git.sesse.net Git - ffmpeg/blob - libswscale/swscale_template.c

   1 /*
   2  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
   3  *
   4  * This file is part of FFmpeg.
   5  *
   6  * FFmpeg is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  *
  11  * FFmpeg is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with FFmpeg; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  *
  20  * The C code (not assembly, MMX, ...) of this file can be used
  21  * under the LGPL license.
  22  */
  23
  24 #undef REAL_MOVNTQ
  25 #undef MOVNTQ
  26 #undef PAVGB
  27 #undef PREFETCH
  28 #undef PREFETCHW
  29
  30 #if COMPILE_TEMPLATE_AMD3DNOW
  31 #define PREFETCH  "prefetch"
  32 #define PREFETCHW "prefetchw"
  33 #elif COMPILE_TEMPLATE_MMX2
  34 #define PREFETCH "prefetchnta"
  35 #define PREFETCHW "prefetcht0"
  36 #else
  37 #define PREFETCH  " # nop"
  38 #define PREFETCHW " # nop"
  39 #endif
  40
  41 #if COMPILE_TEMPLATE_MMX2
  42 #define PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
  43 #elif COMPILE_TEMPLATE_AMD3DNOW
  44 #define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
  45 #endif
  46
  47 #if COMPILE_TEMPLATE_MMX2
  48 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
  49 #else
  50 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
  51 #endif
  52 #define MOVNTQ(a,b)  REAL_MOVNTQ(a,b)
  53
  54 #if COMPILE_TEMPLATE_ALTIVEC
  55 #include "ppc/swscale_altivec_template.c"
  56 #endif
  57
  58 #define YSCALEYUV2YV12X(x, offset, dest, width) \
  59     __asm__ volatile(\
  60         "xor                          %%"REG_a", %%"REG_a"  \n\t"\
  61         "movq             "VROUNDER_OFFSET"(%0), %%mm3      \n\t"\
  62         "movq                             %%mm3, %%mm4      \n\t"\
  63         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
  64         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  65         ASMALIGN(4) /* FIXME Unroll? */\
  66         "1:                                                 \n\t"\
  67         "movq                      8(%%"REG_d"), %%mm0      \n\t" /* filterCoeff */\
  68         "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm2      \n\t" /* srcData */\
  69         "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm5      \n\t" /* srcData */\
  70         "add                                $16, %%"REG_d"  \n\t"\
  71         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  72         "test                         %%"REG_S", %%"REG_S"  \n\t"\
  73         "pmulhw                           %%mm0, %%mm2      \n\t"\
  74         "pmulhw                           %%mm0, %%mm5      \n\t"\
  75         "paddw                            %%mm2, %%mm3      \n\t"\
  76         "paddw                            %%mm5, %%mm4      \n\t"\
  77         " jnz                                1b             \n\t"\
  78         "psraw                               $3, %%mm3      \n\t"\
  79         "psraw                               $3, %%mm4      \n\t"\
  80         "packuswb                         %%mm4, %%mm3      \n\t"\
  81         MOVNTQ(%%mm3, (%1, %%REGa))\
  82         "add                                 $8, %%"REG_a"  \n\t"\
  83         "cmp                                 %2, %%"REG_a"  \n\t"\
  84         "movq             "VROUNDER_OFFSET"(%0), %%mm3      \n\t"\
  85         "movq                             %%mm3, %%mm4      \n\t"\
  86         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
  87         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  88         "jb                                  1b             \n\t"\
  89         :: "r" (&c->redDither),\
  90         "r" (dest), "g" (width)\
  91         : "%"REG_a, "%"REG_d, "%"REG_S\
  92     );
  93
  94 #define YSCALEYUV2YV12X_ACCURATE(x, offset, dest, width) \
  95     __asm__ volatile(\
  96         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
  97         "xor                          %%"REG_a", %%"REG_a"  \n\t"\
  98         "pxor                             %%mm4, %%mm4      \n\t"\
  99         "pxor                             %%mm5, %%mm5      \n\t"\
 100         "pxor                             %%mm6, %%mm6      \n\t"\
 101         "pxor                             %%mm7, %%mm7      \n\t"\
 102         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 103         ASMALIGN(4) \
 104         "1:                                                 \n\t"\
 105         "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm0      \n\t" /* srcData */\
 106         "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm2      \n\t" /* srcData */\
 107         "mov        "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"  \n\t"\
 108         "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm1      \n\t" /* srcData */\
 109         "movq                             %%mm0, %%mm3      \n\t"\
 110         "punpcklwd                        %%mm1, %%mm0      \n\t"\
 111         "punpckhwd                        %%mm1, %%mm3      \n\t"\
 112         "movq       "STR(APCK_COEF)"(%%"REG_d"), %%mm1      \n\t" /* filterCoeff */\
 113         "pmaddwd                          %%mm1, %%mm0      \n\t"\
 114         "pmaddwd                          %%mm1, %%mm3      \n\t"\
 115         "paddd                            %%mm0, %%mm4      \n\t"\
 116         "paddd                            %%mm3, %%mm5      \n\t"\
 117         "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm3      \n\t" /* srcData */\
 118         "mov        "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"  \n\t"\
 119         "add                  $"STR(APCK_SIZE)", %%"REG_d"  \n\t"\
 120         "test                         %%"REG_S", %%"REG_S"  \n\t"\
 121         "movq                             %%mm2, %%mm0      \n\t"\
 122         "punpcklwd                        %%mm3, %%mm2      \n\t"\
 123         "punpckhwd                        %%mm3, %%mm0      \n\t"\
 124         "pmaddwd                          %%mm1, %%mm2      \n\t"\
 125         "pmaddwd                          %%mm1, %%mm0      \n\t"\
 126         "paddd                            %%mm2, %%mm6      \n\t"\
 127         "paddd                            %%mm0, %%mm7      \n\t"\
 128         " jnz                                1b             \n\t"\
 129         "psrad                              $16, %%mm4      \n\t"\
 130         "psrad                              $16, %%mm5      \n\t"\
 131         "psrad                              $16, %%mm6      \n\t"\
 132         "psrad                              $16, %%mm7      \n\t"\
 133         "movq             "VROUNDER_OFFSET"(%0), %%mm0      \n\t"\
 134         "packssdw                         %%mm5, %%mm4      \n\t"\
 135         "packssdw                         %%mm7, %%mm6      \n\t"\
 136         "paddw                            %%mm0, %%mm4      \n\t"\
 137         "paddw                            %%mm0, %%mm6      \n\t"\
 138         "psraw                               $3, %%mm4      \n\t"\
 139         "psraw                               $3, %%mm6      \n\t"\
 140         "packuswb                         %%mm6, %%mm4      \n\t"\
 141         MOVNTQ(%%mm4, (%1, %%REGa))\
 142         "add                                 $8, %%"REG_a"  \n\t"\
 143         "cmp                                 %2, %%"REG_a"  \n\t"\
 144         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
 145         "pxor                             %%mm4, %%mm4      \n\t"\
 146         "pxor                             %%mm5, %%mm5      \n\t"\
 147         "pxor                             %%mm6, %%mm6      \n\t"\
 148         "pxor                             %%mm7, %%mm7      \n\t"\
 149         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 150         "jb                                  1b             \n\t"\
 151         :: "r" (&c->redDither),\
 152         "r" (dest), "g" (width)\
 153         : "%"REG_a, "%"REG_d, "%"REG_S\
 154     );
 155
 156 #define YSCALEYUV2YV121 \
 157     "mov %2, %%"REG_a"                    \n\t"\
 158     ASMALIGN(4) /* FIXME Unroll? */\
 159     "1:                                   \n\t"\
 160     "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"\
 161     "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"\
 162     "psraw                 $7, %%mm0      \n\t"\
 163     "psraw                 $7, %%mm1      \n\t"\
 164     "packuswb           %%mm1, %%mm0      \n\t"\
 165     MOVNTQ(%%mm0, (%1, %%REGa))\
 166     "add                   $8, %%"REG_a"  \n\t"\
 167     "jnc                   1b             \n\t"
 168
 169 #define YSCALEYUV2YV121_ACCURATE \
 170     "mov %2, %%"REG_a"                    \n\t"\
 171     "pcmpeqw %%mm7, %%mm7                 \n\t"\
 172     "psrlw                 $15, %%mm7     \n\t"\
 173     "psllw                  $6, %%mm7     \n\t"\
 174     ASMALIGN(4) /* FIXME Unroll? */\
 175     "1:                                   \n\t"\
 176     "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"\
 177     "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"\
 178     "paddsw             %%mm7, %%mm0      \n\t"\
 179     "paddsw             %%mm7, %%mm1      \n\t"\
 180     "psraw                 $7, %%mm0      \n\t"\
 181     "psraw                 $7, %%mm1      \n\t"\
 182     "packuswb           %%mm1, %%mm0      \n\t"\
 183     MOVNTQ(%%mm0, (%1, %%REGa))\
 184     "add                   $8, %%"REG_a"  \n\t"\
 185     "jnc                   1b             \n\t"
 186
 187 /*
 188     :: "m" (-lumFilterSize), "m" (-chrFilterSize),
 189        "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
 190        "r" (dest), "m" (dstW),
 191        "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
 192     : "%eax", "%ebx", "%ecx", "%edx", "%esi"
 193 */
 194 #define YSCALEYUV2PACKEDX_UV \
 195     __asm__ volatile(\
 196         "xor                   %%"REG_a", %%"REG_a"     \n\t"\
 197         ASMALIGN(4)\
 198         "nop                                            \n\t"\
 199         "1:                                             \n\t"\
 200         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 201         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 202         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\
 203         "movq                      %%mm3, %%mm4         \n\t"\
 204         ASMALIGN(4)\
 205         "2:                                             \n\t"\
 206         "movq               8(%%"REG_d"), %%mm0         \n\t" /* filterCoeff */\
 207         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* UsrcData */\
 208         "movq "AV_STRINGIFY(VOF)"(%%"REG_S", %%"REG_a"), %%mm5         \n\t" /* VsrcData */\
 209         "add                         $16, %%"REG_d"     \n\t"\
 210         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 211         "pmulhw                    %%mm0, %%mm2         \n\t"\
 212         "pmulhw                    %%mm0, %%mm5         \n\t"\
 213         "paddw                     %%mm2, %%mm3         \n\t"\
 214         "paddw                     %%mm5, %%mm4         \n\t"\
 215         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 216         " jnz                         2b                \n\t"\
 217
 218 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
 219     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 220     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 221     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\
 222     "movq                    "#dst1", "#dst2"       \n\t"\
 223     ASMALIGN(4)\
 224     "2:                                             \n\t"\
 225     "movq               8(%%"REG_d"), "#coeff"      \n\t" /* filterCoeff */\
 226     "movq  (%%"REG_S", %%"REG_a", 2), "#src1"       \n\t" /* Y1srcData */\
 227     "movq 8(%%"REG_S", %%"REG_a", 2), "#src2"       \n\t" /* Y2srcData */\
 228     "add                         $16, %%"REG_d"            \n\t"\
 229     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 230     "pmulhw                 "#coeff", "#src1"       \n\t"\
 231     "pmulhw                 "#coeff", "#src2"       \n\t"\
 232     "paddw                   "#src1", "#dst1"       \n\t"\
 233     "paddw                   "#src2", "#dst2"       \n\t"\
 234     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 235     " jnz                         2b                \n\t"\
 236
 237 #define YSCALEYUV2PACKEDX \
 238     YSCALEYUV2PACKEDX_UV \
 239     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
 240
 241 #define YSCALEYUV2PACKEDX_END                     \
 242         :: "r" (&c->redDither),                   \
 243             "m" (dummy), "m" (dummy), "m" (dummy),\
 244             "r" (dest), "m" (dstW)                \
 245         : "%"REG_a, "%"REG_d, "%"REG_S            \
 246     );
 247
 248 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
 249     __asm__ volatile(\
 250         "xor %%"REG_a", %%"REG_a"                       \n\t"\
 251         ASMALIGN(4)\
 252         "nop                                            \n\t"\
 253         "1:                                             \n\t"\
 254         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 255         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 256         "pxor                      %%mm4, %%mm4         \n\t"\
 257         "pxor                      %%mm5, %%mm5         \n\t"\
 258         "pxor                      %%mm6, %%mm6         \n\t"\
 259         "pxor                      %%mm7, %%mm7         \n\t"\
 260         ASMALIGN(4)\
 261         "2:                                             \n\t"\
 262         "movq     (%%"REG_S", %%"REG_a"), %%mm0         \n\t" /* UsrcData */\
 263         "movq "AV_STRINGIFY(VOF)"(%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* VsrcData */\
 264         "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 265         "movq     (%%"REG_S", %%"REG_a"), %%mm1         \n\t" /* UsrcData */\
 266         "movq                      %%mm0, %%mm3         \n\t"\
 267         "punpcklwd                 %%mm1, %%mm0         \n\t"\
 268         "punpckhwd                 %%mm1, %%mm3         \n\t"\
 269         "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1         \n\t" /* filterCoeff */\
 270         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 271         "pmaddwd                   %%mm1, %%mm3         \n\t"\
 272         "paddd                     %%mm0, %%mm4         \n\t"\
 273         "paddd                     %%mm3, %%mm5         \n\t"\
 274         "movq "AV_STRINGIFY(VOF)"(%%"REG_S", %%"REG_a"), %%mm3         \n\t" /* VsrcData */\
 275         "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 276         "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 277         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 278         "movq                      %%mm2, %%mm0         \n\t"\
 279         "punpcklwd                 %%mm3, %%mm2         \n\t"\
 280         "punpckhwd                 %%mm3, %%mm0         \n\t"\
 281         "pmaddwd                   %%mm1, %%mm2         \n\t"\
 282         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 283         "paddd                     %%mm2, %%mm6         \n\t"\
 284         "paddd                     %%mm0, %%mm7         \n\t"\
 285         " jnz                         2b                \n\t"\
 286         "psrad                       $16, %%mm4         \n\t"\
 287         "psrad                       $16, %%mm5         \n\t"\
 288         "psrad                       $16, %%mm6         \n\t"\
 289         "psrad                       $16, %%mm7         \n\t"\
 290         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 291         "packssdw                  %%mm5, %%mm4         \n\t"\
 292         "packssdw                  %%mm7, %%mm6         \n\t"\
 293         "paddw                     %%mm0, %%mm4         \n\t"\
 294         "paddw                     %%mm0, %%mm6         \n\t"\
 295         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\
 296         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\
 297
 298 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
 299     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 300     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 301     "pxor                      %%mm1, %%mm1         \n\t"\
 302     "pxor                      %%mm5, %%mm5         \n\t"\
 303     "pxor                      %%mm7, %%mm7         \n\t"\
 304     "pxor                      %%mm6, %%mm6         \n\t"\
 305     ASMALIGN(4)\
 306     "2:                                             \n\t"\
 307     "movq  (%%"REG_S", %%"REG_a", 2), %%mm0         \n\t" /* Y1srcData */\
 308     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y2srcData */\
 309     "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 310     "movq  (%%"REG_S", %%"REG_a", 2), %%mm4         \n\t" /* Y1srcData */\
 311     "movq                      %%mm0, %%mm3         \n\t"\
 312     "punpcklwd                 %%mm4, %%mm0         \n\t"\
 313     "punpckhwd                 %%mm4, %%mm3         \n\t"\
 314     "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4         \n\t" /* filterCoeff */\
 315     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 316     "pmaddwd                   %%mm4, %%mm3         \n\t"\
 317     "paddd                     %%mm0, %%mm1         \n\t"\
 318     "paddd                     %%mm3, %%mm5         \n\t"\
 319     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3         \n\t" /* Y2srcData */\
 320     "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 321     "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 322     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 323     "movq                      %%mm2, %%mm0         \n\t"\
 324     "punpcklwd                 %%mm3, %%mm2         \n\t"\
 325     "punpckhwd                 %%mm3, %%mm0         \n\t"\
 326     "pmaddwd                   %%mm4, %%mm2         \n\t"\
 327     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 328     "paddd                     %%mm2, %%mm7         \n\t"\
 329     "paddd                     %%mm0, %%mm6         \n\t"\
 330     " jnz                         2b                \n\t"\
 331     "psrad                       $16, %%mm1         \n\t"\
 332     "psrad                       $16, %%mm5         \n\t"\
 333     "psrad                       $16, %%mm7         \n\t"\
 334     "psrad                       $16, %%mm6         \n\t"\
 335     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 336     "packssdw                  %%mm5, %%mm1         \n\t"\
 337     "packssdw                  %%mm6, %%mm7         \n\t"\
 338     "paddw                     %%mm0, %%mm1         \n\t"\
 339     "paddw                     %%mm0, %%mm7         \n\t"\
 340     "movq               "U_TEMP"(%0), %%mm3         \n\t"\
 341     "movq               "V_TEMP"(%0), %%mm4         \n\t"\
 342
 343 #define YSCALEYUV2PACKEDX_ACCURATE \
 344     YSCALEYUV2PACKEDX_ACCURATE_UV \
 345     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
 346
 347 #define YSCALEYUV2RGBX \
 348     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\
 349     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\
 350     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\
 351     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\
 352     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\
 353     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\
 354     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 355     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\
 356     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\
 357     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\
 358     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\
 359     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\
 360     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\
 361     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 362     "paddw           %%mm3, %%mm4       \n\t"\
 363     "movq            %%mm2, %%mm0       \n\t"\
 364     "movq            %%mm5, %%mm6       \n\t"\
 365     "movq            %%mm4, %%mm3       \n\t"\
 366     "punpcklwd       %%mm2, %%mm2       \n\t"\
 367     "punpcklwd       %%mm5, %%mm5       \n\t"\
