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.
33139 Commits
32 Branches
511MB
Tree: a75b254c74
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a75b254c74'
${ noResults }
FFmpeg/libswscale/x86/swscale_template.c

2205 lines
99KB
Raw Blame History 

  1. /*

  2.  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

  6.  * FFmpeg is free software; you can redistribute it and/or

  7.  * modify it under the terms of the GNU Lesser General Public

  8.  * License as published by the Free Software Foundation; either

  9.  * version 2.1 of the License, or (at your option) any later version.

  10.  *

  11.  * FFmpeg is distributed in the hope that it will be useful,

  12.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  13.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  14.  * Lesser General Public License for more details.

  15.  *

  16.  * You should have received a copy of the GNU Lesser General Public

  17.  * License along with FFmpeg; if not, write to the Free Software

  18.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  19.  */

  20. 

  21. #undef REAL_MOVNTQ

  22. #undef MOVNTQ

  23. #undef PREFETCH

  24. 

  25. #if COMPILE_TEMPLATE_MMX2

  26. #define PREFETCH "prefetchnta"

  27. #else

  28. #define PREFETCH  " # nop"

  29. #endif

  30. 

  31. #if COMPILE_TEMPLATE_MMX2

  32. #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"

  33. #else

  34. #define REAL_MOVNTQ(a,b) "movq " #a ", " #b " \n\t"

  35. #endif

  36. #define MOVNTQ(a,b)  REAL_MOVNTQ(a,b)

  37. 

  38. #define YSCALEYUV2YV12X(offset, dest, end, pos) \

  39.     __asm__ volatile(\

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

  41.         "movq                  "DITHER16"+8(%0), %%mm4      \n\t"\

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

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

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

  45.         "1:                                                 \n\t"\

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

  47.         "movq                (%%"REG_S", %3, 2), %%mm2      \n\t" /* srcData */\

  48.         "movq               8(%%"REG_S", %3, 2), %%mm5      \n\t" /* srcData */\

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

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

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

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

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

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

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

  56.         " jnz                                1b             \n\t"\

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

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

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

  60.         MOVNTQ(%%mm3, (%1, %3))\

  61.         "add                                 $8, %3         \n\t"\

  62.         "cmp                                 %2, %3         \n\t"\

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

  64.         "movq                  "DITHER16"+8(%0), %%mm4      \n\t"\

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

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

  67.         "jb                                  1b             \n\t"\

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

  69.            "r" (dest), "g" ((x86_reg)(end)), "r"((x86_reg)(pos))\

  70.         : "%"REG_d, "%"REG_S\

  71.     );

  72. 

  73. #if !COMPILE_TEMPLATE_MMX2

  74. static av_always_inline void

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

  76. {

  77.     if (rot) {

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  92.                          );

  93.     } else {

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

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

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

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

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

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

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

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

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

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

  104.                          );

  105.     }

  106. }

  107. #endif

  108. 

  109. static void RENAME(yuv2yuvX)(SwsContext *c, const int16_t *lumFilter,

  110.                              const int16_t **lumSrc, int lumFilterSize,

  111.                              const int16_t *chrFilter, const int16_t **chrUSrc,

  112.                              const int16_t **chrVSrc,

  113.                              int chrFilterSize, const int16_t **alpSrc,

  114.                              uint8_t *dest[4], int dstW, int chrDstW)

  115. {

  116.     uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2],

  117.             *aDest = CONFIG_SWSCALE_ALPHA ? dest[3] : NULL;

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

  119. 

  120.     if (uDest) {

  121.         x86_reg uv_off = c->uv_offx2 >> 1;

  122.         dither_8to16(c, chrDither, 0);

  123.         YSCALEYUV2YV12X(CHR_MMX_FILTER_OFFSET, uDest, chrDstW, 0)

  124.         dither_8to16(c, chrDither, 1);

  125.         YSCALEYUV2YV12X(CHR_MMX_FILTER_OFFSET, vDest - uv_off, chrDstW + uv_off, uv_off)

  126.     }

  127.     dither_8to16(c, lumDither, 0);

  128.     if (CONFIG_SWSCALE_ALPHA && aDest) {

  129.         YSCALEYUV2YV12X(ALP_MMX_FILTER_OFFSET, aDest, dstW, 0)

  130.     }

  131. 

  132.     YSCALEYUV2YV12X(LUM_MMX_FILTER_OFFSET, yDest, dstW, 0)

  133. }

  134. 

  135. #define YSCALEYUV2YV12X_ACCURATE(offset, dest, end, pos) \

  136.     __asm__ volatile(\

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

  138.         "movq                  "DITHER32"+0(%0), %%mm4      \n\t"\

  139.         "movq                  "DITHER32"+8(%0), %%mm5      \n\t"\

  140.         "movq                 "DITHER32"+16(%0), %%mm6      \n\t"\

  141.         "movq                 "DITHER32"+24(%0), %%mm7      \n\t"\

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

  143.         ".p2align                             4             \n\t"\

  144.         "1:                                                 \n\t"\

  145.         "movq                (%%"REG_S", %3, 2), %%mm0      \n\t" /* srcData */\

  146.         "movq               8(%%"REG_S", %3, 2), %%mm2      \n\t" /* srcData */\

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

  148.         "movq                (%%"REG_S", %3, 2), %%mm1      \n\t" /* srcData */\

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

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

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

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

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

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

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

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

  157.         "movq               8(%%"REG_S", %3, 2), %%mm3      \n\t" /* srcData */\

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

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

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

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

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

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

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

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

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

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

  168.         " jnz                                1b             \n\t"\

  169.         "psrad                              $19, %%mm4      \n\t"\

  170.         "psrad                              $19, %%mm5      \n\t"\

  171.         "psrad                              $19, %%mm6      \n\t"\

  172.         "psrad                              $19, %%mm7      \n\t"\

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

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

  175.         "packuswb                         %%mm6, %%mm4      \n\t"\

  176.         MOVNTQ(%%mm4, (%1, %3))\

  177.         "add                                 $8, %3         \n\t"\

  178.         "cmp                                 %2, %3         \n\t"\

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

  180.         "movq                  "DITHER32"+0(%0), %%mm4      \n\t"\

  181.         "movq                  "DITHER32"+8(%0), %%mm5      \n\t"\

  182.         "movq                 "DITHER32"+16(%0), %%mm6      \n\t"\

  183.         "movq                 "DITHER32"+24(%0), %%mm7      \n\t"\

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

  185.         "jb                                  1b             \n\t"\

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

  187.         "r" (dest), "g" ((x86_reg)(end)), "r"((x86_reg)(pos))\

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

  189.     );

  190. 

  191. #if !COMPILE_TEMPLATE_MMX2

  192. static av_always_inline void

  193. dither_8to32(SwsContext *c, const uint8_t *srcDither, int rot)

  194. {

  195.     if (rot) {

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

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

  198.                          "movq      %%mm4, %%mm5\n\t"

  199.                          "psrlq       $24, %%mm4\n\t"

  200.                          "psllq       $40, %%mm5\n\t"

  201.                          "por       %%mm5, %%mm4\n\t"

  202.                          "movq      %%mm4, %%mm6\n\t"

  203.                          "punpcklbw %%mm0, %%mm4\n\t"

  204.                          "punpckhbw %%mm0, %%mm6\n\t"

  205.                          "movq      %%mm4, %%mm5\n\t"

  206.                          "movq      %%mm6, %%mm7\n\t"

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

  208.                          "punpckhwd %%mm0, %%mm5\n\t"

  209.                          "punpcklwd %%mm0, %%mm6\n\t"

  210.                          "punpckhwd %%mm0, %%mm7\n\t"

  211.                          "pslld       $12, %%mm4\n\t"

  212.                          "pslld       $12, %%mm5\n\t"

  213.                          "pslld       $12, %%mm6\n\t"

  214.                          "pslld       $12, %%mm7\n\t"

  215.                          "movq      %%mm4, "DITHER32"+0(%1)\n\t"

  216.                          "movq      %%mm5, "DITHER32"+8(%1)\n\t"

  217.                          "movq      %%mm6, "DITHER32"+16(%1)\n\t"

  218.                          "movq      %%mm7, "DITHER32"+24(%1)\n\t"

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

  220.                          );

  221.     } else {

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

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

  224.                          "movq      %%mm4, %%mm6\n\t"

  225.                          "punpcklbw %%mm0, %%mm4\n\t"

  226.                          "punpckhbw %%mm0, %%mm6\n\t"

  227.                          "movq      %%mm4, %%mm5\n\t"

  228.                          "movq      %%mm6, %%mm7\n\t"

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

  230.                          "punpckhwd %%mm0, %%mm5\n\t"

  231.                          "punpcklwd %%mm0, %%mm6\n\t"

  232.                          "punpckhwd %%mm0, %%mm7\n\t"

  233.                          "pslld       $12, %%mm4\n\t"

  234.                          "pslld       $12, %%mm5\n\t"

  235.                          "pslld       $12, %%mm6\n\t"

  236.                          "pslld       $12, %%mm7\n\t"

  237.                          "movq      %%mm4, "DITHER32"+0(%1)\n\t"

  238.                          "movq      %%mm5, "DITHER32"+8(%1)\n\t"

  239.                          "movq      %%mm6, "DITHER32"+16(%1)\n\t"

  240.                          "movq      %%mm7, "DITHER32"+24(%1)\n\t"

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

  242.                          );

  243.     }

  244. }

  245. #endif

  246. 

