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

   1 /*
   2  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav 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  * Libav 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 Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <stdint.h>
  22
  23 #include "libavutil/x86/asm.h"
  24 #include "libswscale/swscale_internal.h"
  25
  26 #undef REAL_MOVNTQ
  27 #undef MOVNTQ
  28 #undef PREFETCH
  29
  30 #if COMPILE_TEMPLATE_MMXEXT
  31 #define PREFETCH "prefetchnta"
  32 #else
  33 #define PREFETCH  " # nop"
  34 #endif
  35
  36 #if COMPILE_TEMPLATE_MMXEXT
  37 #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
  38 #else
  39 #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
  40 #endif
  41 #define MOVNTQ(a,b)  REAL_MOVNTQ(a,b)
  42
  43 #define YSCALEYUV2PACKEDX_UV \
  44     __asm__ volatile(\
  45         "xor                   %%"REG_a", %%"REG_a"     \n\t"\
  46         ".p2align                      4                \n\t"\
  47         "nop                                            \n\t"\
  48         "1:                                             \n\t"\
  49         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
  50         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
  51         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\
  52         "movq                      %%mm3, %%mm4         \n\t"\
  53         ".p2align                      4                \n\t"\
  54         "2:                                             \n\t"\
  55         "movq               8(%%"REG_d"), %%mm0         \n\t" /* filterCoeff */\
  56         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* UsrcData */\
  57         "add                          %6, %%"REG_S"     \n\t" \
  58         "movq     (%%"REG_S", %%"REG_a"), %%mm5         \n\t" /* VsrcData */\
  59         "add                         $16, %%"REG_d"     \n\t"\
  60         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
  61         "pmulhw                    %%mm0, %%mm2         \n\t"\
  62         "pmulhw                    %%mm0, %%mm5         \n\t"\
  63         "paddw                     %%mm2, %%mm3         \n\t"\
  64         "paddw                     %%mm5, %%mm4         \n\t"\
  65         "test                  %%"REG_S", %%"REG_S"     \n\t"\
  66         " jnz                         2b                \n\t"\
  67
  68 #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \
  69     "lea                "offset"(%0), %%"REG_d"     \n\t"\
  70     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
  71     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\
  72     "movq                    "#dst1", "#dst2"       \n\t"\
  73     ".p2align                      4                \n\t"\
  74     "2:                                             \n\t"\
  75     "movq               8(%%"REG_d"), "#coeff"      \n\t" /* filterCoeff */\
  76     "movq  (%%"REG_S", %%"REG_a", 2), "#src1"       \n\t" /* Y1srcData */\
  77     "movq 8(%%"REG_S", %%"REG_a", 2), "#src2"       \n\t" /* Y2srcData */\
  78     "add                         $16, %%"REG_d"            \n\t"\
  79     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
  80     "pmulhw                 "#coeff", "#src1"       \n\t"\
  81     "pmulhw                 "#coeff", "#src2"       \n\t"\
  82     "paddw                   "#src1", "#dst1"       \n\t"\
  83     "paddw                   "#src2", "#dst2"       \n\t"\
  84     "test                  %%"REG_S", %%"REG_S"     \n\t"\
  85     " jnz                         2b                \n\t"\
  86
  87 #define YSCALEYUV2PACKEDX \
  88     YSCALEYUV2PACKEDX_UV \
  89     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \
  90
  91 #define YSCALEYUV2PACKEDX_END                     \
  92         :: "r" (&c->redDither),                   \
  93             "m" (dummy), "m" (dummy), "m" (dummy),\
  94             "r" (dest), "m" (dstW_reg), "m"(uv_off) \
  95         : "%"REG_a, "%"REG_d, "%"REG_S            \
  96     );
  97
  98 #define YSCALEYUV2PACKEDX_ACCURATE_UV \
  99     __asm__ volatile(\
 100         "xor %%"REG_a", %%"REG_a"                       \n\t"\
 101         ".p2align                      4                \n\t"\
 102         "nop                                            \n\t"\
 103         "1:                                             \n\t"\
 104         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\
 105         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 106         "pxor                      %%mm4, %%mm4         \n\t"\
 107         "pxor                      %%mm5, %%mm5         \n\t"\
 108         "pxor                      %%mm6, %%mm6         \n\t"\
 109         "pxor                      %%mm7, %%mm7         \n\t"\
 110         ".p2align                      4                \n\t"\
 111         "2:                                             \n\t"\
 112         "movq     (%%"REG_S", %%"REG_a"), %%mm0         \n\t" /* UsrcData */\
 113         "add                          %6, %%"REG_S"      \n\t" \
 114         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* VsrcData */\
 115         "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 116         "movq     (%%"REG_S", %%"REG_a"), %%mm1         \n\t" /* UsrcData */\
 117         "movq                      %%mm0, %%mm3         \n\t"\
 118         "punpcklwd                 %%mm1, %%mm0         \n\t"\
 119         "punpckhwd                 %%mm1, %%mm3         \n\t"\
 120         "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1         \n\t" /* filterCoeff */\
 121         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 122         "pmaddwd                   %%mm1, %%mm3         \n\t"\
 123         "paddd                     %%mm0, %%mm4         \n\t"\
 124         "paddd                     %%mm3, %%mm5         \n\t"\
 125         "add                          %6, %%"REG_S"      \n\t" \
 126         "movq     (%%"REG_S", %%"REG_a"), %%mm3         \n\t" /* VsrcData */\
 127         "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 128         "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 129         "test                  %%"REG_S", %%"REG_S"     \n\t"\
 130         "movq                      %%mm2, %%mm0         \n\t"\
 131         "punpcklwd                 %%mm3, %%mm2         \n\t"\
 132         "punpckhwd                 %%mm3, %%mm0         \n\t"\
 133         "pmaddwd                   %%mm1, %%mm2         \n\t"\
 134         "pmaddwd                   %%mm1, %%mm0         \n\t"\
 135         "paddd                     %%mm2, %%mm6         \n\t"\
 136         "paddd                     %%mm0, %%mm7         \n\t"\
 137         " jnz                         2b                \n\t"\
 138         "psrad                       $16, %%mm4         \n\t"\
 139         "psrad                       $16, %%mm5         \n\t"\
 140         "psrad                       $16, %%mm6         \n\t"\
 141         "psrad                       $16, %%mm7         \n\t"\
 142         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 143         "packssdw                  %%mm5, %%mm4         \n\t"\
 144         "packssdw                  %%mm7, %%mm6         \n\t"\
 145         "paddw                     %%mm0, %%mm4         \n\t"\
 146         "paddw                     %%mm0, %%mm6         \n\t"\
 147         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\
 148         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\
 149
 150 #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \
 151     "lea                "offset"(%0), %%"REG_d"     \n\t"\
 152     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\
 153     "pxor                      %%mm1, %%mm1         \n\t"\
 154     "pxor                      %%mm5, %%mm5         \n\t"\
 155     "pxor                      %%mm7, %%mm7         \n\t"\
 156     "pxor                      %%mm6, %%mm6         \n\t"\
 157     ".p2align                      4                \n\t"\
 158     "2:                                             \n\t"\
 159     "movq  (%%"REG_S", %%"REG_a", 2), %%mm0         \n\t" /* Y1srcData */\
 160     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y2srcData */\
 161     "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\
 162     "movq  (%%"REG_S", %%"REG_a", 2), %%mm4         \n\t" /* Y1srcData */\
 163     "movq                      %%mm0, %%mm3         \n\t"\
 164     "punpcklwd                 %%mm4, %%mm0         \n\t"\
 165     "punpckhwd                 %%mm4, %%mm3         \n\t"\
 166     "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4         \n\t" /* filterCoeff */\
 167     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 168     "pmaddwd                   %%mm4, %%mm3         \n\t"\
 169     "paddd                     %%mm0, %%mm1         \n\t"\
 170     "paddd                     %%mm3, %%mm5         \n\t"\
 171     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3         \n\t" /* Y2srcData */\
 172     "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\
 173     "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\
 174     "test                  %%"REG_S", %%"REG_S"     \n\t"\
 175     "movq                      %%mm2, %%mm0         \n\t"\
 176     "punpcklwd                 %%mm3, %%mm2         \n\t"\
 177     "punpckhwd                 %%mm3, %%mm0         \n\t"\
 178     "pmaddwd                   %%mm4, %%mm2         \n\t"\
 179     "pmaddwd                   %%mm4, %%mm0         \n\t"\
 180     "paddd                     %%mm2, %%mm7         \n\t"\
 181     "paddd                     %%mm0, %%mm6         \n\t"\
 182     " jnz                         2b                \n\t"\
