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.
34093 Commits
32 Branches
320MB
Tree: 5ed032fb8b
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '5ed032fb8b'
${ noResults }
FFmpeg/libswscale/x86/swscale_template.c

1633 lines
71KB
Raw Blame History 

  1. /*

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

  3.  *

  4.  * This file is part of Libav.

  5.  *

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

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

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

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

  10.  *

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

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

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

  14.  * Lesser General Public License for more details.

  15.  *

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

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

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

  19.  */

  20. 

  21. #undef REAL_MOVNTQ

  22. #undef MOVNTQ

  23. #undef PREFETCH

  24. 

  25. #if COMPILE_TEMPLATE_MMX2

  26. #define PREFETCH "prefetchnta"

  27. #else

  28. #define PREFETCH  " # nop"

  29. #endif

  30. 

  31. #if COMPILE_TEMPLATE_MMX2

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

  33. #else

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

  35. #endif

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

  37. 

  38. #define YSCALEYUV2PACKEDX_UV \

  39.     __asm__ volatile(\

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

  41.         ".p2align                      4                \n\t"\

  42.         "nop                                            \n\t"\

  43.         "1:                                             \n\t"\

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

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

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

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

  48.         ".p2align                      4                \n\t"\

  49.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

  62. 

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

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

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

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

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

  68.     ".p2align                      4                \n\t"\

  69.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

  81. 

  82. #define YSCALEYUV2PACKEDX \

  83.     YSCALEYUV2PACKEDX_UV \

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

  85. 

  86. #define YSCALEYUV2PACKEDX_END                     \

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

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

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

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

  91.     );

  92. 

  93. #define YSCALEYUV2PACKEDX_ACCURATE_UV \

  94.     __asm__ volatile(\

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

  96.         ".p2align                      4                \n\t"\

  97.         "nop                                            \n\t"\

  98.         "1:                                             \n\t"\

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

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

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

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

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

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

  105.         ".p2align                      4                \n\t"\

  106.         "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  144. 

  145. #define YSCALEYUV2PACKEDX_ACCURATE_YA(offset) \

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

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

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

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

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

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

  152.     ".p2align                      4                \n\t"\

  153.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  189. 

  190. #define YSCALEYUV2PACKEDX_ACCURATE \

  191.     YSCALEYUV2PACKEDX_ACCURATE_UV \

  192.     YSCALEYUV2PACKEDX_ACCURATE_YA(LUM_MMX_FILTER_OFFSET)

  193. 

  194. #define YSCALEYUV2RGBX \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  229. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  243. \

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

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

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

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

  248. \

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

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

  251.     " jb      1b                \n\t"

  252. #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)

  253. 

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

  255.                                    const int16_t **lumSrc, int lumFilterSize,

  256.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  257.                                    const int16_t **chrVSrc,

  258.                                    int chrFilterSize, const int16_t **alpSrc,

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

  260. {

  261.     x86_reg dummy=0;

  262.     x86_reg dstW_reg = dstW;

  263.     x86_reg uv_off = c->uv_off_byte;

  264. 

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

  266.         YSCALEYUV2PACKEDX_ACCURATE

  267.         YSCALEYUV2RGBX

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

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

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

  271.         YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)

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

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

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

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

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

  277.         YSCALEYUV2PACKEDX_END

  278.     } else {

  279.         YSCALEYUV2PACKEDX_ACCURATE

  280.         YSCALEYUV2RGBX

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

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

  283.         YSCALEYUV2PACKEDX_END

  284.     }

  285. }

  286. 

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

  288.                                 const int16_t **lumSrc, int lumFilterSize,

  289.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  290.                                 const int16_t **chrVSrc,

  291.                                 int chrFilterSize, const int16_t **alpSrc,

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

  293. {

  294.     x86_reg dummy=0;

  295.     x86_reg dstW_reg = dstW;

  296.     x86_reg uv_off = c->uv_off_byte;

  297. 

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

  299.         YSCALEYUV2PACKEDX

  300.         YSCALEYUV2RGBX

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

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

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

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

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

  306.         YSCALEYUV2PACKEDX_END

  307.     } else {

  308.         YSCALEYUV2PACKEDX

  309.         YSCALEYUV2RGBX

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

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

  312.         YSCALEYUV2PACKEDX_END

  313.     }

  314. }

  315. 

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

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

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

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

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

  321. \

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

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

  324. \

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

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

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

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

  329. \

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

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

  332. \

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

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

  335. \

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

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

  338. \

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

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

  341.     " jb             1b             \n\t"

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

  343. 

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

  345.                                     const int16_t **lumSrc, int lumFilterSize,

  346.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  347.                                     const int16_t **chrVSrc,

  348.                                     int chrFilterSize, const int16_t **alpSrc,

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

  350. {

  351.     x86_reg dummy=0;

  352.     x86_reg dstW_reg = dstW;

  353.     x86_reg uv_off = c->uv_off_byte;

  354. 

  355.     YSCALEYUV2PACKEDX_ACCURATE

  356.     YSCALEYUV2RGBX

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

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

  359. #ifdef DITHER1XBPP

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

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

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

  363. #endif

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

  365.     YSCALEYUV2PACKEDX_END

  366. }

  367. 

