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.
30535 Commits
32 Branches
325MB
Tree: 87f40364d1
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '87f40364d1'
${ noResults }
FFmpeg/libswscale/x86/swscale_template.c

2470 lines
113KB
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. static inline void RENAME(yuv2yuvX)(SwsContext *c, const int16_t *lumFilter,

  74.                                     const int16_t **lumSrc, int lumFilterSize,

  75.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  76.                                     const int16_t **chrVSrc,

  77.                                     int chrFilterSize, const int16_t **alpSrc,

  78.                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest,

  79.                                     uint8_t *aDest, int dstW, int chrDstW,

  80.                                     const uint8_t *lumDither, const uint8_t *chrDither)

  81. {

  82.     int i;

  83.     if (uDest) {

  84.         x86_reg uv_off = c->uv_off;

  85.         for(i=0; i<8; i++) c->dither16[i] = chrDither[i]>>4;

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

  87.         for(i=0; i<8; i++) c->dither16[i] = chrDither[(i+3)&7]>>4;

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

  89.     }

  90.     for(i=0; i<8; i++) c->dither16[i] = lumDither[i]>>4;

  91.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  93.     }

  94. 

  95.     YSCALEYUV2YV12X(LUM_MMX_FILTER_OFFSET, dest, dstW, 0)

  96. }

  97. 

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

  99.     __asm__ volatile(\

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

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

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

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

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

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

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

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

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

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

  110.         ".p2align                             4             \n\t"\

  111.         "1:                                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  156.     );

  157. 

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

  159.                                        const int16_t **lumSrc, int lumFilterSize,

  160.                                        const int16_t *chrFilter, const int16_t **chrUSrc,

  161.                                        const int16_t **chrVSrc,

  162.                                        int chrFilterSize, const int16_t **alpSrc,

  163.                                        uint8_t *dest, uint8_t *uDest, uint8_t *vDest,

  164.                                        uint8_t *aDest, int dstW, int chrDstW,

  165.                                        const uint8_t *lumDither, const uint8_t *chrDither)

  166. {

  167.     int i;

  168.     if (uDest) {

  169.         x86_reg uv_off = c->uv_off;

  170.         for(i=0; i<8; i++) c->dither32[i] = chrDither[i]<<12;

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

  172.         for(i=0; i<8; i++) c->dither32[i] = chrDither[(i+3)&7]<<12;

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

  174.     }

  175.     for(i=0; i<8; i++) c->dither32[i] = lumDither[i]<<12;

  176.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  178.     }

  179. 

  180.     YSCALEYUV2YV12X_ACCURATE(LUM_MMX_FILTER_OFFSET, dest, dstW, 0)

  181. }

  182. 

  183. #define YSCALEYUV2YV121 \

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

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

  186.     "1:                                   \n\t"\

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

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

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

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

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

  192.     MOVNTQ(%%mm0, (%1, %%REGa))\

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

  194.     "jnc                   1b             \n\t"

  195. 

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

  197.                                     const int16_t *chrUSrc, const int16_t *chrVSrc,

  198.                                     const int16_t *alpSrc,

  199.                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest,

  200.                                     uint8_t *aDest, int dstW, int chrDstW,

  201.                                     const uint8_t *lumDither, const uint8_t *chrDither)

  202. {

  203.     int p= 4;

  204.     const int16_t *src[4]= { alpSrc + dstW, lumSrc + dstW, chrUSrc + chrDstW, chrVSrc + chrDstW };

  205.     uint8_t *dst[4]= { aDest, dest, uDest, vDest };

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

  207. 

  208.     while (p--) {

  209.         if (dst[p]) {

  210.             __asm__ volatile(

  211.                YSCALEYUV2YV121

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

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

  214.                : "%"REG_a

  215.             );

  216.         }

  217.     }

  218. }

  219. 

  220. #define YSCALEYUV2YV121_ACCURATE \

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

  222.     "movq               0(%3), %%mm6      \n\t"\

  223.     "movq               8(%3), %%mm7      \n\t"\

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

  225.     "1:                                   \n\t"\

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

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

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

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

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

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

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

  233.     MOVNTQ(%%mm0, (%1, %%REGa))\

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

  235.     "jnc                   1b             \n\t"

  236. 

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

  238.                                        const int16_t *chrUSrc, const int16_t *chrVSrc,

  239.                                        const int16_t *alpSrc,

  240.                                        uint8_t *dest, uint8_t *uDest, uint8_t *vDest,

  241.                                        uint8_t *aDest, int dstW, int chrDstW,

  242.                                        const uint8_t *lumDither, const uint8_t *chrDither)

  243. {

  244.     int p= 4;

  245.     const int16_t *src[4]= { alpSrc + dstW, lumSrc + dstW, chrUSrc + chrDstW, chrVSrc + chrDstW };

  246.     uint8_t *dst[4]= { aDest, dest, uDest, vDest };

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

  248. 

  249.     while (p--) {

  250.         if (dst[p]) {

  251.             int i;

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

  253.             __asm__ volatile(

  254.                 YSCALEYUV2YV121_ACCURATE

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

  256.                    "g" (-counter[p]), "r"(c->dither16)

  257.                 : "%"REG_a

  258.             );

  259.         }

  260.     }

  261. }

  262. 

  263. #define YSCALEYUV2PACKEDX_UV \

  264.     __asm__ volatile(\

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

  266.         ".p2align                      4                \n\t"\

  267.         "nop                                            \n\t"\

  268.         "1:                                             \n\t"\

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

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

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

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

  273.         ".p2align                      4                \n\t"\

  274.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

  287. 

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

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

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

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

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

  293.     ".p2align                      4                \n\t"\

  294.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

  306. 

  307. #define YSCALEYUV2PACKEDX \

  308.     YSCALEYUV2PACKEDX_UV \

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

  310. 

  311. #define YSCALEYUV2PACKEDX_END                     \

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

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

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

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

  316.     );

  317. 

  318. #define YSCALEYUV2PACKEDX_ACCURATE_UV \

  319.     __asm__ volatile(\

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

  321.         ".p2align                      4                \n\t"\

  322.         "nop                                            \n\t"\

  323.         "1:                                             \n\t"\

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

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

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

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

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

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

  330.         ".p2align                      4                \n\t"\

  331.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  369. 

  370. #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \

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

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

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

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

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

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

  377.     ".p2align                      4                \n\t"\

  378.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  414. 

  415. #define YSCALEYUV2PACKEDX_ACCURATE \

  416.     YSCALEYUV2PACKEDX_ACCURATE_UV \

  417.     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)

  418. 

  419. #define YSCALEYUV2RGBX \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  454. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  468. \

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

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

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

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

  473. \

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

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

  476.     " jb      1b                \n\t"

  477. #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)

  478. 

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

  480.                                           const int16_t **lumSrc, int lumFilterSize,

  481.                                           const int16_t *chrFilter, const int16_t **chrUSrc,

  482.                                           const int16_t **chrVSrc,

  483.                                           int chrFilterSize, const int16_t **alpSrc,

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

  485. {

  486.     x86_reg dummy=0;

  487.     x86_reg dstW_reg = dstW;

  488.     x86_reg uv_off = c->uv_off << 1;

  489. 

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

  491.         YSCALEYUV2PACKEDX_ACCURATE

  492.         YSCALEYUV2RGBX

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

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

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

  496.         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)

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

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

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

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

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

  502.         YSCALEYUV2PACKEDX_END

  503.     } else {

  504.         YSCALEYUV2PACKEDX_ACCURATE

  505.         YSCALEYUV2RGBX

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

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

  508.         YSCALEYUV2PACKEDX_END

  509.     }

  510. }

  511. 

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

  513.                                        const int16_t **lumSrc, int lumFilterSize,

  514.                                        const int16_t *chrFilter, const int16_t **chrUSrc,

  515.                                        const int16_t **chrVSrc,

  516.                                        int chrFilterSize, const int16_t **alpSrc,

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

  518. {

  519.     x86_reg dummy=0;

  520.     x86_reg dstW_reg = dstW;

  521.     x86_reg uv_off = c->uv_off << 1;

  522. 

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

  524.         YSCALEYUV2PACKEDX

  525.         YSCALEYUV2RGBX

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

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

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

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

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

  531.         YSCALEYUV2PACKEDX_END

  532.     } else {

  533.         YSCALEYUV2PACKEDX

  534.         YSCALEYUV2RGBX

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

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

  537.         YSCALEYUV2PACKEDX_END

  538.     }

  539. }

  540. 

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

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

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

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

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

  546. \

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

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

  549. \

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

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

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

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

  554. \

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

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

  557. \

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

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

  560. \

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

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

  563. \

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

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

  566.     " jb             1b             \n\t"

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

  568. 

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

  570.                                            const int16_t **lumSrc, int lumFilterSize,

  571.                                            const int16_t *chrFilter, const int16_t **chrUSrc,

  572.                                            const int16_t **chrVSrc,

  573.                                            int chrFilterSize, const int16_t **alpSrc,

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

  575. {

  576.     x86_reg dummy=0;

  577.     x86_reg dstW_reg = dstW;

  578.     x86_reg uv_off = c->uv_off << 1;

  579. 

  580.     YSCALEYUV2PACKEDX_ACCURATE

  581.     YSCALEYUV2RGBX

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

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

  584. #ifdef DITHER1XBPP

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

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

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

  588. #endif

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

  590.     YSCALEYUV2PACKEDX_END

  591. }

  592. 

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

  594.                                         const int16_t **lumSrc, int lumFilterSize,

  595.                                         const int16_t *chrFilter, const int16_t **chrUSrc,

  596.                                         const int16_t **chrVSrc,

  597.                                         int chrFilterSize, const int16_t **alpSrc,

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

  599. {

  600.     x86_reg dummy=0;

  601.     x86_reg dstW_reg = dstW;

  602.     x86_reg uv_off = c->uv_off << 1;

  603. 

  604.     YSCALEYUV2PACKEDX

  605.     YSCALEYUV2RGBX

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

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

  608. #ifdef DITHER1XBPP

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

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

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

  612. #endif

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

  614.     YSCALEYUV2PACKEDX_END

  615. }

  616. 