  247. static void RENAME(yuv2yuvX_ar)(SwsContext *c, const int16_t *lumFilter,

  248.                                 const int16_t **lumSrc, int lumFilterSize,

  249.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  250.                                 const int16_t **chrVSrc,

  251.                                 int chrFilterSize, const int16_t **alpSrc,

  252.                                 uint8_t *dest[4], int dstW, int chrDstW)

  253. {

  254.     uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2],

  255.             *aDest = CONFIG_SWSCALE_ALPHA ? dest[3] : NULL;

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

  257. 

  258.     if (uDest) {

  259.         x86_reg uv_off = c->uv_offx2 >> 1;

  260.         dither_8to32(c, chrDither, 0);

  261.         YSCALEYUV2YV12X_ACCURATE(CHR_MMX_FILTER_OFFSET, uDest, chrDstW, 0)

  262.         dither_8to32(c, chrDither, 1);

  263.         YSCALEYUV2YV12X_ACCURATE(CHR_MMX_FILTER_OFFSET, vDest - uv_off, chrDstW + uv_off, uv_off)

  264.     }

  265.     dither_8to32(c, lumDither, 0);

  266.     if (CONFIG_SWSCALE_ALPHA && aDest) {

  267.         YSCALEYUV2YV12X_ACCURATE(ALP_MMX_FILTER_OFFSET, aDest, dstW, 0)

  268.     }

  269. 

  270.     YSCALEYUV2YV12X_ACCURATE(LUM_MMX_FILTER_OFFSET, yDest, dstW, 0)

  271. }

  272. 

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

  274.                              const int16_t *chrUSrc, const int16_t *chrVSrc,

  275.                              const int16_t *alpSrc,

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

  277. {

  278.     int p= 4;

  279.     const int16_t *src[4]= {

  280.         lumSrc + dstW,     chrUSrc + chrDstW,

  281.         chrVSrc + chrDstW, alpSrc + dstW

  282.     };

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

  284. 

  285.     while (p--) {

  286.         if (dst[p]) {

  287.             __asm__ volatile(

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

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

  290.                 "1:                                   \n\t"

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

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

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

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

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

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

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

  298.                 "jnc                   1b             \n\t"

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

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

  301.                 : "%"REG_a

  302.             );

  303.         }

  304.     }

  305. }

  306. 

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

  308.                                 const int16_t *chrUSrc, const int16_t *chrVSrc,

  309.                                 const int16_t *alpSrc,

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

  311. {

  312.     int p= 4;

  313.     const int16_t *src[4]= {

  314.         lumSrc + dstW,     chrUSrc + chrDstW,

  315.         chrVSrc + chrDstW, alpSrc + dstW

  316.     };

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

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

  319. 

  320.     while (p--) {

  321.         if (dst[p]) {

  322.             int i;

  323.             for(i=0; i<8; i++) c->dither16[i] = (p == 2 || p == 3) ? lumDither[i] : chrDither[i];

  324.             __asm__ volatile(

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

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

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

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

  329.                 "1:                                   \n\t"

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

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

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

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

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

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

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

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

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

  339.                 "jnc                   1b             \n\t"

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

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

  342.                 : "%"REG_a

  343.             );

  344.         }

  345.     }

  346. }

  347. 

  348. #define YSCALEYUV2PACKEDX_UV \

  349.     __asm__ volatile(\

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

  351.         ".p2align                      4                \n\t"\

  352.         "nop                                            \n\t"\

  353.         "1:                                             \n\t"\

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

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

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

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

  358.         ".p2align                      4                \n\t"\

  359.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

  372. 

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

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

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

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

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

  378.     ".p2align                      4                \n\t"\

  379.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

  391. 

  392. #define YSCALEYUV2PACKEDX \

  393.     YSCALEYUV2PACKEDX_UV \

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

  395. 

  396. #define YSCALEYUV2PACKEDX_END                     \

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

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

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

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

  401.     );

  402. 

  403. #define YSCALEYUV2PACKEDX_ACCURATE_UV \

  404.     __asm__ volatile(\

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

  406.         ".p2align                      4                \n\t"\

  407.         "nop                                            \n\t"\

  408.         "1:                                             \n\t"\

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

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

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

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

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

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

  415.         ".p2align                      4                \n\t"\

  416.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  454. 

  455. #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \

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

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

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

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

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

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

  462.     ".p2align                      4                \n\t"\

  463.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  499. 

  500. #define YSCALEYUV2PACKEDX_ACCURATE \

  501.     YSCALEYUV2PACKEDX_ACCURATE_UV \

  502.     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)

  503. 

  504. #define YSCALEYUV2RGBX \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  539. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  553. \

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

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

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

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

  558. \

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

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

  561.     " jb      1b                \n\t"

  562. #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)

  563. 

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

  565.                                    const int16_t **lumSrc, int lumFilterSize,

  566.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  567.                                    const int16_t **chrVSrc,

  568.                                    int chrFilterSize, const int16_t **alpSrc,

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

  570. {

  571.     x86_reg dummy=0;

  572.     x86_reg dstW_reg = dstW;

  573.     x86_reg uv_off = c->uv_offx2;

  574. 

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

  576.         YSCALEYUV2PACKEDX_ACCURATE

  577.         YSCALEYUV2RGBX

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

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

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

  581.         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)

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

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

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

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

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

  587.         YSCALEYUV2PACKEDX_END

  588.     } else {

  589.         YSCALEYUV2PACKEDX_ACCURATE

  590.         YSCALEYUV2RGBX

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

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

  593.         YSCALEYUV2PACKEDX_END

  594.     }

  595. }

  596. 

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

  598.                                 const int16_t **lumSrc, int lumFilterSize,

  599.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  600.                                 const int16_t **chrVSrc,

  601.                                 int chrFilterSize, const int16_t **alpSrc,

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

  603. {

  604.     x86_reg dummy=0;

  605.     x86_reg dstW_reg = dstW;

  606.     x86_reg uv_off = c->uv_offx2;

  607. 

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

  609.         YSCALEYUV2PACKEDX

  610.         YSCALEYUV2RGBX

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

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

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

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

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

  616.         YSCALEYUV2PACKEDX_END

  617.     } else {

  618.         YSCALEYUV2PACKEDX

  619.         YSCALEYUV2RGBX

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

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

  622.         YSCALEYUV2PACKEDX_END

  623.     }

  624. }

  625. 

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

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

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

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

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

  631. \

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

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

  634. \

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

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

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

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

  639. \

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

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

  642. \

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

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

  645. \

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

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

  648. \

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

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

  651.     " jb             1b             \n\t"

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

  653. 

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

  655.                                     const int16_t **lumSrc, int lumFilterSize,

  656.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  657.                                     const int16_t **chrVSrc,

  658.                                     int chrFilterSize, const int16_t **alpSrc,

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

  660. {

  661.     x86_reg dummy=0;

  662.     x86_reg dstW_reg = dstW;

  663.     x86_reg uv_off = c->uv_offx2;

  664. 

  665.     YSCALEYUV2PACKEDX_ACCURATE

  666.     YSCALEYUV2RGBX

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

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

  669. #ifdef DITHER1XBPP

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

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

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

  673. #endif

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

  675.     YSCALEYUV2PACKEDX_END

  676. }

  677. 

