git.sesse.net Git - ffmpeg/blob - libswscale/x86/swscale_template.c

   1 /*
   2  * Copyright (C) 2001-2011 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
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (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 GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License 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
  21 #undef REAL_MOVNTQ
  22 #undef MOVNTQ
  23 #undef MOVNTQ2
  24 #undef PREFETCH
  25
  26 #if COMPILE_TEMPLATE_MMXEXT
  27 #define PREFETCH "prefetchnta"
  28 #else
  29 #define PREFETCH  " # nop"
  30 #endif
  31
  32 #if COMPILE_TEMPLATE_MMXEXT
  33 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
  34 #define MOVNTQ2 "movntq "
  35 #else
  36 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
  37 #define MOVNTQ2 "movq "
  38 #endif
  39 #define MOVNTQ(a,b)  REAL_MOVNTQ(a,b)
  40
  41 #if !COMPILE_TEMPLATE_MMXEXT
  42 static av_always_inline void
  43 dither_8to16(const uint8_t *srcDither, int rot)
  44 {
  45     if (rot) {
  46         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
  47                          "movq       (%0), %%mm3\n\t"
  48                          "movq      %%mm3, %%mm4\n\t"
  49                          "psrlq       $24, %%mm3\n\t"
  50                          "psllq       $40, %%mm4\n\t"
  51                          "por       %%mm4, %%mm3\n\t"
  52                          "movq      %%mm3, %%mm4\n\t"
  53                          "punpcklbw %%mm0, %%mm3\n\t"
  54                          "punpckhbw %%mm0, %%mm4\n\t"
  55                          :: "r"(srcDither)
  56                          );
  57     } else {
  58         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
  59                          "movq       (%0), %%mm3\n\t"
  60                          "movq      %%mm3, %%mm4\n\t"
  61                          "punpcklbw %%mm0, %%mm3\n\t"
  62                          "punpckhbw %%mm0, %%mm4\n\t"
  63                          :: "r"(srcDither)
  64                          );
  65     }
  66 }
  67 #endif
  68
  69 static void RENAME(yuv2yuvX)(const int16_t *filter, int filterSize,
  70                            const int16_t **src, uint8_t *dest, int dstW,
  71                            const uint8_t *dither, int offset)
  72 {
  73     dither_8to16(dither, offset);
  74     filterSize--;
  75     __asm__ volatile(
  76         "movd %0, %%mm1\n\t"
  77         "punpcklwd %%mm1, %%mm1\n\t"
  78         "punpckldq %%mm1, %%mm1\n\t"
  79         "psllw        $3, %%mm1\n\t"
  80         "paddw     %%mm1, %%mm3\n\t"
  81         "paddw     %%mm1, %%mm4\n\t"
  82         "psraw        $4, %%mm3\n\t"
  83         "psraw        $4, %%mm4\n\t"
  84         ::"m"(filterSize)
  85      );
  86
  87     __asm__ volatile(\
  88         "movq    %%mm3, %%mm6\n\t"
  89         "movq    %%mm4, %%mm7\n\t"
  90         "movl %3, %%ecx\n\t"
  91         "mov                                 %0, %%"REG_d"  \n\t"\
  92         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  93         ".p2align                             4             \n\t" /* FIXME Unroll? */\
  94         "1:                                                 \n\t"\
  95         "movq                      8(%%"REG_d"), %%mm0      \n\t" /* filterCoeff */\
  96         "movq                (%%"REG_S", %%"REG_c", 2), %%mm2      \n\t" /* srcData */\
  97         "movq               8(%%"REG_S", %%"REG_c", 2), %%mm5      \n\t" /* srcData */\
  98         "add                                $16, %%"REG_d"  \n\t"\
  99         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 100         "test                         %%"REG_S", %%"REG_S"  \n\t"\
 101         "pmulhw                           %%mm0, %%mm2      \n\t"\
 102         "pmulhw                           %%mm0, %%mm5      \n\t"\
 103         "paddw                            %%mm2, %%mm3      \n\t"\
 104         "paddw                            %%mm5, %%mm4      \n\t"\
 105         " jnz                                1b             \n\t"\
 106         "psraw                               $3, %%mm3      \n\t"\
 107         "psraw                               $3, %%mm4      \n\t"\
 108         "packuswb                         %%mm4, %%mm3      \n\t"
 109         MOVNTQ2 "                         %%mm3, (%1, %%"REG_c")\n\t"
 110         "add                          $8, %%"REG_c"         \n\t"\
 111         "cmp                          %2, %%"REG_c"         \n\t"\
 112         "movq    %%mm6, %%mm3\n\t"
 113         "movq    %%mm7, %%mm4\n\t"
 114         "mov                                 %0, %%"REG_d"  \n\t"\
 115         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 116         "jb                                  1b             \n\t"\
 117         :: "g" (filter),
 118            "r" (dest-offset), "g" ((x86_reg)(dstW+offset)), "m" (offset)
 119         : "%"REG_d, "%"REG_S, "%"REG_c
 120     );
 121 }
 122
 123 #define YSCALEYUV2PACKEDX_UV \
 124     __asm__ volatile(\
 125         "xor                   %%"REG_a", %%"REG_a"     \n\t"\
 126         ".p2align                      4                \n\t"\
 127         "nop                                            \n\t"\
 128         "1:                                             \n\t"\
 129         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 130         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 131         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\
 132         "movq                      %%mm3, %%mm4         \n\t"\
 133         ".p2align                      4                \n\t"\
 134         "2:                                             \n\t"\
 135         "movq               8(%%"REG_d"), %%mm0         \n\t" /* filterCoeff */\
 136         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* UsrcData */\
 137         "add                          %6, %%"REG_S"     \n\t" \
 138         "movq     (%%"REG_S", %%"REG_a"), %%mm5         \n\t" /* VsrcData */\
 139         "add                         $16, %%"REG_d"     \n\t"\
 140         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 141         "pmulhw                    %%mm0, %%mm2         \n\t"\
 142         "pmulhw                    %%mm0, %%mm5         \n\t"\
 143         "paddw                     %%mm2, %%mm3         \n\t"\
 144         "paddw                     %%mm5, %%mm4         \n\t"\
 145         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 146         " jnz                         2b                \n\t"\
 147
 148 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
 149     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 150     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 151     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\
 152     "movq                    "#dst1", "#dst2"       \n\t"\
 153     ".p2align                      4                \n\t"\
 154     "2:                                             \n\t"\
 155     "movq               8(%%"REG_d"), "#coeff"      \n\t" /* filterCoeff */\
 156     "movq  (%%"REG_S", %%"REG_a", 2), "#src1"       \n\t" /* Y1srcData */\
 157     "movq 8(%%"REG_S", %%"REG_a", 2), "#src2"       \n\t" /* Y2srcData */\
 158     "add                         $16, %%"REG_d"            \n\t"\
 159     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 160     "pmulhw                 "#coeff", "#src1"       \n\t"\
 161     "pmulhw                 "#coeff", "#src2"       \n\t"\
 162     "paddw                   "#src1", "#dst1"       \n\t"\
 163     "paddw                   "#src2", "#dst2"       \n\t"\
 164     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 165     " jnz                         2b                \n\t"\
 166
 167 #define YSCALEYUV2PACKEDX \
 168     YSCALEYUV2PACKEDX_UV \
 169     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
 170
 171 #define YSCALEYUV2PACKEDX_END                     \
 172         :: "r" (&c->redDither),                   \
 173             "m" (dummy), "m" (dummy), "m" (dummy),\
 174             "r" (dest), "m" (dstW_reg), "m"(uv_off) \
 175         : "%"REG_a, "%"REG_d, "%"REG_S            \
 176     );
 177
 178 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
 179     __asm__ volatile(\
 180         "xor %%"REG_a", %%"REG_a"                       \n\t"\
 181         ".p2align                      4                \n\t"\
 182         "nop                                            \n\t"\
 183         "1:                                             \n\t"\
 184         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 185         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 186         "pxor                      %%mm4, %%mm4         \n\t"\
 187         "pxor                      %%mm5, %%mm5         \n\t"\
 188         "pxor                      %%mm6, %%mm6         \n\t"\
 189         "pxor                      %%mm7, %%mm7         \n\t"\
 190         ".p2align                      4                \n\t"\
 191         "2:                                             \n\t"\
 192         "movq     (%%"REG_S", %%"REG_a"), %%mm0         \n\t" /* UsrcData */\
 193         "add                          %6, %%"REG_S"      \n\t" \
 194         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* VsrcData */\
 195         "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 196         "movq     (%%"REG_S", %%"REG_a"), %%mm1         \n\t" /* UsrcData */\
 197         "movq                      %%mm0, %%mm3         \n\t"\
 198         "punpcklwd                 %%mm1, %%mm0         \n\t"\
 199         "punpckhwd                 %%mm1, %%mm3         \n\t"\
 200         "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1         \n\t" /* filterCoeff */\
 201         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 202         "pmaddwd                   %%mm1, %%mm3         \n\t"\
 203         "paddd                     %%mm0, %%mm4         \n\t"\
 204         "paddd                     %%mm3, %%mm5         \n\t"\
 205         "add                          %6, %%"REG_S"      \n\t" \
 206         "movq     (%%"REG_S", %%"REG_a"), %%mm3         \n\t" /* VsrcData */\