 368     "punpcklwd       %%mm4, %%mm4       \n\t"\
 369     "paddw           %%mm1, %%mm2       \n\t"\
 370     "paddw           %%mm1, %%mm5       \n\t"\
 371     "paddw           %%mm1, %%mm4       \n\t"\
 372     "punpckhwd       %%mm0, %%mm0       \n\t"\
 373     "punpckhwd       %%mm6, %%mm6       \n\t"\
 374     "punpckhwd       %%mm3, %%mm3       \n\t"\
 375     "paddw           %%mm7, %%mm0       \n\t"\
 376     "paddw           %%mm7, %%mm6       \n\t"\
 377     "paddw           %%mm7, %%mm3       \n\t"\
 378     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 379     "packuswb        %%mm0, %%mm2       \n\t"\
 380     "packuswb        %%mm6, %%mm5       \n\t"\
 381     "packuswb        %%mm3, %%mm4       \n\t"\
 382
 383 #define REAL_YSCALEYUV2PACKED(index, c) \
 384     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\
 385     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\
 386     "psraw                $3, %%mm0                           \n\t"\
 387     "psraw                $3, %%mm1                           \n\t"\
 388     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
 389     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
 390     "xor            "#index", "#index"                        \n\t"\
 391     ASMALIGN(4)\
 392     "1:                                 \n\t"\
 393     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 394     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 395     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 396     "movq "AV_STRINGIFY(VOF)"(%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 397     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
 398     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
 399     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
 400     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
 401     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
 402     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 403     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 404     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
 405     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
 406     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
 407     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
 408     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
 409     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
 410     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
 411     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
 412     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 413     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 414     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 415     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 416     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 417     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 418
 419 #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)
 420
 421 #define REAL_YSCALEYUV2RGB_UV(index, c) \
 422     "xor            "#index", "#index"  \n\t"\
 423     ASMALIGN(4)\
 424     "1:                                 \n\t"\
 425     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 426     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 427     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 428     "movq "AV_STRINGIFY(VOF)"(%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 429     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
 430     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
 431     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
 432     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
 433     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
 434     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 435     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 436     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
 437     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
 438     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 439     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 440     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 441     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 442     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 443     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 444     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 445
 446 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
 447     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
 448     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
 449     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
 450     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
 451     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
 452     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
 453     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 454     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 455     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 456     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 457     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 458     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 459
 460 #define REAL_YSCALEYUV2RGB_COEFF(c) \
 461     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 462     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
 463     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
 464     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
 465     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
 466     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
 467     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 468     "paddw             %%mm3, %%mm4     \n\t"\
 469     "movq              %%mm2, %%mm0     \n\t"\
 470     "movq              %%mm5, %%mm6     \n\t"\
 471     "movq              %%mm4, %%mm3     \n\t"\
 472     "punpcklwd         %%mm2, %%mm2     \n\t"\
 473     "punpcklwd         %%mm5, %%mm5     \n\t"\
 474     "punpcklwd         %%mm4, %%mm4     \n\t"\
 475     "paddw             %%mm1, %%mm2     \n\t"\
 476     "paddw             %%mm1, %%mm5     \n\t"\
 477     "paddw             %%mm1, %%mm4     \n\t"\
 478     "punpckhwd         %%mm0, %%mm0     \n\t"\
 479     "punpckhwd         %%mm6, %%mm6     \n\t"\
 480     "punpckhwd         %%mm3, %%mm3     \n\t"\
 481     "paddw             %%mm7, %%mm0     \n\t"\
 482     "paddw             %%mm7, %%mm6     \n\t"\
 483     "paddw             %%mm7, %%mm3     \n\t"\
 484     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 485     "packuswb          %%mm0, %%mm2     \n\t"\
 486     "packuswb          %%mm6, %%mm5     \n\t"\
 487     "packuswb          %%mm3, %%mm4     \n\t"\
 488
 489 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
 490
 491 #define YSCALEYUV2RGB(index, c) \
 492     REAL_YSCALEYUV2RGB_UV(index, c) \
 493     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
 494     REAL_YSCALEYUV2RGB_COEFF(c)
 495
 496 #define REAL_YSCALEYUV2PACKED1(index, c) \
 497     "xor            "#index", "#index"  \n\t"\
 498     ASMALIGN(4)\
 499     "1:                                 \n\t"\
 500     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
 501     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
 502     "psraw                $7, %%mm3     \n\t" \
 503     "psraw                $7, %%mm4     \n\t" \
 504     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
 505     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
 506     "psraw                $7, %%mm1     \n\t" \
 507     "psraw                $7, %%mm7     \n\t" \
 508
 509 #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)
 510
 511 #define REAL_YSCALEYUV2RGB1(index, c) \
 512     "xor            "#index", "#index"  \n\t"\
 513     ASMALIGN(4)\
 514     "1:                                 \n\t"\
 515     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
 516     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
 517     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 518     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 519     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 520     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 521     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 522     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 523     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 524     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 525     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 526     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
 527     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
 528     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 529     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 530     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 531     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
 532     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
 533     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
 534     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
 535     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
 536     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 537     "paddw             %%mm3, %%mm4     \n\t"\
 538     "movq              %%mm2, %%mm0     \n\t"\
 539     "movq              %%mm5, %%mm6     \n\t"\
 540     "movq              %%mm4, %%mm3     \n\t"\
 541     "punpcklwd         %%mm2, %%mm2     \n\t"\
 542     "punpcklwd         %%mm5, %%mm5     \n\t"\
 543     "punpcklwd         %%mm4, %%mm4     \n\t"\
 544     "paddw             %%mm1, %%mm2     \n\t"\
 545     "paddw             %%mm1, %%mm5     \n\t"\
 546     "paddw             %%mm1, %%mm4     \n\t"\
 547     "punpckhwd         %%mm0, %%mm0     \n\t"\
 548     "punpckhwd         %%mm6, %%mm6     \n\t"\
 549     "punpckhwd         %%mm3, %%mm3     \n\t"\
 550     "paddw             %%mm7, %%mm0     \n\t"\
 551     "paddw             %%mm7, %%mm6     \n\t"\
 552     "paddw             %%mm7, %%mm3     \n\t"\
 553     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 554     "packuswb          %%mm0, %%mm2     \n\t"\
 555     "packuswb          %%mm6, %%mm5     \n\t"\
 556     "packuswb          %%mm3, %%mm4     \n\t"\
 557
 558 #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)
 559
 560 #define REAL_YSCALEYUV2PACKED1b(index, c) \
 561     "xor "#index", "#index"             \n\t"\
 562     ASMALIGN(4)\
 563     "1:                                 \n\t"\
 564     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 565     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 566     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 567     "movq "AV_STRINGIFY(VOF)"(%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 568     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
 569     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
 570     "psrlw                $8, %%mm3     \n\t" \
 571     "psrlw                $8, %%mm4     \n\t" \
 572     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
 573     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
 574     "psraw                $7, %%mm1     \n\t" \
 575     "psraw                $7, %%mm7     \n\t"
 576 #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)
 577
 578 // do vertical chrominance interpolation
 579 #define REAL_YSCALEYUV2RGB1b(index, c) \
 580     "xor            "#index", "#index"  \n\t"\
 581     ASMALIGN(4)\
 582     "1:                                 \n\t"\
 583     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 584     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 585     "movq "AV_STRINGIFY(VOF)"(%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 586     "movq "AV_STRINGIFY(VOF)"(%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 587     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
 588     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
 589     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
 590     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\
 591     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 592     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 593     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 594     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 595     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 596     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 597     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 598     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
 599     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
 600     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 601     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 602     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 603     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
 604     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
 605     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
 606     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
 607     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
 608     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 609     "paddw             %%mm3, %%mm4     \n\t"\
 610     "movq              %%mm2, %%mm0     \n\t"\
 611     "movq              %%mm5, %%mm6     \n\t"\
 612     "movq              %%mm4, %%mm3     \n\t"\
 613     "punpcklwd         %%mm2, %%mm2     \n\t"\
 614     "punpcklwd         %%mm5, %%mm5     \n\t"\
 615     "punpcklwd         %%mm4, %%mm4     \n\t"\
 616     "paddw             %%mm1, %%mm2     \n\t"\
 617     "paddw             %%mm1, %%mm5     \n\t"\
 618     "paddw             %%mm1, %%mm4     \n\t"\
 619     "punpckhwd         %%mm0, %%mm0     \n\t"\
 620     "punpckhwd         %%mm6, %%mm6     \n\t"\
 621     "punpckhwd         %%mm3, %%mm3     \n\t"\
 622     "paddw             %%mm7, %%mm0     \n\t"\
 623     "paddw             %%mm7, %%mm6     \n\t"\
 624     "paddw             %%mm7, %%mm3     \n\t"\
 625     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 626     "packuswb          %%mm0, %%mm2     \n\t"\
 627     "packuswb          %%mm6, %%mm5     \n\t"\
 628     "packuswb          %%mm3, %%mm4     \n\t"\
 629
 630 #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)
 631
 632 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
 633     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\
 634     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\
 635     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\
 636     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\
 637     "packuswb          %%mm1, %%mm7     \n\t"
 638 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
 639
 640 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
 641     "movq       "#b", "#q2"     \n\t" /* B */\
 642     "movq       "#r", "#t"      \n\t" /* R */\
 643     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\
 644     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\
 645     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\
 646     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\
 647     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\
 648     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\
 649     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\
 650     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\
 651     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\
 652     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\
 653 \
 654     MOVNTQ(   q0,   (dst, index, 4))\
 655     MOVNTQ(    b,  8(dst, index, 4))\
 656     MOVNTQ(   q2, 16(dst, index, 4))\
 657     MOVNTQ(   q3, 24(dst, index, 4))\
 658 \
 659     "add      $8, "#index"      \n\t"\
 660     "cmp "#dstw", "#index"      \n\t"\
 661     " jb      1b                \n\t"
 662 #define WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t)  REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t)
 663
 664 #define REAL_WRITERGB16(dst, dstw, index) \
 665     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 666     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
 667     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 668     "psrlq           $3, %%mm2  \n\t"\
 669 \
 670     "movq         %%mm2, %%mm1  \n\t"\
 671     "movq         %%mm4, %%mm3  \n\t"\
 672 \
 673     "punpcklbw    %%mm7, %%mm3  \n\t"\
 674     "punpcklbw    %%mm5, %%mm2  \n\t"\
 675     "punpckhbw    %%mm7, %%mm4  \n\t"\
 676     "punpckhbw    %%mm5, %%mm1  \n\t"\
 677 \
 678     "psllq           $3, %%mm3  \n\t"\
 679     "psllq           $3, %%mm4  \n\t"\
 680 \
 681     "por          %%mm3, %%mm2  \n\t"\
 682     "por          %%mm4, %%mm1  \n\t"\
 683 \
 684     MOVNTQ(%%mm2,  (dst, index, 2))\
 685     MOVNTQ(%%mm1, 8(dst, index, 2))\
 686 \
 687     "add             $8, "#index"   \n\t"\
 688     "cmp        "#dstw", "#index"   \n\t"\
 689     " jb             1b             \n\t"
 690 #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)
 691
 692 #define REAL_WRITERGB15(dst, dstw, index) \
 693     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 694     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\
 695     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 696     "psrlq           $3, %%mm2  \n\t"\
 697     "psrlq           $1, %%mm5  \n\t"\
 698 \
 699     "movq         %%mm2, %%mm1  \n\t"\
 700     "movq         %%mm4, %%mm3  \n\t"\
 701 \
 702     "punpcklbw    %%mm7, %%mm3  \n\t"\
 703     "punpcklbw    %%mm5, %%mm2  \n\t"\
 704     "punpckhbw    %%mm7, %%mm4  \n\t"\
 705     "punpckhbw    %%mm5, %%mm1  \n\t"\
 706 \
 707     "psllq           $2, %%mm3  \n\t"\
 708     "psllq           $2, %%mm4  \n\t"\
 709 \
 710     "por          %%mm3, %%mm2  \n\t"\
 711     "por          %%mm4, %%mm1  \n\t"\
 712 \
 713     MOVNTQ(%%mm2,  (dst, index, 2))\
 714     MOVNTQ(%%mm1, 8(dst, index, 2))\
 715 \
 716     "add             $8, "#index"   \n\t"\
 717     "cmp        "#dstw", "#index"   \n\t"\
 718     " jb             1b             \n\t"
 719 #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)
 720
 721 #define WRITEBGR24OLD(dst, dstw, index) \
 722     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 723     "movq      %%mm2, %%mm1             \n\t" /* B */\
 724     "movq      %%mm5, %%mm6             \n\t" /* R */\
 725     "punpcklbw %%mm4, %%mm2             \n\t" /* GBGBGBGB 0 */\
 726     "punpcklbw %%mm7, %%mm5             \n\t" /* 0R0R0R0R 0 */\
 727     "punpckhbw %%mm4, %%mm1             \n\t" /* GBGBGBGB 2 */\
 728     "punpckhbw %%mm7, %%mm6             \n\t" /* 0R0R0R0R 2 */\
 729     "movq      %%mm2, %%mm0             \n\t" /* GBGBGBGB 0 */\
 730     "movq      %%mm1, %%mm3             \n\t" /* GBGBGBGB 2 */\
 731     "punpcklwd %%mm5, %%mm0             \n\t" /* 0RGB0RGB 0 */\
 732     "punpckhwd %%mm5, %%mm2             \n\t" /* 0RGB0RGB 1 */\
 733     "punpcklwd %%mm6, %%mm1             \n\t" /* 0RGB0RGB 2 */\
 734     "punpckhwd %%mm6, %%mm3             \n\t" /* 0RGB0RGB 3 */\
 735 \
 736     "movq      %%mm0, %%mm4             \n\t" /* 0RGB0RGB 0 */\
 737     "psrlq        $8, %%mm0             \n\t" /* 00RGB0RG 0 */\
 738     "pand "MANGLE(bm00000111)", %%mm4   \n\t" /* 00000RGB 0 */\
 739     "pand "MANGLE(bm11111000)", %%mm0   \n\t" /* 00RGB000 0.5 */\
 740     "por       %%mm4, %%mm0             \n\t" /* 00RGBRGB 0 */\
 741     "movq      %%mm2, %%mm4             \n\t" /* 0RGB0RGB 1 */\
 742     "psllq       $48, %%mm2             \n\t" /* GB000000 1 */\
 743     "por       %%mm2, %%mm0             \n\t" /* GBRGBRGB 0 */\
 744 \
 745     "movq      %%mm4, %%mm2             \n\t" /* 0RGB0RGB 1 */\
 746     "psrld       $16, %%mm4             \n\t" /* 000R000R 1 */\
 747     "psrlq       $24, %%mm2             \n\t" /* 0000RGB0 1.5 */\
 748     "por       %%mm4, %%mm2             \n\t" /* 000RRGBR 1 */\
 749     "pand "MANGLE(bm00001111)", %%mm2   \n\t" /* 0000RGBR 1 */\
 750     "movq      %%mm1, %%mm4             \n\t" /* 0RGB0RGB 2 */\
 751     "psrlq        $8, %%mm1             \n\t" /* 00RGB0RG 2 */\
 752     "pand "MANGLE(bm00000111)", %%mm4   \n\t" /* 00000RGB 2 */\
 753     "pand "MANGLE(bm11111000)", %%mm1   \n\t" /* 00RGB000 2.5 */\
 754     "por       %%mm4, %%mm1             \n\t" /* 00RGBRGB 2 */\