 183     "psrad                       $16, %%mm1         \n\t"\
 184     "psrad                       $16, %%mm5         \n\t"\
 185     "psrad                       $16, %%mm7         \n\t"\
 186     "psrad                       $16, %%mm6         \n\t"\
 187     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\
 188     "packssdw                  %%mm5, %%mm1         \n\t"\
 189     "packssdw                  %%mm6, %%mm7         \n\t"\
 190     "paddw                     %%mm0, %%mm1         \n\t"\
 191     "paddw                     %%mm0, %%mm7         \n\t"\
 192     "movq               "U_TEMP"(%0), %%mm3         \n\t"\
 193     "movq               "V_TEMP"(%0), %%mm4         \n\t"\
 194
 195 #define YSCALEYUV2PACKEDX_ACCURATE \
 196     YSCALEYUV2PACKEDX_ACCURATE_UV \
 197     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)
 198
 199 #define YSCALEYUV2RGBX \
 200     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\
 201     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\
 202     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\
 203     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\
 204     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\
 205     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\
 206     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 207     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\
 208     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\
 209     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\
 210     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\
 211     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\
 212     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\
 213     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 214     "paddw           %%mm3, %%mm4       \n\t"\
 215     "movq            %%mm2, %%mm0       \n\t"\
 216     "movq            %%mm5, %%mm6       \n\t"\
 217     "movq            %%mm4, %%mm3       \n\t"\
 218     "punpcklwd       %%mm2, %%mm2       \n\t"\
 219     "punpcklwd       %%mm5, %%mm5       \n\t"\
 220     "punpcklwd       %%mm4, %%mm4       \n\t"\
 221     "paddw           %%mm1, %%mm2       \n\t"\
 222     "paddw           %%mm1, %%mm5       \n\t"\
 223     "paddw           %%mm1, %%mm4       \n\t"\
 224     "punpckhwd       %%mm0, %%mm0       \n\t"\
 225     "punpckhwd       %%mm6, %%mm6       \n\t"\
 226     "punpckhwd       %%mm3, %%mm3       \n\t"\
 227     "paddw           %%mm7, %%mm0       \n\t"\
 228     "paddw           %%mm7, %%mm6       \n\t"\
 229     "paddw           %%mm7, %%mm3       \n\t"\
 230     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 231     "packuswb        %%mm0, %%mm2       \n\t"\
 232     "packuswb        %%mm6, %%mm5       \n\t"\
 233     "packuswb        %%mm3, %%mm4       \n\t"\
 234
 235 #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \
 236     "movq       "#b", "#q2"     \n\t" /* B */\
 237     "movq       "#r", "#t"      \n\t" /* R */\
 238     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\
 239     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\
 240     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\
 241     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\
 242     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\
 243     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\
 244     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\
 245     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\
 246     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\
 247     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\
 248 \
 249     MOVNTQ(   q0,   (dst, index, 4))\
 250     MOVNTQ(    b,  8(dst, index, 4))\
 251     MOVNTQ(   q2, 16(dst, index, 4))\
 252     MOVNTQ(   q3, 24(dst, index, 4))\
 253 \
 254     "add      $8, "#index"      \n\t"\
 255     "cmp "#dstw", "#index"      \n\t"\
 256     " jb      1b                \n\t"
 257 #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)
 258
 259 static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,
 260                                    const int16_t **lumSrc, int lumFilterSize,
 261                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 262                                    const int16_t **chrVSrc,
 263                                    int chrFilterSize, const int16_t **alpSrc,
 264                                    uint8_t *dest, int dstW, int dstY)
 265 {
 266     x86_reg dummy=0;
 267     x86_reg dstW_reg = dstW;
 268     x86_reg uv_off = c->uv_off_byte;
 269
 270     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 271         YSCALEYUV2PACKEDX_ACCURATE
 272         YSCALEYUV2RGBX
 273         "movq                      %%mm2, "U_TEMP"(%0)  \n\t"
 274         "movq                      %%mm4, "V_TEMP"(%0)  \n\t"
 275         "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"
 276         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
 277         "movq               "Y_TEMP"(%0), %%mm5         \n\t"
 278         "psraw                        $3, %%mm1         \n\t"
 279         "psraw                        $3, %%mm7         \n\t"
 280         "packuswb                  %%mm7, %%mm1         \n\t"
 281         WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)
 282         YSCALEYUV2PACKEDX_END
 283     } else {
 284         YSCALEYUV2PACKEDX_ACCURATE
 285         YSCALEYUV2RGBX
 286         "pcmpeqd %%mm7, %%mm7 \n\t"
 287         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 288         YSCALEYUV2PACKEDX_END
 289     }
 290 }
 291
 292 static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,
 293                                 const int16_t **lumSrc, int lumFilterSize,
 294                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 295                                 const int16_t **chrVSrc,
 296                                 int chrFilterSize, const int16_t **alpSrc,
 297                                 uint8_t *dest, int dstW, int dstY)
 298 {
 299     x86_reg dummy=0;
 300     x86_reg dstW_reg = dstW;
 301     x86_reg uv_off = c->uv_off_byte;
 302
 303     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 304         YSCALEYUV2PACKEDX
 305         YSCALEYUV2RGBX
 306         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)
 307         "psraw                        $3, %%mm1         \n\t"
 308         "psraw                        $3, %%mm7         \n\t"
 309         "packuswb                  %%mm7, %%mm1         \n\t"
 310         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 311         YSCALEYUV2PACKEDX_END
 312     } else {
 313         YSCALEYUV2PACKEDX
 314         YSCALEYUV2RGBX
 315         "pcmpeqd %%mm7, %%mm7 \n\t"
 316         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 317         YSCALEYUV2PACKEDX_END
 318     }
 319 }
 320
 321 #define REAL_WRITERGB16(dst, dstw, index) \
 322     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 323     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\
 324     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 325     "psrlq           $3, %%mm2  \n\t"\
 326 \
 327     "movq         %%mm2, %%mm1  \n\t"\
 328     "movq         %%mm4, %%mm3  \n\t"\
 329 \
 330     "punpcklbw    %%mm7, %%mm3  \n\t"\
 331     "punpcklbw    %%mm5, %%mm2  \n\t"\
 332     "punpckhbw    %%mm7, %%mm4  \n\t"\
 333     "punpckhbw    %%mm5, %%mm1  \n\t"\
 334 \
 335     "psllq           $3, %%mm3  \n\t"\
 336     "psllq           $3, %%mm4  \n\t"\
 337 \
 338     "por          %%mm3, %%mm2  \n\t"\
 339     "por          %%mm4, %%mm1  \n\t"\
 340 \
 341     MOVNTQ(%%mm2,  (dst, index, 2))\
 342     MOVNTQ(%%mm1, 8(dst, index, 2))\
 343 \
 344     "add             $8, "#index"   \n\t"\
 345     "cmp        "#dstw", "#index"   \n\t"\
 346     " jb             1b             \n\t"
 347 #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)
 348
 349 static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,
 350                                     const int16_t **lumSrc, int lumFilterSize,
 351                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 352                                     const int16_t **chrVSrc,
 353                                     int chrFilterSize, const int16_t **alpSrc,
 354                                     uint8_t *dest, int dstW, int dstY)
 355 {
 356     x86_reg dummy=0;
 357     x86_reg dstW_reg = dstW;
 358     x86_reg uv_off = c->uv_off_byte;
 359
 360     YSCALEYUV2PACKEDX_ACCURATE
 361     YSCALEYUV2RGBX
 362     "pxor %%mm7, %%mm7 \n\t"
 363     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 364 #ifdef DITHER1XBPP
 365     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 366     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 367     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 368 #endif
 369     WRITERGB16(%4, %5, %%REGa)
 370     YSCALEYUV2PACKEDX_END
 371 }
 372
 373 static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,
 374                                  const int16_t **lumSrc, int lumFilterSize,
 375                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 376                                  const int16_t **chrVSrc,
 377                                  int chrFilterSize, const int16_t **alpSrc,
 378                                  uint8_t *dest, int dstW, int dstY)
 379 {
 380     x86_reg dummy=0;
 381     x86_reg dstW_reg = dstW;
 382     x86_reg uv_off = c->uv_off_byte;
 383
 384     YSCALEYUV2PACKEDX
 385     YSCALEYUV2RGBX
 386     "pxor %%mm7, %%mm7 \n\t"
 387     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 388 #ifdef DITHER1XBPP
 389     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 390     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 391     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 392 #endif
