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.
15934 Commits
32 Branches
1.0GB
Tree: 2e889ae4b9
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '2e889ae4b9'
${ noResults }
FFmpeg/libswscale/swscale_template.c

3255 lines
137KB
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 modify

  7.  * it under the terms of the GNU General Public License as published by

  8.  * the Free Software Foundation; either version 2 of the License, or

  9.  * (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

  14.  * GNU General Public License for more details.

  15.  *

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

  17.  * 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.  * The C code (not assembly, MMX, ...) of this file can be used

  21.  * under the LGPL license.

  22.  */

  23. 

  24. #undef REAL_MOVNTQ

  25. #undef MOVNTQ

  26. #undef PAVGB

  27. #undef PREFETCH

  28. #undef PREFETCHW

  29. #undef EMMS

  30. #undef SFENCE

  31. 

  32. #ifdef HAVE_3DNOW

  33. /* On K6 femms is faster than emms. On K7 femms is directly mapped on emms. */

  34. #define EMMS     "femms"

  35. #else

  36. #define EMMS     "emms"

  37. #endif

  38. 

  39. #ifdef HAVE_3DNOW

  40. #define PREFETCH  "prefetch"

  41. #define PREFETCHW "prefetchw"

  42. #elif defined (HAVE_MMX2)

  43. #define PREFETCH "prefetchnta"

  44. #define PREFETCHW "prefetcht0"

  45. #else

  46. #define PREFETCH  " # nop"

  47. #define PREFETCHW " # nop"

  48. #endif

  49. 

  50. #ifdef HAVE_MMX2

  51. #define SFENCE "sfence"

  52. #else

  53. #define SFENCE " # nop"

  54. #endif

  55. 

  56. #ifdef HAVE_MMX2

  57. #define PAVGB(a,b) "pavgb " #a ", " #b " \n\t"

  58. #elif defined (HAVE_3DNOW)

  59. #define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"

  60. #endif

  61. 

  62. #ifdef HAVE_MMX2

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

  64. #else

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

  66. #endif

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

  68. 

  69. #ifdef HAVE_ALTIVEC

  70. #include "swscale_altivec_template.c"

  71. #endif

  72. 

  73. #define YSCALEYUV2YV12X(x, offset, dest, width) \

  74.     asm volatile(\

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

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

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

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

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

  80.     ASMALIGN(4) /* FIXME Unroll? */\

  81.     "1:                                                 \n\t"\

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

  83.     "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm2      \n\t" /* srcData */\

  84.     "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm5      \n\t" /* srcData */\

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

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

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

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

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

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

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

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

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

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

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

  96.     MOVNTQ(%%mm3, (%1, %%REGa))\

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

  98.     "cmp                                 %2, %%"REG_a"  \n\t"\

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

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

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

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

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

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

  105.     "r" (dest), "g" (width)\

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

  107.     );

  108. 

  109. #define YSCALEYUV2YV12X_ACCURATE(x, offset, dest, width) \

  110.     asm volatile(\

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

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

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

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

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

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

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

  118.     ASMALIGN(4) \

  119.     "1:                                                 \n\t"\

  120.     "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm0      \n\t" /* srcData */\

  121.     "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm2      \n\t" /* srcData */\

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

  123.     "movq   "  x "(%%"REG_S", %%"REG_a", 2), %%mm1      \n\t" /* srcData */\

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

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

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

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

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

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

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

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

  132.     "movq 8+"  x "(%%"REG_S", %%"REG_a", 2), %%mm3      \n\t" /* srcData */\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  154.     "psraw                               $3, %%mm6      \n\t"\

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

  156.     MOVNTQ(%%mm4, (%1, %%REGa))\

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

  158.     "cmp                                 %2, %%"REG_a"  \n\t"\

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

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

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

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

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

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

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

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

  167.     "r" (dest), "g" (width)\

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

  169.     );

  170. 

  171. #define YSCALEYUV2YV121 \

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

  173.     ASMALIGN(4) /* FIXME Unroll? */\

  174.     "1:                                   \n\t"\

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

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

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

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

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

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

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

  182.     "jnc                   1b             \n\t"

  183. 

  184. #define YSCALEYUV2YV121_ACCURATE \

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

  186.     "pcmpeqw %%mm7, %%mm7                 \n\t"\

  187.     "psrlw                 $15, %%mm7     \n\t"\

  188.     "psllw                  $6, %%mm7     \n\t"\

  189.     ASMALIGN(4) /* FIXME Unroll? */\

  190.     "1:                                   \n\t"\

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

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

  193.     "paddsw             %%mm7, %%mm0      \n\t"\

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

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

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

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

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

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

  200.     "jnc                   1b             \n\t"

  201. 

  202. /*

  203.     :: "m" (-lumFilterSize), "m" (-chrFilterSize),

  204.        "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),

  205.        "r" (dest), "m" (dstW),

  206.        "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)

  207.     : "%eax", "%ebx", "%ecx", "%edx", "%esi"

  208. */

  209. #define YSCALEYUV2PACKEDX \

  210.     asm volatile(\

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

  212.     ASMALIGN(4)\

  213.     "nop                                            \n\t"\

  214.     "1:                                             \n\t"\

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

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

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

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

  219.     ASMALIGN(4)\

  220.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

  232. \

  233.     "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\

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

  235.     "movq      "VROUNDER_OFFSET"(%0), %%mm1         \n\t"\

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

  237.     ASMALIGN(4)\

  238.     "2:                                             \n\t"\

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

  240.     "movq  (%%"REG_S", %%"REG_a", 2), %%mm2         \n\t" /* Y1srcData */\

  241.     "movq 8(%%"REG_S", %%"REG_a", 2), %%mm5         \n\t" /* Y2srcData */\

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

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

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

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

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

  247.     "paddw                     %%mm5, %%mm7         \n\t"\

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

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

  250. 

  251. #define YSCALEYUV2PACKEDX_END                 \

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

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

  254.         "r" (dest), "m" (dstW)                \

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

  256.     );

  257. 

  258. #define YSCALEYUV2PACKEDX_ACCURATE \

  259.     asm volatile(\

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

  261.     ASMALIGN(4)\

  262.     "nop                                            \n\t"\

  263.     "1:                                             \n\t"\

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

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

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

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

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

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

  270.     ASMALIGN(4)\

  271.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  307. \

  308.     "lea "LUM_MMX_FILTER_OFFSET"(%0), %%"REG_d"     \n\t"\

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

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

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

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

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

  314.     ASMALIGN(4)\

  315.     "2:                                             \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  351. 

  352. #define YSCALEYUV2RGBX \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  388. #if 0

  389. #define FULL_YSCALEYUV2RGB \

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

  391.     "movd                    %6, %%mm6  \n\t" /*yalpha1*/\

  392.     "punpcklwd            %%mm6, %%mm6  \n\t"\

  393.     "punpcklwd            %%mm6, %%mm6  \n\t"\

  394.     "movd                    %7, %%mm5  \n\t" /*uvalpha1*/\

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

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

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

  398.     ASMALIGN(4)\

  399.     "1:                                 \n\t"\

  400.     "movq     (%0, %%"REG_a",2), %%mm0  \n\t" /*buf0[eax]*/\

  401.     "movq     (%1, %%"REG_a",2), %%mm1  \n\t" /*buf1[eax]*/\

  402.     "movq     (%2, %%"REG_a",2), %%mm2  \n\t" /* uvbuf0[eax]*/\

  403.     "movq     (%3, %%"REG_a",2), %%mm3  \n\t" /* uvbuf1[eax]*/\

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

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

  406.     "pmulhw               %%mm6, %%mm0  \n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\

  407.     "pmulhw               %%mm5, %%mm2  \n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\

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

  409.     "movq "AV_STRINGIFY(VOF)"(%2, %%"REG_a",2), %%mm4  \n\t" /* uvbuf0[eax+2048]*/\

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

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

  412.     "movq "AV_STRINGIFY(VOF)"(%3, %%"REG_a",2), %%mm0  \n\t" /* uvbuf1[eax+2048]*/\

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

  414.     "psubw                %%mm0, %%mm4  \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\

  415.     "psubw        "MANGLE(w80)", %%mm1  \n\t" /* 8(Y-16)*/\

  416.     "psubw       "MANGLE(w400)", %%mm3  \n\t" /* 8(U-128)*/\

  417.     "pmulhw    "MANGLE(yCoeff)", %%mm1  \n\t"\

  418. \

  419. \

  420.     "pmulhw               %%mm5, %%mm4  \n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\

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

  422.     "pmulhw   "MANGLE(ubCoeff)", %%mm3  \n\t"\

  423.     "psraw                   $4, %%mm0  \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\

  424.     "pmulhw   "MANGLE(ugCoeff)", %%mm2  \n\t"\

  425.     "paddw                %%mm4, %%mm0  \n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\

  426.     "psubw       "MANGLE(w400)", %%mm0  \n\t" /* (V-128)8*/\

  427. \

  428. \

  429.     "movq                 %%mm0, %%mm4  \n\t" /* (V-128)8*/\

  430.     "pmulhw   "MANGLE(vrCoeff)", %%mm0  \n\t"\

  431.     "pmulhw   "MANGLE(vgCoeff)", %%mm4  \n\t"\

  432.     "paddw                %%mm1, %%mm3  \n\t" /* B*/\

  433.     "paddw                %%mm1, %%mm0  \n\t" /* R*/\

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

  435. \

  436.     "packuswb             %%mm0, %%mm0  \n\t"\

  437.     "paddw                %%mm4, %%mm2  \n\t"\

  438.     "paddw                %%mm2, %%mm1  \n\t" /* G*/\

  439. \

  440.     "packuswb             %%mm1, %%mm1  \n\t"

  441. #endif

  442. 

  443. #define REAL_YSCALEYUV2PACKED(index, c) \

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

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

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

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

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

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

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

  451.     ASMALIGN(4)\

  452.     "1:                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  478. 

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

  480. 

  481. #define REAL_YSCALEYUV2RGB(index, c) \

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

  483.     ASMALIGN(4)\

  484.     "1:                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  545. #define YSCALEYUV2RGB(index, c)  REAL_YSCALEYUV2RGB(index, c)

  546. 

  547. #define REAL_YSCALEYUV2PACKED1(index, c) \

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

  549.     ASMALIGN(4)\

  550.     "1:                                 \n\t"\

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

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

  553.     "psraw                $7, %%mm3     \n\t" \

  554.     "psraw                $7, %%mm4     \n\t" \

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

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

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

  558.     "psraw                $7, %%mm7     \n\t" \

  559. 

  560. #define YSCALEYUV2PACKED1(index, c)  REAL_YSCALEYUV2PACKED1(index, c)

  561. 

  562. #define REAL_YSCALEYUV2RGB1(index, c) \

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

  564.     ASMALIGN(4)\

  565.     "1:                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  610. 

  611. #define REAL_YSCALEYUV2PACKED1b(index, c) \

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

  613.     ASMALIGN(4)\

  614.     "1:                                 \n\t"\

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

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

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

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

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

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

  621.     "psrlw                $8, %%mm3     \n\t" \

  622.     "psrlw                $8, %%mm4     \n\t" \

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

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

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

  626.     "psraw                $7, %%mm7     \n\t"

  627. #define YSCALEYUV2PACKED1b(index, c)  REAL_YSCALEYUV2PACKED1b(index, c)

  628. 

  629. // do vertical chrominance interpolation

  630. #define REAL_YSCALEYUV2RGB1b(index, c) \

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

  632.     ASMALIGN(4)\

  633.     "1:                                 \n\t"\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  682. 

  683. #define REAL_WRITEBGR32(dst, dstw, index) \

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

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

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

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

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

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

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

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

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

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

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

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

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

  697. \

  698.     MOVNTQ(%%mm0,   (dst, index, 4))\

  699.     MOVNTQ(%%mm2,  8(dst, index, 4))\

  700.     MOVNTQ(%%mm1, 16(dst, index, 4))\

  701.     MOVNTQ(%%mm3, 24(dst, index, 4))\

  702. \

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

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

  705.     " jb      1b                \n\t"

  706. #define WRITEBGR32(dst, dstw, index)  REAL_WRITEBGR32(dst, dstw, index)

  707. 

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

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

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

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

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

  713. \

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

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

  716. \

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

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

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

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

  721. \

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

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

  724. \

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

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

  727. \

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

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

  730. \

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

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

  733.     " jb             1b             \n\t"

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

  735. 

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

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

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

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

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

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

  742. \

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

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

  745. \

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

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

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

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

  750. \

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

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

  753. \

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

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

  756. \

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

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

  759. \

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

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

  762.     " jb             1b             \n\t"

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

  764. 