 755     "movq      %%mm1, %%mm4             \n\t" /* 00RGBRGB 2 */\
 756     "psllq       $32, %%mm1             \n\t" /* BRGB0000 2 */\
 757     "por       %%mm1, %%mm2             \n\t" /* BRGBRGBR 1 */\
 758 \
 759     "psrlq       $32, %%mm4             \n\t" /* 000000RG 2.5 */\
 760     "movq      %%mm3, %%mm5             \n\t" /* 0RGB0RGB 3 */\
 761     "psrlq        $8, %%mm3             \n\t" /* 00RGB0RG 3 */\
 762     "pand "MANGLE(bm00000111)", %%mm5   \n\t" /* 00000RGB 3 */\
 763     "pand "MANGLE(bm11111000)", %%mm3   \n\t" /* 00RGB000 3.5 */\
 764     "por       %%mm5, %%mm3             \n\t" /* 00RGBRGB 3 */\
 765     "psllq       $16, %%mm3             \n\t" /* RGBRGB00 3 */\
 766     "por       %%mm4, %%mm3             \n\t" /* RGBRGBRG 2.5 */\
 767 \
 768     MOVNTQ(%%mm0,   (dst))\
 769     MOVNTQ(%%mm2,  8(dst))\
 770     MOVNTQ(%%mm3, 16(dst))\
 771     "add         $24, "#dst"            \n\t"\
 772 \
 773     "add          $8, "#index"          \n\t"\
 774     "cmp     "#dstw", "#index"          \n\t"\
 775     " jb          1b                    \n\t"
 776
 777 #define WRITEBGR24MMX(dst, dstw, index) \
 778     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 779     "movq      %%mm2, %%mm1     \n\t" /* B */\
 780     "movq      %%mm5, %%mm6     \n\t" /* R */\
 781     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\
 782     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\
 783     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\
 784     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\
 785     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\
 786     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\
 787     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\
 788     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\
 789     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\
 790     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\
 791 \
 792     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\
 793     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\
 794     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\
 795     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\
 796 \
 797     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\
 798     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\
 799     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\
 800     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\
 801 \
 802     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\
 803     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\
 804     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\
 805     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\
 806 \
 807     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\
 808     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\
 809     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\
 810     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\
 811     MOVNTQ(%%mm0, (dst))\
 812 \
 813     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\
 814     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\
 815     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\
 816     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\
 817     MOVNTQ(%%mm6, 8(dst))\
 818 \
 819     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\
 820     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\
 821     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\
 822     MOVNTQ(%%mm5, 16(dst))\
 823 \
 824     "add         $24, "#dst"    \n\t"\
 825 \
 826     "add          $8, "#index"  \n\t"\
 827     "cmp     "#dstw", "#index"  \n\t"\
 828     " jb          1b            \n\t"
 829
 830 #define WRITEBGR24MMX2(dst, dstw, index) \
 831     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 832     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
 833     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
 834     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
 835     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
 836     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
 837 \
 838     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\
 839     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\
 840     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\
 841 \
 842     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\
 843     "por    %%mm1, %%mm6        \n\t"\
 844     "por    %%mm3, %%mm6        \n\t"\
 845     MOVNTQ(%%mm6, (dst))\
 846 \
 847     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
 848     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
 849     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
 850     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
 851 \
 852     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\
 853     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\
 854     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\
 855 \
 856     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\
 857     "por    %%mm3, %%mm6        \n\t"\
 858     MOVNTQ(%%mm6, 8(dst))\
 859 \
 860     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
 861     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
 862     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
 863 \
 864     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\
 865     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\
 866     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\
 867 \
 868     "por    %%mm1, %%mm3        \n\t"\
 869     "por    %%mm3, %%mm6        \n\t"\
 870     MOVNTQ(%%mm6, 16(dst))\
 871 \
 872     "add      $24, "#dst"       \n\t"\
 873 \
 874     "add       $8, "#index"     \n\t"\
 875     "cmp  "#dstw", "#index"     \n\t"\
 876     " jb       1b               \n\t"
 877
 878 #if COMPILE_TEMPLATE_MMX2
 879 #undef WRITEBGR24
 880 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX2(dst, dstw, index)
 881 #else
 882 #undef WRITEBGR24
 883 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)
 884 #endif
 885
 886 #define REAL_WRITEYUY2(dst, dstw, index) \
 887     "packuswb  %%mm3, %%mm3     \n\t"\
 888     "packuswb  %%mm4, %%mm4     \n\t"\
 889     "packuswb  %%mm7, %%mm1     \n\t"\
 890     "punpcklbw %%mm4, %%mm3     \n\t"\
 891     "movq      %%mm1, %%mm7     \n\t"\
 892     "punpcklbw %%mm3, %%mm1     \n\t"\
 893     "punpckhbw %%mm3, %%mm7     \n\t"\
 894 \
 895     MOVNTQ(%%mm1, (dst, index, 2))\
 896     MOVNTQ(%%mm7, 8(dst, index, 2))\
 897 \
 898     "add          $8, "#index"  \n\t"\
 899     "cmp     "#dstw", "#index"  \n\t"\
 900     " jb          1b            \n\t"
 901 #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)
 902
 903
 904 static inline void RENAME(yuv2yuvX)(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 905                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize, const int16_t **alpSrc,
 906                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, long dstW, long chrDstW)
 907 {
 908 #if COMPILE_TEMPLATE_MMX
 909     if(!(c->flags & SWS_BITEXACT)) {
 910         if (c->flags & SWS_ACCURATE_RND) {
 911             if (uDest) {
 912                 YSCALEYUV2YV12X_ACCURATE(   "0", CHR_MMX_FILTER_OFFSET, uDest, chrDstW)
 913                 YSCALEYUV2YV12X_ACCURATE(AV_STRINGIFY(VOF), CHR_MMX_FILTER_OFFSET, vDest, chrDstW)
 914             }
 915             if (CONFIG_SWSCALE_ALPHA && aDest) {
 916                 YSCALEYUV2YV12X_ACCURATE(   "0", ALP_MMX_FILTER_OFFSET, aDest, dstW)
 917             }
 918
 919             YSCALEYUV2YV12X_ACCURATE("0", LUM_MMX_FILTER_OFFSET, dest, dstW)
 920         } else {
 921             if (uDest) {
 922                 YSCALEYUV2YV12X(   "0", CHR_MMX_FILTER_OFFSET, uDest, chrDstW)
 923                 YSCALEYUV2YV12X(AV_STRINGIFY(VOF), CHR_MMX_FILTER_OFFSET, vDest, chrDstW)
 924             }
 925             if (CONFIG_SWSCALE_ALPHA && aDest) {
 926                 YSCALEYUV2YV12X(   "0", ALP_MMX_FILTER_OFFSET, aDest, dstW)
 927             }
 928
 929             YSCALEYUV2YV12X("0", LUM_MMX_FILTER_OFFSET, dest, dstW)
 930         }
 931         return;
 932     }
 933 #endif
 934 #if COMPILE_TEMPLATE_ALTIVEC
 935     yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize,
 936                           chrFilter, chrSrc, chrFilterSize,
 937                           dest, uDest, vDest, dstW, chrDstW);
 938 #else //COMPILE_TEMPLATE_ALTIVEC
 939     yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize,
 940                 chrFilter, chrSrc, chrFilterSize,
 941                 alpSrc, dest, uDest, vDest, aDest, dstW, chrDstW);
 942 #endif //!COMPILE_TEMPLATE_ALTIVEC
 943 }
 944
 945 static inline void RENAME(yuv2nv12X)(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
 946                                      const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
 947                                      uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, enum PixelFormat dstFormat)
 948 {
 949     yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize,
 950                  chrFilter, chrSrc, chrFilterSize,
 951                  dest, uDest, dstW, chrDstW, dstFormat);
 952 }
 953
 954 static inline void RENAME(yuv2yuv1)(SwsContext *c, const int16_t *lumSrc, const int16_t *chrSrc, const int16_t *alpSrc,
 955                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, long dstW, long chrDstW)
 956 {
 957     int i;
 958 #if COMPILE_TEMPLATE_MMX
 959     if(!(c->flags & SWS_BITEXACT)) {
 960         long p= 4;
 961         uint8_t *src[4]= {alpSrc + dstW, lumSrc + dstW, chrSrc + chrDstW, chrSrc + VOFW + chrDstW};
 962         uint8_t *dst[4]= {aDest, dest, uDest, vDest};
 963         x86_reg counter[4]= {dstW, dstW, chrDstW, chrDstW};
 964
 965         if (c->flags & SWS_ACCURATE_RND) {
 966             while(p--) {
 967                 if (dst[p]) {
 968                     __asm__ volatile(
 969                         YSCALEYUV2YV121_ACCURATE
 970                         :: "r" (src[p]), "r" (dst[p] + counter[p]),
 971                         "g" (-counter[p])
 972                         : "%"REG_a
 973                     );
 974                 }
 975             }
 976         } else {
 977             while(p--) {
 978                 if (dst[p]) {
 979                     __asm__ volatile(
 980                         YSCALEYUV2YV121
 981                         :: "r" (src[p]), "r" (dst[p] + counter[p]),
 982                         "g" (-counter[p])
 983                         : "%"REG_a
 984                     );
 985                 }
 986             }
 987         }
 988         return;
 989     }
 990 #endif
 991     for (i=0; i<dstW; i++) {
 992         int val= (lumSrc[i]+64)>>7;
 993
 994         if (val&256) {
 995             if (val<0) val=0;
 996             else       val=255;
 997         }
 998
 999         dest[i]= val;
1000     }
1001
1002     if (uDest)
1003         for (i=0; i<chrDstW; i++) {
1004             int u=(chrSrc[i       ]+64)>>7;
1005             int v=(chrSrc[i + VOFW]+64)>>7;
1006
1007             if ((u|v)&256) {
1008                 if (u<0)        u=0;
1009                 else if (u>255) u=255;
1010                 if (v<0)        v=0;
1011                 else if (v>255) v=255;
1012             }
1013
1014             uDest[i]= u;
1015             vDest[i]= v;
1016         }
1017
1018     if (CONFIG_SWSCALE_ALPHA && aDest)
1019         for (i=0; i<dstW; i++) {
1020             int val= (alpSrc[i]+64)>>7;
1021             aDest[i]= av_clip_uint8(val);
1022         }
1023 }
1024
1025
1026 /**
1027  * vertical scale YV12 to RGB
1028  */
1029 static inline void RENAME(yuv2packedX)(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1030                                        const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1031                                        const int16_t **alpSrc, uint8_t *dest, long dstW, long dstY)
1032 {
1033 #if COMPILE_TEMPLATE_MMX
1034     x86_reg dummy=0;
1035     if(!(c->flags & SWS_BITEXACT)) {
1036         if (c->flags & SWS_ACCURATE_RND) {
1037             switch(c->dstFormat) {
1038             case PIX_FMT_RGB32:
1039                 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1040                     YSCALEYUV2PACKEDX_ACCURATE
1041                     YSCALEYUV2RGBX
1042                     "movq                      %%mm2, "U_TEMP"(%0)  \n\t"
1043                     "movq                      %%mm4, "V_TEMP"(%0)  \n\t"
1044                     "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"
1045                     YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
1046                     "movq               "Y_TEMP"(%0), %%mm5         \n\t"
1047                     "psraw                        $3, %%mm1         \n\t"
1048                     "psraw                        $3, %%mm7         \n\t"
1049                     "packuswb                  %%mm7, %%mm1         \n\t"
1050                     WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
1051
1052                     YSCALEYUV2PACKEDX_END
1053                 } else {
1054                     YSCALEYUV2PACKEDX_ACCURATE
1055                     YSCALEYUV2RGBX
1056                     "pcmpeqd %%mm7, %%mm7 \n\t"
1057                     WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1058
1059                     YSCALEYUV2PACKEDX_END
1060                 }
1061                 return;
1062             case PIX_FMT_BGR24:
1063                 YSCALEYUV2PACKEDX_ACCURATE
1064                 YSCALEYUV2RGBX
1065                 "pxor %%mm7, %%mm7 \n\t"
1066                 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
1067                 "add %4, %%"REG_c"                        \n\t"
1068                 WRITEBGR24(%%REGc, %5, %%REGa)
1069
1070
1071                 :: "r" (&c->redDither),
1072                 "m" (dummy), "m" (dummy), "m" (dummy),
1073                 "r" (dest), "m" (dstW)
1074                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
1075                 );
1076                 return;
1077             case PIX_FMT_RGB555:
1078                 YSCALEYUV2PACKEDX_ACCURATE
1079                 YSCALEYUV2RGBX
1080                 "pxor %%mm7, %%mm7 \n\t"
1081                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1082 #ifdef DITHER1XBPP
1083                 "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
1084                 "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
1085                 "paddusb "RED_DITHER"(%0), %%mm5\n\t"
1086 #endif
1087
1088                 WRITERGB15(%4, %5, %%REGa)
1089                 YSCALEYUV2PACKEDX_END
1090                 return;
1091             case PIX_FMT_RGB565:
1092                 YSCALEYUV2PACKEDX_ACCURATE
1093                 YSCALEYUV2RGBX
1094                 "pxor %%mm7, %%mm7 \n\t"
1095                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1096 #ifdef DITHER1XBPP
1097                 "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
1098                 "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
1099                 "paddusb "RED_DITHER"(%0), %%mm5\n\t"
1100 #endif
1101
1102                 WRITERGB16(%4, %5, %%REGa)
1103                 YSCALEYUV2PACKEDX_END
1104                 return;
1105             case PIX_FMT_YUYV422:
1106                 YSCALEYUV2PACKEDX_ACCURATE
1107                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1108
1109                 "psraw $3, %%mm3    \n\t"
1110                 "psraw $3, %%mm4    \n\t"
1111                 "psraw $3, %%mm1    \n\t"
1112                 "psraw $3, %%mm7    \n\t"
1113                 WRITEYUY2(%4, %5, %%REGa)
1114                 YSCALEYUV2PACKEDX_END
1115                 return;
1116             }
1117         } else {
1118             switch(c->dstFormat) {
1119             case PIX_FMT_RGB32:
1120                 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1121                     YSCALEYUV2PACKEDX
1122                     YSCALEYUV2RGBX
1123                     YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
1124                     "psraw                        $3, %%mm1         \n\t"
1125                     "psraw                        $3, %%mm7         \n\t"
1126                     "packuswb                  %%mm7, %%mm1         \n\t"
1127                     WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
1128                     YSCALEYUV2PACKEDX_END
1129                 } else {
1130                     YSCALEYUV2PACKEDX
1131                     YSCALEYUV2RGBX
1132                     "pcmpeqd %%mm7, %%mm7 \n\t"
1133                     WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1134                     YSCALEYUV2PACKEDX_END
1135                 }
1136                 return;
1137             case PIX_FMT_BGR24:
1138                 YSCALEYUV2PACKEDX
1139                 YSCALEYUV2RGBX
1140                 "pxor                    %%mm7, %%mm7       \n\t"
1141                 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize
1142                 "add                        %4, %%"REG_c"   \n\t"
1143                 WRITEBGR24(%%REGc, %5, %%REGa)
1144
1145                 :: "r" (&c->redDither),
1146                 "m" (dummy), "m" (dummy), "m" (dummy),
1147                 "r" (dest),  "m" (dstW)
1148                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
1149                 );
1150                 return;
1151             case PIX_FMT_RGB555:
1152                 YSCALEYUV2PACKEDX
1153                 YSCALEYUV2RGBX
1154                 "pxor %%mm7, %%mm7 \n\t"
1155                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1156 #ifdef DITHER1XBPP
1157                 "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
1158                 "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
1159                 "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
1160 #endif
1161
1162                 WRITERGB15(%4, %5, %%REGa)
1163                 YSCALEYUV2PACKEDX_END
1164                 return;
1165             case PIX_FMT_RGB565:
1166                 YSCALEYUV2PACKEDX
1167                 YSCALEYUV2RGBX
1168                 "pxor %%mm7, %%mm7 \n\t"
1169                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1170 #ifdef DITHER1XBPP
1171                 "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
1172                 "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
1173                 "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
1174 #endif
1175
1176                 WRITERGB16(%4, %5, %%REGa)
1177                 YSCALEYUV2PACKEDX_END
1178                 return;
1179             case PIX_FMT_YUYV422:
1180                 YSCALEYUV2PACKEDX
1181                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1182
1183                 "psraw $3, %%mm3    \n\t"
1184                 "psraw $3, %%mm4    \n\t"
1185                 "psraw $3, %%mm1    \n\t"
1186                 "psraw $3, %%mm7    \n\t"
1187                 WRITEYUY2(%4, %5, %%REGa)
1188                 YSCALEYUV2PACKEDX_END
1189                 return;
1190             }
1191         }
1192     }
1193 #endif /* COMPILE_TEMPLATE_MMX */
1194 #if COMPILE_TEMPLATE_ALTIVEC
1195     /* The following list of supported dstFormat values should
1196        match what's found in the body of ff_yuv2packedX_altivec() */
1197     if (!(c->flags & SWS_BITEXACT) && !c->alpPixBuf &&
1198          (c->dstFormat==PIX_FMT_ABGR  || c->dstFormat==PIX_FMT_BGRA  ||
1199           c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 ||
1200           c->dstFormat==PIX_FMT_RGBA  || c->dstFormat==PIX_FMT_ARGB))
1201             ff_yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize,
1202                                    chrFilter, chrSrc, chrFilterSize,
1203                                    dest, dstW, dstY);
1204     else
1205 #endif
1206         yuv2packedXinC(c, lumFilter, lumSrc, lumFilterSize,
1207                        chrFilter, chrSrc, chrFilterSize,
1208                        alpSrc, dest, dstW, dstY);
1209 }
1210
1211 /**
1212  * vertical bilinear scale YV12 to RGB
1213  */