 393     WRITERGB16(%4, %5, %%REGa)
 394     YSCALEYUV2PACKEDX_END
 395 }
 396
 397 #define REAL_WRITERGB15(dst, dstw, index) \
 398     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\
 399     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\
 400     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\
 401     "psrlq           $3, %%mm2  \n\t"\
 402     "psrlq           $1, %%mm5  \n\t"\
 403 \
 404     "movq         %%mm2, %%mm1  \n\t"\
 405     "movq         %%mm4, %%mm3  \n\t"\
 406 \
 407     "punpcklbw    %%mm7, %%mm3  \n\t"\
 408     "punpcklbw    %%mm5, %%mm2  \n\t"\
 409     "punpckhbw    %%mm7, %%mm4  \n\t"\
 410     "punpckhbw    %%mm5, %%mm1  \n\t"\
 411 \
 412     "psllq           $2, %%mm3  \n\t"\
 413     "psllq           $2, %%mm4  \n\t"\
 414 \
 415     "por          %%mm3, %%mm2  \n\t"\
 416     "por          %%mm4, %%mm1  \n\t"\
 417 \
 418     MOVNTQ(%%mm2,  (dst, index, 2))\
 419     MOVNTQ(%%mm1, 8(dst, index, 2))\
 420 \
 421     "add             $8, "#index"   \n\t"\
 422     "cmp        "#dstw", "#index"   \n\t"\
 423     " jb             1b             \n\t"
 424 #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)
 425
 426 static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,
 427                                     const int16_t **lumSrc, int lumFilterSize,
 428                                     const int16_t *chrFilter, const int16_t **chrUSrc,
 429                                     const int16_t **chrVSrc,
 430                                     int chrFilterSize, const int16_t **alpSrc,
 431                                     uint8_t *dest, int dstW, int dstY)
 432 {
 433     x86_reg dummy=0;
 434     x86_reg dstW_reg = dstW;
 435     x86_reg uv_off = c->uv_off_byte;
 436
 437     YSCALEYUV2PACKEDX_ACCURATE
 438     YSCALEYUV2RGBX
 439     "pxor %%mm7, %%mm7 \n\t"
 440     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 441 #ifdef DITHER1XBPP
 442     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"
 443     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"
 444     "paddusb "RED_DITHER"(%0), %%mm5\n\t"
 445 #endif
 446     WRITERGB15(%4, %5, %%REGa)
 447     YSCALEYUV2PACKEDX_END
 448 }
 449
 450 static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
 451                                  const int16_t **lumSrc, int lumFilterSize,
 452                                  const int16_t *chrFilter, const int16_t **chrUSrc,
 453                                  const int16_t **chrVSrc,
 454                                  int chrFilterSize, const int16_t **alpSrc,
 455                                  uint8_t *dest, int dstW, int dstY)
 456 {
 457     x86_reg dummy=0;
 458     x86_reg dstW_reg = dstW;
 459     x86_reg uv_off = c->uv_off_byte;
 460
 461     YSCALEYUV2PACKEDX
 462     YSCALEYUV2RGBX
 463     "pxor %%mm7, %%mm7 \n\t"
 464     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 465 #ifdef DITHER1XBPP
 466     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"
 467     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"
 468     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"
 469 #endif
 470     WRITERGB15(%4, %5, %%REGa)
 471     YSCALEYUV2PACKEDX_END
 472 }
 473
 474 #define WRITEBGR24MMX(dst, dstw, index) \
 475     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 476     "movq      %%mm2, %%mm1     \n\t" /* B */\
 477     "movq      %%mm5, %%mm6     \n\t" /* R */\
 478     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\
 479     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\
 480     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\
 481     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\
 482     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\
 483     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\
 484     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\
 485     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\
 486     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\
 487     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\
 488 \
 489     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\
 490     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\
 491     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\
 492     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\
 493 \
 494     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\
 495     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\
 496     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\
 497     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\
 498 \
 499     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\
 500     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\
 501     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\
 502     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\
 503 \
 504     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\
 505     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\
 506     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\
 507     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\
 508     MOVNTQ(%%mm0, (dst))\
 509 \
 510     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\
 511     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\
 512     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\
 513     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\
 514     MOVNTQ(%%mm6, 8(dst))\
 515 \
 516     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\
 517     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\
 518     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\
 519     MOVNTQ(%%mm5, 16(dst))\
 520 \
 521     "add         $24, "#dst"    \n\t"\
 522 \
 523     "add          $8, "#index"  \n\t"\
 524     "cmp     "#dstw", "#index"  \n\t"\
 525     " jb          1b            \n\t"
 526
 527 #define WRITEBGR24MMXEXT(dst, dstw, index) \
 528     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
 529     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\
 530     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\
 531     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
 532     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
 533     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
 534 \
 535     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\
 536     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\
 537     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\
 538 \
 539     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\
 540     "por    %%mm1, %%mm6        \n\t"\
 541     "por    %%mm3, %%mm6        \n\t"\
 542     MOVNTQ(%%mm6, (dst))\
 543 \
 544     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
 545     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
 546     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
 547     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
 548 \
 549     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\
 550     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\
 551     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\
 552 \
 553     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\
 554     "por    %%mm3, %%mm6        \n\t"\
 555     MOVNTQ(%%mm6, 8(dst))\
 556 \
 557     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
 558     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
 559     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
 560 \
 561     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\
 562     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\
 563     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\
 564 \
 565     "por    %%mm1, %%mm3        \n\t"\
 566     "por    %%mm3, %%mm6        \n\t"\
 567     MOVNTQ(%%mm6, 16(dst))\
 568 \
 569     "add      $24, "#dst"       \n\t"\
 570 \
 571     "add       $8, "#index"     \n\t"\
 572     "cmp  "#dstw", "#index"     \n\t"\
 573     " jb       1b               \n\t"
 574
 575 #if COMPILE_TEMPLATE_MMXEXT
 576 #undef WRITEBGR24
 577 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMXEXT(dst, dstw, index)
 578 #else
 579 #undef WRITEBGR24
 580 #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)
 581 #endif
 582
 583 static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
 584                                    const int16_t **lumSrc, int lumFilterSize,
 585                                    const int16_t *chrFilter, const int16_t **chrUSrc,
 586                                    const int16_t **chrVSrc,
 587                                    int chrFilterSize, const int16_t **alpSrc,
 588                                    uint8_t *dest, int dstW, int dstY)
 589 {
 590     x86_reg dummy=0;
 591     x86_reg dstW_reg = dstW;
 592     x86_reg uv_off = c->uv_off_byte;
 593
 594     YSCALEYUV2PACKEDX_ACCURATE
 595     YSCALEYUV2RGBX
 596     "pxor %%mm7, %%mm7 \n\t"
 597     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize
 598     "add %4, %%"REG_c"                        \n\t"
 599     WRITEBGR24(%%REGc, %5, %%REGa)
 600     :: "r" (&c->redDither),
 601        "m" (dummy), "m" (dummy), "m" (dummy),
 602        "r" (dest), "m" (dstW_reg), "m"(uv_off)
 603     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 604     );
 605 }
 606
 607 static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
 608                                 const int16_t **lumSrc, int lumFilterSize,
 609                                 const int16_t *chrFilter, const int16_t **chrUSrc,
 610                                 const int16_t **chrVSrc,
 611                                 int chrFilterSize, const int16_t **alpSrc,
 612                                 uint8_t *dest, int dstW, int dstY)
