libswscale/x86/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
   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 PREFETCH
  24
  25 #if COMPILE_TEMPLATE_MMX2
  26 #define PREFETCH "prefetchnta"
  27 #else
  28 #define PREFETCH  " # nop"
  29 #endif
  30
  31 #if COMPILE_TEMPLATE_MMX2
  32 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
  33 #else
  34 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
  35 #endif
  36 #define MOVNTQ(a,b)  REAL_MOVNTQ(a,b)
  37
  38 #define YSCALEYUV2YV12X(offset, dest, end, pos) \
  39     __asm__ volatile(\
  40         "movq                  "DITHER16"+0(%0), %%mm3      \n\t"\
  41         "movq                  "DITHER16"+8(%0), %%mm4      \n\t"\
  42         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
  43         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  44         ".p2align                             4             \n\t" /* FIXME Unroll? */\
  45         "1:                                                 \n\t"\
  46         "movq                      8(%%"REG_d"), %%mm0      \n\t" /* filterCoeff */\
  47         "movq                (%%"REG_S", %3, 2), %%mm2      \n\t" /* srcData */\
  48         "movq               8(%%"REG_S", %3, 2), %%mm5      \n\t" /* srcData */\
  49         "add                                $16, %%"REG_d"  \n\t"\
  50         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  51         "test                         %%"REG_S", %%"REG_S"  \n\t"\
  52         "pmulhw                           %%mm0, %%mm2      \n\t"\
  53         "pmulhw                           %%mm0, %%mm5      \n\t"\
  54         "paddw                            %%mm2, %%mm3      \n\t"\
  55         "paddw                            %%mm5, %%mm4      \n\t"\
  56         " jnz                                1b             \n\t"\
  57         "psraw                               $3, %%mm3      \n\t"\
  58         "psraw                               $3, %%mm4      \n\t"\
  59         "packuswb                         %%mm4, %%mm3      \n\t"\
  60         MOVNTQ(%%mm3, (%1, %3))\
  61         "add                                 $8, %3         \n\t"\
  62         "cmp                                 %2, %3         \n\t"\
  63         "movq                  "DITHER16"+0(%0), %%mm3      \n\t"\
  64         "movq                  "DITHER16"+8(%0), %%mm4      \n\t"\
  65         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
  66         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
  67         "jb                                  1b             \n\t"\
  68         :: "r" (&c->redDither),\
  69            "r" (dest), "g" ((x86_reg)(end)), "r"((x86_reg)(pos))\
  70         : "%"REG_d, "%"REG_S\
  71     );
  72
  73 #if !COMPILE_TEMPLATE_MMX2
  74 static av_always_inline void
  75 dither_8to16(SwsContext *c, const uint8_t *srcDither, int rot)
  76 {
  77     if (rot) {
  78         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
  79                          "movq       (%0), %%mm3\n\t"
  80                          "movq      %%mm3, %%mm4\n\t"
  81                          "psrlq       $24, %%mm3\n\t"
  82                          "psllq       $40, %%mm4\n\t"
  83                          "por       %%mm4, %%mm3\n\t"
  84                          "movq      %%mm3, %%mm4\n\t"
  85                          "punpcklbw %%mm0, %%mm3\n\t"
  86                          "punpckhbw %%mm0, %%mm4\n\t"
  87                          "psraw        $4, %%mm3\n\t"
  88                          "psraw        $4, %%mm4\n\t"
  89                          "movq      %%mm3, "DITHER16"+0(%1)\n\t"
  90                          "movq      %%mm4, "DITHER16"+8(%1)\n\t"
  91                          :: "r"(srcDither), "r"(&c->redDither)
  92                          );
  93     } else {
  94         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
  95                          "movq       (%0), %%mm3\n\t"
  96                          "movq      %%mm3, %%mm4\n\t"
  97                          "punpcklbw %%mm0, %%mm3\n\t"
  98                          "punpckhbw %%mm0, %%mm4\n\t"
  99                          "psraw        $4, %%mm3\n\t"
 100                          "psraw        $4, %%mm4\n\t"
 101                          "movq      %%mm3, "DITHER16"+0(%1)\n\t"
 102                          "movq      %%mm4, "DITHER16"+8(%1)\n\t"
 103                          :: "r"(srcDither), "r"(&c->redDither)
 104                          );
 105     }
 106 }
 107 #endif
 108
 109 static void RENAME(yuv2yuvX)(SwsContext *c, const int16_t *lumFilter,
 110                              const int16_t **lumSrc, int lumFilterSize,
 111                              const int16_t *chrFilter, const int16_t **chrUSrc,
 112                              const int16_t **chrVSrc,
 113                              int chrFilterSize, const int16_t **alpSrc,
 114                              uint8_t *dest[4], int dstW, int chrDstW)
 115 {
 116     int i;
 117     uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2],
 118             *aDest = CONFIG_SWSCALE_ALPHA ? dest[3] : NULL;
 119     const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;
 120
 121     if (uDest) {
 122         x86_reg uv_off = c->uv_offx2 >> 1;
 123         dither_8to16(c, chrDither, 0);
 124         YSCALEYUV2YV12X(CHR_MMX_FILTER_OFFSET, uDest, chrDstW, 0)
 125         dither_8to16(c, chrDither, 1);
 126         YSCALEYUV2YV12X(CHR_MMX_FILTER_OFFSET, vDest - uv_off, chrDstW + uv_off, uv_off)
 127     }
 128     dither_8to16(c, lumDither, 0);
 129     if (CONFIG_SWSCALE_ALPHA && aDest) {
 130         YSCALEYUV2YV12X(ALP_MMX_FILTER_OFFSET, aDest, dstW, 0)
 131     }
 132
 133     YSCALEYUV2YV12X(LUM_MMX_FILTER_OFFSET, yDest, dstW, 0)
 134 }
 135
 136 #define YSCALEYUV2YV12X_ACCURATE(offset, dest, end, pos) \
 137     __asm__ volatile(\
 138         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
 139         "movq                  "DITHER32"+0(%0), %%mm4      \n\t"\
 140         "movq                  "DITHER32"+8(%0), %%mm5      \n\t"\
 141         "movq                 "DITHER32"+16(%0), %%mm6      \n\t"\
 142         "movq                 "DITHER32"+24(%0), %%mm7      \n\t"\
 143         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 144         ".p2align                             4             \n\t"\
 145         "1:                                                 \n\t"\
 146         "movq                (%%"REG_S", %3, 2), %%mm0      \n\t" /* srcData */\
 147         "movq               8(%%"REG_S", %3, 2), %%mm2      \n\t" /* srcData */\
 148         "mov        "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"  \n\t"\
 149         "movq                (%%"REG_S", %3, 2), %%mm1      \n\t" /* srcData */\
 150         "movq                             %%mm0, %%mm3      \n\t"\
 151         "punpcklwd                        %%mm1, %%mm0      \n\t"\
 152         "punpckhwd                        %%mm1, %%mm3      \n\t"\
 153         "movq       "STR(APCK_COEF)"(%%"REG_d"), %%mm1      \n\t" /* filterCoeff */\
 154         "pmaddwd                          %%mm1, %%mm0      \n\t"\
 155         "pmaddwd                          %%mm1, %%mm3      \n\t"\
 156         "paddd                            %%mm0, %%mm4      \n\t"\
 157         "paddd                            %%mm3, %%mm5      \n\t"\
 158         "movq               8(%%"REG_S", %3, 2), %%mm3      \n\t" /* srcData */\
 159         "mov        "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"  \n\t"\
 160         "add                  $"STR(APCK_SIZE)", %%"REG_d"  \n\t"\
 161         "test                         %%"REG_S", %%"REG_S"  \n\t"\
 162         "movq                             %%mm2, %%mm0      \n\t"\
 163         "punpcklwd                        %%mm3, %%mm2      \n\t"\
 164         "punpckhwd                        %%mm3, %%mm0      \n\t"\
 165         "pmaddwd                          %%mm1, %%mm2      \n\t"\
 166         "pmaddwd                          %%mm1, %%mm0      \n\t"\
 167         "paddd                            %%mm2, %%mm6      \n\t"\
 168         "paddd                            %%mm0, %%mm7      \n\t"\
 169         " jnz                                1b             \n\t"\
 170         "psrad                              $19, %%mm4      \n\t"\
 171         "psrad                              $19, %%mm5      \n\t"\
 172         "psrad                              $19, %%mm6      \n\t"\
 173         "psrad                              $19, %%mm7      \n\t"\
 174         "packssdw                         %%mm5, %%mm4      \n\t"\
 175         "packssdw                         %%mm7, %%mm6      \n\t"\
 176         "packuswb                         %%mm6, %%mm4      \n\t"\
 177         MOVNTQ(%%mm4, (%1, %3))\
 178         "add                                 $8, %3         \n\t"\
 179         "cmp                                 %2, %3         \n\t"\
 180         "lea                     " offset "(%0), %%"REG_d"  \n\t"\
 181         "movq                  "DITHER32"+0(%0), %%mm4      \n\t"\
 182         "movq                  "DITHER32"+8(%0), %%mm5      \n\t"\
 183         "movq                 "DITHER32"+16(%0), %%mm6      \n\t"\
 184         "movq                 "DITHER32"+24(%0), %%mm7      \n\t"\
 185         "mov                        (%%"REG_d"), %%"REG_S"  \n\t"\
 186         "jb                                  1b             \n\t"\
 187         :: "r" (&c->redDither),\
 188         "r" (dest), "g" ((x86_reg)(end)), "r"((x86_reg)(pos))\
 189         : "%"REG_a, "%"REG_d, "%"REG_S\
 190     );
 191
 192 #if !COMPILE_TEMPLATE_MMX2
 193 static av_always_inline void
 194 dither_8to32(SwsContext *c, const uint8_t *srcDither, int rot)
 195 {
 196 int i;
 197 if(rot) for(i=0; i<8; i++) c->dither32[i] = srcDither[(i+3)&7]<<12;
 198 else    for(i=0; i<8; i++) c->dither32[i] = srcDither[i&7]<<12;
 199 return;
 200
 201     if (rot) {
 202         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
 203                          "movq       (%0), %%mm4\n\t"
 204                          "movq      %%mm4, %%mm5\n\t"
 205                          "psrlq       $24, %%mm4\n\t"
 206                          "psllq       $40, %%mm5\n\t"
 207                          "por       %%mm5, %%mm4\n\t"
 208                          "movq      %%mm4, %%mm6\n\t"
 209                          "punpcklbw %%mm0, %%mm4\n\t"
 210                          "punpckhbw %%mm0, %%mm6\n\t"
 211                          "movq      %%mm4, %%mm5\n\t"
 212                          "movq      %%mm6, %%mm7\n\t"
 213                          "punpcklwd %%mm0, %%mm4\n\t"
 214                          "punpckhwd %%mm0, %%mm5\n\t"
 215                          "punpcklwd %%mm0, %%mm6\n\t"
 216                          "punpckhwd %%mm0, %%mm7\n\t"
 217                          "psllw       $12, %%mm4\n\t"
 218                          "psllw       $12, %%mm5\n\t"
 219                          "psllw       $12, %%mm6\n\t"
 220                          "psllw       $12, %%mm7\n\t"
 221                          "movq      %%mm4, "DITHER32"+0(%1)\n\t"
 222                          "movq      %%mm5, "DITHER32"+8(%1)\n\t"
 223                          "movq      %%mm6, "DITHER32"+16(%1)\n\t"
 224                          "movq      %%mm7, "DITHER32"+24(%1)\n\t"
 225                          :: "r"(srcDither), "r"(&c->redDither)
 226                          );
 227     } else {
 228         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"
 229                          "movq       (%0), %%mm4\n\t"
 230                          "movq      %%mm4, %%mm6\n\t"
 231                          "punpcklbw %%mm0, %%mm4\n\t"
 232                          "punpckhbw %%mm0, %%mm6\n\t"
 233                          "movq      %%mm4, %%mm5\n\t"
 234                          "movq      %%mm6, %%mm7\n\t"
 235                          "punpcklwd %%mm0, %%mm4\n\t"
 236                          "punpckhwd %%mm0, %%mm5\n\t"
 237                          "punpcklwd %%mm0, %%mm6\n\t"
 238                          "punpckhwd %%mm0, %%mm7\n\t"
 239                          "psllw       $12, %%mm4\n\t"
 240                          "psllw       $12, %%mm5\n\t"
 241                          "psllw       $12, %%mm6\n\t"
 242                          "psllw       $12, %%mm7\n\t"
 243                          "movq      %%mm4, "DITHER32"+0(%1)\n\t"
 244                          "movq      %%mm5, "DITHER32"+8(%1)\n\t"
 245                          "movq      %%mm6, "DITHER32"+16(%1)\n\t"
 246                          "movq      %%mm7, "DITHER32"+24(%1)\n\t"
 247                          :: "r"(srcDither), "r"(&c->redDither)
 248                          );
 249     }
 250 }
 251 #endif
 252
 253 static void RENAME(yuv2yuvX_ar)(SwsContext *c, const int16_t *lumFilter,
 254                                 const int16_t **lumSrc, int lumFilterSize,
 255                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 256                                 const int16_t **chrVSrc,
 257                                 int chrFilterSize, const int16_t **alpSrc,
 258                                 uint8_t *dest[4], int dstW, int chrDstW)
 259 {
 260     int i;
 261     uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2],
 262             *aDest = CONFIG_SWSCALE_ALPHA ? dest[3] : NULL;
 263     const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;
 264
 265     if (uDest) {
 266         x86_reg uv_off = c->uv_offx2 >> 1;
 267         dither_8to32(c, chrDither, 0);
 268         YSCALEYUV2YV12X_ACCURATE(CHR_MMX_FILTER_OFFSET, uDest, chrDstW, 0)
 269         dither_8to32(c, chrDither, 1);
 270         YSCALEYUV2YV12X_ACCURATE(CHR_MMX_FILTER_OFFSET, vDest - uv_off, chrDstW + uv_off, uv_off)
 271     }
 272     dither_8to32(c, lumDither, 0);
 273     if (CONFIG_SWSCALE_ALPHA && aDest) {
 274         YSCALEYUV2YV12X_ACCURATE(ALP_MMX_FILTER_OFFSET, aDest, dstW, 0)
 275     }
 276
 277     YSCALEYUV2YV12X_ACCURATE(LUM_MMX_FILTER_OFFSET, yDest, dstW, 0)
 278 }
 279
 280 static void RENAME(yuv2yuv1)(SwsContext *c, const int16_t *lumSrc,
 281                              const int16_t *chrUSrc, const int16_t *chrVSrc,
 282                              const int16_t *alpSrc,
 283                              uint8_t *dst[4], int dstW, int chrDstW)
 284 {
 285     int p= 4;
 286     const int16_t *src[4]= {
 287         lumSrc + dstW,     chrUSrc + chrDstW,
 288         chrVSrc + chrDstW, alpSrc + dstW
 289     };
 290     x86_reg counter[4]= { dstW, chrDstW, chrDstW, dstW };
 291
 292     while (p--) {
 293         if (dst[p]) {
 294             __asm__ volatile(
 295                 "mov %2, %%"REG_a"                    \n\t"
 296                 ".p2align               4             \n\t" /* FIXME Unroll? */
 297                 "1:                                   \n\t"
 298                 "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"
 299                 "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"
 300                 "psraw                 $7, %%mm0      \n\t"
 301                 "psraw                 $7, %%mm1      \n\t"
 302                 "packuswb           %%mm1, %%mm0      \n\t"
 303                 MOVNTQ(%%mm0, (%1, %%REGa))
 304                 "add                   $8, %%"REG_a"  \n\t"
 305                 "jnc                   1b             \n\t"
 306                 :: "r" (src[p]), "r" (dst[p] + counter[p]),
 307                    "g" (-counter[p])
 308                 : "%"REG_a
 309             );
 310         }
 311     }
 312 }
 313
 314 static void RENAME(yuv2yuv1_ar)(SwsContext *c, const int16_t *lumSrc,
 315                                 const int16_t *chrUSrc, const int16_t *chrVSrc,
 316                                 const int16_t *alpSrc,
 317                                 uint8_t *dst[4], int dstW, int chrDstW)
 318 {
 319     int p= 4;
 320     const int16_t *src[4]= {
 321         lumSrc + dstW,     chrUSrc + chrDstW,
 322         chrVSrc + chrDstW, alpSrc + dstW
