libvo/osd_template.c

   1 /*
   2  * generic alpha renderers for all YUV modes and RGB depths
   3  * Optimized by Nick and Michael.
   4  *
   5  * This file is part of MPlayer.
   6  *
   7  * MPlayer is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * MPlayer is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with MPlayer; if not, write to the Free Software Foundation, Inc.,
  19  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  20  */
  21
  22 #undef PREFETCH
  23 #undef EMMS
  24 #undef PREFETCHW
  25 #undef PAVGB
  26
  27 #if HAVE_AMD3DNOW
  28 #define PREFETCH  "prefetch"
  29 #define PREFETCHW "prefetchw"
  30 #define PAVGB     "pavgusb"
  31 #elif HAVE_MMX2
  32 #define PREFETCH "prefetchnta"
  33 #define PREFETCHW "prefetcht0"
  34 #define PAVGB     "pavgb"
  35 #else
  36 #define PREFETCH " # nop"
  37 #define PREFETCHW " # nop"
  38 #endif
  39
  40 #if HAVE_AMD3DNOW
  41 /* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */
  42 #define EMMS     "femms"
  43 #else
  44 #define EMMS     "emms"
  45 #endif
  46
  47 static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
  48     int y;
  49 #if defined(FAST_OSD) && !HAVE_MMX
  50     w=w>>1;
  51 #endif
  52 #if HAVE_MMX
  53     __asm__ volatile(
  54         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
  55         "movq %%mm5, %%mm4\n\t"
  56         "movq %%mm5, %%mm7\n\t"
  57         "psllw $8, %%mm5\n\t" //FF00FF00FF00
  58         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
  59         ::);
  60 #endif
  61     for(y=0;y<h;y++){
  62         register int x;
  63 #if HAVE_MMX
  64     __asm__ volatile(
  65         PREFETCHW" %0\n\t"
  66         PREFETCH" %1\n\t"
  67         PREFETCH" %2\n\t"
  68         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
  69     for(x=0;x<w;x+=8){
  70         __asm__ volatile(
  71                 "movl %1, %%eax\n\t"
  72                 "orl 4%1, %%eax\n\t"
  73                 " jz 1f\n\t"
  74                 PREFETCHW" 32%0\n\t"
  75                 PREFETCH" 32%1\n\t"
  76                 PREFETCH" 32%2\n\t"
  77                 "movq   %0, %%mm0\n\t" // dstbase
  78                 "movq   %%mm0, %%mm1\n\t"
  79                 "pand %%mm4, %%mm0\n\t"         //0Y0Y0Y0Y
  80                 "psrlw $8, %%mm1\n\t"           //0Y0Y0Y0Y
  81                 "movq   %1, %%mm2\n\t"          //srca HGFEDCBA
  82                 "paddb  %%mm7, %%mm2\n\t"
  83                 "movq %%mm2, %%mm3\n\t"
  84                 "pand %%mm4, %%mm2\n\t"         //0G0E0C0A
  85                 "psrlw $8, %%mm3\n\t"           //0H0F0D0B
  86                 "pmullw %%mm2, %%mm0\n\t"
  87                 "pmullw %%mm3, %%mm1\n\t"
  88                 "psrlw  $8, %%mm0\n\t"
  89                 "pand %%mm5, %%mm1\n\t"
  90                 "por %%mm1, %%mm0\n\t"
  91                 "paddb  %2, %%mm0\n\t"
  92                 "movq   %%mm0, %0\n\t"
  93                 "1:\n\t"
  94                 :: "m" (dstbase[x]), "m" (srca[x]), "m" (src[x])
  95                 : "%eax");
  96         }
  97 #else
  98         for(x=0;x<w;x++){
  99 #ifdef FAST_OSD
 100             if(srca[2*x+0]) dstbase[2*x+0]=src[2*x+0];
 101             if(srca[2*x+1]) dstbase[2*x+1]=src[2*x+1];
 102 #else
 103             if(srca[x]) dstbase[x]=((dstbase[x]*srca[x])>>8)+src[x];
 104 #endif
 105         }
 106 #endif
 107         src+=srcstride;
 108         srca+=srcstride;
 109         dstbase+=dststride;
 110     }
 111 #if HAVE_MMX
 112         __asm__ volatile(EMMS:::"memory");
 113 #endif
 114     return;