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

  369.                                  const int16_t **lumSrc, int lumFilterSize,

  370.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  371.                                  const int16_t **chrVSrc,

  372.                                  int chrFilterSize, const int16_t **alpSrc,

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

  374. {

  375.     x86_reg dummy=0;

  376.     x86_reg dstW_reg = dstW;

  377.     x86_reg uv_off = c->uv_off_byte;

  378. 

  379.     YSCALEYUV2PACKEDX

  380.     YSCALEYUV2RGBX

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

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

  383. #ifdef DITHER1XBPP

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

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

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

  387. #endif

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

  389.     YSCALEYUV2PACKEDX_END

  390. }

  391. 

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

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

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

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

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

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

  398. \

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

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

  401. \

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

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

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

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

  406. \

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

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

  409. \

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

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

  412. \

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

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

  415. \

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

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

  418.     " jb             1b             \n\t"

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

  420. 

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

  422.                                     const int16_t **lumSrc, int lumFilterSize,

  423.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  424.                                     const int16_t **chrVSrc,

  425.                                     int chrFilterSize, const int16_t **alpSrc,

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

  427. {

  428.     x86_reg dummy=0;

  429.     x86_reg dstW_reg = dstW;

  430.     x86_reg uv_off = c->uv_off_byte;

  431. 

  432.     YSCALEYUV2PACKEDX_ACCURATE

  433.     YSCALEYUV2RGBX

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

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

  436. #ifdef DITHER1XBPP

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

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

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

  440. #endif

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

  442.     YSCALEYUV2PACKEDX_END

  443. }

  444. 

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

  446.                                  const int16_t **lumSrc, int lumFilterSize,

  447.                                  const int16_t *chrFilter, const int16_t **chrUSrc,

  448.                                  const int16_t **chrVSrc,

  449.                                  int chrFilterSize, const int16_t **alpSrc,

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

  451. {

  452.     x86_reg dummy=0;

  453.     x86_reg dstW_reg = dstW;

  454.     x86_reg uv_off = c->uv_off_byte;

  455. 

  456.     YSCALEYUV2PACKEDX

  457.     YSCALEYUV2RGBX

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

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

  460. #ifdef DITHER1XBPP

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

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

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

  464. #endif

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

  466.     YSCALEYUV2PACKEDX_END

  467. }

  468. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  483. \

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

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

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

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

  488. \

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

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

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

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

  493. \

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

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

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

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

  498. \

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

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

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

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

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

  504. \

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

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

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

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

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

  510. \

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

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

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

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

  515. \

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

  517. \

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

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

  520.     " jb          1b            \n\t"

  521. 

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

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

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

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

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

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

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

  529. \

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

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

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

  533. \

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

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

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

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

  538. \

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

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

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

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

  543. \

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

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

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

  547. \

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

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

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

  551. \

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

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

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

  555. \

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

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

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

  559. \

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

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

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

  563. \

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

  565. \

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

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

  568.     " jb       1b               \n\t"

  569. 

  570. #if COMPILE_TEMPLATE_MMX2

  571. #undef WRITEBGR24

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

  573. #else

  574. #undef WRITEBGR24

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

  576. #endif

  577. 

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

  579.                                    const int16_t **lumSrc, int lumFilterSize,

  580.                                    const int16_t *chrFilter, const int16_t **chrUSrc,

  581.                                    const int16_t **chrVSrc,

  582.                                    int chrFilterSize, const int16_t **alpSrc,

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

  584. {

  585.     x86_reg dummy=0;

  586.     x86_reg dstW_reg = dstW;

  587.     x86_reg uv_off = c->uv_off_byte;

  588. 

  589.     YSCALEYUV2PACKEDX_ACCURATE

  590.     YSCALEYUV2RGBX

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

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

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

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

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

  596.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  599.     );

  600. }

  601. 

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

  603.                                 const int16_t **lumSrc, int lumFilterSize,

  604.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  605.                                 const int16_t **chrVSrc,

  606.                                 int chrFilterSize, const int16_t **alpSrc,

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

  608. {

  609.     x86_reg dummy=0;

  610.     x86_reg dstW_reg = dstW;

  611.     x86_reg uv_off = c->uv_off_byte;

  612. 

  613.     YSCALEYUV2PACKEDX

  614.     YSCALEYUV2RGBX

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

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

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

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

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

  620.        "m" (dummy), "m" (dummy), "m" (dummy),

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

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

  623.     );

  624. }

  625. 

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

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

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

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

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

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

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

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

  634. \

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

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

  637. \

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

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

  640.     " jb          1b            \n\t"

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

  642. 

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

  644.                                      const int16_t **lumSrc, int lumFilterSize,

  645.                                      const int16_t *chrFilter, const int16_t **chrUSrc,

  646.                                      const int16_t **chrVSrc,

  647.                                      int chrFilterSize, const int16_t **alpSrc,

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

  649. {

  650.     x86_reg dummy=0;

  651.     x86_reg dstW_reg = dstW;

  652.     x86_reg uv_off = c->uv_off_byte;

  653. 

  654.     YSCALEYUV2PACKEDX_ACCURATE

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

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

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

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

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

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

  661.     YSCALEYUV2PACKEDX_END

  662. }

  663. 

