falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
30594 Commits
32 Branches
303MB
Tree: 6b94711f15
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '6b94711f15'
${ noResults }
FFmpeg/libswscale/x86/swscale_template.c

2007 lines
88KB
Raw Blame History 

  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. #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. #if !COMPILE_TEMPLATE_MMX2

  39. static av_always_inline void

  40. dither_8to16(SwsContext *c, const uint8_t *srcDither, int rot)

  41. {

  42.     if (rot) {

  43.         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"

  44.                          "movq       (%0), %%mm3\n\t"

  45.                          "movq      %%mm3, %%mm4\n\t"

  46.                          "psrlq       $24, %%mm3\n\t"

  47.                          "psllq       $40, %%mm4\n\t"

  48.                          "por       %%mm4, %%mm3\n\t"

  49.                          "movq      %%mm3, %%mm4\n\t"

  50.                          "punpcklbw %%mm0, %%mm3\n\t"

  51.                          "punpckhbw %%mm0, %%mm4\n\t"

  52.                          "psraw        $4, %%mm3\n\t"

  53.                          "psraw        $4, %%mm4\n\t"

  54.                          "movq      %%mm3, "DITHER16"+0(%1)\n\t"

  55.                          "movq      %%mm4, "DITHER16"+8(%1)\n\t"

  56.                          :: "r"(srcDither), "r"(&c->redDither)

  57.                          );

  58.     } else {

  59.         __asm__ volatile("pxor      %%mm0, %%mm0\n\t"

  60.                          "movq       (%0), %%mm3\n\t"

  61.                          "movq      %%mm3, %%mm4\n\t"

  62.                          "punpcklbw %%mm0, %%mm3\n\t"

  63.                          "punpckhbw %%mm0, %%mm4\n\t"

  64.                          "psraw        $4, %%mm3\n\t"

  65.                          "psraw        $4, %%mm4\n\t"

  66.                          "movq      %%mm3, "DITHER16"+0(%1)\n\t"

  67.                          "movq      %%mm4, "DITHER16"+8(%1)\n\t"

  68.                          :: "r"(srcDither), "r"(&c->redDither)

  69.                          );

  70.     }

  71. }

  72. #endif

  73. 

  74. static void RENAME(yuv2yuv1)(SwsContext *c, const int16_t *lumSrc,

  75.                              const int16_t *chrUSrc, const int16_t *chrVSrc,

  76.                              const int16_t *alpSrc,

  77.                              uint8_t *dst[4], int dstW, int chrDstW)

  78. {

  79.     int p= 4;

  80.     const int16_t *src[4]= {

  81.         lumSrc + dstW,     chrUSrc + chrDstW,

  82.         chrVSrc + chrDstW, alpSrc + dstW

  83.     };

  84.     x86_reg counter[4]= { dstW, chrDstW, chrDstW, dstW };

  85. 

  86.     while (p--) {

  87.         if (dst[p]) {

  88.             __asm__ volatile(

  89.                 "mov %2, %%"REG_a"                    \n\t"

  90.                 ".p2align               4             \n\t" /* FIXME Unroll? */

  91.                 "1:                                   \n\t"

  92.                 "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"

  93.                 "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"

  94.                 "psraw                 $7, %%mm0      \n\t"

  95.                 "psraw                 $7, %%mm1      \n\t"

  96.                 "packuswb           %%mm1, %%mm0      \n\t"

  97.                 MOVNTQ(%%mm0, (%1, %%REGa))

  98.                 "add                   $8, %%"REG_a"  \n\t"

  99.                 "jnc                   1b             \n\t"

  100.                 :: "r" (src[p]), "r" (dst[p] + counter[p]),

  101.                    "g" (-counter[p])

  102.                 : "%"REG_a

  103.             );

  104.         }

  105.     }

  106. }

  107. 

  108. static void RENAME(yuv2yuv1_ar)(SwsContext *c, const int16_t *lumSrc,

  109.                                 const int16_t *chrUSrc, const int16_t *chrVSrc,

  110.                                 const int16_t *alpSrc,

  111.                                 uint8_t *dst[4], int dstW, int chrDstW)

  112. {

  113.     int p= 4;

  114.     const int16_t *src[4]= {

  115.         lumSrc + dstW,     chrUSrc + chrDstW,

  116.         chrVSrc + chrDstW, alpSrc + dstW

  117.     };

  118.     x86_reg counter[4]= { dstW, chrDstW, chrDstW, dstW };

  119.     const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;

  120. 

  121.     while (p--) {

  122.         if (dst[p]) {

  123.             dither_8to16(c, (p == 2 || p == 3) ? chrDither : lumDither, p == 2);

  124.             __asm__ volatile(

  125.                 "mov %2, %%"REG_a"                    \n\t"

  126.                 "movq    "DITHER16"+0(%3), %%mm6      \n\t"

  127.                 "movq    "DITHER16"+8(%3), %%mm7      \n\t"

  128.                 ".p2align                4            \n\t" /* FIXME Unroll? */

  129.                 "1:                                   \n\t"

  130.                 "movq  (%0, %%"REG_a", 2), %%mm0      \n\t"

  131.                 "movq 8(%0, %%"REG_a", 2), %%mm1      \n\t"

  132.                 "paddsw             %%mm6, %%mm0      \n\t"

  133.                 "paddsw             %%mm7, %%mm1      \n\t"

  134.                 "psraw                 $7, %%mm0      \n\t"

  135.                 "psraw                 $7, %%mm1      \n\t"

  136.                 "packuswb           %%mm1, %%mm0      \n\t"

  137.                 MOVNTQ(%%mm0, (%1, %%REGa))

  138.                 "add                   $8, %%"REG_a"  \n\t"

  139.                 "jnc                   1b             \n\t"

  140.                 :: "r" (src[p]), "r" (dst[p] + counter[p]),

  141.                    "g" (-counter[p]), "r"(&c->redDither)

  142.                 : "%"REG_a

  143.             );

  144.         }

  145.     }

  146. }

  147. 

  148. #define YSCALEYUV2PACKEDX_UV \

  149.     __asm__ volatile(\

  150.         "xor                   %%"REG_a", %%"REG_a"     \n\t"\

  151.         ".p2align                      4                \n\t"\

  152.         "nop                                            \n\t"\

  153.         "1:                                             \n\t"\

  154.         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\

  155.         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  156.         "movq      "VROUNDER_OFFSET"(%0), %%mm3         \n\t"\

  157.         "movq                      %%mm3, %%mm4         \n\t"\

  158.         ".p2align                      4                \n\t"\

  159.         "2:                                             \n\t"\

  160.         "movq               8(%%"REG_d"), %%mm0         \n\t" /* filterCoeff */\

  161.         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* UsrcData */\

  162.         "add                          %6, %%"REG_S"     \n\t" \

  163.         "movq     (%%"REG_S", %%"REG_a"), %%mm5         \n\t" /* VsrcData */\

  164.         "add                         $16, %%"REG_d"     \n\t"\

  165.         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  166.         "pmulhw                    %%mm0, %%mm2         \n\t"\

  167.         "pmulhw                    %%mm0, %%mm5         \n\t"\

  168.         "paddw                     %%mm2, %%mm3         \n\t"\

  169.         "paddw                     %%mm5, %%mm4         \n\t"\

  170.         "test                  %%"REG_S", %%"REG_S"     \n\t"\

  171.         " jnz                         2b                \n\t"\

  172. 

  173. #define YSCALEYUV2PACKEDX_YA(offset,coeff,src1,src2,dst1,dst2) \

  174.     "lea                "offset"(%0), %%"REG_d"     \n\t"\

  175.     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  176.     "movq      "VROUNDER_OFFSET"(%0), "#dst1"       \n\t"\

  177.     "movq                    "#dst1", "#dst2"       \n\t"\

  178.     ".p2align                      4                \n\t"\

  179.     "2:                                             \n\t"\

  180.     "movq               8(%%"REG_d"), "#coeff"      \n\t" /* filterCoeff */\

  181.     "movq  (%%"REG_S", %%"REG_a", 2), "#src1"       \n\t" /* Y1srcData */\

  182.     "movq 8(%%"REG_S", %%"REG_a", 2), "#src2"       \n\t" /* Y2srcData */\

  183.     "add                         $16, %%"REG_d"            \n\t"\

  184.     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  185.     "pmulhw                 "#coeff", "#src1"       \n\t"\

  186.     "pmulhw                 "#coeff", "#src2"       \n\t"\

  187.     "paddw                   "#src1", "#dst1"       \n\t"\

  188.     "paddw                   "#src2", "#dst2"       \n\t"\

  189.     "test                  %%"REG_S", %%"REG_S"     \n\t"\

  190.     " jnz                         2b                \n\t"\

  191. 

  192. #define YSCALEYUV2PACKEDX \

  193.     YSCALEYUV2PACKEDX_UV \

  194.     YSCALEYUV2PACKEDX_YA(LUM_MMX_FILTER_OFFSET,%%mm0,%%mm2,%%mm5,%%mm1,%%mm7) \

  195. 

  196. #define YSCALEYUV2PACKEDX_END                     \

  197.         :: "r" (&c->redDither),                   \

  198.             "m" (dummy), "m" (dummy), "m" (dummy),\

  199.             "r" (dest), "m" (dstW_reg), "m"(uv_off) \

  200.         : "%"REG_a, "%"REG_d, "%"REG_S            \

  201.     );

  202. 