1214 static inline void RENAME(yuv2packed2)(SwsContext *c, const uint16_t *buf0, const uint16_t *buf1, const uint16_t *uvbuf0, const uint16_t *uvbuf1,
1215                           const uint16_t *abuf0, const uint16_t *abuf1, uint8_t *dest, int dstW, int yalpha, int uvalpha, int y)
1216 {
1217     int  yalpha1=4095- yalpha;
1218     int uvalpha1=4095-uvalpha;
1219     int i;
1220
1221 #if COMPILE_TEMPLATE_MMX
1222     if(!(c->flags & SWS_BITEXACT)) {
1223         switch(c->dstFormat) {
1224         //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1225         case PIX_FMT_RGB32:
1226             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1227 #if ARCH_X86_64
1228                 __asm__ volatile(
1229                     YSCALEYUV2RGB(%%REGBP, %5)
1230                     YSCALEYUV2RGB_YA(%%REGBP, %5, %6, %7)
1231                     "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1232                     "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1233                     "packuswb            %%mm7, %%mm1       \n\t"
1234                     WRITEBGR32(%4, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
1235
1236                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "r" (dest),
1237                     "a" (&c->redDither)
1238                     ,"r" (abuf0), "r" (abuf1)
1239                     : "%"REG_BP
1240                 );
1241 #else
1242                 *(uint16_t **)(&c->u_temp)=abuf0;
1243                 *(uint16_t **)(&c->v_temp)=abuf1;
1244                 __asm__ volatile(
1245                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1246                     "mov        %4, %%"REG_b"               \n\t"
1247                     "push %%"REG_BP"                        \n\t"
1248                     YSCALEYUV2RGB(%%REGBP, %5)
1249                     "push                   %0              \n\t"
1250                     "push                   %1              \n\t"
1251                     "mov          "U_TEMP"(%5), %0          \n\t"
1252                     "mov          "V_TEMP"(%5), %1          \n\t"
1253                     YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
1254                     "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1255                     "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1256                     "packuswb            %%mm7, %%mm1       \n\t"
1257                     "pop                    %1              \n\t"
1258                     "pop                    %0              \n\t"
1259                     WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
1260                     "pop %%"REG_BP"                         \n\t"
1261                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1262
1263                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1264                     "a" (&c->redDither)
1265                 );
1266 #endif
1267             } else {
1268                 __asm__ volatile(
1269                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1270                     "mov        %4, %%"REG_b"               \n\t"
1271                     "push %%"REG_BP"                        \n\t"
1272                     YSCALEYUV2RGB(%%REGBP, %5)
1273                     "pcmpeqd %%mm7, %%mm7                   \n\t"
1274                     WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1275                     "pop %%"REG_BP"                         \n\t"
1276                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1277
1278                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1279                     "a" (&c->redDither)
1280                 );
1281             }
1282             return;
1283         case PIX_FMT_BGR24:
1284             __asm__ volatile(
1285                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1286                 "mov        %4, %%"REG_b"               \n\t"
1287                 "push %%"REG_BP"                        \n\t"
1288                 YSCALEYUV2RGB(%%REGBP, %5)
1289                 "pxor    %%mm7, %%mm7                   \n\t"
1290                 WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1291                 "pop %%"REG_BP"                         \n\t"
1292                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1293                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1294                 "a" (&c->redDither)
1295             );
1296             return;
1297         case PIX_FMT_RGB555:
1298             __asm__ volatile(
1299                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1300                 "mov        %4, %%"REG_b"               \n\t"
1301                 "push %%"REG_BP"                        \n\t"
1302                 YSCALEYUV2RGB(%%REGBP, %5)
1303                 "pxor    %%mm7, %%mm7                   \n\t"
1304                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1305 #ifdef DITHER1XBPP
1306                 "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1307                 "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1308                 "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1309 #endif
1310
1311                 WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1312                 "pop %%"REG_BP"                         \n\t"
1313                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1314
1315                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1316                 "a" (&c->redDither)
1317             );
1318             return;
1319         case PIX_FMT_RGB565:
1320             __asm__ volatile(
1321                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1322                 "mov        %4, %%"REG_b"               \n\t"
1323                 "push %%"REG_BP"                        \n\t"
1324                 YSCALEYUV2RGB(%%REGBP, %5)
1325                 "pxor    %%mm7, %%mm7                   \n\t"
1326                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1327 #ifdef DITHER1XBPP
1328                 "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1329                 "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1330                 "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1331 #endif
1332
1333                 WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1334                 "pop %%"REG_BP"                         \n\t"
1335                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1336                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1337                 "a" (&c->redDither)
1338             );
1339             return;
1340         case PIX_FMT_YUYV422:
1341             __asm__ volatile(
1342                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1343                 "mov %4, %%"REG_b"                        \n\t"
1344                 "push %%"REG_BP"                        \n\t"
1345                 YSCALEYUV2PACKED(%%REGBP, %5)
1346                 WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1347                 "pop %%"REG_BP"                         \n\t"
1348                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1349                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1350                 "a" (&c->redDither)
1351             );
1352             return;
1353         default: break;
1354         }
1355     }
1356 #endif //COMPILE_TEMPLATE_MMX
1357     YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C(void,0), YSCALE_YUV_2_GRAY16_2_C, YSCALE_YUV_2_MONO2_C)
1358 }
1359
1360 /**
1361  * YV12 to RGB without scaling or interpolating
1362  */
1363 static inline void RENAME(yuv2packed1)(SwsContext *c, const uint16_t *buf0, const uint16_t *uvbuf0, const uint16_t *uvbuf1,
1364                           const uint16_t *abuf0, uint8_t *dest, int dstW, int uvalpha, enum PixelFormat dstFormat, int flags, int y)
1365 {
1366     const int yalpha1=0;
1367     int i;
1368
1369     const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1370     const int yalpha= 4096; //FIXME ...
1371
1372     if (flags&SWS_FULL_CHR_H_INT) {
1373         c->yuv2packed2(c, buf0, buf0, uvbuf0, uvbuf1, abuf0, abuf0, dest, dstW, 0, uvalpha, y);
1374         return;
1375     }
1376
1377 #if COMPILE_TEMPLATE_MMX
1378     if(!(flags & SWS_BITEXACT)) {
1379         if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1380             switch(dstFormat) {
1381             case PIX_FMT_RGB32:
1382                 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1383                     __asm__ volatile(
1384                         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1385                         "mov        %4, %%"REG_b"               \n\t"
1386                         "push %%"REG_BP"                        \n\t"
1387                         YSCALEYUV2RGB1(%%REGBP, %5)
1388                         YSCALEYUV2RGB1_ALPHA(%%REGBP)
1389                         WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1390                         "pop %%"REG_BP"                         \n\t"
1391                         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1392
1393                         :: "c" (buf0), "d" (abuf0), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1394                         "a" (&c->redDither)
1395                     );
1396                 } else {
1397                     __asm__ volatile(
1398                         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1399                         "mov        %4, %%"REG_b"               \n\t"
1400                         "push %%"REG_BP"                        \n\t"
1401                         YSCALEYUV2RGB1(%%REGBP, %5)
1402                         "pcmpeqd %%mm7, %%mm7                   \n\t"
1403                         WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1404                         "pop %%"REG_BP"                         \n\t"
1405                         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1406
1407                         :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1408                         "a" (&c->redDither)
1409                     );
1410                 }
1411                 return;
1412             case PIX_FMT_BGR24:
1413                 __asm__ volatile(
1414                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1415                     "mov        %4, %%"REG_b"               \n\t"
1416                     "push %%"REG_BP"                        \n\t"
1417                     YSCALEYUV2RGB1(%%REGBP, %5)
1418                     "pxor    %%mm7, %%mm7                   \n\t"
1419                     WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1420                     "pop %%"REG_BP"                         \n\t"
1421                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1422
1423                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1424                     "a" (&c->redDither)
1425                 );
1426                 return;
1427             case PIX_FMT_RGB555:
1428                 __asm__ volatile(
1429                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1430                     "mov        %4, %%"REG_b"               \n\t"
1431                     "push %%"REG_BP"                        \n\t"
1432                     YSCALEYUV2RGB1(%%REGBP, %5)
1433                     "pxor    %%mm7, %%mm7                   \n\t"
1434                     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1435 #ifdef DITHER1XBPP
1436                     "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1437                     "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1438                     "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1439 #endif
1440                     WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1441                     "pop %%"REG_BP"                         \n\t"
1442                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1443
1444                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1445                     "a" (&c->redDither)
1446                 );
1447                 return;
1448             case PIX_FMT_RGB565:
1449                 __asm__ volatile(
1450                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1451                     "mov        %4, %%"REG_b"               \n\t"
1452                     "push %%"REG_BP"                        \n\t"
1453                     YSCALEYUV2RGB1(%%REGBP, %5)
1454                     "pxor    %%mm7, %%mm7                   \n\t"
1455                     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1456 #ifdef DITHER1XBPP
1457                     "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1458                     "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1459                     "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1460 #endif
1461
1462                     WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1463                     "pop %%"REG_BP"                         \n\t"
1464                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1465
1466                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1467                     "a" (&c->redDither)
1468                 );
1469                 return;
1470             case PIX_FMT_YUYV422:
1471                 __asm__ volatile(
1472                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1473                     "mov        %4, %%"REG_b"               \n\t"
1474                     "push %%"REG_BP"                        \n\t"
1475                     YSCALEYUV2PACKED1(%%REGBP, %5)
1476                     WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1477                     "pop %%"REG_BP"                         \n\t"
1478                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1479
1480                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1481                     "a" (&c->redDither)
1482                 );
1483                 return;
1484             }
1485         } else {
1486             switch(dstFormat) {
1487             case PIX_FMT_RGB32:
1488                 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1489                     __asm__ volatile(
1490                         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1491                         "mov        %4, %%"REG_b"               \n\t"
1492                         "push %%"REG_BP"                        \n\t"
1493                         YSCALEYUV2RGB1b(%%REGBP, %5)
1494                         YSCALEYUV2RGB1_ALPHA(%%REGBP)
1495                         WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1496                         "pop %%"REG_BP"                         \n\t"
1497                         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1498
1499                         :: "c" (buf0), "d" (abuf0), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1500                         "a" (&c->redDither)
1501                     );
1502                 } else {
1503                     __asm__ volatile(
1504                         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1505                         "mov        %4, %%"REG_b"               \n\t"
1506                         "push %%"REG_BP"                        \n\t"
1507                         YSCALEYUV2RGB1b(%%REGBP, %5)
1508                         "pcmpeqd %%mm7, %%mm7                   \n\t"
1509                         WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1510                         "pop %%"REG_BP"                         \n\t"
1511                         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1512
1513                         :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1514                         "a" (&c->redDither)
1515                     );
1516                 }
1517                 return;
1518             case PIX_FMT_BGR24:
1519                 __asm__ volatile(
1520                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1521                     "mov        %4, %%"REG_b"               \n\t"
1522                     "push %%"REG_BP"                        \n\t"
1523                     YSCALEYUV2RGB1b(%%REGBP, %5)
1524                     "pxor    %%mm7, %%mm7                   \n\t"
1525                     WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1526                     "pop %%"REG_BP"                         \n\t"
1527                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1528
1529                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1530                     "a" (&c->redDither)
1531                 );
1532                 return;
1533             case PIX_FMT_RGB555:
1534                 __asm__ volatile(
1535                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1536                     "mov        %4, %%"REG_b"               \n\t"
1537                     "push %%"REG_BP"                        \n\t"
1538                     YSCALEYUV2RGB1b(%%REGBP, %5)
1539                     "pxor    %%mm7, %%mm7                   \n\t"
1540                     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1541 #ifdef DITHER1XBPP
1542                     "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1543                     "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1544                     "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1545 #endif
1546                     WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1547                     "pop %%"REG_BP"                         \n\t"
1548                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1549
1550                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1551                     "a" (&c->redDither)
1552                 );
1553                 return;
1554             case PIX_FMT_RGB565:
1555                 __asm__ volatile(
1556                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1557                     "mov        %4, %%"REG_b"               \n\t"
1558                     "push %%"REG_BP"                        \n\t"
1559                     YSCALEYUV2RGB1b(%%REGBP, %5)
1560                     "pxor    %%mm7, %%mm7                   \n\t"
1561                     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1562 #ifdef DITHER1XBPP
1563                     "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1564                     "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1565                     "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1566 #endif
1567
1568                     WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1569                     "pop %%"REG_BP"                         \n\t"
1570                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1571
1572                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1573                     "a" (&c->redDither)
1574                 );
1575                 return;
1576             case PIX_FMT_YUYV422:
1577                 __asm__ volatile(
1578                     "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1579                     "mov        %4, %%"REG_b"               \n\t"
1580                     "push %%"REG_BP"                        \n\t"
1581                     YSCALEYUV2PACKED1b(%%REGBP, %5)
1582                     WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1583                     "pop %%"REG_BP"                         \n\t"
1584                     "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1585
1586                     :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),
1587                     "a" (&c->redDither)
1588                 );
1589                 return;
1590             }
1591         }
1592     }
1593 #endif /* COMPILE_TEMPLATE_MMX */
1594     if (uvalpha < 2048) {
1595         YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C, YSCALE_YUV_2_PACKED1_C(void,0), YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)
1596     } else {
1597         YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C, YSCALE_YUV_2_PACKED1B_C(void,0), YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)