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

  679.                                  const int16_t **lumSrc, int lumFilterSize,

  680.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  681.                                  const int16_t **chrVSrc,

  682.                                  int chrFilterSize, const int16_t **alpSrc,

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

  684. {

  685.     x86_reg dummy=0;

  686.     x86_reg dstW_reg = dstW;

  687.     x86_reg uv_off = c->uv_offx2;

  688. 

  689.     YSCALEYUV2PACKEDX

  690.     YSCALEYUV2RGBX

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

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

  693. #ifdef DITHER1XBPP

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

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

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

  697. #endif

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

  699.     YSCALEYUV2PACKEDX_END

  700. }

  701. 

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

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

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

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

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

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

  708. \

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

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

  711. \

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

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

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

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

  716. \

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

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

  719. \

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

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

  722. \

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

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

  725. \

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

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

  728.     " jb             1b             \n\t"

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

  730. 

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

  732.                                     const int16_t **lumSrc, int lumFilterSize,

  733.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  734.                                     const int16_t **chrVSrc,

  735.                                     int chrFilterSize, const int16_t **alpSrc,

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

  737. {

  738.     x86_reg dummy=0;

  739.     x86_reg dstW_reg = dstW;

  740.     x86_reg uv_off = c->uv_offx2;

  741. 

  742.     YSCALEYUV2PACKEDX_ACCURATE

  743.     YSCALEYUV2RGBX

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

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

  746. #ifdef DITHER1XBPP

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

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

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

  750. #endif

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

  752.     YSCALEYUV2PACKEDX_END

  753. }

  754. 

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

  756.                                  const int16_t **lumSrc, int lumFilterSize,

  757.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  758.                                  const int16_t **chrVSrc,

  759.                                  int chrFilterSize, const int16_t **alpSrc,

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

  761. {

  762.     x86_reg dummy=0;

  763.     x86_reg dstW_reg = dstW;

  764.     x86_reg uv_off = c->uv_offx2;

  765. 

  766.     YSCALEYUV2PACKEDX

  767.     YSCALEYUV2RGBX

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

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

  770. #ifdef DITHER1XBPP

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

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

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

  774. #endif

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

  776.     YSCALEYUV2PACKEDX_END

  777. }

  778. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  793. \

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

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

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

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

  798. \

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

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

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

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

  803. \

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

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

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

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

  808. \

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

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

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

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

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

  814. \

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

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

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

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

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

  820. \

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

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

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

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

  825. \

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

  827. \

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

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

  830.     " jb          1b            \n\t"

  831. 

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

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

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

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

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

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

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

  839. \

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

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

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

  843. \

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

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

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

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

  848. \

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

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

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

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

  853. \

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

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

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

  857. \

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

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

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

  861. \

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

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

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

  865. \

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

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

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

  869. \

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

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

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

  873. \

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

  875. \

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

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

  878.     " jb       1b               \n\t"

  879. 

  880. #if COMPILE_TEMPLATE_MMX2

  881. #undef WRITEBGR24

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

  883. #else

  884. #undef WRITEBGR24

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

  886. #endif

  887. 

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

  889.                                    const int16_t **lumSrc, int lumFilterSize,

  890.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  891.                                    const int16_t **chrVSrc,

  892.                                    int chrFilterSize, const int16_t **alpSrc,

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

  894. {

  895.     x86_reg dummy=0;

  896.     x86_reg dstW_reg = dstW;

  897.     x86_reg uv_off = c->uv_offx2;

  898. 

  899.     YSCALEYUV2PACKEDX_ACCURATE

  900.     YSCALEYUV2RGBX

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

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

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

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

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

  906.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  909.     );

  910. }

  911. 

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

  913.                                 const int16_t **lumSrc, int lumFilterSize,

  914.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  915.                                 const int16_t **chrVSrc,

  916.                                 int chrFilterSize, const int16_t **alpSrc,

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

  918. {

  919.     x86_reg dummy=0;

  920.     x86_reg dstW_reg = dstW;

  921.     x86_reg uv_off = c->uv_offx2;

  922. 

  923.     YSCALEYUV2PACKEDX

  924.     YSCALEYUV2RGBX

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

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

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

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

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

  930.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  933.     );

  934. }

  935. 

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

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

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

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

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

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

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

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

  944. \

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

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

  947. \

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

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

  950.     " jb          1b            \n\t"

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

  952. 

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

  954.                                      const int16_t **lumSrc, int lumFilterSize,

  955.                                      const int16_t *chrFilter, const int16_t **chrUSrc,

  956.                                      const int16_t **chrVSrc,

  957.                                      int chrFilterSize, const int16_t **alpSrc,

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

  959. {

  960.     x86_reg dummy=0;

  961.     x86_reg dstW_reg = dstW;

  962.     x86_reg uv_off = c->uv_offx2;

  963. 

  964.     YSCALEYUV2PACKEDX_ACCURATE

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

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

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

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

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

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

  971.     YSCALEYUV2PACKEDX_END

  972. }

  973. 

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

  975.                                   const int16_t **lumSrc, int lumFilterSize,

  976.                                   const int16_t *chrFilter, const int16_t **chrUSrc,

  977.                                   const int16_t **chrVSrc,

  978.                                   int chrFilterSize, const int16_t **alpSrc,

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

  980. {

  981.     x86_reg dummy=0;

  982.     x86_reg dstW_reg = dstW;

  983.     x86_reg uv_off = c->uv_offx2;

  984. 

  985.     YSCALEYUV2PACKEDX

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

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

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

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

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

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

  992.     YSCALEYUV2PACKEDX_END

  993. }

  994. 

  995. #define REAL_YSCALEYUV2RGB_UV(index, c) \

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

  997.     ".p2align              4            \n\t"\

  998.     "1:                                 \n\t"\

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

  1000.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1001.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1002.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1003.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1004.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1005.     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\

  1006.     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  1007.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\

  1008.     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

  1009.     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

  1010.     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1011.     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1012.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\

  1013.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  1014.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1015.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1016.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1017.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1018.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1019.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1020.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1021. 

  1022. #define REAL_YSCALEYUV2RGB_YA(index, c, b1, b2) \

  1023.     "movq  ("#b1", "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\

  1024.     "movq  ("#b2", "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\

  1025.     "movq 8("#b1", "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\

  1026.     "movq 8("#b2", "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\

  1027.     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\

  1028.     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\

  1029.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1030.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1031.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1032.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1033.     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1034.     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1035. 

  1036. #define REAL_YSCALEYUV2RGB_COEFF(c) \

  1037.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1038.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1039.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1040.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1041.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1042.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1043.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1044.     "paddw             %%mm3, %%mm4     \n\t"\

  1045.     "movq              %%mm2, %%mm0     \n\t"\

  1046.     "movq              %%mm5, %%mm6     \n\t"\

  1047.     "movq              %%mm4, %%mm3     \n\t"\

  1048.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1049.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1050.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1051.     "paddw             %%mm1, %%mm2     \n\t"\

  1052.     "paddw             %%mm1, %%mm5     \n\t"\

  1053.     "paddw             %%mm1, %%mm4     \n\t"\

  1054.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1055.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1056.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1057.     "paddw             %%mm7, %%mm0     \n\t"\

  1058.     "paddw             %%mm7, %%mm6     \n\t"\

  1059.     "paddw             %%mm7, %%mm3     \n\t"\

  1060.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1061.     "packuswb          %%mm0, %%mm2     \n\t"\

  1062.     "packuswb          %%mm6, %%mm5     \n\t"\

  1063.     "packuswb          %%mm3, %%mm4     \n\t"\

  1064. 

  1065. #define YSCALEYUV2RGB_YA(index, c, b1, b2) REAL_YSCALEYUV2RGB_YA(index, c, b1, b2)

  1066. 

  1067. #define YSCALEYUV2RGB(index, c) \

  1068.     REAL_YSCALEYUV2RGB_UV(index, c) \

  1069.     REAL_YSCALEYUV2RGB_YA(index, c, %0, %1) \

  1070.     REAL_YSCALEYUV2RGB_COEFF(c)

  1071. 

  1072. /**

  1073.  * vertical bilinear scale YV12 to RGB

  1074.  */

  1075. static void RENAME(yuv2rgb32_2)(SwsContext *c, const int16_t *buf[2],

  1076.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  1077.                                 const int16_t *abuf[2], uint8_t *dest,

  1078.                                 int dstW, int yalpha, int uvalpha, int y)

  1079. {

  1080.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1081.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1082. 