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

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

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

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

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

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

  623. \

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

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

  626. \

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

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

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

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

  631. \

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

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

  634. \

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

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

  637. \

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

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

  640. \

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

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

  643.     " jb             1b             \n\t"

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

  645. 

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

  647.                                            const int16_t **lumSrc, int lumFilterSize,

  648.                                            const int16_t *chrFilter, const int16_t **chrUSrc,

  649.                                            const int16_t **chrVSrc,

  650.                                            int chrFilterSize, const int16_t **alpSrc,

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

  652. {

  653.     x86_reg dummy=0;

  654.     x86_reg dstW_reg = dstW;

  655.     x86_reg uv_off = c->uv_off << 1;

  656. 

  657.     YSCALEYUV2PACKEDX_ACCURATE

  658.     YSCALEYUV2RGBX

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

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

  661. #ifdef DITHER1XBPP

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

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

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

  665. #endif

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

  667.     YSCALEYUV2PACKEDX_END

  668. }

  669. 

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

  671.                                         const int16_t **lumSrc, int lumFilterSize,

  672.                                         const int16_t *chrFilter, const int16_t **chrUSrc,

  673.                                         const int16_t **chrVSrc,

  674.                                         int chrFilterSize, const int16_t **alpSrc,

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

  676. {

  677.     x86_reg dummy=0;

  678.     x86_reg dstW_reg = dstW;

  679.     x86_reg uv_off = c->uv_off << 1;

  680. 

  681.     YSCALEYUV2PACKEDX

  682.     YSCALEYUV2RGBX

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

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

  685. #ifdef DITHER1XBPP

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

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

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

  689. #endif

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

  691.     YSCALEYUV2PACKEDX_END

  692. }

  693. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  708. \

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

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

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

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

  713. \

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

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

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

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

  718. \

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

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

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

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

  723. \

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

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

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

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

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

  729. \

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

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

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

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

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

  735. \

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

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

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

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

  740. \

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

  742. \

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

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

  745.     " jb          1b            \n\t"

  746. 

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

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

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

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

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

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

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

  754. \

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

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

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

  758. \

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

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

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

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

  763. \

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

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

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

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

  768. \

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

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

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

  772. \

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

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

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

  776. \

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

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

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

  780. \

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

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

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

  784. \

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

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

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

  788. \

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

  790. \

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

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

  793.     " jb       1b               \n\t"

  794. 

  795. #if COMPILE_TEMPLATE_MMX2

  796. #undef WRITEBGR24

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

  798. #else

  799. #undef WRITEBGR24

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

  801. #endif

  802. 

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

  804.                                           const int16_t **lumSrc, int lumFilterSize,

  805.                                           const int16_t *chrFilter, const int16_t **chrUSrc,

  806.                                           const int16_t **chrVSrc,

  807.                                           int chrFilterSize, const int16_t **alpSrc,

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

  809. {

  810.     x86_reg dummy=0;

  811.     x86_reg dstW_reg = dstW;

  812.     x86_reg uv_off = c->uv_off << 1;

  813. 

  814.     YSCALEYUV2PACKEDX_ACCURATE

  815.     YSCALEYUV2RGBX

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

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

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

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

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

  821.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  824.     );

  825. }

  826. 

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

  828.                                        const int16_t **lumSrc, int lumFilterSize,

  829.                                        const int16_t *chrFilter, const int16_t **chrUSrc,

  830.                                        const int16_t **chrVSrc,

  831.                                        int chrFilterSize, const int16_t **alpSrc,

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

  833. {

  834.     x86_reg dummy=0;

  835.     x86_reg dstW_reg = dstW;

  836.     x86_reg uv_off = c->uv_off << 1;

  837. 

  838.     YSCALEYUV2PACKEDX

  839.     YSCALEYUV2RGBX

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

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

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

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

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

  845.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  848.     );

  849. }

  850. 

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

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

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

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

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

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

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

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

  859. \

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

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

  862. \

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

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

  865.     " jb          1b            \n\t"

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

  867. 

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

  869.                                             const int16_t **lumSrc, int lumFilterSize,

  870.                                             const int16_t *chrFilter, const int16_t **chrUSrc,

  871.                                             const int16_t **chrVSrc,

  872.                                             int chrFilterSize, const int16_t **alpSrc,

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

  874. {

  875.     x86_reg dummy=0;

  876.     x86_reg dstW_reg = dstW;

  877.     x86_reg uv_off = c->uv_off << 1;

  878. 

  879.     YSCALEYUV2PACKEDX_ACCURATE

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

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

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

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

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

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

  886.     YSCALEYUV2PACKEDX_END

  887. }

  888. 

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

  890.                                          const int16_t **lumSrc, int lumFilterSize,

  891.                                          const int16_t *chrFilter, const int16_t **chrUSrc,

  892.                                          const int16_t **chrVSrc,

  893.                                          int chrFilterSize, const int16_t **alpSrc,

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

  895. {

  896.     x86_reg dummy=0;

  897.     x86_reg dstW_reg = dstW;

  898.     x86_reg uv_off = c->uv_off << 1;

  899. 

  900.     YSCALEYUV2PACKEDX

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

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

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

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

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

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

  907.     YSCALEYUV2PACKEDX_END

  908. }

  909. 

  910. #define REAL_YSCALEYUV2RGB_UV(index, c) \

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

  912.     ".p2align              4            \n\t"\

  913.     "1:                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  936. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  950. 

  951. #define REAL_YSCALEYUV2RGB_COEFF(c) \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  979. 

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

  981. 

  982. #define YSCALEYUV2RGB(index, c) \

  983.     REAL_YSCALEYUV2RGB_UV(index, c) \

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

  985.     REAL_YSCALEYUV2RGB_COEFF(c)

  986. 

  987. /**

  988.  * vertical bilinear scale YV12 to RGB

  989.  */

  990. static inline void RENAME(yuv2rgb32_2)(SwsContext *c, const uint16_t *buf0,

  991.                                        const uint16_t *buf1, const uint16_t *ubuf0,

  992.                                        const uint16_t *ubuf1, const uint16_t *vbuf0,

  993.                                        const uint16_t *vbuf1, const uint16_t *abuf0,

  994.                                        const uint16_t *abuf1, uint8_t *dest,

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

  996. {

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

  998. #if ARCH_X86_64

  999.         __asm__ volatile(

  1000.             YSCALEYUV2RGB(%%r8, %5)

  1001.             YSCALEYUV2RGB_YA(%%r8, %5, %6, %7)

  1002.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1003.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1004.             "packuswb            %%mm7, %%mm1       \n\t"

  1005.             WRITEBGR32(%4, 8280(%5), %%r8, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  1006.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "r" (dest),

  1007.                "a" (&c->redDither),

  1008.                "r" (abuf0), "r" (abuf1)

  1009.             : "%r8"

  1010.         );

  1011. #else

  1012.         c->u_temp=(intptr_t)abuf0;

  1013.         c->v_temp=(intptr_t)abuf1;

  1014.         __asm__ volatile(

  1015.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1016.             "mov        %4, %%"REG_b"               \n\t"

  1017.             "push %%"REG_BP"                        \n\t"

  1018.             YSCALEYUV2RGB(%%REGBP, %5)

  1019.             "push                   %0              \n\t"

  1020.             "push                   %1              \n\t"

  1021.             "mov          "U_TEMP"(%5), %0          \n\t"

  1022.             "mov          "V_TEMP"(%5), %1          \n\t"

  1023.             YSCALEYUV2RGB_YA(%%REGBP, %5, %0, %1)

  1024.             "psraw                  $3, %%mm1       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1025.             "psraw                  $3, %%mm7       \n\t" /* abuf0[eax] - abuf1[eax] >>7*/

  1026.             "packuswb            %%mm7, %%mm1       \n\t"

  1027.             "pop                    %1              \n\t"

  1028.             "pop                    %0              \n\t"

  1029.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm1, %%mm0, %%mm7, %%mm3, %%mm6)

  1030.             "pop %%"REG_BP"                         \n\t"

  1031.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1032.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1033.                "a" (&c->redDither)

  1034.         );

  1035. #endif

  1036.     } else {

  1037.         __asm__ volatile(

  1038.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1039.             "mov        %4, %%"REG_b"               \n\t"

  1040.             "push %%"REG_BP"                        \n\t"

  1041.             YSCALEYUV2RGB(%%REGBP, %5)

  1042.             "pcmpeqd %%mm7, %%mm7                   \n\t"

  1043.             WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1044.             "pop %%"REG_BP"                         \n\t"

  1045.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1046.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1047.                "a" (&c->redDither)

  1048.         );

  1049.     }

  1050. }

  1051. 

  1052. static inline void RENAME(yuv2bgr24_2)(SwsContext *c, const uint16_t *buf0,

  1053.                                        const uint16_t *buf1, const uint16_t *ubuf0,

  1054.                                        const uint16_t *ubuf1, const uint16_t *vbuf0,

  1055.                                        const uint16_t *vbuf1, const uint16_t *abuf0,

  1056.                                        const uint16_t *abuf1, uint8_t *dest,

  1057.                                        int dstW, int yalpha, int uvalpha, int y)

  1058. {

  1059.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1060.     __asm__ volatile(

  1061.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1062.         "mov        %4, %%"REG_b"               \n\t"

  1063.         "push %%"REG_BP"                        \n\t"

  1064.         YSCALEYUV2RGB(%%REGBP, %5)

  1065.         "pxor    %%mm7, %%mm7                   \n\t"

  1066.         WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1067.         "pop %%"REG_BP"                         \n\t"

  1068.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1069.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1070.            "a" (&c->redDither)

  1071.     );

  1072. }

  1073. 