1598     }
1599 }
1600
1601 //FIXME yuy2* can read up to 7 samples too much
1602
1603 static inline void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1604 {
1605 #if COMPILE_TEMPLATE_MMX
1606     __asm__ volatile(
1607         "movq "MANGLE(bm01010101)", %%mm2           \n\t"
1608         "mov                    %0, %%"REG_a"       \n\t"
1609         "1:                                         \n\t"
1610         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1611         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1612         "pand                %%mm2, %%mm0           \n\t"
1613         "pand                %%mm2, %%mm1           \n\t"
1614         "packuswb            %%mm1, %%mm0           \n\t"
1615         "movq                %%mm0, (%2, %%"REG_a") \n\t"
1616         "add                    $8, %%"REG_a"       \n\t"
1617         " js                    1b                  \n\t"
1618         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
1619         : "%"REG_a
1620     );
1621 #else
1622     int i;
1623     for (i=0; i<width; i++)
1624         dst[i]= src[2*i];
1625 #endif
1626 }
1627
1628 static inline void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1629 {
1630 #if COMPILE_TEMPLATE_MMX
1631     __asm__ volatile(
1632         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1633         "mov                    %0, %%"REG_a"       \n\t"
1634         "1:                                         \n\t"
1635         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"
1636         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
1637         "psrlw                  $8, %%mm0           \n\t"
1638         "psrlw                  $8, %%mm1           \n\t"
1639         "packuswb            %%mm1, %%mm0           \n\t"
1640         "movq                %%mm0, %%mm1           \n\t"
1641         "psrlw                  $8, %%mm0           \n\t"
1642         "pand                %%mm4, %%mm1           \n\t"
1643         "packuswb            %%mm0, %%mm0           \n\t"
1644         "packuswb            %%mm1, %%mm1           \n\t"
1645         "movd                %%mm0, (%3, %%"REG_a") \n\t"
1646         "movd                %%mm1, (%2, %%"REG_a") \n\t"
1647         "add                    $4, %%"REG_a"       \n\t"
1648         " js                    1b                  \n\t"
1649         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
1650         : "%"REG_a
1651     );
1652 #else
1653     int i;
1654     for (i=0; i<width; i++) {
1655         dstU[i]= src1[4*i + 1];
1656         dstV[i]= src1[4*i + 3];
1657     }
1658 #endif
1659     assert(src1 == src2);
1660 }
1661
1662 static inline void RENAME(LEToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1663 {
1664 #if COMPILE_TEMPLATE_MMX
1665     __asm__ volatile(
1666         "mov                    %0, %%"REG_a"       \n\t"
1667         "1:                                         \n\t"
1668         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1669         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1670         "movq    (%2, %%"REG_a",2), %%mm2           \n\t"
1671         "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"
1672         "psrlw                  $8, %%mm0           \n\t"
1673         "psrlw                  $8, %%mm1           \n\t"
1674         "psrlw                  $8, %%mm2           \n\t"
1675         "psrlw                  $8, %%mm3           \n\t"
1676         "packuswb            %%mm1, %%mm0           \n\t"
1677         "packuswb            %%mm3, %%mm2           \n\t"
1678         "movq                %%mm0, (%3, %%"REG_a") \n\t"
1679         "movq                %%mm2, (%4, %%"REG_a") \n\t"
1680         "add                    $8, %%"REG_a"       \n\t"
1681         " js                    1b                  \n\t"
1682         : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)
1683         : "%"REG_a
1684     );
1685 #else
1686     int i;
1687     for (i=0; i<width; i++) {
1688         dstU[i]= src1[2*i + 1];
1689         dstV[i]= src2[2*i + 1];
1690     }
1691 #endif
1692 }
1693
1694 /* This is almost identical to the previous, end exists only because
1695  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */
1696 static inline void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1697 {
1698 #if COMPILE_TEMPLATE_MMX
1699     __asm__ volatile(
1700         "mov                  %0, %%"REG_a"         \n\t"
1701         "1:                                         \n\t"
1702         "movq  (%1, %%"REG_a",2), %%mm0             \n\t"
1703         "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"
1704         "psrlw                $8, %%mm0             \n\t"
1705         "psrlw                $8, %%mm1             \n\t"
1706         "packuswb          %%mm1, %%mm0             \n\t"
1707         "movq              %%mm0, (%2, %%"REG_a")   \n\t"
1708         "add                  $8, %%"REG_a"         \n\t"
1709         " js                  1b                    \n\t"
1710         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
1711         : "%"REG_a
1712     );
1713 #else
1714     int i;
1715     for (i=0; i<width; i++)
1716         dst[i]= src[2*i+1];
1717 #endif
1718 }
1719
1720 static inline void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1721 {
1722 #if COMPILE_TEMPLATE_MMX
1723     __asm__ volatile(
1724         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1725         "mov                    %0, %%"REG_a"       \n\t"
1726         "1:                                         \n\t"
1727         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"
1728         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
1729         "pand                %%mm4, %%mm0           \n\t"
1730         "pand                %%mm4, %%mm1           \n\t"
1731         "packuswb            %%mm1, %%mm0           \n\t"
1732         "movq                %%mm0, %%mm1           \n\t"
1733         "psrlw                  $8, %%mm0           \n\t"
1734         "pand                %%mm4, %%mm1           \n\t"
1735         "packuswb            %%mm0, %%mm0           \n\t"
1736         "packuswb            %%mm1, %%mm1           \n\t"
1737         "movd                %%mm0, (%3, %%"REG_a") \n\t"
1738         "movd                %%mm1, (%2, %%"REG_a") \n\t"
1739         "add                    $4, %%"REG_a"       \n\t"
1740         " js                    1b                  \n\t"
1741         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
1742         : "%"REG_a
1743     );
1744 #else
1745     int i;
1746     for (i=0; i<width; i++) {
1747         dstU[i]= src1[4*i + 0];
1748         dstV[i]= src1[4*i + 2];
1749     }
1750 #endif
1751     assert(src1 == src2);
1752 }
1753
1754 static inline void RENAME(BEToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1755 {
1756 #if COMPILE_TEMPLATE_MMX
1757     __asm__ volatile(
1758         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1759         "mov                    %0, %%"REG_a"       \n\t"
1760         "1:                                         \n\t"
1761         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1762         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1763         "movq    (%2, %%"REG_a",2), %%mm2           \n\t"
1764         "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"
1765         "pand                %%mm4, %%mm0           \n\t"
1766         "pand                %%mm4, %%mm1           \n\t"
1767         "pand                %%mm4, %%mm2           \n\t"
1768         "pand                %%mm4, %%mm3           \n\t"
1769         "packuswb            %%mm1, %%mm0           \n\t"
1770         "packuswb            %%mm3, %%mm2           \n\t"
1771         "movq                %%mm0, (%3, %%"REG_a") \n\t"
1772         "movq                %%mm2, (%4, %%"REG_a") \n\t"
1773         "add                    $8, %%"REG_a"       \n\t"
1774         " js                    1b                  \n\t"
1775         : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)
1776         : "%"REG_a
1777     );
1778 #else
1779     int i;
1780     for (i=0; i<width; i++) {
1781         dstU[i]= src1[2*i];
1782         dstV[i]= src2[2*i];
1783     }
1784 #endif
1785 }
1786
1787 static inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,
1788                                     const uint8_t *src, long width)
1789 {
1790 #if COMPILE_TEMPLATE_MMX
1791     __asm__ volatile(
1792         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1793         "mov                    %0, %%"REG_a"       \n\t"
1794         "1:                                         \n\t"
1795         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1796         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1797         "movq                %%mm0, %%mm2           \n\t"
1798         "movq                %%mm1, %%mm3           \n\t"
1799         "pand                %%mm4, %%mm0           \n\t"
1800         "pand                %%mm4, %%mm1           \n\t"
1801         "psrlw                  $8, %%mm2           \n\t"
1802         "psrlw                  $8, %%mm3           \n\t"
1803         "packuswb            %%mm1, %%mm0           \n\t"
1804         "packuswb            %%mm3, %%mm2           \n\t"
1805         "movq                %%mm0, (%2, %%"REG_a") \n\t"
1806         "movq                %%mm2, (%3, %%"REG_a") \n\t"
1807         "add                    $8, %%"REG_a"       \n\t"
1808         " js                    1b                  \n\t"
1809         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)
1810         : "%"REG_a
1811     );
1812 #else
1813     int i;
1814     for (i = 0; i < width; i++) {
1815         dst1[i] = src[2*i+0];
1816         dst2[i] = src[2*i+1];
1817     }
1818 #endif
1819 }
1820
1821 static inline void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,
1822                                     const uint8_t *src1, const uint8_t *src2,
1823                                     long width, uint32_t *unused)
1824 {
1825     RENAME(nvXXtoUV)(dstU, dstV, src1, width);
1826 }
1827
1828 static inline void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,
1829                                     const uint8_t *src1, const uint8_t *src2,
1830                                     long width, uint32_t *unused)
1831 {
1832     RENAME(nvXXtoUV)(dstV, dstU, src1, width);
1833 }
1834
1835 #if COMPILE_TEMPLATE_MMX
1836 static inline void RENAME(bgr24ToY_mmx)(uint8_t *dst, const uint8_t *src, long width, enum PixelFormat srcFormat)
1837 {
1838
1839     if(srcFormat == PIX_FMT_BGR24) {
1840         __asm__ volatile(
1841             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"
1842             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"
1843             :
1844         );
1845     } else {
1846         __asm__ volatile(
1847             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"
1848             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"
1849             :
1850         );
1851     }
1852
1853     __asm__ volatile(
1854         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"
1855         "mov                        %2, %%"REG_a"   \n\t"
1856         "pxor                    %%mm7, %%mm7       \n\t"
1857         "1:                                         \n\t"
1858         PREFETCH"               64(%0)              \n\t"
1859         "movd                     (%0), %%mm0       \n\t"
1860         "movd                    2(%0), %%mm1       \n\t"
1861         "movd                    6(%0), %%mm2       \n\t"
1862         "movd                    8(%0), %%mm3       \n\t"
1863         "add                       $12, %0          \n\t"
1864         "punpcklbw               %%mm7, %%mm0       \n\t"
1865         "punpcklbw               %%mm7, %%mm1       \n\t"
1866         "punpcklbw               %%mm7, %%mm2       \n\t"
1867         "punpcklbw               %%mm7, %%mm3       \n\t"
1868         "pmaddwd                 %%mm5, %%mm0       \n\t"
1869         "pmaddwd                 %%mm6, %%mm1       \n\t"
1870         "pmaddwd                 %%mm5, %%mm2       \n\t"
1871         "pmaddwd                 %%mm6, %%mm3       \n\t"
1872         "paddd                   %%mm1, %%mm0       \n\t"
1873         "paddd                   %%mm3, %%mm2       \n\t"
1874         "paddd                   %%mm4, %%mm0       \n\t"
1875         "paddd                   %%mm4, %%mm2       \n\t"
1876         "psrad                     $15, %%mm0       \n\t"
1877         "psrad                     $15, %%mm2       \n\t"
1878         "packssdw                %%mm2, %%mm0       \n\t"
1879         "packuswb                %%mm0, %%mm0       \n\t"
1880         "movd                %%mm0, (%1, %%"REG_a") \n\t"
1881         "add                        $4, %%"REG_a"   \n\t"
1882         " js                        1b              \n\t"
1883     : "+r" (src)
1884     : "r" (dst+width), "g" ((x86_reg)-width)
1885     : "%"REG_a
1886     );
1887 }
1888
1889 static inline void RENAME(bgr24ToUV_mmx)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, long width, enum PixelFormat srcFormat)
1890 {
1891     __asm__ volatile(
1892         "movq                    24+%4, %%mm6       \n\t"
1893         "mov                        %3, %%"REG_a"   \n\t"
1894         "pxor                    %%mm7, %%mm7       \n\t"
1895         "1:                                         \n\t"
1896         PREFETCH"               64(%0)              \n\t"
1897         "movd                     (%0), %%mm0       \n\t"
1898         "movd                    2(%0), %%mm1       \n\t"
1899         "punpcklbw               %%mm7, %%mm0       \n\t"
1900         "punpcklbw               %%mm7, %%mm1       \n\t"
1901         "movq                    %%mm0, %%mm2       \n\t"
1902         "movq                    %%mm1, %%mm3       \n\t"
1903         "pmaddwd                    %4, %%mm0       \n\t"
1904         "pmaddwd                  8+%4, %%mm1       \n\t"
1905         "pmaddwd                 16+%4, %%mm2       \n\t"
1906         "pmaddwd                 %%mm6, %%mm3       \n\t"
1907         "paddd                   %%mm1, %%mm0       \n\t"
1908         "paddd                   %%mm3, %%mm2       \n\t"
1909
1910         "movd                    6(%0), %%mm1       \n\t"
1911         "movd                    8(%0), %%mm3       \n\t"
1912         "add                       $12, %0          \n\t"
1913         "punpcklbw               %%mm7, %%mm1       \n\t"
1914         "punpcklbw               %%mm7, %%mm3       \n\t"
1915         "movq                    %%mm1, %%mm4       \n\t"
1916         "movq                    %%mm3, %%mm5       \n\t"
1917         "pmaddwd                    %4, %%mm1       \n\t"
1918         "pmaddwd                  8+%4, %%mm3       \n\t"
1919         "pmaddwd                 16+%4, %%mm4       \n\t"
1920         "pmaddwd                 %%mm6, %%mm5       \n\t"
1921         "paddd                   %%mm3, %%mm1       \n\t"
1922         "paddd                   %%mm5, %%mm4       \n\t"
1923
1924         "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3       \n\t"
1925         "paddd                   %%mm3, %%mm0       \n\t"
1926         "paddd                   %%mm3, %%mm2       \n\t"
1927         "paddd                   %%mm3, %%mm1       \n\t"
1928         "paddd                   %%mm3, %%mm4       \n\t"
1929         "psrad                     $15, %%mm0       \n\t"
1930         "psrad                     $15, %%mm2       \n\t"
1931         "psrad                     $15, %%mm1       \n\t"
1932         "psrad                     $15, %%mm4       \n\t"
1933         "packssdw                %%mm1, %%mm0       \n\t"
1934         "packssdw                %%mm4, %%mm2       \n\t"
1935         "packuswb                %%mm0, %%mm0       \n\t"
1936         "packuswb                %%mm2, %%mm2       \n\t"
1937         "movd                %%mm0, (%1, %%"REG_a") \n\t"
1938         "movd                %%mm2, (%2, %%"REG_a") \n\t"
1939         "add                        $4, %%"REG_a"   \n\t"
1940         " js                        1b              \n\t"
1941     : "+r" (src)
1942     : "r" (dstU+width), "r" (dstV+width), "g" ((x86_reg)-width), "m"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24][0])
1943     : "%"REG_a
1944     );
1945 }
1946 #endif
1947
1948 static inline void RENAME(bgr24ToY)(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1949 {
1950 #if COMPILE_TEMPLATE_MMX
1951     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);
1952 #else
1953     int i;
1954     for (i=0; i<width; i++) {
1955         int b= src[i*3+0];
1956         int g= src[i*3+1];
1957         int r= src[i*3+2];
1958
1959         dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1960     }
1961 #endif /* COMPILE_TEMPLATE_MMX */
1962 }
1963
1964 static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1965 {
1966 #if COMPILE_TEMPLATE_MMX
1967     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);
1968 #else
1969     int i;
1970     for (i=0; i<width; i++) {
1971         int b= src1[3*i + 0];
1972         int g= src1[3*i + 1];
1973         int r= src1[3*i + 2];
1974
1975         dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
1976         dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
1977     }
1978 #endif /* COMPILE_TEMPLATE_MMX */
1979     assert(src1 == src2);
1980 }
1981
1982 static inline void RENAME(bgr24ToUV_half)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
1983 {
1984     int i;
1985     for (i=0; i<width; i++) {
1986         int b= src1[6*i + 0] + src1[6*i + 3];
1987         int g= src1[6*i + 1] + src1[6*i + 4];
1988         int r= src1[6*i + 2] + src1[6*i + 5];
1989
1990         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);
1991         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);
1992     }
1993     assert(src1 == src2);
1994 }
1995
1996 static inline void RENAME(rgb24ToY)(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1997 {
1998 #if COMPILE_TEMPLATE_MMX
1999     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);
2000 #else
2001     int i;
2002     for (i=0; i<width; i++) {
2003         int r= src[i*3+0];
2004         int g= src[i*3+1];
2005         int b= src[i*3+2];
2006
2007         dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2008     }
2009 #endif
2010 }
2011
2012 static inline void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
2013 {
2014 #if COMPILE_TEMPLATE_MMX
2015     assert(src1==src2);
2016     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);
2017 #else
2018     int i;
2019     assert(src1==src2);
2020     for (i=0; i<width; i++) {
2021         int r= src1[3*i + 0];
2022         int g= src1[3*i + 1];
2023         int b= src1[3*i + 2];
2024
2025         dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
2026         dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;
2027     }
2028 #endif
2029 }
2030
2031 static inline void RENAME(rgb24ToUV_half)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, long width, uint32_t *unused)
2032 {
2033     int i;
2034     assert(src1==src2);
2035     for (i=0; i<width; i++) {
2036         int r= src1[6*i + 0] + src1[6*i + 3];
2037         int g= src1[6*i + 1] + src1[6*i + 4];
2038         int b= src1[6*i + 2] + src1[6*i + 5];
2039
2040         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);
2041         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);
2042     }
2043 }
2044
2045
2046 // bilinear / bicubic scaling
2047 static inline void RENAME(hScale)(int16_t *dst, int dstW, const uint8_t *src, int srcW, int xInc,
2048                                   const int16_t *filter, const int16_t *filterPos, long filterSize)
2049 {
2050 #if COMPILE_TEMPLATE_MMX
2051     assert(filterSize % 4 == 0 && filterSize>0);
2052     if (filterSize==4) { // Always true for upscaling, sometimes for down, too.
2053         x86_reg counter= -2*dstW;
2054         filter-= counter*2;
2055         filterPos-= counter/2;
2056         dst-= counter/2;
2057         __asm__ volatile(
2058 #if defined(PIC)
2059             "push            %%"REG_b"              \n\t"
2060 #endif
2061             "pxor                %%mm7, %%mm7       \n\t"
2062             "push           %%"REG_BP"              \n\t" // we use 7 regs here ...