  1083.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1084.         const int16_t *abuf0 = abuf[0], *abuf1 = abuf[1];

  1085. #if ARCH_X86_64

  1086.         __asm__ volatile(

  1087.             YSCALEYUV2RGB(%%r8, %5)

  1088.             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)

  1089.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1090.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1091.             "packuswb            %%mm7, %%mm1       \n\t"

  1092.             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  1093.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),

  1094.                "a" (&c->redDither),

  1095.                "r" (abuf0), "r" (abuf1)

  1096.             : "%r8"

  1097.         );

  1098. #else

  1099.         c->u_temp=(intptr_t)abuf0;

  1100.         c->v_temp=(intptr_t)abuf1;

  1101.         __asm__ volatile(

  1102.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1103.             "mov        %4, %%"REG_b"               \n\t"

  1104.             "push %%"REG_BP"                        \n\t"

  1105.             YSCALEYUV2RGB(%%REGBP, %5)

  1106.             "push                   %0              \n\t"

  1107.             "push                   %1              \n\t"

  1108.             "mov          "U_TEMP"(%5), %0          \n\t"

  1109.             "mov          "V_TEMP"(%5), %1          \n\t"

  1110.             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)

  1111.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1112.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1113.             "packuswb            %%mm7, %%mm1       \n\t"

  1114.             "pop                    %1              \n\t"

  1115.             "pop                    %0              \n\t"

  1116.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  1117.             "pop %%"REG_BP"                         \n\t"

  1118.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1119.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1120.                "a" (&c->redDither)

  1121.         );

  1122. #endif

  1123.     } else {

  1124.         __asm__ volatile(

  1125.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1126.             "mov        %4, %%"REG_b"               \n\t"

  1127.             "push %%"REG_BP"                        \n\t"

  1128.             YSCALEYUV2RGB(%%REGBP, %5)

  1129.             "pcmpeqd %%mm7, %%mm7                   \n\t"

  1130.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1131.             "pop %%"REG_BP"                         \n\t"

  1132.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1133.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1134.                "a" (&c->redDither)

  1135.         );

  1136.     }

  1137. }

  1138. 

  1139. static void RENAME(yuv2bgr24_2)(SwsContext *c, const int16_t *buf[2],

  1140.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  1141.                                 const int16_t *abuf[2], uint8_t *dest,

  1142.                                 int dstW, int yalpha, int uvalpha, int y)

  1143. {

  1144.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1145.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1146. 

  1147.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1148.     __asm__ volatile(

  1149.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1150.         "mov        %4, %%"REG_b"               \n\t"

  1151.         "push %%"REG_BP"                        \n\t"

  1152.         YSCALEYUV2RGB(%%REGBP, %5)

  1153.         "pxor    %%mm7, %%mm7                   \n\t"

  1154.         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1155.         "pop %%"REG_BP"                         \n\t"

  1156.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1157.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1158.            "a" (&c->redDither)

  1159.     );

  1160. }

  1161. 

  1162. static void RENAME(yuv2rgb555_2)(SwsContext *c, const int16_t *buf[2],

  1163.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  1164.                                  const int16_t *abuf[2], uint8_t *dest,

  1165.                                  int dstW, int yalpha, int uvalpha, int y)

  1166. {

  1167.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1168.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1169. 

  1170.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1171.     __asm__ volatile(

  1172.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1173.         "mov        %4, %%"REG_b"               \n\t"

  1174.         "push %%"REG_BP"                        \n\t"

  1175.         YSCALEYUV2RGB(%%REGBP, %5)

  1176.         "pxor    %%mm7, %%mm7                   \n\t"

  1177.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1178. #ifdef DITHER1XBPP

  1179.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1180.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1181.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1182. #endif

  1183.         WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1184.         "pop %%"REG_BP"                         \n\t"

  1185.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1186.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1187.            "a" (&c->redDither)

  1188.     );

  1189. }

  1190. 

  1191. static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],

  1192.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  1193.                                  const int16_t *abuf[2], uint8_t *dest,

  1194.                                  int dstW, int yalpha, int uvalpha, int y)

  1195. {

  1196.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1197.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1198. 

  1199.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1200.     __asm__ volatile(

  1201.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1202.         "mov        %4, %%"REG_b"               \n\t"

  1203.         "push %%"REG_BP"                        \n\t"

  1204.         YSCALEYUV2RGB(%%REGBP, %5)

  1205.         "pxor    %%mm7, %%mm7                   \n\t"

  1206.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1207. #ifdef DITHER1XBPP

  1208.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1209.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1210.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1211. #endif

  1212.         WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1213.         "pop %%"REG_BP"                         \n\t"

  1214.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1215.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1216.            "a" (&c->redDither)

  1217.     );

  1218. }

  1219. 

  1220. #define REAL_YSCALEYUV2PACKED(index, c) \

  1221.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\

  1222.     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\

  1223.     "psraw                $3, %%mm0                           \n\t"\

  1224.     "psraw                $3, %%mm1                           \n\t"\

  1225.     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1226.     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1227.     "xor            "#index", "#index"                        \n\t"\

  1228.     ".p2align              4            \n\t"\

  1229.     "1:                                 \n\t"\

  1230.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1231.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1232.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1233.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1234.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1235.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1236.     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\

  1237.     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  1238.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\

  1239.     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

  1240.     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

  1241.     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1242.     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1243.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\

  1244.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  1245.     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\

  1246.     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\

  1247.     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\

  1248.     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\

  1249.     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\

  1250.     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\

  1251.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1252.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1253.     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1254.     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1255.     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1256.     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1257. 

  1258. #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)

  1259. 

  1260. static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],

  1261.                                   const int16_t *ubuf[2], const int16_t *vbuf[2],

  1262.                                   const int16_t *abuf[2], uint8_t *dest,

  1263.                                   int dstW, int yalpha, int uvalpha, int y)

  1264. {

  1265.     const int16_t *buf0  = buf[0],  *buf1  = buf[1],

  1266.                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1267. 

  1268.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1269.     __asm__ volatile(

  1270.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1271.         "mov %4, %%"REG_b"                        \n\t"

  1272.         "push %%"REG_BP"                        \n\t"

  1273.         YSCALEYUV2PACKED(%%REGBP, %5)

  1274.         WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1275.         "pop %%"REG_BP"                         \n\t"

  1276.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1277.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1278.            "a" (&c->redDither)

  1279.     );

  1280. }

  1281. 

  1282. #define REAL_YSCALEYUV2RGB1(index, c) \

  1283.     "xor            "#index", "#index"  \n\t"\

  1284.     ".p2align              4            \n\t"\

  1285.     "1:                                 \n\t"\

  1286.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1287.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1288.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1289.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1290.     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1291.     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1292.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1293.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1294.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1295.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1296.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1297.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1298.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1299.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1300.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1301.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1302.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1303.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1304.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1305.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1306.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1307.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1308.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1309.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1310.     "paddw             %%mm3, %%mm4     \n\t"\

  1311.     "movq              %%mm2, %%mm0     \n\t"\

  1312.     "movq              %%mm5, %%mm6     \n\t"\

  1313.     "movq              %%mm4, %%mm3     \n\t"\

  1314.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1315.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1316.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1317.     "paddw             %%mm1, %%mm2     \n\t"\

  1318.     "paddw             %%mm1, %%mm5     \n\t"\

  1319.     "paddw             %%mm1, %%mm4     \n\t"\

  1320.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1321.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1322.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1323.     "paddw             %%mm7, %%mm0     \n\t"\

  1324.     "paddw             %%mm7, %%mm6     \n\t"\

  1325.     "paddw             %%mm7, %%mm3     \n\t"\

  1326.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1327.     "packuswb          %%mm0, %%mm2     \n\t"\

  1328.     "packuswb          %%mm6, %%mm5     \n\t"\

  1329.     "packuswb          %%mm3, %%mm4     \n\t"\

  1330. 

  1331. #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)

  1332. 