 207         "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 208         "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 209         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 210         "movq                      %%mm2, %%mm0         \n\t"\
 211         "punpcklwd                 %%mm3, %%mm2         \n\t"\
 212         "punpckhwd                 %%mm3, %%mm0         \n\t"\
 213         "pmaddwd                   %%mm1, %%mm2         \n\t"\
 214         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 215         "paddd                     %%mm2, %%mm6         \n\t"\
 216         "paddd                     %%mm0, %%mm7         \n\t"\
 217         " jnz                         2b                \n\t"\
 218         "psrad                       $16, %%mm4         \n\t"\
 219         "psrad                       $16, %%mm5         \n\t"\
 220         "psrad                       $16, %%mm6         \n\t"\
 221         "psrad                       $16, %%mm7         \n\t"\
 222         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 223         "packssdw                  %%mm5, %%mm4         \n\t"\
 224         "packssdw                  %%mm7, %%mm6         \n\t"\
 225         "paddw                     %%mm0, %%mm4         \n\t"\
 226         "paddw                     %%mm0, %%mm6         \n\t"\
 227         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\
 228         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\
 229
 230 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
 231     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 232     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 233     "pxor                      %%mm1, %%mm1         \n\t"\
 234     "pxor                      %%mm5, %%mm5         \n\t"\
 235     "pxor                      %%mm7, %%mm7         \n\t"\
 236     "pxor                      %%mm6, %%mm6         \n\t"\
 237     ".p2align                      4                \n\t"\
 238     "2:                                             \n\t"\
 239     "movq  (%%"REG_S", %%"REG_a", 2), %%mm0         \n\t" /* Y1srcData */\
 240     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y2srcData */\
 241     "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 242     "movq  (%%"REG_S", %%"REG_a", 2), %%mm4         \n\t" /* Y1srcData */\
 243     "movq                      %%mm0, %%mm3         \n\t"\
 244     "punpcklwd                 %%mm4, %%mm0         \n\t"\
 245     "punpckhwd                 %%mm4, %%mm3         \n\t"\
 246     "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4         \n\t" /* filterCoeff */\
 247     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 248     "pmaddwd                   %%mm4, %%mm3         \n\t"\
 249     "paddd                     %%mm0, %%mm1         \n\t"\
 250     "paddd                     %%mm3, %%mm5         \n\t"\
 251     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3         \n\t" /* Y2srcData */\
 252     "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 253     "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 254     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 255     "movq                      %%mm2, %%mm0         \n\t"\
 256     "punpcklwd                 %%mm3, %%mm2         \n\t"\
 257     "punpckhwd                 %%mm3, %%mm0         \n\t"\
 258     "pmaddwd                   %%mm4, %%mm2         \n\t"\
 259     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 260     "paddd                     %%mm2, %%mm7         \n\t"\
 261     "paddd                     %%mm0, %%mm6         \n\t"\
 262     " jnz                         2b                \n\t"\
 263     "psrad                       $16, %%mm1         \n\t"\
 264     "psrad                       $16, %%mm5         \n\t"\
 265     "psrad                       $16, %%mm7         \n\t"\
 266     "psrad                       $16, %%mm6         \n\t"\
 267     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 268     "packssdw                  %%mm5, %%mm1         \n\t"\
 269     "packssdw                  %%mm6, %%mm7         \n\t"\
 270     "paddw                     %%mm0, %%mm1         \n\t"\
 271     "paddw                     %%mm0, %%mm7         \n\t"\
 272     "movq               "U_TEMP"(%0), %%mm3         \n\t"\
 273     "movq               "V_TEMP"(%0), %%mm4         \n\t"\
 274
 275 #define YSCALEYUV2PACKEDX_ACCURATE \
 276     YSCALEYUV2PACKEDX_ACCURATE_UV \
 277     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
 278
 279 #define YSCALEYUV2RGBX \
 280     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\
 281     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\
 282     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\
 283     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\
 284     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\
 285     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\
 286     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 287     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\
 288     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\
 289     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\
 290     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\
 291     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\
 292     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\
 293     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 294     "paddw           %%mm3, %%mm4       \n\t"\
 295     "movq            %%mm2, %%mm0       \n\t"\
 296     "movq            %%mm5, %%mm6       \n\t"\
 297     "movq            %%mm4, %%mm3       \n\t"\
 298     "punpcklwd       %%mm2, %%mm2       \n\t"\
 299     "punpcklwd       %%mm5, %%mm5       \n\t"\
 300     "punpcklwd       %%mm4, %%mm4       \n\t"\
 301     "paddw           %%mm1, %%mm2       \n\t"\
 302     "paddw           %%mm1, %%mm5       \n\t"\
 303     "paddw           %%mm1, %%mm4       \n\t"\
 304     "punpckhwd       %%mm0, %%mm0       \n\t"\
 305     "punpckhwd       %%mm6, %%mm6       \n\t"\
 306     "punpckhwd       %%mm3, %%mm3       \n\t"\
 307     "paddw           %%mm7, %%mm0       \n\t"\
 308     "paddw           %%mm7, %%mm6       \n\t"\
 309     "paddw           %%mm7, %%mm3       \n\t"\
 310     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 311     "packuswb        %%mm0, %%mm2       \n\t"\
 312     "packuswb        %%mm6, %%mm5       \n\t"\
 313     "packuswb        %%mm3, %%mm4       \n\t"\
 314
 315 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
 316     "movq       "#b", "#q2"     \n\t" /* B */\
 317     "movq       "#r", "#t"      \n\t" /* R */\
 318     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\
 319     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\
 320     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\
 321     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\
 322     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\
 323     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\
 324     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\
 325     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\
 326     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\
 327     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\
 328 \
 329     MOVNTQ(   q0,   (dst, index, 4))\
 330     MOVNTQ(    b,  8(dst, index, 4))\
 331     MOVNTQ(   q2, 16(dst, index, 4))\
 332     MOVNTQ(   q3, 24(dst, index, 4))\
 333 \
 334     "add      $8, "#index"      \n\t"\
 335     "cmp "#dstw", "#index"      \n\t"\
 336     " jb      1b                \n\t"
 337 #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)
 338
 339 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
 340                                    const int16_t **lumSrc, int lumFilterSize,
 341                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 342                                    const int16_t **chrVSrc,
 343                                    int chrFilterSize, const int16_t **alpSrc,
 344                                    uint8_t *dest, int dstW, int dstY)
 345 {
 346     x86_reg dummy=0;
 347     x86_reg dstW_reg = dstW;
 348     x86_reg uv_off = c->uv_offx2;
 349
 350     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 351         YSCALEYUV2PACKEDX_ACCURATE
 352         YSCALEYUV2RGBX
 353         "movq                      %%mm2, "U_TEMP"(%0)  \n\t"
 354         "movq                      %%mm4, "V_TEMP"(%0)  \n\t"
 355         "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"
 356         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
 357         "movq               "Y_TEMP"(%0), %%mm5         \n\t"
 358         "psraw                        $3, %%mm1         \n\t"
 359         "psraw                        $3, %%mm7         \n\t"
 360         "packuswb                  %%mm7, %%mm1         \n\t"
 361         WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
 362         YSCALEYUV2PACKEDX_END
 363     } else {
 364         YSCALEYUV2PACKEDX_ACCURATE
 365         YSCALEYUV2RGBX
 366         "pcmpeqd %%mm7, %%mm7 \n\t"
 367         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 368         YSCALEYUV2PACKEDX_END
 369     }
 370 }
 371
 372 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
 373                                 const int16_t **lumSrc, int lumFilterSize,
 374                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 375                                 const int16_t **chrVSrc,
 376                                 int chrFilterSize, const int16_t **alpSrc,
 377                                 uint8_t *dest, int dstW, int dstY)
 378 {
 379     x86_reg dummy=0;
 380     x86_reg dstW_reg = dstW;
 381     x86_reg uv_off = c->uv_offx2;
 382
 383     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 384         YSCALEYUV2PACKEDX
 385         YSCALEYUV2RGBX
 386         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
 387         "psraw                        $3, %%mm1         \n\t"
 388         "psraw                        $3, %%mm7         \n\t"
 389         "packuswb                  %%mm7, %%mm1         \n\t"
 390         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 391         YSCALEYUV2PACKEDX_END