 323     };
 324     x86_reg counter[4]= { dstW, chrDstW, chrDstW, dstW };
 325     const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;
 326
 327     while (p--) {
 328         if (dst[p]) {
 329             int i;
 330             for(i=0; i<8; i++) c->dither16[i] = (p == 2 || p == 3) ? lumDither[i] : chrDither[i];
 331             __asm__ volatile(
 332                 "mov %2, %%"REG_a"                    \n\t"
 333                 "movq    "DITHER16"+0(%3), %%mm6      \n\t"
 334                 "movq    "DITHER16"+8(%3), %%mm7      \n\t"
 335                 ".p2align                4            \n\t" /* FIXME Unroll? */
 336                 "1:                                   \n\t"
 337                 "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"
 338                 "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"
 339                 "paddsw             %%mm6, %%mm0      \n\t"
 340                 "paddsw             %%mm7, %%mm1      \n\t"
 341                 "psraw                 $7, %%mm0      \n\t"
 342                 "psraw                 $7, %%mm1      \n\t"
 343                 "packuswb           %%mm1, %%mm0      \n\t"
 344                 MOVNTQ(%%mm0, (%1, %%REGa))
 345                 "add                   $8, %%"REG_a"  \n\t"
 346                 "jnc                   1b             \n\t"
 347                 :: "r" (src[p]), "r" (dst[p] + counter[p]),
 348                    "g" (-counter[p]), "r"(&c->redDither)
 349                 : "%"REG_a
 350             );
 351         }
 352     }
 353 }
 354
 355 #define YSCALEYUV2PACKEDX_UV \
 356     __asm__ volatile(\
 357         "xor                   %%"REG_a", %%"REG_a"     \n\t"\
 358         ".p2align                      4                \n\t"\
 359         "nop                                            \n\t"\
 360         "1:                                             \n\t"\
 361         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 362         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 363         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\
 364         "movq                      %%mm3, %%mm4         \n\t"\
 365         ".p2align                      4                \n\t"\
 366         "2:                                             \n\t"\
 367         "movq               8(%%"REG_d"), %%mm0         \n\t" /* filterCoeff */\
 368         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* UsrcData */\
 369         "add                          %6, %%"REG_S"     \n\t" \
 370         "movq     (%%"REG_S", %%"REG_a"), %%mm5         \n\t" /* VsrcData */\
 371         "add                         $16, %%"REG_d"     \n\t"\
 372         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 373         "pmulhw                    %%mm0, %%mm2         \n\t"\
 374         "pmulhw                    %%mm0, %%mm5         \n\t"\
 375         "paddw                     %%mm2, %%mm3         \n\t"\
 376         "paddw                     %%mm5, %%mm4         \n\t"\
 377         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 378         " jnz                         2b                \n\t"\
 379
 380 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
 381     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 382     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 383     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\
 384     "movq                    "#dst1", "#dst2"       \n\t"\
 385     ".p2align                      4                \n\t"\
 386     "2:                                             \n\t"\
 387     "movq               8(%%"REG_d"), "#coeff"      \n\t" /* filterCoeff */\
 388     "movq  (%%"REG_S", %%"REG_a", 2), "#src1"       \n\t" /* Y1srcData */\
 389     "movq 8(%%"REG_S", %%"REG_a", 2), "#src2"       \n\t" /* Y2srcData */\
 390     "add                         $16, %%"REG_d"            \n\t"\
 391     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 392     "pmulhw                 "#coeff", "#src1"       \n\t"\
 393     "pmulhw                 "#coeff", "#src2"       \n\t"\
 394     "paddw                   "#src1", "#dst1"       \n\t"\
 395     "paddw                   "#src2", "#dst2"       \n\t"\
 396     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 397     " jnz                         2b                \n\t"\
 398
 399 #define YSCALEYUV2PACKEDX \
 400     YSCALEYUV2PACKEDX_UV \
 401     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
 402
 403 #define YSCALEYUV2PACKEDX_END                     \
 404         :: "r" (&c->redDither),                   \
 405             "m" (dummy), "m" (dummy), "m" (dummy),\
 406             "r" (dest), "m" (dstW_reg), "m"(uv_off) \
 407         : "%"REG_a, "%"REG_d, "%"REG_S            \
 408     );
 409
 410 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
 411     __asm__ volatile(\
 412         "xor %%"REG_a", %%"REG_a"                       \n\t"\
 413         ".p2align                      4                \n\t"\
 414         "nop                                            \n\t"\
 415         "1:                                             \n\t"\
 416         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 417         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 418         "pxor                      %%mm4, %%mm4         \n\t"\
 419         "pxor                      %%mm5, %%mm5         \n\t"\
 420         "pxor                      %%mm6, %%mm6         \n\t"\
 421         "pxor                      %%mm7, %%mm7         \n\t"\
 422         ".p2align                      4                \n\t"\
 423         "2:                                             \n\t"\
 424         "movq     (%%"REG_S", %%"REG_a"), %%mm0         \n\t" /* UsrcData */\
 425         "add                          %6, %%"REG_S"      \n\t" \
 426         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* VsrcData */\
 427         "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 428         "movq     (%%"REG_S", %%"REG_a"), %%mm1         \n\t" /* UsrcData */\
 429         "movq                      %%mm0, %%mm3         \n\t"\
 430         "punpcklwd                 %%mm1, %%mm0         \n\t"\
 431         "punpckhwd                 %%mm1, %%mm3         \n\t"\
 432         "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1         \n\t" /* filterCoeff */\
 433         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 434         "pmaddwd                   %%mm1, %%mm3         \n\t"\
 435         "paddd                     %%mm0, %%mm4         \n\t"\
 436         "paddd                     %%mm3, %%mm5         \n\t"\
 437         "add                          %6, %%"REG_S"      \n\t" \
 438         "movq     (%%"REG_S", %%"REG_a"), %%mm3         \n\t" /* VsrcData */\
 439         "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 440         "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 441         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 442         "movq                      %%mm2, %%mm0         \n\t"\
 443         "punpcklwd                 %%mm3, %%mm2         \n\t"\
 444         "punpckhwd                 %%mm3, %%mm0         \n\t"\
 445         "pmaddwd                   %%mm1, %%mm2         \n\t"\
 446         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 447         "paddd                     %%mm2, %%mm6         \n\t"\
 448         "paddd                     %%mm0, %%mm7         \n\t"\
 449         " jnz                         2b                \n\t"\
 450         "psrad                       $16, %%mm4         \n\t"\
 451         "psrad                       $16, %%mm5         \n\t"\
 452         "psrad                       $16, %%mm6         \n\t"\
 453         "psrad                       $16, %%mm7         \n\t"\
 454         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 455         "packssdw                  %%mm5, %%mm4         \n\t"\
 456         "packssdw                  %%mm7, %%mm6         \n\t"\
 457         "paddw                     %%mm0, %%mm4         \n\t"\
 458         "paddw                     %%mm0, %%mm6         \n\t"\
 459         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\
 460         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\
 461
 462 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
 463     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 464     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 465     "pxor                      %%mm1, %%mm1         \n\t"\
 466     "pxor                      %%mm5, %%mm5         \n\t"\
 467     "pxor                      %%mm7, %%mm7         \n\t"\
 468     "pxor                      %%mm6, %%mm6         \n\t"\
 469     ".p2align                      4                \n\t"\
 470     "2:                                             \n\t"\
 471     "movq  (%%"REG_S", %%"REG_a", 2), %%mm0         \n\t" /* Y1srcData */\
 472     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y2srcData */\
 473     "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 474     "movq  (%%"REG_S", %%"REG_a", 2), %%mm4         \n\t" /* Y1srcData */\
 475     "movq                      %%mm0, %%mm3         \n\t"\
 476     "punpcklwd                 %%mm4, %%mm0         \n\t"\
 477     "punpckhwd                 %%mm4, %%mm3         \n\t"\
 478     "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4         \n\t" /* filterCoeff */\
 479     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 480     "pmaddwd                   %%mm4, %%mm3         \n\t"\
 481     "paddd                     %%mm0, %%mm1         \n\t"\
 482     "paddd                     %%mm3, %%mm5         \n\t"\
 483     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3         \n\t" /* Y2srcData */\
 484     "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 485     "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 486     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 487     "movq                      %%mm2, %%mm0         \n\t"\
 488     "punpcklwd                 %%mm3, %%mm2         \n\t"\
 489     "punpckhwd                 %%mm3, %%mm0         \n\t"\
 490     "pmaddwd                   %%mm4, %%mm2         \n\t"\
 491     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 492     "paddd                     %%mm2, %%mm7         \n\t"\
 493     "paddd                     %%mm0, %%mm6         \n\t"\
 494     " jnz                         2b                \n\t"\
 495     "psrad                       $16, %%mm1         \n\t"\
 496     "psrad                       $16, %%mm5         \n\t"\
 497     "psrad                       $16, %%mm7         \n\t"\
 498     "psrad                       $16, %%mm6         \n\t"\
 499     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 500     "packssdw                  %%mm5, %%mm1         \n\t"\
 501     "packssdw                  %%mm6, %%mm7         \n\t"\
 502     "paddw                     %%mm0, %%mm1         \n\t"\
 503     "paddw                     %%mm0, %%mm7         \n\t"\
 504     "movq               "U_TEMP"(%0), %%mm3         \n\t"\
 505     "movq               "V_TEMP"(%0), %%mm4         \n\t"\
 506
 507 #define YSCALEYUV2PACKEDX_ACCURATE \
 508     YSCALEYUV2PACKEDX_ACCURATE_UV \
 509     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
 510
 511 #define YSCALEYUV2RGBX \
 512     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\
 513     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\
 514     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\
 515     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\
 516     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\
 517     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\
 518     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 519     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\
 520     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\
 521     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\
 522     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\
 523     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\
 524     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\
 525     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 526     "paddw           %%mm3, %%mm4       \n\t"\
 527     "movq            %%mm2, %%mm0       \n\t"\
 528     "movq            %%mm5, %%mm6       \n\t"\
 529     "movq            %%mm4, %%mm3       \n\t"\
 530     "punpcklwd       %%mm2, %%mm2       \n\t"\
 531     "punpcklwd       %%mm5, %%mm5       \n\t"\
 532     "punpcklwd       %%mm4, %%mm4       \n\t"\
 533     "paddw           %%mm1, %%mm2       \n\t"\
 534     "paddw           %%mm1, %%mm5       \n\t"\
 535     "paddw           %%mm1, %%mm4       \n\t"\
 536     "punpckhwd       %%mm0, %%mm0       \n\t"\
 537     "punpckhwd       %%mm6, %%mm6       \n\t"\
 538     "punpckhwd       %%mm3, %%mm3       \n\t"\
 539     "paddw           %%mm7, %%mm0       \n\t"\
 540     "paddw           %%mm7, %%mm6       \n\t"\
 541     "paddw           %%mm7, %%mm3       \n\t"\
 542     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 543     "packuswb        %%mm0, %%mm2       \n\t"\
 544     "packuswb        %%mm6, %%mm5       \n\t"\
 545     "packuswb        %%mm3, %%mm4       \n\t"\
 546
 547 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
 548     "movq       "#b", "#q2"     \n\t" /* B */\
 549     "movq       "#r", "#t"      \n\t" /* R */\
 550     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\
 551     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\
 552     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\
 553     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\
 554     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\
 555     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\
 556     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\
 557     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\
 558     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\
 559     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\
 560 \
 561     MOVNTQ(   q0,   (dst, index, 4))\
 562     MOVNTQ(    b,  8(dst, index, 4))\
 563     MOVNTQ(   q2, 16(dst, index, 4))\
 564     MOVNTQ(   q3, 24(dst, index, 4))\
 565 \
 566     "add      $8, "#index"      \n\t"\
 567     "cmp "#dstw", "#index"      \n\t"\
 568     " jb      1b                \n\t"
 569 #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)
 570
 571 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
 572                                    const int16_t **lumSrc, int lumFilterSize,
 573                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 574                                    const int16_t **chrVSrc,
 575                                    int chrFilterSize, const int16_t **alpSrc,
 576                                    uint8_t *dest, int dstW, int dstY)
 577 {
 578     x86_reg dummy=0;
 579     x86_reg dstW_reg = dstW;
 580     x86_reg uv_off = c->uv_offx2;
 581
 582     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 583         YSCALEYUV2PACKEDX_ACCURATE
 584         YSCALEYUV2RGBX
 585         "movq                      %%mm2, "U_TEMP"(%0)  \n\t"
 586         "movq                      %%mm4, "V_TEMP"(%0)  \n\t"
 587         "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"
 588         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
 589         "movq               "Y_TEMP"(%0), %%mm5         \n\t"
 590         "psraw                        $3, %%mm1         \n\t"
 591         "psraw                        $3, %%mm7         \n\t"
 592         "packuswb                  %%mm7, %%mm1         \n\t"
 593         WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
 594         YSCALEYUV2PACKEDX_END
 595     } else {
 596         YSCALEYUV2PACKEDX_ACCURATE
 597         YSCALEYUV2RGBX
 598         "pcmpeqd %%mm7, %%mm7 \n\t"
 599         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 600         YSCALEYUV2PACKEDX_END