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

  665.                                   const int16_t **lumSrc, int lumFilterSize,

  666.                                   const int16_t *chrFilter, const int16_t **chrUSrc,

  667.                                   const int16_t **chrVSrc,

  668.                                   int chrFilterSize, const int16_t **alpSrc,

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

  670. {

  671.     x86_reg dummy=0;

  672.     x86_reg dstW_reg = dstW;

  673.     x86_reg uv_off = c->uv_off_byte;

  674. 

  675.     YSCALEYUV2PACKEDX

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

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

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

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

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

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

  682.     YSCALEYUV2PACKEDX_END

  683. }

  684. 

  685. #define REAL_YSCALEYUV2RGB_UV(index, c) \

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

  687.     ".p2align              4            \n\t"\

  688.     "1:                                 \n\t"\

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

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

  691.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

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

  694.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  711. 

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

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

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

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

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

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

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

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

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

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

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

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

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

  725. 

  726. #define REAL_YSCALEYUV2RGB_COEFF(c) \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  754. 

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

  756. 

  757. #define YSCALEYUV2RGB(index, c) \

  758.     REAL_YSCALEYUV2RGB_UV(index, c) \

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

  760.     REAL_YSCALEYUV2RGB_COEFF(c)

  761. 

  762. /**

  763.  * vertical bilinear scale YV12 to RGB

  764.  */

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

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

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

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

  769. {

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

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

  772. 

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

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

  775. #if ARCH_X86_64

  776.         __asm__ volatile(

  777.             YSCALEYUV2RGB(%%r8, %5)

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

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

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

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

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

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

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

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

  786.             : "%r8"

  787.         );

  788. #else

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

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

  791.         __asm__ volatile(

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

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

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

  795.             YSCALEYUV2RGB(%%REGBP, %5)

  796.             "push                   %0              \n\t"

  797.             "push                   %1              \n\t"

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

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

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

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

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

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

  804.             "pop                    %1              \n\t"

  805.             "pop                    %0              \n\t"

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

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

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

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

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

  811.         );

  812. #endif

  813.     } else {

  814.         __asm__ volatile(

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

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

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

  818.             YSCALEYUV2RGB(%%REGBP, %5)

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

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

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

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

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

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

  825.         );

  826.     }

  827. }

  828. 

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

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

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

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

  833. {

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

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

  836. 

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

  838.     __asm__ volatile(

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

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

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

  842.         YSCALEYUV2RGB(%%REGBP, %5)

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

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

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

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

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

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

  849.     );

  850. }

  851. 

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

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

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

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

  856. {

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

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

  859. 

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

  861.     __asm__ volatile(

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

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

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

  865.         YSCALEYUV2RGB(%%REGBP, %5)

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

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

  868. #ifdef DITHER1XBPP

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

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

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

  872. #endif

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

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

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

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

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

  878.     );

  879. }

  880. 

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

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

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

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

  885. {

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

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

  888. 

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

  890.     __asm__ volatile(

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

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

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

  894.         YSCALEYUV2RGB(%%REGBP, %5)

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

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

  897. #ifdef DITHER1XBPP

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

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

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

  901. #endif

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

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

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

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

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

  907.     );

  908. }

  909. 

  910. #define REAL_YSCALEYUV2PACKED(index, c) \

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

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

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

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

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

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

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

  918.     ".p2align              4            \n\t"\

  919.     "1:                                 \n\t"\

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

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

  922.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

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

  925.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  947. 

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

  949. 

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

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

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

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

  954. {

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

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

  957. 

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

  959.     __asm__ volatile(

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

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

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

  963.         YSCALEYUV2PACKED(%%REGBP, %5)

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

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

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

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

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

  969.     );

  970. }

  971. 

  972. #define REAL_YSCALEYUV2RGB1(index, c) \

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

  974.     ".p2align              4            \n\t"\

  975.     "1:                                 \n\t"\

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

  977.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

  979.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  1000.     "paddw             %%mm3, %%mm4     \n\t"\

  1001.     "movq              %%mm2, %%mm0     \n\t"\

  1002.     "movq              %%mm5, %%mm6     \n\t"\

  1003.     "movq              %%mm4, %%mm3     \n\t"\

  1004.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1005.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1006.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1007.     "paddw             %%mm1, %%mm2     \n\t"\

  1008.     "paddw             %%mm1, %%mm5     \n\t"\

  1009.     "paddw             %%mm1, %%mm4     \n\t"\

  1010.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1011.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1012.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1013.     "paddw             %%mm7, %%mm0     \n\t"\

  1014.     "paddw             %%mm7, %%mm6     \n\t"\

  1015.     "paddw             %%mm7, %%mm3     \n\t"\

  1016.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1017.     "packuswb          %%mm0, %%mm2     \n\t"\

  1018.     "packuswb          %%mm6, %%mm5     \n\t"\

  1019.     "packuswb          %%mm3, %%mm4     \n\t"\

  1020. 

  1021. #define YSCALEYUV2RGB1(index, c)  REAL_YSCALEYUV2RGB1(index, c)

  1022. 