 613 {
 614     x86_reg dummy=0;
 615     x86_reg dstW_reg = dstW;
 616     x86_reg uv_off = c->uv_off_byte;
 617
 618     YSCALEYUV2PACKEDX
 619     YSCALEYUV2RGBX
 620     "pxor                    %%mm7, %%mm7       \n\t"
 621     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize
 622     "add                        %4, %%"REG_c"   \n\t"
 623     WRITEBGR24(%%REGc, %5, %%REGa)
 624     :: "r" (&c->redDither),
 625        "m" (dummy), "m" (dummy), "m" (dummy),
 626        "r" (dest),  "m" (dstW_reg), "m"(uv_off)
 627     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
 628     );
 629 }
 630
 631 #define REAL_WRITEYUY2(dst, dstw, index) \
 632     "packuswb  %%mm3, %%mm3     \n\t"\
 633     "packuswb  %%mm4, %%mm4     \n\t"\
 634     "packuswb  %%mm7, %%mm1     \n\t"\
 635     "punpcklbw %%mm4, %%mm3     \n\t"\
 636     "movq      %%mm1, %%mm7     \n\t"\
 637     "punpcklbw %%mm3, %%mm1     \n\t"\
 638     "punpckhbw %%mm3, %%mm7     \n\t"\
 639 \
 640     MOVNTQ(%%mm1, (dst, index, 2))\
 641     MOVNTQ(%%mm7, 8(dst, index, 2))\
 642 \
 643     "add          $8, "#index"  \n\t"\
 644     "cmp     "#dstw", "#index"  \n\t"\
 645     " jb          1b            \n\t"
 646 #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)
 647
 648 static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,
 649                                      const int16_t **lumSrc, int lumFilterSize,
 650                                      const int16_t *chrFilter, const int16_t **chrUSrc,
 651                                      const int16_t **chrVSrc,
 652                                      int chrFilterSize, const int16_t **alpSrc,
 653                                      uint8_t *dest, int dstW, int dstY)
 654 {
 655     x86_reg dummy=0;
 656     x86_reg dstW_reg = dstW;
 657     x86_reg uv_off = c->uv_off_byte;
 658
 659     YSCALEYUV2PACKEDX_ACCURATE
 660     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 661     "psraw $3, %%mm3    \n\t"
 662     "psraw $3, %%mm4    \n\t"
 663     "psraw $3, %%mm1    \n\t"
 664     "psraw $3, %%mm7    \n\t"
 665     WRITEYUY2(%4, %5, %%REGa)
 666     YSCALEYUV2PACKEDX_END
 667 }
 668
 669 static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
 670                                   const int16_t **lumSrc, int lumFilterSize,
 671                                   const int16_t *chrFilter, const int16_t **chrUSrc,
 672                                   const int16_t **chrVSrc,
 673                                   int chrFilterSize, const int16_t **alpSrc,
 674                                   uint8_t *dest, int dstW, int dstY)
 675 {
 676     x86_reg dummy=0;
 677     x86_reg dstW_reg = dstW;
 678     x86_reg uv_off = c->uv_off_byte;
 679
 680     YSCALEYUV2PACKEDX
 681     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 682     "psraw $3, %%mm3    \n\t"
 683     "psraw $3, %%mm4    \n\t"
 684     "psraw $3, %%mm1    \n\t"
 685     "psraw $3, %%mm7    \n\t"
 686     WRITEYUY2(%4, %5, %%REGa)
 687     YSCALEYUV2PACKEDX_END
 688 }
 689
 690 #define REAL_YSCALEYUV2RGB_UV(index, c) \
 691     "xor            "#index", "#index"  \n\t"\
 692     ".p2align              4            \n\t"\
 693     "1:                                 \n\t"\
 694     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 695     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 696     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 697     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 698     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 699     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 700     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
 701     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
 702     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
 703     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
 704     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
 705     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 706     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 707     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
 708     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
 709     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 710     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 711     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 712     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 713     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 714     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 715     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 716
 717 #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \
 718     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
 719     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
 720     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
 721     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
 722     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
 723     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
 724     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 725     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 726     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 727     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 728     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 729     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 730
 731 #define REAL_YSCALEYUV2RGB_COEFF(c) \
 732     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 733     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
 734     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
 735     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
 736     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
 737     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
 738     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
 739     "paddw             %%mm3, %%mm4     \n\t"\
 740     "movq              %%mm2, %%mm0     \n\t"\
 741     "movq              %%mm5, %%mm6     \n\t"\
 742     "movq              %%mm4, %%mm3     \n\t"\
 743     "punpcklwd         %%mm2, %%mm2     \n\t"\
 744     "punpcklwd         %%mm5, %%mm5     \n\t"\
 745     "punpcklwd         %%mm4, %%mm4     \n\t"\
 746     "paddw             %%mm1, %%mm2     \n\t"\
 747     "paddw             %%mm1, %%mm5     \n\t"\
 748     "paddw             %%mm1, %%mm4     \n\t"\
 749     "punpckhwd         %%mm0, %%mm0     \n\t"\
 750     "punpckhwd         %%mm6, %%mm6     \n\t"\
 751     "punpckhwd         %%mm3, %%mm3     \n\t"\
 752     "paddw             %%mm7, %%mm0     \n\t"\
 753     "paddw             %%mm7, %%mm6     \n\t"\
 754     "paddw             %%mm7, %%mm3     \n\t"\
 755     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
 756     "packuswb          %%mm0, %%mm2     \n\t"\
 757     "packuswb          %%mm6, %%mm5     \n\t"\
 758     "packuswb          %%mm3, %%mm4     \n\t"\
 759
 760 #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)
 761
 762 #define YSCALEYUV2RGB(index, c) \
 763     REAL_YSCALEYUV2RGB_UV(index, c) \
 764     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \
 765     REAL_YSCALEYUV2RGB_COEFF(c)
 766
 767 /**
 768  * vertical bilinear scale YV12 to RGB
 769  */
 770 static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],
 771                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
 772                                 const int16_t *abuf[2], uint8_t *dest,
 773                                 int dstW, int yalpha, int uvalpha, int y)
 774 {
 775     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 776                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 777
 778     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
 779         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];
 780 #if ARCH_X86_64
 781         __asm__ volatile(
 782             YSCALEYUV2RGB(%%r8, %5)
 783             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)
 784             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 785             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 786             "packuswb            %%mm7, %%mm1       \n\t"
 787             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 788             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),
 789                "a" (&c->redDither),
 790                "r" (abuf0), "r" (abuf1)
 791             : "%r8"
 792         );
 793 #else
 794         *(const uint16_t **)(&c->u_temp)=abuf0;
 795         *(const uint16_t **)(&c->v_temp)=abuf1;
 796         __asm__ volatile(
 797             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 798             "mov        %4, %%"REG_b"               \n\t"
 799             "push %%"REG_BP"                        \n\t"
 800             YSCALEYUV2RGB(%%REGBP, %5)
 801             "push                   %0              \n\t"
 802             "push                   %1              \n\t"
 803             "mov          "U_TEMP"(%5), %0          \n\t"
 804             "mov          "V_TEMP"(%5), %1          \n\t"
 805             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)
 806             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 807             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/
 808             "packuswb            %%mm7, %%mm1       \n\t"
 809             "pop                    %1              \n\t"
 810             "pop                    %0              \n\t"
 811             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)
 812             "pop %%"REG_BP"                         \n\t"
 813             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 814             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 815                "a" (&c->redDither)