 601     }
 602 }
 603
 604 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
 605                                 const int16_t **lumSrc, int lumFilterSize,
 606                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 607                                 const int16_t **chrVSrc,
 608                                 int chrFilterSize, const int16_t **alpSrc,
 609                                 uint8_t *dest, int dstW, int dstY)
 610 {
 611     x86_reg dummy=0;
 612     x86_reg dstW_reg = dstW;
 613     x86_reg uv_off = c->uv_offx2;
 614
 615     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 616         YSCALEYUV2PACKEDX
 617         YSCALEYUV2RGBX
 618         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
 619         "psraw                        $3, %%mm1         \n\t"
 620         "psraw                        $3, %%mm7         \n\t"
 621         "packuswb                  %%mm7, %%mm1         \n\t"
 622         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 623         YSCALEYUV2PACKEDX_END
 624     } else {
 625         YSCALEYUV2PACKEDX
 626         YSCALEYUV2RGBX
 627         "pcmpeqd %%mm7, %%mm7 \n\t"
 628         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 629         YSCALEYUV2PACKEDX_END
 630     }
 631 }
 632
 633 #define REAL_WRITERGB16(dst, dstw, index) \
 634     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 635     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
 636     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 637     "psrlq           $3, %%mm2  \n\t"\
 638 \
 639     "movq         %%mm2, %%mm1  \n\t"\
 640     "movq         %%mm4, %%mm3  \n\t"\
 641 \
 642     "punpcklbw    %%mm7, %%mm3  \n\t"\
 643     "punpcklbw    %%mm5, %%mm2  \n\t"\
 644     "punpckhbw    %%mm7, %%mm4  \n\t"\
 645     "punpckhbw    %%mm5, %%mm1  \n\t"\
 646 \
 647     "psllq           $3, %%mm3  \n\t"\
 648     "psllq           $3, %%mm4  \n\t"\
 649 \
 650     "por          %%mm3, %%mm2  \n\t"\
 651     "por          %%mm4, %%mm1  \n\t"\
 652 \
 653     MOVNTQ(%%mm2,  (dst, index, 2))\
 654     MOVNTQ(%%mm1, 8(dst, index, 2))\
 655 \
 656     "add             $8, "#index"   \n\t"\
 657     "cmp        "#dstw", "#index"   \n\t"\
 658     " jb             1b             \n\t"
 659 #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)
 660
 661 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
 662                                     const int16_t **lumSrc, int lumFilterSize,
 663                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 664                                     const int16_t **chrVSrc,
 665                                     int chrFilterSize, const int16_t **alpSrc,
 666                                     uint8_t *dest, int dstW, int dstY)
 667 {
 668     x86_reg dummy=0;
 669     x86_reg dstW_reg = dstW;
 670     x86_reg uv_off = c->uv_offx2;
 671
 672     YSCALEYUV2PACKEDX_ACCURATE
 673     YSCALEYUV2RGBX
 674     "pxor %%mm7, %%mm7 \n\t"
 675     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 676 #ifdef DITHER1XBPP
 677     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 678     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 679     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 680 #endif
 681     WRITERGB16(%4, %5, %%REGa)
 682     YSCALEYUV2PACKEDX_END
 683 }
 684
 685 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
 686                                  const int16_t **lumSrc, int lumFilterSize,
 687                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 688                                  const int16_t **chrVSrc,
 689                                  int chrFilterSize, const int16_t **alpSrc,
 690                                  uint8_t *dest, int dstW, int dstY)
 691 {
 692     x86_reg dummy=0;
 693     x86_reg dstW_reg = dstW;
 694     x86_reg uv_off = c->uv_offx2;
 695
 696     YSCALEYUV2PACKEDX
 697     YSCALEYUV2RGBX
 698     "pxor %%mm7, %%mm7 \n\t"
 699     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 700 #ifdef DITHER1XBPP
 701     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 702     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 703     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 704 #endif
 705     WRITERGB16(%4, %5, %%REGa)
 706     YSCALEYUV2PACKEDX_END
 707 }
 708
 709 #define REAL_WRITERGB15(dst, dstw, index) \
 710     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 711     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\
 712     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 713     "psrlq           $3, %%mm2  \n\t"\
 714     "psrlq           $1, %%mm5  \n\t"\
 715 \
 716     "movq         %%mm2, %%mm1  \n\t"\
 717     "movq         %%mm4, %%mm3  \n\t"\
 718 \
 719     "punpcklbw    %%mm7, %%mm3  \n\t"\
 720     "punpcklbw    %%mm5, %%mm2  \n\t"\
 721     "punpckhbw    %%mm7, %%mm4  \n\t"\
 722     "punpckhbw    %%mm5, %%mm1  \n\t"\
 723 \
 724     "psllq           $2, %%mm3  \n\t"\
 725     "psllq           $2, %%mm4  \n\t"\
 726 \
 727     "por          %%mm3, %%mm2  \n\t"\
 728     "por          %%mm4, %%mm1  \n\t"\
 729 \
 730     MOVNTQ(%%mm2,  (dst, index, 2))\
 731     MOVNTQ(%%mm1, 8(dst, index, 2))\
 732 \
 733     "add             $8, "#index"   \n\t"\
 734     "cmp        "#dstw", "#index"   \n\t"\
 735     " jb             1b             \n\t"
 736 #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)
 737
 738 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
 739                                     const int16_t **lumSrc, int lumFilterSize,
 740                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 741                                     const int16_t **chrVSrc,
 742                                     int chrFilterSize, const int16_t **alpSrc,
 743                                     uint8_t *dest, int dstW, int dstY)
 744 {
 745     x86_reg dummy=0;
 746     x86_reg dstW_reg = dstW;
 747     x86_reg uv_off = c->uv_offx2;
 748
 749     YSCALEYUV2PACKEDX_ACCURATE
 750     YSCALEYUV2RGBX
 751     "pxor %%mm7, %%mm7 \n\t"
 752     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 753 #ifdef DITHER1XBPP
 754     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 755     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 756     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 757 #endif
 758     WRITERGB15(%4, %5, %%REGa)
 759     YSCALEYUV2PACKEDX_END
 760 }
 761
 762 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
 763                                  const int16_t **lumSrc, int lumFilterSize,
 764                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 765                                  const int16_t **chrVSrc,
 766                                  int chrFilterSize, const int16_t **alpSrc,
 767                                  uint8_t *dest, int dstW, int dstY)
 768 {
 769     x86_reg dummy=0;
 770     x86_reg dstW_reg = dstW;
 771     x86_reg uv_off = c->uv_offx2;
 772
 773     YSCALEYUV2PACKEDX
 774     YSCALEYUV2RGBX
 775     "pxor %%mm7, %%mm7 \n\t"
 776     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 777 #ifdef DITHER1XBPP
 778     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 779     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 780     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 781 #endif
 782     WRITERGB15(%4, %5, %%REGa)
 783     YSCALEYUV2PACKEDX_END
 784 }
 785
 786 #define WRITEBGR24MMX(dst, dstw, index) \
 787     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 788     "movq      %%mm2, %%mm1     \n\t" /* B */\
 789     "movq      %%mm5, %%mm6     \n\t" /* R */\
 790     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\
 791     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\
 792     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\
 793     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\
 794     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\
 795     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\
 796     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\
 797     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\
 798     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\
 799     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\
 800 \
 801     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\
 802     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\
 803     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\
 804     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\
 805 \
 806     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\
 807     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\
 808     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\
 809     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\
 810 \
 811     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\
 812     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\
 813     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\
 814     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\
 815 \
 816     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\
 817     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\
 818     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\
 819     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\
 820     MOVNTQ(%%mm0, (dst))\
 821 \
 822     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\
 823     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\
 824     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\
 825     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\
 826     MOVNTQ(%%mm6, 8(dst))\
 827 \
 828     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\
 829     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\
 830     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\
 831     MOVNTQ(%%mm5, 16(dst))\
 832 \
 833     "add         $24, "#dst"    \n\t"\
 834 \
 835     "add          $8, "#index"  \n\t"\
 836     "cmp     "#dstw", "#index"  \n\t"\
 837     " jb          1b            \n\t"
 838
 839 #define WRITEBGR24MMX2(dst, dstw, index) \
 840     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 841     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
 842     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
 843     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
 844     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
 845     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
 846 \
 847     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\
 848     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\
 849     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\
 850 \
 851     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\
 852     "por    %%mm1, %%mm6        \n\t"\
 853     "por    %%mm3, %%mm6        \n\t"\
 854     MOVNTQ(%%mm6, (dst))\
 855 \
 856     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
 857     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
 858     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
 859     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
 860 \
 861     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\
 862     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\
 863     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\
 864 \
 865     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\
 866     "por    %%mm3, %%mm6        \n\t"\
 867     MOVNTQ(%%mm6, 8(dst))\
 868 \
 869     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
 870     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
 871     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
 872 \
 873     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\
 874     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\
 875     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\
 876 \
 877     "por    %%mm1, %%mm3        \n\t"\
 878     "por    %%mm3, %%mm6        \n\t"\
 879     MOVNTQ(%%mm6, 16(dst))\
 880 \
 881     "add      $24, "#dst"       \n\t"\
 882 \
 883     "add       $8, "#index"     \n\t"\
 884     "cmp  "#dstw", "#index"     \n\t"\
 885     " jb       1b               \n\t"
 886
 887 #if COMPILE_TEMPLATE_MMX2
 888 #undef WRITEBGR24
 889 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX2(dst, dstw, index)
 890 #else
 891 #undef WRITEBGR24
 892 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)
 893 #endif
 894
 895 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
 896                                    const int16_t **lumSrc, int lumFilterSize,
 897                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 898                                    const int16_t **chrVSrc,
 899                                    int chrFilterSize, const int16_t **alpSrc,
 900                                    uint8_t *dest, int dstW, int dstY)
 901 {
 902     x86_reg dummy=0;
 903     x86_reg dstW_reg = dstW;
 904     x86_reg uv_off = c->uv_offx2;
 905
 906     YSCALEYUV2PACKEDX_ACCURATE
 907     YSCALEYUV2RGBX
 908     "pxor %%mm7, %%mm7 \n\t"
 909     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
 910     "add %4, %%"REG_c"                        \n\t"
 911     WRITEBGR24(%%REGc, %5, %%REGa)
 912     :: "r" (&c->redDither),
 913        "m" (dummy), "m" (dummy), "m" (dummy),
 914        "r" (dest), "m" (dstW_reg), "m"(uv_off)
 915     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 916     );
 917 }
 918
 919 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
 920                                 const int16_t **lumSrc, int lumFilterSize,
 921                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 922                                 const int16_t **chrVSrc,
 923                                 int chrFilterSize, const int16_t **alpSrc,
 924                                 uint8_t *dest, int dstW, int dstY)
 925 {
 926     x86_reg dummy=0;
 927     x86_reg dstW_reg = dstW;
 928     x86_reg uv_off = c->uv_offx2;
 929
 930     YSCALEYUV2PACKEDX
 931     YSCALEYUV2RGBX
 932     "pxor                    %%mm7, %%mm7       \n\t"
 933     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize
 934     "add                        %4, %%"REG_c"   \n\t"
 935     WRITEBGR24(%%REGc, %5, %%REGa)
 936     :: "r" (&c->redDither),
 937        "m" (dummy), "m" (dummy), "m" (dummy),
 938        "r" (dest),  "m" (dstW_reg), "m"(uv_off)
 939     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 940     );
 941 }
 942
 943 #define REAL_WRITEYUY2(dst, dstw, index) \
 944     "packuswb  %%mm3, %%mm3     \n\t"\
 945     "packuswb  %%mm4, %%mm4     \n\t"\
 946     "packuswb  %%mm7, %%mm1     \n\t"\
 947     "punpcklbw %%mm4, %%mm3     \n\t"\
 948     "movq      %%mm1, %%mm7     \n\t"\
 949     "punpcklbw %%mm3, %%mm1     \n\t"\
 950     "punpckhbw %%mm3, %%mm7     \n\t"\
 951 \
 952     MOVNTQ(%%mm1, (dst, index, 2))\
 953     MOVNTQ(%%mm7, 8(dst, index, 2))\
 954 \
 955     "add          $8, "#index"  \n\t"\
 956     "cmp     "#dstw", "#index"  \n\t"\
 957     " jb          1b            \n\t"
 958 #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)
 959
 960 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
 961                                      const int16_t **lumSrc, int lumFilterSize,
 962                                      const int16_t *chrFilter, const int16_t **chrUSrc,
 963                                      const int16_t **chrVSrc,
 964                                      int chrFilterSize, const int16_t **alpSrc,
 965                                      uint8_t *dest, int dstW, int dstY)
 966 {
 967     x86_reg dummy=0;
 968     x86_reg dstW_reg = dstW;
 969     x86_reg uv_off = c->uv_offx2;
 970
 971     YSCALEYUV2PACKEDX_ACCURATE
 972     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 973     "psraw $3, %%mm3    \n\t"
 974     "psraw $3, %%mm4    \n\t"
 975     "psraw $3, %%mm1    \n\t"
 976     "psraw $3, %%mm7    \n\t"
 977     WRITEYUY2(%4, %5, %%REGa)
