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 HAVE_MMX
  50     __asm__ volatile(
  51         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
  52         "movq %%mm5, %%mm4\n\t"
  53         "movq %%mm5, %%mm7\n\t"
  54         "psllw $8, %%mm5\n\t" //FF00FF00FF00
  55         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
  56         ::);
  57 #endif
  58     for(y=0;y<h;y++){
  59         register int x;
  60 #if HAVE_MMX
  61     __asm__ volatile(
  62         PREFETCHW" %0\n\t"
  63         PREFETCH" %1\n\t"
  64         PREFETCH" %2\n\t"
  65         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
  66     for(x=0;x<w;x+=8){
  67         __asm__ volatile(
  68                 "movl %1, %%eax\n\t"
  69                 "orl 4%1, %%eax\n\t"
  70                 " jz 1f\n\t"
  71                 PREFETCHW" 32%0\n\t"
  72                 PREFETCH" 32%1\n\t"
  73                 PREFETCH" 32%2\n\t"
  74                 "movq   %0, %%mm0\n\t" // dstbase
  75                 "movq   %%mm0, %%mm1\n\t"
  76                 "pand %%mm4, %%mm0\n\t"         //0Y0Y0Y0Y
  77                 "psrlw $8, %%mm1\n\t"           //0Y0Y0Y0Y
  78                 "movq   %1, %%mm2\n\t"          //srca HGFEDCBA
  79                 "paddb  %%mm7, %%mm2\n\t"
  80                 "movq %%mm2, %%mm3\n\t"
  81                 "pand %%mm4, %%mm2\n\t"         //0G0E0C0A
  82                 "psrlw $8, %%mm3\n\t"           //0H0F0D0B
  83                 "pmullw %%mm2, %%mm0\n\t"
  84                 "pmullw %%mm3, %%mm1\n\t"
  85                 "psrlw  $8, %%mm0\n\t"
  86                 "pand %%mm5, %%mm1\n\t"
  87                 "por %%mm1, %%mm0\n\t"
  88                 "paddb  %2, %%mm0\n\t"
  89                 "movq   %%mm0, %0\n\t"
  90                 "1:\n\t"
  91                 :: "m" (dstbase[x]), "m" (srca[x]), "m" (src[x])
  92                 : "%eax");
  93         }
  94 #else
  95         for(x=0;x<w;x++){
  96             if(srca[x]) dstbase[x]=((dstbase[x]*srca[x])>>8)+src[x];
  97         }
  98 #endif
  99         src+=srcstride;
 100         srca+=srcstride;
 101         dstbase+=dststride;
 102     }
 103 #if HAVE_MMX
 104         __asm__ volatile(EMMS:::"memory");
 105 #endif
 106     return;
 107 }
 108
 109 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){
 110     int y;
 111 #if HAVE_MMX
 112     __asm__ volatile(
 113         "pxor %%mm7, %%mm7\n\t"
 114         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
 115         "movq %%mm5, %%mm6\n\t"
 116         "movq %%mm5, %%mm4\n\t"
 117         "psllw $8, %%mm5\n\t" //FF00FF00FF00
 118         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
 119         ::);
 120 #endif
 121     for(y=0;y<h;y++){
 122         register int x;
 123 #if HAVE_MMX
 124     __asm__ volatile(
 125         PREFETCHW" %0\n\t"
 126         PREFETCH" %1\n\t"
 127         PREFETCH" %2\n\t"
 128         ::"m"(*dstbase),"m"(*srca),"m"(*src));
 129     for(x=0;x<w;x+=4){
 130         __asm__ volatile(
 131                 "movl %1, %%eax\n\t"
 132                 "orl %%eax, %%eax\n\t"
 133                 " jz 1f\n\t"
 134                 PREFETCHW" 32%0\n\t"
 135                 PREFETCH" 32%1\n\t"
 136                 PREFETCH" 32%2\n\t"
 137                 "movq   %0, %%mm0\n\t" // dstbase
 138                 "movq   %%mm0, %%mm1\n\t"
 139                 "pand %%mm4, %%mm0\n\t"         //0Y0Y0Y0Y
 140                 "movd   %%eax, %%mm2\n\t"       //srca 0000DCBA
 141                 "paddb  %%mm6, %%mm2\n\t"
 142                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0D0C0B0A
 143                 "pmullw %%mm2, %%mm0\n\t"
 144                 "psrlw  $8, %%mm0\n\t"
 145                 "pand %%mm5, %%mm1\n\t"         //U0V0U0V0