  1074. static inline void RENAME(yuv2rgb555_2)(SwsContext *c, const uint16_t *buf0,

  1075.                                         const uint16_t *buf1, const uint16_t *ubuf0,

  1076.                                         const uint16_t *ubuf1, const uint16_t *vbuf0,

  1077.                                         const uint16_t *vbuf1, const uint16_t *abuf0,

  1078.                                         const uint16_t *abuf1, uint8_t *dest,

  1079.                                         int dstW, int yalpha, int uvalpha, int y)

  1080. {

  1081.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1082.     __asm__ volatile(

  1083.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1084.         "mov        %4, %%"REG_b"               \n\t"

  1085.         "push %%"REG_BP"                        \n\t"

  1086.         YSCALEYUV2RGB(%%REGBP, %5)

  1087.         "pxor    %%mm7, %%mm7                   \n\t"

  1088.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1089. #ifdef DITHER1XBPP

  1090.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1091.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1092.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1093. #endif

  1094.         WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1095.         "pop %%"REG_BP"                         \n\t"

  1096.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1097.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1098.            "a" (&c->redDither)

  1099.     );

  1100. }

  1101. 

  1102. static inline void RENAME(yuv2rgb565_2)(SwsContext *c, const uint16_t *buf0,

  1103.                                         const uint16_t *buf1, const uint16_t *ubuf0,

  1104.                                         const uint16_t *ubuf1, const uint16_t *vbuf0,

  1105.                                         const uint16_t *vbuf1, const uint16_t *abuf0,

  1106.                                         const uint16_t *abuf1, uint8_t *dest,

  1107.                                         int dstW, int yalpha, int uvalpha, int y)

  1108. {

  1109.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1110.     __asm__ volatile(

  1111.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1112.         "mov        %4, %%"REG_b"               \n\t"

  1113.         "push %%"REG_BP"                        \n\t"

  1114.         YSCALEYUV2RGB(%%REGBP, %5)

  1115.         "pxor    %%mm7, %%mm7                   \n\t"

  1116.         /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1117. #ifdef DITHER1XBPP

  1118.         "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1119.         "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1120.         "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1121. #endif

  1122.         WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1123.         "pop %%"REG_BP"                         \n\t"

  1124.         "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1125.         :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1126.            "a" (&c->redDither)

  1127.     );

  1128. }

  1129. 

  1130. #define REAL_YSCALEYUV2PACKED(index, c) \

  1131.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0              \n\t"\

  1132.     "movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1              \n\t"\

  1133.     "psraw                $3, %%mm0                           \n\t"\

  1134.     "psraw                $3, %%mm1                           \n\t"\

  1135.     "movq              %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1136.     "movq              %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c") \n\t"\

  1137.     "xor            "#index", "#index"                        \n\t"\

  1138.     ".p2align              4            \n\t"\

  1139.     "1:                                 \n\t"\

  1140.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1141.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1142.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1143.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1144.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1145.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1146.     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\

  1147.     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  1148.     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\

  1149.     "pmulhw            %%mm0, %%mm2     \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

  1150.     "pmulhw            %%mm0, %%mm5     \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

  1151.     "psraw                $7, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1152.     "psraw                $7, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1153.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\

  1154.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  1155.     "movq  (%0, "#index", 2), %%mm0     \n\t" /*buf0[eax]*/\

  1156.     "movq  (%1, "#index", 2), %%mm1     \n\t" /*buf1[eax]*/\

  1157.     "movq 8(%0, "#index", 2), %%mm6     \n\t" /*buf0[eax]*/\

  1158.     "movq 8(%1, "#index", 2), %%mm7     \n\t" /*buf1[eax]*/\

  1159.     "psubw             %%mm1, %%mm0     \n\t" /* buf0[eax] - buf1[eax]*/\

  1160.     "psubw             %%mm7, %%mm6     \n\t" /* buf0[eax] - buf1[eax]*/\

  1161.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1162.     "pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  1163.     "psraw                $7, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1164.     "psraw                $7, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1165.     "paddw             %%mm0, %%mm1     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1166.     "paddw             %%mm6, %%mm7     \n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\

  1167. 

  1168. #define YSCALEYUV2PACKED(index, c)  REAL_YSCALEYUV2PACKED(index, c)

  1169. 

  1170. static inline void RENAME(yuv2yuyv422_2)(SwsContext *c, const uint16_t *buf0,

  1171.                                          const uint16_t *buf1, const uint16_t *ubuf0,

  1172.                                          const uint16_t *ubuf1, const uint16_t *vbuf0,

  1173.                                          const uint16_t *vbuf1, const uint16_t *abuf0,

  1174.                                          const uint16_t *abuf1, uint8_t *dest,

  1175.                                          int dstW, int yalpha, int uvalpha, int y)

  1176. {

  1177.     //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1178.     __asm__ volatile(

  1179.         "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1180.         "mov %4, %%"REG_b"                        \n\t"

  1181.         "push %%"REG_BP"                        \n\t"

  1182.         YSCALEYUV2PACKED(%%REGBP, %5)

  1183.         WRITEYUY2(%%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. #define REAL_YSCALEYUV2RGB1(index, c) \

  1192.     "xor            "#index", "#index"  \n\t"\

  1193.     ".p2align              4            \n\t"\

  1194.     "1:                                 \n\t"\

  1195.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1196.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1197.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1198.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1199.     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\

  1200.     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  1201.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1202.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1203.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1204.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1205.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1206.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1207.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1208.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1209.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1210.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1211.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1212.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1213.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1214.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1215.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1216.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1217.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1218.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1219.     "paddw             %%mm3, %%mm4     \n\t"\

  1220.     "movq              %%mm2, %%mm0     \n\t"\

  1221.     "movq              %%mm5, %%mm6     \n\t"\

  1222.     "movq              %%mm4, %%mm3     \n\t"\

  1223.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1224.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1225.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1226.     "paddw             %%mm1, %%mm2     \n\t"\

  1227.     "paddw             %%mm1, %%mm5     \n\t"\

  1228.     "paddw             %%mm1, %%mm4     \n\t"\

  1229.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1230.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1231.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1232.     "paddw             %%mm7, %%mm0     \n\t"\

  1233.     "paddw             %%mm7, %%mm6     \n\t"\

  1234.     "paddw             %%mm7, %%mm3     \n\t"\

  1235.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1236.     "packuswb          %%mm0, %%mm2     \n\t"\

  1237.     "packuswb          %%mm6, %%mm5     \n\t"\

  1238.     "packuswb          %%mm3, %%mm4     \n\t"\

  1239. 

  1240. #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)

  1241. 

  1242. // do vertical chrominance interpolation

  1243. #define REAL_YSCALEYUV2RGB1b(index, c) \

  1244.     "xor            "#index", "#index"  \n\t"\

  1245.     ".p2align              4            \n\t"\

  1246.     "1:                                 \n\t"\

  1247.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1248.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1249.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1250.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1251.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1252.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1253.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1254.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1255.     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\

  1256.     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\

  1257.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1258.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1259.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1260.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1261.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1262.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1263.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1264.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1265.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1266.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1267.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1268.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1269.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1270.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1271.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1272.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1273.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1274.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1275.     "paddw             %%mm3, %%mm4     \n\t"\

  1276.     "movq              %%mm2, %%mm0     \n\t"\

  1277.     "movq              %%mm5, %%mm6     \n\t"\

  1278.     "movq              %%mm4, %%mm3     \n\t"\

  1279.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1280.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1281.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1282.     "paddw             %%mm1, %%mm2     \n\t"\

  1283.     "paddw             %%mm1, %%mm5     \n\t"\

  1284.     "paddw             %%mm1, %%mm4     \n\t"\

  1285.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1286.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1287.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1288.     "paddw             %%mm7, %%mm0     \n\t"\

  1289.     "paddw             %%mm7, %%mm6     \n\t"\

  1290.     "paddw             %%mm7, %%mm3     \n\t"\

  1291.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1292.     "packuswb          %%mm0, %%mm2     \n\t"\

  1293.     "packuswb          %%mm6, %%mm5     \n\t"\

  1294.     "packuswb          %%mm3, %%mm4     \n\t"\

  1295. 

  1296. #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)

  1297. 

  1298. #define REAL_YSCALEYUV2RGB1_ALPHA(index) \

  1299.     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\

  1300.     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\

  1301.     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\

  1302.     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\

  1303.     "packuswb          %%mm1, %%mm7     \n\t"

  1304. #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)

  1305. 

  1306. /**

  1307.  * YV12 to RGB without scaling or interpolating

  1308.  */

  1309. static inline void RENAME(yuv2rgb32_1)(SwsContext *c, const uint16_t *buf0,

  1310.                                        const uint16_t *ubuf0, const uint16_t *ubuf1,

  1311.                                        const uint16_t *vbuf0, const uint16_t *vbuf1,

  1312.                                        const uint16_t *abuf0, uint8_t *dest,

  1313.                                        int dstW, int uvalpha, enum PixelFormat dstFormat,

  1314.                                        int flags, int y)

  1315. {

  1316.     const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1317. 

  1318.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1319.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1320.             __asm__ volatile(

  1321.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1322.                 "mov        %4, %%"REG_b"               \n\t"

  1323.                 "push %%"REG_BP"                        \n\t"

  1324.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1325.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1326.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1327.                 "pop %%"REG_BP"                         \n\t"

  1328.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1329.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1330.                    "a" (&c->redDither)

  1331.             );

  1332.         } else {

  1333.             __asm__ volatile(

  1334.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1335.                 "mov        %4, %%"REG_b"               \n\t"

  1336.                 "push %%"REG_BP"                        \n\t"

  1337.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1338.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1339.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1340.                 "pop %%"REG_BP"                         \n\t"

  1341.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1342.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1343.                    "a" (&c->redDither)

  1344.             );

  1345.         }

  1346.     } else {

  1347.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1348.             __asm__ volatile(

  1349.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1350.                 "mov        %4, %%"REG_b"               \n\t"

  1351.                 "push %%"REG_BP"                        \n\t"

  1352.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1353.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1354.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1355.                 "pop %%"REG_BP"                         \n\t"

  1356.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1357.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1358.                    "a" (&c->redDither)

  1359.             );

  1360.         } else {

  1361.             __asm__ volatile(

  1362.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1363.                 "mov        %4, %%"REG_b"               \n\t"

  1364.                 "push %%"REG_BP"                        \n\t"

  1365.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1366.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1367.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1368.                 "pop %%"REG_BP"                         \n\t"

  1369.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1370.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1371.                    "a" (&c->redDither)

  1372.             );

  1373.         }

  1374.     }

  1375. }

  1376. 