  1023. // do vertical chrominance interpolation

  1024. #define REAL_YSCALEYUV2RGB1b(index, c) \

  1025.     "xor            "#index", "#index"  \n\t"\

  1026.     ".p2align              4            \n\t"\

  1027.     "1:                                 \n\t"\

  1028.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1029.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1030.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1031.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1032.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1033.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1034.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1035.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1036.     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\

  1037.     "psrlw                $5, %%mm4     \n\t" /*FIXME might overflow*/\

  1038.     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\

  1039.     "psubw  "V_OFFSET"("#c"), %%mm4     \n\t" /* (V-128)8*/\

  1040.     "movq              %%mm3, %%mm2     \n\t" /* (U-128)8*/\

  1041.     "movq              %%mm4, %%mm5     \n\t" /* (V-128)8*/\

  1042.     "pmulhw "UG_COEFF"("#c"), %%mm3     \n\t"\

  1043.     "pmulhw "VG_COEFF"("#c"), %%mm4     \n\t"\

  1044.     /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\

  1045.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1046.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1047.     "psraw                $4, %%mm1     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1048.     "psraw                $4, %%mm7     \n\t" /* buf0[eax] - buf1[eax] >>4*/\

  1049.     "pmulhw "UB_COEFF"("#c"), %%mm2     \n\t"\

  1050.     "pmulhw "VR_COEFF"("#c"), %%mm5     \n\t"\

  1051.     "psubw  "Y_OFFSET"("#c"), %%mm1     \n\t" /* 8(Y-16)*/\

  1052.     "psubw  "Y_OFFSET"("#c"), %%mm7     \n\t" /* 8(Y-16)*/\

  1053.     "pmulhw  "Y_COEFF"("#c"), %%mm1     \n\t"\

  1054.     "pmulhw  "Y_COEFF"("#c"), %%mm7     \n\t"\

  1055.     /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\

  1056.     "paddw             %%mm3, %%mm4     \n\t"\

  1057.     "movq              %%mm2, %%mm0     \n\t"\

  1058.     "movq              %%mm5, %%mm6     \n\t"\

  1059.     "movq              %%mm4, %%mm3     \n\t"\

  1060.     "punpcklwd         %%mm2, %%mm2     \n\t"\

  1061.     "punpcklwd         %%mm5, %%mm5     \n\t"\

  1062.     "punpcklwd         %%mm4, %%mm4     \n\t"\

  1063.     "paddw             %%mm1, %%mm2     \n\t"\

  1064.     "paddw             %%mm1, %%mm5     \n\t"\

  1065.     "paddw             %%mm1, %%mm4     \n\t"\

  1066.     "punpckhwd         %%mm0, %%mm0     \n\t"\

  1067.     "punpckhwd         %%mm6, %%mm6     \n\t"\

  1068.     "punpckhwd         %%mm3, %%mm3     \n\t"\

  1069.     "paddw             %%mm7, %%mm0     \n\t"\

  1070.     "paddw             %%mm7, %%mm6     \n\t"\

  1071.     "paddw             %%mm7, %%mm3     \n\t"\

  1072.     /* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\

  1073.     "packuswb          %%mm0, %%mm2     \n\t"\

  1074.     "packuswb          %%mm6, %%mm5     \n\t"\

  1075.     "packuswb          %%mm3, %%mm4     \n\t"\

  1076. 

  1077. #define YSCALEYUV2RGB1b(index, c)  REAL_YSCALEYUV2RGB1b(index, c)

  1078. 

  1079. #define REAL_YSCALEYUV2RGB1_ALPHA(index) \

  1080.     "movq  (%1, "#index", 2), %%mm7     \n\t" /* abuf0[index  ]     */\

  1081.     "movq 8(%1, "#index", 2), %%mm1     \n\t" /* abuf0[index+4]     */\

  1082.     "psraw                $7, %%mm7     \n\t" /* abuf0[index  ] >>7 */\

  1083.     "psraw                $7, %%mm1     \n\t" /* abuf0[index+4] >>7 */\

  1084.     "packuswb          %%mm1, %%mm7     \n\t"

  1085. #define YSCALEYUV2RGB1_ALPHA(index) REAL_YSCALEYUV2RGB1_ALPHA(index)

  1086. 

  1087. /**

  1088.  * YV12 to RGB without scaling or interpolating

  1089.  */

  1090. static void RENAME(yuv2rgb32_1)(SwsContext *c, const int16_t *buf0,

  1091.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  1092.                                 const int16_t *abuf0, uint8_t *dest,

  1093.                                 int dstW, int uvalpha, int y)

  1094. {

  1095.     const int16_t *ubuf0 = ubuf[0];

  1096.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1097. 

  1098.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1099.         const int16_t *ubuf1 = ubuf[0];

  1100.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1101.             __asm__ volatile(

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

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

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

  1105.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1106.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1107.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1108.                 "pop %%"REG_BP"                         \n\t"

  1109.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1110.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1111.                    "a" (&c->redDither)

  1112.             );

  1113.         } else {

  1114.             __asm__ volatile(

  1115.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1116.                 "mov        %4, %%"REG_b"               \n\t"

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

  1118.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1119.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1120.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1121.                 "pop %%"REG_BP"                         \n\t"

  1122.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1123.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1124.                    "a" (&c->redDither)

  1125.             );

  1126.         }

  1127.     } else {

  1128.         const int16_t *ubuf1 = ubuf[1];

  1129.         if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1130.             __asm__ volatile(

  1131.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1132.                 "mov        %4, %%"REG_b"               \n\t"

  1133.                 "push %%"REG_BP"                        \n\t"