  203. #define YSCALEYUV2PACKEDX_ACCURATE_UV \

  204.     __asm__ volatile(\

  205.         "xor %%"REG_a", %%"REG_a"                       \n\t"\

  206.         ".p2align                      4                \n\t"\

  207.         "nop                                            \n\t"\

  208.         "1:                                             \n\t"\

  209.         "lea "CHR_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\

  210.         "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  211.         "pxor                      %%mm4, %%mm4         \n\t"\

  212.         "pxor                      %%mm5, %%mm5         \n\t"\

  213.         "pxor                      %%mm6, %%mm6         \n\t"\

  214.         "pxor                      %%mm7, %%mm7         \n\t"\

  215.         ".p2align                      4                \n\t"\

  216.         "2:                                             \n\t"\

  217.         "movq     (%%"REG_S", %%"REG_a"), %%mm0         \n\t" /* UsrcData */\

  218.         "add                          %6, %%"REG_S"      \n\t" \

  219.         "movq     (%%"REG_S", %%"REG_a"), %%mm2         \n\t" /* VsrcData */\

  220.         "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\

  221.         "movq     (%%"REG_S", %%"REG_a"), %%mm1         \n\t" /* UsrcData */\

  222.         "movq                      %%mm0, %%mm3         \n\t"\

  223.         "punpcklwd                 %%mm1, %%mm0         \n\t"\

  224.         "punpckhwd                 %%mm1, %%mm3         \n\t"\

  225.         "movq "STR(APCK_COEF)"(%%"REG_d"),%%mm1         \n\t" /* filterCoeff */\

  226.         "pmaddwd                   %%mm1, %%mm0         \n\t"\

  227.         "pmaddwd                   %%mm1, %%mm3         \n\t"\

  228.         "paddd                     %%mm0, %%mm4         \n\t"\

  229.         "paddd                     %%mm3, %%mm5         \n\t"\

  230.         "add                          %6, %%"REG_S"      \n\t" \

  231.         "movq     (%%"REG_S", %%"REG_a"), %%mm3         \n\t" /* VsrcData */\

  232.         "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\

  233.         "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\

  234.         "test                  %%"REG_S", %%"REG_S"     \n\t"\

  235.         "movq                      %%mm2, %%mm0         \n\t"\

  236.         "punpcklwd                 %%mm3, %%mm2         \n\t"\

  237.         "punpckhwd                 %%mm3, %%mm0         \n\t"\

  238.         "pmaddwd                   %%mm1, %%mm2         \n\t"\

  239.         "pmaddwd                   %%mm1, %%mm0         \n\t"\

  240.         "paddd                     %%mm2, %%mm6         \n\t"\

  241.         "paddd                     %%mm0, %%mm7         \n\t"\

  242.         " jnz                         2b                \n\t"\

  243.         "psrad                       $16, %%mm4         \n\t"\

  244.         "psrad                       $16, %%mm5         \n\t"\

  245.         "psrad                       $16, %%mm6         \n\t"\

  246.         "psrad                       $16, %%mm7         \n\t"\

  247.         "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\

  248.         "packssdw                  %%mm5, %%mm4         \n\t"\

  249.         "packssdw                  %%mm7, %%mm6         \n\t"\

  250.         "paddw                     %%mm0, %%mm4         \n\t"\

  251.         "paddw                     %%mm0, %%mm6         \n\t"\

  252.         "movq                      %%mm4, "U_TEMP"(%0)  \n\t"\

  253.         "movq                      %%mm6, "V_TEMP"(%0)  \n\t"\

  254. 

  255. #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \

  256.     "lea                "offset"(%0), %%"REG_d"     \n\t"\

  257.     "mov                 (%%"REG_d"), %%"REG_S"     \n\t"\

  258.     "pxor                      %%mm1, %%mm1         \n\t"\

  259.     "pxor                      %%mm5, %%mm5         \n\t"\

  260.     "pxor                      %%mm7, %%mm7         \n\t"\

  261.     "pxor                      %%mm6, %%mm6         \n\t"\

  262.     ".p2align                      4                \n\t"\

  263.     "2:                                             \n\t"\

  264.     "movq  (%%"REG_S", %%"REG_a", 2), %%mm0         \n\t" /* Y1srcData */\

  265.     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y2srcData */\

  266.     "mov "STR(APCK_PTR2)"(%%"REG_d"), %%"REG_S"     \n\t"\

  267.     "movq  (%%"REG_S", %%"REG_a", 2), %%mm4         \n\t" /* Y1srcData */\

  268.     "movq                      %%mm0, %%mm3         \n\t"\

  269.     "punpcklwd                 %%mm4, %%mm0         \n\t"\

  270.     "punpckhwd                 %%mm4, %%mm3         \n\t"\

  271.     "movq "STR(APCK_COEF)"(%%"REG_d"), %%mm4         \n\t" /* filterCoeff */\

  272.     "pmaddwd                   %%mm4, %%mm0         \n\t"\

  273.     "pmaddwd                   %%mm4, %%mm3         \n\t"\

  274.     "paddd                     %%mm0, %%mm1         \n\t"\

  275.     "paddd                     %%mm3, %%mm5         \n\t"\

  276.     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm3         \n\t" /* Y2srcData */\

  277.     "mov "STR(APCK_SIZE)"(%%"REG_d"), %%"REG_S"     \n\t"\

  278.     "add           $"STR(APCK_SIZE)", %%"REG_d"     \n\t"\

  279.     "test                  %%"REG_S", %%"REG_S"     \n\t"\

  280.     "movq                      %%mm2, %%mm0         \n\t"\

  281.     "punpcklwd                 %%mm3, %%mm2         \n\t"\

  282.     "punpckhwd                 %%mm3, %%mm0         \n\t"\

  283.     "pmaddwd                   %%mm4, %%mm2         \n\t"\

  284.     "pmaddwd                   %%mm4, %%mm0         \n\t"\

  285.     "paddd                     %%mm2, %%mm7         \n\t"\

  286.     "paddd                     %%mm0, %%mm6         \n\t"\

  287.     " jnz                         2b                \n\t"\

  288.     "psrad                       $16, %%mm1         \n\t"\

  289.     "psrad                       $16, %%mm5         \n\t"\

  290.     "psrad                       $16, %%mm7         \n\t"\

  291.     "psrad                       $16, %%mm6         \n\t"\

  292.     "movq      "VROUNDER_OFFSET"(%0), %%mm0         \n\t"\

  293.     "packssdw                  %%mm5, %%mm1         \n\t"\

  294.     "packssdw                  %%mm6, %%mm7         \n\t"\

  295.     "paddw                     %%mm0, %%mm1         \n\t"\

  296.     "paddw                     %%mm0, %%mm7         \n\t"\

  297.     "movq               "U_TEMP"(%0), %%mm3         \n\t"\

  298.     "movq               "V_TEMP"(%0), %%mm4         \n\t"\

  299. 

  300. #define YSCALEYUV2PACKEDX_ACCURATE \

  301.     YSCALEYUV2PACKEDX_ACCURATE_UV \

  302.     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)

  303. 

  304. #define YSCALEYUV2RGBX \

  305.     "psubw  "U_OFFSET"(%0), %%mm3       \n\t" /* (U-128)8*/\

  306.     "psubw  "V_OFFSET"(%0), %%mm4       \n\t" /* (V-128)8*/\

  307.     "movq            %%mm3, %%mm2       \n\t" /* (U-128)8*/\

  308.     "movq            %%mm4, %%mm5       \n\t" /* (V-128)8*/\

  309.     "pmulhw "UG_COEFF"(%0), %%mm3       \n\t"\

  310.     "pmulhw "VG_COEFF"(%0), %%mm4       \n\t"\

  311.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  312.     "pmulhw "UB_COEFF"(%0), %%mm2       \n\t"\

  313.     "pmulhw "VR_COEFF"(%0), %%mm5       \n\t"\

  314.     "psubw  "Y_OFFSET"(%0), %%mm1       \n\t" /* 8(Y-16)*/\

  315.     "psubw  "Y_OFFSET"(%0), %%mm7       \n\t" /* 8(Y-16)*/\

  316.     "pmulhw  "Y_COEFF"(%0), %%mm1       \n\t"\

  317.     "pmulhw  "Y_COEFF"(%0), %%mm7       \n\t"\

  318.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  319.     "paddw           %%mm3, %%mm4       \n\t"\

  320.     "movq            %%mm2, %%mm0       \n\t"\

  321.     "movq            %%mm5, %%mm6       \n\t"\

  322.     "movq            %%mm4, %%mm3       \n\t"\

  323.     "punpcklwd       %%mm2, %%mm2       \n\t"\

  324.     "punpcklwd       %%mm5, %%mm5       \n\t"\

  325.     "punpcklwd       %%mm4, %%mm4       \n\t"\

  326.     "paddw           %%mm1, %%mm2       \n\t"\

  327.     "paddw           %%mm1, %%mm5       \n\t"\

  328.     "paddw           %%mm1, %%mm4       \n\t"\

  329.     "punpckhwd       %%mm0, %%mm0       \n\t"\

  330.     "punpckhwd       %%mm6, %%mm6       \n\t"\

  331.     "punpckhwd       %%mm3, %%mm3       \n\t"\

  332.     "paddw           %%mm7, %%mm0       \n\t"\

  333.     "paddw           %%mm7, %%mm6       \n\t"\

  334.     "paddw           %%mm7, %%mm3       \n\t"\

  335.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  336.     "packuswb        %%mm0, %%mm2       \n\t"\

  337.     "packuswb        %%mm6, %%mm5       \n\t"\

  338.     "packuswb        %%mm3, %%mm4       \n\t"\

  339. 

  340. #define REAL_WRITEBGR32(dst, dstw, index, b, g, r, a, q0, q2, q3, t) \

  341.     "movq       "#b", "#q2"     \n\t" /* B */\

  342.     "movq       "#r", "#t"      \n\t" /* R */\

  343.     "punpcklbw  "#g", "#b"      \n\t" /* GBGBGBGB 0 */\

  344.     "punpcklbw  "#a", "#r"      \n\t" /* ARARARAR 0 */\

  345.     "punpckhbw  "#g", "#q2"     \n\t" /* GBGBGBGB 2 */\

  346.     "punpckhbw  "#a", "#t"      \n\t" /* ARARARAR 2 */\

  347.     "movq       "#b", "#q0"     \n\t" /* GBGBGBGB 0 */\

  348.     "movq      "#q2", "#q3"     \n\t" /* GBGBGBGB 2 */\

  349.     "punpcklwd  "#r", "#q0"     \n\t" /* ARGBARGB 0 */\

  350.     "punpckhwd  "#r", "#b"      \n\t" /* ARGBARGB 1 */\

  351.     "punpcklwd  "#t", "#q2"     \n\t" /* ARGBARGB 2 */\

  352.     "punpckhwd  "#t", "#q3"     \n\t" /* ARGBARGB 3 */\

  353. \

  354.     MOVNTQ(   q0,   (dst, index, 4))\

  355.     MOVNTQ(    b,  8(dst, index, 4))\

  356.     MOVNTQ(   q2, 16(dst, index, 4))\

  357.     MOVNTQ(   q3, 24(dst, index, 4))\

  358. \

  359.     "add      $8, "#index"      \n\t"\

  360.     "cmp "#dstw", "#index"      \n\t"\

  361.     " jb      1b                \n\t"

  362. #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)

  363. 

  364. static void RENAME(yuv2rgb32_X_ar)(SwsContext *c, const int16_t *lumFilter,

  365.                                    const int16_t **lumSrc, int lumFilterSize,

  366.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  367.                                    const int16_t **chrVSrc,

  368.                                    int chrFilterSize, const int16_t **alpSrc,

  369.                                    uint8_t *dest, int dstW, int dstY)

  370. {

  371.     x86_reg dummy=0;

  372.     x86_reg dstW_reg = dstW;

  373.     x86_reg uv_off = c->uv_off_byte;

  374. 