 392     } else {
 393         YSCALEYUV2PACKEDX
 394         YSCALEYUV2RGBX
 395         "pcmpeqd %%mm7, %%mm7 \n\t"
 396         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 397         YSCALEYUV2PACKEDX_END
 398     }
 399 }
 400
 401 #define REAL_WRITERGB16(dst, dstw, index) \
 402     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 403     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
 404     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 405     "psrlq           $3, %%mm2  \n\t"\
 406 \
 407     "movq         %%mm2, %%mm1  \n\t"\
 408     "movq         %%mm4, %%mm3  \n\t"\
 409 \
 410     "punpcklbw    %%mm7, %%mm3  \n\t"\
 411     "punpcklbw    %%mm5, %%mm2  \n\t"\
 412     "punpckhbw    %%mm7, %%mm4  \n\t"\
 413     "punpckhbw    %%mm5, %%mm1  \n\t"\
 414 \
 415     "psllq           $3, %%mm3  \n\t"\
 416     "psllq           $3, %%mm4  \n\t"\
 417 \
 418     "por          %%mm3, %%mm2  \n\t"\
 419     "por          %%mm4, %%mm1  \n\t"\
 420 \
 421     MOVNTQ(%%mm2,  (dst, index, 2))\
 422     MOVNTQ(%%mm1, 8(dst, index, 2))\
 423 \
 424     "add             $8, "#index"   \n\t"\
 425     "cmp        "#dstw", "#index"   \n\t"\
 426     " jb             1b             \n\t"
 427 #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)
 428
 429 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
 430                                     const int16_t **lumSrc, int lumFilterSize,
 431                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 432                                     const int16_t **chrVSrc,
 433                                     int chrFilterSize, const int16_t **alpSrc,
 434                                     uint8_t *dest, int dstW, int dstY)
 435 {
 436     x86_reg dummy=0;
 437     x86_reg dstW_reg = dstW;
 438     x86_reg uv_off = c->uv_offx2;
 439
 440     YSCALEYUV2PACKEDX_ACCURATE
 441     YSCALEYUV2RGBX
 442     "pxor %%mm7, %%mm7 \n\t"
 443     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 444 #ifdef DITHER1XBPP
 445     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 446     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 447     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 448 #endif
 449     WRITERGB16(%4, %5, %%REGa)
 450     YSCALEYUV2PACKEDX_END
 451 }
 452
 453 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
 454                                  const int16_t **lumSrc, int lumFilterSize,
 455                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 456                                  const int16_t **chrVSrc,
 457                                  int chrFilterSize, const int16_t **alpSrc,
 458                                  uint8_t *dest, int dstW, int dstY)
 459 {
 460     x86_reg dummy=0;
 461     x86_reg dstW_reg = dstW;
 462     x86_reg uv_off = c->uv_offx2;
 463
 464     YSCALEYUV2PACKEDX
 465     YSCALEYUV2RGBX
 466     "pxor %%mm7, %%mm7 \n\t"
 467     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 468 #ifdef DITHER1XBPP
 469     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 470     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 471     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 472 #endif
 473     WRITERGB16(%4, %5, %%REGa)
 474     YSCALEYUV2PACKEDX_END
 475 }
 476
 477 #define REAL_WRITERGB15(dst, dstw, index) \
 478     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 479     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\
 480     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 481     "psrlq           $3, %%mm2  \n\t"\
 482     "psrlq           $1, %%mm5  \n\t"\
 483 \
 484     "movq         %%mm2, %%mm1  \n\t"\
 485     "movq         %%mm4, %%mm3  \n\t"\
 486 \
 487     "punpcklbw    %%mm7, %%mm3  \n\t"\
 488     "punpcklbw    %%mm5, %%mm2  \n\t"\
 489     "punpckhbw    %%mm7, %%mm4  \n\t"\
 490     "punpckhbw    %%mm5, %%mm1  \n\t"\
 491 \
 492     "psllq           $2, %%mm3  \n\t"\
 493     "psllq           $2, %%mm4  \n\t"\
 494 \
 495     "por          %%mm3, %%mm2  \n\t"\
 496     "por          %%mm4, %%mm1  \n\t"\
 497 \
 498     MOVNTQ(%%mm2,  (dst, index, 2))\
 499     MOVNTQ(%%mm1, 8(dst, index, 2))\
 500 \
 501     "add             $8, "#index"   \n\t"\
 502     "cmp        "#dstw", "#index"   \n\t"\
 503     " jb             1b             \n\t"
 504 #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)
 505
 506 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
 507                                     const int16_t **lumSrc, int lumFilterSize,
 508                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 509                                     const int16_t **chrVSrc,
 510                                     int chrFilterSize, const int16_t **alpSrc,
 511                                     uint8_t *dest, int dstW, int dstY)
 512 {
 513     x86_reg dummy=0;
 514     x86_reg dstW_reg = dstW;
 515     x86_reg uv_off = c->uv_offx2;
 516
 517     YSCALEYUV2PACKEDX_ACCURATE
 518     YSCALEYUV2RGBX
 519     "pxor %%mm7, %%mm7 \n\t"
 520     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 521 #ifdef DITHER1XBPP
 522     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 523     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 524     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 525 #endif
 526     WRITERGB15(%4, %5, %%REGa)
 527     YSCALEYUV2PACKEDX_END
 528 }
 529
 530 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
 531                                  const int16_t **lumSrc, int lumFilterSize,
 532                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 533                                  const int16_t **chrVSrc,
 534                                  int chrFilterSize, const int16_t **alpSrc,
 535                                  uint8_t *dest, int dstW, int dstY)
 536 {
 537     x86_reg dummy=0;
 538     x86_reg dstW_reg = dstW;
 539     x86_reg uv_off = c->uv_offx2;
 540
 541     YSCALEYUV2PACKEDX
 542     YSCALEYUV2RGBX
 543     "pxor %%mm7, %%mm7 \n\t"
 544     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 545 #ifdef DITHER1XBPP
 546     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 547     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 548     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 549 #endif
 550     WRITERGB15(%4, %5, %%REGa)
 551     YSCALEYUV2PACKEDX_END
 552 }
 553
 554 #define WRITEBGR24MMX(dst, dstw, index) \
 555     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 556     "movq      %%mm2, %%mm1     \n\t" /* B */\
 557     "movq      %%mm5, %%mm6     \n\t" /* R */\
 558     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\
 559     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\
 560     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\
 561     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\
 562     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\
 563     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\
 564     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\
 565     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\
 566     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\
 567     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\
 568 \
 569     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\
 570     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\
 571     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\
 572     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\
 573 \
 574     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\
 575     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\
 576     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\
 577     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\
 578 \
 579     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\
 580     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\
 581     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\
 582     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\
 583 \
 584     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\
 585     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\
 586     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\
 587     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\
 588     MOVNTQ(%%mm0, (dst))\
 589 \
 590     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\
 591     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\
 592     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\
 593     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\
 594     MOVNTQ(%%mm6, 8(dst))\
 595 \
 596     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\
 597     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\
 598     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\
 599     MOVNTQ(%%mm5, 16(dst))\
 600 \
 601     "add         $24, "#dst"    \n\t"\
 602 \
 603     "add          $8, "#index"  \n\t"\
 604     "cmp     "#dstw", "#index"  \n\t"\
 605     " jb          1b            \n\t"
 606
 607 #define WRITEBGR24MMXEXT(dst, dstw, index) \
 608     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 609     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
 610     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
 611     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
 612     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
 613     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
 614 \
 615     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\
 616     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\
 617     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\
 618 \
 619     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\
 620     "por    %%mm1, %%mm6        \n\t"\
 621     "por    %%mm3, %%mm6        \n\t"\
 622     MOVNTQ(%%mm6, (dst))\
 623 \
 624     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
 625     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
 626     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
 627     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
 628 \
 629     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\
 630     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\