  1134.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1135.                 YSCALEYUV2RGB1_ALPHA(%%REGBP)

  1136.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1137.                 "pop %%"REG_BP"                         \n\t"

  1138.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1139.                 :: "c" (buf0), "d" (abuf0), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1140.                    "a" (&c->redDither)

  1141.             );

  1142.         } else {

  1143.             __asm__ volatile(

  1144.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1145.                 "mov        %4, %%"REG_b"               \n\t"

  1146.                 "push %%"REG_BP"                        \n\t"

  1147.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1148.                 "pcmpeqd %%mm7, %%mm7                   \n\t"

  1149.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP, %%mm2, %%mm4, %%mm5, %%mm7, %%mm0, %%mm1, %%mm3, %%mm6)

  1150.                 "pop %%"REG_BP"                         \n\t"

  1151.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1152.                 :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1153.                    "a" (&c->redDither)

  1154.             );

  1155.         }

  1156.     }

  1157. }

  1158. 

  1159. static void RENAME(yuv2bgr24_1)(SwsContext *c, const int16_t *buf0,

  1160.                                 const int16_t *ubuf[2], const int16_t *vbuf[2],

  1161.                                 const int16_t *abuf0, uint8_t *dest,

  1162.                                 int dstW, int uvalpha, int y)

  1163. {

  1164.     const int16_t *ubuf0 = ubuf[0];

  1165.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1166. 

  1167.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1168.         const int16_t *ubuf1 = ubuf[0];

  1169.         __asm__ volatile(

  1170.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1171.             "mov        %4, %%"REG_b"               \n\t"

  1172.             "push %%"REG_BP"                        \n\t"

  1173.             YSCALEYUV2RGB1(%%REGBP, %5)

  1174.             "pxor    %%mm7, %%mm7                   \n\t"

  1175.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1176.             "pop %%"REG_BP"                         \n\t"

  1177.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1178.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1179.                "a" (&c->redDither)

  1180.         );

  1181.     } else {

  1182.         const int16_t *ubuf1 = ubuf[1];

  1183.         __asm__ volatile(

  1184.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1185.             "mov        %4, %%"REG_b"               \n\t"

  1186.             "push %%"REG_BP"                        \n\t"

  1187.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1188.             "pxor    %%mm7, %%mm7                   \n\t"

  1189.             WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1190.             "pop %%"REG_BP"                         \n\t"

  1191.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1192.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1193.                "a" (&c->redDither)

  1194.         );

  1195.     }

  1196. }

  1197. 

  1198. static void RENAME(yuv2rgb555_1)(SwsContext *c, const int16_t *buf0,

  1199.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  1200.                                  const int16_t *abuf0, uint8_t *dest,

  1201.                                  int dstW, int uvalpha, int y)

  1202. {

  1203.     const int16_t *ubuf0 = ubuf[0];

  1204.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1205. 

  1206.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1207.         const int16_t *ubuf1 = ubuf[0];

  1208.         __asm__ volatile(

  1209.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1210.             "mov        %4, %%"REG_b"               \n\t"

  1211.             "push %%"REG_BP"                        \n\t"

  1212.             YSCALEYUV2RGB1(%%REGBP, %5)

  1213.             "pxor    %%mm7, %%mm7                   \n\t"

  1214.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1215. #ifdef DITHER1XBPP

  1216.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1217.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1218.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1219. #endif

  1220.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1221.             "pop %%"REG_BP"                         \n\t"

  1222.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1223.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1224.                "a" (&c->redDither)

  1225.         );

  1226.     } else {

  1227.         const int16_t *ubuf1 = ubuf[1];

  1228.         __asm__ volatile(

  1229.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1230.             "mov        %4, %%"REG_b"               \n\t"

  1231.             "push %%"REG_BP"                        \n\t"

  1232.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1233.             "pxor    %%mm7, %%mm7                   \n\t"

  1234.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1235. #ifdef DITHER1XBPP

  1236.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1237.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1238.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1239. #endif

  1240.             WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1241.             "pop %%"REG_BP"                         \n\t"

  1242.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1243.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1244.                "a" (&c->redDither)

  1245.         );

  1246.     }

  1247. }

  1248. 

  1249. static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,

  1250.                                  const int16_t *ubuf[2], const int16_t *vbuf[2],

  1251.                                  const int16_t *abuf0, uint8_t *dest,

  1252.                                  int dstW, int uvalpha, int y)

  1253. {

  1254.     const int16_t *ubuf0 = ubuf[0];

  1255.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1256. 

  1257.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1258.         const int16_t *ubuf1 = ubuf[0];

  1259.         __asm__ volatile(

  1260.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1261.             "mov        %4, %%"REG_b"               \n\t"

  1262.             "push %%"REG_BP"                        \n\t"

  1263.             YSCALEYUV2RGB1(%%REGBP, %5)

  1264.             "pxor    %%mm7, %%mm7                   \n\t"

  1265.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1266. #ifdef DITHER1XBPP

  1267.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1268.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1269.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1270. #endif

  1271.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1272.             "pop %%"REG_BP"                         \n\t"

  1273.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1274.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1275.                "a" (&c->redDither)

  1276.         );

  1277.     } else {

  1278.         const int16_t *ubuf1 = ubuf[1];