  375.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  376.         YSCALEYUV2PACKEDX_ACCURATE

  377.         YSCALEYUV2RGBX

  378.         "movq                      %%mm2, "U_TEMP"(%0)  \n\t"

  379.         "movq                      %%mm4, "V_TEMP"(%0)  \n\t"

  380.         "movq                      %%mm5, "Y_TEMP"(%0)  \n\t"

  381.         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)

  382.         "movq               "Y_TEMP"(%0), %%mm5         \n\t"

  383.         "psraw                        $3, %%mm1         \n\t"

  384.         "psraw                        $3, %%mm7         \n\t"

  385.         "packuswb                  %%mm7, %%mm1         \n\t"

  386.         WRITEBGR32(%4, %5, %%REGa, %%mm3, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm2, %%mm6)

  387.         YSCALEYUV2PACKEDX_END

  388.     } else {

  389.         YSCALEYUV2PACKEDX_ACCURATE

  390.         YSCALEYUV2RGBX

  391.         "pcmpeqd %%mm7, %%mm7 \n\t"

  392.         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  393.         YSCALEYUV2PACKEDX_END

  394.     }

  395. }

  396. 

  397. static void RENAME(yuv2rgb32_X)(SwsContext *c, const int16_t *lumFilter,

  398.                                 const int16_t **lumSrc, int lumFilterSize,

  399.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  400.                                 const int16_t **chrVSrc,

  401.                                 int chrFilterSize, const int16_t **alpSrc,

  402.                                 uint8_t *dest, int dstW, int dstY)

  403. {

  404.     x86_reg dummy=0;

  405.     x86_reg dstW_reg = dstW;

  406.     x86_reg uv_off = c->uv_off_byte;

  407. 

  408.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  409.         YSCALEYUV2PACKEDX

  410.         YSCALEYUV2RGBX

  411.         YSCALEYUV2PACKEDX_YA(ALP_MMX_FILTER_OFFSET, %%mm0, %%mm3, %%mm6, %%mm1, %%mm7)

  412.         "psraw                        $3, %%mm1         \n\t"

  413.         "psraw                        $3, %%mm7         \n\t"

  414.         "packuswb                  %%mm7, %%mm1         \n\t"

  415.         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  416.         YSCALEYUV2PACKEDX_END

  417.     } else {

  418.         YSCALEYUV2PACKEDX

  419.         YSCALEYUV2RGBX

  420.         "pcmpeqd %%mm7, %%mm7 \n\t"

  421.         WRITEBGR32(%4, %5, %%REGa, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  422.         YSCALEYUV2PACKEDX_END

  423.     }

  424. }

  425. 

  426. #define REAL_WRITERGB16(dst, dstw, index) \

  427.     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\

  428.     "pand "MANGLE(bFC)", %%mm4  \n\t" /* G */\

  429.     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\

  430.     "psrlq           $3, %%mm2  \n\t"\

  431. \

  432.     "movq         %%mm2, %%mm1  \n\t"\

  433.     "movq         %%mm4, %%mm3  \n\t"\

  434. \

  435.     "punpcklbw    %%mm7, %%mm3  \n\t"\

  436.     "punpcklbw    %%mm5, %%mm2  \n\t"\

  437.     "punpckhbw    %%mm7, %%mm4  \n\t"\

  438.     "punpckhbw    %%mm5, %%mm1  \n\t"\

  439. \

  440.     "psllq           $3, %%mm3  \n\t"\

  441.     "psllq           $3, %%mm4  \n\t"\

  442. \

  443.     "por          %%mm3, %%mm2  \n\t"\

  444.     "por          %%mm4, %%mm1  \n\t"\

  445. \

  446.     MOVNTQ(%%mm2,  (dst, index, 2))\

  447.     MOVNTQ(%%mm1, 8(dst, index, 2))\

  448. \

  449.     "add             $8, "#index"   \n\t"\

  450.     "cmp        "#dstw", "#index"   \n\t"\

  451.     " jb             1b             \n\t"

  452. #define WRITERGB16(dst, dstw, index)  REAL_WRITERGB16(dst, dstw, index)

  453. 

  454. static void RENAME(yuv2rgb565_X_ar)(SwsContext *c, const int16_t *lumFilter,

  455.                                     const int16_t **lumSrc, int lumFilterSize,

  456.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  457.                                     const int16_t **chrVSrc,

  458.                                     int chrFilterSize, const int16_t **alpSrc,

  459.                                     uint8_t *dest, int dstW, int dstY)

  460. {

  461.     x86_reg dummy=0;

  462.     x86_reg dstW_reg = dstW;

  463.     x86_reg uv_off = c->uv_off_byte;

  464. 

  465.     YSCALEYUV2PACKEDX_ACCURATE

  466.     YSCALEYUV2RGBX

  467.     "pxor %%mm7, %%mm7 \n\t"

  468.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  469. #ifdef DITHER1XBPP

  470.     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"

  471.     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"

  472.     "paddusb "RED_DITHER"(%0), %%mm5\n\t"

  473. #endif

  474.     WRITERGB16(%4, %5, %%REGa)

  475.     YSCALEYUV2PACKEDX_END

  476. }

  477. 

  478. static void RENAME(yuv2rgb565_X)(SwsContext *c, const int16_t *lumFilter,

  479.                                  const int16_t **lumSrc, int lumFilterSize,

  480.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  481.                                  const int16_t **chrVSrc,

  482.                                  int chrFilterSize, const int16_t **alpSrc,

  483.                                  uint8_t *dest, int dstW, int dstY)

  484. {

  485.     x86_reg dummy=0;

  486.     x86_reg dstW_reg = dstW;

  487.     x86_reg uv_off = c->uv_off_byte;

  488. 

  489.     YSCALEYUV2PACKEDX

  490.     YSCALEYUV2RGBX

  491.     "pxor %%mm7, %%mm7 \n\t"

  492.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  493. #ifdef DITHER1XBPP

  494.     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"

  495.     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"

  496.     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"

  497. #endif

  498.     WRITERGB16(%4, %5, %%REGa)

  499.     YSCALEYUV2PACKEDX_END

  500. }

  501. 

  502. #define REAL_WRITERGB15(dst, dstw, index) \

  503.     "pand "MANGLE(bF8)", %%mm2  \n\t" /* B */\

  504.     "pand "MANGLE(bF8)", %%mm4  \n\t" /* G */\

  505.     "pand "MANGLE(bF8)", %%mm5  \n\t" /* R */\

  506.     "psrlq           $3, %%mm2  \n\t"\

  507.     "psrlq           $1, %%mm5  \n\t"\

  508. \

  509.     "movq         %%mm2, %%mm1  \n\t"\

  510.     "movq         %%mm4, %%mm3  \n\t"\

  511. \

  512.     "punpcklbw    %%mm7, %%mm3  \n\t"\

  513.     "punpcklbw    %%mm5, %%mm2  \n\t"\

  514.     "punpckhbw    %%mm7, %%mm4  \n\t"\

  515.     "punpckhbw    %%mm5, %%mm1  \n\t"\

  516. \

  517.     "psllq           $2, %%mm3  \n\t"\

  518.     "psllq           $2, %%mm4  \n\t"\

  519. \

  520.     "por          %%mm3, %%mm2  \n\t"\

  521.     "por          %%mm4, %%mm1  \n\t"\

  522. \

  523.     MOVNTQ(%%mm2,  (dst, index, 2))\

  524.     MOVNTQ(%%mm1, 8(dst, index, 2))\

  525. \

  526.     "add             $8, "#index"   \n\t"\

  527.     "cmp        "#dstw", "#index"   \n\t"\

  528.     " jb             1b             \n\t"

  529. #define WRITERGB15(dst, dstw, index)  REAL_WRITERGB15(dst, dstw, index)

  530. 

  531. static void RENAME(yuv2rgb555_X_ar)(SwsContext *c, const int16_t *lumFilter,

  532.                                     const int16_t **lumSrc, int lumFilterSize,

  533.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  534.                                     const int16_t **chrVSrc,

  535.                                     int chrFilterSize, const int16_t **alpSrc,

  536.                                     uint8_t *dest, int dstW, int dstY)

  537. {

  538.     x86_reg dummy=0;

  539.     x86_reg dstW_reg = dstW;

  540.     x86_reg uv_off = c->uv_off_byte;

  541. 

  542.     YSCALEYUV2PACKEDX_ACCURATE

  543.     YSCALEYUV2RGBX

  544.     "pxor %%mm7, %%mm7 \n\t"

  545.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  546. #ifdef DITHER1XBPP

  547.     "paddusb "BLUE_DITHER"(%0), %%mm2\n\t"

  548.     "paddusb "GREEN_DITHER"(%0), %%mm4\n\t"

  549.     "paddusb "RED_DITHER"(%0), %%mm5\n\t"

  550. #endif

  551.     WRITERGB15(%4, %5, %%REGa)

  552.     YSCALEYUV2PACKEDX_END

  553. }

  554. 

  555. static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,

  556.                                  const int16_t **lumSrc, int lumFilterSize,

  557.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  558.                                  const int16_t **chrVSrc,

  559.                                  int chrFilterSize, const int16_t **alpSrc,

  560.                                  uint8_t *dest, int dstW, int dstY)

  561. {

  562.     x86_reg dummy=0;

  563.     x86_reg dstW_reg = dstW;

  564.     x86_reg uv_off = c->uv_off_byte;

  565. 

  566.     YSCALEYUV2PACKEDX

  567.     YSCALEYUV2RGBX

  568.     "pxor %%mm7, %%mm7 \n\t"

  569.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  570. #ifdef DITHER1XBPP

  571.     "paddusb "BLUE_DITHER"(%0), %%mm2  \n\t"

  572.     "paddusb "GREEN_DITHER"(%0), %%mm4  \n\t"

  573.     "paddusb "RED_DITHER"(%0), %%mm5  \n\t"

  574. #endif

  575.     WRITERGB15(%4, %5, %%REGa)

  576.     YSCALEYUV2PACKEDX_END

  577. }

  578. 