  765. #define WRITEBGR24OLD(dst, dstw, index) \

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

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

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

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

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

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

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

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

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

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

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

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

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

  779. \

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

  781.     "psrlq        $8, %%mm0             \n\t" /* 00RGB0RG 0 */\

  782.     "pand "MANGLE(bm00000111)", %%mm4   \n\t" /* 00000RGB 0 */\

  783.     "pand "MANGLE(bm11111000)", %%mm0   \n\t" /* 00RGB000 0.5 */\

  784.     "por       %%mm4, %%mm0             \n\t" /* 00RGBRGB 0 */\

  785.     "movq      %%mm2, %%mm4             \n\t" /* 0RGB0RGB 1 */\

  786.     "psllq       $48, %%mm2             \n\t" /* GB000000 1 */\

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

  788. \

  789.     "movq      %%mm4, %%mm2             \n\t" /* 0RGB0RGB 1 */\

  790.     "psrld       $16, %%mm4             \n\t" /* 000R000R 1 */\

  791.     "psrlq       $24, %%mm2             \n\t" /* 0000RGB0 1.5 */\

  792.     "por       %%mm4, %%mm2             \n\t" /* 000RRGBR 1 */\

  793.     "pand "MANGLE(bm00001111)", %%mm2   \n\t" /* 0000RGBR 1 */\

  794.     "movq      %%mm1, %%mm4             \n\t" /* 0RGB0RGB 2 */\

  795.     "psrlq        $8, %%mm1             \n\t" /* 00RGB0RG 2 */\

  796.     "pand "MANGLE(bm00000111)", %%mm4   \n\t" /* 00000RGB 2 */\

  797.     "pand "MANGLE(bm11111000)", %%mm1   \n\t" /* 00RGB000 2.5 */\

  798.     "por       %%mm4, %%mm1             \n\t" /* 00RGBRGB 2 */\

  799.     "movq      %%mm1, %%mm4             \n\t" /* 00RGBRGB 2 */\

  800.     "psllq       $32, %%mm1             \n\t" /* BRGB0000 2 */\

  801.     "por       %%mm1, %%mm2             \n\t" /* BRGBRGBR 1 */\

  802. \

  803.     "psrlq       $32, %%mm4             \n\t" /* 000000RG 2.5 */\

  804.     "movq      %%mm3, %%mm5             \n\t" /* 0RGB0RGB 3 */\

  805.     "psrlq        $8, %%mm3             \n\t" /* 00RGB0RG 3 */\

  806.     "pand "MANGLE(bm00000111)", %%mm5   \n\t" /* 00000RGB 3 */\

  807.     "pand "MANGLE(bm11111000)", %%mm3   \n\t" /* 00RGB000 3.5 */\

  808.     "por       %%mm5, %%mm3             \n\t" /* 00RGBRGB 3 */\

  809.     "psllq       $16, %%mm3             \n\t" /* RGBRGB00 3 */\

  810.     "por       %%mm4, %%mm3             \n\t" /* RGBRGBRG 2.5 */\

  811. \

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

  813.     MOVNTQ(%%mm2,  8(dst))\

  814.     MOVNTQ(%%mm3, 16(dst))\

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

  816. \

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

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

  819.     " jb          1b                    \n\t"

  820. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  835. \

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

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

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

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

  840. \

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

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

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

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

  845. \

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

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

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

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

  850. \

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

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

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

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

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

  856. \

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

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

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

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

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

  862. \

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

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

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

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

  867. \

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

  869. \

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

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

  872.     " jb          1b            \n\t"

  873. 

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

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

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

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

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

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

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

  881. \

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

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

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

  885. \

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

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

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

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

  890. \

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

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

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

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

  895. \

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

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

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

  899. \

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

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

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

  903. \

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

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

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

  907. \

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

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

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

  911. \

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

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

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

  915. \

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

  917. \

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

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

  920.     " jb       1b               \n\t"

  921. 

  922. #ifdef HAVE_MMX2

  923. #undef WRITEBGR24

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

  925. #else

  926. #undef WRITEBGR24

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

  928. #endif

  929. 

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

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

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

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

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

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

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

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

  938. \

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

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

  941. \

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

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

  944.     " jb          1b            \n\t"

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

  946. 

  947. 

  948. static inline void RENAME(yuv2yuvX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,

  949.                                     int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

  950.                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW)

  951. {

  952. #ifdef HAVE_MMX

  953.     if(!(c->flags & SWS_BITEXACT)){

  954.         if (c->flags & SWS_ACCURATE_RND){

  955.             if (uDest){

  956.                 YSCALEYUV2YV12X_ACCURATE(   "0", CHR_MMX_FILTER_OFFSET, uDest, chrDstW)

  957.                 YSCALEYUV2YV12X_ACCURATE(AV_STRINGIFY(VOF), CHR_MMX_FILTER_OFFSET, vDest, chrDstW)

  958.             }

  959. 

  960.             YSCALEYUV2YV12X_ACCURATE("0", LUM_MMX_FILTER_OFFSET, dest, dstW)

  961.         }else{

  962.             if (uDest){

  963.                 YSCALEYUV2YV12X(   "0", CHR_MMX_FILTER_OFFSET, uDest, chrDstW)

  964.                 YSCALEYUV2YV12X(AV_STRINGIFY(VOF), CHR_MMX_FILTER_OFFSET, vDest, chrDstW)

  965.             }

  966. 

  967.             YSCALEYUV2YV12X("0", LUM_MMX_FILTER_OFFSET, dest, dstW)

  968.         }

  969.         return;

  970.     }

  971. #endif

  972. #ifdef HAVE_ALTIVEC

  973. yuv2yuvX_altivec_real(lumFilter, lumSrc, lumFilterSize,

  974.                       chrFilter, chrSrc, chrFilterSize,

  975.                       dest, uDest, vDest, dstW, chrDstW);

  976. #else //HAVE_ALTIVEC

  977. yuv2yuvXinC(lumFilter, lumSrc, lumFilterSize,

  978.             chrFilter, chrSrc, chrFilterSize,

  979.             dest, uDest, vDest, dstW, chrDstW);

  980. #endif //!HAVE_ALTIVEC

  981. }

  982. 

  983. static inline void RENAME(yuv2nv12X)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,

  984.                                      int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

  985.                                      uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)

  986. {

  987. yuv2nv12XinC(lumFilter, lumSrc, lumFilterSize,

  988.              chrFilter, chrSrc, chrFilterSize,

  989.              dest, uDest, dstW, chrDstW, dstFormat);

  990. }

  991. 

  992. static inline void RENAME(yuv2yuv1)(SwsContext *c, int16_t *lumSrc, int16_t *chrSrc,

  993.                                     uint8_t *dest, uint8_t *uDest, uint8_t *vDest, long dstW, long chrDstW)

  994. {

  995.     int i;

  996. #ifdef HAVE_MMX

  997.     if(!(c->flags & SWS_BITEXACT)){

  998.         long p= uDest ? 3 : 1;

  999.         uint8_t *src[3]= {lumSrc + dstW, chrSrc + chrDstW, chrSrc + VOFW + chrDstW};

  1000.         uint8_t *dst[3]= {dest, uDest, vDest};

  1001.         long counter[3] = {dstW, chrDstW, chrDstW};

  1002. 

  1003.         if (c->flags & SWS_ACCURATE_RND){

  1004.             while(p--){

  1005.                 asm volatile(

  1006.                     YSCALEYUV2YV121_ACCURATE

  1007.                     :: "r" (src[p]), "r" (dst[p] + counter[p]),

  1008.                     "g" (-counter[p])

  1009.                     : "%"REG_a

  1010.                 );

  1011.             }

  1012.         }else{

  1013.             while(p--){

  1014.                 asm volatile(

  1015.                     YSCALEYUV2YV121

  1016.                     :: "r" (src[p]), "r" (dst[p] + counter[p]),

  1017.                     "g" (-counter[p])

  1018.                     : "%"REG_a

  1019.                 );

  1020.             }

  1021.         }

  1022.         return;

  1023.     }

  1024. #endif

  1025.     for (i=0; i<dstW; i++)

  1026.     {

  1027.         int val= (lumSrc[i]+64)>>7;

  1028. 

  1029.         if (val&256){

  1030.             if (val<0) val=0;

  1031.             else       val=255;

  1032.         }

  1033. 

  1034.         dest[i]= val;

  1035.     }

  1036. 

  1037.     if (uDest)

  1038.         for (i=0; i<chrDstW; i++)

  1039.         {

  1040.             int u=(chrSrc[i       ]+64)>>7;

  1041.             int v=(chrSrc[i + VOFW]+64)>>7;

  1042. 

  1043.             if ((u|v)&256){

  1044.                 if (u<0)        u=0;

  1045.                 else if (u>255) u=255;

  1046.                 if (v<0)        v=0;

  1047.                 else if (v>255) v=255;

  1048.             }

  1049. 

  1050.             uDest[i]= u;

  1051.             vDest[i]= v;

  1052.         }

  1053. }

  1054. 

  1055. 

  1056. /**

  1057.  * vertical scale YV12 to RGB

  1058.  */

  1059. static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,

  1060.                                        int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,

  1061.                                        uint8_t *dest, long dstW, long dstY)

  1062. {

  1063. #ifdef HAVE_MMX

  1064.     long dummy=0;

  1065.     if(!(c->flags & SWS_BITEXACT)){

  1066.         if (c->flags & SWS_ACCURATE_RND){

  1067.             switch(c->dstFormat){

  1068.             case PIX_FMT_RGB32:

  1069.                 YSCALEYUV2PACKEDX_ACCURATE

  1070.                 YSCALEYUV2RGBX

  1071.                 WRITEBGR32(%4, %5, %%REGa)

  1072. 

  1073.                 YSCALEYUV2PACKEDX_END

  1074.                 return;

  1075.             case PIX_FMT_BGR24:

  1076.                 YSCALEYUV2PACKEDX_ACCURATE

  1077.                 YSCALEYUV2RGBX

  1078.                 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"\n\t" //FIXME optimize

  1079.                 "add %4, %%"REG_c"                        \n\t"

  1080.                 WRITEBGR24(%%REGc, %5, %%REGa)

  1081. 

  1082. 

  1083.                 :: "r" (&c->redDither),

  1084.                 "m" (dummy), "m" (dummy), "m" (dummy),

  1085.                 "r" (dest), "m" (dstW)

  1086.                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S

  1087.                 );

  1088.                 return;

  1089.             case PIX_FMT_RGB555:

  1090.                 YSCALEYUV2PACKEDX_ACCURATE

  1091.                 YSCALEYUV2RGBX

  1092.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1093. #ifdef DITHER1XBPP

  1094.                 "paddusb "MANGLE(b5Dither)", %%mm2\n\t"

  1095.                 "paddusb "MANGLE(g5Dither)", %%mm4\n\t"

  1096.                 "paddusb "MANGLE(r5Dither)", %%mm5\n\t"

  1097. #endif

  1098. 

  1099.                 WRITERGB15(%4, %5, %%REGa)

  1100.                 YSCALEYUV2PACKEDX_END

  1101.                 return;

  1102.             case PIX_FMT_RGB565:

  1103.                 YSCALEYUV2PACKEDX_ACCURATE

  1104.                 YSCALEYUV2RGBX

  1105.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1106. #ifdef DITHER1XBPP

  1107.                 "paddusb "MANGLE(b5Dither)", %%mm2\n\t"

  1108.                 "paddusb "MANGLE(g6Dither)", %%mm4\n\t"

  1109.                 "paddusb "MANGLE(r5Dither)", %%mm5\n\t"

  1110. #endif

  1111. 

  1112.                 WRITERGB16(%4, %5, %%REGa)

  1113.                 YSCALEYUV2PACKEDX_END

  1114.                 return;

  1115.             case PIX_FMT_YUYV422:

  1116.                 YSCALEYUV2PACKEDX_ACCURATE

  1117.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1118. 

  1119.                 "psraw $3, %%mm3    \n\t"

  1120.                 "psraw $3, %%mm4    \n\t"

  1121.                 "psraw $3, %%mm1    \n\t"

  1122.                 "psraw $3, %%mm7    \n\t"

  1123.                 WRITEYUY2(%4, %5, %%REGa)

  1124.                 YSCALEYUV2PACKEDX_END

  1125.                 return;

  1126.             }

  1127.         }else{

  1128.             switch(c->dstFormat)

  1129.             {

  1130.             case PIX_FMT_RGB32:

  1131.                 YSCALEYUV2PACKEDX

  1132.                 YSCALEYUV2RGBX

  1133.                 WRITEBGR32(%4, %5, %%REGa)

  1134.                 YSCALEYUV2PACKEDX_END

  1135.                 return;

  1136.             case PIX_FMT_BGR24:

  1137.                 YSCALEYUV2PACKEDX

  1138.                 YSCALEYUV2RGBX

  1139.                 "lea (%%"REG_a", %%"REG_a", 2), %%"REG_c"   \n\t" //FIXME optimize

  1140.                 "add                        %4, %%"REG_c"   \n\t"

  1141.                 WRITEBGR24(%%REGc, %5, %%REGa)

  1142. 

  1143.                 :: "r" (&c->redDither),

  1144.                 "m" (dummy), "m" (dummy), "m" (dummy),

  1145.                 "r" (dest),  "m" (dstW)

  1146.                 : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S

  1147.                 );

  1148.                 return;

  1149.             case PIX_FMT_RGB555:

  1150.                 YSCALEYUV2PACKEDX

  1151.                 YSCALEYUV2RGBX

  1152.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1153. #ifdef DITHER1XBPP

  1154.                 "paddusb "MANGLE(b5Dither)", %%mm2  \n\t"

  1155.                 "paddusb "MANGLE(g5Dither)", %%mm4  \n\t"

  1156.                 "paddusb "MANGLE(r5Dither)", %%mm5  \n\t"

  1157. #endif

  1158. 

  1159.                 WRITERGB15(%4, %5, %%REGa)

  1160.                 YSCALEYUV2PACKEDX_END

  1161.                 return;

  1162.             case PIX_FMT_RGB565:

  1163.                 YSCALEYUV2PACKEDX

  1164.                 YSCALEYUV2RGBX

  1165.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1166. #ifdef DITHER1XBPP

  1167.                 "paddusb "MANGLE(b5Dither)", %%mm2  \n\t"

  1168.                 "paddusb "MANGLE(g6Dither)", %%mm4  \n\t"

  1169.                 "paddusb "MANGLE(r5Dither)", %%mm5  \n\t"

  1170. #endif

  1171. 

  1172.                 WRITERGB16(%4, %5, %%REGa)

  1173.                 YSCALEYUV2PACKEDX_END

  1174.                 return;

  1175.             case PIX_FMT_YUYV422:

  1176.                 YSCALEYUV2PACKEDX

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

  1178. 