  1377. static inline void RENAME(yuv2bgr24_1)(SwsContext *c, const uint16_t *buf0,

  1378.                                        const uint16_t *ubuf0, const uint16_t *ubuf1,

  1379.                                        const uint16_t *vbuf0, const uint16_t *vbuf1,

  1380.                                        const uint16_t *abuf0, uint8_t *dest,

  1381.                                        int dstW, int uvalpha, enum PixelFormat dstFormat,

  1382.                                        int flags, int y)

  1383. {

  1384.     const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1385. 

  1386.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1387.         __asm__ volatile(

  1388.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1389.             "mov        %4, %%"REG_b"               \n\t"

  1390.             "push %%"REG_BP"                        \n\t"

  1391.             YSCALEYUV2RGB1(%%REGBP, %5)

  1392.             "pxor    %%mm7, %%mm7                   \n\t"

  1393.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1394.             "pop %%"REG_BP"                         \n\t"

  1395.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1396.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1397.                "a" (&c->redDither)

  1398.         );

  1399.     } else {

  1400.         __asm__ volatile(

  1401.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1402.             "mov        %4, %%"REG_b"               \n\t"

  1403.             "push %%"REG_BP"                        \n\t"

  1404.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1405.             "pxor    %%mm7, %%mm7                   \n\t"

  1406.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1407.             "pop %%"REG_BP"                         \n\t"

  1408.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1409.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1410.                "a" (&c->redDither)

  1411.         );

  1412.     }

  1413. }

  1414. 

  1415. static inline void RENAME(yuv2rgb555_1)(SwsContext *c, const uint16_t *buf0,

  1416.                                         const uint16_t *ubuf0, const uint16_t *ubuf1,

  1417.                                         const uint16_t *vbuf0, const uint16_t *vbuf1,

  1418.                                         const uint16_t *abuf0, uint8_t *dest,

  1419.                                         int dstW, int uvalpha, enum PixelFormat dstFormat,

  1420.                                         int flags, int y)

  1421. {

  1422.     const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1423. 

  1424.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1425.         __asm__ volatile(

  1426.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1427.             "mov        %4, %%"REG_b"               \n\t"

  1428.             "push %%"REG_BP"                        \n\t"

  1429.             YSCALEYUV2RGB1(%%REGBP, %5)

  1430.             "pxor    %%mm7, %%mm7                   \n\t"

  1431.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1432. #ifdef DITHER1XBPP

  1433.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1434.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1435.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1436. #endif

  1437.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1438.             "pop %%"REG_BP"                         \n\t"

  1439.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1440.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1441.                "a" (&c->redDither)

  1442.         );

  1443.     } else {

  1444.         __asm__ volatile(

  1445.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1446.             "mov        %4, %%"REG_b"               \n\t"

  1447.             "push %%"REG_BP"                        \n\t"

  1448.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1449.             "pxor    %%mm7, %%mm7                   \n\t"

  1450.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1451. #ifdef DITHER1XBPP

  1452.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1453.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1454.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1455. #endif

  1456.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1457.             "pop %%"REG_BP"                         \n\t"

  1458.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1459.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1460.                "a" (&c->redDither)

  1461.         );

  1462.     }

  1463. }

  1464. 

  1465. static inline void RENAME(yuv2rgb565_1)(SwsContext *c, const uint16_t *buf0,

  1466.                                         const uint16_t *ubuf0, const uint16_t *ubuf1,

  1467.                                         const uint16_t *vbuf0, const uint16_t *vbuf1,

  1468.                                         const uint16_t *abuf0, uint8_t *dest,

  1469.                                         int dstW, int uvalpha, enum PixelFormat dstFormat,

  1470.                                         int flags, int y)

  1471. {

  1472.     const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1473. 

  1474.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1475.         __asm__ volatile(

  1476.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1477.             "mov        %4, %%"REG_b"               \n\t"

  1478.             "push %%"REG_BP"                        \n\t"

  1479.             YSCALEYUV2RGB1(%%REGBP, %5)

  1480.             "pxor    %%mm7, %%mm7                   \n\t"

  1481.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1482. #ifdef DITHER1XBPP

  1483.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1484.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1485.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1486. #endif

  1487.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1488.             "pop %%"REG_BP"                         \n\t"

  1489.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1490.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1491.                "a" (&c->redDither)

  1492.         );

  1493.     } else {

  1494.         __asm__ volatile(

  1495.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1496.             "mov        %4, %%"REG_b"               \n\t"

  1497.             "push %%"REG_BP"                        \n\t"

  1498.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1499.             "pxor    %%mm7, %%mm7                   \n\t"

  1500.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1501. #ifdef DITHER1XBPP

  1502.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1503.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1504.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1505. #endif

  1506.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1507.             "pop %%"REG_BP"                         \n\t"

  1508.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1509.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1510.                "a" (&c->redDither)

  1511.         );

  1512.     }

  1513. }

  1514. 

  1515. #define REAL_YSCALEYUV2PACKED1(index, c) \

  1516.     "xor            "#index", "#index"  \n\t"\

  1517.     ".p2align              4            \n\t"\

  1518.     "1:                                 \n\t"\

  1519.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1520.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1521.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1522.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1523.     "psraw                $7, %%mm3     \n\t" \

  1524.     "psraw                $7, %%mm4     \n\t" \

  1525.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1526.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1527.     "psraw                $7, %%mm1     \n\t" \

  1528.     "psraw                $7, %%mm7     \n\t" \

  1529. 

  1530. #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)

  1531. 

  1532. #define REAL_YSCALEYUV2PACKED1b(index, c) \

  1533.     "xor "#index", "#index"             \n\t"\

  1534.     ".p2align              4            \n\t"\

  1535.     "1:                                 \n\t"\

  1536.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1537.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1538.     "add           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1539.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1540.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1541.     "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \

  1542.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1543.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1544.     "psrlw                $8, %%mm3     \n\t" \

  1545.     "psrlw                $8, %%mm4     \n\t" \

  1546.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1547.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1548.     "psraw                $7, %%mm1     \n\t" \

  1549.     "psraw                $7, %%mm7     \n\t"

  1550. #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)

  1551. 

  1552. static inline void RENAME(yuv2yuyv422_1)(SwsContext *c, const uint16_t *buf0,

  1553.                                          const uint16_t *ubuf0, const uint16_t *ubuf1,

  1554.                                          const uint16_t *vbuf0, const uint16_t *vbuf1,

  1555.                                          const uint16_t *abuf0, uint8_t *dest,

  1556.                                          int dstW, int uvalpha, enum PixelFormat dstFormat,

  1557.                                          int flags, int y)

  1558. {

  1559.     const uint16_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.             YSCALEYUV2PACKED1(%%REGBP, %5)

  1567.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1568.             "pop %%"REG_BP"                         \n\t"

  1569.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1570.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1571.                "a" (&c->redDither)

  1572.         );

  1573.     } else {

  1574.         __asm__ volatile(

  1575.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1576.             "mov        %4, %%"REG_b"               \n\t"

  1577.             "push %%"REG_BP"                        \n\t"

  1578.             YSCALEYUV2PACKED1b(%%REGBP, %5)

  1579.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1580.             "pop %%"REG_BP"                         \n\t"

  1581.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1582.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1583.                "a" (&c->redDither)

  1584.         );

  1585.     }

  1586. }

  1587. 

  1588. #if !COMPILE_TEMPLATE_MMX2

  1589. //FIXME yuy2* can read up to 7 samples too much

  1590. 

  1591. static inline void RENAME(yuy2ToY)(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1592. {

  1593.     __asm__ volatile(

  1594.         "movq "MANGLE(bm01010101)", %%mm2           \n\t"

  1595.         "mov                    %0, %%"REG_a"       \n\t"

  1596.         "1:                                         \n\t"

  1597.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1598.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1599.         "pand                %%mm2, %%mm0           \n\t"

  1600.         "pand                %%mm2, %%mm1           \n\t"

  1601.         "packuswb            %%mm1, %%mm0           \n\t"

  1602.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1603.         "add                    $8, %%"REG_a"       \n\t"

  1604.         " js                    1b                  \n\t"

  1605.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1606.         : "%"REG_a

  1607.     );

  1608. }

  1609. 

  1610. static inline void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1611. {

  1612.     __asm__ volatile(

  1613.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1614.         "mov                    %0, %%"REG_a"       \n\t"

  1615.         "1:                                         \n\t"

  1616.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1617.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1618.         "psrlw                  $8, %%mm0           \n\t"

  1619.         "psrlw                  $8, %%mm1           \n\t"

  1620.         "packuswb            %%mm1, %%mm0           \n\t"

  1621.         "movq                %%mm0, %%mm1           \n\t"

  1622.         "psrlw                  $8, %%mm0           \n\t"

  1623.         "pand                %%mm4, %%mm1           \n\t"

  1624.         "packuswb            %%mm0, %%mm0           \n\t"

  1625.         "packuswb            %%mm1, %%mm1           \n\t"

  1626.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1627.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1628.         "add                    $4, %%"REG_a"       \n\t"

  1629.         " js                    1b                  \n\t"

  1630.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1631.         : "%"REG_a

  1632.     );

  1633.     assert(src1 == src2);

  1634. }

  1635. 

  1636. static inline void RENAME(LEToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1637. {

  1638.     __asm__ volatile(

  1639.         "mov                    %0, %%"REG_a"       \n\t"

  1640.         "1:                                         \n\t"

  1641.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1642.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1643.         "movq    (%2, %%"REG_a",2), %%mm2           \n\t"

  1644.         "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"

  1645.         "psrlw                  $8, %%mm0           \n\t"

  1646.         "psrlw                  $8, %%mm1           \n\t"

  1647.         "psrlw                  $8, %%mm2           \n\t"

  1648.         "psrlw                  $8, %%mm3           \n\t"

  1649.         "packuswb            %%mm1, %%mm0           \n\t"

  1650.         "packuswb            %%mm3, %%mm2           \n\t"

  1651.         "movq                %%mm0, (%3, %%"REG_a") \n\t"

  1652.         "movq                %%mm2, (%4, %%"REG_a") \n\t"

  1653.         "add                    $8, %%"REG_a"       \n\t"

  1654.         " js                    1b                  \n\t"

  1655.         : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)

  1656.         : "%"REG_a

  1657.     );

  1658. }

  1659. 