 146                 "movd %2, %%mm2\n\t"            //src 0000DCBA
 147                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0D0C0B0A
 148                 "por %%mm1, %%mm0\n\t"
 149                 "paddb  %%mm2, %%mm0\n\t"
 150                 "movq   %%mm0, %0\n\t"
 151                 "1:\n\t"
 152                 :: "m" (dstbase[x*2]), "m" (srca[x]), "m" (src[x])
 153                 : "%eax");
 154         }
 155 #else
 156         for(x=0;x<w;x++){
 157             if(srca[x]) {
 158                dstbase[2*x]=((dstbase[2*x]*srca[x])>>8)+src[x];
 159                dstbase[2*x+1]=((((signed)dstbase[2*x+1]-128)*srca[x])>>8)+128;
 160            }
 161         }
 162 #endif
 163         src+=srcstride;
 164         srca+=srcstride;
 165         dstbase+=dststride;
 166     }
 167 #if HAVE_MMX
 168         __asm__ volatile(EMMS:::"memory");
 169 #endif
 170     return;
 171 }
 172
 173 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){
 174     int y;
 175 #if HAVE_MMX
 176     __asm__ volatile(
 177         "pxor %%mm7, %%mm7\n\t"
 178         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 179         ::);
 180 #endif
 181     for(y=0;y<h;y++){
 182         register unsigned char *dst = dstbase;
 183         register int x;
 184 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 185 #if HAVE_MMX
 186     __asm__ volatile(
 187         PREFETCHW" %0\n\t"
 188         PREFETCH" %1\n\t"
 189         PREFETCH" %2\n\t"
 190         ::"m"(*dst),"m"(*srca),"m"(*src):"memory");
 191     for(x=0;x<w;x+=2){
 192      if(srca[x] || srca[x+1])
 193         __asm__ volatile(
 194                 PREFETCHW" 32%0\n\t"
 195                 PREFETCH" 32%1\n\t"
 196                 PREFETCH" 32%2\n\t"
 197                 "movq   %0, %%mm0\n\t" // dstbase
 198                 "movq   %%mm0, %%mm1\n\t"
 199                 "movq   %%mm0, %%mm5\n\t"
 200                 "punpcklbw %%mm7, %%mm0\n\t"
 201                 "punpckhbw %%mm7, %%mm1\n\t"
 202                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 203                 "paddb  %%mm6, %%mm2\n\t"
 204                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 205                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 206                 "psrlq  $8, %%mm2\n\t" // srca AAABBBB0
 207                 "movq   %%mm2, %%mm3\n\t"
 208                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0B
 209                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B00
 210                 "pmullw %%mm2, %%mm0\n\t"
 211                 "pmullw %%mm3, %%mm1\n\t"
 212                 "psrlw  $8, %%mm0\n\t"
 213                 "psrlw  $8, %%mm1\n\t"
 214                 "packuswb %%mm1, %%mm0\n\t"
 215                 "movd %2, %%mm2 \n\t" // src ABCD0000
 216                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 217                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 218                 "psrlq  $8, %%mm2\n\t" // src AAABBBB0
 219                 "paddb  %%mm2, %%mm0\n\t"
 220                 "pand   %4, %%mm5\n\t"
 221                 "pand   %3, %%mm0\n\t"
 222                 "por    %%mm0, %%mm5\n\t"
 223                 "movq   %%mm5, %0\n\t"
 224                 :: "m" (dst[0]), "m" (srca[x]), "m" (src[x]), "m"(mask24hl), "m"(mask24lh));
 225                 dst += 6;
 226         }
 227 #else /* HAVE_MMX */
 228     for(x=0;x<w;x++){
 229         if(srca[x]){
 230             __asm__ volatile(
 231                 "movzbl (%0), %%ecx\n\t"
 232                 "movzbl 1(%0), %%eax\n\t"
 233
 234                 "imull %1, %%ecx\n\t"
 235                 "imull %1, %%eax\n\t"
 236
 237                 "addl %2, %%ecx\n\t"
 238                 "addl %2, %%eax\n\t"
 239
 240                 "movb %%ch, (%0)\n\t"
