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_rgb24)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
 190     int y;
 191 #if HAVE_MMX
 192     __asm__ volatile(
 193         "pxor %%mm7, %%mm7\n\t"
 194         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 195         ::);
 196 #endif
 197     for(y=0;y<h;y++){
 198         register unsigned char *dst = dstbase;
 199         register int x;
 200 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 201 #if HAVE_MMX
 202     __asm__ volatile(
 203         PREFETCHW" %0\n\t"
 204         PREFETCH" %1\n\t"
 205         PREFETCH" %2\n\t"
 206         ::"m"(*dst),"m"(*srca),"m"(*src):"memory");
 207     for(x=0;x<w;x+=2){
 208      if(srca[x] || srca[x+1])
 209         __asm__ volatile(
 210                 PREFETCHW" 32%0\n\t"
 211                 PREFETCH" 32%1\n\t"
 212                 PREFETCH" 32%2\n\t"
 213                 "movq   %0, %%mm0\n\t" // dstbase
 214                 "movq   %%mm0, %%mm1\n\t"
 215                 "movq   %%mm0, %%mm5\n\t"
 216                 "punpcklbw %%mm7, %%mm0\n\t"
 217                 "punpckhbw %%mm7, %%mm1\n\t"
 218                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 219                 "paddb  %%mm6, %%mm2\n\t"
 220                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 221                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 222                 "psrlq  $8, %%mm2\n\t" // srca AAABBBB0
 223                 "movq   %%mm2, %%mm3\n\t"
 224                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0B
 225                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B00
 226                 "pmullw %%mm2, %%mm0\n\t"
 227                 "pmullw %%mm3, %%mm1\n\t"
 228                 "psrlw  $8, %%mm0\n\t"
 229                 "psrlw  $8, %%mm1\n\t"
 230                 "packuswb %%mm1, %%mm0\n\t"
 231                 "movd %2, %%mm2 \n\t" // src ABCD0000
 232                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 233                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 234                 "psrlq  $8, %%mm2\n\t" // src AAABBBB0
 235                 "paddb  %%mm2, %%mm0\n\t"
 236                 "pand   %4, %%mm5\n\t"
 237                 "pand   %3, %%mm0\n\t"
 238                 "por    %%mm0, %%mm5\n\t"
 239                 "movq   %%mm5, %0\n\t"
 240                 :: "m" (dst[0]), "m" (srca[x]), "m" (src[x]), "m"(mask24hl), "m"(mask24lh));
 241                 dst += 6;
 242         }
 243 #else /* HAVE_MMX */
 244     for(x=0;x<w;x++){
 245         if(srca[x]){
 246             __asm__ volatile(
 247                 "movzbl (%0), %%ecx\n\t"
 248                 "movzbl 1(%0), %%eax\n\t"
 249
 250                 "imull %1, %%ecx\n\t"
 251                 "imull %1, %%eax\n\t"
 252
 253                 "addl %2, %%ecx\n\t"
 254                 "addl %2, %%eax\n\t"
 255
 256                 "movb %%ch, (%0)\n\t"
 257                 "movb %%ah, 1(%0)\n\t"
 258
 259                 "movzbl 2(%0), %%eax\n\t"
 260                 "imull %1, %%eax\n\t"
 261                 "addl %2, %%eax\n\t"
 262                 "movb %%ah, 2(%0)\n\t"
 263                 :
 264                 :"D" (dst),
 265                  "r" ((unsigned)srca[x]),
 266                  "r" (((unsigned)src[x])<<8)
 267                 :"%eax", "%ecx"
 268                 );
 269             }
 270             dst += 3;
 271         }
 272 #endif /* !HAVE_MMX */
 273 #else /*non x86 arch or x86_64 with MMX disabled */
 274         for(x=0;x<w;x++){
 275             if(srca[x]){
 276 #ifdef FAST_OSD