 816         );
 817 #endif
 818     } else {
 819         __asm__ volatile(
 820             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 821             "mov        %4, %%"REG_b"               \n\t"
 822             "push %%"REG_BP"                        \n\t"
 823             YSCALEYUV2RGB(%%REGBP, %5)
 824             "pcmpeqd %%mm7, %%mm7                   \n\t"
 825             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
 826             "pop %%"REG_BP"                         \n\t"
 827             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 828             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 829                "a" (&c->redDither)
 830         );
 831     }
 832 }
 833
 834 static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],
 835                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
 836                                 const int16_t *abuf[2], uint8_t *dest,
 837                                 int dstW, int yalpha, int uvalpha, int y)
 838 {
 839     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 840                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 841
 842     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 843     __asm__ volatile(
 844         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 845         "mov        %4, %%"REG_b"               \n\t"
 846         "push %%"REG_BP"                        \n\t"
 847         YSCALEYUV2RGB(%%REGBP, %5)
 848         "pxor    %%mm7, %%mm7                   \n\t"
 849         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
 850         "pop %%"REG_BP"                         \n\t"
 851         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 852         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 853            "a" (&c->redDither)
 854     );
 855 }
 856
 857 static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],
 858                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
 859                                  const int16_t *abuf[2], uint8_t *dest,
 860                                  int dstW, int yalpha, int uvalpha, int y)
 861 {
 862     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 863                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 864
 865     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 866     __asm__ volatile(
 867         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 868         "mov        %4, %%"REG_b"               \n\t"
 869         "push %%"REG_BP"                        \n\t"
 870         YSCALEYUV2RGB(%%REGBP, %5)
 871         "pxor    %%mm7, %%mm7                   \n\t"
 872         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 873 #ifdef DITHER1XBPP
 874         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
 875         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
 876         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
 877 #endif
 878         WRITERGB15(%%REGb, 8280(%5), %%REGBP)
 879         "pop %%"REG_BP"                         \n\t"
 880         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 881         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 882            "a" (&c->redDither)
 883     );
 884 }
 885
 886 static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
 887                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
 888                                  const int16_t *abuf[2], uint8_t *dest,
 889                                  int dstW, int yalpha, int uvalpha, int y)
 890 {
 891     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 892                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 893
 894     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 895     __asm__ volatile(
 896         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 897         "mov        %4, %%"REG_b"               \n\t"
 898         "push %%"REG_BP"                        \n\t"
 899         YSCALEYUV2RGB(%%REGBP, %5)
 900         "pxor    %%mm7, %%mm7                   \n\t"
 901         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
 902 #ifdef DITHER1XBPP
 903         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
 904         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
 905         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
 906 #endif
 907         WRITERGB16(%%REGb, 8280(%5), %%REGBP)
 908         "pop %%"REG_BP"                         \n\t"
 909         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 910         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 911            "a" (&c->redDither)
 912     );
 913 }
 914
 915 #define REAL_YSCALEYUV2PACKED(index, c) \
 916     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\
 917     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\
 918     "psraw                $3, %%mm0                           \n\t"\
 919     "psraw                $3, %%mm1                           \n\t"\
 920     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\
 921     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\
 922     "xor            "#index", "#index"                        \n\t"\
 923     ".p2align              4            \n\t"\
 924     "1:                                 \n\t"\
 925     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
 926     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
 927     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 928     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
 929     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
 930     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 931     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
 932     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
 933     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
 934     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
 935     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
 936     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 937     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 938     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
 939     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
 940     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\
 941     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\
 942     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\
 943     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\
 944     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\
 945     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\
 946     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 947     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
 948     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 949     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 950     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 951     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
 952
 953 #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)
 954
 955 static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
 956                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
 957                                   const int16_t *abuf[2], uint8_t *dest,
 958                                   int dstW, int yalpha, int uvalpha, int y)
 959 {
 960     const int16_t *buf0  = buf[0],  *buf1  = buf[1],
 961                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];
 962
 963     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(
 964     __asm__ volatile(
 965         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
 966         "mov %4, %%"REG_b"                        \n\t"
 967         "push %%"REG_BP"                        \n\t"
 968         YSCALEYUV2PACKED(%%REGBP, %5)
 969         WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
 970         "pop %%"REG_BP"                         \n\t"
 971         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
 972         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
 973            "a" (&c->redDither)
 974     );
 975 }
 976
 977 #define REAL_YSCALEYUV2RGB1(index, c) \
 978     "xor            "#index", "#index"  \n\t"\
 979     ".p2align              4            \n\t"\
 980     "1:                                 \n\t"\
 981     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
 982     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 983     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
 984     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
 985     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
 986     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
 987     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
 988     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
 989     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
 990     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
 991     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
 992     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
 993     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
 994     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
 995     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
 996     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 997     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
 998     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