 115 }
 116
 117 static inline void RENAME(vo_draw_alpha_yuy2)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
 118     int y;
 119 #if defined(FAST_OSD) && !HAVE_MMX
 120     w=w>>1;
 121 #endif
 122 #if HAVE_MMX
 123     __asm__ volatile(
 124         "pxor %%mm7, %%mm7\n\t"
 125         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
 126         "movq %%mm5, %%mm6\n\t"
 127         "movq %%mm5, %%mm4\n\t"
 128         "psllw $8, %%mm5\n\t" //FF00FF00FF00
 129         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
 130         ::);
 131 #endif
 132     for(y=0;y<h;y++){
 133         register int x;
 134 #if HAVE_MMX
 135     __asm__ volatile(
 136         PREFETCHW" %0\n\t"
 137         PREFETCH" %1\n\t"
 138         PREFETCH" %2\n\t"
 139         ::"m"(*dstbase),"m"(*srca),"m"(*src));
 140     for(x=0;x<w;x+=4){
 141         __asm__ volatile(
 142                 "movl %1, %%eax\n\t"
 143                 "orl %%eax, %%eax\n\t"
 144                 " jz 1f\n\t"
 145                 PREFETCHW" 32%0\n\t"
 146                 PREFETCH" 32%1\n\t"
 147                 PREFETCH" 32%2\n\t"
 148                 "movq   %0, %%mm0\n\t" // dstbase
 149                 "movq   %%mm0, %%mm1\n\t"
 150                 "pand %%mm4, %%mm0\n\t"         //0Y0Y0Y0Y
 151                 "movd   %%eax, %%mm2\n\t"       //srca 0000DCBA
 152                 "paddb  %%mm6, %%mm2\n\t"
 153                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0D0C0B0A
 154                 "pmullw %%mm2, %%mm0\n\t"
 155                 "psrlw  $8, %%mm0\n\t"
 156                 "pand %%mm5, %%mm1\n\t"         //U0V0U0V0
 157                 "movd %2, %%mm2\n\t"            //src 0000DCBA
 158                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0D0C0B0A
 159                 "por %%mm1, %%mm0\n\t"
 160                 "paddb  %%mm2, %%mm0\n\t"
 161                 "movq   %%mm0, %0\n\t"
 162                 "1:\n\t"
 163                 :: "m" (dstbase[x*2]), "m" (srca[x]), "m" (src[x])
 164                 : "%eax");
 165         }
 166 #else
 167         for(x=0;x<w;x++){
 168 #ifdef FAST_OSD
 169             if(srca[2*x+0]) dstbase[4*x+0]=src[2*x+0];
 170             if(srca[2*x+1]) dstbase[4*x+2]=src[2*x+1];
 171 #else
 172             if(srca[x]) {
 173                dstbase[2*x]=((dstbase[2*x]*srca[x])>>8)+src[x];
 174                dstbase[2*x+1]=((((signed)dstbase[2*x+1]-128)*srca[x])>>8)+128;
 175            }
 176 #endif
 177         }
 178 #endif
 179         src+=srcstride;
 180         srca+=srcstride;
 181         dstbase+=dststride;
 182     }
 183 #if HAVE_MMX
 184         __asm__ volatile(EMMS:::"memory");
 185 #endif
 186     return;
 187 }
 188
 189 static inline void RENAME(vo_draw_alpha_uyvy)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
 190   int y;
 191 #if defined(FAST_OSD)
 192   w=w>>1;
 193 #endif
 194   for(y=0;y<h;y++){
 195     register int x;
 196     for(x=0;x<w;x++){
 197 #ifdef FAST_OSD
 198       if(srca[2*x+0]) dstbase[4*x+2]=src[2*x+0];
 199       if(srca[2*x+1]) dstbase[4*x+0]=src[2*x+1];
 200 #else
 201       if(srca[x]) {
 202         dstbase[2*x+1]=((dstbase[2*x+1]*srca[x])>>8)+src[x];
 203         dstbase[2*x]=((((signed)dstbase[2*x]-128)*srca[x])>>8)+128;
 204       }
 205 #endif
 206     }
 207     src+=srcstride;
 208     srca+=srcstride;
 209     dstbase+=dststride;
 210   }
 211 }
 212
 213 static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
 214     int y;
 215 #if HAVE_MMX
 216     __asm__ volatile(
 217         "pxor %%mm7, %%mm7\n\t"