 631     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\
 632 \
 633     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\
 634     "por    %%mm3, %%mm6        \n\t"\
 635     MOVNTQ(%%mm6, 8(dst))\
 636 \
 637     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
 638     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
 639     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
 640 \
 641     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\
 642     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\
 643     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\
 644 \
 645     "por    %%mm1, %%mm3        \n\t"\
 646     "por    %%mm3, %%mm6        \n\t"\
 647     MOVNTQ(%%mm6, 16(dst))\
 648 \
 649     "add      $24, "#dst"       \n\t"\
 650 \
 651     "add       $8, "#index"     \n\t"\
 652     "cmp  "#dstw", "#index"     \n\t"\
 653     " jb       1b               \n\t"
 654
 655 #if COMPILE_TEMPLATE_MMXEXT
 656 #undef WRITEBGR24
 657 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMXEXT(dst, dstw, index)
 658 #else
 659 #undef WRITEBGR24
 660 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)
 661 #endif
 662
 663 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
 664                                    const int16_t **lumSrc, int lumFilterSize,
 665                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 666                                    const int16_t **chrVSrc,
 667                                    int chrFilterSize, const int16_t **alpSrc,
 668                                    uint8_t *dest, int dstW, int dstY)
 669 {
 670     x86_reg dummy=0;
 671     x86_reg dstW_reg = dstW;
 672     x86_reg uv_off = c->uv_offx2;
 673
 674     YSCALEYUV2PACKEDX_ACCURATE
 675     YSCALEYUV2RGBX
 676     "pxor %%mm7, %%mm7 \n\t"
 677     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
 678     "add %4, %%"REG_c"                        \n\t"
 679     WRITEBGR24(%%REGc, %5, %%REGa)
 680     :: "r" (&c->redDither),
 681        "m" (dummy), "m" (dummy), "m" (dummy),
 682        "r" (dest), "m" (dstW_reg), "m"(uv_off)
 683     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 684     );
 685 }
 686
 687 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
 688                                 const int16_t **lumSrc, int lumFilterSize,
 689                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 690                                 const int16_t **chrVSrc,
 691                                 int chrFilterSize, const int16_t **alpSrc,
 692                                 uint8_t *dest, int dstW, int dstY)
 693 {
 694     x86_reg dummy=0;
 695     x86_reg dstW_reg = dstW;
 696     x86_reg uv_off = c->uv_offx2;
 697
 698     YSCALEYUV2PACKEDX
 699     YSCALEYUV2RGBX
 700     "pxor                    %%mm7, %%mm7       \n\t"
 701     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize
 702     "add                        %4, %%"REG_c"   \n\t"
 703     WRITEBGR24(%%REGc, %5, %%REGa)
 704     :: "r" (&c->redDither),
 705        "m" (dummy), "m" (dummy), "m" (dummy),
 706        "r" (dest),  "m" (dstW_reg), "m"(uv_off)
 707     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 708     );
 709 }
 710
 711 #define REAL_WRITEYUY2(dst, dstw, index) \
 712     "packuswb  %%mm3, %%mm3     \n\t"\
 713     "packuswb  %%mm4, %%mm4     \n\t"\
 714     "packuswb  %%mm7, %%mm1     \n\t"\
 715     "punpcklbw %%mm4, %%mm3     \n\t"\
 716     "movq      %%mm1, %%mm7     \n\t"\
 717     "punpcklbw %%mm3, %%mm1     \n\t"\
 718     "punpckhbw %%mm3, %%mm7     \n\t"\
 719 \
 720     MOVNTQ(%%mm1, (dst, index, 2))\
 721     MOVNTQ(%%mm7, 8(dst, index, 2))\
 722 \
 723     "add          $8, "#index"  \n\t"\
 724     "cmp     "#dstw", "#index"  \n\t"\
 725     " jb          1b            \n\t"
 726 #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)
 727
 728 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
 729                                      const int16_t **lumSrc, int lumFilterSize,
 730                                      const int16_t *chrFilter, const int16_t **chrUSrc,
 731                                      const int16_t **chrVSrc,
 732                                      int chrFilterSize, const int16_t **alpSrc,
 733                                      uint8_t *dest, int dstW, int dstY)
 734 {
 735     x86_reg dummy=0;
 736     x86_reg dstW_reg = dstW;
 737     x86_reg uv_off = c->uv_offx2;
 738
 739     YSCALEYUV2PACKEDX_ACCURATE
 740     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 741     "psraw $3, %%mm3    \n\t"
 742     "psraw $3, %%mm4    \n\t"
 743     "psraw $3, %%mm1    \n\t"
 744     "psraw $3, %%mm7    \n\t"
 745     WRITEYUY2(%4, %5, %%REGa)
 746     YSCALEYUV2PACKEDX_END
 747 }
 748
 749 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
 750                                   const int16_t **lumSrc, int lumFilterSize,
 751                                   const int16_t *chrFilter, const int16_t **chrUSrc,
 752                                   const int16_t **chrVSrc,
 753                                   int chrFilterSize, const int16_t **alpSrc,
 754                                   uint8_t *dest, int dstW, int dstY)
 755 {
 756     x86_reg dummy=0;
 757     x86_reg dstW_reg = dstW;
 758     x86_reg uv_off = c->uv_offx2;
 759
 760     YSCALEYUV2PACKEDX
 761     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 762     "psraw $3, %%mm3    \n\t"
 763     "psraw $3, %%mm4    \n\t"
 764     "psraw $3, %%mm1    \n\t"
 765     "psraw $3, %%mm7    \n\t"
 766     WRITEYUY2(%4, %5, %%REGa)
 767     YSCALEYUV2PACKEDX_END
 768 }
 769
 770 #define REAL_YSCALEYUV2RGB_UV(index, c) \
 771     "xor            "#index", "#index"  \n\t"\
 772     ".p2align              4            \n\t"\
 773     "1:                                 \n\t"\
 774     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 775     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 776     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 777     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 778     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 779     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 780     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
 781     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
 782     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
 783     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
 784     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
 785     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 786     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 787     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
 788     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
 789     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 790     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 791     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 792     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 793     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 794     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 795     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 796
 797 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
 798     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
 799     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
 800     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
 801     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
 802     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
 803     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
 804     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 805     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 806     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 807     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 808     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 809     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 810
 811 #define REAL_YSCALEYUV2RGB_COEFF(c) \
 812     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 813     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
 814     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
 815     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
 816     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
 817     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
 818     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 819     "paddw             %%mm3, %%mm4     \n\t"\
 820     "movq              %%mm2, %%mm0     \n\t"\
 821     "movq              %%mm5, %%mm6     \n\t"\
 822     "movq              %%mm4, %%mm3     \n\t"\
 823     "punpcklwd         %%mm2, %%mm2     \n\t"\
 824     "punpcklwd         %%mm5, %%mm5     \n\t"\
 825     "punpcklwd         %%mm4, %%mm4     \n\t"\
 826     "paddw             %%mm1, %%mm2     \n\t"\
 827     "paddw             %%mm1, %%mm5     \n\t"\
 828     "paddw             %%mm1, %%mm4     \n\t"\
 829     "punpckhwd         %%mm0, %%mm0     \n\t"\
 830     "punpckhwd         %%mm6, %%mm6     \n\t"\
 831     "punpckhwd         %%mm3, %%mm3     \n\t"\
 832     "paddw             %%mm7, %%mm0     \n\t"\
 833     "paddw             %%mm7, %%mm6     \n\t"\
 834     "paddw             %%mm7, %%mm3     \n\t"\
 835     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 836     "packuswb          %%mm0, %%mm2     \n\t"\
 837     "packuswb          %%mm6, %%mm5     \n\t"\
 838     "packuswb          %%mm3, %%mm4     \n\t"\
 839
 840 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
 841
 842 #define YSCALEYUV2RGB(index, c) \
 843     REAL_YSCALEYUV2RGB_UV(index, c) \
 844     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
 845     REAL_YSCALEYUV2RGB_COEFF(c)
 846
 847 /**
 848  * vertical bilinear scale YV12 to RGB
 849  */
 850 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
 851                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
 852                                 const int16_t *abuf[2], uint8_t *dest,
 853                                 int dstW, int yalpha, int uvalpha, int y)
 854 {