 999     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1000     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1001     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1002     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1003     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1004     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1005     "paddw             %%mm3, %%mm4     \n\t"\
1006     "movq              %%mm2, %%mm0     \n\t"\
1007     "movq              %%mm5, %%mm6     \n\t"\
1008     "movq              %%mm4, %%mm3     \n\t"\
1009     "punpcklwd         %%mm2, %%mm2     \n\t"\
1010     "punpcklwd         %%mm5, %%mm5     \n\t"\
1011     "punpcklwd         %%mm4, %%mm4     \n\t"\
1012     "paddw             %%mm1, %%mm2     \n\t"\
1013     "paddw             %%mm1, %%mm5     \n\t"\
1014     "paddw             %%mm1, %%mm4     \n\t"\
1015     "punpckhwd         %%mm0, %%mm0     \n\t"\
1016     "punpckhwd         %%mm6, %%mm6     \n\t"\
1017     "punpckhwd         %%mm3, %%mm3     \n\t"\
1018     "paddw             %%mm7, %%mm0     \n\t"\
1019     "paddw             %%mm7, %%mm6     \n\t"\
1020     "paddw             %%mm7, %%mm3     \n\t"\
1021     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1022     "packuswb          %%mm0, %%mm2     \n\t"\
1023     "packuswb          %%mm6, %%mm5     \n\t"\
1024     "packuswb          %%mm3, %%mm4     \n\t"\
1025
1026 #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)
1027
1028 // do vertical chrominance interpolation
1029 #define REAL_YSCALEYUV2RGB1b(index, c) \
1030     "xor            "#index", "#index"  \n\t"\
1031     ".p2align              4            \n\t"\
1032     "1:                                 \n\t"\
1033     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1034     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1035     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1036     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1037     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1038     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1039     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1040     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1041     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
1042     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\
1043     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
1044     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\
1045     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\
1046     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\
1047     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\
1048     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\
1049     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
1050     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1051     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1052     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1053     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\
1054     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\
1055     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\
1056     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\
1057     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\
1058     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\
1059     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\
1060     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
1061     "paddw             %%mm3, %%mm4     \n\t"\
1062     "movq              %%mm2, %%mm0     \n\t"\
1063     "movq              %%mm5, %%mm6     \n\t"\
1064     "movq              %%mm4, %%mm3     \n\t"\
1065     "punpcklwd         %%mm2, %%mm2     \n\t"\
1066     "punpcklwd         %%mm5, %%mm5     \n\t"\
1067     "punpcklwd         %%mm4, %%mm4     \n\t"\
1068     "paddw             %%mm1, %%mm2     \n\t"\
1069     "paddw             %%mm1, %%mm5     \n\t"\
1070     "paddw             %%mm1, %%mm4     \n\t"\
1071     "punpckhwd         %%mm0, %%mm0     \n\t"\
1072     "punpckhwd         %%mm6, %%mm6     \n\t"\
1073     "punpckhwd         %%mm3, %%mm3     \n\t"\
1074     "paddw             %%mm7, %%mm0     \n\t"\
1075     "paddw             %%mm7, %%mm6     \n\t"\
1076     "paddw             %%mm7, %%mm3     \n\t"\
1077     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
1078     "packuswb          %%mm0, %%mm2     \n\t"\
1079     "packuswb          %%mm6, %%mm5     \n\t"\
1080     "packuswb          %%mm3, %%mm4     \n\t"\
1081
1082 #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)
1083
1084 #define REAL_YSCALEYUV2RGB1_ALPHA(index) \
1085     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\
1086     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\
1087     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\
1088     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\
1089     "packuswb          %%mm1, %%mm7     \n\t"
1090 #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)
1091
1092 /**
1093  * YV12 to RGB without scaling or interpolating
1094  */
1095 static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,
1096                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1097                                 const int16_t *abuf0, uint8_t *dest,
1098                                 int dstW, int uvalpha, int y)
1099 {
1100     const int16_t *ubuf0 = ubuf[0];
1101     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1102
1103     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1104         const int16_t *ubuf1 = ubuf[0];
1105         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1106             __asm__ volatile(
1107                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1108                 "mov        %4, %%"REG_b"               \n\t"
1109                 "push %%"REG_BP"                        \n\t"
1110                 YSCALEYUV2RGB1(%%REGBP, %5)
1111                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1112                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1113                 "pop %%"REG_BP"                         \n\t"
1114                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1115                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1116                    "a" (&c->redDither)
1117             );
1118         } else {
1119             __asm__ volatile(
1120                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1121                 "mov        %4, %%"REG_b"               \n\t"
1122                 "push %%"REG_BP"                        \n\t"
1123                 YSCALEYUV2RGB1(%%REGBP, %5)
1124                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1125                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1126                 "pop %%"REG_BP"                         \n\t"
1127                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1128                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1129                    "a" (&c->redDither)
1130             );
1131         }
1132     } else {
1133         const int16_t *ubuf1 = ubuf[1];
1134         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1135             __asm__ volatile(
1136                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1137                 "mov        %4, %%"REG_b"               \n\t"
1138                 "push %%"REG_BP"                        \n\t"
1139                 YSCALEYUV2RGB1b(%%REGBP, %5)
1140                 YSCALEYUV2RGB1_ALPHA(%%REGBP)
1141                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1142                 "pop %%"REG_BP"                         \n\t"
1143                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1144                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1145                    "a" (&c->redDither)
1146             );
1147         } else {
1148             __asm__ volatile(
1149                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1150                 "mov        %4, %%"REG_b"               \n\t"
1151                 "push %%"REG_BP"                        \n\t"
1152                 YSCALEYUV2RGB1b(%%REGBP, %5)
1153                 "pcmpeqd %%mm7, %%mm7                   \n\t"
1154                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)
1155                 "pop %%"REG_BP"                         \n\t"
1156                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1157                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1158                    "a" (&c->redDither)
1159             );
1160         }
1161     }
1162 }
1163
1164 static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,
1165                                 const int16_t *ubuf[2], const int16_t *vbuf[2],
1166                                 const int16_t *abuf0, uint8_t *dest,
1167                                 int dstW, int uvalpha, int y)
1168 {
1169     const int16_t *ubuf0 = ubuf[0];
1170     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1171
1172     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1173         const int16_t *ubuf1 = ubuf[0];
1174         __asm__ volatile(
1175             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1176             "mov        %4, %%"REG_b"               \n\t"
1177             "push %%"REG_BP"                        \n\t"
1178             YSCALEYUV2RGB1(%%REGBP, %5)
1179             "pxor    %%mm7, %%mm7                   \n\t"
1180             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1181             "pop %%"REG_BP"                         \n\t"
1182             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1183             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1184                "a" (&c->redDither)
1185         );
1186     } else {
1187         const int16_t *ubuf1 = ubuf[1];