  579. #define WRITEBGR24MMX(dst, dstw, index) \

  580.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\

  581.     "movq      %%mm2, %%mm1     \n\t" /* B */\

  582.     "movq      %%mm5, %%mm6     \n\t" /* R */\

  583.     "punpcklbw %%mm4, %%mm2     \n\t" /* GBGBGBGB 0 */\

  584.     "punpcklbw %%mm7, %%mm5     \n\t" /* 0R0R0R0R 0 */\

  585.     "punpckhbw %%mm4, %%mm1     \n\t" /* GBGBGBGB 2 */\

  586.     "punpckhbw %%mm7, %%mm6     \n\t" /* 0R0R0R0R 2 */\

  587.     "movq      %%mm2, %%mm0     \n\t" /* GBGBGBGB 0 */\

  588.     "movq      %%mm1, %%mm3     \n\t" /* GBGBGBGB 2 */\

  589.     "punpcklwd %%mm5, %%mm0     \n\t" /* 0RGB0RGB 0 */\

  590.     "punpckhwd %%mm5, %%mm2     \n\t" /* 0RGB0RGB 1 */\

  591.     "punpcklwd %%mm6, %%mm1     \n\t" /* 0RGB0RGB 2 */\

  592.     "punpckhwd %%mm6, %%mm3     \n\t" /* 0RGB0RGB 3 */\

  593. \

  594.     "movq      %%mm0, %%mm4     \n\t" /* 0RGB0RGB 0 */\

  595.     "movq      %%mm2, %%mm6     \n\t" /* 0RGB0RGB 1 */\

  596.     "movq      %%mm1, %%mm5     \n\t" /* 0RGB0RGB 2 */\

  597.     "movq      %%mm3, %%mm7     \n\t" /* 0RGB0RGB 3 */\

  598. \

  599.     "psllq       $40, %%mm0     \n\t" /* RGB00000 0 */\

  600.     "psllq       $40, %%mm2     \n\t" /* RGB00000 1 */\

  601.     "psllq       $40, %%mm1     \n\t" /* RGB00000 2 */\

  602.     "psllq       $40, %%mm3     \n\t" /* RGB00000 3 */\

  603. \

  604.     "punpckhdq %%mm4, %%mm0     \n\t" /* 0RGBRGB0 0 */\

  605.     "punpckhdq %%mm6, %%mm2     \n\t" /* 0RGBRGB0 1 */\

  606.     "punpckhdq %%mm5, %%mm1     \n\t" /* 0RGBRGB0 2 */\

  607.     "punpckhdq %%mm7, %%mm3     \n\t" /* 0RGBRGB0 3 */\

  608. \

  609.     "psrlq        $8, %%mm0     \n\t" /* 00RGBRGB 0 */\

  610.     "movq      %%mm2, %%mm6     \n\t" /* 0RGBRGB0 1 */\

  611.     "psllq       $40, %%mm2     \n\t" /* GB000000 1 */\

  612.     "por       %%mm2, %%mm0     \n\t" /* GBRGBRGB 0 */\

  613.     MOVNTQ(%%mm0, (dst))\

  614. \

  615.     "psrlq       $24, %%mm6     \n\t" /* 0000RGBR 1 */\

  616.     "movq      %%mm1, %%mm5     \n\t" /* 0RGBRGB0 2 */\

  617.     "psllq       $24, %%mm1     \n\t" /* BRGB0000 2 */\

  618.     "por       %%mm1, %%mm6     \n\t" /* BRGBRGBR 1 */\

  619.     MOVNTQ(%%mm6, 8(dst))\

  620. \

  621.     "psrlq       $40, %%mm5     \n\t" /* 000000RG 2 */\

  622.     "psllq        $8, %%mm3     \n\t" /* RGBRGB00 3 */\

  623.     "por       %%mm3, %%mm5     \n\t" /* RGBRGBRG 2 */\

  624.     MOVNTQ(%%mm5, 16(dst))\

  625. \

  626.     "add         $24, "#dst"    \n\t"\

  627. \

  628.     "add          $8, "#index"  \n\t"\

  629.     "cmp     "#dstw", "#index"  \n\t"\

  630.     " jb          1b            \n\t"

  631. 

  632. #define WRITEBGR24MMX2(dst, dstw, index) \

  633.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\

  634.     "movq "MANGLE(ff_M24A)", %%mm0 \n\t"\

  635.     "movq "MANGLE(ff_M24C)", %%mm7 \n\t"\

  636.     "pshufw $0x50, %%mm2, %%mm1 \n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\

  637.     "pshufw $0x50, %%mm4, %%mm3 \n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\

  638.     "pshufw $0x00, %%mm5, %%mm6 \n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\

  639. \

  640.     "pand   %%mm0, %%mm1        \n\t" /*    B2        B1       B0 */\

  641.     "pand   %%mm0, %%mm3        \n\t" /*    G2        G1       G0 */\

  642.     "pand   %%mm7, %%mm6        \n\t" /*       R1        R0       */\

  643. \

  644.     "psllq     $8, %%mm3        \n\t" /* G2        G1       G0    */\

  645.     "por    %%mm1, %%mm6        \n\t"\

  646.     "por    %%mm3, %%mm6        \n\t"\

  647.     MOVNTQ(%%mm6, (dst))\

  648. \

  649.     "psrlq     $8, %%mm4        \n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\

  650.     "pshufw $0xA5, %%mm2, %%mm1 \n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\

  651.     "pshufw $0x55, %%mm4, %%mm3 \n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\

  652.     "pshufw $0xA5, %%mm5, %%mm6 \n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\

  653. \

  654.     "pand "MANGLE(ff_M24B)", %%mm1 \n\t" /* B5       B4        B3    */\

  655.     "pand   %%mm7, %%mm3        \n\t" /*       G4        G3       */\

  656.     "pand   %%mm0, %%mm6        \n\t" /*    R4        R3       R2 */\

  657. \

  658.     "por    %%mm1, %%mm3        \n\t" /* B5    G4 B4     G3 B3    */\

  659.     "por    %%mm3, %%mm6        \n\t"\

  660.     MOVNTQ(%%mm6, 8(dst))\

  661. \

  662.     "pshufw $0xFF, %%mm2, %%mm1 \n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\

  663.     "pshufw $0xFA, %%mm4, %%mm3 \n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\

  664.     "pshufw $0xFA, %%mm5, %%mm6 \n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\

  665. \

  666.     "pand   %%mm7, %%mm1        \n\t" /*       B7        B6       */\

  667.     "pand   %%mm0, %%mm3        \n\t" /*    G7        G6       G5 */\

  668.     "pand "MANGLE(ff_M24B)", %%mm6 \n\t" /* R7       R6        R5    */\

  669. \

  670.     "por    %%mm1, %%mm3        \n\t"\

  671.     "por    %%mm3, %%mm6        \n\t"\

  672.     MOVNTQ(%%mm6, 16(dst))\

  673. \

  674.     "add      $24, "#dst"       \n\t"\

  675. \

  676.     "add       $8, "#index"     \n\t"\

  677.     "cmp  "#dstw", "#index"     \n\t"\

  678.     " jb       1b               \n\t"

  679. 

  680. #if COMPILE_TEMPLATE_MMX2

  681. #undef WRITEBGR24

  682. #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX2(dst, dstw, index)

  683. #else

  684. #undef WRITEBGR24

  685. #define WRITEBGR24(dst, dstw, index)  WRITEBGR24MMX(dst, dstw, index)

  686. #endif

  687. 

  688. static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,

  689.                                    const int16_t **lumSrc, int lumFilterSize,

  690.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  691.                                    const int16_t **chrVSrc,

  692.                                    int chrFilterSize, const int16_t **alpSrc,

  693.                                    uint8_t *dest, int dstW, int dstY)

  694. {

  695.     x86_reg dummy=0;

  696.     x86_reg dstW_reg = dstW;

  697.     x86_reg uv_off = c->uv_off_byte;

  698. 

  699.     YSCALEYUV2PACKEDX_ACCURATE

  700.     YSCALEYUV2RGBX

  701.     "pxor %%mm7, %%mm7 \n\t"

  702.     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize

  703.     "add %4, %%"REG_c"                        \n\t"

  704.     WRITEBGR24(%%REGc, %5, %%REGa)

  705.     :: "r" (&c->redDither),

  706.        "m" (dummy), "m" (dummy), "m" (dummy),

  707.        "r" (dest), "m" (dstW_reg), "m"(uv_off)

  708.     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S

  709.     );

  710. }

  711. 

  712. static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,

  713.                                 const int16_t **lumSrc, int lumFilterSize,

  714.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  715.                                 const int16_t **chrVSrc,

  716.                                 int chrFilterSize, const int16_t **alpSrc,

  717.                                 uint8_t *dest, int dstW, int dstY)

  718. {

  719.     x86_reg dummy=0;

  720.     x86_reg dstW_reg = dstW;

  721.     x86_reg uv_off = c->uv_off_byte;

  722. 

  723.     YSCALEYUV2PACKEDX

  724.     YSCALEYUV2RGBX

  725.     "pxor                    %%mm7, %%mm7       \n\t"

  726.     "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize

  727.     "add                        %4, %%"REG_c"   \n\t"

  728.     WRITEBGR24(%%REGc, %5, %%REGa)

  729.     :: "r" (&c->redDither),

  730.        "m" (dummy), "m" (dummy), "m" (dummy),

  731.        "r" (dest),  "m" (dstW_reg), "m"(uv_off)

  732.     : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S

  733.     );

  734. }

  735. 

  736. #define REAL_WRITEYUY2(dst, dstw, index) \

  737.     "packuswb  %%mm3, %%mm3     \n\t"\

  738.     "packuswb  %%mm4, %%mm4     \n\t"\

  739.     "packuswb  %%mm7, %%mm1     \n\t"\

  740.     "punpcklbw %%mm4, %%mm3     \n\t"\

  741.     "movq      %%mm1, %%mm7     \n\t"\

  742.     "punpcklbw %%mm3, %%mm1     \n\t"\

  743.     "punpckhbw %%mm3, %%mm7     \n\t"\

  744. \

  745.     MOVNTQ(%%mm1, (dst, index, 2))\

  746.     MOVNTQ(%%mm7, 8(dst, index, 2))\

  747. \

  748.     "add          $8, "#index"  \n\t"\

  749.     "cmp     "#dstw", "#index"  \n\t"\

  750.     " jb          1b            \n\t"

  751. #define WRITEYUY2(dst, dstw, index)  REAL_WRITEYUY2(dst, dstw, index)

  752. 

  753. static void RENAME(yuv2yuyv422_X_ar)(SwsContext *c, const int16_t *lumFilter,

  754.                                      const int16_t **lumSrc, int lumFilterSize,

  755.                                      const int16_t *chrFilter, const int16_t **chrUSrc,

  756.                                      const int16_t **chrVSrc,

  757.                                      int chrFilterSize, const int16_t **alpSrc,

  758.                                      uint8_t *dest, int dstW, int dstY)

  759. {

  760.     x86_reg dummy=0;

  761.     x86_reg dstW_reg = dstW;

  762.     x86_reg uv_off = c->uv_off_byte;

  763. 

  764.     YSCALEYUV2PACKEDX_ACCURATE

  765.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  766.     "psraw $3, %%mm3    \n\t"

  767.     "psraw $3, %%mm4    \n\t"

  768.     "psraw $3, %%mm1    \n\t"

  769.     "psraw $3, %%mm7    \n\t"

  770.     WRITEYUY2(%4, %5, %%REGa)

  771.     YSCALEYUV2PACKEDX_END

  772. }

  773. 

  774. static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,

  775.                                   const int16_t **lumSrc, int lumFilterSize,

  776.                                   const int16_t *chrFilter, const int16_t **chrUSrc,

  777.                                   const int16_t **chrVSrc,

  778.                                   int chrFilterSize, const int16_t **alpSrc,

  779.                                   uint8_t *dest, int dstW, int dstY)

  780. {

  781.     x86_reg dummy=0;

  782.     x86_reg dstW_reg = dstW;

  783.     x86_reg uv_off = c->uv_off_byte;

  784. 