 978     YSCALEYUV2PACKEDX_END
 979 }
 980
 981 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
 982                                   const int16_t **lumSrc, int lumFilterSize,
 983                                   const int16_t *chrFilter, const int16_t **chrUSrc,
 984                                   const int16_t **chrVSrc,
 985                                   int chrFilterSize, const int16_t **alpSrc,
 986                                   uint8_t *dest, int dstW, int dstY)
 987 {
 988     x86_reg dummy=0;
 989     x86_reg dstW_reg = dstW;
 990     x86_reg uv_off = c->uv_offx2;
 991
 992     YSCALEYUV2PACKEDX
 993     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 994     "psraw $3, %%mm3    \n\t"
 995     "psraw $3, %%mm4    \n\t"
 996     "psraw $3, %%mm1    \n\t"
 997     "psraw $3, %%mm7    \n\t"
 998     WRITEYUY2(%4, %5, %%REGa)
 999     YSCALEYUV2PACKEDX_END
1000 }
1001
1002 #define REAL_YSCALEYUV2RGB_UV(index, c) \
1003     "xor            "#index", "#index"  \n\t"\
1004     ".p2align              4            \n\t"\
1005     "1:                                 \n\t"\
1006     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1007     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1008     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1009     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1010     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1011     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1012     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1013     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1014     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
1015     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1016     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1017     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1018     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1019     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1020     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1021     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1022     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1023     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1024     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1025     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1026     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1027     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1028
1029 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
1030     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
1031     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
1032     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
1033     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
1034     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
1035     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
1036     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1037     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1038     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1039     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1040     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1041     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1042
1043 #define REAL_YSCALEYUV2RGB_COEFF(c) \
1044     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1045     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1046     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1047     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1048     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1049     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1050     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1051     "paddw             %%mm3, %%mm4     \n\t"\
1052     "movq              %%mm2, %%mm0     \n\t"\
1053     "movq              %%mm5, %%mm6     \n\t"\
1054     "movq              %%mm4, %%mm3     \n\t"\
1055     "punpcklwd         %%mm2, %%mm2     \n\t"\
1056     "punpcklwd         %%mm5, %%mm5     \n\t"\
1057     "punpcklwd         %%mm4, %%mm4     \n\t"\
1058     "paddw             %%mm1, %%mm2     \n\t"\
1059     "paddw             %%mm1, %%mm5     \n\t"\
1060     "paddw             %%mm1, %%mm4     \n\t"\
1061     "punpckhwd         %%mm0, %%mm0     \n\t"\
1062     "punpckhwd         %%mm6, %%mm6     \n\t"\
1063     "punpckhwd         %%mm3, %%mm3     \n\t"\
1064     "paddw             %%mm7, %%mm0     \n\t"\
1065     "paddw             %%mm7, %%mm6     \n\t"\
1066     "paddw             %%mm7, %%mm3     \n\t"\
1067     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1068     "packuswb          %%mm0, %%mm2     \n\t"\
1069     "packuswb          %%mm6, %%mm5     \n\t"\
1070     "packuswb          %%mm3, %%mm4     \n\t"\
1071
1072 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
1073
1074 #define YSCALEYUV2RGB(index, c) \
1075     REAL_YSCALEYUV2RGB_UV(index, c) \
1076     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
1077     REAL_YSCALEYUV2RGB_COEFF(c)
1078
1079 /**
1080  * vertical bilinear scale YV12 to RGB
1081  */
1082 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
1083                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1084                                 const int16_t *abuf[2], uint8_t *dest,
1085                                 int dstW, int yalpha, int uvalpha, int y)
1086 {
1087     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1088                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1089
1090     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1091         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
1092 #if ARCH_X86_64
1093         __asm__ volatile(
1094             YSCALEYUV2RGB(%%r8, %5)
1095             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
1096             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1097             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1098             "packuswb            %%mm7, %%mm1       \n\t"
1099             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
1100             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
1101                "a" (&c->redDither),
1102                "r" (abuf0), "r" (abuf1)
1103             : "%r8"
1104         );
1105 #else
1106         c->u_temp=(intptr_t)abuf0;
1107         c->v_temp=(intptr_t)abuf1;
1108         __asm__ volatile(
1109             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1110             "mov        %4, %%"REG_b"               \n\t"
1111             "push %%"REG_BP"                        \n\t"
1112             YSCALEYUV2RGB(%%REGBP, %5)
1113             "push                   %0              \n\t"
1114             "push                   %1              \n\t"
1115             "mov          "U_TEMP"(%5), %0          \n\t"
1116             "mov          "V_TEMP"(%5), %1          \n\t"
1117             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
1118             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1119             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
1120             "packuswb            %%mm7, %%mm1       \n\t"
1121             "pop                    %1              \n\t"
1122             "pop                    %0              \n\t"
1123             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
1124             "pop %%"REG_BP"                         \n\t"
1125             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1126             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1127                "a" (&c->redDither)
1128         );
1129 #endif
1130     } else {
1131         __asm__ volatile(
1132             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1133             "mov        %4, %%"REG_b"               \n\t"
1134             "push %%"REG_BP"                        \n\t"
1135             YSCALEYUV2RGB(%%REGBP, %5)
1136             "pcmpeqd %%mm7, %%mm7                   \n\t"
1137             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1138             "pop %%"REG_BP"                         \n\t"
1139             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1140             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1141                "a" (&c->redDither)
1142         );
1143     }
1144 }
1145
1146 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
1147                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1148                                 const int16_t *abuf[2], uint8_t *dest,
1149                                 int dstW, int yalpha, int uvalpha, int y)
1150 {
1151     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1152                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1153
1154     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1155     __asm__ volatile(
1156         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1157         "mov        %4, %%"REG_b"               \n\t"
1158         "push %%"REG_BP"                        \n\t"
1159         YSCALEYUV2RGB(%%REGBP, %5)
1160         "pxor    %%mm7, %%mm7                   \n\t"
1161         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1162         "pop %%"REG_BP"                         \n\t"
1163         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1164         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1165            "a" (&c->redDither)
1166     );
1167 }
1168
1169 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
1170                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1171                                  const int16_t *abuf[2], uint8_t *dest,
1172                                  int dstW, int yalpha, int uvalpha, int y)
1173 {
1174     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1175                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1176
1177     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1178     __asm__ volatile(
1179         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1180         "mov        %4, %%"REG_b"               \n\t"
1181         "push %%"REG_BP"                        \n\t"
1182         YSCALEYUV2RGB(%%REGBP, %5)
1183         "pxor    %%mm7, %%mm7                   \n\t"
1184         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1185 #ifdef DITHER1XBPP
1186         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1187         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1188         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1189 #endif
1190         WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1191         "pop %%"REG_BP"                         \n\t"
1192         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1193         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1194            "a" (&c->redDither)
1195     );
1196 }
1197
1198 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
1199                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1200                                  const int16_t *abuf[2], uint8_t *dest,
1201                                  int dstW, int yalpha, int uvalpha, int y)
1202 {
1203     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1204                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1205
1206     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1207     __asm__ volatile(
1208         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1209         "mov        %4, %%"REG_b"               \n\t"
1210         "push %%"REG_BP"                        \n\t"
1211         YSCALEYUV2RGB(%%REGBP, %5)
1212         "pxor    %%mm7, %%mm7                   \n\t"
1213         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1214 #ifdef DITHER1XBPP
1215         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1216         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1217         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1218 #endif
1219         WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1220         "pop %%"REG_BP"                         \n\t"
1221         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1222         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1223            "a" (&c->redDither)
1224     );
1225 }
1226
1227 #define REAL_YSCALEYUV2PACKED(index, c) \
1228     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\
1229     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\
1230     "psraw                $3, %%mm0                           \n\t"\
1231     "psraw                $3, %%mm1                           \n\t"\
1232     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1233     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
1234     "xor            "#index", "#index"                        \n\t"\
1235     ".p2align              4            \n\t"\
1236     "1:                                 \n\t"\
1237     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1238     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1239     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1240     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1241     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1242     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1243     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
1244     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
1245     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
1246     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
1247     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
1248     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1249     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1250     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
1251     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
1252     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
1253     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
1254     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
1255     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
1256     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
1257     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
1258     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1259     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
1260     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1261     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1262     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1263     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
1264
1265 #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)
1266
1267 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
1268                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
1269                                   const int16_t *abuf[2], uint8_t *dest,
1270                                   int dstW, int yalpha, int uvalpha, int y)
1271 {
1272     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
1273                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1274
1275     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
1276     __asm__ volatile(
1277         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1278         "mov %4, %%"REG_b"                        \n\t"
1279         "push %%"REG_BP"                        \n\t"
1280         YSCALEYUV2PACKED(%%REGBP, %5)
1281         WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1282         "pop %%"REG_BP"                         \n\t"
1283         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1284         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1285            "a" (&c->redDither)
1286     );
1287 }
1288
1289 #define REAL_YSCALEYUV2RGB1(index, c) \
1290     "xor            "#index", "#index"  \n\t"\
1291     ".p2align              4            \n\t"\
1292     "1:                                 \n\t"\