 218         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 219         ::);
 220 #endif
 221     for(y=0;y<h;y++){
 222         register unsigned char *dst = dstbase;
 223         register int x;
 224 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 225 #if HAVE_MMX
 226     __asm__ volatile(
 227         PREFETCHW" %0\n\t"
 228         PREFETCH" %1\n\t"
 229         PREFETCH" %2\n\t"
 230         ::"m"(*dst),"m"(*srca),"m"(*src):"memory");
 231     for(x=0;x<w;x+=2){
 232      if(srca[x] || srca[x+1])
 233         __asm__ volatile(
 234                 PREFETCHW" 32%0\n\t"
 235                 PREFETCH" 32%1\n\t"
 236                 PREFETCH" 32%2\n\t"
 237                 "movq   %0, %%mm0\n\t" // dstbase
 238                 "movq   %%mm0, %%mm1\n\t"
 239                 "movq   %%mm0, %%mm5\n\t"
 240                 "punpcklbw %%mm7, %%mm0\n\t"
 241                 "punpckhbw %%mm7, %%mm1\n\t"
 242                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 243                 "paddb  %%mm6, %%mm2\n\t"
 244                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 245                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 246                 "psrlq  $8, %%mm2\n\t" // srca AAABBBB0
 247                 "movq   %%mm2, %%mm3\n\t"
 248                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0B
 249                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B00
 250                 "pmullw %%mm2, %%mm0\n\t"
 251                 "pmullw %%mm3, %%mm1\n\t"
 252                 "psrlw  $8, %%mm0\n\t"
 253                 "psrlw  $8, %%mm1\n\t"
 254                 "packuswb %%mm1, %%mm0\n\t"
 255                 "movd %2, %%mm2 \n\t" // src ABCD0000
 256                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 257                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 258                 "psrlq  $8, %%mm2\n\t" // src AAABBBB0
 259                 "paddb  %%mm2, %%mm0\n\t"
 260                 "pand   %4, %%mm5\n\t"
 261                 "pand   %3, %%mm0\n\t"
 262                 "por    %%mm0, %%mm5\n\t"
 263                 "movq   %%mm5, %0\n\t"
 264                 :: "m" (dst[0]), "m" (srca[x]), "m" (src[x]), "m"(mask24hl), "m"(mask24lh));
 265                 dst += 6;
 266         }
 267 #else /* HAVE_MMX */
 268     for(x=0;x<w;x++){
 269         if(srca[x]){
 270             __asm__ volatile(
 271                 "movzbl (%0), %%ecx\n\t"
 272                 "movzbl 1(%0), %%eax\n\t"
 273
 274                 "imull %1, %%ecx\n\t"
 275                 "imull %1, %%eax\n\t"
 276
 277                 "addl %2, %%ecx\n\t"
 278                 "addl %2, %%eax\n\t"
 279
 280                 "movb %%ch, (%0)\n\t"
 281                 "movb %%ah, 1(%0)\n\t"
 282
 283                 "movzbl 2(%0), %%eax\n\t"
 284                 "imull %1, %%eax\n\t"
 285                 "addl %2, %%eax\n\t"
 286                 "movb %%ah, 2(%0)\n\t"
 287                 :
 288                 :"D" (dst),
 289                  "r" ((unsigned)srca[x]),
 290                  "r" (((unsigned)src[x])<<8)
 291                 :"%eax", "%ecx"
 292                 );
 293             }
 294             dst += 3;
 295         }
 296 #endif /* !HAVE_MMX */
 297 #else /*non x86 arch or x86_64 with MMX disabled */
 298         for(x=0;x<w;x++){
 299             if(srca[x]){
 300 #ifdef FAST_OSD
 301                 dst[0]=dst[1]=dst[2]=src[x];
 302 #else
 303                 dst[0]=((dst[0]*srca[x])>>8)+src[x];
 304                 dst[1]=((dst[1]*srca[x])>>8)+src[x];