  1660. /* This is almost identical to the previous, end exists only because

  1661.  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */

  1662. static inline void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1663. {

  1664.     __asm__ volatile(

  1665.         "mov                  %0, %%"REG_a"         \n\t"

  1666.         "1:                                         \n\t"

  1667.         "movq  (%1, %%"REG_a",2), %%mm0             \n\t"

  1668.         "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"

  1669.         "psrlw                $8, %%mm0             \n\t"

  1670.         "psrlw                $8, %%mm1             \n\t"

  1671.         "packuswb          %%mm1, %%mm0             \n\t"

  1672.         "movq              %%mm0, (%2, %%"REG_a")   \n\t"

  1673.         "add                  $8, %%"REG_a"         \n\t"

  1674.         " js                  1b                    \n\t"

  1675.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst+width)

  1676.         : "%"REG_a

  1677.     );

  1678. }

  1679. 

  1680. static inline void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1681. {

  1682.     __asm__ volatile(

  1683.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1684.         "mov                    %0, %%"REG_a"       \n\t"

  1685.         "1:                                         \n\t"

  1686.         "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1687.         "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1688.         "pand                %%mm4, %%mm0           \n\t"

  1689.         "pand                %%mm4, %%mm1           \n\t"

  1690.         "packuswb            %%mm1, %%mm0           \n\t"

  1691.         "movq                %%mm0, %%mm1           \n\t"

  1692.         "psrlw                  $8, %%mm0           \n\t"

  1693.         "pand                %%mm4, %%mm1           \n\t"

  1694.         "packuswb            %%mm0, %%mm0           \n\t"

  1695.         "packuswb            %%mm1, %%mm1           \n\t"

  1696.         "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1697.         "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1698.         "add                    $4, %%"REG_a"       \n\t"

  1699.         " js                    1b                  \n\t"

  1700.         : : "g" ((x86_reg)-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1701.         : "%"REG_a

  1702.     );

  1703.     assert(src1 == src2);

  1704. }

  1705. 

  1706. static inline void RENAME(BEToUV)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1707. {

  1708.     __asm__ volatile(

  1709.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1710.         "mov                    %0, %%"REG_a"       \n\t"

  1711.         "1:                                         \n\t"

  1712.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1713.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1714.         "movq    (%2, %%"REG_a",2), %%mm2           \n\t"

  1715.         "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"

  1716.         "pand                %%mm4, %%mm0           \n\t"

  1717.         "pand                %%mm4, %%mm1           \n\t"

  1718.         "pand                %%mm4, %%mm2           \n\t"

  1719.         "pand                %%mm4, %%mm3           \n\t"

  1720.         "packuswb            %%mm1, %%mm0           \n\t"

  1721.         "packuswb            %%mm3, %%mm2           \n\t"

  1722.         "movq                %%mm0, (%3, %%"REG_a") \n\t"

  1723.         "movq                %%mm2, (%4, %%"REG_a") \n\t"

  1724.         "add                    $8, %%"REG_a"       \n\t"

  1725.         " js                    1b                  \n\t"

  1726.         : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)

  1727.         : "%"REG_a

  1728.     );

  1729. }

  1730. 

  1731. static inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,

  1732.                                     const uint8_t *src, int width)

  1733. {

  1734.     __asm__ volatile(

  1735.         "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1736.         "mov                    %0, %%"REG_a"       \n\t"

  1737.         "1:                                         \n\t"

  1738.         "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1739.         "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1740.         "movq                %%mm0, %%mm2           \n\t"

  1741.         "movq                %%mm1, %%mm3           \n\t"

  1742.         "pand                %%mm4, %%mm0           \n\t"

  1743.         "pand                %%mm4, %%mm1           \n\t"

  1744.         "psrlw                  $8, %%mm2           \n\t"

  1745.         "psrlw                  $8, %%mm3           \n\t"

  1746.         "packuswb            %%mm1, %%mm0           \n\t"

  1747.         "packuswb            %%mm3, %%mm2           \n\t"

  1748.         "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1749.         "movq                %%mm2, (%3, %%"REG_a") \n\t"

  1750.         "add                    $8, %%"REG_a"       \n\t"

  1751.         " js                    1b                  \n\t"

  1752.         : : "g" ((x86_reg)-width), "r" (src+width*2), "r" (dst1+width), "r" (dst2+width)

  1753.         : "%"REG_a

  1754.     );

  1755. }

  1756. 

  1757. static inline void RENAME(nv12ToUV)(uint8_t *dstU, uint8_t *dstV,

  1758.                                     const uint8_t *src1, const uint8_t *src2,

  1759.                                     int width, uint32_t *unused)

  1760. {

  1761.     RENAME(nvXXtoUV)(dstU, dstV, src1, width);

  1762. }

  1763. 

  1764. static inline void RENAME(nv21ToUV)(uint8_t *dstU, uint8_t *dstV,

  1765.                                     const uint8_t *src1, const uint8_t *src2,

  1766.                                     int width, uint32_t *unused)

  1767. {

  1768.     RENAME(nvXXtoUV)(dstV, dstU, src1, width);

  1769. }

  1770. #endif /* !COMPILE_TEMPLATE_MMX2 */

  1771. 

  1772. static inline void RENAME(bgr24ToY_mmx)(int16_t *dst, const uint8_t *src, int width, enum PixelFormat srcFormat)

  1773. {

  1774. 

  1775.     if(srcFormat == PIX_FMT_BGR24) {

  1776.         __asm__ volatile(

  1777.             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"

  1778.             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"

  1779.             :

  1780.         );

  1781.     } else {

  1782.         __asm__ volatile(

  1783.             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"

  1784.             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"

  1785.             :

  1786.         );

  1787.     }

  1788. 

  1789.     __asm__ volatile(

  1790.         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"

  1791.         "mov                        %2, %%"REG_a"   \n\t"

  1792.         "pxor                    %%mm7, %%mm7       \n\t"

  1793.         "1:                                         \n\t"

  1794.         PREFETCH"               64(%0)              \n\t"

  1795.         "movd                     (%0), %%mm0       \n\t"

  1796.         "movd                    2(%0), %%mm1       \n\t"

  1797.         "movd                    6(%0), %%mm2       \n\t"

  1798.         "movd                    8(%0), %%mm3       \n\t"

  1799.         "add                       $12, %0          \n\t"

  1800.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1801.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1802.         "punpcklbw               %%mm7, %%mm2       \n\t"

  1803.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1804.         "pmaddwd                 %%mm5, %%mm0       \n\t"

  1805.         "pmaddwd                 %%mm6, %%mm1       \n\t"

  1806.         "pmaddwd                 %%mm5, %%mm2       \n\t"

  1807.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1808.         "paddd                   %%mm1, %%mm0       \n\t"

  1809.         "paddd                   %%mm3, %%mm2       \n\t"

  1810.         "paddd                   %%mm4, %%mm0       \n\t"

  1811.         "paddd                   %%mm4, %%mm2       \n\t"

  1812.         "psrad                     $9, %%mm0       \n\t"

  1813.         "psrad                     $9, %%mm2       \n\t"

  1814.         "packssdw                %%mm2, %%mm0       \n\t"

  1815.         "movq                %%mm0, (%1, %%"REG_a") \n\t"

  1816.         "add                        $8, %%"REG_a"   \n\t"

  1817.         " js                        1b              \n\t"

  1818.     : "+r" (src)

  1819.     : "r" (dst+width), "g" ((x86_reg)-2*width)

  1820.     : "%"REG_a

  1821.     );

  1822. }

  1823. 

  1824. static inline void RENAME(bgr24ToUV_mmx)(int16_t *dstU, int16_t *dstV, const uint8_t *src, int width, enum PixelFormat srcFormat)

  1825. {

  1826.     __asm__ volatile(

  1827.         "movq                    24(%4), %%mm6       \n\t"

  1828.         "mov                        %3, %%"REG_a"   \n\t"

  1829.         "pxor                    %%mm7, %%mm7       \n\t"

  1830.         "1:                                         \n\t"

  1831.         PREFETCH"               64(%0)              \n\t"

  1832.         "movd                     (%0), %%mm0       \n\t"

  1833.         "movd                    2(%0), %%mm1       \n\t"

  1834.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1835.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1836.         "movq                    %%mm0, %%mm2       \n\t"

  1837.         "movq                    %%mm1, %%mm3       \n\t"

  1838.         "pmaddwd                  (%4), %%mm0       \n\t"

  1839.         "pmaddwd                 8(%4), %%mm1       \n\t"

  1840.         "pmaddwd                16(%4), %%mm2       \n\t"

  1841.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1842.         "paddd                   %%mm1, %%mm0       \n\t"

  1843.         "paddd                   %%mm3, %%mm2       \n\t"

  1844. 

  1845.         "movd                    6(%0), %%mm1       \n\t"

  1846.         "movd                    8(%0), %%mm3       \n\t"

  1847.         "add                       $12, %0          \n\t"

  1848.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1849.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1850.         "movq                    %%mm1, %%mm4       \n\t"

  1851.         "movq                    %%mm3, %%mm5       \n\t"

  1852.         "pmaddwd                  (%4), %%mm1       \n\t"

  1853.         "pmaddwd                 8(%4), %%mm3       \n\t"

  1854.         "pmaddwd                16(%4), %%mm4       \n\t"

  1855.         "pmaddwd                 %%mm6, %%mm5       \n\t"

  1856.         "paddd                   %%mm3, %%mm1       \n\t"

  1857.         "paddd                   %%mm5, %%mm4       \n\t"

  1858. 