  1179.                 "psraw $3, %%mm3    \n\t"

  1180.                 "psraw $3, %%mm4    \n\t"

  1181.                 "psraw $3, %%mm1    \n\t"

  1182.                 "psraw $3, %%mm7    \n\t"

  1183.                 WRITEYUY2(%4, %5, %%REGa)

  1184.                 YSCALEYUV2PACKEDX_END

  1185.                 return;

  1186.             }

  1187.         }

  1188.     }

  1189. #endif /* HAVE_MMX */

  1190. #ifdef HAVE_ALTIVEC

  1191.     /* The following list of supported dstFormat values should

  1192.        match what's found in the body of altivec_yuv2packedX() */

  1193.     if (c->dstFormat==PIX_FMT_ABGR  || c->dstFormat==PIX_FMT_BGRA  ||

  1194.         c->dstFormat==PIX_FMT_BGR24 || c->dstFormat==PIX_FMT_RGB24 ||

  1195.         c->dstFormat==PIX_FMT_RGBA  || c->dstFormat==PIX_FMT_ARGB)

  1196.             altivec_yuv2packedX (c, lumFilter, lumSrc, lumFilterSize,

  1197.                                  chrFilter, chrSrc, chrFilterSize,

  1198.                                  dest, dstW, dstY);

  1199.     else

  1200. #endif

  1201.         yuv2packedXinC(c, lumFilter, lumSrc, lumFilterSize,

  1202.                        chrFilter, chrSrc, chrFilterSize,

  1203.                        dest, dstW, dstY);

  1204. }

  1205. 

  1206. /**

  1207.  * vertical bilinear scale YV12 to RGB

  1208.  */

  1209. static inline void RENAME(yuv2packed2)(SwsContext *c, uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1,

  1210.                           uint8_t *dest, int dstW, int yalpha, int uvalpha, int y)

  1211. {

  1212.     int  yalpha1=4095- yalpha;

  1213.     int uvalpha1=4095-uvalpha;

  1214.     int i;

  1215. 

  1216. #if 0 //isn't used

  1217.     if (flags&SWS_FULL_CHR_H_INT)

  1218.     {

  1219.         switch(dstFormat)

  1220.         {

  1221. #ifdef HAVE_MMX

  1222.         case PIX_FMT_RGB32:

  1223.             asm volatile(

  1224. 

  1225. 

  1226. FULL_YSCALEYUV2RGB

  1227.             "punpcklbw %%mm1, %%mm3    \n\t" // BGBGBGBG

  1228.             "punpcklbw %%mm7, %%mm0    \n\t" // R0R0R0R0

  1229. 

  1230.             "movq      %%mm3, %%mm1    \n\t"

  1231.             "punpcklwd %%mm0, %%mm3    \n\t" // BGR0BGR0

  1232.             "punpckhwd %%mm0, %%mm1    \n\t" // BGR0BGR0

  1233. 

  1234.             MOVNTQ(%%mm3,  (%4, %%REGa, 4))

  1235.             MOVNTQ(%%mm1, 8(%4, %%REGa, 4))

  1236. 

  1237.             "add $4, %%"REG_a"  \n\t"

  1238.             "cmp %5, %%"REG_a"  \n\t"

  1239.             " jb 1b             \n\t"

  1240. 

  1241.             :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" ((long)dstW),

  1242.             "m" (yalpha1), "m" (uvalpha1)

  1243.             : "%"REG_a

  1244.             );

  1245.             break;

  1246.         case PIX_FMT_BGR24:

  1247.             asm volatile(

  1248. 

  1249. FULL_YSCALEYUV2RGB

  1250. 

  1251.                                               // lsb ... msb

  1252.             "punpcklbw %%mm1, %%mm3     \n\t" // BGBGBGBG

  1253.             "punpcklbw %%mm7, %%mm0     \n\t" // R0R0R0R0

  1254. 

  1255.             "movq      %%mm3, %%mm1     \n\t"

  1256.             "punpcklwd %%mm0, %%mm3     \n\t" // BGR0BGR0

  1257.             "punpckhwd %%mm0, %%mm1     \n\t" // BGR0BGR0

  1258. 

  1259.             "movq      %%mm3, %%mm2     \n\t" // BGR0BGR0

  1260.             "psrlq        $8, %%mm3     \n\t" // GR0BGR00

  1261.             "pand "MANGLE(bm00000111)", %%mm2   \n\t" // BGR00000

  1262.             "pand "MANGLE(bm11111000)", %%mm3   \n\t" // 000BGR00

  1263.             "por       %%mm2, %%mm3     \n\t" // BGRBGR00

  1264.             "movq      %%mm1, %%mm2     \n\t"

  1265.             "psllq       $48, %%mm1     \n\t" // 000000BG

  1266.             "por       %%mm1, %%mm3     \n\t" // BGRBGRBG

  1267. 

  1268.             "movq      %%mm2, %%mm1     \n\t" // BGR0BGR0

  1269.             "psrld       $16, %%mm2     \n\t" // R000R000

  1270.             "psrlq       $24, %%mm1     \n\t" // 0BGR0000

  1271.             "por       %%mm2, %%mm1     \n\t" // RBGRR000

  1272. 

  1273.             "mov          %4, %%"REG_b" \n\t"

  1274.             "add   %%"REG_a", %%"REG_b" \n\t"

  1275. 

  1276. #ifdef HAVE_MMX2

  1277.             //FIXME Alignment

  1278.             "movntq %%mm3,  (%%"REG_b", %%"REG_a", 2)   \n\t"

  1279.             "movntq %%mm1, 8(%%"REG_b", %%"REG_a", 2)   \n\t"

  1280. #else

  1281.             "movd %%mm3,  (%%"REG_b", %%"REG_a", 2)     \n\t"

  1282.             "psrlq  $32, %%mm3                          \n\t"

  1283.             "movd %%mm3, 4(%%"REG_b", %%"REG_a", 2)     \n\t"

  1284.             "movd %%mm1, 8(%%"REG_b", %%"REG_a", 2)     \n\t"

  1285. #endif

  1286.             "add     $4, %%"REG_a"                      \n\t"

  1287.             "cmp     %5, %%"REG_a"                      \n\t"

  1288.             " jb     1b                                 \n\t"

  1289. 

  1290.             :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstW),

  1291.             "m" (yalpha1), "m" (uvalpha1)

  1292.             : "%"REG_a, "%"REG_b

  1293.             );

  1294.             break;

  1295.         case PIX_FMT_BGR555:

  1296.             asm volatile(

  1297. 

  1298. FULL_YSCALEYUV2RGB

  1299. #ifdef DITHER1XBPP

  1300.             "paddusb "MANGLE(g5Dither)", %%mm1  \n\t"

  1301.             "paddusb "MANGLE(r5Dither)", %%mm0  \n\t"

  1302.             "paddusb "MANGLE(b5Dither)", %%mm3  \n\t"

  1303. #endif

  1304.             "punpcklbw            %%mm7, %%mm1  \n\t" // 0G0G0G0G

  1305.             "punpcklbw            %%mm7, %%mm3  \n\t" // 0B0B0B0B

  1306.             "punpcklbw            %%mm7, %%mm0  \n\t" // 0R0R0R0R

  1307. 

  1308.             "psrlw                   $3, %%mm3  \n\t"

  1309.             "psllw                   $2, %%mm1  \n\t"

  1310.             "psllw                   $7, %%mm0  \n\t"

  1311.             "pand     "MANGLE(g15Mask)", %%mm1  \n\t"

  1312.             "pand     "MANGLE(r15Mask)", %%mm0  \n\t"

  1313. 

  1314.             "por                  %%mm3, %%mm1  \n\t"

  1315.             "por                  %%mm1, %%mm0  \n\t"

  1316. 

  1317.             MOVNTQ(%%mm0, (%4, %%REGa, 2))

  1318. 

  1319.             "add $4, %%"REG_a"  \n\t"

  1320.             "cmp %5, %%"REG_a"  \n\t"

  1321.             " jb 1b             \n\t"

  1322. 

  1323.             :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW),

  1324.             "m" (yalpha1), "m" (uvalpha1)

  1325.             : "%"REG_a

  1326.             );

  1327.             break;

  1328.         case PIX_FMT_BGR565:

  1329.             asm volatile(

  1330. 

  1331. FULL_YSCALEYUV2RGB

  1332. #ifdef DITHER1XBPP

  1333.             "paddusb "MANGLE(g6Dither)", %%mm1  \n\t"

  1334.             "paddusb "MANGLE(r5Dither)", %%mm0  \n\t"

  1335.             "paddusb "MANGLE(b5Dither)", %%mm3  \n\t"

  1336. #endif

  1337.             "punpcklbw            %%mm7, %%mm1  \n\t" // 0G0G0G0G

  1338.             "punpcklbw            %%mm7, %%mm3  \n\t" // 0B0B0B0B

  1339.             "punpcklbw            %%mm7, %%mm0  \n\t" // 0R0R0R0R

  1340. 

  1341.             "psrlw                   $3, %%mm3  \n\t"

  1342.             "psllw                   $3, %%mm1  \n\t"

  1343.             "psllw                   $8, %%mm0  \n\t"

  1344.             "pand     "MANGLE(g16Mask)", %%mm1  \n\t"

  1345.             "pand     "MANGLE(r16Mask)", %%mm0  \n\t"

  1346. 

  1347.             "por                  %%mm3, %%mm1  \n\t"

  1348.             "por                  %%mm1, %%mm0  \n\t"

  1349. 

  1350.             MOVNTQ(%%mm0, (%4, %%REGa, 2))

  1351. 

  1352.             "add $4, %%"REG_a"  \n\t"

  1353.             "cmp %5, %%"REG_a"  \n\t"

  1354.             " jb 1b             \n\t"

  1355. 

  1356.             :: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstW),

  1357.             "m" (yalpha1), "m" (uvalpha1)

  1358.             : "%"REG_a

  1359.             );

  1360.             break;

  1361. #endif /* HAVE_MMX */

  1362.         case PIX_FMT_BGR32:

  1363. #ifndef HAVE_MMX

  1364.         case PIX_FMT_RGB32:

  1365. #endif

  1366.             if (dstFormat==PIX_FMT_RGB32)

  1367.             {

  1368.                 int i;

  1369. #ifdef WORDS_BIGENDIAN

  1370.                 dest++;

  1371. #endif

  1372.                 for (i=0;i<dstW;i++){

  1373.                     // vertical linear interpolation && yuv2rgb in a single step:

  1374.                     int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];

  1375.                     int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);

  1376.                     int V=((uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19);

  1377.                     dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)];

  1378.                     dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)];

  1379.                     dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)];

  1380.                     dest+= 4;

  1381.                 }

  1382.             }

  1383.             else if (dstFormat==PIX_FMT_BGR24)

  1384.             {

  1385.                 int i;

  1386.                 for (i=0;i<dstW;i++){

  1387.                     // vertical linear interpolation && yuv2rgb in a single step:

  1388.                     int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];

  1389.                     int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);

  1390.                     int V=((uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19);

  1391.                     dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)];

  1392.                     dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)];

  1393.                     dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)];

  1394.                     dest+= 3;

  1395.                 }

  1396.             }

  1397.             else if (dstFormat==PIX_FMT_BGR565)

  1398.             {

  1399.                 int i;

  1400.                 for (i=0;i<dstW;i++){

  1401.                     // vertical linear interpolation && yuv2rgb in a single step:

  1402.                     int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];

  1403.                     int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);

  1404.                     int V=((uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19);

  1405. 

  1406.                     ((uint16_t*)dest)[i] =

  1407.                         clip_table16b[(Y + yuvtab_40cf[U]) >>13] |

  1408.                         clip_table16g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] |

  1409.                         clip_table16r[(Y + yuvtab_3343[V]) >>13];

  1410.                 }

  1411.             }

  1412.             else if (dstFormat==PIX_FMT_BGR555)

  1413.             {

  1414.                 int i;

  1415.                 for (i=0;i<dstW;i++){

  1416.                     // vertical linear interpolation && yuv2rgb in a single step:

  1417.                     int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];

  1418.                     int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);

  1419.                     int V=((uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19);

  1420. 

  1421.                     ((uint16_t*)dest)[i] =

  1422.                         clip_table15b[(Y + yuvtab_40cf[U]) >>13] |

  1423.                         clip_table15g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] |

  1424.                         clip_table15r[(Y + yuvtab_3343[V]) >>13];

  1425.                 }

  1426.             }

  1427.         }//FULL_UV_IPOL

  1428.     else

  1429.     {

  1430. #endif // if 0

  1431. #ifdef HAVE_MMX

  1432.     if(!(c->flags & SWS_BITEXACT)){

  1433.         switch(c->dstFormat)

  1434.         {

  1435.             //Note 8280 == DSTW_OFFSET but the preprocessor can't handle that there :(

  1436.             case PIX_FMT_RGB32:

  1437.                 asm volatile(

  1438.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1439.                 "mov        %4, %%"REG_b"               \n\t"

  1440.                 "push %%"REG_BP"                        \n\t"

  1441.                 YSCALEYUV2RGB(%%REGBP, %5)

  1442.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP)

  1443.                 "pop %%"REG_BP"                         \n\t"

  1444.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1445. 

  1446.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1447.                 "a" (&c->redDither)

  1448.                 );

  1449.                 return;

  1450.             case PIX_FMT_BGR24:

  1451.                 asm volatile(

  1452.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1453.                 "mov        %4, %%"REG_b"               \n\t"

  1454.                 "push %%"REG_BP"                        \n\t"

  1455.                 YSCALEYUV2RGB(%%REGBP, %5)

  1456.                 WRITEBGR24(%%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" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1460.                 "a" (&c->redDither)

  1461.                 );

  1462.                 return;

  1463.             case PIX_FMT_RGB555:

  1464.                 asm volatile(

  1465.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1466.                 "mov        %4, %%"REG_b"               \n\t"

  1467.                 "push %%"REG_BP"                        \n\t"

  1468.                 YSCALEYUV2RGB(%%REGBP, %5)

  1469.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1470. #ifdef DITHER1XBPP

  1471.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1472.                 "paddusb "MANGLE(g5Dither)", %%mm4      \n\t"

  1473.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1474. #endif

  1475. 