1293     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1294     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1295     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1296     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1297     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
1298     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
1299     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1300     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1301     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1302     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1303     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1304     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1305     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1306     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1307     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1308     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1309     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1310     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1311     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1312     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1313     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1314     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1315     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1316     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1317     "paddw             %%mm3, %%mm4     \n\t"\
1318     "movq              %%mm2, %%mm0     \n\t"\
1319     "movq              %%mm5, %%mm6     \n\t"\
1320     "movq              %%mm4, %%mm3     \n\t"\
1321     "punpcklwd         %%mm2, %%mm2     \n\t"\
1322     "punpcklwd         %%mm5, %%mm5     \n\t"\
1323     "punpcklwd         %%mm4, %%mm4     \n\t"\
1324     "paddw             %%mm1, %%mm2     \n\t"\
1325     "paddw             %%mm1, %%mm5     \n\t"\
1326     "paddw             %%mm1, %%mm4     \n\t"\
1327     "punpckhwd         %%mm0, %%mm0     \n\t"\
1328     "punpckhwd         %%mm6, %%mm6     \n\t"\
1329     "punpckhwd         %%mm3, %%mm3     \n\t"\
1330     "paddw             %%mm7, %%mm0     \n\t"\
1331     "paddw             %%mm7, %%mm6     \n\t"\
1332     "paddw             %%mm7, %%mm3     \n\t"\
1333     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1334     "packuswb          %%mm0, %%mm2     \n\t"\
1335     "packuswb          %%mm6, %%mm5     \n\t"\
1336     "packuswb          %%mm3, %%mm4     \n\t"\
1337
1338 #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)
1339
1340 // do vertical chrominance interpolation
1341 #define REAL_YSCALEYUV2RGB1b(index, c) \
1342     "xor            "#index", "#index"  \n\t"\
1343     ".p2align              4            \n\t"\
1344     "1:                                 \n\t"\
1345     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1346     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1347     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1348     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1349     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1350     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1351     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1352     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1353     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
1354     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\
1355     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1356     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1357     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1358     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1359     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1360     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1361     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1362     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1363     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1364     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1365     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1366     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1367     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1368     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1369     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1370     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1371     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1372     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1373     "paddw             %%mm3, %%mm4     \n\t"\
1374     "movq              %%mm2, %%mm0     \n\t"\
1375     "movq              %%mm5, %%mm6     \n\t"\
1376     "movq              %%mm4, %%mm3     \n\t"\
1377     "punpcklwd         %%mm2, %%mm2     \n\t"\
1378     "punpcklwd         %%mm5, %%mm5     \n\t"\
1379     "punpcklwd         %%mm4, %%mm4     \n\t"\
1380     "paddw             %%mm1, %%mm2     \n\t"\
1381     "paddw             %%mm1, %%mm5     \n\t"\
1382     "paddw             %%mm1, %%mm4     \n\t"\
1383     "punpckhwd         %%mm0, %%mm0     \n\t"\
1384     "punpckhwd         %%mm6, %%mm6     \n\t"\
1385     "punpckhwd         %%mm3, %%mm3     \n\t"\
1386     "paddw             %%mm7, %%mm0     \n\t"\
1387     "paddw             %%mm7, %%mm6     \n\t"\
1388     "paddw             %%mm7, %%mm3     \n\t"\
1389     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1390     "packuswb          %%mm0, %%mm2     \n\t"\
1391     "packuswb          %%mm6, %%mm5     \n\t"\
1392     "packuswb          %%mm3, %%mm4     \n\t"\
1393
1394 #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)
1395
1396 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1397     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\
1398     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\
1399     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\
1400     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\
1401     "packuswb          %%mm1, %%mm7     \n\t"
1402 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1403
1404 /**
1405  * YV12 to RGB without scaling or interpolating
1406  */
1407 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1408                                 const int16_t *ubuf[2], const int16_t *bguf[2],
1409                                 const int16_t *abuf0, uint8_t *dest,
1410                                 int dstW, int uvalpha, int y)
1411 {
1412     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1413     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1414
1415     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1416         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1417             __asm__ volatile(
1418                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1419                 "mov        %4, %%"REG_b"               \n\t"
1420                 "push %%"REG_BP"                        \n\t"
1421                 YSCALEYUV2RGB1(%%REGBP, %5)
1422                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1423                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1424                 "pop %%"REG_BP"                         \n\t"
1425                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1426                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1427                    "a" (&c->redDither)
1428             );
1429         } else {
1430             __asm__ volatile(
1431                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1432                 "mov        %4, %%"REG_b"               \n\t"
1433                 "push %%"REG_BP"                        \n\t"
1434                 YSCALEYUV2RGB1(%%REGBP, %5)
1435                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1436                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1437                 "pop %%"REG_BP"                         \n\t"
1438                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1439                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1440                    "a" (&c->redDither)
1441             );
1442         }
1443     } else {
1444         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
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                 YSCALEYUV2RGB1b(%%REGBP, %5)
1450                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1451                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1452                 "pop %%"REG_BP"                         \n\t"
1453                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1454                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1455                    "a" (&c->redDither)
1456             );
1457         } else {
1458             __asm__ volatile(
1459                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1460                 "mov        %4, %%"REG_b"               \n\t"
1461                 "push %%"REG_BP"                        \n\t"
1462                 YSCALEYUV2RGB1b(%%REGBP, %5)
1463                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1464                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1465                 "pop %%"REG_BP"                         \n\t"
1466                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1467                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1468                    "a" (&c->redDither)
1469             );
1470         }
1471     }
1472 }
1473
1474 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1475                                 const int16_t *ubuf[2], const int16_t *bguf[2],
1476                                 const int16_t *abuf0, uint8_t *dest,
1477                                 int dstW, int uvalpha, int y)
1478 {
1479     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1480     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1481
1482     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1483         __asm__ volatile(
1484             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1485             "mov        %4, %%"REG_b"               \n\t"
1486             "push %%"REG_BP"                        \n\t"
1487             YSCALEYUV2RGB1(%%REGBP, %5)
1488             "pxor    %%mm7, %%mm7                   \n\t"
1489             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1490             "pop %%"REG_BP"                         \n\t"
1491             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1492             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1493                "a" (&c->redDither)
1494         );
1495     } else {
1496         __asm__ volatile(
1497             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1498             "mov        %4, %%"REG_b"               \n\t"
1499             "push %%"REG_BP"                        \n\t"
1500             YSCALEYUV2RGB1b(%%REGBP, %5)
1501             "pxor    %%mm7, %%mm7                   \n\t"
1502             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1503             "pop %%"REG_BP"                         \n\t"
1504             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1505             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1506                "a" (&c->redDither)
1507         );
1508     }
1509 }
1510
1511 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1512                                  const int16_t *ubuf[2], const int16_t *bguf[2],
1513                                  const int16_t *abuf0, uint8_t *dest,
1514                                  int dstW, int uvalpha, int y)
1515 {
1516     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1517     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1518
1519     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1520         __asm__ volatile(
1521             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1522             "mov        %4, %%"REG_b"               \n\t"
1523             "push %%"REG_BP"                        \n\t"
1524             YSCALEYUV2RGB1(%%REGBP, %5)
1525             "pxor    %%mm7, %%mm7                   \n\t"
1526             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1527 #ifdef DITHER1XBPP
1528             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1529             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1530             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1531 #endif
1532             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1533             "pop %%"REG_BP"                         \n\t"
1534             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1535             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1536                "a" (&c->redDither)
1537         );
1538     } else {
1539         __asm__ volatile(
1540             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1541             "mov        %4, %%"REG_b"               \n\t"
1542             "push %%"REG_BP"                        \n\t"
1543             YSCALEYUV2RGB1b(%%REGBP, %5)
1544             "pxor    %%mm7, %%mm7                   \n\t"
1545             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1546 #ifdef DITHER1XBPP
1547             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1548             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1549             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1550 #endif
1551             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1552             "pop %%"REG_BP"                         \n\t"
1553             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1554             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1555                "a" (&c->redDither)
1556         );
1557     }
1558 }
1559
1560 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1561                                  const int16_t *ubuf[2], const int16_t *bguf[2],
1562                                  const int16_t *abuf0, uint8_t *dest,
1563                                  int dstW, int uvalpha, int y)
1564 {
1565     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1566     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1567
1568     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1569         __asm__ volatile(
1570             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1571             "mov        %4, %%"REG_b"               \n\t"
1572             "push %%"REG_BP"                        \n\t"
1573             YSCALEYUV2RGB1(%%REGBP, %5)
1574             "pxor    %%mm7, %%mm7                   \n\t"
1575             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1576 #ifdef DITHER1XBPP
1577             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1578             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1579             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1580 #endif
1581             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1582             "pop %%"REG_BP"                         \n\t"
1583             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1584             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1585                "a" (&c->redDither)
1586         );
1587     } else {
1588         __asm__ volatile(
1589             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1590             "mov        %4, %%"REG_b"               \n\t"
1591             "push %%"REG_BP"                        \n\t"
1592             YSCALEYUV2RGB1b(%%REGBP, %5)
1593             "pxor    %%mm7, %%mm7                   \n\t"
1594             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1595 #ifdef DITHER1XBPP
1596             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1597             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1598             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1599 #endif