  1333. // do vertical chrominance interpolation

  1334. #define REAL_YSCALEYUV2RGB1b(index, c) \

  1335.     "xor            "#index", "#index"  \n\t"\

  1336.     ".p2align              4            \n\t"\

  1337.     "1:                                 \n\t"\

  1338.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1339.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1340.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1341.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1342.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1343.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1344.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1345.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1346.     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\

  1347.     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\

  1348.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1349.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1350.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1351.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1352.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1353.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1354.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1355.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1356.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1357.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1358.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1359.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1360.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1361.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1362.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1363.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1364.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1365.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1366.     "paddw             %%mm3, %%mm4     \n\t"\

  1367.     "movq              %%mm2, %%mm0     \n\t"\

  1368.     "movq              %%mm5, %%mm6     \n\t"\

  1369.     "movq              %%mm4, %%mm3     \n\t"\

  1370.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1371.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1372.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1373.     "paddw             %%mm1, %%mm2     \n\t"\

  1374.     "paddw             %%mm1, %%mm5     \n\t"\

  1375.     "paddw             %%mm1, %%mm4     \n\t"\

  1376.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1377.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1378.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1379.     "paddw             %%mm7, %%mm0     \n\t"\

  1380.     "paddw             %%mm7, %%mm6     \n\t"\

  1381.     "paddw             %%mm7, %%mm3     \n\t"\

  1382.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1383.     "packuswb          %%mm0, %%mm2     \n\t"\

  1384.     "packuswb          %%mm6, %%mm5     \n\t"\

  1385.     "packuswb          %%mm3, %%mm4     \n\t"\

  1386. 

  1387. #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)

  1388. 

  1389. #define REAL_YSCALEYUV2RGB1_ALPHA(index) \

  1390.     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\

  1391.     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\

  1392.     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\

  1393.     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\

  1394.     "packuswb          %%mm1, %%mm7     \n\t"

  1395. #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)

  1396. 

  1397. /**

  1398.  * YV12 to RGB without scaling or interpolating

  1399.  */

  1400. static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,

  1401.                                 const int16_t *ubuf[2], const int16_t *bguf[2],

  1402.                                 const int16_t *abuf0, uint8_t *dest,

  1403.                                 int dstW, int uvalpha, int y)

  1404. {

  1405.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1406.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1407. 

  1408.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1409.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1410.             __asm__ volatile(

  1411.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1412.                 "mov        %4, %%"REG_b"               \n\t"

  1413.                 "push %%"REG_BP"                        \n\t"

  1414.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1415.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1416.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1417.                 "pop %%"REG_BP"                         \n\t"

  1418.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1419.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1420.                    "a" (&c->redDither)

  1421.             );

  1422.         } else {

  1423.             __asm__ volatile(

  1424.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1425.                 "mov        %4, %%"REG_b"               \n\t"

  1426.                 "push %%"REG_BP"                        \n\t"

  1427.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1428.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1429.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1430.                 "pop %%"REG_BP"                         \n\t"

  1431.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1432.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1433.                    "a" (&c->redDither)

  1434.             );

  1435.         }

  1436.     } else {

  1437.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1438.             __asm__ volatile(

  1439.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1440.                 "mov        %4, %%"REG_b"               \n\t"

  1441.                 "push %%"REG_BP"                        \n\t"

  1442.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1443.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1444.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1445.                 "pop %%"REG_BP"                         \n\t"

  1446.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1447.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1448.                    "a" (&c->redDither)

  1449.             );

  1450.         } else {

  1451.             __asm__ volatile(

  1452.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1453.                 "mov        %4, %%"REG_b"               \n\t"

  1454.                 "push %%"REG_BP"                        \n\t"

  1455.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1456.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1457.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1458.                 "pop %%"REG_BP"                         \n\t"

  1459.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1460.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1461.                    "a" (&c->redDither)

  1462.             );

  1463.         }

  1464.     }

  1465. }

  1466. 

  1467. static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,

  1468.                                 const int16_t *ubuf[2], const int16_t *bguf[2],

  1469.                                 const int16_t *abuf0, uint8_t *dest,

  1470.                                 int dstW, int uvalpha, int y)

  1471. {

  1472.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1473.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1474. 

  1475.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1476.         __asm__ volatile(

  1477.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1478.             "mov        %4, %%"REG_b"               \n\t"

  1479.             "push %%"REG_BP"                        \n\t"

  1480.             YSCALEYUV2RGB1(%%REGBP, %5)

  1481.             "pxor    %%mm7, %%mm7                   \n\t"

  1482.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1483.             "pop %%"REG_BP"                         \n\t"

  1484.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1485.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1486.                "a" (&c->redDither)

  1487.         );

  1488.     } else {

  1489.         __asm__ volatile(

  1490.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1491.             "mov        %4, %%"REG_b"               \n\t"

  1492.             "push %%"REG_BP"                        \n\t"

  1493.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1494.             "pxor    %%mm7, %%mm7                   \n\t"

  1495.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1496.             "pop %%"REG_BP"                         \n\t"

  1497.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1498.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1499.                "a" (&c->redDither)

  1500.         );

  1501.     }

  1502. }

  1503. 

  1504. static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,

  1505.                                  const int16_t *ubuf[2], const int16_t *bguf[2],

  1506.                                  const int16_t *abuf0, uint8_t *dest,

  1507.                                  int dstW, int uvalpha, int y)

  1508. {

  1509.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1510.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1511. 

  1512.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1513.         __asm__ volatile(

  1514.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1515.             "mov        %4, %%"REG_b"               \n\t"

  1516.             "push %%"REG_BP"                        \n\t"

  1517.             YSCALEYUV2RGB1(%%REGBP, %5)

  1518.             "pxor    %%mm7, %%mm7                   \n\t"

  1519.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1520. #ifdef DITHER1XBPP

  1521.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1522.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1523.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1524. #endif

  1525.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1526.             "pop %%"REG_BP"                         \n\t"

  1527.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1528.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1529.                "a" (&c->redDither)

  1530.         );

  1531.     } else {

  1532.         __asm__ volatile(

  1533.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1534.             "mov        %4, %%"REG_b"               \n\t"

  1535.             "push %%"REG_BP"                        \n\t"

  1536.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1537.             "pxor    %%mm7, %%mm7                   \n\t"

  1538.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1539. #ifdef DITHER1XBPP

  1540.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1541.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1542.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1543. #endif

  1544.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1545.             "pop %%"REG_BP"                         \n\t"

  1546.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1547.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1548.                "a" (&c->redDither)

  1549.         );

  1550.     }

  1551. }

  1552. 

  1553. static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,

  1554.                                  const int16_t *ubuf[2], const int16_t *bguf[2],

  1555.                                  const int16_t *abuf0, uint8_t *dest,

  1556.                                  int dstW, int uvalpha, int y)

  1557. {

  1558.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1559.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1560. 

  1561.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1562.         __asm__ volatile(

  1563.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1564.             "mov        %4, %%"REG_b"               \n\t"

  1565.             "push %%"REG_BP"                        \n\t"

  1566.             YSCALEYUV2RGB1(%%REGBP, %5)

  1567.             "pxor    %%mm7, %%mm7                   \n\t"

  1568.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1569. #ifdef DITHER1XBPP

  1570.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1571.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1572.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1573. #endif

  1574.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1575.             "pop %%"REG_BP"                         \n\t"

  1576.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1577.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1578.                "a" (&c->redDither)

  1579.         );

  1580.     } else {

  1581.         __asm__ volatile(

  1582.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1583.             "mov        %4, %%"REG_b"               \n\t"

  1584.             "push %%"REG_BP"                        \n\t"

  1585.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1586.             "pxor    %%mm7, %%mm7                   \n\t"

  1587.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1588. #ifdef DITHER1XBPP

  1589.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1590.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1591.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1592. #endif

  1593.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1594.             "pop %%"REG_BP"                         \n\t"

  1595.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1596.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1597.                "a" (&c->redDither)

  1598.         );

  1599.     }

  1600. }

  1601. 

  1602. #define REAL_YSCALEYUV2PACKED1(index, c) \

  1603.     "xor            "#index", "#index"  \n\t"\

  1604.     ".p2align              4            \n\t"\

  1605.     "1:                                 \n\t"\

  1606.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1607.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1608.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1609.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1610.     "psraw                $7, %%mm3     \n\t" \

  1611.     "psraw                $7, %%mm4     \n\t" \

  1612.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1613.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1614.     "psraw                $7, %%mm1     \n\t" \

  1615.     "psraw                $7, %%mm7     \n\t" \

  1616. 

  1617. #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)

  1618. 