  1476.                 WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1477.                 "pop %%"REG_BP"                         \n\t"

  1478.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1479. 

  1480.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1481.                 "a" (&c->redDither)

  1482.                 );

  1483.                 return;

  1484.             case PIX_FMT_RGB565:

  1485.                 asm volatile(

  1486.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1487.                 "mov        %4, %%"REG_b"               \n\t"

  1488.                 "push %%"REG_BP"                        \n\t"

  1489.                 YSCALEYUV2RGB(%%REGBP, %5)

  1490.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1491. #ifdef DITHER1XBPP

  1492.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1493.                 "paddusb "MANGLE(g6Dither)", %%mm4      \n\t"

  1494.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1495. #endif

  1496. 

  1497.                 WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1498.                 "pop %%"REG_BP"                         \n\t"

  1499.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1500.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1501.                 "a" (&c->redDither)

  1502.                 );

  1503.                 return;

  1504.             case PIX_FMT_YUYV422:

  1505.                 asm volatile(

  1506.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1507.                 "mov %4, %%"REG_b"                        \n\t"

  1508.                 "push %%"REG_BP"                        \n\t"

  1509.                 YSCALEYUV2PACKED(%%REGBP, %5)

  1510.                 WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1511.                 "pop %%"REG_BP"                         \n\t"

  1512.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1513.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1514.                 "a" (&c->redDither)

  1515.                 );

  1516.                 return;

  1517.             default: break;

  1518.         }

  1519.     }

  1520. #endif //HAVE_MMX

  1521. YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C, YSCALE_YUV_2_GRAY16_2_C, YSCALE_YUV_2_MONO2_C)

  1522. }

  1523. 

  1524. /**

  1525.  * YV12 to RGB without scaling or interpolating

  1526.  */

  1527. static inline void RENAME(yuv2packed1)(SwsContext *c, uint16_t *buf0, uint16_t *uvbuf0, uint16_t *uvbuf1,

  1528.                           uint8_t *dest, int dstW, int uvalpha, int dstFormat, int flags, int y)

  1529. {

  1530.     const int yalpha1=0;

  1531.     int i;

  1532. 

  1533.     uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1

  1534.     const int yalpha= 4096; //FIXME ...

  1535. 

  1536.     if (flags&SWS_FULL_CHR_H_INT)

  1537.     {

  1538.         RENAME(yuv2packed2)(c, buf0, buf0, uvbuf0, uvbuf1, dest, dstW, 0, uvalpha, y);

  1539.         return;

  1540.     }

  1541. 

  1542. #ifdef HAVE_MMX

  1543.     if(!(flags & SWS_BITEXACT)){

  1544.         if (uvalpha < 2048) // note this is not correct (shifts chrominance by 0.5 pixels) but it is a bit faster

  1545.         {

  1546.             switch(dstFormat)

  1547.             {

  1548.             case PIX_FMT_RGB32:

  1549.                 asm volatile(

  1550.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1551.                 "mov        %4, %%"REG_b"               \n\t"

  1552.                 "push %%"REG_BP"                        \n\t"

  1553.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1554.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP)

  1555.                 "pop %%"REG_BP"                         \n\t"

  1556.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1557. 

  1558.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1559.                 "a" (&c->redDither)

  1560.                 );

  1561.                 return;

  1562.             case PIX_FMT_BGR24:

  1563.                 asm volatile(

  1564.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1565.                 "mov        %4, %%"REG_b"               \n\t"

  1566.                 "push %%"REG_BP"                        \n\t"

  1567.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1568.                 WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1569.                 "pop %%"REG_BP"                         \n\t"

  1570.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1571. 

  1572.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1573.                 "a" (&c->redDither)

  1574.                 );

  1575.                 return;

  1576.             case PIX_FMT_RGB555:

  1577.                 asm volatile(

  1578.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1579.                 "mov        %4, %%"REG_b"               \n\t"

  1580.                 "push %%"REG_BP"                        \n\t"

  1581.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1582.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1583. #ifdef DITHER1XBPP

  1584.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1585.                 "paddusb "MANGLE(g5Dither)", %%mm4      \n\t"

  1586.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1587. #endif

  1588.                 WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1589.                 "pop %%"REG_BP"                         \n\t"

  1590.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1591. 

  1592.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1593.                 "a" (&c->redDither)

  1594.                 );

  1595.                 return;

  1596.             case PIX_FMT_RGB565:

  1597.                 asm volatile(

  1598.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1599.                 "mov        %4, %%"REG_b"               \n\t"

  1600.                 "push %%"REG_BP"                        \n\t"

  1601.                 YSCALEYUV2RGB1(%%REGBP, %5)

  1602.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1603. #ifdef DITHER1XBPP

  1604.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1605.                 "paddusb "MANGLE(g6Dither)", %%mm4      \n\t"

  1606.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1607. #endif

  1608. 

  1609.                 WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1610.                 "pop %%"REG_BP"                         \n\t"

  1611.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1612. 

  1613.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1614.                 "a" (&c->redDither)

  1615.                 );

  1616.                 return;

  1617.             case PIX_FMT_YUYV422:

  1618.                 asm volatile(

  1619.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1620.                 "mov        %4, %%"REG_b"               \n\t"

  1621.                 "push %%"REG_BP"                        \n\t"

  1622.                 YSCALEYUV2PACKED1(%%REGBP, %5)

  1623.                 WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1624.                 "pop %%"REG_BP"                         \n\t"

  1625.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1626. 

  1627.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1628.                 "a" (&c->redDither)

  1629.                 );

  1630.                 return;

  1631.             }

  1632.         }

  1633.         else

  1634.         {

  1635.             switch(dstFormat)

  1636.             {

  1637.             case PIX_FMT_RGB32:

  1638.                 asm volatile(

  1639.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1640.                 "mov        %4, %%"REG_b"               \n\t"

  1641.                 "push %%"REG_BP"                        \n\t"

  1642.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1643.                 WRITEBGR32(%%REGb, 8280(%5), %%REGBP)

  1644.                 "pop %%"REG_BP"                         \n\t"

  1645.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1646. 

  1647.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1648.                 "a" (&c->redDither)

  1649.                 );

  1650.                 return;

  1651.             case PIX_FMT_BGR24:

  1652.                 asm volatile(

  1653.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1654.                 "mov        %4, %%"REG_b"               \n\t"

  1655.                 "push %%"REG_BP"                        \n\t"

  1656.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1657.                 WRITEBGR24(%%REGb, 8280(%5), %%REGBP)

  1658.                 "pop %%"REG_BP"                         \n\t"

  1659.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1660. 

  1661.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1662.                 "a" (&c->redDither)

  1663.                 );

  1664.                 return;

  1665.             case PIX_FMT_RGB555:

  1666.                 asm volatile(

  1667.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1668.                 "mov        %4, %%"REG_b"               \n\t"

  1669.                 "push %%"REG_BP"                        \n\t"

  1670.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1671.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1672. #ifdef DITHER1XBPP

  1673.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1674.                 "paddusb "MANGLE(g5Dither)", %%mm4      \n\t"

  1675.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1676. #endif

  1677.                 WRITERGB15(%%REGb, 8280(%5), %%REGBP)

  1678.                 "pop %%"REG_BP"                         \n\t"

  1679.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1680. 

  1681.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1682.                 "a" (&c->redDither)

  1683.                 );

  1684.                 return;

  1685.             case PIX_FMT_RGB565:

  1686.                 asm volatile(

  1687.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1688.                 "mov        %4, %%"REG_b"               \n\t"

  1689.                 "push %%"REG_BP"                        \n\t"

  1690.                 YSCALEYUV2RGB1b(%%REGBP, %5)

  1691.                 /* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */

  1692. #ifdef DITHER1XBPP

  1693.                 "paddusb "MANGLE(b5Dither)", %%mm2      \n\t"

  1694.                 "paddusb "MANGLE(g6Dither)", %%mm4      \n\t"

  1695.                 "paddusb "MANGLE(r5Dither)", %%mm5      \n\t"

  1696. #endif

  1697. 

  1698.                 WRITERGB16(%%REGb, 8280(%5), %%REGBP)

  1699.                 "pop %%"REG_BP"                         \n\t"

  1700.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1701. 

  1702.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1703.                 "a" (&c->redDither)

  1704.                 );

  1705.                 return;

  1706.             case PIX_FMT_YUYV422:

  1707.                 asm volatile(

  1708.                 "mov %%"REG_b", "ESP_OFFSET"(%5)        \n\t"

  1709.                 "mov        %4, %%"REG_b"               \n\t"

  1710.                 "push %%"REG_BP"                        \n\t"

  1711.                 YSCALEYUV2PACKED1b(%%REGBP, %5)

  1712.                 WRITEYUY2(%%REGb, 8280(%5), %%REGBP)

  1713.                 "pop %%"REG_BP"                         \n\t"

  1714.                 "mov "ESP_OFFSET"(%5), %%"REG_b"        \n\t"

  1715. 

  1716.                 :: "c" (buf0), "d" (buf1), "S" (uvbuf0), "D" (uvbuf1), "m" (dest),

  1717.                 "a" (&c->redDither)

  1718.                 );

  1719.                 return;

  1720.             }

  1721.         }

  1722.     }

  1723. #endif /* HAVE_MMX */

  1724.     if (uvalpha < 2048)

  1725.     {

  1726.         YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C, YSCALE_YUV_2_PACKED1_C, YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)

  1727.     }else{

  1728.         YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C, YSCALE_YUV_2_PACKED1B_C, YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)

  1729.     }

  1730. }

  1731. 

  1732. //FIXME yuy2* can read up to 7 samples too much

  1733. 

  1734. static inline void RENAME(yuy2ToY)(uint8_t *dst, uint8_t *src, long width)

  1735. {

  1736. #ifdef HAVE_MMX

  1737.     asm volatile(

  1738.     "movq "MANGLE(bm01010101)", %%mm2           \n\t"

  1739.     "mov                    %0, %%"REG_a"       \n\t"

  1740.     "1:                                         \n\t"

  1741.     "movq    (%1, %%"REG_a",2), %%mm0           \n\t"

  1742.     "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"

  1743.     "pand                %%mm2, %%mm0           \n\t"

  1744.     "pand                %%mm2, %%mm1           \n\t"

  1745.     "packuswb            %%mm1, %%mm0           \n\t"

  1746.     "movq                %%mm0, (%2, %%"REG_a") \n\t"

  1747.     "add                    $8, %%"REG_a"       \n\t"

  1748.     " js                    1b                  \n\t"

  1749.     : : "g" (-width), "r" (src+width*2), "r" (dst+width)

  1750.     : "%"REG_a

  1751.     );

  1752. #else

  1753.     int i;

  1754.     for (i=0; i<width; i++)

  1755.         dst[i]= src[2*i];

  1756. #endif

  1757. }

  1758. 

  1759. static inline void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  1760. {

  1761. #ifdef HAVE_MMX

  1762.     asm volatile(

  1763.     "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1764.     "mov                    %0, %%"REG_a"       \n\t"

  1765.     "1:                                         \n\t"

  1766.     "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1767.     "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1768.     "psrlw                  $8, %%mm0           \n\t"

  1769.     "psrlw                  $8, %%mm1           \n\t"

  1770.     "packuswb            %%mm1, %%mm0           \n\t"

  1771.     "movq                %%mm0, %%mm1           \n\t"

  1772.     "psrlw                  $8, %%mm0           \n\t"

  1773.     "pand                %%mm4, %%mm1           \n\t"

  1774.     "packuswb            %%mm0, %%mm0           \n\t"

  1775.     "packuswb            %%mm1, %%mm1           \n\t"

  1776.     "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1777.     "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1778.     "add                    $4, %%"REG_a"       \n\t"

  1779.     " js                    1b                  \n\t"

  1780.     : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1781.     : "%"REG_a

  1782.     );

  1783. #else

  1784.     int i;

  1785.     for (i=0; i<width; i++)

  1786.     {

  1787.         dstU[i]= src1[4*i + 1];

  1788.         dstV[i]= src1[4*i + 3];

  1789.     }

  1790. #endif

  1791.     assert(src1 == src2);

  1792. }

  1793. 

  1794. /* This is almost identical to the previous, end exists only because

  1795.  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */

  1796. static inline void RENAME(uyvyToY)(uint8_t *dst, uint8_t *src, long width)

  1797. {

  1798. #ifdef HAVE_MMX

  1799.     asm volatile(

  1800.     "mov                  %0, %%"REG_a"         \n\t"

  1801.     "1:                                         \n\t"

  1802.     "movq  (%1, %%"REG_a",2), %%mm0             \n\t"

  1803.     "movq 8(%1, %%"REG_a",2), %%mm1             \n\t"

  1804.     "psrlw                $8, %%mm0             \n\t"

  1805.     "psrlw                $8, %%mm1             \n\t"

  1806.     "packuswb          %%mm1, %%mm0             \n\t"

  1807.     "movq              %%mm0, (%2, %%"REG_a")   \n\t"

  1808.     "add                  $8, %%"REG_a"         \n\t"

  1809.     " js                  1b                    \n\t"

  1810.     : : "g" (-width), "r" (src+width*2), "r" (dst+width)

  1811.     : "%"REG_a

  1812.     );

  1813. #else

  1814.     int i;

  1815.     for (i=0; i<width; i++)

  1816.         dst[i]= src[2*i+1];

  1817. #endif

  1818. }

  1819. 