1600             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1601             "pop %%"REG_BP"                         \n\t"
1602             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1603             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1604                "a" (&c->redDither)
1605         );
1606     }
1607 }
1608
1609 #define REAL_YSCALEYUV2PACKED1(index, c) \
1610     "xor            "#index", "#index"  \n\t"\
1611     ".p2align              4            \n\t"\
1612     "1:                                 \n\t"\
1613     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1614     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1615     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1616     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1617     "psraw                $7, %%mm3     \n\t" \
1618     "psraw                $7, %%mm4     \n\t" \
1619     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1620     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1621     "psraw                $7, %%mm1     \n\t" \
1622     "psraw                $7, %%mm7     \n\t" \
1623
1624 #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)
1625
1626 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1627     "xor "#index", "#index"             \n\t"\
1628     ".p2align              4            \n\t"\
1629     "1:                                 \n\t"\
1630     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1631     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1632     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
1633     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1634     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1635     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
1636     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1637     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1638     "psrlw                $8, %%mm3     \n\t" \
1639     "psrlw                $8, %%mm4     \n\t" \
1640     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1641     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1642     "psraw                $7, %%mm1     \n\t" \
1643     "psraw                $7, %%mm7     \n\t"
1644 #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)
1645
1646 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1647                                   const int16_t *ubuf[2], const int16_t *bguf[2],
1648                                   const int16_t *abuf0, uint8_t *dest,
1649                                   int dstW, int uvalpha, int y)
1650 {
1651     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
1652     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1653
1654     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1655         __asm__ volatile(
1656             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1657             "mov        %4, %%"REG_b"               \n\t"
1658             "push %%"REG_BP"                        \n\t"
1659             YSCALEYUV2PACKED1(%%REGBP, %5)
1660             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1661             "pop %%"REG_BP"                         \n\t"
1662             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1663             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1664                "a" (&c->redDither)
1665         );
1666     } else {
1667         __asm__ volatile(
1668             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1669             "mov        %4, %%"REG_b"               \n\t"
1670             "push %%"REG_BP"                        \n\t"
1671             YSCALEYUV2PACKED1b(%%REGBP, %5)
1672             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1673             "pop %%"REG_BP"                         \n\t"
1674             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1675             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1676                "a" (&c->redDither)
1677         );
1678     }
1679 }
1680
1681 #if !COMPILE_TEMPLATE_MMX2
1682 //FIXME yuy2* can read up to 7 samples too much
1683
1684 static void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src,
1685                             int width, uint32_t *unused)
1686 {
1687     __asm__ volatile(
1688         "movq "MANGLE(bm01010101)", %%mm2           \n\t"
1689         "mov                    %0, %%"REG_a"       \n\t"
1690         "1:                                         \n\t"
1691         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1692         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1693         "pand                %%mm2, %%mm0           \n\t"
1694         "pand                %%mm2, %%mm1           \n\t"
1695         "packuswb            %%mm1, %%mm0           \n\t"
1696         "movq                %%mm0, (%2, %%"REG_a") \n\t"
1697         "add                    $8, %%"REG_a"       \n\t"
1698         " js                    1b                  \n\t"
1699         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
1700         : "%"REG_a
1701     );
1702 }
1703
1704 static void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV,
1705                              const uint8_t *src1, const uint8_t *src2,
1706                              int width, uint32_t *unused)
1707 {
1708     __asm__ volatile(
1709         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1710         "mov                    %0, %%"REG_a"       \n\t"
1711         "1:                                         \n\t"
1712         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"
1713         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
1714         "psrlw                  $8, %%mm0           \n\t"
1715         "psrlw                  $8, %%mm1           \n\t"
1716         "packuswb            %%mm1, %%mm0           \n\t"
1717         "movq                %%mm0, %%mm1           \n\t"
1718         "psrlw                  $8, %%mm0           \n\t"
1719         "pand                %%mm4, %%mm1           \n\t"
1720         "packuswb            %%mm0, %%mm0           \n\t"
1721         "packuswb            %%mm1, %%mm1           \n\t"
1722         "movd                %%mm0, (%3, %%"REG_a") \n\t"
1723         "movd                %%mm1, (%2, %%"REG_a") \n\t"
1724         "add                    $4, %%"REG_a"       \n\t"
1725         " js                    1b                  \n\t"
1726         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
1727         : "%"REG_a
1728     );
1729     assert(src1 == src2);
1730 }
1731
1732 /* This is almost identical to the previous, end exists only because
1733  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */
1734 static void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src,
1735                             int width, uint32_t *unused)
1736 {
1737     __asm__ volatile(
1738         "mov                  %0, %%"REG_a"         \n\t"
1739         "1:                                         \n\t"
1740         "movq  (%1, %%"REG_a",2), %%mm0             \n\t"
1741         "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"
1742         "psrlw                $8, %%mm0             \n\t"
1743         "psrlw                $8, %%mm1             \n\t"
1744         "packuswb          %%mm1, %%mm0             \n\t"
1745         "movq              %%mm0, (%2, %%"REG_a")   \n\t"
1746         "add                  $8, %%"REG_a"         \n\t"
1747         " js                  1b                    \n\t"
1748         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)
1749         : "%"REG_a
1750     );
1751 }
1752
1753 static void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV,
1754                              const uint8_t *src1, const uint8_t *src2,
1755                              int width, uint32_t *unused)
1756 {
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",4), %%mm0           \n\t"
1762         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"
1763         "pand                %%mm4, %%mm0           \n\t"
1764         "pand                %%mm4, %%mm1           \n\t"
1765         "packuswb            %%mm1, %%mm0           \n\t"
1766         "movq                %%mm0, %%mm1           \n\t"
1767         "psrlw                  $8, %%mm0           \n\t"
1768         "pand                %%mm4, %%mm1           \n\t"
1769         "packuswb            %%mm0, %%mm0           \n\t"
1770         "packuswb            %%mm1, %%mm1           \n\t"
1771         "movd                %%mm0, (%3, %%"REG_a") \n\t"
1772         "movd                %%mm1, (%2, %%"REG_a") \n\t"
1773         "add                    $4, %%"REG_a"       \n\t"
1774         " js                    1b                  \n\t"
1775         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)
1776         : "%"REG_a
1777     );
1778     assert(src1 == src2);
1779 }
1780
1781 static av_always_inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,
1782                                               const uint8_t *src, int width)
1783 {
1784     __asm__ volatile(
1785         "movq "MANGLE(bm01010101)", %%mm4           \n\t"
1786         "mov                    %0, %%"REG_a"       \n\t"
1787         "1:                                         \n\t"
1788         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
1789         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
1790         "movq                %%mm0, %%mm2           \n\t"
1791         "movq                %%mm1, %%mm3           \n\t"
1792         "pand                %%mm4, %%mm0           \n\t"
1793         "pand                %%mm4, %%mm1           \n\t"
1794         "psrlw                  $8, %%mm2           \n\t"
1795         "psrlw                  $8, %%mm3           \n\t"
1796         "packuswb            %%mm1, %%mm0           \n\t"
1797         "packuswb            %%mm3, %%mm2           \n\t"
1798         "movq                %%mm0, (%2, %%"REG_a") \n\t"
1799         "movq                %%mm2, (%3, %%"REG_a") \n\t"
1800         "add                    $8, %%"REG_a"       \n\t"
1801         " js                    1b                  \n\t"
1802         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)
1803         : "%"REG_a
1804     );
1805 }
1806
1807 static void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,
1808                              const uint8_t *src1, const uint8_t *src2,
1809                              int width, uint32_t *unused)
1810 {
1811     RENAME(nvXXtoUV)(dstU, dstV, src1, width);
1812 }
1813
1814 static void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,
1815                              const uint8_t *src1, const uint8_t *src2,
1816                              int width, uint32_t *unused)
1817 {
1818     RENAME(nvXXtoUV)(dstV, dstU, src1, width);
1819 }
1820 #endif /* !COMPILE_TEMPLATE_MMX2 */
1821
1822 static av_always_inline void RENAME(bgr24ToY_mmx)(int16_t *dst, const uint8_t *src,
1823                                                   int width, enum PixelFormat srcFormat)
1824 {
1825
1826     if(srcFormat == PIX_FMT_BGR24) {
1827         __asm__ volatile(
1828             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"
1829             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"
1830             :
1831         );
1832     } else {
1833         __asm__ volatile(
1834             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"
1835             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"
1836             :
1837         );
1838     }
1839
1840     __asm__ volatile(
1841         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"
1842         "mov                        %2, %%"REG_a"   \n\t"
1843         "pxor                    %%mm7, %%mm7       \n\t"
1844         "1:                                         \n\t"
1845         PREFETCH"               64(%0)              \n\t"
1846         "movd                     (%0), %%mm0       \n\t"
1847         "movd                    2(%0), %%mm1       \n\t"
1848         "movd                    6(%0), %%mm2       \n\t"
1849         "movd                    8(%0), %%mm3       \n\t"
1850         "add                       $12, %0          \n\t"
1851         "punpcklbw               %%mm7, %%mm0       \n\t"
1852         "punpcklbw               %%mm7, %%mm1       \n\t"
1853         "punpcklbw               %%mm7, %%mm2       \n\t"
1854         "punpcklbw               %%mm7, %%mm3       \n\t"
1855         "pmaddwd                 %%mm5, %%mm0       \n\t"
1856         "pmaddwd                 %%mm6, %%mm1       \n\t"
1857         "pmaddwd                 %%mm5, %%mm2       \n\t"
1858         "pmaddwd                 %%mm6, %%mm3       \n\t"
1859         "paddd                   %%mm1, %%mm0       \n\t"
1860         "paddd                   %%mm3, %%mm2       \n\t"
1861         "paddd                   %%mm4, %%mm0       \n\t"
1862         "paddd                   %%mm4, %%mm2       \n\t"
1863         "psrad                     $9, %%mm0       \n\t"
1864         "psrad                     $9, %%mm2       \n\t"
1865         "packssdw                %%mm2, %%mm0       \n\t"
1866         "movq                %%mm0, (%1, %%"REG_a") \n\t"
1867         "add                        $8, %%"REG_a"   \n\t"
1868         " js                        1b              \n\t"
1869     : "+r" (src)
1870     : "r" (dst+width), "g" ((x86_reg)-2*width)
1871     : "%"REG_a
1872     );
1873 }
1874
1875 static void RENAME(bgr24ToY)(int16_t *dst, const uint8_t *src,
1876                              int width, uint32_t *unused)
1877 {
1878     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);
1879 }
1880
1881 static void RENAME(rgb24ToY)(int16_t *dst, const uint8_t *src,
1882                              int width, uint32_t *unused)
1883 {
1884     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);
1885 }
1886
1887 static av_always_inline void RENAME(bgr24ToUV_mmx)(int16_t *dstU, int16_t *dstV,
1888                                                    const uint8_t *src, int width,
1889                                                    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                     $9, %%mm0       \n\t"
1930         "psrad                     $9, %%mm2       \n\t"
1931         "psrad                     $9, %%mm1       \n\t"
1932         "psrad                     $9, %%mm4       \n\t"
1933         "packssdw                %%mm1, %%mm0       \n\t"
1934         "packssdw                %%mm4, %%mm2       \n\t"
1935         "movq                %%mm0, (%1, %%"REG_a") \n\t"
1936         "movq                %%mm2, (%2, %%"REG_a") \n\t"
1937         "add                        $8, %%"REG_a"   \n\t"
1938         " js                        1b              \n\t"
1939     : "+r" (src)
1940     : "r" (dstU+width), "r" (dstV+width), "g" ((x86_reg)-2*width), "r"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24])
1941     : "%"REG_a
1942     );
1943 }
1944
1945 static void RENAME(bgr24ToUV)(int16_t *dstU, int16_t *dstV,
1946                               const uint8_t *src1, const uint8_t *src2,
1947                               int width, uint32_t *unused)
1948 {
1949     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);
1950     assert(src1 == src2);
1951 }
1952
1953 static void RENAME(rgb24ToUV)(int16_t *dstU, int16_t *dstV,
1954                               const uint8_t *src1, const uint8_t *src2,
1955                               int width, uint32_t *unused)
1956 {
1957     assert(src1==src2);
1958     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);
1959 }
1960
1961 #if !COMPILE_TEMPLATE_MMX2
1962 // bilinear / bicubic scaling
1963 static void RENAME(hScale)(SwsContext *c, int16_t *dst, int dstW,
1964                            const uint8_t *src, const int16_t *filter,
1965                            const int16_t *filterPos, int filterSize)
1966 {
1967     assert(filterSize % 4 == 0 && filterSize>0);
1968     if (filterSize==4) { // Always true for upscaling, sometimes for down, too.