  1859.         "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3       \n\t"

  1860.         "paddd                   %%mm3, %%mm0       \n\t"

  1861.         "paddd                   %%mm3, %%mm2       \n\t"

  1862.         "paddd                   %%mm3, %%mm1       \n\t"

  1863.         "paddd                   %%mm3, %%mm4       \n\t"

  1864.         "psrad                     $9, %%mm0       \n\t"

  1865.         "psrad                     $9, %%mm2       \n\t"

  1866.         "psrad                     $9, %%mm1       \n\t"

  1867.         "psrad                     $9, %%mm4       \n\t"

  1868.         "packssdw                %%mm1, %%mm0       \n\t"

  1869.         "packssdw                %%mm4, %%mm2       \n\t"

  1870.         "movq                %%mm0, (%1, %%"REG_a") \n\t"

  1871.         "movq                %%mm2, (%2, %%"REG_a") \n\t"

  1872.         "add                        $8, %%"REG_a"   \n\t"

  1873.         " js                        1b              \n\t"

  1874.     : "+r" (src)

  1875.     : "r" (dstU+width), "r" (dstV+width), "g" ((x86_reg)-2*width), "r"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24])

  1876.     : "%"REG_a

  1877.     );

  1878. }

  1879. 

  1880. static inline void RENAME(bgr24ToY)(int16_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1881. {

  1882.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);

  1883. }

  1884. 

  1885. static inline void RENAME(bgr24ToUV)(int16_t *dstU, int16_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1886. {

  1887.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);

  1888.     assert(src1 == src2);

  1889. }

  1890. 

  1891. static inline void RENAME(rgb24ToY)(int16_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1892. {

  1893.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);

  1894. }

  1895. 

  1896. static inline void RENAME(rgb24ToUV)(int16_t *dstU, int16_t *dstV, const uint8_t *src1, const uint8_t *src2, int width, uint32_t *unused)

  1897. {

  1898.     assert(src1==src2);

  1899.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);

  1900. }

  1901. 

  1902. #if !COMPILE_TEMPLATE_MMX2

  1903. // bilinear / bicubic scaling

  1904. static inline void RENAME(hScale)(int16_t *dst, int dstW, const uint8_t *src, int srcW, int xInc,

  1905.                                   const int16_t *filter, const int16_t *filterPos, int filterSize)

  1906. {

  1907.     assert(filterSize % 4 == 0 && filterSize>0);

  1908.     if (filterSize==4) { // Always true for upscaling, sometimes for down, too.

  1909.         x86_reg counter= -2*dstW;

  1910.         filter-= counter*2;

  1911.         filterPos-= counter/2;

  1912.         dst-= counter/2;

  1913.         __asm__ volatile(

  1914. #if defined(PIC)

  1915.             "push            %%"REG_b"              \n\t"

  1916. #endif

  1917.             "pxor                %%mm7, %%mm7       \n\t"

  1918.             "push           %%"REG_BP"              \n\t" // we use 7 regs here ...

  1919.             "mov             %%"REG_a", %%"REG_BP"  \n\t"

  1920.             ".p2align                4              \n\t"

  1921.             "1:                                     \n\t"

  1922.             "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"

  1923.             "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"

  1924.             "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"

  1925.             "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"

  1926.             "movd      (%3, %%"REG_a"), %%mm0       \n\t"

  1927.             "movd      (%3, %%"REG_b"), %%mm2       \n\t"

  1928.             "punpcklbw           %%mm7, %%mm0       \n\t"

  1929.             "punpcklbw           %%mm7, %%mm2       \n\t"

  1930.             "pmaddwd             %%mm1, %%mm0       \n\t"

  1931.             "pmaddwd             %%mm2, %%mm3       \n\t"

  1932.             "movq                %%mm0, %%mm4       \n\t"

  1933.             "punpckldq           %%mm3, %%mm0       \n\t"

  1934.             "punpckhdq           %%mm3, %%mm4       \n\t"

  1935.             "paddd               %%mm4, %%mm0       \n\t"

  1936.             "psrad                  $7, %%mm0       \n\t"

  1937.             "packssdw            %%mm0, %%mm0       \n\t"

  1938.             "movd                %%mm0, (%4, %%"REG_BP")    \n\t"

  1939.             "add                    $4, %%"REG_BP"  \n\t"

  1940.             " jnc                   1b              \n\t"

  1941. 

  1942.             "pop            %%"REG_BP"              \n\t"

  1943. #if defined(PIC)

  1944.             "pop             %%"REG_b"              \n\t"

  1945. #endif

  1946.             : "+a" (counter)

  1947.             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)

  1948. #if !defined(PIC)

  1949.             : "%"REG_b

  1950. #endif

  1951.         );

  1952.     } else if (filterSize==8) {

  1953.         x86_reg counter= -2*dstW;

  1954.         filter-= counter*4;

  1955.         filterPos-= counter/2;

  1956.         dst-= counter/2;

  1957.         __asm__ volatile(

  1958. #if defined(PIC)

  1959.             "push             %%"REG_b"             \n\t"

  1960. #endif

  1961.             "pxor                 %%mm7, %%mm7      \n\t"

  1962.             "push            %%"REG_BP"             \n\t" // we use 7 regs here ...

  1963.             "mov              %%"REG_a", %%"REG_BP" \n\t"

  1964.             ".p2align                 4             \n\t"

  1965.             "1:                                     \n\t"

  1966.             "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"

  1967.             "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"

  1968.             "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"

  1969.             "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"

  1970.             "movd       (%3, %%"REG_a"), %%mm0      \n\t"

  1971.             "movd       (%3, %%"REG_b"), %%mm2      \n\t"

  1972.             "punpcklbw            %%mm7, %%mm0      \n\t"

  1973.             "punpcklbw            %%mm7, %%mm2      \n\t"

  1974.             "pmaddwd              %%mm1, %%mm0      \n\t"

  1975.             "pmaddwd              %%mm2, %%mm3      \n\t"

  1976. 

  1977.             "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"

  1978.             "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"

  1979.             "movd      4(%3, %%"REG_a"), %%mm4      \n\t"

  1980.             "movd      4(%3, %%"REG_b"), %%mm2      \n\t"

  1981.             "punpcklbw            %%mm7, %%mm4      \n\t"

  1982.             "punpcklbw            %%mm7, %%mm2      \n\t"

  1983.             "pmaddwd              %%mm1, %%mm4      \n\t"

  1984.             "pmaddwd              %%mm2, %%mm5      \n\t"

  1985.             "paddd                %%mm4, %%mm0      \n\t"

  1986.             "paddd                %%mm5, %%mm3      \n\t"

  1987.             "movq                 %%mm0, %%mm4      \n\t"

  1988.             "punpckldq            %%mm3, %%mm0      \n\t"

  1989.             "punpckhdq            %%mm3, %%mm4      \n\t"

  1990.             "paddd                %%mm4, %%mm0      \n\t"

  1991.             "psrad                   $7, %%mm0      \n\t"

  1992.             "packssdw             %%mm0, %%mm0      \n\t"

  1993.             "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"

  1994.             "add                     $4, %%"REG_BP" \n\t"

  1995.             " jnc                    1b             \n\t"

  1996. 

  1997.             "pop             %%"REG_BP"             \n\t"

  1998. #if defined(PIC)

  1999.             "pop              %%"REG_b"             \n\t"

  2000. #endif

  2001.             : "+a" (counter)

  2002.             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)

  2003. #if !defined(PIC)

  2004.             : "%"REG_b

  2005. #endif

  2006.         );

  2007.     } else {

  2008.         const uint8_t *offset = src+filterSize;

  2009.         x86_reg counter= -2*dstW;

  2010.         //filter-= counter*filterSize/2;

  2011.         filterPos-= counter/2;

  2012.         dst-= counter/2;

  2013.         __asm__ volatile(

  2014.             "pxor                  %%mm7, %%mm7     \n\t"

  2015.             ".p2align                  4            \n\t"

  2016.             "1:                                     \n\t"

  2017.             "mov                      %2, %%"REG_c" \n\t"

  2018.             "movzwl      (%%"REG_c", %0), %%eax     \n\t"

  2019.             "movzwl     2(%%"REG_c", %0), %%edx     \n\t"

  2020.             "mov                      %5, %%"REG_c" \n\t"

  2021.             "pxor                  %%mm4, %%mm4     \n\t"

  2022.             "pxor                  %%mm5, %%mm5     \n\t"

  2023.             "2:                                     \n\t"

  2024.             "movq                   (%1), %%mm1     \n\t"

  2025.             "movq               (%1, %6), %%mm3     \n\t"

  2026.             "movd (%%"REG_c", %%"REG_a"), %%mm0     \n\t"

  2027.             "movd (%%"REG_c", %%"REG_d"), %%mm2     \n\t"

  2028.             "punpcklbw             %%mm7, %%mm0     \n\t"

  2029.             "punpcklbw             %%mm7, %%mm2     \n\t"

  2030.             "pmaddwd               %%mm1, %%mm0     \n\t"

  2031.             "pmaddwd               %%mm2, %%mm3     \n\t"

  2032.             "paddd                 %%mm3, %%mm5     \n\t"

  2033.             "paddd                 %%mm0, %%mm4     \n\t"

  2034.             "add                      $8, %1        \n\t"

  2035.             "add                      $4, %%"REG_c" \n\t"

  2036.             "cmp                      %4, %%"REG_c" \n\t"

  2037.             " jb                      2b            \n\t"

  2038.             "add                      %6, %1        \n\t"

  2039.             "movq                  %%mm4, %%mm0     \n\t"

  2040.             "punpckldq             %%mm5, %%mm4     \n\t"

  2041.             "punpckhdq             %%mm5, %%mm0     \n\t"

  2042.             "paddd                 %%mm0, %%mm4     \n\t"

  2043.             "psrad                    $7, %%mm4     \n\t"

  2044.             "packssdw              %%mm4, %%mm4     \n\t"

  2045.             "mov                      %3, %%"REG_a" \n\t"

  2046.             "movd                  %%mm4, (%%"REG_a", %0)   \n\t"

  2047.             "add                      $4, %0        \n\t"

  2048.             " jnc                     1b            \n\t"

  2049. 