  1820. static inline void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  1821. {

  1822. #ifdef HAVE_MMX

  1823.     asm volatile(

  1824.     "movq "MANGLE(bm01010101)", %%mm4           \n\t"

  1825.     "mov                    %0, %%"REG_a"       \n\t"

  1826.     "1:                                         \n\t"

  1827.     "movq    (%1, %%"REG_a",4), %%mm0           \n\t"

  1828.     "movq   8(%1, %%"REG_a",4), %%mm1           \n\t"

  1829.     "pand                %%mm4, %%mm0           \n\t"

  1830.     "pand                %%mm4, %%mm1           \n\t"

  1831.     "packuswb            %%mm1, %%mm0           \n\t"

  1832.     "movq                %%mm0, %%mm1           \n\t"

  1833.     "psrlw                  $8, %%mm0           \n\t"

  1834.     "pand                %%mm4, %%mm1           \n\t"

  1835.     "packuswb            %%mm0, %%mm0           \n\t"

  1836.     "packuswb            %%mm1, %%mm1           \n\t"

  1837.     "movd                %%mm0, (%3, %%"REG_a") \n\t"

  1838.     "movd                %%mm1, (%2, %%"REG_a") \n\t"

  1839.     "add                    $4, %%"REG_a"       \n\t"

  1840.     " js                    1b                  \n\t"

  1841.     : : "g" (-width), "r" (src1+width*4), "r" (dstU+width), "r" (dstV+width)

  1842.     : "%"REG_a

  1843.     );

  1844. #else

  1845.     int i;

  1846.     for (i=0; i<width; i++)

  1847.     {

  1848.         dstU[i]= src1[4*i + 0];

  1849.         dstV[i]= src1[4*i + 2];

  1850.     }

  1851. #endif

  1852.     assert(src1 == src2);

  1853. }

  1854. 

  1855. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1856. static inline void RENAME(name)(uint8_t *dst, uint8_t *src, long width)\

  1857. {\

  1858.     int i;\

  1859.     for (i=0; i<width; i++)\

  1860.     {\

  1861.         int b= (((type*)src)[i]>>shb)&maskb;\

  1862.         int g= (((type*)src)[i]>>shg)&maskg;\

  1863.         int r= (((type*)src)[i]>>shr)&maskr;\

  1864. \

  1865.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1866.     }\

  1867. }

  1868. 

  1869. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1870. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1871. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1872. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1873. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1874. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1875. 

  1876. #define BGR2UV(type, name, shr, shg, shb, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\

  1877. static inline void RENAME(name)(uint8_t *dstU, uint8_t *dstV, uint8_t *src, uint8_t *dummy, long width)\

  1878. {\

  1879.     int i;\

  1880.     for (i=0; i<width; i++)\

  1881.     {\

  1882.         int b= (((type*)src)[i]&maskb)>>shb;\

  1883.         int g= (((type*)src)[i]&maskg)>>shg;\

  1884.         int r= (((type*)src)[i]&maskr)>>shr;\

  1885. \

  1886.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1887.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1888.     }\

  1889. }\

  1890. static inline void RENAME(name ## _half)(uint8_t *dstU, uint8_t *dstV, uint8_t *src, uint8_t *dummy, long width)\

  1891. {\

  1892.     int i;\

  1893.     for (i=0; i<width; i++)\

  1894.     {\

  1895.         int pix0= ((type*)src)[2*i+0];\

  1896.         int pix1= ((type*)src)[2*i+1];\

  1897.         int g= (pix0&maskg)+(pix1&maskg);\

  1898.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1899.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1900. \

  1901.         g>>=shg;\

  1902. \

  1903.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1904.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1905.     }\

  1906. }

  1907. 

  1908. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1909. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1910. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1911. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1912. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1913. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1914. 

  1915. #ifdef HAVE_MMX

  1916. static inline void RENAME(bgr24ToY_mmx)(uint8_t *dst, uint8_t *src, long width, int srcFormat)

  1917. {

  1918. 

  1919.     if(srcFormat == PIX_FMT_BGR24){

  1920.         asm volatile(

  1921.             "movq  "MANGLE(ff_bgr24toY1Coeff)", %%mm5       \n\t"

  1922.             "movq  "MANGLE(ff_bgr24toY2Coeff)", %%mm6       \n\t"

  1923.             :

  1924.         );

  1925.     }else{

  1926.         asm volatile(

  1927.             "movq  "MANGLE(ff_rgb24toY1Coeff)", %%mm5       \n\t"

  1928.             "movq  "MANGLE(ff_rgb24toY2Coeff)", %%mm6       \n\t"

  1929.             :

  1930.         );

  1931.     }

  1932. 

  1933.     asm volatile(

  1934.         "movq  "MANGLE(ff_bgr24toYOffset)", %%mm4   \n\t"

  1935.         "mov                        %2, %%"REG_a"   \n\t"

  1936.         "pxor                    %%mm7, %%mm7       \n\t"

  1937.         "1:                                         \n\t"

  1938.         PREFETCH"               64(%0)              \n\t"

  1939.         "movd                     (%0), %%mm0       \n\t"

  1940.         "movd                    2(%0), %%mm1       \n\t"

  1941.         "movd                    6(%0), %%mm2       \n\t"

  1942.         "movd                    8(%0), %%mm3       \n\t"

  1943.         "add                       $12, %0          \n\t"

  1944.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1945.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1946.         "punpcklbw               %%mm7, %%mm2       \n\t"

  1947.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1948.         "pmaddwd                 %%mm5, %%mm0       \n\t"

  1949.         "pmaddwd                 %%mm6, %%mm1       \n\t"

  1950.         "pmaddwd                 %%mm5, %%mm2       \n\t"

  1951.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1952.         "paddd                   %%mm1, %%mm0       \n\t"

  1953.         "paddd                   %%mm3, %%mm2       \n\t"

  1954.         "paddd                   %%mm4, %%mm0       \n\t"

  1955.         "paddd                   %%mm4, %%mm2       \n\t"

  1956.         "psrad                     $15, %%mm0       \n\t"

  1957.         "psrad                     $15, %%mm2       \n\t"

  1958.         "packssdw                %%mm2, %%mm0       \n\t"

  1959.         "packuswb                %%mm0, %%mm0       \n\t"

  1960.         "movd                %%mm0, (%1, %%"REG_a") \n\t"

  1961.         "add                        $4, %%"REG_a"   \n\t"

  1962.         " js                        1b              \n\t"

  1963.     : "+r" (src)

  1964.     : "r" (dst+width), "g" (-width)

  1965.     : "%"REG_a

  1966.     );

  1967. }

  1968. 

  1969. static inline void RENAME(bgr24ToUV_mmx)(uint8_t *dstU, uint8_t *dstV, uint8_t *src, long width, int srcFormat)

  1970. {

  1971.     asm volatile(

  1972.         "movq                    24+%4, %%mm6       \n\t"

  1973.         "mov                        %3, %%"REG_a"   \n\t"

  1974.         "pxor                    %%mm7, %%mm7       \n\t"

  1975.         "1:                                         \n\t"

  1976.         PREFETCH"               64(%0)              \n\t"

  1977.         "movd                     (%0), %%mm0       \n\t"

  1978.         "movd                    2(%0), %%mm1       \n\t"

  1979.         "punpcklbw               %%mm7, %%mm0       \n\t"

  1980.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1981.         "movq                    %%mm0, %%mm2       \n\t"

  1982.         "movq                    %%mm1, %%mm3       \n\t"

  1983.         "pmaddwd                    %4, %%mm0       \n\t"

  1984.         "pmaddwd                  8+%4, %%mm1       \n\t"

  1985.         "pmaddwd                 16+%4, %%mm2       \n\t"

  1986.         "pmaddwd                 %%mm6, %%mm3       \n\t"

  1987.         "paddd                   %%mm1, %%mm0       \n\t"

  1988.         "paddd                   %%mm3, %%mm2       \n\t"

  1989. 

  1990.         "movd                    6(%0), %%mm1       \n\t"

  1991.         "movd                    8(%0), %%mm3       \n\t"

  1992.         "add                       $12, %0          \n\t"

  1993.         "punpcklbw               %%mm7, %%mm1       \n\t"

  1994.         "punpcklbw               %%mm7, %%mm3       \n\t"

  1995.         "movq                    %%mm1, %%mm4       \n\t"

  1996.         "movq                    %%mm3, %%mm5       \n\t"

  1997.         "pmaddwd                    %4, %%mm1       \n\t"

  1998.         "pmaddwd                  8+%4, %%mm3       \n\t"

  1999.         "pmaddwd                 16+%4, %%mm4       \n\t"

  2000.         "pmaddwd                 %%mm6, %%mm5       \n\t"

  2001.         "paddd                   %%mm3, %%mm1       \n\t"

  2002.         "paddd                   %%mm5, %%mm4       \n\t"

  2003. 

  2004.         "movq "MANGLE(ff_bgr24toUVOffset)", %%mm3       \n\t"

  2005.         "paddd                   %%mm3, %%mm0       \n\t"

  2006.         "paddd                   %%mm3, %%mm2       \n\t"

  2007.         "paddd                   %%mm3, %%mm1       \n\t"

  2008.         "paddd                   %%mm3, %%mm4       \n\t"

  2009.         "psrad                     $15, %%mm0       \n\t"

  2010.         "psrad                     $15, %%mm2       \n\t"

  2011.         "psrad                     $15, %%mm1       \n\t"

  2012.         "psrad                     $15, %%mm4       \n\t"

  2013.         "packssdw                %%mm1, %%mm0       \n\t"

  2014.         "packssdw                %%mm4, %%mm2       \n\t"

  2015.         "packuswb                %%mm0, %%mm0       \n\t"

  2016.         "packuswb                %%mm2, %%mm2       \n\t"

  2017.         "movd                %%mm0, (%1, %%"REG_a") \n\t"

  2018.         "movd                %%mm2, (%2, %%"REG_a") \n\t"

  2019.         "add                        $4, %%"REG_a"   \n\t"

  2020.         " js                        1b              \n\t"

  2021.     : "+r" (src)

  2022.     : "r" (dstU+width), "r" (dstV+width), "g" (-width), "m"(ff_bgr24toUV[srcFormat == PIX_FMT_RGB24][0])

  2023.     : "%"REG_a

  2024.     );

  2025. }

  2026. #endif

  2027. 

  2028. static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, long width)

  2029. {

  2030. #ifdef HAVE_MMX

  2031.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_BGR24);

  2032. #else

  2033.     int i;

  2034.     for (i=0; i<width; i++)

  2035.     {

  2036.         int b= src[i*3+0];

  2037.         int g= src[i*3+1];

  2038.         int r= src[i*3+2];

  2039. 

  2040.         dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2041.     }

  2042. #endif /* HAVE_MMX */

  2043. }

  2044. 

  2045. static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  2046. {

  2047. #ifdef HAVE_MMX

  2048.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_BGR24);

  2049. #else

  2050.     int i;

  2051.     for (i=0; i<width; i++)

  2052.     {

  2053.         int b= src1[3*i + 0];

  2054.         int g= src1[3*i + 1];

  2055.         int r= src1[3*i + 2];

  2056. 

  2057.         dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;

  2058.         dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;

  2059.     }

  2060. #endif /* HAVE_MMX */

  2061.     assert(src1 == src2);

  2062. }

  2063. 

  2064. static inline void RENAME(bgr24ToUV_half)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  2065. {

  2066.     int i;

  2067.     for (i=0; i<width; i++)

  2068.     {

  2069.         int b= src1[6*i + 0] + src1[6*i + 3];

  2070.         int g= src1[6*i + 1] + src1[6*i + 4];

  2071.         int r= src1[6*i + 2] + src1[6*i + 5];

  2072. 

  2073.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);

  2074.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);

  2075.     }

  2076.     assert(src1 == src2);

  2077. }

  2078. 

  2079. static inline void RENAME(rgb24ToY)(uint8_t *dst, uint8_t *src, long width)

  2080. {

  2081. #ifdef HAVE_MMX

  2082.     RENAME(bgr24ToY_mmx)(dst, src, width, PIX_FMT_RGB24);

  2083. #else

  2084.     int i;

  2085.     for (i=0; i<width; i++)

  2086.     {

  2087.         int r= src[i*3+0];

  2088.         int g= src[i*3+1];

  2089.         int b= src[i*3+2];

  2090. 

  2091.         dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2092.     }

  2093. #endif

  2094. }

  2095. 

  2096. static inline void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  2097. {

  2098.     int i;

  2099.     assert(src1==src2);

  2100. #ifdef HAVE_MMX

  2101.     RENAME(bgr24ToUV_mmx)(dstU, dstV, src1, width, PIX_FMT_RGB24);

  2102. #else

  2103.     for (i=0; i<width; i++)

  2104.     {

  2105.         int r= src1[3*i + 0];

  2106.         int g= src1[3*i + 1];

  2107.         int b= src1[3*i + 2];

  2108. 

  2109.         dstU[i]= (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;

  2110.         dstV[i]= (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT;

  2111.     }

  2112. #endif

  2113. }

  2114. 

  2115. static inline void RENAME(rgb24ToUV_half)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width)

  2116. {

  2117.     int i;

  2118.     assert(src1==src2);

  2119.     for (i=0; i<width; i++)

  2120.     {

  2121.         int r= src1[6*i + 0] + src1[6*i + 3];

  2122.         int g= src1[6*i + 1] + src1[6*i + 4];

  2123.         int b= src1[6*i + 2] + src1[6*i + 5];

  2124. 

  2125.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);

  2126.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT))>>(RGB2YUV_SHIFT+1);

  2127.     }

  2128. }

  2129. 

  2130. 