  785.     YSCALEYUV2PACKEDX

  786.     /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  787.     "psraw $3, %%mm3    \n\t"

  788.     "psraw $3, %%mm4    \n\t"

  789.     "psraw $3, %%mm1    \n\t"

  790.     "psraw $3, %%mm7    \n\t"

  791.     WRITEYUY2(%4, %5, %%REGa)

  792.     YSCALEYUV2PACKEDX_END

  793. }

  794. 

  795. #define REAL_YSCALEYUV2RGB_UV(index, c) \

  796.     "xor            "#index", "#index"  \n\t"\

  797.     ".p2align              4            \n\t"\

  798.     "1:                                 \n\t"\

  799.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  800.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  801.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  802.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  803.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  804.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  805.     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\

  806.     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  807.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\

  808.     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

  809.     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

  810.     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  811.     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  812.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\

  813.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  814.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  815.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  816.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  817.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  818.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  819.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  820.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  821. 

  822. #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \

  823.     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\

  824.     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\

  825.     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\

  826.     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\

  827.     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\

  828.     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\

  829.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  830.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  831.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  832.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  833.     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  834.     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  835. 

  836. #define REAL_YSCALEYUV2RGB_COEFF(c) \

  837.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  838.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  839.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  840.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  841.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  842.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  843.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  844.     "paddw             %%mm3, %%mm4     \n\t"\

  845.     "movq              %%mm2, %%mm0     \n\t"\

  846.     "movq              %%mm5, %%mm6     \n\t"\

  847.     "movq              %%mm4, %%mm3     \n\t"\

  848.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  849.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  850.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  851.     "paddw             %%mm1, %%mm2     \n\t"\

  852.     "paddw             %%mm1, %%mm5     \n\t"\

  853.     "paddw             %%mm1, %%mm4     \n\t"\

  854.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  855.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  856.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  857.     "paddw             %%mm7, %%mm0     \n\t"\

  858.     "paddw             %%mm7, %%mm6     \n\t"\

  859.     "paddw             %%mm7, %%mm3     \n\t"\

  860.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  861.     "packuswb          %%mm0, %%mm2     \n\t"\

  862.     "packuswb          %%mm6, %%mm5     \n\t"\

  863.     "packuswb          %%mm3, %%mm4     \n\t"\

  864. 

  865. #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)

  866. 

  867. #define YSCALEYUV2RGB(index, c) \

  868.     REAL_YSCALEYUV2RGB_UV(index, c) \

  869.     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \

  870.     REAL_YSCALEYUV2RGB_COEFF(c)

  871. 

  872. /**

  873.  * vertical bilinear scale YV12 to RGB

  874.  */

  875. static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],

  876.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  877.                                 const int16_t *abuf[2], uint8_t *dest,

  878.                                 int dstW, int yalpha, int uvalpha, int y)

  879. {

  880.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  881.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  882. 

  883.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  884.         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];

  885. #if ARCH_X86_64

  886.         __asm__ volatile(

  887.             YSCALEYUV2RGB(%%r8, %5)

  888.             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)

  889.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  890.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  891.             "packuswb            %%mm7, %%mm1       \n\t"

  892.             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  893.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),

  894.                "a" (&c->redDither),

  895.                "r" (abuf0), "r" (abuf1)

  896.             : "%r8"

  897.         );

  898. #else

  899.         *(const uint16_t **)(&c->u_temp)=abuf0;

  900.         *(const uint16_t **)(&c->v_temp)=abuf1;

  901.         __asm__ volatile(

  902.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  903.             "mov        %4, %%"REG_b"               \n\t"

  904.             "push %%"REG_BP"                        \n\t"

  905.             YSCALEYUV2RGB(%%REGBP, %5)

  906.             "push                   %0              \n\t"

  907.             "push                   %1              \n\t"

  908.             "mov          "U_TEMP"(%5), %0          \n\t"

  909.             "mov          "V_TEMP"(%5), %1          \n\t"

  910.             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)

  911.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  912.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  913.             "packuswb            %%mm7, %%mm1       \n\t"

  914.             "pop                    %1              \n\t"

  915.             "pop                    %0              \n\t"

  916.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  917.             "pop %%"REG_BP"                         \n\t"

  918.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  919.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  920.                "a" (&c->redDither)

  921.         );

  922. #endif

  923.     } else {

  924.         __asm__ volatile(

  925.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  926.             "mov        %4, %%"REG_b"               \n\t"

  927.             "push %%"REG_BP"                        \n\t"

  928.             YSCALEYUV2RGB(%%REGBP, %5)

  929.             "pcmpeqd %%mm7, %%mm7                   \n\t"

  930.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  931.             "pop %%"REG_BP"                         \n\t"

  932.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  933.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  934.                "a" (&c->redDither)

  935.         );

  936.     }

  937. }

  938. 

  939. static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],

  940.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  941.                                 const int16_t *abuf[2], uint8_t *dest,

  942.                                 int dstW, int yalpha, int uvalpha, int y)

  943. {

  944.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  945.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  946. 

  947.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  948.     __asm__ volatile(

  949.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  950.         "mov        %4, %%"REG_b"               \n\t"

  951.         "push %%"REG_BP"                        \n\t"

  952.         YSCALEYUV2RGB(%%REGBP, %5)

  953.         "pxor    %%mm7, %%mm7                   \n\t"

  954.         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  955.         "pop %%"REG_BP"                         \n\t"

  956.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  957.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  958.            "a" (&c->redDither)

  959.     );

  960. }

  961. 

  962. static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],

  963.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  964.                                  const int16_t *abuf[2], uint8_t *dest,

  965.                                  int dstW, int yalpha, int uvalpha, int y)

  966. {

  967.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  968.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  969. 

  970.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  971.     __asm__ volatile(

  972.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  973.         "mov        %4, %%"REG_b"               \n\t"

  974.         "push %%"REG_BP"                        \n\t"

  975.         YSCALEYUV2RGB(%%REGBP, %5)

  976.         "pxor    %%mm7, %%mm7                   \n\t"

  977.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  978. #ifdef DITHER1XBPP

  979.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  980.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  981.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  982. #endif

  983.         WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  984.         "pop %%"REG_BP"                         \n\t"

  985.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  986.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  987.            "a" (&c->redDither)

  988.     );

  989. }

  990. 

  991. static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],

  992.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  993.                                  const int16_t *abuf[2], uint8_t *dest,

  994.                                  int dstW, int yalpha, int uvalpha, int y)

  995. {

  996.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  997.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  998. 

  999.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1000.     __asm__ volatile(

  1001.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1002.         "mov        %4, %%"REG_b"               \n\t"

  1003.         "push %%"REG_BP"                        \n\t"

  1004.         YSCALEYUV2RGB(%%REGBP, %5)

  1005.         "pxor    %%mm7, %%mm7                   \n\t"

  1006.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1007. #ifdef DITHER1XBPP

  1008.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1009.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1010.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1011. #endif

  1012.         WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1013.         "pop %%"REG_BP"                         \n\t"

  1014.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1015.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1016.            "a" (&c->redDither)

  1017.     );

  1018. }

  1019. 

  1020. #define REAL_YSCALEYUV2PACKED(index, c) \

  1021.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\

  1022.     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\

  1023.     "psraw                $3, %%mm0                           \n\t"\

  1024.     "psraw                $3, %%mm1                           \n\t"\

  1025.     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1026.     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1027.     "xor            "#index", "#index"                        \n\t"\

  1028.     ".p2align              4            \n\t"\

  1029.     "1:                                 \n\t"\

  1030.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1031.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1032.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1033.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1034.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1035.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1036.     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\

  1037.     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  1038.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\

  1039.     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

  1040.     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

  1041.     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1042.     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1043.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\

  1044.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  1045.     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\

  1046.     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\

  1047.     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\

  1048.     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\

  1049.     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\

  1050.     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\

  1051.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1052.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1053.     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1054.     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1055.     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1056.     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1057. 

  1058. #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)

  1059. 

  1060. static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],

  1061.                                   const int16_t *ubuf[2], const int16_t *vbuf[2],

  1062.                                   const int16_t *abuf[2], uint8_t *dest,

  1063.                                   int dstW, int yalpha, int uvalpha, int y)

  1064. {

  1065.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1066.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1067. 

  1068.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1069.     __asm__ volatile(

  1070.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1071.         "mov %4, %%"REG_b"                        \n\t"

  1072.         "push %%"REG_BP"                        \n\t"

  1073.         YSCALEYUV2PACKED(%%REGBP, %5)

  1074.         WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1075.         "pop %%"REG_BP"                         \n\t"

  1076.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1077.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1078.            "a" (&c->redDither)

  1079.     );

  1080. }

  1081. 

  1082. #define REAL_YSCALEYUV2RGB1(index, c) \

  1083.     "xor            "#index", "#index"  \n\t"\

  1084.     ".p2align              4            \n\t"\

  1085.     "1:                                 \n\t"\

  1086.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1087.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1088.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1089.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1090.     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1091.     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1092.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1093.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1094.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1095.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1096.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1097.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1098.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1099.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1100.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1101.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1102.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1103.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1104.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1105.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1106.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1107.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1108.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1109.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1110.     "paddw             %%mm3, %%mm4     \n\t"\

  1111.     "movq              %%mm2, %%mm0     \n\t"\

  1112.     "movq              %%mm5, %%mm6     \n\t"\

  1113.     "movq              %%mm4, %%mm3     \n\t"\

  1114.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1115.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1116.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1117.     "paddw             %%mm1, %%mm2     \n\t"\

  1118.     "paddw             %%mm1, %%mm5     \n\t"\

  1119.     "paddw             %%mm1, %%mm4     \n\t"\

  1120.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1121.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1122.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1123.     "paddw             %%mm7, %%mm0     \n\t"\

  1124.     "paddw             %%mm7, %%mm6     \n\t"\

  1125.     "paddw             %%mm7, %%mm3     \n\t"\

  1126.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1127.     "packuswb          %%mm0, %%mm2     \n\t"\

  1128.     "packuswb          %%mm6, %%mm5     \n\t"\

  1129.     "packuswb          %%mm3, %%mm4     \n\t"\

  1130. 

  1131. #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)

  1132. 