 855     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 856                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 857
 858     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 859         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
 860 #if ARCH_X86_64
 861         __asm__ volatile(
 862             YSCALEYUV2RGB(%%r8, %5)
 863             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
 864             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 865             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 866             "packuswb            %%mm7, %%mm1       \n\t"
 867             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 868             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
 869                "a" (&c->redDither),
 870                "r" (abuf0), "r" (abuf1)
 871             : "%r8"
 872         );
 873 #else
 874         c->u_temp=(intptr_t)abuf0;
 875         c->v_temp=(intptr_t)abuf1;
 876         __asm__ volatile(
 877             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 878             "mov        %4, %%"REG_b"               \n\t"
 879             "push %%"REG_BP"                        \n\t"
 880             YSCALEYUV2RGB(%%REGBP, %5)
 881             "push                   %0              \n\t"
 882             "push                   %1              \n\t"
 883             "mov          "U_TEMP"(%5), %0          \n\t"
 884             "mov          "V_TEMP"(%5), %1          \n\t"
 885             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
 886             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 887             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 888             "packuswb            %%mm7, %%mm1       \n\t"
 889             "pop                    %1              \n\t"
 890             "pop                    %0              \n\t"
 891             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 892             "pop %%"REG_BP"                         \n\t"
 893             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 894             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 895                "a" (&c->redDither)
 896         );
 897 #endif
 898     } else {
 899         __asm__ volatile(
 900             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 901             "mov        %4, %%"REG_b"               \n\t"
 902             "push %%"REG_BP"                        \n\t"
 903             YSCALEYUV2RGB(%%REGBP, %5)
 904             "pcmpeqd %%mm7, %%mm7                   \n\t"
 905             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 906             "pop %%"REG_BP"                         \n\t"
 907             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 908             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 909                "a" (&c->redDither)
 910         );
 911     }
 912 }
 913
 914 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
 915                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
 916                                 const int16_t *abuf[2], uint8_t *dest,
 917                                 int dstW, int yalpha, int uvalpha, int y)
 918 {
 919     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 920                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 921
 922     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 923     __asm__ volatile(
 924         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 925         "mov        %4, %%"REG_b"               \n\t"
 926         "push %%"REG_BP"                        \n\t"
 927         YSCALEYUV2RGB(%%REGBP, %5)
 928         "pxor    %%mm7, %%mm7                   \n\t"
 929         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
 930         "pop %%"REG_BP"                         \n\t"
 931         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 932         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 933            "a" (&c->redDither)
 934     );
 935 }
 936
 937 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
 938                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
 939                                  const int16_t *abuf[2], uint8_t *dest,
 940                                  int dstW, int yalpha, int uvalpha, int y)
 941 {
 942     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 943                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 944
 945     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 946     __asm__ volatile(
 947         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 948         "mov        %4, %%"REG_b"               \n\t"
 949         "push %%"REG_BP"                        \n\t"
 950         YSCALEYUV2RGB(%%REGBP, %5)
 951         "pxor    %%mm7, %%mm7                   \n\t"
 952         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 953 #ifdef DITHER1XBPP
 954         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
 955         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
 956         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
 957 #endif
 958         WRITERGB15(%%REGb, 8280(%5), %%REGBP)
 959         "pop %%"REG_BP"                         \n\t"
 960         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 961         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 962            "a" (&c->redDither)
 963     );
 964 }
 965
 966 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
 967                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
 968                                  const int16_t *abuf[2], uint8_t *dest,
 969                                  int dstW, int yalpha, int uvalpha, int y)
 970 {
 971     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 972                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 973
 974     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 975     __asm__ volatile(
 976         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 977         "mov        %4, %%"REG_b"               \n\t"
 978         "push %%"REG_BP"                        \n\t"
 979         YSCALEYUV2RGB(%%REGBP, %5)
 980         "pxor    %%mm7, %%mm7                   \n\t"
 981         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 982 #ifdef DITHER1XBPP
 983         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
 984         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
 985         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
 986 #endif
 987         WRITERGB16(%%REGb, 8280(%5), %%REGBP)
 988         "pop %%"REG_BP"                         \n\t"
 989         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 990         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 991            "a" (&c->redDither)
 992     );
 993 }
 994
 995 #define REAL_YSCALEYUV2PACKED(index, c) \
 996     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\
 997     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\
 998     "psraw                $3, %%mm0                           \n\t"\
 999     "psraw                $3, %%mm1                           \n\t"\
1000     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1001     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1002     "xor            "#index", "#index"                        \n\t"\
1003     ".p2align              4            \n\t"\
1004     "1:                                 \n\t"\
1005     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1006     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1007     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1008     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1009     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1010     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1011     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1012     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1013     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
1014     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1015     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1016     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1017     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1018     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1019     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1020     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
1021     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
1022     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
1023     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
1024     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
1025     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
1026     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1027     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1028     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1029     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1030     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1031     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1032
1033 #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)
1034
1035 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1036                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
1037                                   const int16_t *abuf[2], uint8_t *dest,
1038                                   int dstW, int yalpha, int uvalpha, int y)
1039 {
1040     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1041                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1042
1043     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1044     __asm__ volatile(
1045         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1046         "mov %4, %%"REG_b"                        \n\t"
1047         "push %%"REG_BP"                        \n\t"
1048         YSCALEYUV2PACKED(%%REGBP, %5)
1049         WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1050         "pop %%"REG_BP"                         \n\t"
1051         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1052         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1053            "a" (&c->redDither)