  2131. static inline void RENAME(palToY)(uint8_t *dst, uint8_t *src, long width, uint32_t *pal)

  2132. {

  2133.     int i;

  2134.     for (i=0; i<width; i++)

  2135.     {

  2136.         int d= src[i];

  2137. 

  2138.         dst[i]= pal[d] & 0xFF;

  2139.     }

  2140. }

  2141. 

  2142. static inline void RENAME(palToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, long width, uint32_t *pal)

  2143. {

  2144.     int i;

  2145.     assert(src1 == src2);

  2146.     for (i=0; i<width; i++)

  2147.     {

  2148.         int p= pal[src1[i]];

  2149. 

  2150.         dstU[i]= p>>8;

  2151.         dstV[i]= p>>16;

  2152.     }

  2153. }

  2154. 

  2155. static inline void RENAME(mono2Y)(uint8_t *dst, uint8_t *src, long width, int format)

  2156. {

  2157.     int i, j;

  2158.     for (i=0; i<width/8; i++){

  2159.         int d= format == PIX_FMT_MONOBLACK ? src[i] : ~src[i];

  2160.         for(j=0; j<8; j++)

  2161.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  2162.     }

  2163. }

  2164. 

  2165. // bilinear / bicubic scaling

  2166. static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc,

  2167.                                   int16_t *filter, int16_t *filterPos, long filterSize)

  2168. {

  2169. #ifdef HAVE_MMX

  2170.     assert(filterSize % 4 == 0 && filterSize>0);

  2171.     if (filterSize==4) // Always true for upscaling, sometimes for down, too.

  2172.     {

  2173.         long counter= -2*dstW;

  2174.         filter-= counter*2;

  2175.         filterPos-= counter/2;

  2176.         dst-= counter/2;

  2177.         asm volatile(

  2178. #if defined(PIC)

  2179.         "push            %%"REG_b"              \n\t"

  2180. #endif

  2181.         "pxor                %%mm7, %%mm7       \n\t"

  2182.         "push           %%"REG_BP"              \n\t" // we use 7 regs here ...

  2183.         "mov             %%"REG_a", %%"REG_BP"  \n\t"

  2184.         ASMALIGN(4)

  2185.         "1:                                     \n\t"

  2186.         "movzwl   (%2, %%"REG_BP"), %%eax       \n\t"

  2187.         "movzwl  2(%2, %%"REG_BP"), %%ebx       \n\t"

  2188.         "movq  (%1, %%"REG_BP", 4), %%mm1       \n\t"

  2189.         "movq 8(%1, %%"REG_BP", 4), %%mm3       \n\t"

  2190.         "movd      (%3, %%"REG_a"), %%mm0       \n\t"

  2191.         "movd      (%3, %%"REG_b"), %%mm2       \n\t"

  2192.         "punpcklbw           %%mm7, %%mm0       \n\t"

  2193.         "punpcklbw           %%mm7, %%mm2       \n\t"

  2194.         "pmaddwd             %%mm1, %%mm0       \n\t"

  2195.         "pmaddwd             %%mm2, %%mm3       \n\t"

  2196.         "movq                %%mm0, %%mm4       \n\t"

  2197.         "punpckldq           %%mm3, %%mm0       \n\t"

  2198.         "punpckhdq           %%mm3, %%mm4       \n\t"

  2199.         "paddd               %%mm4, %%mm0       \n\t"

  2200.         "psrad                  $7, %%mm0       \n\t"

  2201.         "packssdw            %%mm0, %%mm0       \n\t"

  2202.         "movd                %%mm0, (%4, %%"REG_BP")    \n\t"

  2203.         "add                    $4, %%"REG_BP"  \n\t"

  2204.         " jnc                   1b              \n\t"

  2205. 

  2206.         "pop            %%"REG_BP"              \n\t"

  2207. #if defined(PIC)

  2208.         "pop             %%"REG_b"              \n\t"

  2209. #endif

  2210.         : "+a" (counter)

  2211.         : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)

  2212. #if !defined(PIC)

  2213.         : "%"REG_b

  2214. #endif

  2215.         );

  2216.     }

  2217.     else if (filterSize==8)

  2218.     {

  2219.         long counter= -2*dstW;

  2220.         filter-= counter*4;

  2221.         filterPos-= counter/2;

  2222.         dst-= counter/2;

  2223.         asm volatile(

  2224. #if defined(PIC)

  2225.         "push             %%"REG_b"             \n\t"

  2226. #endif

  2227.         "pxor                 %%mm7, %%mm7      \n\t"

  2228.         "push            %%"REG_BP"             \n\t" // we use 7 regs here ...

  2229.         "mov              %%"REG_a", %%"REG_BP" \n\t"

  2230.         ASMALIGN(4)

  2231.         "1:                                     \n\t"

  2232.         "movzwl    (%2, %%"REG_BP"), %%eax      \n\t"

  2233.         "movzwl   2(%2, %%"REG_BP"), %%ebx      \n\t"

  2234.         "movq   (%1, %%"REG_BP", 8), %%mm1      \n\t"

  2235.         "movq 16(%1, %%"REG_BP", 8), %%mm3      \n\t"

  2236.         "movd       (%3, %%"REG_a"), %%mm0      \n\t"

  2237.         "movd       (%3, %%"REG_b"), %%mm2      \n\t"

  2238.         "punpcklbw            %%mm7, %%mm0      \n\t"

  2239.         "punpcklbw            %%mm7, %%mm2      \n\t"

  2240.         "pmaddwd              %%mm1, %%mm0      \n\t"

  2241.         "pmaddwd              %%mm2, %%mm3      \n\t"

  2242. 

  2243.         "movq  8(%1, %%"REG_BP", 8), %%mm1      \n\t"

  2244.         "movq 24(%1, %%"REG_BP", 8), %%mm5      \n\t"

  2245.         "movd      4(%3, %%"REG_a"), %%mm4      \n\t"

  2246.         "movd      4(%3, %%"REG_b"), %%mm2      \n\t"

  2247.         "punpcklbw            %%mm7, %%mm4      \n\t"

  2248.         "punpcklbw            %%mm7, %%mm2      \n\t"

  2249.         "pmaddwd              %%mm1, %%mm4      \n\t"

  2250.         "pmaddwd              %%mm2, %%mm5      \n\t"

  2251.         "paddd                %%mm4, %%mm0      \n\t"

  2252.         "paddd                %%mm5, %%mm3      \n\t"

  2253.         "movq                 %%mm0, %%mm4      \n\t"

  2254.         "punpckldq            %%mm3, %%mm0      \n\t"

  2255.         "punpckhdq            %%mm3, %%mm4      \n\t"

  2256.         "paddd                %%mm4, %%mm0      \n\t"

  2257.         "psrad                   $7, %%mm0      \n\t"

  2258.         "packssdw             %%mm0, %%mm0      \n\t"

  2259.         "movd                 %%mm0, (%4, %%"REG_BP")   \n\t"

  2260.         "add                     $4, %%"REG_BP" \n\t"

  2261.         " jnc                    1b             \n\t"

  2262. 

  2263.         "pop             %%"REG_BP"             \n\t"

  2264. #if defined(PIC)

  2265.         "pop              %%"REG_b"             \n\t"

  2266. #endif

  2267.         : "+a" (counter)

  2268.         : "c" (filter), "d" (filterPos), "S" (src), "D" (dst)

  2269. #if !defined(PIC)

  2270.         : "%"REG_b

  2271. #endif

  2272.         );

  2273.     }

  2274.     else

  2275.     {

  2276.         uint8_t *offset = src+filterSize;

  2277.         long counter= -2*dstW;

  2278.         //filter-= counter*filterSize/2;

  2279.         filterPos-= counter/2;

  2280.         dst-= counter/2;

  2281.         asm volatile(

  2282.         "pxor                  %%mm7, %%mm7     \n\t"

  2283.         ASMALIGN(4)

  2284.         "1:                                     \n\t"

  2285.         "mov                      %2, %%"REG_c" \n\t"

  2286.         "movzwl      (%%"REG_c", %0), %%eax     \n\t"

  2287.         "movzwl     2(%%"REG_c", %0), %%edx     \n\t"

  2288.         "mov                      %5, %%"REG_c" \n\t"

  2289.         "pxor                  %%mm4, %%mm4     \n\t"

  2290.         "pxor                  %%mm5, %%mm5     \n\t"

  2291.         "2:                                     \n\t"

  2292.         "movq                   (%1), %%mm1     \n\t"

  2293.         "movq               (%1, %6), %%mm3     \n\t"

  2294.         "movd (%%"REG_c", %%"REG_a"), %%mm0     \n\t"

  2295.         "movd (%%"REG_c", %%"REG_d"), %%mm2     \n\t"

  2296.         "punpcklbw             %%mm7, %%mm0     \n\t"

  2297.         "punpcklbw             %%mm7, %%mm2     \n\t"

  2298.         "pmaddwd               %%mm1, %%mm0     \n\t"

  2299.         "pmaddwd               %%mm2, %%mm3     \n\t"

  2300.         "paddd                 %%mm3, %%mm5     \n\t"

  2301.         "paddd                 %%mm0, %%mm4     \n\t"

  2302.         "add                      $8, %1        \n\t"

  2303.         "add                      $4, %%"REG_c" \n\t"

  2304.         "cmp                      %4, %%"REG_c" \n\t"

  2305.         " jb                      2b            \n\t"

  2306.         "add                      %6, %1        \n\t"

  2307.         "movq                  %%mm4, %%mm0     \n\t"

  2308.         "punpckldq             %%mm5, %%mm4     \n\t"

  2309.         "punpckhdq             %%mm5, %%mm0     \n\t"

  2310.         "paddd                 %%mm0, %%mm4     \n\t"

  2311.         "psrad                    $7, %%mm4     \n\t"

  2312.         "packssdw              %%mm4, %%mm4     \n\t"

  2313.         "mov                      %3, %%"REG_a" \n\t"

  2314.         "movd                  %%mm4, (%%"REG_a", %0)   \n\t"

  2315.         "add                      $4, %0        \n\t"

  2316.         " jnc                     1b            \n\t"

  2317. 

  2318.         : "+r" (counter), "+r" (filter)

  2319.         : "m" (filterPos), "m" (dst), "m"(offset),

  2320.           "m" (src), "r" (filterSize*2)

  2321.         : "%"REG_a, "%"REG_c, "%"REG_d

  2322.         );

  2323.     }

  2324. #else

  2325. #ifdef HAVE_ALTIVEC

  2326.     hScale_altivec_real(dst, dstW, src, srcW, xInc, filter, filterPos, filterSize);

  2327. #else

  2328.     int i;

  2329.     for (i=0; i<dstW; i++)

  2330.     {

  2331.         int j;

  2332.         int srcPos= filterPos[i];

  2333.         int val=0;

  2334.         //printf("filterPos: %d\n", filterPos[i]);

  2335.         for (j=0; j<filterSize; j++)

  2336.         {

  2337.             //printf("filter: %d, src: %d\n", filter[i], src[srcPos + j]);

  2338.             val += ((int)src[srcPos + j])*filter[filterSize*i + j];

  2339.         }

  2340.         //filter += hFilterSize;

  2341.         dst[i] = FFMIN(val>>7, (1<<15)-1); // the cubic equation does overflow ...

  2342.         //dst[i] = val>>7;

  2343.     }

  2344. #endif /* HAVE_ALTIVEC */

  2345. #endif /* HAVE_MMX */

  2346. }

  2347.       // *** horizontal scale Y line to temp buffer

  2348. static inline void RENAME(hyscale)(SwsContext *c, uint16_t *dst, long dstWidth, uint8_t *src, int srcW, int xInc,

  2349.                                    int flags, int canMMX2BeUsed, int16_t *hLumFilter,

  2350.                                    int16_t *hLumFilterPos, int hLumFilterSize, void *funnyYCode,

  2351.                                    int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter,

  2352.                                    int32_t *mmx2FilterPos, uint8_t *pal)

  2353. {

  2354.     if (srcFormat==PIX_FMT_YUYV422 || srcFormat==PIX_FMT_GRAY16BE)

  2355.     {

  2356.         RENAME(yuy2ToY)(formatConvBuffer, src, srcW);

  2357.         src= formatConvBuffer;

  2358.     }

  2359.     else if (srcFormat==PIX_FMT_UYVY422 || srcFormat==PIX_FMT_GRAY16LE)

  2360.     {

  2361.         RENAME(uyvyToY)(formatConvBuffer, src, srcW);

  2362.         src= formatConvBuffer;

  2363.     }

  2364.     else if (srcFormat==PIX_FMT_RGB32)

  2365.     {

  2366.         RENAME(bgr32ToY)(formatConvBuffer, src, srcW);

  2367.         src= formatConvBuffer;

  2368.     }

  2369.     else if (srcFormat==PIX_FMT_RGB32_1)

  2370.     {

  2371.         RENAME(bgr32ToY)(formatConvBuffer, src+ALT32_CORR, srcW);

  2372.         src= formatConvBuffer;

  2373.     }

  2374.     else if (srcFormat==PIX_FMT_BGR24)

  2375.     {

  2376.         RENAME(bgr24ToY)(formatConvBuffer, src, srcW);

  2377.         src= formatConvBuffer;

  2378.     }

  2379.     else if (srcFormat==PIX_FMT_BGR565)

  2380.     {

  2381.         RENAME(bgr16ToY)(formatConvBuffer, src, srcW);

  2382.         src= formatConvBuffer;

  2383.     }

  2384.     else if (srcFormat==PIX_FMT_BGR555)

  2385.     {

  2386.         RENAME(bgr15ToY)(formatConvBuffer, src, srcW);

  2387.         src= formatConvBuffer;

  2388.     }

  2389.     else if (srcFormat==PIX_FMT_BGR32)

  2390.     {

  2391.         RENAME(rgb32ToY)(formatConvBuffer, src, srcW);

  2392.         src= formatConvBuffer;

  2393.     }

  2394.     else if (srcFormat==PIX_FMT_BGR32_1)

  2395.     {

  2396.         RENAME(rgb32ToY)(formatConvBuffer, src+ALT32_CORR, srcW);

  2397.         src= formatConvBuffer;

  2398.     }

  2399.     else if (srcFormat==PIX_FMT_RGB24)

  2400.     {

  2401.         RENAME(rgb24ToY)(formatConvBuffer, src, srcW);

  2402.         src= formatConvBuffer;

  2403.     }

  2404.     else if (srcFormat==PIX_FMT_RGB565)

  2405.     {

  2406.         RENAME(rgb16ToY)(formatConvBuffer, src, srcW);

  2407.         src= formatConvBuffer;

  2408.     }

  2409.     else if (srcFormat==PIX_FMT_RGB555)

  2410.     {

  2411.         RENAME(rgb15ToY)(formatConvBuffer, src, srcW);

  2412.         src= formatConvBuffer;

  2413.     }

  2414.     else if (srcFormat==PIX_FMT_RGB8 || srcFormat==PIX_FMT_BGR8 || srcFormat==PIX_FMT_PAL8 || srcFormat==PIX_FMT_BGR4_BYTE  || srcFormat==PIX_FMT_RGB4_BYTE)

  2415.     {

  2416.         RENAME(palToY)(formatConvBuffer, src, srcW, (uint32_t*)pal);

  2417.         src= formatConvBuffer;

  2418.     }

  2419.     else if (srcFormat==PIX_FMT_MONOBLACK ||srcFormat==PIX_FMT_MONOWHITE)

  2420.     {

  2421.         RENAME(mono2Y)(formatConvBuffer, src, srcW, srcFormat);

  2422.         src= formatConvBuffer;

  2423.     }

  2424. 