 241                 "movb %%ah, 1(%0)\n\t"
 242
 243                 "movzbl 2(%0), %%eax\n\t"
 244                 "imull %1, %%eax\n\t"
 245                 "addl %2, %%eax\n\t"
 246                 "movb %%ah, 2(%0)\n\t"
 247                 :
 248                 :"D" (dst),
 249                  "r" ((unsigned)srca[x]),
 250                  "r" (((unsigned)src[x])<<8)
 251                 :"%eax", "%ecx"
 252                 );
 253             }
 254             dst += 3;
 255         }
 256 #endif /* !HAVE_MMX */
 257 #else /*non x86 arch or x86_64 with MMX disabled */
 258         for(x=0;x<w;x++){
 259             if(srca[x]){
 260                 dst[0]=((dst[0]*srca[x])>>8)+src[x];
 261                 dst[1]=((dst[1]*srca[x])>>8)+src[x];
 262                 dst[2]=((dst[2]*srca[x])>>8)+src[x];
 263             }
 264             dst+=3; // 24bpp
 265         }
 266 #endif /* arch_x86 */
 267         src+=srcstride;
 268         srca+=srcstride;
 269         dstbase+=dststride;
 270     }
 271 #if HAVE_MMX
 272         __asm__ volatile(EMMS:::"memory");
 273 #endif
 274     return;
 275 }
 276
 277 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){
 278     int y;
 279 #if HAVE_BIGENDIAN
 280     dstbase++;
 281 #endif
 282 #if HAVE_MMX
 283 #if HAVE_AMD3DNOW
 284     __asm__ volatile(
 285         "pxor %%mm7, %%mm7\n\t"
 286         "pcmpeqb %%mm6, %%mm6\n\t" // F..F
 287         ::);
 288 #else /* HAVE_AMD3DNOW */
 289     __asm__ volatile(
 290         "pxor %%mm7, %%mm7\n\t"
 291         "pcmpeqb %%mm5, %%mm5\n\t" // F..F
 292         "movq %%mm5, %%mm4\n\t"
 293         "psllw $8, %%mm5\n\t" //FF00FF00FF00
 294         "psrlw $8, %%mm4\n\t" //00FF00FF00FF
 295         ::);
 296 #endif /* HAVE_AMD3DNOW */
 297 #endif /* HAVE_MMX */
 298     for(y=0;y<h;y++){
 299         register int x;
 300 #if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX)
 301 #if HAVE_MMX
 302 #if HAVE_AMD3DNOW
 303     __asm__ volatile(
 304         PREFETCHW" %0\n\t"
 305         PREFETCH" %1\n\t"
 306         PREFETCH" %2\n\t"
 307         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 308     for(x=0;x<w;x+=2){
 309      if(srca[x] || srca[x+1])
 310         __asm__ volatile(
 311                 PREFETCHW" 32%0\n\t"
 312                 PREFETCH" 32%1\n\t"
 313                 PREFETCH" 32%2\n\t"
 314                 "movq   %0, %%mm0\n\t" // dstbase
 315                 "movq   %%mm0, %%mm1\n\t"
 316                 "punpcklbw %%mm7, %%mm0\n\t"
 317                 "punpckhbw %%mm7, %%mm1\n\t"
 318                 "movd   %1, %%mm2\n\t" // srca ABCD0000
 319                 "paddb  %%mm6, %%mm2\n\t"
 320                 "punpcklbw %%mm2, %%mm2\n\t" // srca AABBCCDD
 321                 "punpcklbw %%mm2, %%mm2\n\t" // srca AAAABBBB
 322                 "movq   %%mm2, %%mm3\n\t"
 323                 "punpcklbw %%mm7, %%mm2\n\t" // srca 0A0A0A0A
 324                 "punpckhbw %%mm7, %%mm3\n\t" // srca 0B0B0B0B
 325                 "pmullw %%mm2, %%mm0\n\t"
 326                 "pmullw %%mm3, %%mm1\n\t"
 327                 "psrlw  $8, %%mm0\n\t"
 328                 "psrlw  $8, %%mm1\n\t"
 329                 "packuswb %%mm1, %%mm0\n\t"
 330                 "movd %2, %%mm2 \n\t" // src ABCD0000
 331                 "punpcklbw %%mm2, %%mm2\n\t" // src AABBCCDD
 332                 "punpcklbw %%mm2, %%mm2\n\t" // src AAAABBBB
 333                 "paddb  %%mm2, %%mm0\n\t"
 334                 "movq   %%mm0, %0\n\t"
 335                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]));
 336         }
 337 #else //this is faster for intels crap
 338     __asm__ volatile(
 339         PREFETCHW" %0\n\t"
 340         PREFETCH" %1\n\t"
 341         PREFETCH" %2\n\t"
 342         ::"m"(*dstbase),"m"(*srca),"m"(*src):"memory");
 343     for(x=0;x<w;x+=4){