1188         __asm__ volatile(
1189             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1190             "mov        %4, %%"REG_b"               \n\t"
1191             "push %%"REG_BP"                        \n\t"
1192             YSCALEYUV2RGB1b(%%REGBP, %5)
1193             "pxor    %%mm7, %%mm7                   \n\t"
1194             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)
1195             "pop %%"REG_BP"                         \n\t"
1196             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1197             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1198                "a" (&c->redDither)
1199         );
1200     }
1201 }
1202
1203 static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,
1204                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1205                                  const int16_t *abuf0, uint8_t *dest,
1206                                  int dstW, int uvalpha, int y)
1207 {
1208     const int16_t *ubuf0 = ubuf[0];
1209     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1210
1211     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1212         const int16_t *ubuf1 = ubuf[0];
1213         __asm__ volatile(
1214             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1215             "mov        %4, %%"REG_b"               \n\t"
1216             "push %%"REG_BP"                        \n\t"
1217             YSCALEYUV2RGB1(%%REGBP, %5)
1218             "pxor    %%mm7, %%mm7                   \n\t"
1219             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1220 #ifdef DITHER1XBPP
1221             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1222             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1223             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1224 #endif
1225             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1226             "pop %%"REG_BP"                         \n\t"
1227             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1228             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1229                "a" (&c->redDither)
1230         );
1231     } else {
1232         const int16_t *ubuf1 = ubuf[1];
1233         __asm__ volatile(
1234             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1235             "mov        %4, %%"REG_b"               \n\t"
1236             "push %%"REG_BP"                        \n\t"
1237             YSCALEYUV2RGB1b(%%REGBP, %5)
1238             "pxor    %%mm7, %%mm7                   \n\t"
1239             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1240 #ifdef DITHER1XBPP
1241             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1242             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1243             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1244 #endif
1245             WRITERGB15(%%REGb, 8280(%5), %%REGBP)
1246             "pop %%"REG_BP"                         \n\t"
1247             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1248             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1249                "a" (&c->redDither)
1250         );
1251     }
1252 }
1253
1254 static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
1255                                  const int16_t *ubuf[2], const int16_t *vbuf[2],
1256                                  const int16_t *abuf0, uint8_t *dest,
1257                                  int dstW, int uvalpha, int y)
1258 {
1259     const int16_t *ubuf0 = ubuf[0];
1260     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1261
1262     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1263         const int16_t *ubuf1 = ubuf[0];
1264         __asm__ volatile(
1265             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1266             "mov        %4, %%"REG_b"               \n\t"
1267             "push %%"REG_BP"                        \n\t"
1268             YSCALEYUV2RGB1(%%REGBP, %5)
1269             "pxor    %%mm7, %%mm7                   \n\t"
1270             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1271 #ifdef DITHER1XBPP
1272             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1273             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1274             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1275 #endif
1276             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1277             "pop %%"REG_BP"                         \n\t"
1278             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1279             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1280                "a" (&c->redDither)
1281         );
1282     } else {
1283         const int16_t *ubuf1 = ubuf[1];
1284         __asm__ volatile(
1285             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1286             "mov        %4, %%"REG_b"               \n\t"
1287             "push %%"REG_BP"                        \n\t"
1288             YSCALEYUV2RGB1b(%%REGBP, %5)
1289             "pxor    %%mm7, %%mm7                   \n\t"
1290             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
1291 #ifdef DITHER1XBPP
1292             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"
1293             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"
1294             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"
1295 #endif
1296             WRITERGB16(%%REGb, 8280(%5), %%REGBP)
1297             "pop %%"REG_BP"                         \n\t"
1298             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1299             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1300                "a" (&c->redDither)
1301         );
1302     }
1303 }
1304
1305 #define REAL_YSCALEYUV2PACKED1(index, c) \
1306     "xor            "#index", "#index"  \n\t"\
1307     ".p2align              4            \n\t"\
1308     "1:                                 \n\t"\
1309     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
1310     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1311     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
1312     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1313     "psraw                $7, %%mm3     \n\t" \
1314     "psraw                $7, %%mm4     \n\t" \
1315     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1316     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1317     "psraw                $7, %%mm1     \n\t" \
1318     "psraw                $7, %%mm7     \n\t" \
1319
1320 #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)
1321
1322 #define REAL_YSCALEYUV2PACKED1b(index, c) \
1323     "xor "#index", "#index"             \n\t"\
1324     ".p2align              4            \n\t"\
1325     "1:                                 \n\t"\
1326     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
1327     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
1328     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1329     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
1330     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
1331     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
1332     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
1333     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
1334     "psrlw                $8, %%mm3     \n\t" \
1335     "psrlw                $8, %%mm4     \n\t" \
1336     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
1337     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\
1338     "psraw                $7, %%mm1     \n\t" \
1339     "psraw                $7, %%mm7     \n\t"
1340 #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)
1341
1342 static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,
1343                                   const int16_t *ubuf[2], const int16_t *vbuf[2],
1344                                   const int16_t *abuf0, uint8_t *dest,
1345                                   int dstW, int uvalpha, int y)
1346 {
1347     const int16_t *ubuf0 = ubuf[0];
1348     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
1349
1350     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster
1351         const int16_t *ubuf1 = ubuf[0];
1352         __asm__ volatile(
1353             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1354             "mov        %4, %%"REG_b"               \n\t"
1355             "push %%"REG_BP"                        \n\t"
1356             YSCALEYUV2PACKED1(%%REGBP, %5)
1357             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1358             "pop %%"REG_BP"                         \n\t"
1359             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1360             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1361                "a" (&c->redDither)
1362         );
1363     } else {
1364         const int16_t *ubuf1 = ubuf[1];
1365         __asm__ volatile(
1366             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"
1367             "mov        %4, %%"REG_b"               \n\t"
1368             "push %%"REG_BP"                        \n\t"
1369             YSCALEYUV2PACKED1b(%%REGBP, %5)
1370             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)
1371             "pop %%"REG_BP"                         \n\t"
1372             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"
1373             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),
1374                "a" (&c->redDither)
1375         );
1376     }
1377 }
1378
1379 #if COMPILE_TEMPLATE_MMXEXT
1380 static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
1381                                  int dstWidth, const uint8_t *src,
1382                                  int srcW, int xInc)
1383 {