  1279.         __asm__ volatile(

  1280.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1281.             "mov        %4, %%"REG_b"               \n\t"

  1282.             "push %%"REG_BP"                        \n\t"

  1283.             YSCALEYUV2RGB1b(%%REGBP, %5)

  1284.             "pxor    %%mm7, %%mm7                   \n\t"

  1285.             /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1286. #ifdef DITHER1XBPP

  1287.             "paddusb "BLUE_DITHER"(%5), %%mm2      \n\t"

  1288.             "paddusb "GREEN_DITHER"(%5), %%mm4      \n\t"

  1289.             "paddusb "RED_DITHER"(%5), %%mm5      \n\t"

  1290. #endif

  1291.             WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1292.             "pop %%"REG_BP"                         \n\t"

  1293.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1294.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1295.                "a" (&c->redDither)

  1296.         );

  1297.     }

  1298. }

  1299. 

  1300. #define REAL_YSCALEYUV2PACKED1(index, c) \

  1301.     "xor            "#index", "#index"  \n\t"\

  1302.     ".p2align              4            \n\t"\

  1303.     "1:                                 \n\t"\

  1304.     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\

  1305.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1306.     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\

  1307.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1308.     "psraw                $7, %%mm3     \n\t" \

  1309.     "psraw                $7, %%mm4     \n\t" \

  1310.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1311.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1312.     "psraw                $7, %%mm1     \n\t" \

  1313.     "psraw                $7, %%mm7     \n\t" \

  1314. 

  1315. #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)

  1316. 

  1317. #define REAL_YSCALEYUV2PACKED1b(index, c) \

  1318.     "xor "#index", "#index"             \n\t"\

  1319.     ".p2align              4            \n\t"\

  1320.     "1:                                 \n\t"\

  1321.     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\

  1322.     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\

  1323.     "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1324.     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\

  1325.     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\

  1326.     "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \

  1327.     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\

  1328.     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\

  1329.     "psrlw                $8, %%mm3     \n\t" \

  1330.     "psrlw                $8, %%mm4     \n\t" \

  1331.     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\

  1332.     "movq 8(%0, "#index", 2), %%mm7     \n\t" /*buf0[eax]*/\

  1333.     "psraw                $7, %%mm1     \n\t" \

  1334.     "psraw                $7, %%mm7     \n\t"

  1335. #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)

  1336. 

  1337. static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0,

  1338.                                   const int16_t *ubuf[2], const int16_t *vbuf[2],

  1339.                                   const int16_t *abuf0, uint8_t *dest,

  1340.                                   int dstW, int uvalpha, int y)

  1341. {

  1342.     const int16_t *ubuf0 = ubuf[0];

  1343.     const int16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1344. 

  1345.     if (uvalpha < 2048) { // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1346.         const int16_t *ubuf1 = ubuf[0];

  1347.         __asm__ volatile(

  1348.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1349.             "mov        %4, %%"REG_b"               \n\t"

  1350.             "push %%"REG_BP"                        \n\t"

  1351.             YSCALEYUV2PACKED1(%%REGBP, %5)

  1352.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1353.             "pop %%"REG_BP"                         \n\t"

  1354.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1355.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1356.                "a" (&c->redDither)

  1357.         );

  1358.     } else {

  1359.         const int16_t *ubuf1 = ubuf[1];

  1360.         __asm__ volatile(

  1361.             "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1362.             "mov        %4, %%"REG_b"               \n\t"

  1363.             "push %%"REG_BP"                        \n\t"

  1364.             YSCALEYUV2PACKED1b(%%REGBP, %5)

  1365.             WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1366.             "pop %%"REG_BP"                         \n\t"

  1367.             "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1368.             :: "c" (buf0), "d" (buf1), "S" (ubuf0), "D" (ubuf1), "m" (dest),

  1369.                "a" (&c->redDither)

  1370.         );

  1371.     }

  1372. }

  1373. 

  1374. #if COMPILE_TEMPLATE_MMX2

  1375. static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,

  1376.                                  int dstWidth, const uint8_t *src,

  1377.                                  int srcW, int xInc)

  1378. {

  1379.     int32_t *filterPos = c->hLumFilterPos;

  1380.     int16_t *filter    = c->hLumFilter;

  1381.     void    *mmx2FilterCode= c->lumMmx2FilterCode;

  1382.     int i;

  1383. #if defined(PIC)

  1384.     uint64_t ebxsave;

  1385. #endif

  1386. #if ARCH_X86_64

  1387.     uint64_t retsave;

  1388. #endif

  1389. 

  1390.     __asm__ volatile(

  1391. #if defined(PIC)

  1392.         "mov               %%"REG_b", %5        \n\t"

  1393. #if ARCH_X86_64

  1394.         "mov               -8(%%rsp), %%"REG_a" \n\t"

  1395.         "mov               %%"REG_a", %6        \n\t"

  1396. #endif

  1397. #else

  1398. #if ARCH_X86_64

  1399.         "mov               -8(%%rsp), %%"REG_a" \n\t"

  1400.         "mov               %%"REG_a", %5        \n\t"

  1401. #endif

  1402. #endif

  1403.         "pxor                  %%mm7, %%mm7     \n\t"

  1404.         "mov                      %0, %%"REG_c" \n\t"

  1405.         "mov                      %1, %%"REG_D" \n\t"

  1406.         "mov                      %2, %%"REG_d" \n\t"

  1407.         "mov                      %3, %%"REG_b" \n\t"

  1408.         "xor               %%"REG_a", %%"REG_a" \n\t" // i

  1409.         PREFETCH"        (%%"REG_c")            \n\t"

  1410.         PREFETCH"      32(%%"REG_c")            \n\t"

  1411.         PREFETCH"      64(%%"REG_c")            \n\t"

  1412. 