2063             "mov             %%"REG_a", %%"REG_BP"  \n\t"
2064             ASMALIGN(4)
2065             "1:                                     \n\t"
2066             "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"
2067             "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"
2068             "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"
2069             "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"
2070             "movd      (%3, %%"REG_a"), %%mm0       \n\t"
2071             "movd      (%3, %%"REG_b"), %%mm2       \n\t"
2072             "punpcklbw           %%mm7, %%mm0       \n\t"
2073             "punpcklbw           %%mm7, %%mm2       \n\t"
2074             "pmaddwd             %%mm1, %%mm0       \n\t"
2075             "pmaddwd             %%mm2, %%mm3       \n\t"
2076             "movq                %%mm0, %%mm4       \n\t"
2077             "punpckldq           %%mm3, %%mm0       \n\t"
2078             "punpckhdq           %%mm3, %%mm4       \n\t"
2079             "paddd               %%mm4, %%mm0       \n\t"
2080             "psrad                  $7, %%mm0       \n\t"
2081             "packssdw            %%mm0, %%mm0       \n\t"
2082             "movd                %%mm0, (%4, %%"REG_BP")    \n\t"
2083             "add                    $4, %%"REG_BP"  \n\t"
2084             " jnc                   1b              \n\t"
2085
2086             "pop            %%"REG_BP"              \n\t"
2087 #if defined(PIC)
2088             "pop             %%"REG_b"              \n\t"
2089 #endif
2090             : "+a" (counter)
2091             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
2092 #if !defined(PIC)
2093             : "%"REG_b
2094 #endif
2095         );
2096     } else if (filterSize==8) {
2097         x86_reg counter= -2*dstW;
2098         filter-= counter*4;
2099         filterPos-= counter/2;
2100         dst-= counter/2;
2101         __asm__ volatile(
2102 #if defined(PIC)
2103             "push             %%"REG_b"             \n\t"
2104 #endif
2105             "pxor                 %%mm7, %%mm7      \n\t"
2106             "push            %%"REG_BP"             \n\t" // we use 7 regs here ...
2107             "mov              %%"REG_a", %%"REG_BP" \n\t"
2108             ASMALIGN(4)
2109             "1:                                     \n\t"
2110             "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"
2111             "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"
2112             "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"
2113             "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"
2114             "movd       (%3, %%"REG_a"), %%mm0      \n\t"
2115             "movd       (%3, %%"REG_b"), %%mm2      \n\t"
2116             "punpcklbw            %%mm7, %%mm0      \n\t"
2117             "punpcklbw            %%mm7, %%mm2      \n\t"
2118             "pmaddwd              %%mm1, %%mm0      \n\t"
2119             "pmaddwd              %%mm2, %%mm3      \n\t"
2120
2121             "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"
2122             "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"
2123             "movd      4(%3, %%"REG_a"), %%mm4      \n\t"
2124             "movd      4(%3, %%"REG_b"), %%mm2      \n\t"
2125             "punpcklbw            %%mm7, %%mm4      \n\t"
2126             "punpcklbw            %%mm7, %%mm2      \n\t"
2127             "pmaddwd              %%mm1, %%mm4      \n\t"
2128             "pmaddwd              %%mm2, %%mm5      \n\t"
2129             "paddd                %%mm4, %%mm0      \n\t"
2130             "paddd                %%mm5, %%mm3      \n\t"
2131             "movq                 %%mm0, %%mm4      \n\t"
2132             "punpckldq            %%mm3, %%mm0      \n\t"
2133             "punpckhdq            %%mm3, %%mm4      \n\t"
2134             "paddd                %%mm4, %%mm0      \n\t"
2135             "psrad                   $7, %%mm0      \n\t"
2136             "packssdw             %%mm0, %%mm0      \n\t"
2137             "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"
2138             "add                     $4, %%"REG_BP" \n\t"
2139             " jnc                    1b             \n\t"
2140
2141             "pop             %%"REG_BP"             \n\t"
2142 #if defined(PIC)
2143             "pop              %%"REG_b"             \n\t"
2144 #endif
2145             : "+a" (counter)
2146             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
2147 #if !defined(PIC)
2148             : "%"REG_b
2149 #endif
2150         );
2151     } else {
2152         uint8_t *offset = src+filterSize;
2153         x86_reg counter= -2*dstW;
2154         //filter-= counter*filterSize/2;
2155         filterPos-= counter/2;
2156         dst-= counter/2;
2157         __asm__ volatile(
2158             "pxor                  %%mm7, %%mm7     \n\t"
2159             ASMALIGN(4)
2160             "1:                                     \n\t"
2161             "mov                      %2, %%"REG_c" \n\t"
2162             "movzwl      (%%"REG_c", %0), %%eax     \n\t"
2163             "movzwl     2(%%"REG_c", %0), %%edx     \n\t"
2164             "mov                      %5, %%"REG_c" \n\t"
2165             "pxor                  %%mm4, %%mm4     \n\t"
2166             "pxor                  %%mm5, %%mm5     \n\t"
2167             "2:                                     \n\t"
2168             "movq                   (%1), %%mm1     \n\t"
2169             "movq               (%1, %6), %%mm3     \n\t"
2170             "movd (%%"REG_c", %%"REG_a"), %%mm0     \n\t"
2171             "movd (%%"REG_c", %%"REG_d"), %%mm2     \n\t"
2172             "punpcklbw             %%mm7, %%mm0     \n\t"
2173             "punpcklbw             %%mm7, %%mm2     \n\t"
2174             "pmaddwd               %%mm1, %%mm0     \n\t"
2175             "pmaddwd               %%mm2, %%mm3     \n\t"
2176             "paddd                 %%mm3, %%mm5     \n\t"
2177             "paddd                 %%mm0, %%mm4     \n\t"
2178             "add                      $8, %1        \n\t"
2179             "add                      $4, %%"REG_c" \n\t"
2180             "cmp                      %4, %%"REG_c" \n\t"
2181             " jb                      2b            \n\t"
2182             "add                      %6, %1        \n\t"
2183             "movq                  %%mm4, %%mm0     \n\t"
2184             "punpckldq             %%mm5, %%mm4     \n\t"
2185             "punpckhdq             %%mm5, %%mm0     \n\t"
2186             "paddd                 %%mm0, %%mm4     \n\t"
2187             "psrad                    $7, %%mm4     \n\t"
2188             "packssdw              %%mm4, %%mm4     \n\t"
2189             "mov                      %3, %%"REG_a" \n\t"
2190             "movd                  %%mm4, (%%"REG_a", %0)   \n\t"
2191             "add                      $4, %0        \n\t"
2192             " jnc                     1b            \n\t"
2193
2194             : "+r" (counter), "+r" (filter)
2195             : "m" (filterPos), "m" (dst), "m"(offset),
2196             "m" (src), "r" ((x86_reg)filterSize*2)
2197             : "%"REG_a, "%"REG_c, "%"REG_d
2198         );
2199     }
2200 #else
2201 #if COMPILE_TEMPLATE_ALTIVEC
2202     hScale_altivec_real(dst, dstW, src, srcW, xInc, filter, filterPos, filterSize);
2203 #else
2204     int i;
2205     for (i=0; i<dstW; i++) {
2206         int j;
2207         int srcPos= filterPos[i];
2208         int val=0;
2209         //printf("filterPos: %d\n", filterPos[i]);
2210         for (j=0; j<filterSize; j++) {
2211             //printf("filter: %d, src: %d\n", filter[i], src[srcPos + j]);
2212             val += ((int)src[srcPos + j])*filter[filterSize*i + j];
2213         }
2214         //filter += hFilterSize;
2215         dst[i] = FFMIN(val>>7, (1<<15)-1); // the cubic equation does overflow ...
2216         //dst[i] = val>>7;
2217     }
2218 #endif /* COMPILE_ALTIVEC */
2219 #endif /* COMPILE_MMX */
2220 }
2221
2222 #define FAST_BILINEAR_X86 \
2223     "subl    %%edi, %%esi    \n\t" /*  src[xx+1] - src[xx] */                   \
2224     "imull   %%ecx, %%esi    \n\t" /* (src[xx+1] - src[xx])*xalpha */           \
2225     "shll      $16, %%edi    \n\t"                                              \
2226     "addl    %%edi, %%esi    \n\t" /* src[xx+1]*xalpha + src[xx]*(1-xalpha) */  \
2227     "mov        %1, %%"REG_D"\n\t"                                              \
2228     "shrl       $9, %%esi    \n\t"                                              \
2229
2230 static inline void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
2231                                         int dstWidth, const uint8_t *src, int srcW,
2232                                         int xInc)
2233 {
2234     int i;
2235     unsigned int xpos=0;
2236     for (i=0;i<dstWidth;i++) {
2237         register unsigned int xx=xpos>>16;
2238         register unsigned int xalpha=(xpos&0xFFFF)>>9;
2239         dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
2240         xpos+=xInc;
2241     }
2242 }
2243
2244       // *** horizontal scale Y line to temp buffer
2245 static inline void RENAME(hyscale)(SwsContext *c, uint16_t *dst, long dstWidth, const uint8_t *src, int srcW, int xInc,
2246                                    int flags, const int16_t *hLumFilter,
2247                                    const int16_t *hLumFilterPos, int hLumFilterSize,
2248                                    enum PixelFormat srcFormat, uint8_t *formatConvBuffer,
2249                                    uint32_t *pal, int isAlpha)
2250 {
2251     int32_t av_unused *mmx2FilterPos = c->lumMmx2FilterPos;
2252     int16_t av_unused *mmx2Filter    = c->lumMmx2Filter;
2253     int     av_unused canMMX2BeUsed  = c->canMMX2BeUsed;
2254     void    av_unused *mmx2FilterCode= c->lumMmx2FilterCode;
2255     void (*internal_func)(uint8_t *, const uint8_t *, long, uint32_t *) = isAlpha ? c->hascale_internal : c->hyscale_internal;
2256
2257     if (isAlpha) {
2258         if (srcFormat == PIX_FMT_RGB32   || srcFormat == PIX_FMT_BGR32  )
2259             src += 3;
2260     } else {
2261         if (srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2262             src += ALT32_CORR;
2263     }
2264
2265     if (srcFormat == PIX_FMT_RGB48LE)
2266         src++;
2267
2268     if (internal_func) {
2269         internal_func(formatConvBuffer, src, srcW, pal);
2270         src= formatConvBuffer;
2271     }
2272
2273 #if COMPILE_TEMPLATE_MMX
2274     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).
2275     if (!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed))
2276 #else
2277     if (!(flags&SWS_FAST_BILINEAR))
2278 #endif
2279     {
2280         c->hScale(dst, dstWidth, src, srcW, xInc, hLumFilter, hLumFilterPos, hLumFilterSize);
2281     } else { // fast bilinear upscale / crap downscale
2282 #if ARCH_X86 && CONFIG_GPL
2283 #if COMPILE_TEMPLATE_MMX2
2284         int i;
2285 #if defined(PIC)
2286         DECLARE_ALIGNED(8, uint64_t, ebxsave);
2287 #endif
2288         if (canMMX2BeUsed) {
2289             __asm__ volatile(
2290 #if defined(PIC)
2291                 "mov               %%"REG_b", %5        \n\t"
2292 #endif
2293                 "pxor                  %%mm7, %%mm7     \n\t"
2294                 "mov                      %0, %%"REG_c" \n\t"
2295                 "mov                      %1, %%"REG_D" \n\t"
2296                 "mov                      %2, %%"REG_d" \n\t"
2297                 "mov                      %3, %%"REG_b" \n\t"
2298                 "xor               %%"REG_a", %%"REG_a" \n\t" // i
2299                 PREFETCH"        (%%"REG_c")            \n\t"
2300                 PREFETCH"      32(%%"REG_c")            \n\t"
2301                 PREFETCH"      64(%%"REG_c")            \n\t"
2302
2303 #if ARCH_X86_64
2304
2305 #define CALL_MMX2_FILTER_CODE \
2306                 "movl            (%%"REG_b"), %%esi     \n\t"\
2307                 "call                    *%4            \n\t"\
2308                 "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\
2309                 "add               %%"REG_S", %%"REG_c" \n\t"\
2310                 "add               %%"REG_a", %%"REG_D" \n\t"\
2311                 "xor               %%"REG_a", %%"REG_a" \n\t"\
2312
2313 #else
2314
2315 #define CALL_MMX2_FILTER_CODE \
2316                 "movl (%%"REG_b"), %%esi        \n\t"\
2317                 "call         *%4                       \n\t"\
2318                 "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
2319                 "add               %%"REG_a", %%"REG_D" \n\t"\
2320                 "xor               %%"REG_a", %%"REG_a" \n\t"\
2321
2322 #endif /* ARCH_X86_64 */
2323
2324                 CALL_MMX2_FILTER_CODE
2325                 CALL_MMX2_FILTER_CODE
2326                 CALL_MMX2_FILTER_CODE
2327                 CALL_MMX2_FILTER_CODE
2328                 CALL_MMX2_FILTER_CODE
2329                 CALL_MMX2_FILTER_CODE
2330                 CALL_MMX2_FILTER_CODE
2331                 CALL_MMX2_FILTER_CODE
2332
2333 #if defined(PIC)
2334                 "mov                      %5, %%"REG_b" \n\t"
2335 #endif
2336                 :: "m" (src), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),
2337                 "m" (mmx2FilterCode)
2338 #if defined(PIC)
2339                 ,"m" (ebxsave)
2340 #endif
2341                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
2342 #if !defined(PIC)
2343                 ,"%"REG_b
2344 #endif
2345             );
2346             for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) dst[i] = src[srcW-1]*128;
2347         } else {
2348 #endif /* COMPILE_TEMPLATE_MMX2 */
2349         x86_reg xInc_shr16 = xInc >> 16;
2350         uint16_t xInc_mask = xInc & 0xffff;
2351         //NO MMX just normal asm ...
2352         __asm__ volatile(
2353             "xor %%"REG_a", %%"REG_a"            \n\t" // i
2354             "xor %%"REG_d", %%"REG_d"            \n\t" // xx
2355             "xorl    %%ecx, %%ecx                \n\t" // xalpha
2356             ASMALIGN(4)
2357             "1:                                  \n\t"
2358             "movzbl    (%0, %%"REG_d"), %%edi    \n\t" //src[xx]
2359             "movzbl   1(%0, %%"REG_d"), %%esi    \n\t" //src[xx+1]
2360             FAST_BILINEAR_X86
2361             "movw     %%si, (%%"REG_D", %%"REG_a", 2)   \n\t"
2362             "addw       %4, %%cx                 \n\t" //xalpha += xInc&0xFFFF
2363             "adc        %3, %%"REG_d"            \n\t" //xx+= xInc>>16 + carry
2364
2365             "movzbl    (%0, %%"REG_d"), %%edi    \n\t" //src[xx]
2366             "movzbl   1(%0, %%"REG_d"), %%esi    \n\t" //src[xx+1]
2367             FAST_BILINEAR_X86
2368             "movw     %%si, 2(%%"REG_D", %%"REG_a", 2)  \n\t"
2369             "addw       %4, %%cx                 \n\t" //xalpha += xInc&0xFFFF
2370             "adc        %3, %%"REG_d"            \n\t" //xx+= xInc>>16 + carry
2371
2372
2373             "add        $2, %%"REG_a"            \n\t"
2374             "cmp        %2, %%"REG_a"            \n\t"
2375             " jb        1b                       \n\t"
2376
2377
2378             :: "r" (src), "m" (dst), "m" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask)
2379             : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi"
2380         );
2381 #if COMPILE_TEMPLATE_MMX2
2382         } //if MMX2 can't be used
2383 #endif
2384 #else
2385         c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc);
2386 #endif /* ARCH_X86 */
2387     }
2388
2389     if(!isAlpha && c->srcRange != c->dstRange && !(isRGB(c->dstFormat) || isBGR(c->dstFormat))) {
2390         int i;
2391         //FIXME all pal and rgb srcFormats could do this convertion as well
2392         //FIXME all scalers more complex than bilinear could do half of this transform
2393         if(c->srcRange) {
2394             for (i=0; i<dstWidth; i++)
2395                 dst[i]= (dst[i]*14071 + 33561947)>>14;
2396         } else {
2397             for (i=0; i<dstWidth; i++)
2398                 dst[i]= (FFMIN(dst[i],30189)*19077 - 39057361)>>14;
2399         }
2400     }
2401 }
2402
2403 static inline void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst,
2404                                         int dstWidth, const uint8_t *src1,
2405                                         const uint8_t *src2, int srcW, int xInc)
2406 {
2407     int i;
2408     unsigned int xpos=0;
2409     for (i=0;i<dstWidth;i++) {
2410         register unsigned int xx=xpos>>16;
2411         register unsigned int xalpha=(xpos&0xFFFF)>>9;
2412         dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);
2413         dst[i+VOFW]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
2414         /* slower
2415         dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha;
2416         dst[i+VOFW]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha;
2417         */
2418         xpos+=xInc;
2419     }
2420 }
2421
2422 inline static void RENAME(hcscale)(SwsContext *c, uint16_t *dst, long dstWidth, const uint8_t *src1, const uint8_t *src2,
2423                                    int srcW, int xInc, int flags, const int16_t *hChrFilter,
2424                                    const int16_t *hChrFilterPos, int hChrFilterSize,
2425                                    enum PixelFormat srcFormat, uint8_t *formatConvBuffer,
2426                                    uint32_t *pal)
2427 {
2428     int32_t av_unused *mmx2FilterPos = c->chrMmx2FilterPos;
2429     int16_t av_unused *mmx2Filter    = c->chrMmx2Filter;
2430     int     av_unused canMMX2BeUsed  = c->canMMX2BeUsed;
2431     void    av_unused *mmx2FilterCode= c->chrMmx2FilterCode;
2432
2433     if (isGray(srcFormat) || srcFormat==PIX_FMT_MONOBLACK || srcFormat==PIX_FMT_MONOWHITE)
2434         return;
2435
2436     if (srcFormat==PIX_FMT_RGB32_1 || srcFormat==PIX_FMT_BGR32_1) {
2437         src1 += ALT32_CORR;
2438         src2 += ALT32_CORR;
2439     }
2440
2441     if (srcFormat==PIX_FMT_RGB48LE) {
2442         src1++;
2443         src2++;
2444     }
2445
2446     if (c->hcscale_internal) {
2447         c->hcscale_internal(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW, pal);
2448         src1= formatConvBuffer;
2449         src2= formatConvBuffer+VOFW;
2450     }
2451
2452 #if COMPILE_TEMPLATE_MMX
2453     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).