 344         __asm__ volatile(
 345                 "movl %1, %%eax\n\t"
 346                 "orl %%eax, %%eax\n\t"
 347                 " jz 1f\n\t"
 348                 PREFETCHW" 32%0\n\t"
 349                 PREFETCH" 32%1\n\t"
 350                 PREFETCH" 32%2\n\t"
 351                 "movq   %0, %%mm0\n\t" // dstbase
 352                 "movq   %%mm0, %%mm1\n\t"
 353                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 354                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 355                 "movd   %%eax, %%mm2\n\t"       //srca 0000DCBA
 356                 "paddb  %3, %%mm2\n\t"
 357                 "punpcklbw %%mm2, %%mm2\n\t"    //srca DDCCBBAA
 358                 "movq %%mm2, %%mm3\n\t"
 359                 "punpcklbw %%mm7, %%mm2\n\t"    //srca 0B0B0A0A
 360                 "pmullw %%mm2, %%mm0\n\t"
 361                 "pmullw %%mm2, %%mm1\n\t"
 362                 "psrlw  $8, %%mm0\n\t"
 363                 "pand %%mm5, %%mm1\n\t"
 364                 "por %%mm1, %%mm0\n\t"
 365                 "movd %2, %%mm2 \n\t"           //src 0000DCBA
 366                 "punpcklbw %%mm2, %%mm2\n\t"    //src DDCCBBAA
 367                 "movq %%mm2, %%mm6\n\t"
 368                 "punpcklbw %%mm2, %%mm2\n\t"    //src BBBBAAAA
 369                 "paddb  %%mm2, %%mm0\n\t"
 370                 "movq   %%mm0, %0\n\t"
 371
 372                 "movq   8%0, %%mm0\n\t" // dstbase
 373                 "movq   %%mm0, %%mm1\n\t"
 374                 "pand %%mm4, %%mm0\n\t"         //0R0B0R0B
 375                 "psrlw $8, %%mm1\n\t"           //0?0G0?0G
 376                 "punpckhbw %%mm7, %%mm3\n\t"    //srca 0D0D0C0C
 377                 "pmullw %%mm3, %%mm0\n\t"
 378                 "pmullw %%mm3, %%mm1\n\t"
 379                 "psrlw  $8, %%mm0\n\t"
 380                 "pand %%mm5, %%mm1\n\t"
 381                 "por %%mm1, %%mm0\n\t"
 382                 "punpckhbw %%mm6, %%mm6\n\t"    //src DDDDCCCC
 383                 "paddb  %%mm6, %%mm0\n\t"
 384                 "movq   %%mm0, 8%0\n\t"
 385                 "1:\n\t"
 386                 :: "m" (dstbase[4*x]), "m" (srca[x]), "m" (src[x]), "m" (bFF)
 387                 : "%eax");
 388         }
 389 #endif
 390 #else /* HAVE_MMX */
 391     for(x=0;x<w;x++){
 392         if(srca[x]){
 393             __asm__ volatile(
 394                 "movzbl (%0), %%ecx\n\t"
 395                 "movzbl 1(%0), %%eax\n\t"
 396                 "movzbl 2(%0), %%edx\n\t"
 397
 398                 "imull %1, %%ecx\n\t"
 399                 "imull %1, %%eax\n\t"
 400                 "imull %1, %%edx\n\t"
 401
 402                 "addl %2, %%ecx\n\t"
 403                 "addl %2, %%eax\n\t"
 404                 "addl %2, %%edx\n\t"
 405
 406                 "movb %%ch, (%0)\n\t"
 407                 "movb %%ah, 1(%0)\n\t"
 408                 "movb %%dh, 2(%0)\n\t"
 409
 410                 :
 411                 :"r" (&dstbase[4*x]),
 412                  "r" ((unsigned)srca[x]),
 413                  "r" (((unsigned)src[x])<<8)
 414                 :"%eax", "%ecx", "%edx"
 415                 );
 416             }
 417         }
 418 #endif /* HAVE_MMX */
 419 #else /*non x86 arch or x86_64 with MMX disabled */
 420         for(x=0;x<w;x++){
 421             if(srca[x]){
 422                 dstbase[4*x+0]=((dstbase[4*x+0]*srca[x])>>8)+src[x];
 423                 dstbase[4*x+1]=((dstbase[4*x+1]*srca[x])>>8)+src[x];
 424                 dstbase[4*x+2]=((dstbase[4*x+2]*srca[x])>>8)+src[x];
 425             }
 426         }
 427 #endif /* arch_x86 */
 428         src+=srcstride;
 429         srca+=srcstride;
 430         dstbase+=dststride;
 431     }
 432 #if HAVE_MMX
 433         __asm__ volatile(EMMS:::"memory");
 434 #endif
 435     return;
 436 }