  1133. // do vertical chrominance interpolation

  1134. #define REAL_YSCALEYUV2RGB1b(index, c) \

  1135.     "xor            "#index", "#index"  \n\t"\

  1136.     ".p2align              4            \n\t"\

  1137.     "1:                                 \n\t"\

  1138.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1139.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1140.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1141.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1142.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1143.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1144.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1145.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1146.     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\

  1147.     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\

  1148.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1149.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1150.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1151.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1152.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1153.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1154.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1155.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1156.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1157.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1158.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1159.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1160.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1161.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1162.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1163.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1164.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1165.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1166.     "paddw             %%mm3, %%mm4     \n\t"\

  1167.     "movq              %%mm2, %%mm0     \n\t"\

  1168.     "movq              %%mm5, %%mm6     \n\t"\

  1169.     "movq              %%mm4, %%mm3     \n\t"\

  1170.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1171.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1172.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1173.     "paddw             %%mm1, %%mm2     \n\t"\

  1174.     "paddw             %%mm1, %%mm5     \n\t"\

  1175.     "paddw             %%mm1, %%mm4     \n\t"\

  1176.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1177.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1178.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1179.     "paddw             %%mm7, %%mm0     \n\t"\

  1180.     "paddw             %%mm7, %%mm6     \n\t"\

  1181.     "paddw             %%mm7, %%mm3     \n\t"\

  1182.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1183.     "packuswb          %%mm0, %%mm2     \n\t"\

  1184.     "packuswb          %%mm6, %%mm5     \n\t"\

  1185.     "packuswb          %%mm3, %%mm4     \n\t"\

  1186. 

  1187. #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)

  1188. 

  1189. #define REAL_YSCALEYUV2RGB1_ALPHA(index) \

  1190.     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\

  1191.     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\

  1192.     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\

  1193.     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\

  1194.     "packuswb          %%mm1, %%mm7     \n\t"

  1195. #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)

  1196. 

  1197. /**

  1198.  * YV12 to RGB without scaling or interpolating

  1199.  */

  1200. static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,

  1201.                                 const int16_t *ubuf[2], const int16_t *bguf[2],

  1202.                                 const int16_t *abuf0, uint8_t *dest,

  1203.                                 int dstW, int uvalpha, int y)

  1204. {

  1205.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1206.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1207. 

  1208.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1209.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1210.             __asm__ volatile(

  1211.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1212.                 "mov        %4, %%"REG_b"               \n\t"

  1213.                 "push %%"REG_BP"                        \n\t"

  1214.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1215.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1216.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1217.                 "pop %%"REG_BP"                         \n\t"

  1218.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1219.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1220.                    "a" (&c->redDither)

  1221.             );

  1222.         } else {

  1223.             __asm__ volatile(

  1224.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1225.                 "mov        %4, %%"REG_b"               \n\t"

  1226.                 "push %%"REG_BP"                        \n\t"

  1227.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1228.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1229.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1230.                 "pop %%"REG_BP"                         \n\t"

  1231.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1232.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1233.                    "a" (&c->redDither)

  1234.             );

  1235.         }

  1236.     } else {

  1237.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1238.             __asm__ volatile(

  1239.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1240.                 "mov        %4, %%"REG_b"               \n\t"

  1241.                 "push %%"REG_BP"                        \n\t"

  1242.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1243.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1244.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1245.                 "pop %%"REG_BP"                         \n\t"

  1246.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1247.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1248.                    "a" (&c->redDither)

  1249.             );

  1250.         } else {

  1251.             __asm__ volatile(

  1252.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1253.                 "mov        %4, %%"REG_b"               \n\t"

  1254.                 "push %%"REG_BP"                        \n\t"

  1255.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1256.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1257.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1258.                 "pop %%"REG_BP"                         \n\t"

  1259.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1260.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1261.                    "a" (&c->redDither)

  1262.             );

  1263.         }

  1264.     }

  1265. }

  1266. 

  1267. static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,

  1268.                                 const int16_t *ubuf[2], const int16_t *bguf[2],

  1269.                                 const int16_t *abuf0, uint8_t *dest,

  1270.                                 int dstW, int uvalpha, int y)

  1271. {

  1272.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1273.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1274. 

  1275.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  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.             YSCALEYUV2RGB1(%%REGBP, %5)

  1281.             "pxor    %%mm7, %%mm7                   \n\t"

  1282.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1283.             "pop %%"REG_BP"                         \n\t"

  1284.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1285.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1286.                "a" (&c->redDither)

  1287.         );

  1288.     } else {

  1289.         __asm__ volatile(

  1290.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1291.             "mov        %4, %%"REG_b"               \n\t"

  1292.             "push %%"REG_BP"                        \n\t"

  1293.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1294.             "pxor    %%mm7, %%mm7                   \n\t"

  1295.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1296.             "pop %%"REG_BP"                         \n\t"

  1297.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1298.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1299.                "a" (&c->redDither)

  1300.         );

  1301.     }

  1302. }

  1303. 

  1304. static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,

  1305.                                  const int16_t *ubuf[2], const int16_t *bguf[2],

  1306.                                  const int16_t *abuf0, uint8_t *dest,

  1307.                                  int dstW, int uvalpha, int y)

  1308. {

  1309.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1310.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1311. 

  1312.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1313.         __asm__ volatile(

  1314.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1315.             "mov        %4, %%"REG_b"               \n\t"

  1316.             "push %%"REG_BP"                        \n\t"

  1317.             YSCALEYUV2RGB1(%%REGBP, %5)

  1318.             "pxor    %%mm7, %%mm7                   \n\t"

  1319.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1320. #ifdef DITHER1XBPP

  1321.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1322.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1323.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1324. #endif

  1325.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1326.             "pop %%"REG_BP"                         \n\t"

  1327.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1328.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1329.                "a" (&c->redDither)

  1330.         );

  1331.     } else {

  1332.         __asm__ volatile(

  1333.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1334.             "mov        %4, %%"REG_b"               \n\t"

  1335.             "push %%"REG_BP"                        \n\t"

  1336.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1337.             "pxor    %%mm7, %%mm7                   \n\t"

  1338.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1339. #ifdef DITHER1XBPP

  1340.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1341.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1342.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1343. #endif

  1344.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1345.             "pop %%"REG_BP"                         \n\t"

  1346.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1347.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1348.                "a" (&c->redDither)

  1349.         );

  1350.     }

  1351. }

  1352. 

  1353. static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,

  1354.                                  const int16_t *ubuf[2], const int16_t *bguf[2],

  1355.                                  const int16_t *abuf0, uint8_t *dest,

  1356.                                  int dstW, int uvalpha, int y)

  1357. {

  1358.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1359.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1360. 

  1361.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1362.         __asm__ volatile(

  1363.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1364.             "mov        %4, %%"REG_b"               \n\t"

  1365.             "push %%"REG_BP"                        \n\t"

  1366.             YSCALEYUV2RGB1(%%REGBP, %5)

  1367.             "pxor    %%mm7, %%mm7                   \n\t"

  1368.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1369. #ifdef DITHER1XBPP

  1370.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1371.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1372.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1373. #endif

  1374.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1375.             "pop %%"REG_BP"                         \n\t"

  1376.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1377.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1378.                "a" (&c->redDither)

  1379.         );

  1380.     } else {

  1381.         __asm__ volatile(

  1382.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1383.             "mov        %4, %%"REG_b"               \n\t"

  1384.             "push %%"REG_BP"                        \n\t"

  1385.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1386.             "pxor    %%mm7, %%mm7                   \n\t"

  1387.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1388. #ifdef DITHER1XBPP

  1389.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1390.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1391.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1392. #endif

  1393.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1394.             "pop %%"REG_BP"                         \n\t"

  1395.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1396.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1397.                "a" (&c->redDither)

  1398.         );

  1399.     }

  1400. }

  1401. 

  1402. #define REAL_YSCALEYUV2PACKED1(index, c) \

  1403.     "xor            "#index", "#index"  \n\t"\

  1404.     ".p2align              4            \n\t"\

  1405.     "1:                                 \n\t"\

  1406.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1407.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1408.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1409.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1410.     "psraw                $7, %%mm3     \n\t" \

  1411.     "psraw                $7, %%mm4     \n\t" \

  1412.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1413.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1414.     "psraw                $7, %%mm1     \n\t" \

  1415.     "psraw                $7, %%mm7     \n\t" \

  1416. 

  1417. #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)

  1418. 

  1419. #define REAL_YSCALEYUV2PACKED1b(index, c) \

  1420.     "xor "#index", "#index"             \n\t"\

  1421.     ".p2align              4            \n\t"\

  1422.     "1:                                 \n\t"\

  1423.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1424.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1425.     "add   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1426.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1427.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1428.     "sub   "UV_OFF_PX"("#c"), "#index"  \n\t" \

  1429.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1430.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1431.     "psrlw                $8, %%mm3     \n\t" \

  1432.     "psrlw                $8, %%mm4     \n\t" \

  1433.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1434.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1435.     "psraw                $7, %%mm1     \n\t" \

  1436.     "psraw                $7, %%mm7     \n\t"

  1437. #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)

  1438. 

  1439. static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,

  1440.                                   const int16_t *ubuf[2], const int16_t *bguf[2],

  1441.                                   const int16_t *abuf0, uint8_t *dest,

  1442.                                   int dstW, int uvalpha, int y)

  1443. {

  1444.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1445.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1446. 

  1447.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1448.         __asm__ volatile(

  1449.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1450.             "mov        %4, %%"REG_b"               \n\t"

  1451.             "push %%"REG_BP"                        \n\t"

  1452.             YSCALEYUV2PACKED1(%%REGBP, %5)

  1453.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1454.             "pop %%"REG_BP"                         \n\t"

  1455.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1456.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1457.                "a" (&c->redDither)

  1458.         );

  1459.     } else {

  1460.         __asm__ volatile(

  1461.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1462.             "mov        %4, %%"REG_b"               \n\t"

  1463.             "push %%"REG_BP"                        \n\t"

  1464.             YSCALEYUV2PACKED1b(%%REGBP, %5)

  1465.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1466.             "pop %%"REG_BP"                         \n\t"

  1467.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1468.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1469.                "a" (&c->redDither)

  1470.         );

  1471.     }

  1472. }

  1473. 

  1474. #if !COMPILE_TEMPLATE_MMX2

  1475. //FIXME yuy2* can read up to 7 samples too much

  1476. 

  1477. static void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src,

  1478.                             int width, uint32_t *unused)

  1479. {

  1480.     __asm__ volatile(

  1481.         "movq "MANGLE(bm01010101)", %%mm2           \n\t"

  1482.         "mov                    %0, %%"REG_a"       \n\t"

  1483.         "1:                                         \n\t"

  1484.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1485.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1486.         "pand                %%mm2, %%mm0           \n\t"

  1487.         "pand                %%mm2, %%mm1           \n\t"

  1488.         "packuswb            %%mm1, %%mm0           \n\t"

  1489.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1490.         "add                    $8, %%"REG_a"       \n\t"

  1491.         " js                    1b                  \n\t"

  1492.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1493.         : "%"REG_a

  1494.     );

  1495. }

  1496. 