  2050.             : "+r" (counter), "+r" (filter)

  2051.             : "m" (filterPos), "m" (dst), "m"(offset),

  2052.             "m" (src), "r" ((x86_reg)filterSize*2)

  2053.             : "%"REG_a, "%"REG_c, "%"REG_d

  2054.         );

  2055.     }

  2056. }

  2057. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2058. 

  2059. static inline void RENAME(hScale16)(int16_t *dst, int dstW, const uint16_t *src, int srcW, int xInc,

  2060.                                     const int16_t *filter, const int16_t *filterPos, long filterSize, int shift)

  2061. {

  2062.     int i, j;

  2063. 

  2064.     assert(filterSize % 4 == 0 && filterSize>0);

  2065.     if (filterSize==4 && shift<15) { // Always true for upscaling, sometimes for down, too.

  2066.         x86_reg counter= -2*dstW;

  2067.         filter-= counter*2;

  2068.         filterPos-= counter/2;

  2069.         dst-= counter/2;

  2070.         __asm__ volatile(

  2071.             "movd                   %5, %%mm7       \n\t"

  2072. #if defined(PIC)

  2073.             "push            %%"REG_b"              \n\t"

  2074. #endif

  2075.             "push           %%"REG_BP"              \n\t" // we use 7 regs here ...

  2076.             "mov             %%"REG_a", %%"REG_BP"  \n\t"

  2077.             ".p2align                4              \n\t"

  2078.             "1:                                     \n\t"

  2079.             "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"

  2080.             "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"

  2081.             "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"

  2082.             "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"

  2083.             "movq      (%3, %%"REG_a", 2), %%mm0    \n\t"

  2084.             "movq      (%3, %%"REG_b", 2), %%mm2    \n\t"

  2085.             "pmaddwd             %%mm1, %%mm0       \n\t"

  2086.             "pmaddwd             %%mm2, %%mm3       \n\t"

  2087.             "movq                %%mm0, %%mm4       \n\t"

  2088.             "punpckldq           %%mm3, %%mm0       \n\t"

  2089.             "punpckhdq           %%mm3, %%mm4       \n\t"

  2090.             "paddd               %%mm4, %%mm0       \n\t"

  2091.             "psrad               %%mm7, %%mm0       \n\t"

  2092.             "packssdw            %%mm0, %%mm0       \n\t"

  2093.             "movd                %%mm0, (%4, %%"REG_BP")    \n\t"

  2094.             "add                    $4, %%"REG_BP"  \n\t"

  2095.             " jnc                   1b              \n\t"

  2096. 

  2097.             "pop            %%"REG_BP"              \n\t"

  2098. #if defined(PIC)

  2099.             "pop             %%"REG_b"              \n\t"

  2100. #endif

  2101.             : "+a" (counter)

  2102.             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst), "m"(shift)

  2103. #if !defined(PIC)

  2104.             : "%"REG_b

  2105. #endif

  2106.         );

  2107.     } else if (filterSize==8 && shift<15) {

  2108.         x86_reg counter= -2*dstW;

  2109.         filter-= counter*4;

  2110.         filterPos-= counter/2;

  2111.         dst-= counter/2;

  2112.         __asm__ volatile(

  2113.             "movd                   %5, %%mm7       \n\t"

  2114. #if defined(PIC)

  2115.             "push            %%"REG_b"              \n\t"

  2116. #endif

  2117.             "push            %%"REG_BP"             \n\t" // we use 7 regs here ...

  2118.             "mov              %%"REG_a", %%"REG_BP" \n\t"

  2119.             ".p2align                 4             \n\t"

  2120.             "1:                                     \n\t"

  2121.             "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"

  2122.             "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"

  2123.             "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"

  2124.             "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"

  2125.             "movq       (%3, %%"REG_a", 2), %%mm0   \n\t"

  2126.             "movq       (%3, %%"REG_b", 2), %%mm2   \n\t"

  2127.             "pmaddwd              %%mm1, %%mm0      \n\t"

  2128.             "pmaddwd              %%mm2, %%mm3      \n\t"

  2129. 

  2130.             "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"

  2131.             "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"

  2132.             "movq      8(%3, %%"REG_a", 2), %%mm4   \n\t"

  2133.             "movq      8(%3, %%"REG_b", 2), %%mm2   \n\t"

  2134.             "pmaddwd              %%mm1, %%mm4      \n\t"

  2135.             "pmaddwd              %%mm2, %%mm5      \n\t"

  2136.             "paddd                %%mm4, %%mm0      \n\t"

  2137.             "paddd                %%mm5, %%mm3      \n\t"

  2138.             "movq                 %%mm0, %%mm4      \n\t"

  2139.             "punpckldq            %%mm3, %%mm0      \n\t"

  2140.             "punpckhdq            %%mm3, %%mm4      \n\t"

  2141.             "paddd                %%mm4, %%mm0      \n\t"

  2142.             "psrad                %%mm7, %%mm0      \n\t"

  2143.             "packssdw             %%mm0, %%mm0      \n\t"

  2144.             "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"

  2145.             "add                     $4, %%"REG_BP" \n\t"

  2146.             " jnc                    1b             \n\t"

  2147. 

  2148.             "pop             %%"REG_BP"             \n\t"

  2149. #if defined(PIC)

  2150.             "pop             %%"REG_b"              \n\t"

  2151. #endif

  2152.             : "+a" (counter)

  2153.             : "c" (filter), "d" (filterPos), "S" (src), "D" (dst), "m"(shift)

  2154. #if !defined(PIC)

  2155.             : "%"REG_b

  2156. #endif

  2157.         );

  2158.     } else if (shift<15){

  2159.         const uint16_t *offset = src+filterSize;

  2160.         x86_reg counter= -2*dstW;

  2161.         //filter-= counter*filterSize/2;

  2162.         filterPos-= counter/2;

  2163.         dst-= counter/2;

  2164.         __asm__ volatile(

  2165.             "movd                   %7, %%mm7       \n\t"

  2166.             ".p2align                  4            \n\t"

  2167.             "1:                                     \n\t"

  2168.             "mov                      %2, %%"REG_c" \n\t"

  2169.             "movzwl      (%%"REG_c", %0), %%eax     \n\t"

  2170.             "movzwl     2(%%"REG_c", %0), %%edx     \n\t"

  2171.             "mov                      %5, %%"REG_c" \n\t"

  2172.             "pxor                  %%mm4, %%mm4     \n\t"

  2173.             "pxor                  %%mm5, %%mm5     \n\t"

  2174.             "2:                                     \n\t"

  2175.             "movq                   (%1), %%mm1     \n\t"

  2176.             "movq               (%1, %6), %%mm3     \n\t"

  2177.             "movq (%%"REG_c", %%"REG_a", 2), %%mm0     \n\t"

  2178.             "movq (%%"REG_c", %%"REG_d", 2), %%mm2     \n\t"

  2179.             "pmaddwd               %%mm1, %%mm0     \n\t"

  2180.             "pmaddwd               %%mm2, %%mm3     \n\t"

  2181.             "paddd                 %%mm3, %%mm5     \n\t"

  2182.             "paddd                 %%mm0, %%mm4     \n\t"

  2183.             "add                      $8, %1        \n\t"

  2184.             "add                      $8, %%"REG_c" \n\t"

  2185.             "cmp                      %4, %%"REG_c" \n\t"

  2186.             " jb                      2b            \n\t"

  2187.             "add                      %6, %1        \n\t"

  2188.             "movq                  %%mm4, %%mm0     \n\t"

  2189.             "punpckldq             %%mm5, %%mm4     \n\t"

  2190.             "punpckhdq             %%mm5, %%mm0     \n\t"

  2191.             "paddd                 %%mm0, %%mm4     \n\t"

  2192.             "psrad                 %%mm7, %%mm4     \n\t"

  2193.             "packssdw              %%mm4, %%mm4     \n\t"

  2194.             "mov                      %3, %%"REG_a" \n\t"

  2195.             "movd                  %%mm4, (%%"REG_a", %0)   \n\t"

  2196.             "add                      $4, %0        \n\t"

  2197.             " jnc                     1b            \n\t"

  2198. 

  2199.             : "+r" (counter), "+r" (filter)

  2200.             : "m" (filterPos), "m" (dst), "m"(offset),

  2201.             "m" (src), "r" ((x86_reg)filterSize*2), "m"(shift)

  2202.             : "%"REG_a, "%"REG_c, "%"REG_d

  2203.         );

  2204.     } else

  2205.     for (i=0; i<dstW; i++) {

  2206.         int srcPos= filterPos[i];

  2207.         int val=0;

  2208.         for (j=0; j<filterSize; j++) {

  2209.             val += ((int)src[srcPos + j])*filter[filterSize*i + j];

  2210.         }

  2211.         dst[i] = FFMIN(val>>shift, (1<<15)-1); // the cubic equation does overflow ...

  2212.     }

  2213. }

  2214. 

  2215. 

  2216. #if COMPILE_TEMPLATE_MMX2

  2217. static inline void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,

  2218.                                         int dstWidth, const uint8_t *src, int srcW,

  2219.                                         int xInc)

  2220. {

  2221.     int32_t *filterPos = c->hLumFilterPos;

  2222.     int16_t *filter    = c->hLumFilter;

  2223.     void    *mmx2FilterCode= c->lumMmx2FilterCode;

  2224.     int i;

  2225. #if defined(PIC)

  2226.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  2227. #endif

  2228. 

  2229.     __asm__ volatile(

  2230. #if defined(PIC)

  2231.         "mov               %%"REG_b", %5        \n\t"

  2232. #endif

  2233.         "pxor                  %%mm7, %%mm7     \n\t"

  2234.         "mov                      %0, %%"REG_c" \n\t"

  2235.         "mov                      %1, %%"REG_D" \n\t"

  2236.         "mov                      %2, %%"REG_d" \n\t"

  2237.         "mov                      %3, %%"REG_b" \n\t"

  2238.         "xor               %%"REG_a", %%"REG_a" \n\t" // i

  2239.         PREFETCH"        (%%"REG_c")            \n\t"

  2240.         PREFETCH"      32(%%"REG_c")            \n\t"

  2241.         PREFETCH"      64(%%"REG_c")            \n\t"

  2242. 