1054     );
1055 }
1056
1057 #define REAL_YSCALEYUV2RGB1(index, c) \
1058     "xor            "#index", "#index"  \n\t"\
1059     ".p2align              4            \n\t"\
1060     "1:                                 \n\t"\
1061     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1062     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1063     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1064     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1065     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1066     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1067     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1068     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1069     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1070     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1071     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1072     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1073     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1074     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1075     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1076     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1077     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1078     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1079     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1080     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1081     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1082     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1083     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1084     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1085     "paddw             %%mm3, %%mm4     \n\t"\
1086     "movq              %%mm2, %%mm0     \n\t"\
1087     "movq              %%mm5, %%mm6     \n\t"\
1088     "movq              %%mm4, %%mm3     \n\t"\
1089     "punpcklwd         %%mm2, %%mm2     \n\t"\
1090     "punpcklwd         %%mm5, %%mm5     \n\t"\
1091     "punpcklwd         %%mm4, %%mm4     \n\t"\
1092     "paddw             %%mm1, %%mm2     \n\t"\
1093     "paddw             %%mm1, %%mm5     \n\t"\
1094     "paddw             %%mm1, %%mm4     \n\t"\
1095     "punpckhwd         %%mm0, %%mm0     \n\t"\
1096     "punpckhwd         %%mm6, %%mm6     \n\t"\
1097     "punpckhwd         %%mm3, %%mm3     \n\t"\
1098     "paddw             %%mm7, %%mm0     \n\t"\
1099     "paddw             %%mm7, %%mm6     \n\t"\
1100     "paddw             %%mm7, %%mm3     \n\t"\
1101     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1102     "packuswb          %%mm0, %%mm2     \n\t"\
1103     "packuswb          %%mm6, %%mm5     \n\t"\
1104     "packuswb          %%mm3, %%mm4     \n\t"\
1105
1106 #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)
1107
1108 // do vertical chrominance interpolation
1109 #define REAL_YSCALEYUV2RGB1b(index, c) \
1110     "xor            "#index", "#index"  \n\t"\
1111     ".p2align              4            \n\t"\
1112     "1:                                 \n\t"\
1113     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1114     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1115     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1116     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1117     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1118     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1119     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1120     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1121     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
1122     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\
1123     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1124     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1125     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1126     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1127     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1128     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1129     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1130     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1131     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1132     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1133     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1134     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1135     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1136     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1137     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1138     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1139     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1140     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1141     "paddw             %%mm3, %%mm4     \n\t"\
1142     "movq              %%mm2, %%mm0     \n\t"\
1143     "movq              %%mm5, %%mm6     \n\t"\
1144     "movq              %%mm4, %%mm3     \n\t"\
1145     "punpcklwd         %%mm2, %%mm2     \n\t"\
1146     "punpcklwd         %%mm5, %%mm5     \n\t"\
1147     "punpcklwd         %%mm4, %%mm4     \n\t"\
1148     "paddw             %%mm1, %%mm2     \n\t"\
1149     "paddw             %%mm1, %%mm5     \n\t"\
1150     "paddw             %%mm1, %%mm4     \n\t"\
1151     "punpckhwd         %%mm0, %%mm0     \n\t"\
1152     "punpckhwd         %%mm6, %%mm6     \n\t"\
1153     "punpckhwd         %%mm3, %%mm3     \n\t"\
1154     "paddw             %%mm7, %%mm0     \n\t"\
1155     "paddw             %%mm7, %%mm6     \n\t"\
1156     "paddw             %%mm7, %%mm3     \n\t"\
1157     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1158     "packuswb          %%mm0, %%mm2     \n\t"\
1159     "packuswb          %%mm6, %%mm5     \n\t"\
1160     "packuswb          %%mm3, %%mm4     \n\t"\
1161
1162 #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)
1163
1164 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1165     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\
1166     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\
1167     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\
1168     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\
1169     "packuswb          %%mm1, %%mm7     \n\t"
1170 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1171
1172 /**
1173  * YV12 to RGB without scaling or interpolating
1174  */
1175 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1176                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1177                                 const int16_t *abuf0, uint8_t *dest,
1178                                 int dstW, int uvalpha, int y)
1179 {
1180     const int16_t *ubuf0 = ubuf[0];
1181     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1182
1183     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1184         const int16_t *ubuf1 = ubuf[0];
1185         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1186             __asm__ volatile(
1187                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1188                 "mov        %4, %%"REG_b"               \n\t"
1189                 "push %%"REG_BP"                        \n\t"
1190                 YSCALEYUV2RGB1(%%REGBP, %5)
1191                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1192                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1193                 "pop %%"REG_BP"                         \n\t"
1194                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1195                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1196                    "a" (&c->redDither)
1197             );
1198         } else {
1199             __asm__ volatile(
1200                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1201                 "mov        %4, %%"REG_b"               \n\t"
1202                 "push %%"REG_BP"                        \n\t"
1203                 YSCALEYUV2RGB1(%%REGBP, %5)
1204                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1205                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1206                 "pop %%"REG_BP"                         \n\t"
1207                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1208                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1209                    "a" (&c->redDither)
1210             );
1211         }
1212     } else {
1213         const int16_t *ubuf1 = ubuf[1];
1214         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1215             __asm__ volatile(
1216                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1217                 "mov        %4, %%"REG_b"               \n\t"
1218                 "push %%"REG_BP"                        \n\t"
1219                 YSCALEYUV2RGB1b(%%REGBP, %5)
1220                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1221                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1222                 "pop %%"REG_BP"                         \n\t"
1223                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1224                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1225                    "a" (&c->redDither)
1226             );
1227         } else {
1228             __asm__ volatile(
1229                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1230                 "mov        %4, %%"REG_b"               \n\t"
1231                 "push %%"REG_BP"                        \n\t"
1232                 YSCALEYUV2RGB1b(%%REGBP, %5)
1233                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1234                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1235                 "pop %%"REG_BP"                         \n\t"
1236                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1237                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1238                    "a" (&c->redDither)
1239             );
1240         }
1241     }
1242 }
1243
1244 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1245                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1246                                 const int16_t *abuf0, uint8_t *dest,