  1497. static void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV,

  1498.                              const uint8_t *src1, const uint8_t *src2,

  1499.                              int width, uint32_t *unused)

  1500. {

  1501.     __asm__ volatile(

  1502.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1503.         "mov                    %0, %%"REG_a"       \n\t"

  1504.         "1:                                         \n\t"

  1505.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1506.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1507.         "psrlw                  $8, %%mm0           \n\t"

  1508.         "psrlw                  $8, %%mm1           \n\t"

  1509.         "packuswb            %%mm1, %%mm0           \n\t"

  1510.         "movq                %%mm0, %%mm1           \n\t"

  1511.         "psrlw                  $8, %%mm0           \n\t"

  1512.         "pand                %%mm4, %%mm1           \n\t"

  1513.         "packuswb            %%mm0, %%mm0           \n\t"

  1514.         "packuswb            %%mm1, %%mm1           \n\t"

  1515.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1516.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1517.         "add                    $4, %%"REG_a"       \n\t"

  1518.         " js                    1b                  \n\t"

  1519.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1520.         : "%"REG_a

  1521.     );

  1522.     assert(src1 == src2);

  1523. }

  1524. 

  1525. /* This is almost identical to the previous, end exists only because

  1526.  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */

  1527. static void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src,

  1528.                             int width, uint32_t *unused)

  1529. {

  1530.     __asm__ volatile(

  1531.         "mov                  %0, %%"REG_a"         \n\t"

  1532.         "1:                                         \n\t"

  1533.         "movq  (%1, %%"REG_a",2), %%mm0             \n\t"

  1534.         "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"

  1535.         "psrlw                $8, %%mm0             \n\t"

  1536.         "psrlw                $8, %%mm1             \n\t"

  1537.         "packuswb          %%mm1, %%mm0             \n\t"

  1538.         "movq              %%mm0, (%2, %%"REG_a")   \n\t"

  1539.         "add                  $8, %%"REG_a"         \n\t"

  1540.         " js                  1b                    \n\t"

  1541.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1542.         : "%"REG_a

  1543.     );

  1544. }

  1545. 

  1546. static void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV,

  1547.                              const uint8_t *src1, const uint8_t *src2,

  1548.                              int width, uint32_t *unused)

  1549. {

  1550.     __asm__ volatile(

  1551.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1552.         "mov                    %0, %%"REG_a"       \n\t"

  1553.         "1:                                         \n\t"

  1554.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1555.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1556.         "pand                %%mm4, %%mm0           \n\t"

  1557.         "pand                %%mm4, %%mm1           \n\t"

  1558.         "packuswb            %%mm1, %%mm0           \n\t"

  1559.         "movq                %%mm0, %%mm1           \n\t"

  1560.         "psrlw                  $8, %%mm0           \n\t"

  1561.         "pand                %%mm4, %%mm1           \n\t"

  1562.         "packuswb            %%mm0, %%mm0           \n\t"

  1563.         "packuswb            %%mm1, %%mm1           \n\t"

  1564.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1565.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1566.         "add                    $4, %%"REG_a"       \n\t"

  1567.         " js                    1b                  \n\t"

  1568.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1569.         : "%"REG_a

  1570.     );

  1571.     assert(src1 == src2);

  1572. }

  1573. 

  1574. static av_always_inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,

  1575.                                               const uint8_t *src, int width)

  1576. {

  1577.     __asm__ volatile(

  1578.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1579.         "mov                    %0, %%"REG_a"       \n\t"

  1580.         "1:                                         \n\t"

  1581.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1582.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1583.         "movq                %%mm0, %%mm2           \n\t"

  1584.         "movq                %%mm1, %%mm3           \n\t"

  1585.         "pand                %%mm4, %%mm0           \n\t"

  1586.         "pand                %%mm4, %%mm1           \n\t"

  1587.         "psrlw                  $8, %%mm2           \n\t"

  1588.         "psrlw                  $8, %%mm3           \n\t"

  1589.         "packuswb            %%mm1, %%mm0           \n\t"

  1590.         "packuswb            %%mm3, %%mm2           \n\t"

  1591.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1592.         "movq                %%mm2, (%3, %%"REG_a") \n\t"

  1593.         "add                    $8, %%"REG_a"       \n\t"

  1594.         " js                    1b                  \n\t"

  1595.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)

  1596.         : "%"REG_a

  1597.     );

  1598. }

  1599. 

  1600. static void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,

  1601.                              const uint8_t *src1, const uint8_t *src2,

  1602.                              int width, uint32_t *unused)

  1603. {

  1604.     RENAME(nvXXtoUV)(dstU, dstV, src1, width);

  1605. }

  1606. 

  1607. static void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,

  1608.                              const uint8_t *src1, const uint8_t *src2,

  1609.                              int width, uint32_t *unused)

  1610. {

  1611.     RENAME(nvXXtoUV)(dstV, dstU, src1, width);

  1612. }

  1613. #endif /* !COMPILE_TEMPLATE_MMX2 */

  1614. 

  1615. static av_always_inline void RENAME(bgr24ToY_mmx)(uint8_t *dst, const uint8_t *src,

  1616.                                                   int width, enum PixelFormat srcFormat)

  1617. {

  1618. 

  1619.     if(srcFormat == PIX_FMT_BGR24) {

  1620.         __asm__ volatile(

  1621.             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"

  1622.             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"

  1623.             :

  1624.         );

  1625.     } else {

  1626.         __asm__ volatile(

  1627.             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"

  1628.             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"

  1629.             :

  1630.         );

  1631.     }

  1632. 

  1633.     __asm__ volatile(

  1634.         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"

  1635.         "mov                        %2, %%"REG_a"   \n\t"

  1636.         "pxor                    %%mm7, %%mm7       \n\t"

  1637.         "1:                                         \n\t"

  1638.         PREFETCH"               64(%0)              \n\t"

  1639.         "movd                     (%0), %%mm0       \n\t"

  1640.         "movd                    2(%0), %%mm1       \n\t"

  1641.         "movd                    6(%0), %%mm2       \n\t"

  1642.         "movd                    8(%0), %%mm3       \n\t"

  1643.         "add                       $12, %0          \n\t"

  1644.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1645.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1646.         "punpcklbw               %%mm7, %%mm2       \n\t"

  1647.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1648.         "pmaddwd                 %%mm5, %%mm0       \n\t"

  1649.         "pmaddwd                 %%mm6, %%mm1       \n\t"

  1650.         "pmaddwd                 %%mm5, %%mm2       \n\t"

  1651.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1652.         "paddd                   %%mm1, %%mm0       \n\t"

  1653.         "paddd                   %%mm3, %%mm2       \n\t"

  1654.         "paddd                   %%mm4, %%mm0       \n\t"

  1655.         "paddd                   %%mm4, %%mm2       \n\t"

  1656.         "psrad                     $15, %%mm0       \n\t"

  1657.         "psrad                     $15, %%mm2       \n\t"

  1658.         "packssdw                %%mm2, %%mm0       \n\t"

  1659.         "packuswb                %%mm0, %%mm0       \n\t"

  1660.         "movd                %%mm0, (%1, %%"REG_a") \n\t"

  1661.         "add                        $4, %%"REG_a"   \n\t"

  1662.         " js                        1b              \n\t"

  1663.     : "+r" (src)

  1664.     : "r" (dst+width), "g" ((x86_reg)-width)

  1665.     : "%"REG_a

  1666.     );

  1667. }

  1668. 

  1669. static void RENAME(bgr24ToY)(uint8_t *dst, const uint8_t *src,

  1670.                              int width, uint32_t *unused)

  1671. {

  1672.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);

  1673. }

  1674. 

  1675. static void RENAME(rgb24ToY)(uint8_t *dst, const uint8_t *src,

  1676.                              int width, uint32_t *unused)

  1677. {

  1678.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);

  1679. }

  1680. 

  1681. static av_always_inline void RENAME(bgr24ToUV_mmx)(uint8_t *dstU, uint8_t *dstV,

  1682.                                                    const uint8_t *src, int width,

  1683.                                                    enum PixelFormat srcFormat)

  1684. {

  1685.     __asm__ volatile(

  1686.         "movq                    24(%4), %%mm6       \n\t"

  1687.         "mov                        %3, %%"REG_a"   \n\t"

  1688.         "pxor                    %%mm7, %%mm7       \n\t"

  1689.         "1:                                         \n\t"

  1690.         PREFETCH"               64(%0)              \n\t"

  1691.         "movd                     (%0), %%mm0       \n\t"

  1692.         "movd                    2(%0), %%mm1       \n\t"

  1693.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1694.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1695.         "movq                    %%mm0, %%mm2       \n\t"

  1696.         "movq                    %%mm1, %%mm3       \n\t"

  1697.         "pmaddwd                  (%4), %%mm0       \n\t"

  1698.         "pmaddwd                 8(%4), %%mm1       \n\t"

  1699.         "pmaddwd                16(%4), %%mm2       \n\t"

  1700.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1701.         "paddd                   %%mm1, %%mm0       \n\t"

  1702.         "paddd                   %%mm3, %%mm2       \n\t"

  1703. 

  1704.         "movd                    6(%0), %%mm1       \n\t"

  1705.         "movd                    8(%0), %%mm3       \n\t"

  1706.         "add                       $12, %0          \n\t"

  1707.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1708.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1709.         "movq                    %%mm1, %%mm4       \n\t"

  1710.         "movq                    %%mm3, %%mm5       \n\t"

  1711.         "pmaddwd                  (%4), %%mm1       \n\t"

  1712.         "pmaddwd                 8(%4), %%mm3       \n\t"

  1713.         "pmaddwd                16(%4), %%mm4       \n\t"

  1714.         "pmaddwd                 %%mm6, %%mm5       \n\t"

  1715.         "paddd                   %%mm3, %%mm1       \n\t"

  1716.         "paddd                   %%mm5, %%mm4       \n\t"

  1717. 

  1718.         "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3       \n\t"

  1719.         "paddd                   %%mm3, %%mm0       \n\t"

  1720.         "paddd                   %%mm3, %%mm2       \n\t"

  1721.         "paddd                   %%mm3, %%mm1       \n\t"

  1722.         "paddd                   %%mm3, %%mm4       \n\t"

  1723.         "psrad                     $15, %%mm0       \n\t"

  1724.         "psrad                     $15, %%mm2       \n\t"

  1725.         "psrad                     $15, %%mm1       \n\t"

  1726.         "psrad                     $15, %%mm4       \n\t"

  1727.         "packssdw                %%mm1, %%mm0       \n\t"

  1728.         "packssdw                %%mm4, %%mm2       \n\t"

  1729.         "packuswb                %%mm0, %%mm0       \n\t"

  1730.         "packuswb                %%mm2, %%mm2       \n\t"

  1731.         "movd                %%mm0, (%1, %%"REG_a") \n\t"

  1732.         "movd                %%mm2, (%2, %%"REG_a") \n\t"

  1733.         "add                        $4, %%"REG_a"   \n\t"

  1734.         " js                        1b              \n\t"

  1735.     : "+r" (src)

  1736.     : "r" (dstU+width), "r" (dstV+width), "g" ((x86_reg)-width), "r"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24])

  1737.     : "%"REG_a

  1738.     );

  1739. }

  1740. 

  1741. static void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV,

  1742.                               const uint8_t *src1, const uint8_t *src2,

  1743.                               int width, uint32_t *unused)

  1744. {

  1745.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);

  1746.     assert(src1 == src2);

  1747. }

  1748. 