  1413. #if ARCH_X86_64

  1414. #define CALL_MMX2_FILTER_CODE \

  1415.         "movl            (%%"REG_b"), %%esi     \n\t"\

  1416.         "call                    *%4            \n\t"\

  1417.         "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\

  1418.         "add               %%"REG_S", %%"REG_c" \n\t"\

  1419.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1420.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1421. 

  1422. #else

  1423. #define CALL_MMX2_FILTER_CODE \

  1424.         "movl (%%"REG_b"), %%esi        \n\t"\

  1425.         "call         *%4                       \n\t"\

  1426.         "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  1427.         "add               %%"REG_a", %%"REG_D" \n\t"\

  1428.         "xor               %%"REG_a", %%"REG_a" \n\t"\

  1429. 

  1430. #endif /* ARCH_X86_64 */

  1431. 

  1432.         CALL_MMX2_FILTER_CODE

  1433.         CALL_MMX2_FILTER_CODE

  1434.         CALL_MMX2_FILTER_CODE

  1435.         CALL_MMX2_FILTER_CODE

  1436.         CALL_MMX2_FILTER_CODE

  1437.         CALL_MMX2_FILTER_CODE

  1438.         CALL_MMX2_FILTER_CODE

  1439.         CALL_MMX2_FILTER_CODE

  1440. 

  1441. #if defined(PIC)

  1442.         "mov                      %5, %%"REG_b" \n\t"

  1443. #if ARCH_X86_64

  1444.         "mov                      %6, %%"REG_a" \n\t"

  1445.         "mov               %%"REG_a", -8(%%rsp) \n\t"

  1446. #endif

  1447. #else

  1448. #if ARCH_X86_64

  1449.         "mov                      %5, %%"REG_a" \n\t"

  1450.         "mov               %%"REG_a", -8(%%rsp) \n\t"

  1451. #endif

  1452. #endif

  1453.         :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),

  1454.            "m" (mmx2FilterCode)

  1455. #if defined(PIC)

  1456.           ,"m" (ebxsave)

  1457. #endif

  1458. #if ARCH_X86_64

  1459.           ,"m"(retsave)

  1460. #endif

  1461.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  1462. #if !defined(PIC)

  1463.          ,"%"REG_b

  1464. #endif

  1465.     );

  1466. 

  1467.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)

  1468.         dst[i] = src[srcW-1]*128;

  1469. }

  1470. 

  1471. static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,

  1472.                                  int dstWidth, const uint8_t *src1,

  1473.                                  const uint8_t *src2, int srcW, int xInc)

  1474. {

  1475.     int32_t *filterPos = c->hChrFilterPos;

  1476.     int16_t *filter    = c->hChrFilter;

  1477.     void    *mmx2FilterCode= c->chrMmx2FilterCode;

  1478.     int i;

  1479. #if defined(PIC)

  1480.     DECLARE_ALIGNED(8, uint64_t, ebxsave);

  1481. #endif

  1482. #if ARCH_X86_64

  1483.     DECLARE_ALIGNED(8, uint64_t, retsave);

  1484. #endif

  1485. 

  1486.     __asm__ volatile(

  1487. #if defined(PIC)

  1488.         "mov          %%"REG_b", %7         \n\t"

  1489. #if ARCH_X86_64

  1490.         "mov          -8(%%rsp), %%"REG_a"  \n\t"

  1491.         "mov          %%"REG_a", %8         \n\t"

  1492. #endif

  1493. #else

  1494. #if ARCH_X86_64

  1495.         "mov          -8(%%rsp), %%"REG_a"  \n\t"

  1496.         "mov          %%"REG_a", %7         \n\t"

  1497. #endif

  1498. #endif

  1499.         "pxor             %%mm7, %%mm7      \n\t"

  1500.         "mov                 %0, %%"REG_c"  \n\t"

  1501.         "mov                 %1, %%"REG_D"  \n\t"

  1502.         "mov                 %2, %%"REG_d"  \n\t"

  1503.         "mov                 %3, %%"REG_b"  \n\t"

  1504.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  1505.         PREFETCH"   (%%"REG_c")             \n\t"

  1506.         PREFETCH" 32(%%"REG_c")             \n\t"

  1507.         PREFETCH" 64(%%"REG_c")             \n\t"

  1508. 

  1509.         CALL_MMX2_FILTER_CODE

  1510.         CALL_MMX2_FILTER_CODE

  1511.         CALL_MMX2_FILTER_CODE

  1512.         CALL_MMX2_FILTER_CODE

  1513.         "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  1514.         "mov                 %5, %%"REG_c"  \n\t" // src

  1515.         "mov                 %6, %%"REG_D"  \n\t" // buf2

  1516.         PREFETCH"   (%%"REG_c")             \n\t"

  1517.         PREFETCH" 32(%%"REG_c")             \n\t"

  1518.         PREFETCH" 64(%%"REG_c")             \n\t"

  1519. 