1247                                 int dstW, int uvalpha, int y)
1248 {
1249     const int16_t *ubuf0 = ubuf[0];
1250     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1251
1252     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1253         const int16_t *ubuf1 = ubuf[0];
1254         __asm__ volatile(
1255             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1256             "mov        %4, %%"REG_b"               \n\t"
1257             "push %%"REG_BP"                        \n\t"
1258             YSCALEYUV2RGB1(%%REGBP, %5)
1259             "pxor    %%mm7, %%mm7                   \n\t"
1260             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1261             "pop %%"REG_BP"                         \n\t"
1262             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1263             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1264                "a" (&c->redDither)
1265         );
1266     } else {
1267         const int16_t *ubuf1 = ubuf[1];
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             YSCALEYUV2RGB1b(%%REGBP, %5)
1273             "pxor    %%mm7, %%mm7                   \n\t"
1274             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1275             "pop %%"REG_BP"                         \n\t"
1276             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1277             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1278                "a" (&c->redDither)
1279         );
1280     }
1281 }
1282
1283 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1284                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1285                                  const int16_t *abuf0, uint8_t *dest,
1286                                  int dstW, int uvalpha, int y)
1287 {
1288     const int16_t *ubuf0 = ubuf[0];
1289     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1290
1291     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1292         const int16_t *ubuf1 = ubuf[0];
1293         __asm__ volatile(
1294             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1295             "mov        %4, %%"REG_b"               \n\t"
1296             "push %%"REG_BP"                        \n\t"
1297             YSCALEYUV2RGB1(%%REGBP, %5)
1298             "pxor    %%mm7, %%mm7                   \n\t"
1299             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1300 #ifdef DITHER1XBPP
1301             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1302             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1303             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1304 #endif
1305             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1306             "pop %%"REG_BP"                         \n\t"
1307             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1308             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1309                "a" (&c->redDither)
1310         );
1311     } else {
1312         const int16_t *ubuf1 = ubuf[1];
1313         __asm__ volatile(
1314             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1315             "mov        %4, %%"REG_b"               \n\t"
1316             "push %%"REG_BP"                        \n\t"
1317             YSCALEYUV2RGB1b(%%REGBP, %5)
1318             "pxor    %%mm7, %%mm7                   \n\t"
1319             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1320 #ifdef DITHER1XBPP
1321             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1322             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1323             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1324 #endif
1325             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1326             "pop %%"REG_BP"                         \n\t"
1327             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1328             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1329                "a" (&c->redDither)
1330         );
1331     }
1332 }
1333
1334 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1335                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1336                                  const int16_t *abuf0, uint8_t *dest,
1337                                  int dstW, int uvalpha, int y)
1338 {
1339     const int16_t *ubuf0 = ubuf[0];
1340     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1341
1342     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1343         const int16_t *ubuf1 = ubuf[0];
1344         __asm__ volatile(
1345             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1346             "mov        %4, %%"REG_b"               \n\t"
1347             "push %%"REG_BP"                        \n\t"
1348             YSCALEYUV2RGB1(%%REGBP, %5)
1349             "pxor    %%mm7, %%mm7                   \n\t"
1350             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1351 #ifdef DITHER1XBPP
1352             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1353             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1354             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1355 #endif
1356             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1357             "pop %%"REG_BP"                         \n\t"
1358             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1359             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1360                "a" (&c->redDither)
1361         );
1362     } else {
1363         const int16_t *ubuf1 = ubuf[1];
1364         __asm__ volatile(
1365             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1366             "mov        %4, %%"REG_b"               \n\t"
1367             "push %%"REG_BP"                        \n\t"
1368             YSCALEYUV2RGB1b(%%REGBP, %5)
1369             "pxor    %%mm7, %%mm7                   \n\t"
1370             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1371 #ifdef DITHER1XBPP
1372             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1373             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1374             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1375 #endif
1376             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1377             "pop %%"REG_BP"                         \n\t"
1378             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1379             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1380                "a" (&c->redDither)
1381         );
1382     }
1383 }
1384
1385 #define REAL_YSCALEYUV2PACKED1(index, c) \
1386     "xor            "#index", "#index"  \n\t"\
1387     ".p2align              4            \n\t"\
1388     "1:                                 \n\t"\
1389     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1390     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1391     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1392     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1393     "psraw                $7, %%mm3     \n\t" \
1394     "psraw                $7, %%mm4     \n\t" \
1395     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1396     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1397     "psraw                $7, %%mm1     \n\t" \
1398     "psraw                $7, %%mm7     \n\t" \
1399
1400 #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)
1401
1402 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1403     "xor "#index", "#index"             \n\t"\
1404     ".p2align              4            \n\t"\
1405     "1:                                 \n\t"\
1406     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1407     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1408     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1409     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1410     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1411     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1412     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1413     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1414     "psrlw                $8, %%mm3     \n\t" \
1415     "psrlw                $8, %%mm4     \n\t" \
1416     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1417     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1418     "psraw                $7, %%mm1     \n\t" \
1419     "psraw                $7, %%mm7     \n\t"
1420 #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)
1421
1422 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1423                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
1424                                   const int16_t *abuf0, uint8_t *dest,
1425                                   int dstW, int uvalpha, int y)
1426 {
1427     const int16_t *ubuf0 = ubuf[0];
1428     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1429
1430     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1431         const int16_t *ubuf1 = ubuf[0];
1432         __asm__ volatile(
1433             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1434             "mov        %4, %%"REG_b"               \n\t"
1435             "push %%"REG_BP"                        \n\t"
1436             YSCALEYUV2PACKED1(%%REGBP, %5)
1437             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1438             "pop %%"REG_BP"                         \n\t"
1439             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1440             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1441                "a" (&c->redDither)
1442         );
1443     } else {
1444         const int16_t *ubuf1 = ubuf[1];
1445         __asm__ volatile(
1446             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1447             "mov        %4, %%"REG_b"               \n\t"
1448             "push %%"REG_BP"                        \n\t"
1449             YSCALEYUV2PACKED1b(%%REGBP, %5)
1450             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1451             "pop %%"REG_BP"                         \n\t"