1969         x86_reg counter= -2*dstW;
1970         filter-= counter*2;
1971         filterPos-= counter/2;
1972         dst-= counter/2;
1973         __asm__ volatile(
1974 #if defined(PIC)
1975             "push            %%"REG_b"              \n\t"
1976 #endif
1977             "pxor                %%mm7, %%mm7       \n\t"
1978             "push           %%"REG_BP"              \n\t" // we use 7 regs here ...
1979             "mov             %%"REG_a", %%"REG_BP"  \n\t"
1980             ".p2align                4              \n\t"
1981             "1:                                     \n\t"
1982             "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"
1983             "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"
1984             "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"
1985             "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"
1986             "movd      (%3, %%"REG_a"), %%mm0       \n\t"
1987             "movd      (%3, %%"REG_b"), %%mm2       \n\t"
1988             "punpcklbw           %%mm7, %%mm0       \n\t"
1989             "punpcklbw           %%mm7, %%mm2       \n\t"
1990             "pmaddwd             %%mm1, %%mm0       \n\t"
1991             "pmaddwd             %%mm2, %%mm3       \n\t"
1992             "movq                %%mm0, %%mm4       \n\t"
1993             "punpckldq           %%mm3, %%mm0       \n\t"
1994             "punpckhdq           %%mm3, %%mm4       \n\t"
1995             "paddd               %%mm4, %%mm0       \n\t"
1996             "psrad                  $7, %%mm0       \n\t"
1997             "packssdw            %%mm0, %%mm0       \n\t"
1998             "movd                %%mm0, (%4, %%"REG_BP")    \n\t"
1999             "add                    $4, %%"REG_BP"  \n\t"
2000             " jnc                   1b              \n\t"
2001
2002             "pop            %%"REG_BP"              \n\t"
2003 #if defined(PIC)
2004             "pop             %%"REG_b"              \n\t"
2005 #endif
2006             : "+a" (counter)
2007             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
2008 #if !defined(PIC)
2009             : "%"REG_b
2010 #endif
2011         );
2012     } else if (filterSize==8) {
2013         x86_reg counter= -2*dstW;
2014         filter-= counter*4;
2015         filterPos-= counter/2;
2016         dst-= counter/2;
2017         __asm__ volatile(
2018 #if defined(PIC)
2019             "push             %%"REG_b"             \n\t"
2020 #endif
2021             "pxor                 %%mm7, %%mm7      \n\t"
2022             "push            %%"REG_BP"             \n\t" // we use 7 regs here ...
2023             "mov              %%"REG_a", %%"REG_BP" \n\t"
2024             ".p2align                 4             \n\t"
2025             "1:                                     \n\t"
2026             "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"
2027             "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"
2028             "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"
2029             "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"
2030             "movd       (%3, %%"REG_a"), %%mm0      \n\t"
2031             "movd       (%3, %%"REG_b"), %%mm2      \n\t"
2032             "punpcklbw            %%mm7, %%mm0      \n\t"
2033             "punpcklbw            %%mm7, %%mm2      \n\t"
2034             "pmaddwd              %%mm1, %%mm0      \n\t"
2035             "pmaddwd              %%mm2, %%mm3      \n\t"
2036
2037             "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"
2038             "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"
2039             "movd      4(%3, %%"REG_a"), %%mm4      \n\t"
2040             "movd      4(%3, %%"REG_b"), %%mm2      \n\t"
2041             "punpcklbw            %%mm7, %%mm4      \n\t"
2042             "punpcklbw            %%mm7, %%mm2      \n\t"
2043             "pmaddwd              %%mm1, %%mm4      \n\t"
2044             "pmaddwd              %%mm2, %%mm5      \n\t"
2045             "paddd                %%mm4, %%mm0      \n\t"
2046             "paddd                %%mm5, %%mm3      \n\t"
2047             "movq                 %%mm0, %%mm4      \n\t"
2048             "punpckldq            %%mm3, %%mm0      \n\t"
2049             "punpckhdq            %%mm3, %%mm4      \n\t"
2050             "paddd                %%mm4, %%mm0      \n\t"
2051             "psrad                   $7, %%mm0      \n\t"
2052             "packssdw             %%mm0, %%mm0      \n\t"
2053             "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"
2054             "add                     $4, %%"REG_BP" \n\t"
2055             " jnc                    1b             \n\t"
2056
2057             "pop             %%"REG_BP"             \n\t"
2058 #if defined(PIC)
2059             "pop              %%"REG_b"             \n\t"
2060 #endif
2061             : "+a" (counter)
2062             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
2063 #if !defined(PIC)
2064             : "%"REG_b
2065 #endif
2066         );
2067     } else {
2068         const uint8_t *offset = src+filterSize;
2069         x86_reg counter= -2*dstW;
2070         //filter-= counter*filterSize/2;
2071         filterPos-= counter/2;
2072         dst-= counter/2;
2073         __asm__ volatile(
2074             "pxor                  %%mm7, %%mm7     \n\t"
2075             ".p2align                  4            \n\t"
2076             "1:                                     \n\t"
2077             "mov                      %2, %%"REG_c" \n\t"
2078             "movzwl      (%%"REG_c", %0), %%eax     \n\t"
2079             "movzwl     2(%%"REG_c", %0), %%edx     \n\t"
2080             "mov                      %5, %%"REG_c" \n\t"
2081             "pxor                  %%mm4, %%mm4     \n\t"
2082             "pxor                  %%mm5, %%mm5     \n\t"
2083             "2:                                     \n\t"
2084             "movq                   (%1), %%mm1     \n\t"
2085             "movq               (%1, %6), %%mm3     \n\t"
2086             "movd (%%"REG_c", %%"REG_a"), %%mm0     \n\t"
2087             "movd (%%"REG_c", %%"REG_d"), %%mm2     \n\t"
2088             "punpcklbw             %%mm7, %%mm0     \n\t"
2089             "punpcklbw             %%mm7, %%mm2     \n\t"
2090             "pmaddwd               %%mm1, %%mm0     \n\t"
2091             "pmaddwd               %%mm2, %%mm3     \n\t"
2092             "paddd                 %%mm3, %%mm5     \n\t"
2093             "paddd                 %%mm0, %%mm4     \n\t"
2094             "add                      $8, %1        \n\t"
2095             "add                      $4, %%"REG_c" \n\t"
2096             "cmp                      %4, %%"REG_c" \n\t"
2097             " jb                      2b            \n\t"
2098             "add                      %6, %1        \n\t"
2099             "movq                  %%mm4, %%mm0     \n\t"
2100             "punpckldq             %%mm5, %%mm4     \n\t"
2101             "punpckhdq             %%mm5, %%mm0     \n\t"
2102             "paddd                 %%mm0, %%mm4     \n\t"
2103             "psrad                    $7, %%mm4     \n\t"
2104             "packssdw              %%mm4, %%mm4     \n\t"
2105             "mov                      %3, %%"REG_a" \n\t"
2106             "movd                  %%mm4, (%%"REG_a", %0)   \n\t"
2107             "add                      $4, %0        \n\t"
2108             " jnc                     1b            \n\t"
2109
2110             : "+r" (counter), "+r" (filter)
2111             : "m" (filterPos), "m" (dst), "m"(offset),
2112             "m" (src), "r" ((x86_reg)filterSize*2)
2113             : "%"REG_a, "%"REG_c, "%"REG_d
2114         );
2115     }
2116 }
2117 #endif /* !COMPILE_TEMPLATE_MMX2 */
2118
2119 static inline void RENAME(hScale16)(int16_t *dst, int dstW, const uint16_t *src, int srcW, int xInc,
2120                                     const int16_t *filter, const int16_t *filterPos, long filterSize, int shift)
2121 {
2122     int i, j;
2123
2124     assert(filterSize % 4 == 0 && filterSize>0);
2125     if (filterSize==4 && shift<15) { // Always true for upscaling, sometimes for down, too.
2126         x86_reg counter= -2*dstW;
2127         filter-= counter*2;
2128         filterPos-= counter/2;
2129         dst-= counter/2;
2130         __asm__ volatile(
2131             "movd                   %5, %%mm7       \n\t"
2132 #if defined(PIC)
2133             "push            %%"REG_b"              \n\t"
2134 #endif
2135             "push           %%"REG_BP"              \n\t" // we use 7 regs here ...
2136             "mov             %%"REG_a", %%"REG_BP"  \n\t"
2137             ".p2align                4              \n\t"
2138             "1:                                     \n\t"
2139             "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"
2140             "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"
2141             "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"
2142             "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"
2143             "movq      (%3, %%"REG_a", 2), %%mm0    \n\t"
2144             "movq      (%3, %%"REG_b", 2), %%mm2    \n\t"
2145             "pmaddwd             %%mm1, %%mm0       \n\t"
2146             "pmaddwd             %%mm2, %%mm3       \n\t"
2147             "movq                %%mm0, %%mm4       \n\t"
2148             "punpckldq           %%mm3, %%mm0       \n\t"
2149             "punpckhdq           %%mm3, %%mm4       \n\t"
2150             "paddd               %%mm4, %%mm0       \n\t"
2151             "psrad               %%mm7, %%mm0       \n\t"
2152             "packssdw            %%mm0, %%mm0       \n\t"
2153             "movd                %%mm0, (%4, %%"REG_BP")    \n\t"
2154             "add                    $4, %%"REG_BP"  \n\t"
2155             " jnc                   1b              \n\t"
2156
2157             "pop            %%"REG_BP"              \n\t"
2158 #if defined(PIC)
2159             "pop             %%"REG_b"              \n\t"
2160 #endif
2161             : "+a" (counter)
2162             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst), "m"(shift)
2163 #if !defined(PIC)
2164             : "%"REG_b
2165 #endif
2166         );
2167     } else if (filterSize==8 && shift<15) {
2168         x86_reg counter= -2*dstW;
2169         filter-= counter*4;
2170         filterPos-= counter/2;
2171         dst-= counter/2;
2172         __asm__ volatile(
2173             "movd                   %5, %%mm7       \n\t"
2174 #if defined(PIC)
2175             "push            %%"REG_b"              \n\t"
2176 #endif
2177             "push            %%"REG_BP"             \n\t" // we use 7 regs here ...