  2425. #ifdef HAVE_MMX

  2426.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  2427.     if (!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed))

  2428. #else

  2429.     if (!(flags&SWS_FAST_BILINEAR))

  2430. #endif

  2431.     {

  2432.         RENAME(hScale)(dst, dstWidth, src, srcW, xInc, hLumFilter, hLumFilterPos, hLumFilterSize);

  2433.     }

  2434.     else // fast bilinear upscale / crap downscale

  2435.     {

  2436. #if defined(ARCH_X86)

  2437. #ifdef HAVE_MMX2

  2438.         int i;

  2439. #if defined(PIC)

  2440.         uint64_t ebxsave __attribute__((aligned(8)));

  2441. #endif

  2442.         if (canMMX2BeUsed)

  2443.         {

  2444.             asm volatile(

  2445. #if defined(PIC)

  2446.             "mov               %%"REG_b", %5        \n\t"

  2447. #endif

  2448.             "pxor                  %%mm7, %%mm7     \n\t"

  2449.             "mov                      %0, %%"REG_c" \n\t"

  2450.             "mov                      %1, %%"REG_D" \n\t"

  2451.             "mov                      %2, %%"REG_d" \n\t"

  2452.             "mov                      %3, %%"REG_b" \n\t"

  2453.             "xor               %%"REG_a", %%"REG_a" \n\t" // i

  2454.             PREFETCH"        (%%"REG_c")            \n\t"

  2455.             PREFETCH"      32(%%"REG_c")            \n\t"

  2456.             PREFETCH"      64(%%"REG_c")            \n\t"

  2457. 

  2458. #ifdef ARCH_X86_64

  2459. 

  2460. #define FUNNY_Y_CODE \

  2461.             "movl            (%%"REG_b"), %%esi     \n\t"\

  2462.             "call                    *%4            \n\t"\

  2463.             "movl (%%"REG_b", %%"REG_a"), %%esi     \n\t"\

  2464.             "add               %%"REG_S", %%"REG_c" \n\t"\

  2465.             "add               %%"REG_a", %%"REG_D" \n\t"\

  2466.             "xor               %%"REG_a", %%"REG_a" \n\t"\

  2467. 

  2468. #else

  2469. 

  2470. #define FUNNY_Y_CODE \

  2471.             "movl (%%"REG_b"), %%esi        \n\t"\

  2472.             "call         *%4                       \n\t"\

  2473.             "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  2474.             "add               %%"REG_a", %%"REG_D" \n\t"\

  2475.             "xor               %%"REG_a", %%"REG_a" \n\t"\

  2476. 

  2477. #endif /* ARCH_X86_64 */

  2478. 

  2479. FUNNY_Y_CODE

  2480. FUNNY_Y_CODE

  2481. FUNNY_Y_CODE

  2482. FUNNY_Y_CODE

  2483. FUNNY_Y_CODE

  2484. FUNNY_Y_CODE

  2485. FUNNY_Y_CODE

  2486. FUNNY_Y_CODE

  2487. 

  2488. #if defined(PIC)

  2489.             "mov                      %5, %%"REG_b" \n\t"

  2490. #endif

  2491.             :: "m" (src), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),

  2492.             "m" (funnyYCode)

  2493. #if defined(PIC)

  2494.             ,"m" (ebxsave)

  2495. #endif

  2496.             : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2497. #if !defined(PIC)

  2498.             ,"%"REG_b

  2499. #endif

  2500.             );

  2501.             for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) dst[i] = src[srcW-1]*128;

  2502.         }

  2503.         else

  2504.         {

  2505. #endif /* HAVE_MMX2 */

  2506.         long xInc_shr16 = xInc >> 16;

  2507.         uint16_t xInc_mask = xInc & 0xffff;

  2508.         //NO MMX just normal asm ...

  2509.         asm volatile(

  2510.         "xor %%"REG_a", %%"REG_a"            \n\t" // i

  2511.         "xor %%"REG_d", %%"REG_d"            \n\t" // xx

  2512.         "xorl    %%ecx, %%ecx                \n\t" // 2*xalpha

  2513.         ASMALIGN(4)

  2514.         "1:                                  \n\t"

  2515.         "movzbl    (%0, %%"REG_d"), %%edi    \n\t" //src[xx]

  2516.         "movzbl   1(%0, %%"REG_d"), %%esi    \n\t" //src[xx+1]

  2517.         "subl    %%edi, %%esi                \n\t" //src[xx+1] - src[xx]

  2518.         "imull   %%ecx, %%esi                \n\t" //(src[xx+1] - src[xx])*2*xalpha

  2519.         "shll      $16, %%edi                \n\t"

  2520.         "addl    %%edi, %%esi                \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha)

  2521.         "mov        %1, %%"REG_D"            \n\t"

  2522.         "shrl       $9, %%esi                \n\t"

  2523.         "movw     %%si, (%%"REG_D", %%"REG_a", 2)   \n\t"

  2524.         "addw       %4, %%cx                 \n\t" //2*xalpha += xInc&0xFF

  2525.         "adc        %3, %%"REG_d"            \n\t" //xx+= xInc>>8 + carry

  2526. 

  2527.         "movzbl    (%0, %%"REG_d"), %%edi    \n\t" //src[xx]

  2528.         "movzbl   1(%0, %%"REG_d"), %%esi    \n\t" //src[xx+1]

  2529.         "subl    %%edi, %%esi                \n\t" //src[xx+1] - src[xx]

  2530.         "imull   %%ecx, %%esi                \n\t" //(src[xx+1] - src[xx])*2*xalpha

  2531.         "shll      $16, %%edi                \n\t"

  2532.         "addl    %%edi, %%esi                \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha)

  2533.         "mov        %1, %%"REG_D"            \n\t"

  2534.         "shrl       $9, %%esi                \n\t"

  2535.         "movw     %%si, 2(%%"REG_D", %%"REG_a", 2)  \n\t"

  2536.         "addw       %4, %%cx                 \n\t" //2*xalpha += xInc&0xFF

  2537.         "adc        %3, %%"REG_d"            \n\t" //xx+= xInc>>8 + carry

  2538. 

  2539. 

  2540.         "add        $2, %%"REG_a"            \n\t"

  2541.         "cmp        %2, %%"REG_a"            \n\t"

  2542.         " jb        1b                       \n\t"

  2543. 

  2544. 

  2545.         :: "r" (src), "m" (dst), "m" (dstWidth), "m" (xInc_shr16), "m" (xInc_mask)

  2546.         : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi"

  2547.         );

  2548. #ifdef HAVE_MMX2

  2549.         } //if MMX2 can't be used

  2550. #endif

  2551. #else

  2552.         int i;

  2553.         unsigned int xpos=0;

  2554.         for (i=0;i<dstWidth;i++)

  2555.         {

  2556.             register unsigned int xx=xpos>>16;

  2557.             register unsigned int xalpha=(xpos&0xFFFF)>>9;

  2558.             dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;

  2559.             xpos+=xInc;

  2560.         }

  2561. #endif /* defined(ARCH_X86) */

  2562.     }

  2563. 

  2564.     if(c->srcRange != c->dstRange && !(isRGB(c->dstFormat) || isBGR(c->dstFormat))){

  2565.         int i;

  2566.         //FIXME all pal and rgb srcFormats could do this convertion as well

  2567.         //FIXME all scalers more complex than bilinear could do half of this transform

  2568.         if(c->srcRange){

  2569.             for (i=0; i<dstWidth; i++)

  2570.                 dst[i]= (dst[i]*14071 + 33561947)>>14;

  2571.         }else{

  2572.             for (i=0; i<dstWidth; i++)

  2573.                 dst[i]= (FFMIN(dst[i],30189)*19077 - 39057361)>>14;

  2574.         }

  2575.     }

  2576. }

  2577. 

  2578. inline static void RENAME(hcscale)(SwsContext *c, uint16_t *dst, long dstWidth, uint8_t *src1, uint8_t *src2,

  2579.                                    int srcW, int xInc, int flags, int canMMX2BeUsed, int16_t *hChrFilter,

  2580.                                    int16_t *hChrFilterPos, int hChrFilterSize, void *funnyUVCode,

  2581.                                    int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter,

  2582.                                    int32_t *mmx2FilterPos, uint8_t *pal)

  2583. {

  2584.     if (srcFormat==PIX_FMT_YUYV422)

  2585.     {

  2586.         RENAME(yuy2ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2587.         src1= formatConvBuffer;

  2588.         src2= formatConvBuffer+VOFW;

  2589.     }

  2590.     else if (srcFormat==PIX_FMT_UYVY422)

  2591.     {

  2592.         RENAME(uyvyToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2593.         src1= formatConvBuffer;

  2594.         src2= formatConvBuffer+VOFW;

  2595.     }

  2596.     else if (srcFormat==PIX_FMT_RGB32)

  2597.     {

  2598.         if(c->chrSrcHSubSample)

  2599.             RENAME(bgr32ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2600.         else

  2601.             RENAME(bgr32ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2602.         src1= formatConvBuffer;

  2603.         src2= formatConvBuffer+VOFW;

  2604.     }

  2605.     else if (srcFormat==PIX_FMT_RGB32_1)

  2606.     {

  2607.         if(c->chrSrcHSubSample)

  2608.             RENAME(bgr32ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1+ALT32_CORR, src2+ALT32_CORR, srcW);

  2609.         else

  2610.             RENAME(bgr32ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1+ALT32_CORR, src2+ALT32_CORR, srcW);

  2611.         src1= formatConvBuffer;

  2612.         src2= formatConvBuffer+VOFW;

  2613.     }

  2614.     else if (srcFormat==PIX_FMT_BGR24)

  2615.     {

  2616.         if(c->chrSrcHSubSample)

  2617.             RENAME(bgr24ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2618.         else

  2619.             RENAME(bgr24ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2620.         src1= formatConvBuffer;

  2621.         src2= formatConvBuffer+VOFW;

  2622.     }

  2623.     else if (srcFormat==PIX_FMT_BGR565)

  2624.     {

  2625.         if(c->chrSrcHSubSample)

  2626.             RENAME(bgr16ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2627.         else

  2628.             RENAME(bgr16ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2629.         src1= formatConvBuffer;

  2630.         src2= formatConvBuffer+VOFW;

  2631.     }

  2632.     else if (srcFormat==PIX_FMT_BGR555)

  2633.     {

  2634.         if(c->chrSrcHSubSample)

  2635.             RENAME(bgr15ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2636.         else

  2637.             RENAME(bgr15ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2638.         src1= formatConvBuffer;

  2639.         src2= formatConvBuffer+VOFW;

  2640.     }

  2641.     else if (srcFormat==PIX_FMT_BGR32)

  2642.     {

  2643.         if(c->chrSrcHSubSample)

  2644.             RENAME(rgb32ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2645.         else

  2646.             RENAME(rgb32ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2647.         src1= formatConvBuffer;

  2648.         src2= formatConvBuffer+VOFW;

  2649.     }

  2650.     else if (srcFormat==PIX_FMT_BGR32_1)

  2651.     {

  2652.         if(c->chrSrcHSubSample)

  2653.             RENAME(rgb32ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1+ALT32_CORR, src2+ALT32_CORR, srcW);

  2654.         else

  2655.             RENAME(rgb32ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1+ALT32_CORR, src2+ALT32_CORR, srcW);

  2656.         src1= formatConvBuffer;

  2657.         src2= formatConvBuffer+VOFW;

  2658.     }

  2659.     else if (srcFormat==PIX_FMT_RGB24)

  2660.     {

  2661.         if(c->chrSrcHSubSample)

  2662.             RENAME(rgb24ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2663.         else

  2664.             RENAME(rgb24ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2665.         src1= formatConvBuffer;

  2666.         src2= formatConvBuffer+VOFW;

  2667.     }

  2668.     else if (srcFormat==PIX_FMT_RGB565)

  2669.     {

  2670.         if(c->chrSrcHSubSample)

  2671.             RENAME(rgb16ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2672.         else

  2673.             RENAME(rgb16ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2674.         src1= formatConvBuffer;

  2675.         src2= formatConvBuffer+VOFW;

  2676.     }

  2677.     else if (srcFormat==PIX_FMT_RGB555)

  2678.     {

  2679.         if(c->chrSrcHSubSample)

  2680.             RENAME(rgb15ToUV_half)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2681.         else

  2682.             RENAME(rgb15ToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW);

  2683.         src1= formatConvBuffer;

  2684.         src2= formatConvBuffer+VOFW;

  2685.     }

  2686.     else if (isGray(srcFormat) || srcFormat==PIX_FMT_MONOBLACK || srcFormat==PIX_FMT_MONOWHITE)

  2687.     {

  2688.         return;

  2689.     }

  2690.     else if (srcFormat==PIX_FMT_RGB8 || srcFormat==PIX_FMT_BGR8 || srcFormat==PIX_FMT_PAL8 || srcFormat==PIX_FMT_BGR4_BYTE  || srcFormat==PIX_FMT_RGB4_BYTE)

  2691.     {

  2692.         RENAME(palToUV)(formatConvBuffer, formatConvBuffer+VOFW, src1, src2, srcW, (uint32_t*)pal);

  2693.         src1= formatConvBuffer;

  2694.         src2= formatConvBuffer+VOFW;

  2695.     }

  2696. 