  1520.         CALL_MMX2_FILTER_CODE

  1521.         CALL_MMX2_FILTER_CODE

  1522.         CALL_MMX2_FILTER_CODE

  1523.         CALL_MMX2_FILTER_CODE

  1524. 

  1525. #if defined(PIC)

  1526.         "mov %7, %%"REG_b"    \n\t"

  1527. #if ARCH_X86_64

  1528.         "mov                 %8, %%"REG_a"  \n\t"

  1529.         "mov          %%"REG_a", -8(%%rsp)  \n\t"

  1530. #endif

  1531. #else

  1532. #if ARCH_X86_64

  1533.         "mov                 %7, %%"REG_a"  \n\t"

  1534.         "mov          %%"REG_a", -8(%%rsp)  \n\t"

  1535. #endif

  1536. #endif

  1537.         :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),

  1538.            "m" (mmx2FilterCode), "m" (src2), "m"(dst2)

  1539. #if defined(PIC)

  1540.           ,"m" (ebxsave)

  1541. #endif

  1542. #if ARCH_X86_64

  1543.           ,"m"(retsave)

  1544. #endif

  1545.         : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  1546. #if !defined(PIC)

  1547.          ,"%"REG_b

  1548. #endif

  1549.     );

  1550. 

  1551.     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {

  1552.         dst1[i] = src1[srcW-1]*128;

  1553.         dst2[i] = src2[srcW-1]*128;

  1554.     }

  1555. }

  1556. #endif /* COMPILE_TEMPLATE_MMX2 */

  1557. 

  1558. static av_cold void RENAME(sws_init_swScale)(SwsContext *c)

  1559. {

  1560.     enum PixelFormat dstFormat = c->dstFormat;

  1561. 

  1562.     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&

  1563.         dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21) {

  1564.         if (!(c->flags & SWS_BITEXACT)) {

  1565.             if (c->flags & SWS_ACCURATE_RND) {

  1566.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1567.                     switch (c->dstFormat) {

  1568.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X_ar);   break;

  1569.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X_ar);   break;

  1570.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X_ar);  break;

  1571.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X_ar);  break;

  1572.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;

  1573.                     default: break;

  1574.                     }

  1575.                 }

  1576.             } else {

  1577.                 if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1578.                     switch (c->dstFormat) {

  1579.                     case PIX_FMT_RGB32:   c->yuv2packedX = RENAME(yuv2rgb32_X);   break;

  1580.                     case PIX_FMT_BGR24:   c->yuv2packedX = RENAME(yuv2bgr24_X);   break;

  1581.                     case PIX_FMT_RGB555:  c->yuv2packedX = RENAME(yuv2rgb555_X);  break;

  1582.                     case PIX_FMT_RGB565:  c->yuv2packedX = RENAME(yuv2rgb565_X);  break;

  1583.                     case PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;

  1584.                     default: break;

  1585.                     }

  1586.                 }

  1587.             }

  1588.         }

  1589.         if (!(c->flags & SWS_FULL_CHR_H_INT)) {

  1590.             switch (c->dstFormat) {

  1591.             case PIX_FMT_RGB32:

  1592.                 c->yuv2packed1 = RENAME(yuv2rgb32_1);

  1593.                 c->yuv2packed2 = RENAME(yuv2rgb32_2);

  1594.                 break;

  1595.             case PIX_FMT_BGR24:

  1596.                 c->yuv2packed1 = RENAME(yuv2bgr24_1);

  1597.                 c->yuv2packed2 = RENAME(yuv2bgr24_2);

  1598.                 break;

  1599.             case PIX_FMT_RGB555:

  1600.                 c->yuv2packed1 = RENAME(yuv2rgb555_1);

  1601.                 c->yuv2packed2 = RENAME(yuv2rgb555_2);

  1602.                 break;

  1603.             case PIX_FMT_RGB565:

  1604.                 c->yuv2packed1 = RENAME(yuv2rgb565_1);

  1605.                 c->yuv2packed2 = RENAME(yuv2rgb565_2);

  1606.                 break;

  1607.             case PIX_FMT_YUYV422:

  1608.                 c->yuv2packed1 = RENAME(yuv2yuyv422_1);

  1609.                 c->yuv2packed2 = RENAME(yuv2yuyv422_2);

  1610.                 break;

  1611.             default:

  1612.                 break;

  1613.             }

  1614.         }

  1615.     }

  1616. 

  1617.     if (c->srcBpc == 8 && c->dstBpc <= 10) {

  1618.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  1619. #if COMPILE_TEMPLATE_MMX2

  1620.     if (c->flags & SWS_FAST_BILINEAR && c->canMMX2BeUsed)

  1621.     {

  1622.         c->hyscale_fast = RENAME(hyscale_fast);

  1623.         c->hcscale_fast = RENAME(hcscale_fast);

  1624.     } else {

  1625. #endif /* COMPILE_TEMPLATE_MMX2 */

  1626.         c->hyscale_fast = NULL;

  1627.         c->hcscale_fast = NULL;

  1628. #if COMPILE_TEMPLATE_MMX2

  1629.     }

  1630. #endif /* COMPILE_TEMPLATE_MMX2 */

  1631.     }

  1632. }