 277                 dst[0]=dst[1]=dst[2]=src[x];
 278 #else
 279                 dst[0]=((dst[0]*srca[x])>>8)+src[x];
 280                 dst[1]=((dst[1]*srca[x])>>8)+src[x];
 281                 dst[2]=((dst[2]*srca[x])>>8)+src[x];
 282 #endif
 283             }
 284             dst+=3; // 24bpp
 285         }
 286 #endif /* arch_x86 */
 287         src+=srcstride;
 288         srca+=srcstride;
 289         dstbase+=dststride;
 290     }
 291 #if HAVE_MMX
 292         __asm__ volatile(EMMS:::"memory");
 293 #endif
 294     return;
 295 }
 296
 297 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){
 298     int y;
 299 #if HAVE_BIGENDIAN
 300     dstbase++;
 301 #endif
 302 #if HAVE_MMX
 303 #if HAVE_AMD3DNOW
 304     __asm__ volatile(
 305         "pxor %%mm7, %%mm7\n\t"
 306         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 307         ::);
 308 #else /* HAVE_AMD3DNOW */
 309     __asm__ volatile(
 310         "pxor %%mm7, %%mm7\n\t"
 311         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
 312         "movq %%mm5, %%mm4\n\t"
 313         "psllw $8, %%mm5\n\t" //FF00FF00FF00
 314         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
 315         ::);
 316 #endif /* HAVE_AMD3DNOW */
 317 #endif /* HAVE_MMX */
 318     for(y=0;y<h;y++){
 319         register int x;
 320 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 321 #if HAVE_MMX
 322 #if HAVE_AMD3DNOW
 323     __asm__ volatile(
 324         PREFETCHW" %0\n\t"
 325         PREFETCH" %1\n\t"
 326         PREFETCH" %2\n\t"
 327         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 328     for(x=0;x<w;x+=2){
 329      if(srca[x] || srca[x+1])
 330         __asm__ volatile(
 331                 PREFETCHW" 32%0\n\t"
 332                 PREFETCH" 32%1\n\t"
 333                 PREFETCH" 32%2\n\t"
 334                 "movq   %0, %%mm0\n\t" // dstbase
 335                 "movq   %%mm0, %%mm1\n\t"
 336                 "punpcklbw %%mm7, %%mm0\n\t"
 337                 "punpckhbw %%mm7, %%mm1\n\t"
 338                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 339                 "paddb  %%mm6, %%mm2\n\t"
 340                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 341                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 342                 "movq   %%mm2, %%mm3\n\t"
 343                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0A
 344                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B0B
 345                 "pmullw %%mm2, %%mm0\n\t"
 346                 "pmullw %%mm3, %%mm1\n\t"
 347                 "psrlw  $8, %%mm0\n\t"
 348                 "psrlw  $8, %%mm1\n\t"
 349                 "packuswb %%mm1, %%mm0\n\t"
 350                 "movd %2, %%mm2 \n\t" // src ABCD0000
 351                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 352                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 353                 "paddb  %%mm2, %%mm0\n\t"
 354                 "movq   %%mm0, %0\n\t"
 355                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]));
 356         }
 357 #else //this is faster for intels crap
 358     __asm__ volatile(
 359         PREFETCHW" %0\n\t"
 360         PREFETCH" %1\n\t"
 361         PREFETCH" %2\n\t"
 362         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 363     for(x=0;x<w;x+=4){
 364         __asm__ volatile(
 365                 "movl %1, %%eax\n\t"
 366                 "orl %%eax, %%eax\n\t"
 367                 " jz 1f\n\t"
 368                 PREFETCHW" 32%0\n\t"
 369                 PREFETCH" 32%1\n\t"
 370                 PREFETCH" 32%2\n\t"
 371                 "movq   %0, %%mm0\n\t" // dstbase