  1749. static void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV,

  1750.                               const uint8_t *src1, const uint8_t *src2,

  1751.                               int width, uint32_t *unused)

  1752. {

  1753.     assert(src1==src2);

  1754.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);

  1755. }

  1756. 

  1757. #if COMPILE_TEMPLATE_MMX2

  1758. static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,

  1759.                                  int dstWidth, const uint8_t *src,

  1760.                                  int srcW, int xInc)

  1761. {

  1762.     int16_t *filterPos = c->hLumFilterPos;

  1763.     int16_t *filter    = c->hLumFilter;

  1764.     void    *mmx2FilterCode= c->lumMmx2FilterCode;

  1765.     int i;

  1766. #if defined(PIC)

  1767.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  1768. #endif

  1769. 

  1770.     __asm__ volatile(

  1771. #if defined(PIC)

  1772.         "mov               %%"REG_b", %5        \n\t"

  1773. #endif

  1774.         "pxor                  %%mm7, %%mm7     \n\t"

  1775.         "mov                      %0, %%"REG_c" \n\t"

  1776.         "mov                      %1, %%"REG_D" \n\t"

  1777.         "mov                      %2, %%"REG_d" \n\t"

  1778.         "mov                      %3, %%"REG_b" \n\t"

  1779.         "xor               %%"REG_a", %%"REG_a" \n\t" // i

  1780.         PREFETCH"        (%%"REG_c")            \n\t"

  1781.         PREFETCH"      32(%%"REG_c")            \n\t"

  1782.         PREFETCH"      64(%%"REG_c")            \n\t"

  1783. 

  1784. #if ARCH_X86_64

  1785. #define CALL_MMX2_FILTER_CODE \

  1786.         "movl            (%%"REG_b"), %%esi     \n\t"\

  1787.         "call                    *%4            \n\t"\

  1788.         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\

  1789.         "add               %%"REG_S", %%"REG_c" \n\t"\

  1790.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1791.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1792. 

  1793. #else

  1794. #define CALL_MMX2_FILTER_CODE \

  1795.         "movl (%%"REG_b"), %%esi        \n\t"\

  1796.         "call         *%4                       \n\t"\

  1797.         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  1798.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1799.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1800. 

  1801. #endif /* ARCH_X86_64 */

  1802. 

  1803.         CALL_MMX2_FILTER_CODE

  1804.         CALL_MMX2_FILTER_CODE

  1805.         CALL_MMX2_FILTER_CODE

  1806.         CALL_MMX2_FILTER_CODE

  1807.         CALL_MMX2_FILTER_CODE

  1808.         CALL_MMX2_FILTER_CODE

  1809.         CALL_MMX2_FILTER_CODE

  1810.         CALL_MMX2_FILTER_CODE

  1811. 

  1812. #if defined(PIC)

  1813.         "mov                      %5, %%"REG_b" \n\t"

  1814. #endif

  1815.         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),

  1816.            "m" (mmx2FilterCode)

  1817. #if defined(PIC)

  1818.           ,"m" (ebxsave)

  1819. #endif

  1820.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  1821. #if !defined(PIC)

  1822.          ,"%"REG_b

  1823. #endif

  1824.     );

  1825. 

  1826.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)

  1827.         dst[i] = src[srcW-1]*128;

  1828. }

  1829. 

  1830. static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,

  1831.                                  int dstWidth, const uint8_t *src1,

  1832.                                  const uint8_t *src2, int srcW, int xInc)

  1833. {

  1834.     int16_t *filterPos = c->hChrFilterPos;

  1835.     int16_t *filter    = c->hChrFilter;

  1836.     void    *mmx2FilterCode= c->chrMmx2FilterCode;

  1837.     int i;

  1838. #if defined(PIC)

  1839.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  1840. #endif

  1841. 

  1842.     __asm__ volatile(

  1843. #if defined(PIC)

  1844.         "mov          %%"REG_b", %7         \n\t"

  1845. #endif

  1846.         "pxor             %%mm7, %%mm7      \n\t"

  1847.         "mov                 %0, %%"REG_c"  \n\t"

  1848.         "mov                 %1, %%"REG_D"  \n\t"

  1849.         "mov                 %2, %%"REG_d"  \n\t"

  1850.         "mov                 %3, %%"REG_b"  \n\t"

  1851.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  1852.         PREFETCH"   (%%"REG_c")             \n\t"

  1853.         PREFETCH" 32(%%"REG_c")             \n\t"

  1854.         PREFETCH" 64(%%"REG_c")             \n\t"

  1855. 

  1856.         CALL_MMX2_FILTER_CODE

  1857.         CALL_MMX2_FILTER_CODE

  1858.         CALL_MMX2_FILTER_CODE

  1859.         CALL_MMX2_FILTER_CODE

  1860.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  1861.         "mov                 %5, %%"REG_c"  \n\t" // src

  1862.         "mov                 %6, %%"REG_D"  \n\t" // buf2

  1863.         PREFETCH"   (%%"REG_c")             \n\t"

  1864.         PREFETCH" 32(%%"REG_c")             \n\t"

  1865.         PREFETCH" 64(%%"REG_c")             \n\t"

  1866. 

  1867.         CALL_MMX2_FILTER_CODE

  1868.         CALL_MMX2_FILTER_CODE

  1869.         CALL_MMX2_FILTER_CODE

  1870.         CALL_MMX2_FILTER_CODE

  1871. 

  1872. #if defined(PIC)

  1873.         "mov %7, %%"REG_b"    \n\t"

  1874. #endif

  1875.         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),

  1876.            "m" (mmx2FilterCode), "m" (src2), "m"(dst2)

  1877. #if defined(PIC)

  1878.           ,"m" (ebxsave)

  1879. #endif

  1880.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  1881. #if !defined(PIC)

  1882.          ,"%"REG_b

  1883. #endif

  1884.     );

  1885. 

  1886.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {

  1887.         dst1[i] = src1[srcW-1]*128;

  1888.         dst2[i] = src2[srcW-1]*128;

  1889.     }

  1890. }

  1891. #endif /* COMPILE_TEMPLATE_MMX2 */

  1892. 

  1893. static av_cold void RENAME(sws_init_swScale)(SwsContext *c)

  1894. {

  1895.     enum PixelFormat srcFormat = c->srcFormat,

  1896.                      dstFormat = c->dstFormat;

  1897. 

  1898.     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&

  1899.         dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21) {

  1900.         if (!(c->flags & SWS_BITEXACT)) {

  1901.             if (c->flags & SWS_ACCURATE_RND) {

  1902.                 //c->yuv2yuv1 = RENAME(yuv2yuv1_ar    );

  1903.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1904.                     switch (c->dstFormat) {

  1905.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;

  1906.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;

  1907.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;

  1908.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;

  1909.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;

  1910.                     default: break;

  1911.                     }

  1912.                 }

  1913.             } else {

  1914.                 //c->yuv2yuv1 = RENAME(yuv2yuv1    );

  1915.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1916.                     switch (c->dstFormat) {

  1917.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;

  1918.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;

  1919.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;

  1920.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;

  1921.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;

  1922.                     default: break;

  1923.                     }

  1924.                 }

  1925.             }

  1926.         }

  1927.         if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1928.             switch (c->dstFormat) {

  1929.             case PIX_FMT_RGB32:

  1930.                 c->yuv2packed1 = RENAME(yuv2rgb32_1);

  1931.                 c->yuv2packed2 = RENAME(yuv2rgb32_2);

  1932.                 break;

  1933.             case PIX_FMT_BGR24:

  1934.                 c->yuv2packed1 = RENAME(yuv2bgr24_1);

  1935.                 c->yuv2packed2 = RENAME(yuv2bgr24_2);

  1936.                 break;

  1937.             case PIX_FMT_RGB555:

  1938.                 c->yuv2packed1 = RENAME(yuv2rgb555_1);

  1939.                 c->yuv2packed2 = RENAME(yuv2rgb555_2);

  1940.                 break;

  1941.             case PIX_FMT_RGB565:

  1942.                 c->yuv2packed1 = RENAME(yuv2rgb565_1);

  1943.                 c->yuv2packed2 = RENAME(yuv2rgb565_2);

  1944.                 break;

  1945.             case PIX_FMT_YUYV422:

  1946.                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);

  1947.                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);

  1948.                 break;

  1949.             default:

  1950.                 break;

  1951.             }

  1952.         }

  1953.     }

  1954. 

  1955.     if (c->srcBpc == 8 && c->dstBpc <= 10) {

  1956.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  1957. #if COMPILE_TEMPLATE_MMX2

  1958.     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)

  1959.     {

  1960.         c->hyscale_fast = RENAME(hyscale_fast);

  1961.         c->hcscale_fast = RENAME(hcscale_fast);

  1962.     } else {

  1963. #endif /* COMPILE_TEMPLATE_MMX2 */

  1964.         c->hyscale_fast = NULL;

  1965.         c->hcscale_fast = NULL;

  1966. #if COMPILE_TEMPLATE_MMX2

  1967.     }

  1968. #endif /* COMPILE_TEMPLATE_MMX2 */

  1969.     }

  1970. 

  1971. #if !COMPILE_TEMPLATE_MMX2

  1972.     switch(srcFormat) {

  1973.         case PIX_FMT_YUYV422  : c->chrToYV12 = RENAME(yuy2ToUV); break;

  1974.         case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;

  1975.         case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;

  1976.         case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;

  1977.         default: break;

  1978.     }

  1979. #endif /* !COMPILE_TEMPLATE_MMX2 */

  1980.     if (!c->chrSrcHSubSample) {

  1981.         switch(srcFormat) {

  1982.         case PIX_FMT_BGR24  : c->chrToYV12 = RENAME(bgr24ToUV); break;

  1983.         case PIX_FMT_RGB24  : c->chrToYV12 = RENAME(rgb24ToUV); break;

  1984.         default: break;

  1985.         }

  1986.     }

  1987. 

  1988.     switch (srcFormat) {

  1989. #if !COMPILE_TEMPLATE_MMX2

  1990.     case PIX_FMT_YUYV422  :

  1991.     case PIX_FMT_Y400A    : c->lumToYV12 = RENAME(yuy2ToY); break;

  1992.     case PIX_FMT_UYVY422  : c->lumToYV12 = RENAME(uyvyToY); break;

  1993. #endif /* !COMPILE_TEMPLATE_MMX2 */

  1994.     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;

  1995.     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;

  1996.     default: break;

  1997.     }

  1998. #if !COMPILE_TEMPLATE_MMX2

  1999.     if (c->alpPixBuf) {

  2000.         switch (srcFormat) {

  2001.         case PIX_FMT_Y400A  : c->alpToYV12 = RENAME(yuy2ToY); break;

  2002.         default: break;

  2003.         }

  2004.     }

  2005. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2006. }