  1619. #define REAL_YSCALEYUV2PACKED1b(index, c) \

  1620.     "xor "#index", "#index"             \n\t"\

  1621.     ".p2align              4            \n\t"\

  1622.     "1:                                 \n\t"\

  1623.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1624.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1625.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1626.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1627.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1628.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1629.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1630.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1631.     "psrlw                $8, %%mm3     \n\t" \

  1632.     "psrlw                $8, %%mm4     \n\t" \

  1633.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1634.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1635.     "psraw                $7, %%mm1     \n\t" \

  1636.     "psraw                $7, %%mm7     \n\t"

  1637. #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)

  1638. 

  1639. static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,

  1640.                                   const int16_t *ubuf[2], const int16_t *bguf[2],

  1641.                                   const int16_t *abuf0, uint8_t *dest,

  1642.                                   int dstW, int uvalpha, int y)

  1643. {

  1644.     const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1];

  1645.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1646. 

  1647.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1648.         __asm__ volatile(

  1649.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1650.             "mov        %4, %%"REG_b"               \n\t"

  1651.             "push %%"REG_BP"                        \n\t"

  1652.             YSCALEYUV2PACKED1(%%REGBP, %5)

  1653.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1654.             "pop %%"REG_BP"                         \n\t"

  1655.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1656.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1657.                "a" (&c->redDither)

  1658.         );

  1659.     } else {

  1660.         __asm__ volatile(

  1661.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1662.             "mov        %4, %%"REG_b"               \n\t"

  1663.             "push %%"REG_BP"                        \n\t"

  1664.             YSCALEYUV2PACKED1b(%%REGBP, %5)

  1665.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1666.             "pop %%"REG_BP"                         \n\t"

  1667.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1668.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1669.                "a" (&c->redDither)

  1670.         );

  1671.     }

  1672. }

  1673. 

  1674. #if !COMPILE_TEMPLATE_MMX2

  1675. //FIXME yuy2* can read up to 7 samples too much

  1676. 

  1677. static void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src,

  1678.                             int width, uint32_t *unused)

  1679. {

  1680.     __asm__ volatile(

  1681.         "movq "MANGLE(bm01010101)", %%mm2           \n\t"

  1682.         "mov                    %0, %%"REG_a"       \n\t"

  1683.         "1:                                         \n\t"

  1684.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1685.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1686.         "pand                %%mm2, %%mm0           \n\t"

  1687.         "pand                %%mm2, %%mm1           \n\t"

  1688.         "packuswb            %%mm1, %%mm0           \n\t"

  1689.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1690.         "add                    $8, %%"REG_a"       \n\t"

  1691.         " js                    1b                  \n\t"

  1692.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1693.         : "%"REG_a

  1694.     );

  1695. }

  1696. 

  1697. static void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV,

  1698.                              const uint8_t *src1, const uint8_t *src2,

  1699.                              int width, uint32_t *unused)

  1700. {

  1701.     __asm__ volatile(

  1702.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1703.         "mov                    %0, %%"REG_a"       \n\t"

  1704.         "1:                                         \n\t"

  1705.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1706.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1707.         "psrlw                  $8, %%mm0           \n\t"

  1708.         "psrlw                  $8, %%mm1           \n\t"

  1709.         "packuswb            %%mm1, %%mm0           \n\t"

  1710.         "movq                %%mm0, %%mm1           \n\t"

  1711.         "psrlw                  $8, %%mm0           \n\t"

  1712.         "pand                %%mm4, %%mm1           \n\t"

  1713.         "packuswb            %%mm0, %%mm0           \n\t"

  1714.         "packuswb            %%mm1, %%mm1           \n\t"

  1715.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1716.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1717.         "add                    $4, %%"REG_a"       \n\t"

  1718.         " js                    1b                  \n\t"

  1719.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1720.         : "%"REG_a

  1721.     );

  1722.     assert(src1 == src2);

  1723. }

  1724. 

  1725. /* This is almost identical to the previous, end exists only because

  1726.  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */

  1727. static void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src,

  1728.                             int width, uint32_t *unused)

  1729. {

  1730.     __asm__ volatile(

  1731.         "mov                  %0, %%"REG_a"         \n\t"

  1732.         "1:                                         \n\t"

  1733.         "movq  (%1, %%"REG_a",2), %%mm0             \n\t"

  1734.         "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"

  1735.         "psrlw                $8, %%mm0             \n\t"

  1736.         "psrlw                $8, %%mm1             \n\t"

  1737.         "packuswb          %%mm1, %%mm0             \n\t"

  1738.         "movq              %%mm0, (%2, %%"REG_a")   \n\t"

  1739.         "add                  $8, %%"REG_a"         \n\t"

  1740.         " js                  1b                    \n\t"

  1741.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1742.         : "%"REG_a

  1743.     );

  1744. }

  1745. 

  1746. static void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV,

  1747.                              const uint8_t *src1, const uint8_t *src2,

  1748.                              int width, uint32_t *unused)

  1749. {

  1750.     __asm__ volatile(

  1751.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1752.         "mov                    %0, %%"REG_a"       \n\t"

  1753.         "1:                                         \n\t"

  1754.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1755.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1756.         "pand                %%mm4, %%mm0           \n\t"

  1757.         "pand                %%mm4, %%mm1           \n\t"

  1758.         "packuswb            %%mm1, %%mm0           \n\t"

  1759.         "movq                %%mm0, %%mm1           \n\t"

  1760.         "psrlw                  $8, %%mm0           \n\t"

  1761.         "pand                %%mm4, %%mm1           \n\t"

  1762.         "packuswb            %%mm0, %%mm0           \n\t"

  1763.         "packuswb            %%mm1, %%mm1           \n\t"

  1764.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1765.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1766.         "add                    $4, %%"REG_a"       \n\t"

  1767.         " js                    1b                  \n\t"

  1768.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1769.         : "%"REG_a

  1770.     );

  1771.     assert(src1 == src2);

  1772. }

  1773. 

  1774. static av_always_inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,

  1775.                                               const uint8_t *src, int width)

  1776. {

  1777.     __asm__ volatile(

  1778.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1779.         "mov                    %0, %%"REG_a"       \n\t"

  1780.         "1:                                         \n\t"

  1781.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1782.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1783.         "movq                %%mm0, %%mm2           \n\t"

  1784.         "movq                %%mm1, %%mm3           \n\t"

  1785.         "pand                %%mm4, %%mm0           \n\t"

  1786.         "pand                %%mm4, %%mm1           \n\t"

  1787.         "psrlw                  $8, %%mm2           \n\t"

  1788.         "psrlw                  $8, %%mm3           \n\t"

  1789.         "packuswb            %%mm1, %%mm0           \n\t"

  1790.         "packuswb            %%mm3, %%mm2           \n\t"

  1791.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1792.         "movq                %%mm2, (%3, %%"REG_a") \n\t"

  1793.         "add                    $8, %%"REG_a"       \n\t"

  1794.         " js                    1b                  \n\t"

  1795.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)

  1796.         : "%"REG_a

  1797.     );

  1798. }

  1799. 

  1800. static void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,

  1801.                              const uint8_t *src1, const uint8_t *src2,

  1802.                              int width, uint32_t *unused)

  1803. {

  1804.     RENAME(nvXXtoUV)(dstU, dstV, src1, width);

  1805. }

  1806. 

  1807. static void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,

  1808.                              const uint8_t *src1, const uint8_t *src2,

  1809.                              int width, uint32_t *unused)

  1810. {

  1811.     RENAME(nvXXtoUV)(dstV, dstU, src1, width);

  1812. }

  1813. #endif /* !COMPILE_TEMPLATE_MMX2 */

  1814. 

  1815. static av_always_inline void RENAME(bgr24ToY_mmx)(int16_t *dst, const uint8_t *src,

  1816.                                                   int width, enum PixelFormat srcFormat)

  1817. {

  1818. 

  1819.     if(srcFormat == PIX_FMT_BGR24) {

  1820.         __asm__ volatile(

  1821.             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"

  1822.             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"

  1823.             :

  1824.         );

  1825.     } else {

  1826.         __asm__ volatile(

  1827.             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"

  1828.             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"

  1829.             :

  1830.         );

  1831.     }

  1832. 