 372                 "movq   %%mm0, %%mm1\n\t"
 373                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 374                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 375                 "movd   %%eax, %%mm2\n\t"       //srca 0000DCBA
 376                 "paddb  %3, %%mm2\n\t"
 377                 "punpcklbw %%mm2, %%mm2\n\t"    //srca DDCCBBAA
 378                 "movq %%mm2, %%mm3\n\t"
 379                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0B0B0A0A
 380                 "pmullw %%mm2, %%mm0\n\t"
 381                 "pmullw %%mm2, %%mm1\n\t"
 382                 "psrlw  $8, %%mm0\n\t"
 383                 "pand %%mm5, %%mm1\n\t"
 384                 "por %%mm1, %%mm0\n\t"
 385                 "movd %2, %%mm2 \n\t"           //src 0000DCBA
 386                 "punpcklbw %%mm2, %%mm2\n\t"    //src DDCCBBAA
 387                 "movq %%mm2, %%mm6\n\t"
 388                 "punpcklbw %%mm2, %%mm2\n\t"    //src BBBBAAAA
 389                 "paddb  %%mm2, %%mm0\n\t"
 390                 "movq   %%mm0, %0\n\t"
 391
 392                 "movq   8%0, %%mm0\n\t" // dstbase
 393                 "movq   %%mm0, %%mm1\n\t"
 394                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 395                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 396                 "punpckhbw %%mm7, %%mm3\n\t"    //srca 0D0D0C0C
 397                 "pmullw %%mm3, %%mm0\n\t"
 398                 "pmullw %%mm3, %%mm1\n\t"
 399                 "psrlw  $8, %%mm0\n\t"
 400                 "pand %%mm5, %%mm1\n\t"
 401                 "por %%mm1, %%mm0\n\t"
 402                 "punpckhbw %%mm6, %%mm6\n\t"    //src DDDDCCCC
 403                 "paddb  %%mm6, %%mm0\n\t"
 404                 "movq   %%mm0, 8%0\n\t"
 405                 "1:\n\t"
 406                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]), "m" (bFF)
 407                 : "%eax");
 408         }
 409 #endif
 410 #else /* HAVE_MMX */
 411     for(x=0;x<w;x++){
 412         if(srca[x]){
 413             __asm__ volatile(
 414                 "movzbl (%0), %%ecx\n\t"
 415                 "movzbl 1(%0), %%eax\n\t"
 416                 "movzbl 2(%0), %%edx\n\t"
 417
 418                 "imull %1, %%ecx\n\t"
 419                 "imull %1, %%eax\n\t"
 420                 "imull %1, %%edx\n\t"
 421
 422                 "addl %2, %%ecx\n\t"
 423                 "addl %2, %%eax\n\t"
 424                 "addl %2, %%edx\n\t"
 425
 426                 "movb %%ch, (%0)\n\t"
 427                 "movb %%ah, 1(%0)\n\t"
 428                 "movb %%dh, 2(%0)\n\t"
 429
 430                 :
 431                 :"r" (&dstbase[4*x]),
 432                  "r" ((unsigned)srca[x]),
 433                  "r" (((unsigned)src[x])<<8)
 434                 :"%eax", "%ecx", "%edx"
 435                 );
 436             }
 437         }
 438 #endif /* HAVE_MMX */
 439 #else /*non x86 arch or x86_64 with MMX disabled */
 440         for(x=0;x<w;x++){
 441             if(srca[x]){
 442 #ifdef FAST_OSD
 443                 dstbase[4*x+0]=dstbase[4*x+1]=dstbase[4*x+2]=src[x];
 444 #else
 445                 dstbase[4*x+0]=((dstbase[4*x+0]*srca[x])>>8)+src[x];
 446                 dstbase[4*x+1]=((dstbase[4*x+1]*srca[x])>>8)+src[x];
 447                 dstbase[4*x+2]=((dstbase[4*x+2]*srca[x])>>8)+src[x];
 448 #endif
 449             }
 450         }
 451 #endif /* arch_x86 */
 452         src+=srcstride;
 453         srca+=srcstride;
 454         dstbase+=dststride;
 455     }
 456 #if HAVE_MMX
 457         __asm__ volatile(EMMS:::"memory");
 458 #endif
 459     return;
 460 }