[1], img->i_stride[1],
 103                           i_width / 2, i_height / 2 );
 104         plane_copy_vflip( frm->plane[2], frm->i_stride[2],
 105                           img->plane[2], img->i_stride[2],
 106                           i_width / 2, i_height / 2 );
 107     }
 108     else
 109     {
 110         plane_copy( frm->plane[0], frm->i_stride[0],
 111                     img->plane[0], img->i_stride[0],
 112                     i_width, i_height );
 113         plane_copy( frm->plane[1], frm->i_stride[1],
 114                     img->plane[1], img->i_stride[1],
 115                     i_width / 2, i_height / 2 );
 116         plane_copy( frm->plane[2], frm->i_stride[2],
 117                     img->plane[2], img->i_stride[2],
 118                     i_width / 2, i_height / 2 );
 119     }
 120 }
 121
 122 static void yv12_to_i420( x264_frame_t *frm, x264_image_t *img,
 123                           int i_width, int i_height )
 124 {
 125     if( img->i_csp & X264_CSP_VFLIP )
 126     {
 127         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 128                           img->plane[0], img->i_stride[0],
 129                           i_width, i_height );
 130         plane_copy_vflip( frm->plane[2], frm->i_stride[2],
 131                           img->plane[1], img->i_stride[1],
 132                           i_width / 2, i_height / 2 );
 133         plane_copy_vflip( frm->plane[1], frm->i_stride[1],
 134                           img->plane[2], img->i_stride[2],
 135                           i_width / 2, i_height / 2 );
 136     }
 137     else
 138     {
 139         plane_copy( frm->plane[0], frm->i_stride[0],
 140                     img->plane[0], img->i_stride[0],
 141                     i_width, i_height );
 142         plane_copy( frm->plane[2], frm->i_stride[2],
 143                     img->plane[1], img->i_stride[1],
 144                     i_width / 2, i_height / 2 );
 145         plane_copy( frm->plane[1], frm->i_stride[1],
 146                     img->plane[2], img->i_stride[2],
 147                     i_width / 2, i_height / 2 );
 148     }
 149 }
 150
 151 static void i422_to_i420( x264_frame_t *frm, x264_image_t *img,
 152                           int i_width, int i_height )
 153 {
 154     if( img->i_csp & X264_CSP_VFLIP )
 155     {
 156         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 157                           img->plane[0], img->i_stride[0],
 158                           i_width, i_height );
 159
 160         plane_subsamplev2_vlip( frm->plane[1], frm->i_stride[1],
 161                                 img->plane[1], img->i_stride[1],
 162                                 i_width / 2, i_height / 2 );
 163         plane_subsamplev2_vlip( frm->plane[2], frm->i_stride[2],
 164                                 img->plane[2], img->i_stride[2],
 165                                 i_width / 2, i_height / 2 );
 166     }
 167     else
 168     {
 169         plane_copy( frm->plane[0], frm->i_stride[0],
 170                     img->plane[0], img->i_stride[0],
 171                     i_width, i_height );
 172
 173         plane_subsamplev2( frm->plane[1], frm->i_stride[1],
 174                            img->plane[1], img->i_stride[1],
 175                            i_width / 2, i_height / 2 );
 176         plane_subsamplev2( frm->plane[2], frm->i_stride[2],
 177                            img->plane[2], img->i_stride[2],
 178                            i_width / 2, i_height / 2 );
 179     }
 180 }
 181
 182 static void i444_to_i420( x264_frame_t *frm, x264_image_t *img,
 183                           int i_width, int i_height )
 184 {
 185     if( img->i_csp & X264_CSP_VFLIP )
 186     {
 187         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 188                           img->plane[0], img->i_stride[0],
 189                           i_width, i_height );
 190
 191         plane_subsamplehv2_vlip( frm->plane[1], frm->i_stride[1],
 192                                  img->plane[1], img->i_stride[1],
 193                                  i_width / 2, i_height / 2 );
 194         plane_subsamplehv2_vlip( frm->plane[2], frm->i_stride[2],
 195                                  img->plane[2], img->i_stride[2],
 196                                  i_width / 2, i_height / 2 );
 197     }
 198     else
 199     {
 200         plane_copy( frm->plane[0], frm->i_stride[0],
 201                     img->plane[0], img->i_stride[0],
 202                     i_width, i_height );
 203
 204         plane_subsamplehv2( frm->plane[1], frm->i_stride[1],
 205                             img->plane[1], img->i_stride[1],
 206                             i_width / 2, i_height / 2 );
 207         plane_subsamplehv2( frm->plane[2], frm->i_stride[2],
 208                             img->plane[2], img->i_stride[2],
 209                             i_width / 2, i_height / 2 );
 210     }
 211 }
 212 static void yuyv_to_i420( x264_frame_t *frm, x264_image_t *img,
 213                           int i_width, int i_height )
 214 {
 215     uint8_t *src = img->plane[0];
 216     int     i_src= img->i_stride[0];
 217
 218     uint8_t *y   = frm->plane[0];
 219     uint8_t *u   = frm->plane[1];
 220     uint8_t *v   = frm->plane[2];
 221
 222     if( img->i_csp & X264_CSP_VFLIP )
 223     {
 224         src += ( i_height - 1 ) * i_src;
 225         i_src = -i_src;
 226     }
 227
 228     for( ; i_height > 0; i_height -= 2 )
 229     {
 230         uint8_t *ss = src;
 231         uint8_t *yy = y;
 232         uint8_t *uu = u;
 233         uint8_t *vv = v;
 234         int w;
 235
 236         for( w = i_width; w > 0; w -= 2 )
 237         {
 238             *yy++ = ss[0];
 239             *yy++ = ss[2];
 240
 241             *uu++ = ( ss[1] + ss[1+i_src] + 1 ) >> 1;
 242             *vv++ = ( ss[3] + ss[3+i_src] + 1 ) >> 1;
 243
 244             ss += 4;
 245         }
 246         src += i_src;
 247         y += frm->i_stride[0];
 248         u += frm->i_stride[1];