 305                 dst[2]=((dst[2]*srca[x])>>8)+src[x];
 306 #endif
 307             }
 308             dst+=3; // 24bpp
 309         }
 310 #endif /* arch_x86 */
 311         src+=srcstride;
 312         srca+=srcstride;
 313         dstbase+=dststride;
 314     }
 315 #if HAVE_MMX
 316         __asm__ volatile(EMMS:::"memory");
 317 #endif
 318     return;
 319 }
 320
 321 static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
 322     int y;
 323 #if HAVE_BIGENDIAN
 324     dstbase++;
 325 #endif
 326 #if HAVE_MMX
 327 #if HAVE_AMD3DNOW
 328     __asm__ volatile(
 329         "pxor %%mm7, %%mm7\n\t"
 330         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 331         ::);
 332 #else /* HAVE_AMD3DNOW */
 333     __asm__ volatile(
 334         "pxor %%mm7, %%mm7\n\t"
 335         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
 336         "movq %%mm5, %%mm4\n\t"
 337         "psllw $8, %%mm5\n\t" //FF00FF00FF00
 338         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
 339         ::);
 340 #endif /* HAVE_AMD3DNOW */
 341 #endif /* HAVE_MMX */
 342     for(y=0;y<h;y++){
 343         register int x;
 344 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 345 #if HAVE_MMX
 346 #if HAVE_AMD3DNOW
 347     __asm__ volatile(
 348         PREFETCHW" %0\n\t"
 349         PREFETCH" %1\n\t"
 350         PREFETCH" %2\n\t"
 351         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 352     for(x=0;x<w;x+=2){
 353      if(srca[x] || srca[x+1])
 354         __asm__ volatile(
 355                 PREFETCHW" 32%0\n\t"
 356                 PREFETCH" 32%1\n\t"
 357                 PREFETCH" 32%2\n\t"
 358                 "movq   %0, %%mm0\n\t" // dstbase
 359                 "movq   %%mm0, %%mm1\n\t"
 360                 "punpcklbw %%mm7, %%mm0\n\t"
 361                 "punpckhbw %%mm7, %%mm1\n\t"
 362                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 363                 "paddb  %%mm6, %%mm2\n\t"
 364                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 365                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 366                 "movq   %%mm2, %%mm3\n\t"
 367                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0A
 368                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B0B
 369                 "pmullw %%mm2, %%mm0\n\t"
 370                 "pmullw %%mm3, %%mm1\n\t"
 371                 "psrlw  $8, %%mm0\n\t"
 372                 "psrlw  $8, %%mm1\n\t"
 373                 "packuswb %%mm1, %%mm0\n\t"
 374                 "movd %2, %%mm2 \n\t" // src ABCD0000
 375                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 376                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 377                 "paddb  %%mm2, %%mm0\n\t"
 378                 "movq   %%mm0, %0\n\t"
 379                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]));
 380         }
 381 #else //this is faster for intels crap
 382     __asm__ volatile(
 383         PREFETCHW" %0\n\t"
 384         PREFETCH" %1\n\t"
 385         PREFETCH" %2\n\t"
 386         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 387     for(x=0;x<w;x+=4){
 388         __asm__ volatile(
 389                 "movl %1, %%eax\n\t"
 390                 "orl %%eax, %%eax\n\t"
 391                 " jz 1f\n\t"
 392                 PREFETCHW" 32%0\n\t"
 393                 PREFETCH" 32%1\n\t"
 394                 PREFETCH" 32%2\n\t"
 395                 "movq   %0, %%mm0\n\t" // dstbase
 396                 "movq   %%mm0, %%mm1\n\t"