  1833.     __asm__ volatile(

  1834.         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"

  1835.         "mov                        %2, %%"REG_a"   \n\t"

  1836.         "pxor                    %%mm7, %%mm7       \n\t"

  1837.         "1:                                         \n\t"

  1838.         PREFETCH"               64(%0)              \n\t"

  1839.         "movd                     (%0), %%mm0       \n\t"

  1840.         "movd                    2(%0), %%mm1       \n\t"

  1841.         "movd                    6(%0), %%mm2       \n\t"

  1842.         "movd                    8(%0), %%mm3       \n\t"

  1843.         "add                       $12, %0          \n\t"

  1844.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1845.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1846.         "punpcklbw               %%mm7, %%mm2       \n\t"

  1847.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1848.         "pmaddwd                 %%mm5, %%mm0       \n\t"

  1849.         "pmaddwd                 %%mm6, %%mm1       \n\t"

  1850.         "pmaddwd                 %%mm5, %%mm2       \n\t"

  1851.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1852.         "paddd                   %%mm1, %%mm0       \n\t"

  1853.         "paddd                   %%mm3, %%mm2       \n\t"

  1854.         "paddd                   %%mm4, %%mm0       \n\t"

  1855.         "paddd                   %%mm4, %%mm2       \n\t"

  1856.         "psrad                     $9, %%mm0       \n\t"

  1857.         "psrad                     $9, %%mm2       \n\t"

  1858.         "packssdw                %%mm2, %%mm0       \n\t"

  1859.         "movq                %%mm0, (%1, %%"REG_a") \n\t"

  1860.         "add                        $8, %%"REG_a"   \n\t"

  1861.         " js                        1b              \n\t"

  1862.     : "+r" (src)

  1863.     : "r" (dst+width), "g" ((x86_reg)-2*width)

  1864.     : "%"REG_a

  1865.     );

  1866. }

  1867. 

  1868. static void RENAME(bgr24ToY)(int16_t *dst, const uint8_t *src,

  1869.                              int width, uint32_t *unused)

  1870. {

  1871.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);

  1872. }

  1873. 

  1874. static void RENAME(rgb24ToY)(int16_t *dst, const uint8_t *src,

  1875.                              int width, uint32_t *unused)

  1876. {

  1877.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);

  1878. }

  1879. 

  1880. static av_always_inline void RENAME(bgr24ToUV_mmx)(int16_t *dstU, int16_t *dstV,

  1881.                                                    const uint8_t *src, int width,

  1882.                                                    enum PixelFormat srcFormat)

  1883. {

  1884.     __asm__ volatile(

  1885.         "movq                    24(%4), %%mm6       \n\t"

  1886.         "mov                        %3, %%"REG_a"   \n\t"

  1887.         "pxor                    %%mm7, %%mm7       \n\t"

  1888.         "1:                                         \n\t"

  1889.         PREFETCH"               64(%0)              \n\t"

  1890.         "movd                     (%0), %%mm0       \n\t"

  1891.         "movd                    2(%0), %%mm1       \n\t"

  1892.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1893.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1894.         "movq                    %%mm0, %%mm2       \n\t"

  1895.         "movq                    %%mm1, %%mm3       \n\t"

  1896.         "pmaddwd                  (%4), %%mm0       \n\t"

  1897.         "pmaddwd                 8(%4), %%mm1       \n\t"

  1898.         "pmaddwd                16(%4), %%mm2       \n\t"

  1899.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1900.         "paddd                   %%mm1, %%mm0       \n\t"

  1901.         "paddd                   %%mm3, %%mm2       \n\t"

  1902. 

  1903.         "movd                    6(%0), %%mm1       \n\t"

  1904.         "movd                    8(%0), %%mm3       \n\t"

  1905.         "add                       $12, %0          \n\t"

  1906.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1907.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1908.         "movq                    %%mm1, %%mm4       \n\t"

  1909.         "movq                    %%mm3, %%mm5       \n\t"

  1910.         "pmaddwd                  (%4), %%mm1       \n\t"

  1911.         "pmaddwd                 8(%4), %%mm3       \n\t"

  1912.         "pmaddwd                16(%4), %%mm4       \n\t"

  1913.         "pmaddwd                 %%mm6, %%mm5       \n\t"

  1914.         "paddd                   %%mm3, %%mm1       \n\t"

  1915.         "paddd                   %%mm5, %%mm4       \n\t"

  1916. 

  1917.         "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3       \n\t"

  1918.         "paddd                   %%mm3, %%mm0       \n\t"

  1919.         "paddd                   %%mm3, %%mm2       \n\t"

  1920.         "paddd                   %%mm3, %%mm1       \n\t"

  1921.         "paddd                   %%mm3, %%mm4       \n\t"

  1922.         "psrad                     $9, %%mm0       \n\t"

  1923.         "psrad                     $9, %%mm2       \n\t"

  1924.         "psrad                     $9, %%mm1       \n\t"

  1925.         "psrad                     $9, %%mm4       \n\t"

  1926.         "packssdw                %%mm1, %%mm0       \n\t"

  1927.         "packssdw                %%mm4, %%mm2       \n\t"

  1928.         "movq                %%mm0, (%1, %%"REG_a") \n\t"

  1929.         "movq                %%mm2, (%2, %%"REG_a") \n\t"

  1930.         "add                        $8, %%"REG_a"   \n\t"

  1931.         " js                        1b              \n\t"

  1932.     : "+r" (src)

  1933.     : "r" (dstU+width), "r" (dstV+width), "g" ((x86_reg)-2*width), "r"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24])

  1934.     : "%"REG_a

  1935.     );

  1936. }

  1937. 

  1938. static void RENAME(bgr24ToUV)(int16_t *dstU, int16_t *dstV,

  1939.                               const uint8_t *src1, const uint8_t *src2,

  1940.                               int width, uint32_t *unused)

  1941. {

  1942.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);

  1943.     assert(src1 == src2);

  1944. }

  1945. 

  1946. static void RENAME(rgb24ToUV)(int16_t *dstU, int16_t *dstV,

  1947.                               const uint8_t *src1, const uint8_t *src2,

  1948.                               int width, uint32_t *unused)

  1949. {

  1950.     assert(src1==src2);

  1951.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);

  1952. }

  1953. 

  1954. #if COMPILE_TEMPLATE_MMX2

  1955. static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,

  1956.                                  int dstWidth, const uint8_t *src,

  1957.                                  int srcW, int xInc)

  1958. {

  1959.     int16_t *filterPos = c->hLumFilterPos;

  1960.     int16_t *filter    = c->hLumFilter;

  1961.     void    *mmx2FilterCode= c->lumMmx2FilterCode;

  1962.     int i;

  1963. #if defined(PIC)

  1964.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  1965. #endif

  1966. 

  1967.     __asm__ volatile(

  1968. #if defined(PIC)

  1969.         "mov               %%"REG_b", %5        \n\t"

  1970. #endif

  1971.         "pxor                  %%mm7, %%mm7     \n\t"

  1972.         "mov                      %0, %%"REG_c" \n\t"

  1973.         "mov                      %1, %%"REG_D" \n\t"

  1974.         "mov                      %2, %%"REG_d" \n\t"

  1975.         "mov                      %3, %%"REG_b" \n\t"

  1976.         "xor               %%"REG_a", %%"REG_a" \n\t" // i

  1977.         PREFETCH"        (%%"REG_c")            \n\t"

  1978.         PREFETCH"      32(%%"REG_c")            \n\t"

  1979.         PREFETCH"      64(%%"REG_c")            \n\t"

  1980. 

  1981. #if ARCH_X86_64

  1982. #define CALL_MMX2_FILTER_CODE \

  1983.         "movl            (%%"REG_b"), %%esi     \n\t"\

  1984.         "call                    *%4            \n\t"\

  1985.         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\

  1986.         "add               %%"REG_S", %%"REG_c" \n\t"\

  1987.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1988.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1989. 

  1990. #else

  1991. #define CALL_MMX2_FILTER_CODE \

  1992.         "movl (%%"REG_b"), %%esi        \n\t"\

  1993.         "call         *%4                       \n\t"\

  1994.         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  1995.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1996.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1997. 

  1998. #endif /* ARCH_X86_64 */

  1999. 

  2000.         CALL_MMX2_FILTER_CODE

  2001.         CALL_MMX2_FILTER_CODE

  2002.         CALL_MMX2_FILTER_CODE

  2003.         CALL_MMX2_FILTER_CODE

  2004.         CALL_MMX2_FILTER_CODE

  2005.         CALL_MMX2_FILTER_CODE

  2006.         CALL_MMX2_FILTER_CODE

  2007.         CALL_MMX2_FILTER_CODE

  2008. 