  2243. #if ARCH_X86_64

  2244. #define CALL_MMX2_FILTER_CODE \

  2245.         "movl            (%%"REG_b"), %%esi     \n\t"\

  2246.         "call                    *%4            \n\t"\

  2247.         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\

  2248.         "add               %%"REG_S", %%"REG_c" \n\t"\

  2249.         "add               %%"REG_a", %%"REG_D" \n\t"\

  2250.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  2251. 

  2252. #else

  2253. #define CALL_MMX2_FILTER_CODE \

  2254.         "movl (%%"REG_b"), %%esi        \n\t"\

  2255.         "call         *%4                       \n\t"\

  2256.         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  2257.         "add               %%"REG_a", %%"REG_D" \n\t"\

  2258.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  2259. 

  2260. #endif /* ARCH_X86_64 */

  2261. 

  2262.         CALL_MMX2_FILTER_CODE

  2263.         CALL_MMX2_FILTER_CODE

  2264.         CALL_MMX2_FILTER_CODE

  2265.         CALL_MMX2_FILTER_CODE

  2266.         CALL_MMX2_FILTER_CODE

  2267.         CALL_MMX2_FILTER_CODE

  2268.         CALL_MMX2_FILTER_CODE

  2269.         CALL_MMX2_FILTER_CODE

  2270. 

  2271. #if defined(PIC)

  2272.         "mov                      %5, %%"REG_b" \n\t"

  2273. #endif

  2274.         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),

  2275.            "m" (mmx2FilterCode)

  2276. #if defined(PIC)

  2277.           ,"m" (ebxsave)

  2278. #endif

  2279.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2280. #if !defined(PIC)

  2281.          ,"%"REG_b

  2282. #endif

  2283.     );

  2284. 

  2285.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)

  2286.         dst[i] = src[srcW-1]*128;

  2287. }

  2288. 

  2289. static inline void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,

  2290.                                         int dstWidth, const uint8_t *src1,

  2291.                                         const uint8_t *src2, int srcW, int xInc)

  2292. {

  2293.     int32_t *filterPos = c->hChrFilterPos;

  2294.     int16_t *filter    = c->hChrFilter;

  2295.     void    *mmx2FilterCode= c->chrMmx2FilterCode;

  2296.     int i;

  2297. #if defined(PIC)

  2298.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  2299. #endif

  2300. 

  2301.     __asm__ volatile(

  2302. #if defined(PIC)

  2303.         "mov          %%"REG_b", %7         \n\t"

  2304. #endif

  2305.         "pxor             %%mm7, %%mm7      \n\t"

  2306.         "mov                 %0, %%"REG_c"  \n\t"

  2307.         "mov                 %1, %%"REG_D"  \n\t"

  2308.         "mov                 %2, %%"REG_d"  \n\t"

  2309.         "mov                 %3, %%"REG_b"  \n\t"

  2310.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2311.         PREFETCH"   (%%"REG_c")             \n\t"

  2312.         PREFETCH" 32(%%"REG_c")             \n\t"

  2313.         PREFETCH" 64(%%"REG_c")             \n\t"

  2314. 

  2315.         CALL_MMX2_FILTER_CODE

  2316.         CALL_MMX2_FILTER_CODE

  2317.         CALL_MMX2_FILTER_CODE

  2318.         CALL_MMX2_FILTER_CODE

  2319.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2320.         "mov                 %5, %%"REG_c"  \n\t" // src

  2321.         "mov                 %6, %%"REG_D"  \n\t" // buf2

  2322.         PREFETCH"   (%%"REG_c")             \n\t"

  2323.         PREFETCH" 32(%%"REG_c")             \n\t"

  2324.         PREFETCH" 64(%%"REG_c")             \n\t"

  2325. 

  2326.         CALL_MMX2_FILTER_CODE

  2327.         CALL_MMX2_FILTER_CODE

  2328.         CALL_MMX2_FILTER_CODE

  2329.         CALL_MMX2_FILTER_CODE

  2330. 

  2331. #if defined(PIC)

  2332.         "mov %7, %%"REG_b"    \n\t"

  2333. #endif

  2334.         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),

  2335.            "m" (mmx2FilterCode), "m" (src2), "m"(dst2)

  2336. #if defined(PIC)

  2337.           ,"m" (ebxsave)

  2338. #endif

  2339.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2340. #if !defined(PIC)

  2341.          ,"%"REG_b

  2342. #endif

  2343.     );

  2344. 

  2345.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {

  2346.         dst1[i] = src1[srcW-1]*128;

  2347.         dst2[i] = src2[srcW-1]*128;

  2348.     }

  2349. }

  2350. #endif /* COMPILE_TEMPLATE_MMX2 */

  2351. 

  2352. static void RENAME(sws_init_swScale)(SwsContext *c)

  2353. {

  2354.     enum PixelFormat srcFormat = c->srcFormat;

  2355. 

  2356.     if (!(c->flags & SWS_BITEXACT)) {

  2357.         if (c->flags & SWS_ACCURATE_RND) {

  2358.             c->yuv2yuv1     = RENAME(yuv2yuv1_ar    );

  2359.             c->yuv2yuvX     = RENAME(yuv2yuvX_ar    );

  2360.             switch (c->dstFormat) {

  2361.             case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;

  2362.             case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;

  2363.             case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;

  2364.             case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;

  2365.             case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;

  2366.             default: break;

  2367.             }

  2368.         } else {

  2369.             c->yuv2yuv1     = RENAME(yuv2yuv1    );

  2370.             c->yuv2yuvX     = RENAME(yuv2yuvX    );

  2371.             switch (c->dstFormat) {

  2372.             case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;

  2373.             case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;

  2374.             case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;

  2375.             case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;

  2376.             case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;

  2377.             default: break;

  2378.             }

  2379.         }

  2380.         switch (c->dstFormat) {

  2381.         case PIX_FMT_RGB32:

  2382.                 c->yuv2packed1 = RENAME(yuv2rgb32_1);

  2383.                 c->yuv2packed2 = RENAME(yuv2rgb32_2);

  2384.                 break;

  2385.         case PIX_FMT_BGR24:

  2386.                 c->yuv2packed1 = RENAME(yuv2bgr24_1);

  2387.                 c->yuv2packed2 = RENAME(yuv2bgr24_2);

  2388.                 break;

  2389.         case PIX_FMT_RGB555:

  2390.                 c->yuv2packed1 = RENAME(yuv2rgb555_1);

  2391.                 c->yuv2packed2 = RENAME(yuv2rgb555_2);

  2392.                 break;

  2393.         case PIX_FMT_RGB565:

  2394.                 c->yuv2packed1 = RENAME(yuv2rgb565_1);

  2395.                 c->yuv2packed2 = RENAME(yuv2rgb565_2);

  2396.                 break;

  2397.         case PIX_FMT_YUYV422:

  2398.                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);

  2399.                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);

  2400.                 break;

  2401.         default:

  2402.                 break;

  2403.         }

  2404.     }

  2405. 

  2406. #if !COMPILE_TEMPLATE_MMX2

  2407.     c->hScale       = RENAME(hScale      );

  2408. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2409. 

  2410.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  2411. #if COMPILE_TEMPLATE_MMX2

  2412.     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)

  2413.     {

  2414.         c->hyscale_fast = RENAME(hyscale_fast);

  2415.         c->hcscale_fast = RENAME(hcscale_fast);

  2416.     } else {

  2417. #endif /* COMPILE_TEMPLATE_MMX2 */

  2418.         c->hyscale_fast = NULL;

  2419.         c->hcscale_fast = NULL;

  2420. #if COMPILE_TEMPLATE_MMX2

  2421.     }

  2422. #endif /* COMPILE_TEMPLATE_MMX2 */

  2423. 

  2424. #if !COMPILE_TEMPLATE_MMX2

  2425.     switch(srcFormat) {

  2426.         case PIX_FMT_YUYV422  : c->chrToYV12 = RENAME(yuy2ToUV); break;

  2427.         case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;

  2428.         case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;

  2429.         case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;

  2430.         case PIX_FMT_GRAY16LE :

  2431.         case PIX_FMT_YUV420P9LE:

  2432.         case PIX_FMT_YUV422P10LE:

  2433.         case PIX_FMT_YUV420P10LE:

  2434.         case PIX_FMT_YUV420P16LE:

  2435.         case PIX_FMT_YUV422P16LE:

  2436.         case PIX_FMT_YUV444P16LE: c->hScale16= RENAME(hScale16); break;

  2437.     }

  2438. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2439.     if (!c->chrSrcHSubSample) {

  2440.         switch(srcFormat) {

  2441.         case PIX_FMT_BGR24  : c->chrToYV12 = RENAME(bgr24ToUV); break;

  2442.         case PIX_FMT_RGB24  : c->chrToYV12 = RENAME(rgb24ToUV); break;

  2443.         default: break;

  2444.         }

  2445.     }

  2446. 

  2447.     switch (srcFormat) {

  2448. #if !COMPILE_TEMPLATE_MMX2

  2449.     case PIX_FMT_YUYV422  :

  2450.     case PIX_FMT_Y400A    :

  2451.                             c->lumToYV12 = RENAME(yuy2ToY); break;

  2452.     case PIX_FMT_UYVY422  :

  2453.                             c->lumToYV12 = RENAME(uyvyToY); break;

  2454. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2455.     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;

  2456.     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;

  2457.     default: break;

  2458.     }

  2459. #if !COMPILE_TEMPLATE_MMX2

  2460.     if (c->alpPixBuf) {

  2461.         switch (srcFormat) {

  2462.         case PIX_FMT_Y400A  : c->alpToYV12 = RENAME(yuy2ToY); break;

  2463.         default: break;

  2464.         }

  2465.     }

  2466. #endif /* !COMPILE_TEMPLATE_MMX2 */

  2467.     if(isAnyRGB(c->srcFormat))

  2468.         c->hScale16= RENAME(hScale16);

  2469. }