1452             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1453             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1454                "a" (&c->redDither)
1455         );
1456     }
1457 }
1458
1459 #if COMPILE_TEMPLATE_MMXEXT
1460 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1461                                  int dstWidth, const uint8_t *src,
1462                                  int srcW, int xInc)
1463 {
1464     int32_t *filterPos = c->hLumFilterPos;
1465     int16_t *filter    = c->hLumFilter;
1466     void    *mmxextFilterCode = c->lumMmxextFilterCode;
1467     int i;
1468 #if defined(PIC)
1469     uint64_t ebxsave;
1470 #endif
1471 #if ARCH_X86_64
1472     uint64_t retsave;
1473 #endif
1474
1475     __asm__ volatile(
1476 #if defined(PIC)
1477         "mov               %%"REG_b", %5        \n\t"
1478 #if ARCH_X86_64
1479         "mov               -8(%%rsp), %%"REG_a" \n\t"
1480         "mov               %%"REG_a", %6        \n\t"
1481 #endif
1482 #else
1483 #if ARCH_X86_64
1484         "mov               -8(%%rsp), %%"REG_a" \n\t"
1485         "mov               %%"REG_a", %5        \n\t"
1486 #endif
1487 #endif
1488         "pxor                  %%mm7, %%mm7     \n\t"
1489         "mov                      %0, %%"REG_c" \n\t"
1490         "mov                      %1, %%"REG_D" \n\t"
1491         "mov                      %2, %%"REG_d" \n\t"
1492         "mov                      %3, %%"REG_b" \n\t"
1493         "xor               %%"REG_a", %%"REG_a" \n\t" // i
1494         PREFETCH"        (%%"REG_c")            \n\t"
1495         PREFETCH"      32(%%"REG_c")            \n\t"
1496         PREFETCH"      64(%%"REG_c")            \n\t"
1497
1498 #if ARCH_X86_64
1499 #define CALL_MMXEXT_FILTER_CODE \
1500         "movl            (%%"REG_b"), %%esi     \n\t"\
1501         "call                    *%4            \n\t"\
1502         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\
1503         "add               %%"REG_S", %%"REG_c" \n\t"\
1504         "add               %%"REG_a", %%"REG_D" \n\t"\
1505         "xor               %%"REG_a", %%"REG_a" \n\t"\
1506
1507 #else
1508 #define CALL_MMXEXT_FILTER_CODE \
1509         "movl (%%"REG_b"), %%esi        \n\t"\
1510         "call         *%4                       \n\t"\
1511         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1512         "add               %%"REG_a", %%"REG_D" \n\t"\
1513         "xor               %%"REG_a", %%"REG_a" \n\t"\
1514
1515 #endif /* ARCH_X86_64 */
1516
1517         CALL_MMXEXT_FILTER_CODE
1518         CALL_MMXEXT_FILTER_CODE
1519         CALL_MMXEXT_FILTER_CODE
1520         CALL_MMXEXT_FILTER_CODE
1521         CALL_MMXEXT_FILTER_CODE
1522         CALL_MMXEXT_FILTER_CODE
1523         CALL_MMXEXT_FILTER_CODE
1524         CALL_MMXEXT_FILTER_CODE
1525
1526 #if defined(PIC)
1527         "mov                      %5, %%"REG_b" \n\t"
1528 #if ARCH_X86_64
1529         "mov                      %6, %%"REG_a" \n\t"
1530         "mov               %%"REG_a", -8(%%rsp) \n\t"
1531 #endif
1532 #else
1533 #if ARCH_X86_64
1534         "mov                      %5, %%"REG_a" \n\t"
1535         "mov               %%"REG_a", -8(%%rsp) \n\t"
1536 #endif
1537 #endif
1538         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1539            "m" (mmxextFilterCode)
1540 #if defined(PIC)
1541           ,"m" (ebxsave)
1542 #endif
1543 #if ARCH_X86_64
1544           ,"m"(retsave)
1545 #endif
1546         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1547 #if !defined(PIC)
1548          ,"%"REG_b
1549 #endif
1550     );
1551
1552     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1553         dst[i] = src[srcW-1]*128;
1554 }
1555
1556 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1557                                  int dstWidth, const uint8_t *src1,
1558                                  const uint8_t *src2, int srcW, int xInc)
1559 {
1560     int32_t *filterPos = c->hChrFilterPos;
1561     int16_t *filter    = c->hChrFilter;
1562     void    *mmxextFilterCode = c->chrMmxextFilterCode;
1563     int i;
1564 #if defined(PIC)
1565     DECLARE_ALIGNED(8, uint64_t, ebxsave);
1566 #endif
1567 #if ARCH_X86_64
1568     DECLARE_ALIGNED(8, uint64_t, retsave);
1569 #endif
1570
1571     __asm__ volatile(
1572 #if defined(PIC)
1573         "mov          %%"REG_b", %7         \n\t"
1574 #if ARCH_X86_64
1575         "mov          -8(%%rsp), %%"REG_a"  \n\t"
1576         "mov          %%"REG_a", %8         \n\t"
1577 #endif
1578 #else
1579 #if ARCH_X86_64
1580         "mov          -8(%%rsp), %%"REG_a"  \n\t"
1581         "mov          %%"REG_a", %7         \n\t"
1582 #endif
1583 #endif
1584         "pxor             %%mm7, %%mm7      \n\t"
1585         "mov                 %0, %%"REG_c"  \n\t"
1586         "mov                 %1, %%"REG_D"  \n\t"
1587         "mov                 %2, %%"REG_d"  \n\t"
1588         "mov                 %3, %%"REG_b"  \n\t"
1589         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
1590         PREFETCH"   (%%"REG_c")             \n\t"
1591         PREFETCH" 32(%%"REG_c")             \n\t"
1592         PREFETCH" 64(%%"REG_c")             \n\t"
1593
1594         CALL_MMXEXT_FILTER_CODE
1595         CALL_MMXEXT_FILTER_CODE
1596         CALL_MMXEXT_FILTER_CODE
1597         CALL_MMXEXT_FILTER_CODE
1598         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
1599         "mov                 %5, %%"REG_c"  \n\t" // src
1600         "mov                 %6, %%"REG_D"  \n\t" // buf2
1601         PREFETCH"   (%%"REG_c")             \n\t"
1602         PREFETCH" 32(%%"REG_c")             \n\t"
1603         PREFETCH" 64(%%"REG_c")             \n\t"
1604
1605         CALL_MMXEXT_FILTER_CODE
1606         CALL_MMXEXT_FILTER_CODE
1607         CALL_MMXEXT_FILTER_CODE
1608         CALL_MMXEXT_FILTER_CODE
1609
1610 #if defined(PIC)
1611         "mov %7, %%"REG_b"    \n\t"
1612 #if ARCH_X86_64
1613         "mov                 %8, %%"REG_a"  \n\t"
1614         "mov          %%"REG_a", -8(%%rsp)  \n\t"
1615 #endif
1616 #else
1617 #if ARCH_X86_64
1618         "mov                 %7, %%"REG_a"  \n\t"
1619         "mov          %%"REG_a", -8(%%rsp)  \n\t"
1620 #endif
1621 #endif
1622         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1623            "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1624 #if defined(PIC)
1625           ,"m" (ebxsave)
1626 #endif
1627 #if ARCH_X86_64
1628           ,"m"(retsave)
1629 #endif
1630         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1631 #if !defined(PIC)
1632          ,"%"REG_b
1633 #endif
1634     );
1635
1636     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1637         dst1[i] = src1[srcW-1]*128;
1638         dst2[i] = src2[srcW-1]*128;
1639     }
1640 }
1641 #endif /* COMPILE_TEMPLATE_MMXEXT */
1642
1643 static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
1644 {
1645     enum AVPixelFormat dstFormat = c->dstFormat;
1646
1647     c->use_mmx_vfilter= 0;
1648     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != AV_PIX_FMT_NV12
1649         && dstFormat != AV_PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
1650             if (c->flags & SWS_ACCURATE_RND) {
1651                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1652                     switch (c->dstFormat) {
1653                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;
1654                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;
1655                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;
1656                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;
1657                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1658                     default: break;
1659                     }
1660                 }
1661             } else {
1662                 c->use_mmx_vfilter= 1;
1663                 c->yuv2planeX = RENAME(yuv2yuvX    );
1664                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1665                     switch (c->dstFormat) {
1666                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;
1667                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;
1668                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;
1669                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;
1670                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1671                     default: break;
1672                     }
1673                 }
1674             }
1675         if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1676             switch (c->dstFormat) {
1677             case AV_PIX_FMT_RGB32:
1678                 c->yuv2packed1 = RENAME(yuv2rgb32_1);
1679                 c->yuv2packed2 = RENAME(yuv2rgb32_2);
1680                 break;
1681             case AV_PIX_FMT_BGR24:
1682                 c->yuv2packed1 = RENAME(yuv2bgr24_1);
1683                 c->yuv2packed2 = RENAME(yuv2bgr24_2);
1684                 break;
1685             case AV_PIX_FMT_RGB555:
1686                 c->yuv2packed1 = RENAME(yuv2rgb555_1);
1687                 c->yuv2packed2 = RENAME(yuv2rgb555_2);
1688                 break;
1689             case AV_PIX_FMT_RGB565:
1690                 c->yuv2packed1 = RENAME(yuv2rgb565_1);
1691                 c->yuv2packed2 = RENAME(yuv2rgb565_2);
1692                 break;
1693             case AV_PIX_FMT_YUYV422:
1694                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1695                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1696                 break;
1697             default:
1698                 break;
1699             }
1700         }
1701     }
1702
1703     if (c->srcBpc == 8 && c->dstBpc <= 14) {
1704     // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1705 #if COMPILE_TEMPLATE_MMXEXT
1706     if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1707         c->hyscale_fast = RENAME(hyscale_fast);
1708         c->hcscale_fast = RENAME(hcscale_fast);
1709     } else {
1710 #endif /* COMPILE_TEMPLATE_MMXEXT */
1711         c->hyscale_fast = NULL;
1712         c->hcscale_fast = NULL;
1713 #if COMPILE_TEMPLATE_MMXEXT
1714     }
1715 #endif /* COMPILE_TEMPLATE_MMXEXT */
1716     }
1717 }