2178             "mov              %%"REG_a", %%"REG_BP" \n\t"
2179             ".p2align                 4             \n\t"
2180             "1:                                     \n\t"
2181             "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"
2182             "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"
2183             "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"
2184             "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"
2185             "movq       (%3, %%"REG_a", 2), %%mm0   \n\t"
2186             "movq       (%3, %%"REG_b", 2), %%mm2   \n\t"
2187             "pmaddwd              %%mm1, %%mm0      \n\t"
2188             "pmaddwd              %%mm2, %%mm3      \n\t"
2189
2190             "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"
2191             "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"
2192             "movq      8(%3, %%"REG_a", 2), %%mm4   \n\t"
2193             "movq      8(%3, %%"REG_b", 2), %%mm2   \n\t"
2194             "pmaddwd              %%mm1, %%mm4      \n\t"
2195             "pmaddwd              %%mm2, %%mm5      \n\t"
2196             "paddd                %%mm4, %%mm0      \n\t"
2197             "paddd                %%mm5, %%mm3      \n\t"
2198             "movq                 %%mm0, %%mm4      \n\t"
2199             "punpckldq            %%mm3, %%mm0      \n\t"
2200             "punpckhdq            %%mm3, %%mm4      \n\t"
2201             "paddd                %%mm4, %%mm0      \n\t"
2202             "psrad                %%mm7, %%mm0      \n\t"
2203             "packssdw             %%mm0, %%mm0      \n\t"
2204             "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"
2205             "add                     $4, %%"REG_BP" \n\t"
2206             " jnc                    1b             \n\t"
2207
2208             "pop             %%"REG_BP"             \n\t"
2209 #if defined(PIC)
2210             "pop             %%"REG_b"              \n\t"
2211 #endif
2212             : "+a" (counter)
2213             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst), "m"(shift)
2214 #if !defined(PIC)
2215             : "%"REG_b
2216 #endif
2217         );
2218     } else if (shift<15){
2219         const uint16_t *offset = src+filterSize;
2220         x86_reg counter= -2*dstW;
2221         //filter-= counter*filterSize/2;
2222         filterPos-= counter/2;
2223         dst-= counter/2;
2224         __asm__ volatile(
2225             "movd                   %7, %%mm7       \n\t"
2226             ".p2align                  4            \n\t"
2227             "1:                                     \n\t"
2228             "mov                      %2, %%"REG_c" \n\t"
2229             "movzwl      (%%"REG_c", %0), %%eax     \n\t"
2230             "movzwl     2(%%"REG_c", %0), %%edx     \n\t"
2231             "mov                      %5, %%"REG_c" \n\t"
2232             "pxor                  %%mm4, %%mm4     \n\t"
2233             "pxor                  %%mm5, %%mm5     \n\t"
2234             "2:                                     \n\t"
2235             "movq                   (%1), %%mm1     \n\t"
2236             "movq               (%1, %6), %%mm3     \n\t"
2237             "movq (%%"REG_c", %%"REG_a", 2), %%mm0     \n\t"
2238             "movq (%%"REG_c", %%"REG_d", 2), %%mm2     \n\t"
2239             "pmaddwd               %%mm1, %%mm0     \n\t"
2240             "pmaddwd               %%mm2, %%mm3     \n\t"
2241             "paddd                 %%mm3, %%mm5     \n\t"
2242             "paddd                 %%mm0, %%mm4     \n\t"
2243             "add                      $8, %1        \n\t"
2244             "add                      $8, %%"REG_c" \n\t"
2245             "cmp                      %4, %%"REG_c" \n\t"
2246             " jb                      2b            \n\t"
2247             "add                      %6, %1        \n\t"
2248             "movq                  %%mm4, %%mm0     \n\t"
2249             "punpckldq             %%mm5, %%mm4     \n\t"
2250             "punpckhdq             %%mm5, %%mm0     \n\t"
2251             "paddd                 %%mm0, %%mm4     \n\t"
2252             "psrad                 %%mm7, %%mm4     \n\t"
2253             "packssdw              %%mm4, %%mm4     \n\t"
2254             "mov                      %3, %%"REG_a" \n\t"
2255             "movd                  %%mm4, (%%"REG_a", %0)   \n\t"
2256             "add                      $4, %0        \n\t"
2257             " jnc                     1b            \n\t"
2258
2259             : "+r" (counter), "+r" (filter)
2260             : "m" (filterPos), "m" (dst), "m"(offset),
2261             "m" (src), "r" ((x86_reg)filterSize*2), "m"(shift)
2262             : "%"REG_a, "%"REG_c, "%"REG_d
2263         );
2264     } else
2265     for (i=0; i<dstW; i++) {
2266         int srcPos= filterPos[i];
2267         int val=0;
2268         for (j=0; j<filterSize; j++) {
2269             val += ((int)src[srcPos + j])*filter[filterSize*i + j];
2270         }
2271         dst[i] = FFMIN(val>>shift, (1<<15)-1); // the cubic equation does overflow ...
2272     }
2273 }
2274
2275
2276 #if COMPILE_TEMPLATE_MMX2
2277 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
2278                                  int dstWidth, const uint8_t *src,
2279                                  int srcW, int xInc)
2280 {
2281     int16_t *filterPos = c->hLumFilterPos;
2282     int16_t *filter    = c->hLumFilter;
2283     void    *mmx2FilterCode= c->lumMmx2FilterCode;
2284     int i;
2285 #if defined(PIC)
2286     DECLARE_ALIGNED(8, uint64_t, ebxsave);
2287 #endif
2288
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 #define CALL_MMX2_FILTER_CODE \
2305         "movl            (%%"REG_b"), %%esi     \n\t"\
2306         "call                    *%4            \n\t"\
2307         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\
2308         "add               %%"REG_S", %%"REG_c" \n\t"\
2309         "add               %%"REG_a", %%"REG_D" \n\t"\
2310         "xor               %%"REG_a", %%"REG_a" \n\t"\
2311
2312 #else
2313 #define CALL_MMX2_FILTER_CODE \
2314         "movl (%%"REG_b"), %%esi        \n\t"\
2315         "call         *%4                       \n\t"\
2316         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
2317         "add               %%"REG_a", %%"REG_D" \n\t"\
2318         "xor               %%"REG_a", %%"REG_a" \n\t"\
2319
2320 #endif /* ARCH_X86_64 */
2321
2322         CALL_MMX2_FILTER_CODE
2323         CALL_MMX2_FILTER_CODE
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
2331 #if defined(PIC)
2332         "mov                      %5, %%"REG_b" \n\t"
2333 #endif
2334         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
2335            "m" (mmx2FilterCode)
2336 #if defined(PIC)
2337           ,"m" (ebxsave)
2338 #endif
2339         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
2340 #if !defined(PIC)
2341          ,"%"REG_b
2342 #endif
2343     );
2344
2345     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
2346         dst[i] = src[srcW-1]*128;
2347 }
2348
2349 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
2350                                  int dstWidth, const uint8_t *src1,
2351                                  const uint8_t *src2, int srcW, int xInc)
2352 {
2353     int16_t *filterPos = c->hChrFilterPos;
2354     int16_t *filter    = c->hChrFilter;
2355     void    *mmx2FilterCode= c->chrMmx2FilterCode;
2356     int i;
2357 #if defined(PIC)
2358     DECLARE_ALIGNED(8, uint64_t, ebxsave);
2359 #endif
2360
2361     __asm__ volatile(
2362 #if defined(PIC)
2363         "mov          %%"REG_b", %7         \n\t"
2364 #endif
2365         "pxor             %%mm7, %%mm7      \n\t"
2366         "mov                 %0, %%"REG_c"  \n\t"
2367         "mov                 %1, %%"REG_D"  \n\t"
2368         "mov                 %2, %%"REG_d"  \n\t"
2369         "mov                 %3, %%"REG_b"  \n\t"
2370         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
2371         PREFETCH"   (%%"REG_c")             \n\t"
2372         PREFETCH" 32(%%"REG_c")             \n\t"
2373         PREFETCH" 64(%%"REG_c")             \n\t"
2374
2375         CALL_MMX2_FILTER_CODE
2376         CALL_MMX2_FILTER_CODE
2377         CALL_MMX2_FILTER_CODE
2378         CALL_MMX2_FILTER_CODE
2379         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
2380         "mov                 %5, %%"REG_c"  \n\t" // src
2381         "mov                 %6, %%"REG_D"  \n\t" // buf2
2382         PREFETCH"   (%%"REG_c")             \n\t"
2383         PREFETCH" 32(%%"REG_c")             \n\t"
2384         PREFETCH" 64(%%"REG_c")             \n\t"
2385
2386         CALL_MMX2_FILTER_CODE
2387         CALL_MMX2_FILTER_CODE
2388         CALL_MMX2_FILTER_CODE
2389         CALL_MMX2_FILTER_CODE
2390
2391 #if defined(PIC)
2392         "mov %7, %%"REG_b"    \n\t"
2393 #endif
2394         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
2395            "m" (mmx2FilterCode), "m" (src2), "m"(dst2)
2396 #if defined(PIC)
2397           ,"m" (ebxsave)
2398 #endif
2399         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
2400 #if !defined(PIC)
2401          ,"%"REG_b
2402 #endif
2403     );
2404
2405     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
2406         dst1[i] = src1[srcW-1]*128;
2407         dst2[i] = src2[srcW-1]*128;
2408     }
2409 }
2410 #endif /* COMPILE_TEMPLATE_MMX2 */
2411
2412 static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
2413 {
2414     enum PixelFormat srcFormat = c->srcFormat,
2415                      dstFormat = c->dstFormat;
2416
2417     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != PIX_FMT_NV12
2418         && dstFormat != PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {
2419             if (c->flags & SWS_ACCURATE_RND) {
2420                 c->yuv2yuv1 = RENAME(yuv2yuv1_ar    );
2421                 c->yuv2yuvX = RENAME(yuv2yuvX_ar    );
2422                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
2423                     switch (c->dstFormat) {
2424                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;
2425                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;
2426                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;
2427                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;
2428                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
2429                     default: break;
2430                     }
2431                 }
2432             } else {
2433                 int should_dither= isNBPS(c->srcFormat) || is16BPS(c->srcFormat);
2434                 c->yuv2yuv1 = should_dither ? RENAME(yuv2yuv1_ar    ) : RENAME(yuv2yuv1    );
2435                 c->yuv2yuvX = RENAME(yuv2yuvX    );
2436                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
2437                     switch (c->dstFormat) {
2438                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;
2439                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;
2440                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;
2441                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;
2442                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
2443                     default: break;
2444                     }
2445                 }
2446             }
2447         if (!(c->flags & SWS_FULL_CHR_H_INT)) {
2448             switch (c->dstFormat) {
2449             case PIX_FMT_RGB32:
2450                 c->yuv2packed1 = RENAME(yuv2rgb32_1);
2451                 c->yuv2packed2 = RENAME(yuv2rgb32_2);
2452                 break;
2453             case PIX_FMT_BGR24:
2454                 c->yuv2packed1 = RENAME(yuv2bgr24_1);
2455                 c->yuv2packed2 = RENAME(yuv2bgr24_2);
2456                 break;
2457             case PIX_FMT_RGB555:
2458                 c->yuv2packed1 = RENAME(yuv2rgb555_1);
2459                 c->yuv2packed2 = RENAME(yuv2rgb555_2);
2460                 break;
2461             case PIX_FMT_RGB565:
2462                 c->yuv2packed1 = RENAME(yuv2rgb565_1);
2463                 c->yuv2packed2 = RENAME(yuv2rgb565_2);
2464                 break;
2465             case PIX_FMT_YUYV422:
2466                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);
2467                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);
2468                 break;
2469             default:
2470                 break;
2471             }
2472         }
2473     }
2474
2475     if (c->srcBpc == 8 && c->dstBpc <= 10) {
2476 #if !COMPILE_TEMPLATE_MMX2
2477     c->hyScale = c->hcScale = RENAME(hScale      );
2478 #endif /* !COMPILE_TEMPLATE_MMX2 */
2479
2480     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).
2481 #if COMPILE_TEMPLATE_MMX2
2482     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)
2483     {
2484         c->hyscale_fast = RENAME(hyscale_fast);
2485         c->hcscale_fast = RENAME(hcscale_fast);
2486     } else {
2487 #endif /* COMPILE_TEMPLATE_MMX2 */
2488         c->hyscale_fast = NULL;
2489         c->hcscale_fast = NULL;
2490 #if COMPILE_TEMPLATE_MMX2
2491     }
2492 #endif /* COMPILE_TEMPLATE_MMX2 */
2493     }
2494
2495 #if !COMPILE_TEMPLATE_MMX2
2496     switch(srcFormat) {
2497         case PIX_FMT_YUYV422  : c->chrToYV12 = RENAME(yuy2ToUV); break;
2498         case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;
2499         case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;
2500         case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;
2501         case PIX_FMT_YUV420P9LE:
2502         case PIX_FMT_YUV422P10LE:
2503         case PIX_FMT_YUV420P10LE: c->hScale16= RENAME(hScale16); break;
2504         default: break;
2505     }
2506 #endif /* !COMPILE_TEMPLATE_MMX2 */
2507     if (!c->chrSrcHSubSample) {
2508         switch(srcFormat) {
2509         case PIX_FMT_BGR24  : c->chrToYV12 = RENAME(bgr24ToUV); break;
2510         case PIX_FMT_RGB24  : c->chrToYV12 = RENAME(rgb24ToUV); break;
2511         default: break;
2512         }
2513     }
2514
2515     switch (srcFormat) {
2516 #if !COMPILE_TEMPLATE_MMX2
2517     case PIX_FMT_YUYV422  :
2518     case PIX_FMT_Y400A    : c->lumToYV12 = RENAME(yuy2ToY); break;
2519     case PIX_FMT_UYVY422  : c->lumToYV12 = RENAME(uyvyToY); break;
2520 #endif /* !COMPILE_TEMPLATE_MMX2 */
2521     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;
2522     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;
2523     default: break;
2524     }
2525 #if !COMPILE_TEMPLATE_MMX2
2526     if (c->alpPixBuf) {
2527         switch (srcFormat) {
2528         case PIX_FMT_Y400A  : c->alpToYV12 = RENAME(yuy2ToY); break;
2529         default: break;
2530         }
2531     }
2532 #endif /* !COMPILE_TEMPLATE_MMX2 */
2533     if(isAnyRGB(c->srcFormat) && av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1<15)
2534         c->hScale16= RENAME(hScale16);
2535     if(c->dstBpc > 10)
2536         c->hScale16 = NULL;
2537 }