 249         v += frm->i_stride[2];
 250
 251         ss = src;
 252         yy = y;
 253         for( w = i_width; w > 0; w -= 2 )
 254         {
 255             *yy++ = ss[0];
 256             *yy++ = ss[2];
 257             ss += 4;
 258         }
 259         src += i_src;
 260         y += frm->i_stride[0];
 261     }
 262 }
 263
 264 /* Same value than in XviD */
 265 #define BITS 8
 266 #define FIX(f) ((int)((f) * (1 << BITS) + 0.5))
 267
 268 #define Y_R   FIX(0.257)
 269 #define Y_G   FIX(0.504)
 270 #define Y_B   FIX(0.098)
 271 #define Y_ADD 16
 272
 273 #define U_R   FIX(0.148)
 274 #define U_G   FIX(0.291)
 275 #define U_B   FIX(0.439)
 276 #define U_ADD 128
 277
 278 #define V_R   FIX(0.439)
 279 #define V_G   FIX(0.368)
 280 #define V_B   FIX(0.071)
 281 #define V_ADD 128
 282 #define RGB_TO_I420( name, POS_R, POS_G, POS_B, S_RGB ) \
 283 static void name( x264_frame_t *frm, x264_image_t *img, \
 284                   int i_width, int i_height )           \
 285 {                                                       \
 286     uint8_t *src = img->plane[0];                       \
 287     int     i_src= img->i_stride[0];                    \
 288     int     i_y  = frm->i_stride[0];                    \
 289     uint8_t *y   = frm->plane[0];                       \
 290     uint8_t *u   = frm->plane[1];                       \
 291     uint8_t *v   = frm->plane[2];                       \
 292                                                         \
 293     if( img->i_csp & X264_CSP_VFLIP )                   \
 294     {                                                   \
 295         src += ( i_height - 1 ) * i_src;                \
 296         i_src = -i_src;                                 \
 297     }                                                   \
 298                                                         \
 299     for(  ; i_height > 0; i_height -= 2 )               \
 300     {                                                   \
 301         uint8_t *ss = src;                              \
 302         uint8_t *yy = y;                                \
 303         uint8_t *uu = u;                                \
 304         uint8_t *vv = v;                                \
 305         int w;                                          \
 306                                                         \
 307         for( w = i_width; w > 0; w -= 2 )               \
 308         {                                               \
 309             int cr = 0,cg = 0,cb = 0;                   \
 310             int r, g, b;                                \
 311                                                         \
 312             /* Luma */                                  \
 313             cr = r = ss[POS_R];                         \
 314             cg = g = ss[POS_G];                         \
 315             cb = b = ss[POS_B];                         \
 316                                                         \
 317             yy[0] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);    \
 318                                                         \
 319             cr+= r = ss[POS_R+i_src];                   \
 320             cg+= g = ss[POS_G+i_src];                   \
 321             cb+= b = ss[POS_B+i_src];                   \
 322             yy[i_y] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);  \
 323             yy++;                                       \
 324             ss += S_RGB;                                \
 325                                                         \
 326             cr+= r = ss[POS_R];                         \
 327             cg+= g = ss[POS_G];                         \
 328             cb+= b = ss[POS_B];                         \
 329                                                         \
 330             yy[0] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);    \
 331                                                         \
 332             cr+= r = ss[POS_R+i_src];                   \
 333             cg+= g = ss[POS_G+i_src];                   \
 334             cb+= b = ss[POS_B+i_src];                   \
 335             yy[i_y] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);  \
 336             yy++;                                       \
 337             ss += S_RGB;                                \
 338                                                         \
 339             /* Chroma */                                \
 340             *uu++ = (uint8_t)(U_ADD + ((-U_R * cr - U_G * cg + U_B * cb) >> (BITS+2)) ); \
 341             *vv++ = (uint8_t)(V_ADD + (( V_R * cr - V_G * cg - V_B * cb) >> (BITS+2)) ); \
 342         }                                               \
 343                                                         \
 344         src += 2*i_src;                                   \
 345         y += 2*frm->i_stride[0];                        \
 346         u += frm->i_stride[1];                          \
 347         v += frm->i_stride[2];                          \
 348     }                                                   \
 349 }
 350
 351 RGB_TO_I420( rgb_to_i420,  0, 1, 2, 3 );
 352 RGB_TO_I420( bgr_to_i420,  2, 1, 0, 3 );
 353 RGB_TO_I420( bgra_to_i420, 2, 1, 0, 4 );
 354
 355 void x264_csp_init( int cpu, int i_csp, x264_csp_function_t *pf )
 356 {
 357     switch( i_csp )
 358     {
 359         case X264_CSP_I420:
 360             pf->i420 = i420_to_i420;
 361             pf->i422 = i422_to_i420;
 362             pf->i444 = i444_to_i420;
 363             pf->yv12 = yv12_to_i420;
 364             pf->yuyv = yuyv_to_i420;
 365             pf->rgb  = rgb_to_i420;
 366             pf->bgr  = bgr_to_i420;
 367             pf->bgra = bgra_to_i420;
 368             break;
 369
 370         default:
 371             /* For now, can't happen */
 372             fprintf( stderr, "arg in x264_csp_init\n" );
 373             exit( -1 );
 374             break;
 375     }
 376 }
 377