1384     int32_t *filterPos = c->hLumFilterPos;
1385     int16_t *filter    = c->hLumFilter;
1386     void    *mmxextFilterCode = c->lumMmxextFilterCode;
1387     int i;
1388 #if defined(PIC)
1389     uint64_t ebxsave;
1390 #endif
1391 #if ARCH_X86_64
1392     uint64_t retsave;
1393 #endif
1394
1395     __asm__ volatile(
1396 #if defined(PIC)
1397         "mov               %%"REG_b", %5        \n\t"
1398 #if ARCH_X86_64
1399         "mov               -8(%%rsp), %%"REG_a" \n\t"
1400         "mov               %%"REG_a", %6        \n\t"
1401 #endif
1402 #else
1403 #if ARCH_X86_64
1404         "mov               -8(%%rsp), %%"REG_a" \n\t"
1405         "mov               %%"REG_a", %5        \n\t"
1406 #endif
1407 #endif
1408         "pxor                  %%mm7, %%mm7     \n\t"
1409         "mov                      %0, %%"REG_c" \n\t"
1410         "mov                      %1, %%"REG_D" \n\t"
1411         "mov                      %2, %%"REG_d" \n\t"
1412         "mov                      %3, %%"REG_b" \n\t"
1413         "xor               %%"REG_a", %%"REG_a" \n\t" // i
1414         PREFETCH"        (%%"REG_c")            \n\t"
1415         PREFETCH"      32(%%"REG_c")            \n\t"
1416         PREFETCH"      64(%%"REG_c")            \n\t"
1417
1418 #if ARCH_X86_64
1419 #define CALL_MMXEXT_FILTER_CODE \
1420         "movl            (%%"REG_b"), %%esi     \n\t"\
1421         "call                    *%4            \n\t"\
1422         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\
1423         "add               %%"REG_S", %%"REG_c" \n\t"\
1424         "add               %%"REG_a", %%"REG_D" \n\t"\
1425         "xor               %%"REG_a", %%"REG_a" \n\t"\
1426
1427 #else
1428 #define CALL_MMXEXT_FILTER_CODE \
1429         "movl (%%"REG_b"), %%esi        \n\t"\
1430         "call         *%4                       \n\t"\
1431         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\
1432         "add               %%"REG_a", %%"REG_D" \n\t"\
1433         "xor               %%"REG_a", %%"REG_a" \n\t"\
1434
1435 #endif /* ARCH_X86_64 */
1436
1437         CALL_MMXEXT_FILTER_CODE
1438         CALL_MMXEXT_FILTER_CODE
1439         CALL_MMXEXT_FILTER_CODE
1440         CALL_MMXEXT_FILTER_CODE
1441         CALL_MMXEXT_FILTER_CODE
1442         CALL_MMXEXT_FILTER_CODE
1443         CALL_MMXEXT_FILTER_CODE
1444         CALL_MMXEXT_FILTER_CODE
1445
1446 #if defined(PIC)
1447         "mov                      %5, %%"REG_b" \n\t"
1448 #if ARCH_X86_64
1449         "mov                      %6, %%"REG_a" \n\t"
1450         "mov               %%"REG_a", -8(%%rsp) \n\t"
1451 #endif
1452 #else
1453 #if ARCH_X86_64
1454         "mov                      %5, %%"REG_a" \n\t"
1455         "mov               %%"REG_a", -8(%%rsp) \n\t"
1456 #endif
1457 #endif
1458         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
1459            "m" (mmxextFilterCode)
1460 #if defined(PIC)
1461           ,"m" (ebxsave)
1462 #endif
1463 #if ARCH_X86_64
1464           ,"m"(retsave)
1465 #endif
1466         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1467 #if !defined(PIC)
1468          ,"%"REG_b
1469 #endif
1470     );
1471
1472     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
1473         dst[i] = src[srcW-1]*128;
1474 }
1475
1476 static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
1477                                  int dstWidth, const uint8_t *src1,
1478                                  const uint8_t *src2, int srcW, int xInc)
1479 {
1480     int32_t *filterPos = c->hChrFilterPos;
1481     int16_t *filter    = c->hChrFilter;
1482     void    *mmxextFilterCode = c->chrMmxextFilterCode;
1483     int i;
1484 #if defined(PIC)
1485     DECLARE_ALIGNED(8, uint64_t, ebxsave);
1486 #endif
1487 #if ARCH_X86_64
1488     DECLARE_ALIGNED(8, uint64_t, retsave);
1489 #endif
1490
1491     __asm__ volatile(
1492 #if defined(PIC)
1493         "mov          %%"REG_b", %7         \n\t"
1494 #if ARCH_X86_64
1495         "mov          -8(%%rsp), %%"REG_a"  \n\t"
1496         "mov          %%"REG_a", %8         \n\t"
1497 #endif
1498 #else
1499 #if ARCH_X86_64
1500         "mov          -8(%%rsp), %%"REG_a"  \n\t"
1501         "mov          %%"REG_a", %7         \n\t"
1502 #endif
1503 #endif
1504         "pxor             %%mm7, %%mm7      \n\t"
1505         "mov                 %0, %%"REG_c"  \n\t"
1506         "mov                 %1, %%"REG_D"  \n\t"
1507         "mov                 %2, %%"REG_d"  \n\t"
1508         "mov                 %3, %%"REG_b"  \n\t"
1509         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
1510         PREFETCH"   (%%"REG_c")             \n\t"
1511         PREFETCH" 32(%%"REG_c")             \n\t"
1512         PREFETCH" 64(%%"REG_c")             \n\t"
1513
1514         CALL_MMXEXT_FILTER_CODE
1515         CALL_MMXEXT_FILTER_CODE
1516         CALL_MMXEXT_FILTER_CODE
1517         CALL_MMXEXT_FILTER_CODE
1518         "xor          %%"REG_a", %%"REG_a"  \n\t" // i
1519         "mov                 %5, %%"REG_c"  \n\t" // src
1520         "mov                 %6, %%"REG_D"  \n\t" // buf2
1521         PREFETCH"   (%%"REG_c")             \n\t"
1522         PREFETCH" 32(%%"REG_c")             \n\t"
1523         PREFETCH" 64(%%"REG_c")             \n\t"
1524
1525         CALL_MMXEXT_FILTER_CODE
1526         CALL_MMXEXT_FILTER_CODE
1527         CALL_MMXEXT_FILTER_CODE
1528         CALL_MMXEXT_FILTER_CODE
1529
1530 #if defined(PIC)
1531         "mov %7, %%"REG_b"    \n\t"
1532 #if ARCH_X86_64
1533         "mov                 %8, %%"REG_a"  \n\t"
1534         "mov          %%"REG_a", -8(%%rsp)  \n\t"
1535 #endif
1536 #else
1537 #if ARCH_X86_64
1538         "mov                 %7, %%"REG_a"  \n\t"
1539         "mov          %%"REG_a", -8(%%rsp)  \n\t"
1540 #endif
1541 #endif
1542         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
1543            "m" (mmxextFilterCode), "m" (src2), "m"(dst2)
1544 #if defined(PIC)
1545           ,"m" (ebxsave)
1546 #endif
1547 #if ARCH_X86_64
1548           ,"m"(retsave)
1549 #endif
1550         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D
1551 #if !defined(PIC)
1552          ,"%"REG_b
1553 #endif
1554     );
1555
1556     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
1557         dst1[i] = src1[srcW-1]*128;
1558         dst2[i] = src2[srcW-1]*128;
1559     }
1560 }
1561 #endif /* COMPILE_TEMPLATE_MMXEXT */
1562
1563 static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
1564 {
1565     enum AVPixelFormat dstFormat = c->dstFormat;
1566
1567     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
1568         dstFormat != AV_PIX_FMT_NV12 && dstFormat != AV_PIX_FMT_NV21) {
1569         if (!(c->flags & SWS_BITEXACT)) {
1570             if (c->flags & SWS_ACCURATE_RND) {
1571                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1572                     switch (c->dstFormat) {
1573                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;
1574                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;
1575                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;
1576                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;
1577                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
1578                     default: break;
1579                     }
1580                 }
1581             } else {
1582                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1583                     switch (c->dstFormat) {
1584                     case AV_PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;
1585                     case AV_PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;
1586                     case AV_PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;
1587                     case AV_PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;
1588                     case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
1589                     default: break;
1590                     }
1591                 }
1592             }
1593         }
1594         if (!(c->flags & SWS_FULL_CHR_H_INT)) {
1595             switch (c->dstFormat) {
1596             case AV_PIX_FMT_RGB32:
1597                 c->yuv2packed1 = RENAME(yuv2rgb32_1);
1598                 c->yuv2packed2 = RENAME(yuv2rgb32_2);
1599                 break;
1600             case AV_PIX_FMT_BGR24:
1601                 c->yuv2packed1 = RENAME(yuv2bgr24_1);
1602                 c->yuv2packed2 = RENAME(yuv2bgr24_2);
1603                 break;
1604             case AV_PIX_FMT_RGB555:
1605                 c->yuv2packed1 = RENAME(yuv2rgb555_1);
1606                 c->yuv2packed2 = RENAME(yuv2rgb555_2);
1607                 break;
1608             case AV_PIX_FMT_RGB565:
1609                 c->yuv2packed1 = RENAME(yuv2rgb565_1);
1610                 c->yuv2packed2 = RENAME(yuv2rgb565_2);
1611                 break;
1612             case AV_PIX_FMT_YUYV422:
1613                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);
1614                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);
1615                 break;
1616             default:
1617                 break;
1618             }
1619         }
1620     }
1621
1622     if (c->srcBpc == 8 && c->dstBpc <= 10) {
1623     // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one).
1624 #if COMPILE_TEMPLATE_MMXEXT
1625     if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) {
1626         c->hyscale_fast = RENAME(hyscale_fast);
1627         c->hcscale_fast = RENAME(hcscale_fast);
1628     } else {
1629 #endif /* COMPILE_TEMPLATE_MMXEXT */
1630         c->hyscale_fast = NULL;
1631         c->hcscale_fast = NULL;
1632 #if COMPILE_TEMPLATE_MMXEXT
1633     }
1634 #endif /* COMPILE_TEMPLATE_MMXEXT */
1635     }
1636 }