2454     if (!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed))
2455 #else
2456     if (!(flags&SWS_FAST_BILINEAR))
2457 #endif
2458     {
2459         c->hScale(dst     , dstWidth, src1, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);
2460         c->hScale(dst+VOFW, dstWidth, src2, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);
2461     } else { // fast bilinear upscale / crap downscale
2462 #if ARCH_X86 && CONFIG_GPL
2463 #if COMPILE_TEMPLATE_MMX2
2464         int i;
2465 #if defined(PIC)
2466         DECLARE_ALIGNED(8, uint64_t, ebxsave);
2467 #endif
2468         if (canMMX2BeUsed) {
2469             __asm__ volatile(
2470 #if defined(PIC)
2471                 "mov          %%"REG_b", %6         \n\t"
2472 #endif
2473                 "pxor             %%mm7, %%mm7      \n\t"
2474                 "mov                 %0, %%"REG_c"  \n\t"
2475                 "mov                 %1, %%"REG_D"  \n\t"
2476                 "mov                 %2, %%"REG_d"  \n\t"
2477                 "mov                 %3, %%"REG_b"  \n\t"
2478                 "xor          %%"REG_a", %%"REG_a"  \n\t" // i
2479                 PREFETCH"   (%%"REG_c")             \n\t"
2480                 PREFETCH" 32(%%"REG_c")             \n\t"
2481                 PREFETCH" 64(%%"REG_c")             \n\t"
2482
2483                 CALL_MMX2_FILTER_CODE
2484                 CALL_MMX2_FILTER_CODE
2485                 CALL_MMX2_FILTER_CODE
2486                 CALL_MMX2_FILTER_CODE
2487                 "xor          %%"REG_a", %%"REG_a"  \n\t" // i
2488                 "mov                 %5, %%"REG_c"  \n\t" // src
2489                 "mov                 %1, %%"REG_D"  \n\t" // buf1
2490                 "add              $"AV_STRINGIFY(VOF)", %%"REG_D"  \n\t"
2491                 PREFETCH"   (%%"REG_c")             \n\t"
2492                 PREFETCH" 32(%%"REG_c")             \n\t"
2493                 PREFETCH" 64(%%"REG_c")             \n\t"
2494
2495                 CALL_MMX2_FILTER_CODE
2496                 CALL_MMX2_FILTER_CODE
2497                 CALL_MMX2_FILTER_CODE
2498                 CALL_MMX2_FILTER_CODE
2499
2500 #if defined(PIC)
2501                 "mov %6, %%"REG_b"    \n\t"
2502 #endif
2503                 :: "m" (src1), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),
2504                 "m" (mmx2FilterCode), "m" (src2)
2505 #if defined(PIC)
2506                 ,"m" (ebxsave)
2507 #endif
2508                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
2509 #if !defined(PIC)
2510                 ,"%"REG_b
2511 #endif
2512             );
2513             for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
2514                 //printf("%d %d %d\n", dstWidth, i, srcW);
2515                 dst[i] = src1[srcW-1]*128;
2516                 dst[i+VOFW] = src2[srcW-1]*128;
2517             }
2518         } else {
2519 #endif /* COMPILE_TEMPLATE_MMX2 */
2520             x86_reg xInc_shr16 = (x86_reg) (xInc >> 16);
2521             uint16_t xInc_mask = xInc & 0xffff;
2522             __asm__ volatile(
2523                 "xor %%"REG_a", %%"REG_a"               \n\t" // i
2524                 "xor %%"REG_d", %%"REG_d"               \n\t" // xx
2525                 "xorl    %%ecx, %%ecx                   \n\t" // xalpha
2526                 ASMALIGN(4)
2527                 "1:                                     \n\t"
2528                 "mov        %0, %%"REG_S"               \n\t"
2529                 "movzbl  (%%"REG_S", %%"REG_d"), %%edi  \n\t" //src[xx]
2530                 "movzbl 1(%%"REG_S", %%"REG_d"), %%esi  \n\t" //src[xx+1]
2531                 FAST_BILINEAR_X86
2532                 "movw     %%si, (%%"REG_D", %%"REG_a", 2)   \n\t"
2533
2534                 "movzbl    (%5, %%"REG_d"), %%edi       \n\t" //src[xx]
2535                 "movzbl   1(%5, %%"REG_d"), %%esi       \n\t" //src[xx+1]
2536                 FAST_BILINEAR_X86
2537                 "movw     %%si, "AV_STRINGIFY(VOF)"(%%"REG_D", %%"REG_a", 2)   \n\t"
2538
2539                 "addw       %4, %%cx                    \n\t" //xalpha += xInc&0xFFFF
2540                 "adc        %3, %%"REG_d"               \n\t" //xx+= xInc>>16 + carry
2541                 "add        $1, %%"REG_a"               \n\t"
2542                 "cmp        %2, %%"REG_a"               \n\t"
2543                 " jb        1b                          \n\t"
2544
2545 /* GCC 3.3 makes MPlayer crash on IA-32 machines when using "g" operand here,
2546    which is needed to support GCC 4.0. */
2547 #if ARCH_X86_64 && AV_GCC_VERSION_AT_LEAST(3,4)
2548                 :: "m" (src1), "m" (dst), "g" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask),
2549 #else
2550                 :: "m" (src1), "m" (dst), "m" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask),
2551 #endif
2552                 "r" (src2)
2553                 : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi"
2554             );
2555 #if COMPILE_TEMPLATE_MMX2
2556         } //if MMX2 can't be used
2557 #endif
2558 #else
2559         c->hcscale_fast(c, dst, dstWidth, src1, src2, srcW, xInc);
2560 #endif /* ARCH_X86 */
2561     }
2562     if(c->srcRange != c->dstRange && !(isRGB(c->dstFormat) || isBGR(c->dstFormat))) {
2563         int i;
2564         //FIXME all pal and rgb srcFormats could do this convertion as well
2565         //FIXME all scalers more complex than bilinear could do half of this transform
2566         if(c->srcRange) {
2567             for (i=0; i<dstWidth; i++) {
2568                 dst[i     ]= (dst[i     ]*1799 + 4081085)>>11; //1469
2569                 dst[i+VOFW]= (dst[i+VOFW]*1799 + 4081085)>>11; //1469
2570             }
2571         } else {
2572             for (i=0; i<dstWidth; i++) {
2573                 dst[i     ]= (FFMIN(dst[i     ],30775)*4663 - 9289992)>>12; //-264
2574                 dst[i+VOFW]= (FFMIN(dst[i+VOFW],30775)*4663 - 9289992)>>12; //-264
2575             }
2576         }
2577     }
2578 }
2579
2580 #define DEBUG_SWSCALE_BUFFERS 0
2581 #define DEBUG_BUFFERS(...) if (DEBUG_SWSCALE_BUFFERS) av_log(c, AV_LOG_DEBUG, __VA_ARGS__)
2582
2583 static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2584                            int srcSliceH, uint8_t* dst[], int dstStride[])
2585 {
2586     /* load a few things into local vars to make the code more readable? and faster */
2587     const int srcW= c->srcW;
2588     const int dstW= c->dstW;
2589     const int dstH= c->dstH;
2590     const int chrDstW= c->chrDstW;
2591     const int chrSrcW= c->chrSrcW;
2592     const int lumXInc= c->lumXInc;
2593     const int chrXInc= c->chrXInc;
2594     const enum PixelFormat dstFormat= c->dstFormat;
2595     const enum PixelFormat srcFormat= c->srcFormat;
2596     const int flags= c->flags;
2597     int16_t *vLumFilterPos= c->vLumFilterPos;
2598     int16_t *vChrFilterPos= c->vChrFilterPos;
2599     int16_t *hLumFilterPos= c->hLumFilterPos;
2600     int16_t *hChrFilterPos= c->hChrFilterPos;
2601     int16_t *vLumFilter= c->vLumFilter;
2602     int16_t *vChrFilter= c->vChrFilter;
2603     int16_t *hLumFilter= c->hLumFilter;
2604     int16_t *hChrFilter= c->hChrFilter;
2605     int32_t *lumMmxFilter= c->lumMmxFilter;
2606     int32_t *chrMmxFilter= c->chrMmxFilter;
2607     int32_t *alpMmxFilter= c->alpMmxFilter;
2608     const int vLumFilterSize= c->vLumFilterSize;
2609     const int vChrFilterSize= c->vChrFilterSize;
2610     const int hLumFilterSize= c->hLumFilterSize;
2611     const int hChrFilterSize= c->hChrFilterSize;
2612     int16_t **lumPixBuf= c->lumPixBuf;
2613     int16_t **chrPixBuf= c->chrPixBuf;
2614     int16_t **alpPixBuf= c->alpPixBuf;
2615     const int vLumBufSize= c->vLumBufSize;
2616     const int vChrBufSize= c->vChrBufSize;
2617     uint8_t *formatConvBuffer= c->formatConvBuffer;
2618     const int chrSrcSliceY= srcSliceY >> c->chrSrcVSubSample;
2619     const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample);
2620     int lastDstY;
2621     uint32_t *pal=c->pal_yuv;
2622
2623     /* vars which will change and which we need to store back in the context */
2624     int dstY= c->dstY;
2625     int lumBufIndex= c->lumBufIndex;
2626     int chrBufIndex= c->chrBufIndex;
2627     int lastInLumBuf= c->lastInLumBuf;
2628     int lastInChrBuf= c->lastInChrBuf;
2629
2630     if (isPacked(c->srcFormat)) {
2631         src[0]=
2632         src[1]=
2633         src[2]=
2634         src[3]= src[0];
2635         srcStride[0]=
2636         srcStride[1]=
2637         srcStride[2]=
2638         srcStride[3]= srcStride[0];
2639     }
2640     srcStride[1]<<= c->vChrDrop;
2641     srcStride[2]<<= c->vChrDrop;
2642
2643     DEBUG_BUFFERS("swScale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n",
2644                   src[0], srcStride[0], src[1], srcStride[1], src[2], srcStride[2], src[3], srcStride[3],
2645                   dst[0], dstStride[0], dst[1], dstStride[1], dst[2], dstStride[2], dst[3], dstStride[3]);
2646     DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n",
2647                    srcSliceY,    srcSliceH,    dstY,    dstH);
2648     DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n",
2649                    vLumFilterSize,    vLumBufSize,    vChrFilterSize,    vChrBufSize);
2650
2651     if (dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0 || dstStride[3]%8 != 0) {
2652         static int warnedAlready=0; //FIXME move this into the context perhaps
2653         if (flags & SWS_PRINT_INFO && !warnedAlready) {
2654             av_log(c, AV_LOG_WARNING, "Warning: dstStride is not aligned!\n"
2655                    "         ->cannot do aligned memory accesses anymore\n");
2656             warnedAlready=1;
2657         }
2658     }
2659
2660     /* Note the user might start scaling the picture in the middle so this
2661        will not get executed. This is not really intended but works
2662        currently, so people might do it. */
2663     if (srcSliceY ==0) {
2664         lumBufIndex=-1;
2665         chrBufIndex=-1;
2666         dstY=0;
2667         lastInLumBuf= -1;
2668         lastInChrBuf= -1;
2669     }
2670
2671     lastDstY= dstY;
2672
2673     for (;dstY < dstH; dstY++) {
2674         unsigned char *dest =dst[0]+dstStride[0]*dstY;
2675         const int chrDstY= dstY>>c->chrDstVSubSample;
2676         unsigned char *uDest=dst[1]+dstStride[1]*chrDstY;
2677         unsigned char *vDest=dst[2]+dstStride[2]*chrDstY;
2678         unsigned char *aDest=(CONFIG_SWSCALE_ALPHA && alpPixBuf) ? dst[3]+dstStride[3]*dstY : NULL;
2679
2680         const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input
2681         const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input
2682         int lastLumSrcY= firstLumSrcY + vLumFilterSize -1; // Last line needed as input
2683         int lastChrSrcY= firstChrSrcY + vChrFilterSize -1; // Last line needed as input
2684         int enough_lines;
2685
2686         //handle holes (FAST_BILINEAR & weird filters)
2687         if (firstLumSrcY > lastInLumBuf) lastInLumBuf= firstLumSrcY-1;
2688         if (firstChrSrcY > lastInChrBuf) lastInChrBuf= firstChrSrcY-1;
2689         assert(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);
2690         assert(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);
2691
2692         // Do we have enough lines in this slice to output the dstY line
2693         enough_lines = lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < -((-srcSliceY - srcSliceH)>>c->chrSrcVSubSample);
2694         if (!enough_lines) {
2695             lastLumSrcY = srcSliceY + srcSliceH - 1;
2696             lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;
2697         }
2698
2699         DEBUG_BUFFERS("dstY: %d\n", dstY);
2700         DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n",
2701                          firstLumSrcY,    lastLumSrcY,    lastInLumBuf);
2702         DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n",
2703                          firstChrSrcY,    lastChrSrcY,    lastInChrBuf);
2704
2705         //Do horizontal scaling
2706         while(lastInLumBuf < lastLumSrcY) {
2707             uint8_t *src1= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0];
2708             uint8_t *src2= src[3]+(lastInLumBuf + 1 - srcSliceY)*srcStride[3];
2709             lumBufIndex++;
2710             DEBUG_BUFFERS("\t\tlumBufIndex %d: lastInLumBuf: %d\n",
2711                                lumBufIndex,    lastInLumBuf);
2712             assert(lumBufIndex < 2*vLumBufSize);
2713             assert(lastInLumBuf + 1 - srcSliceY < srcSliceH);
2714             assert(lastInLumBuf + 1 - srcSliceY >= 0);
2715             RENAME(hyscale)(c, lumPixBuf[ lumBufIndex ], dstW, src1, srcW, lumXInc,
2716                             flags, hLumFilter, hLumFilterPos, hLumFilterSize,
2717                             c->srcFormat, formatConvBuffer,
2718                             pal, 0);
2719             if (CONFIG_SWSCALE_ALPHA && alpPixBuf)
2720                 RENAME(hyscale)(c, alpPixBuf[ lumBufIndex ], dstW, src2, srcW, lumXInc,
2721                                 flags, hLumFilter, hLumFilterPos, hLumFilterSize,
2722                                 c->srcFormat, formatConvBuffer,
2723                                 pal, 1);
2724             lastInLumBuf++;
2725         }
2726         while(lastInChrBuf < lastChrSrcY) {
2727             uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1];
2728             uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2];
2729             chrBufIndex++;
2730             DEBUG_BUFFERS("\t\tchrBufIndex %d: lastInChrBuf: %d\n",
2731                                chrBufIndex,    lastInChrBuf);
2732             assert(chrBufIndex < 2*vChrBufSize);
2733             assert(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));
2734             assert(lastInChrBuf + 1 - chrSrcSliceY >= 0);
2735             //FIXME replace parameters through context struct (some at least)
2736
2737             if (!(isGray(srcFormat) || isGray(dstFormat)))
2738                 RENAME(hcscale)(c, chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc,
2739                                 flags, hChrFilter, hChrFilterPos, hChrFilterSize,
2740                                 c->srcFormat, formatConvBuffer,
2741                                 pal);
2742             lastInChrBuf++;
2743         }
2744         //wrap buf index around to stay inside the ring buffer
2745         if (lumBufIndex >= vLumBufSize) lumBufIndex-= vLumBufSize;
2746         if (chrBufIndex >= vChrBufSize) chrBufIndex-= vChrBufSize;
2747         if (!enough_lines)
2748             break; //we can't output a dstY line so let's try with the next slice
2749
2750 #if COMPILE_TEMPLATE_MMX
2751         c->blueDither= ff_dither8[dstY&1];
2752         if (c->dstFormat == PIX_FMT_RGB555 || c->dstFormat == PIX_FMT_BGR555)
2753             c->greenDither= ff_dither8[dstY&1];
2754         else
2755             c->greenDither= ff_dither4[dstY&1];
2756         c->redDither= ff_dither8[(dstY+1)&1];
2757 #endif
2758         if (dstY < dstH-2) {
2759             const int16_t **lumSrcPtr= (const int16_t **) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
2760             const int16_t **chrSrcPtr= (const int16_t **) chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
2761             const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
2762 #if COMPILE_TEMPLATE_MMX
2763             int i;
2764             if (flags & SWS_ACCURATE_RND) {
2765                 int s= APCK_SIZE / 8;
2766                 for (i=0; i<vLumFilterSize; i+=2) {
2767                     *(void**)&lumMmxFilter[s*i              ]= lumSrcPtr[i  ];
2768                     *(void**)&lumMmxFilter[s*i+APCK_PTR2/4  ]= lumSrcPtr[i+(vLumFilterSize>1)];
2769                               lumMmxFilter[s*i+APCK_COEF/4  ]=
2770                               lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i    ]
2771                         + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);
2772                     if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
2773                         *(void**)&alpMmxFilter[s*i              ]= alpSrcPtr[i  ];