 397                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 398                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 399                 "movd   %%eax, %%mm2\n\t"       //srca 0000DCBA
 400                 "paddb  %3, %%mm2\n\t"
 401                 "punpcklbw %%mm2, %%mm2\n\t"    //srca DDCCBBAA
 402                 "movq %%mm2, %%mm3\n\t"
 403                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0B0B0A0A
 404                 "pmullw %%mm2, %%mm0\n\t"
 405                 "pmullw %%mm2, %%mm1\n\t"
 406                 "psrlw  $8, %%mm0\n\t"
 407                 "pand %%mm5, %%mm1\n\t"
 408                 "por %%mm1, %%mm0\n\t"
 409                 "movd %2, %%mm2 \n\t"           //src 0000DCBA
 410                 "punpcklbw %%mm2, %%mm2\n\t"    //src DDCCBBAA
 411                 "movq %%mm2, %%mm6\n\t"
 412                 "punpcklbw %%mm2, %%mm2\n\t"    //src BBBBAAAA
 413                 "paddb  %%mm2, %%mm0\n\t"
 414                 "movq   %%mm0, %0\n\t"
 415
 416                 "movq   8%0, %%mm0\n\t" // dstbase
 417                 "movq   %%mm0, %%mm1\n\t"
 418                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 419                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 420                 "punpckhbw %%mm7, %%mm3\n\t"    //srca 0D0D0C0C
 421                 "pmullw %%mm3, %%mm0\n\t"
 422                 "pmullw %%mm3, %%mm1\n\t"
 423                 "psrlw  $8, %%mm0\n\t"
 424                 "pand %%mm5, %%mm1\n\t"
 425                 "por %%mm1, %%mm0\n\t"
 426                 "punpckhbw %%mm6, %%mm6\n\t"    //src DDDDCCCC
 427                 "paddb  %%mm6, %%mm0\n\t"
 428                 "movq   %%mm0, 8%0\n\t"
 429                 "1:\n\t"
 430                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]), "m" (bFF)
 431                 : "%eax");
 432         }
 433 #endif
 434 #else /* HAVE_MMX */
 435     for(x=0;x<w;x++){
 436         if(srca[x]){
 437             __asm__ volatile(
 438                 "movzbl (%0), %%ecx\n\t"
 439                 "movzbl 1(%0), %%eax\n\t"
 440                 "movzbl 2(%0), %%edx\n\t"
 441
 442                 "imull %1, %%ecx\n\t"
 443                 "imull %1, %%eax\n\t"
 444                 "imull %1, %%edx\n\t"
 445
 446                 "addl %2, %%ecx\n\t"
 447                 "addl %2, %%eax\n\t"
 448                 "addl %2, %%edx\n\t"
 449
 450                 "movb %%ch, (%0)\n\t"
 451                 "movb %%ah, 1(%0)\n\t"
 452                 "movb %%dh, 2(%0)\n\t"
 453
 454                 :
 455                 :"r" (&dstbase[4*x]),
 456                  "r" ((unsigned)srca[x]),
 457                  "r" (((unsigned)src[x])<<8)
 458                 :"%eax", "%ecx", "%edx"
 459                 );
 460             }
 461         }
 462 #endif /* HAVE_MMX */
 463 #else /*non x86 arch or x86_64 with MMX disabled */
 464         for(x=0;x<w;x++){
 465             if(srca[x]){
 466 #ifdef FAST_OSD
 467                 dstbase[4*x+0]=dstbase[4*x+1]=dstbase[4*x+2]=src[x];
 468 #else
 469                 dstbase[4*x+0]=((dstbase[4*x+0]*srca[x])>>8)+src[x];
 470                 dstbase[4*x+1]=((dstbase[4*x+1]*srca[x])>>8)+src[x];
 471                 dstbase[4*x+2]=((dstbase[4*x+2]*srca[x])>>8)+src[x];
 472 #endif
 473             }
 474         }
 475 #endif /* arch_x86 */
 476         src+=srcstride;
 477         srca+=srcstride;
 478         dstbase+=dststride;
 479     }
 480 #if HAVE_MMX
 481         __asm__ volatile(EMMS:::"memory");
 482 #endif
 483     return;
 484 }