  2009. #if defined(PIC)

  2010.         "mov                      %5, %%"REG_b" \n\t"

  2011. #endif

  2012.         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),

  2013.            "m" (mmx2FilterCode)

  2014. #if defined(PIC)

  2015.           ,"m" (ebxsave)

  2016. #endif

  2017.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2018. #if !defined(PIC)

  2019.          ,"%"REG_b

  2020. #endif

  2021.     );

  2022. 

  2023.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)

  2024.         dst[i] = src[srcW-1]*128;

  2025. }

  2026. 

  2027. static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,

  2028.                                  int dstWidth, const uint8_t *src1,

  2029.                                  const uint8_t *src2, int srcW, int xInc)

  2030. {

  2031.     int16_t *filterPos = c->hChrFilterPos;

  2032.     int16_t *filter    = c->hChrFilter;

  2033.     void    *mmx2FilterCode= c->chrMmx2FilterCode;

  2034.     int i;

  2035. #if defined(PIC)

  2036.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  2037. #endif

  2038. 

  2039.     __asm__ volatile(

  2040. #if defined(PIC)

  2041.         "mov          %%"REG_b", %7         \n\t"

  2042. #endif

  2043.         "pxor             %%mm7, %%mm7      \n\t"

  2044.         "mov                 %0, %%"REG_c"  \n\t"

  2045.         "mov                 %1, %%"REG_D"  \n\t"

  2046.         "mov                 %2, %%"REG_d"  \n\t"

  2047.         "mov                 %3, %%"REG_b"  \n\t"

  2048.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2049.         PREFETCH"   (%%"REG_c")             \n\t"

  2050.         PREFETCH" 32(%%"REG_c")             \n\t"

  2051.         PREFETCH" 64(%%"REG_c")             \n\t"

  2052. 

  2053.         CALL_MMX2_FILTER_CODE

  2054.         CALL_MMX2_FILTER_CODE

  2055.         CALL_MMX2_FILTER_CODE

  2056.         CALL_MMX2_FILTER_CODE

  2057.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2058.         "mov                 %5, %%"REG_c"  \n\t" // src

  2059.         "mov                 %6, %%"REG_D"  \n\t" // buf2

  2060.         PREFETCH"   (%%"REG_c")             \n\t"

  2061.         PREFETCH" 32(%%"REG_c")             \n\t"

  2062.         PREFETCH" 64(%%"REG_c")             \n\t"

  2063. 

  2064.         CALL_MMX2_FILTER_CODE

  2065.         CALL_MMX2_FILTER_CODE

  2066.         CALL_MMX2_FILTER_CODE

  2067.         CALL_MMX2_FILTER_CODE

  2068. 

  2069. #if defined(PIC)

  2070.         "mov %7, %%"REG_b"    \n\t"

  2071. #endif

  2072.         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),

  2073.            "m" (mmx2FilterCode), "m" (src2), "m"(dst2)

  2074. #if defined(PIC)

  2075.           ,"m" (ebxsave)

  2076. #endif

  2077.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2078. #if !defined(PIC)

  2079.          ,"%"REG_b

  2080. #endif

  2081.     );

  2082. 

  2083.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {

  2084.         dst1[i] = src1[srcW-1]*128;

  2085.         dst2[i] = src2[srcW-1]*128;

  2086.     }

  2087. }

  2088. #endif /* COMPILE_TEMPLATE_MMX2 */

  2089. 

  2090. static av_cold void RENAME(sws_init_swScale)(SwsContext *c)

  2091. {

  2092.     enum PixelFormat srcFormat = c->srcFormat,

  2093.                      dstFormat = c->dstFormat;

  2094. 

  2095.     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) && dstFormat != PIX_FMT_NV12

  2096.         && dstFormat != PIX_FMT_NV21 && !(c->flags & SWS_BITEXACT)) {

  2097.             if (c->flags & SWS_ACCURATE_RND) {

  2098.                 c->yuv2yuv1 = RENAME(yuv2yuv1_ar    );

  2099.                 c->yuv2yuvX = RENAME(yuv2yuvX_ar    );

  2100.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  2101.                     switch (c->dstFormat) {

  2102.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;

  2103.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;

  2104.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;

  2105.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;

  2106.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;

  2107.                     default: break;

  2108.                     }

  2109.                 }

  2110.             } else {

  2111.                 int should_dither= isNBPS(c->srcFormat) || is16BPS(c->srcFormat);

  2112.                 c->yuv2yuv1 = should_dither ? RENAME(yuv2yuv1_ar    ) : RENAME(yuv2yuv1    );

  2113.                 c->yuv2yuvX = RENAME(yuv2yuvX    );

  2114.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  2115.                     switch (c->dstFormat) {

  2116.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;

  2117.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;

  2118.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;

  2119.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;

  2120.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;

  2121.                     default: break;

  2122.                     }

  2123.                 }

  2124.             }

  2125.         if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  2126.             switch (c->dstFormat) {

  2127.             case PIX_FMT_RGB32:

  2128.                 c->yuv2packed1 = RENAME(yuv2rgb32_1);

  2129.                 c->yuv2packed2 = RENAME(yuv2rgb32_2);

  2130.                 break;

  2131.             case PIX_FMT_BGR24:

  2132.                 c->yuv2packed1 = RENAME(yuv2bgr24_1);

  2133.                 c->yuv2packed2 = RENAME(yuv2bgr24_2);

  2134.                 break;

  2135.             case PIX_FMT_RGB555:

  2136.                 c->yuv2packed1 = RENAME(yuv2rgb555_1);

  2137.                 c->yuv2packed2 = RENAME(yuv2rgb555_2);

  2138.                 break;

  2139.             case PIX_FMT_RGB565:

  2140.                 c->yuv2packed1 = RENAME(yuv2rgb565_1);

  2141.                 c->yuv2packed2 = RENAME(yuv2rgb565_2);

  2142.                 break;

  2143.             case PIX_FMT_YUYV422:

  2144.                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);

  2145.                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);

  2146.                 break;

  2147.             default:

  2148.                 break;

  2149.             }

  2150.         }

  2151.     }

  2152. 

  2153.     if (c->srcBpc == 8 && c->dstBpc <= 10) {

  2154.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  2155. #if COMPILE_TEMPLATE_MMX2

  2156.     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)

  2157.     {

  2158.         c->hyscale_fast = RENAME(hyscale_fast);

  2159.         c->hcscale_fast = RENAME(hcscale_fast);

  2160.     } else {

  2161. #endif /* COMPILE_TEMPLATE_MMX2 */

  2162.         c->hyscale_fast = NULL;

  2163.         c->hcscale_fast = NULL;

  2164. #if COMPILE_TEMPLATE_MMX2

  2165.     }

  2166. #endif /* COMPILE_TEMPLATE_MMX2 */

  2167.     }

  2168. 

  2169. #if !COMPILE_TEMPLATE_MMX2

  2170.     switch(srcFormat) {

  2171.         case PIX_FMT_YUYV422  : c->chrToYV12 = RENAME(yuy2ToUV); break;

  2172.         case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;

  2173.         case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;

  2174.         case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;

  2175.         default: break;

  2176.     }

  2177. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2178.     if (!c->chrSrcHSubSample) {

  2179.         switch(srcFormat) {

  2180.         case PIX_FMT_BGR24  : c->chrToYV12 = RENAME(bgr24ToUV); break;

  2181.         case PIX_FMT_RGB24  : c->chrToYV12 = RENAME(rgb24ToUV); break;

  2182.         default: break;

  2183.         }

  2184.     }

  2185. 

  2186.     switch (srcFormat) {

  2187. #if !COMPILE_TEMPLATE_MMX2

  2188.     case PIX_FMT_YUYV422  :

  2189.     case PIX_FMT_Y400A    : c->lumToYV12 = RENAME(yuy2ToY); break;

  2190.     case PIX_FMT_UYVY422  : c->lumToYV12 = RENAME(uyvyToY); break;

  2191. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2192.     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;

  2193.     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;

  2194.     default: break;

  2195.     }

  2196. #if !COMPILE_TEMPLATE_MMX2

  2197.     if (c->alpPixBuf) {

  2198.         switch (srcFormat) {

  2199.         case PIX_FMT_Y400A  : c->alpToYV12 = RENAME(yuy2ToY); break;

  2200.         default: break;

  2201.         }

  2202.     }

  2203. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2204. }