2774                         *(void**)&alpMmxFilter[s*i+APCK_PTR2/4  ]= alpSrcPtr[i+(vLumFilterSize>1)];
2775                                   alpMmxFilter[s*i+APCK_COEF/4  ]=
2776                                   alpMmxFilter[s*i+APCK_COEF/4+1]= lumMmxFilter[s*i+APCK_COEF/4  ];
2777                     }
2778                 }
2779                 for (i=0; i<vChrFilterSize; i+=2) {
2780                     *(void**)&chrMmxFilter[s*i              ]= chrSrcPtr[i  ];
2781                     *(void**)&chrMmxFilter[s*i+APCK_PTR2/4  ]= chrSrcPtr[i+(vChrFilterSize>1)];
2782                               chrMmxFilter[s*i+APCK_COEF/4  ]=
2783                               chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i    ]
2784                         + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);
2785                 }
2786             } else {
2787                 for (i=0; i<vLumFilterSize; i++) {
2788                     lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i];
2789                     lumMmxFilter[4*i+1]= (uint64_t)lumSrcPtr[i] >> 32;
2790                     lumMmxFilter[4*i+2]=
2791                     lumMmxFilter[4*i+3]=
2792                         ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
2793                     if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
2794                         alpMmxFilter[4*i+0]= (int32_t)alpSrcPtr[i];
2795                         alpMmxFilter[4*i+1]= (uint64_t)alpSrcPtr[i] >> 32;
2796                         alpMmxFilter[4*i+2]=
2797                         alpMmxFilter[4*i+3]= lumMmxFilter[4*i+2];
2798                     }
2799                 }
2800                 for (i=0; i<vChrFilterSize; i++) {
2801                     chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i];
2802                     chrMmxFilter[4*i+1]= (uint64_t)chrSrcPtr[i] >> 32;
2803                     chrMmxFilter[4*i+2]=
2804                     chrMmxFilter[4*i+3]=
2805                         ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
2806                 }
2807             }
2808 #endif
2809             if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21) {
2810                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
2811                 if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi
2812                 c->yuv2nv12X(c,
2813                              vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2814                              vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2815                              dest, uDest, dstW, chrDstW, dstFormat);
2816             } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12 like
2817                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
2818                 if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
2819                 if (is16BPS(dstFormat)) {
2820                     yuv2yuvX16inC(
2821                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2822                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2823                                   alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW,
2824                                   dstFormat);
2825                 } else if (vLumFilterSize == 1 && vChrFilterSize == 1) { // unscaled YV12
2826                     int16_t *lumBuf = lumSrcPtr[0];
2827                     int16_t *chrBuf= chrSrcPtr[0];
2828                     int16_t *alpBuf= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? alpSrcPtr[0] : NULL;
2829                     c->yuv2yuv1(c, lumBuf, chrBuf, alpBuf, dest, uDest, vDest, aDest, dstW, chrDstW);
2830                 } else { //General YV12
2831                     c->yuv2yuvX(c,
2832                                 vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2833                                 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2834                                 alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW);
2835                 }
2836             } else {
2837                 assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);
2838                 assert(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2);
2839                 if (vLumFilterSize == 1 && vChrFilterSize == 2) { //unscaled RGB
2840                     int chrAlpha= vChrFilter[2*dstY+1];
2841                     if(flags & SWS_FULL_CHR_H_INT) {
2842                         yuv2rgbXinC_full(c, //FIXME write a packed1_full function
2843                                          vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2844                                          vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2845                                          alpSrcPtr, dest, dstW, dstY);
2846                     } else {
2847                         c->yuv2packed1(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1),
2848                                        alpPixBuf ? *alpSrcPtr : NULL,
2849                                        dest, dstW, chrAlpha, dstFormat, flags, dstY);
2850                     }
2851                 } else if (vLumFilterSize == 2 && vChrFilterSize == 2) { //bilinear upscale RGB
2852                     int lumAlpha= vLumFilter[2*dstY+1];
2853                     int chrAlpha= vChrFilter[2*dstY+1];
2854                     lumMmxFilter[2]=
2855                     lumMmxFilter[3]= vLumFilter[2*dstY   ]*0x10001;
2856                     chrMmxFilter[2]=
2857                     chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001;
2858                     if(flags & SWS_FULL_CHR_H_INT) {
2859                         yuv2rgbXinC_full(c, //FIXME write a packed2_full function
2860                                          vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2861                                          vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2862                                          alpSrcPtr, dest, dstW, dstY);
2863                     } else {
2864                         c->yuv2packed2(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1),
2865                                        alpPixBuf ? *alpSrcPtr : NULL, alpPixBuf ? *(alpSrcPtr+1) : NULL,
2866                                        dest, dstW, lumAlpha, chrAlpha, dstY);
2867                     }
2868                 } else { //general RGB
2869                     if(flags & SWS_FULL_CHR_H_INT) {
2870                         yuv2rgbXinC_full(c,
2871                                          vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2872                                          vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2873                                          alpSrcPtr, dest, dstW, dstY);
2874                     } else {
2875                         c->yuv2packedX(c,
2876                                        vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2877                                        vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2878                                        alpSrcPtr, dest, dstW, dstY);
2879                     }
2880                 }
2881             }
2882         } else { // hmm looks like we can't use MMX here without overwriting this array's tail
2883             const int16_t **lumSrcPtr= (const int16_t **)lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
2884             const int16_t **chrSrcPtr= (const int16_t **)chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
2885             const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
2886             if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21) {
2887                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
2888                 if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi
2889                 yuv2nv12XinC(
2890                              vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2891                              vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2892                              dest, uDest, dstW, chrDstW, dstFormat);
2893             } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12
2894                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
2895                 if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
2896                 if (is16BPS(dstFormat)) {
2897                     yuv2yuvX16inC(
2898                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2899                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2900                                   alpSrcPtr, (uint16_t *) dest, (uint16_t *) uDest, (uint16_t *) vDest, (uint16_t *) aDest, dstW, chrDstW,
2901                                   dstFormat);
2902                 } else {
2903                     yuv2yuvXinC(
2904                                 vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
2905                                 vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2906                                 alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW);
2907                 }
2908             } else {
2909                 assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);
2910                 assert(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2);
2911                 if(flags & SWS_FULL_CHR_H_INT) {
2912                     yuv2rgbXinC_full(c,
2913                                      vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2914                                      vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2915                                      alpSrcPtr, dest, dstW, dstY);
2916                 } else {
2917                     yuv2packedXinC(c,
2918                                    vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
2919                                    vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
2920                                    alpSrcPtr, dest, dstW, dstY);
2921                 }
2922             }
2923         }
2924     }
2925
2926     if ((dstFormat == PIX_FMT_YUVA420P) && !alpPixBuf)
2927         fillPlane(dst[3], dstStride[3], dstW, dstY-lastDstY, lastDstY, 255);
2928
2929 #if COMPILE_TEMPLATE_MMX
2930     if (flags & SWS_CPU_CAPS_MMX2 )  __asm__ volatile("sfence":::"memory");
2931     /* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */
2932     if (flags & SWS_CPU_CAPS_3DNOW)  __asm__ volatile("femms" :::"memory");
2933     else                             __asm__ volatile("emms"  :::"memory");
2934 #endif
2935     /* store changed local vars back in the context */
2936     c->dstY= dstY;
2937     c->lumBufIndex= lumBufIndex;
2938     c->chrBufIndex= chrBufIndex;
2939     c->lastInLumBuf= lastInLumBuf;
2940     c->lastInChrBuf= lastInChrBuf;
2941
2942     return dstY - lastDstY;
2943 }
2944
2945 static void RENAME(sws_init_swScale)(SwsContext *c)
2946 {
2947     enum PixelFormat srcFormat = c->srcFormat;
2948
2949     c->yuv2nv12X    = RENAME(yuv2nv12X   );
2950     c->yuv2yuv1     = RENAME(yuv2yuv1    );
2951     c->yuv2yuvX     = RENAME(yuv2yuvX    );
2952     c->yuv2packed1  = RENAME(yuv2packed1 );
2953     c->yuv2packed2  = RENAME(yuv2packed2 );
2954     c->yuv2packedX  = RENAME(yuv2packedX );
2955
2956     c->hScale       = RENAME(hScale      );
2957
2958     c->hyscale_fast = RENAME(hyscale_fast);
2959     c->hcscale_fast = RENAME(hcscale_fast);
2960
2961     c->hcscale_internal = NULL;
2962     switch(srcFormat) {
2963         case PIX_FMT_YUYV422  : c->hcscale_internal = RENAME(yuy2ToUV); break;
2964         case PIX_FMT_UYVY422  : c->hcscale_internal = RENAME(uyvyToUV); break;
2965         case PIX_FMT_NV12     : c->hcscale_internal = RENAME(nv12ToUV); break;
2966         case PIX_FMT_NV21     : c->hcscale_internal = RENAME(nv21ToUV); break;
2967         case PIX_FMT_RGB8     :
2968         case PIX_FMT_BGR8     :
2969         case PIX_FMT_PAL8     :
2970         case PIX_FMT_BGR4_BYTE:
2971         case PIX_FMT_RGB4_BYTE: c->hcscale_internal = palToUV; break;
2972         case PIX_FMT_YUV420P16BE:
2973         case PIX_FMT_YUV422P16BE:
2974         case PIX_FMT_YUV444P16BE: c->hcscale_internal = RENAME(BEToUV); break;
2975         case PIX_FMT_YUV420P16LE:
2976         case PIX_FMT_YUV422P16LE:
2977         case PIX_FMT_YUV444P16LE: c->hcscale_internal = RENAME(LEToUV); break;
2978     }
2979     if (c->chrSrcHSubSample) {
2980         switch(srcFormat) {
2981         case PIX_FMT_RGB48BE:
2982         case PIX_FMT_RGB48LE: c->hcscale_internal = rgb48ToUV_half; break;
2983         case PIX_FMT_RGB32  :
2984         case PIX_FMT_RGB32_1: c->hcscale_internal = bgr32ToUV_half; break;
2985         case PIX_FMT_BGR24  : c->hcscale_internal = RENAME(bgr24ToUV_half); break;
2986         case PIX_FMT_BGR565 : c->hcscale_internal = bgr16ToUV_half; break;
2987         case PIX_FMT_BGR555 : c->hcscale_internal = bgr15ToUV_half; break;
2988         case PIX_FMT_BGR32  :
2989         case PIX_FMT_BGR32_1: c->hcscale_internal = rgb32ToUV_half; break;
2990         case PIX_FMT_RGB24  : c->hcscale_internal = RENAME(rgb24ToUV_half); break;
2991         case PIX_FMT_RGB565 : c->hcscale_internal = rgb16ToUV_half; break;
2992         case PIX_FMT_RGB555 : c->hcscale_internal = rgb15ToUV_half; break;
2993         }
2994     } else {
2995         switch(srcFormat) {
2996         case PIX_FMT_RGB48BE:
2997         case PIX_FMT_RGB48LE: c->hcscale_internal = rgb48ToUV; break;
2998         case PIX_FMT_RGB32  :
2999         case PIX_FMT_RGB32_1: c->hcscale_internal = bgr32ToUV; break;
3000         case PIX_FMT_BGR24  : c->hcscale_internal = RENAME(bgr24ToUV); break;
3001         case PIX_FMT_BGR565 : c->hcscale_internal = bgr16ToUV; break;
3002         case PIX_FMT_BGR555 : c->hcscale_internal = bgr15ToUV; break;
3003         case PIX_FMT_BGR32  :
3004         case PIX_FMT_BGR32_1: c->hcscale_internal = rgb32ToUV; break;
3005         case PIX_FMT_RGB24  : c->hcscale_internal = RENAME(rgb24ToUV); break;
3006         case PIX_FMT_RGB565 : c->hcscale_internal = rgb16ToUV; break;
3007         case PIX_FMT_RGB555 : c->hcscale_internal = rgb15ToUV; break;
3008         }
3009     }
3010
3011     c->hyscale_internal = NULL;
3012     c->hascale_internal = NULL;
3013     switch (srcFormat) {
3014     case PIX_FMT_YUYV422  :
3015     case PIX_FMT_YUV420P16BE:
3016     case PIX_FMT_YUV422P16BE:
3017     case PIX_FMT_YUV444P16BE:
3018     case PIX_FMT_GRAY16BE : c->hyscale_internal = RENAME(yuy2ToY); break;
3019     case PIX_FMT_UYVY422  :
3020     case PIX_FMT_YUV420P16LE:
3021     case PIX_FMT_YUV422P16LE:
3022     case PIX_FMT_YUV444P16LE:
3023     case PIX_FMT_GRAY16LE : c->hyscale_internal = RENAME(uyvyToY); break;
3024     case PIX_FMT_BGR24    : c->hyscale_internal = RENAME(bgr24ToY); break;
3025     case PIX_FMT_BGR565   : c->hyscale_internal = bgr16ToY; break;
3026     case PIX_FMT_BGR555   : c->hyscale_internal = bgr15ToY; break;
3027     case PIX_FMT_RGB24    : c->hyscale_internal = RENAME(rgb24ToY); break;
3028     case PIX_FMT_RGB565   : c->hyscale_internal = rgb16ToY; break;
3029     case PIX_FMT_RGB555   : c->hyscale_internal = rgb15ToY; break;
3030     case PIX_FMT_RGB8     :
3031     case PIX_FMT_BGR8     :
3032     case PIX_FMT_PAL8     :
3033     case PIX_FMT_BGR4_BYTE:
3034     case PIX_FMT_RGB4_BYTE: c->hyscale_internal = palToY; break;
3035     case PIX_FMT_MONOBLACK: c->hyscale_internal = monoblack2Y; break;
3036     case PIX_FMT_MONOWHITE: c->hyscale_internal = monowhite2Y; break;
3037     case PIX_FMT_RGB32  :
3038     case PIX_FMT_RGB32_1: c->hyscale_internal = bgr32ToY; break;
3039     case PIX_FMT_BGR32  :
3040     case PIX_FMT_BGR32_1: c->hyscale_internal = rgb32ToY; break;
3041     case PIX_FMT_RGB48BE:
3042     case PIX_FMT_RGB48LE: c->hyscale_internal = rgb48ToY; break;
3043     }
3044     if (c->alpPixBuf) {
3045         switch (srcFormat) {
3046         case PIX_FMT_RGB32  :
3047         case PIX_FMT_RGB32_1:
3048         case PIX_FMT_BGR32  :
3049         case PIX_FMT_BGR32_1: c->hascale_internal = abgrToA; break;
3050         }
3051     }
3052 }