  2697. #ifdef HAVE_MMX

  2698.     // Use the new MMX scaler if the MMX2 one can't be used (it is faster than the x86 ASM one).

  2699.     if (!(flags&SWS_FAST_BILINEAR) || (!canMMX2BeUsed))

  2700. #else

  2701.     if (!(flags&SWS_FAST_BILINEAR))

  2702. #endif

  2703.     {

  2704.         RENAME(hScale)(dst     , dstWidth, src1, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);

  2705.         RENAME(hScale)(dst+VOFW, dstWidth, src2, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);

  2706.     }

  2707.     else // fast bilinear upscale / crap downscale

  2708.     {

  2709. #if defined(ARCH_X86)

  2710. #ifdef HAVE_MMX2

  2711.         int i;

  2712. #if defined(PIC)

  2713.         uint64_t ebxsave __attribute__((aligned(8)));

  2714. #endif

  2715.         if (canMMX2BeUsed)

  2716.         {

  2717.             asm volatile(

  2718. #if defined(PIC)

  2719.             "mov          %%"REG_b", %6         \n\t"

  2720. #endif

  2721.             "pxor             %%mm7, %%mm7      \n\t"

  2722.             "mov                 %0, %%"REG_c"  \n\t"

  2723.             "mov                 %1, %%"REG_D"  \n\t"

  2724.             "mov                 %2, %%"REG_d"  \n\t"

  2725.             "mov                 %3, %%"REG_b"  \n\t"

  2726.             "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2727.             PREFETCH"   (%%"REG_c")             \n\t"

  2728.             PREFETCH" 32(%%"REG_c")             \n\t"

  2729.             PREFETCH" 64(%%"REG_c")             \n\t"

  2730. 

  2731. #ifdef ARCH_X86_64

  2732. 

  2733. #define FUNNY_UV_CODE \

  2734.             "movl       (%%"REG_b"), %%esi      \n\t"\

  2735.             "call               *%4             \n\t"\

  2736.             "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\

  2737.             "add          %%"REG_S", %%"REG_c"  \n\t"\

  2738.             "add          %%"REG_a", %%"REG_D"  \n\t"\

  2739.             "xor          %%"REG_a", %%"REG_a"  \n\t"\

  2740. 

  2741. #else

  2742. 

  2743. #define FUNNY_UV_CODE \

  2744.             "movl       (%%"REG_b"), %%esi      \n\t"\

  2745.             "call               *%4             \n\t"\

  2746.             "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\

  2747.             "add          %%"REG_a", %%"REG_D"  \n\t"\

  2748.             "xor          %%"REG_a", %%"REG_a"  \n\t"\

  2749. 

  2750. #endif /* ARCH_X86_64 */

  2751. 

  2752. FUNNY_UV_CODE

  2753. FUNNY_UV_CODE

  2754. FUNNY_UV_CODE

  2755. FUNNY_UV_CODE

  2756.             "xor          %%"REG_a", %%"REG_a"  \n\t" // i

  2757.             "mov                 %5, %%"REG_c"  \n\t" // src

  2758.             "mov                 %1, %%"REG_D"  \n\t" // buf1

  2759.             "add              $"AV_STRINGIFY(VOF)", %%"REG_D"  \n\t"

  2760.             PREFETCH"   (%%"REG_c")             \n\t"

  2761.             PREFETCH" 32(%%"REG_c")             \n\t"

  2762.             PREFETCH" 64(%%"REG_c")             \n\t"

  2763. 

  2764. FUNNY_UV_CODE

  2765. FUNNY_UV_CODE

  2766. FUNNY_UV_CODE

  2767. FUNNY_UV_CODE

  2768. 

  2769. #if defined(PIC)

  2770.             "mov %6, %%"REG_b"    \n\t"

  2771. #endif

  2772.             :: "m" (src1), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),

  2773.             "m" (funnyUVCode), "m" (src2)

  2774. #if defined(PIC)

  2775.             ,"m" (ebxsave)

  2776. #endif

  2777.             : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D

  2778. #if !defined(PIC)

  2779.              ,"%"REG_b

  2780. #endif

  2781.             );

  2782.             for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)

  2783.             {

  2784.                 //printf("%d %d %d\n", dstWidth, i, srcW);

  2785.                 dst[i] = src1[srcW-1]*128;

  2786.                 dst[i+VOFW] = src2[srcW-1]*128;

  2787.             }

  2788.         }

  2789.         else

  2790.         {

  2791. #endif /* HAVE_MMX2 */

  2792.             long xInc_shr16 = (long) (xInc >> 16);

  2793.             uint16_t xInc_mask = xInc & 0xffff;

  2794.             asm volatile(

  2795.             "xor %%"REG_a", %%"REG_a"               \n\t" // i

  2796.             "xor %%"REG_d", %%"REG_d"               \n\t" // xx

  2797.             "xorl    %%ecx, %%ecx                   \n\t" // 2*xalpha

  2798.             ASMALIGN(4)

  2799.             "1:                                     \n\t"

  2800.             "mov        %0, %%"REG_S"               \n\t"

  2801.             "movzbl  (%%"REG_S", %%"REG_d"), %%edi  \n\t" //src[xx]

  2802.             "movzbl 1(%%"REG_S", %%"REG_d"), %%esi  \n\t" //src[xx+1]

  2803.             "subl    %%edi, %%esi                   \n\t" //src[xx+1] - src[xx]

  2804.             "imull   %%ecx, %%esi                   \n\t" //(src[xx+1] - src[xx])*2*xalpha

  2805.             "shll      $16, %%edi                   \n\t"

  2806.             "addl    %%edi, %%esi                   \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha)

  2807.             "mov        %1, %%"REG_D"               \n\t"

  2808.             "shrl       $9, %%esi                   \n\t"

  2809.             "movw     %%si, (%%"REG_D", %%"REG_a", 2)   \n\t"

  2810. 

  2811.             "movzbl    (%5, %%"REG_d"), %%edi       \n\t" //src[xx]

  2812.             "movzbl   1(%5, %%"REG_d"), %%esi       \n\t" //src[xx+1]

  2813.             "subl    %%edi, %%esi                   \n\t" //src[xx+1] - src[xx]

  2814.             "imull   %%ecx, %%esi                   \n\t" //(src[xx+1] - src[xx])*2*xalpha

  2815.             "shll      $16, %%edi                   \n\t"

  2816.             "addl    %%edi, %%esi                   \n\t" //src[xx+1]*2*xalpha + src[xx]*(1-2*xalpha)

  2817.             "mov        %1, %%"REG_D"               \n\t"

  2818.             "shrl       $9, %%esi                   \n\t"

  2819.             "movw     %%si, "AV_STRINGIFY(VOF)"(%%"REG_D", %%"REG_a", 2)   \n\t"

  2820. 

  2821.             "addw       %4, %%cx                    \n\t" //2*xalpha += xInc&0xFF

  2822.             "adc        %3, %%"REG_d"               \n\t" //xx+= xInc>>8 + carry

  2823.             "add        $1, %%"REG_a"               \n\t"

  2824.             "cmp        %2, %%"REG_a"               \n\t"

  2825.             " jb        1b                          \n\t"

  2826. 

  2827. /* GCC 3.3 makes MPlayer crash on IA-32 machines when using "g" operand here,

  2828.    which is needed to support GCC 4.0. */

  2829. #if defined(ARCH_X86_64) && ((__GNUC__ > 3) || (__GNUC__ == 3 && __GNUC_MINOR__ >= 4))

  2830.             :: "m" (src1), "m" (dst), "g" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask),

  2831. #else

  2832.             :: "m" (src1), "m" (dst), "m" ((long)dstWidth), "m" (xInc_shr16), "m" (xInc_mask),

  2833. #endif

  2834.             "r" (src2)

  2835.             : "%"REG_a, "%"REG_d, "%ecx", "%"REG_D, "%esi"

  2836.             );

  2837. #ifdef HAVE_MMX2

  2838.         } //if MMX2 can't be used

  2839. #endif

  2840. #else

  2841.         int i;

  2842.         unsigned int xpos=0;

  2843.         for (i=0;i<dstWidth;i++)

  2844.         {

  2845.             register unsigned int xx=xpos>>16;

  2846.             register unsigned int xalpha=(xpos&0xFFFF)>>9;

  2847.             dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);

  2848.             dst[i+VOFW]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);

  2849.             /* slower

  2850.             dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha;

  2851.             dst[i+VOFW]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha;

  2852.             */

  2853.             xpos+=xInc;

  2854.         }

  2855. #endif /* defined(ARCH_X86) */

  2856.     }

  2857.     if(c->srcRange != c->dstRange && !(isRGB(c->dstFormat) || isBGR(c->dstFormat))){

  2858.         int i;

  2859.         //FIXME all pal and rgb srcFormats could do this convertion as well

  2860.         //FIXME all scalers more complex than bilinear could do half of this transform

  2861.         if(c->srcRange){

  2862.             for (i=0; i<dstWidth; i++){

  2863.                 dst[i     ]= (dst[i     ]*1799 + 4081085)>>11; //1469

  2864.                 dst[i+VOFW]= (dst[i+VOFW]*1799 + 4081085)>>11; //1469

  2865.             }

  2866.         }else{

  2867.             for (i=0; i<dstWidth; i++){

  2868.                 dst[i     ]= (FFMIN(dst[i     ],30775)*4663 - 9289992)>>12; //-264

  2869.                 dst[i+VOFW]= (FFMIN(dst[i+VOFW],30775)*4663 - 9289992)>>12; //-264

  2870.             }

  2871.         }

  2872.     }

  2873. }

  2874. 

  2875. static int RENAME(swScale)(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2876.                            int srcSliceH, uint8_t* dst[], int dstStride[]){

  2877. 

  2878.     /* load a few things into local vars to make the code more readable? and faster */

  2879.     const int srcW= c->srcW;

  2880.     const int dstW= c->dstW;

  2881.     const int dstH= c->dstH;

  2882.     const int chrDstW= c->chrDstW;

  2883.     const int chrSrcW= c->chrSrcW;

  2884.     const int lumXInc= c->lumXInc;

  2885.     const int chrXInc= c->chrXInc;

  2886.     const int dstFormat= c->dstFormat;

  2887.     const int srcFormat= c->srcFormat;

  2888.     const int flags= c->flags;

  2889.     const int canMMX2BeUsed= c->canMMX2BeUsed;

  2890.     int16_t *vLumFilterPos= c->vLumFilterPos;

  2891.     int16_t *vChrFilterPos= c->vChrFilterPos;

  2892.     int16_t *hLumFilterPos= c->hLumFilterPos;

  2893.     int16_t *hChrFilterPos= c->hChrFilterPos;

  2894.     int16_t *vLumFilter= c->vLumFilter;

  2895.     int16_t *vChrFilter= c->vChrFilter;

  2896.     int16_t *hLumFilter= c->hLumFilter;

  2897.     int16_t *hChrFilter= c->hChrFilter;

  2898.     int32_t *lumMmxFilter= c->lumMmxFilter;

  2899.     int32_t *chrMmxFilter= c->chrMmxFilter;

  2900.     const int vLumFilterSize= c->vLumFilterSize;

  2901.     const int vChrFilterSize= c->vChrFilterSize;

  2902.     const int hLumFilterSize= c->hLumFilterSize;

  2903.     const int hChrFilterSize= c->hChrFilterSize;

  2904.     int16_t **lumPixBuf= c->lumPixBuf;

  2905.     int16_t **chrPixBuf= c->chrPixBuf;

  2906.     const int vLumBufSize= c->vLumBufSize;

  2907.     const int vChrBufSize= c->vChrBufSize;

  2908.     uint8_t *funnyYCode= c->funnyYCode;

  2909.     uint8_t *funnyUVCode= c->funnyUVCode;

  2910.     uint8_t *formatConvBuffer= c->formatConvBuffer;

  2911.     const int chrSrcSliceY= srcSliceY >> c->chrSrcVSubSample;

  2912.     const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample);

  2913.     int lastDstY;

  2914.     uint8_t *pal=NULL;

  2915. 

  2916.     /* vars which will change and which we need to store back in the context */

  2917.     int dstY= c->dstY;

  2918.     int lumBufIndex= c->lumBufIndex;

  2919.     int chrBufIndex= c->chrBufIndex;

  2920.     int lastInLumBuf= c->lastInLumBuf;

  2921.     int lastInChrBuf= c->lastInChrBuf;

  2922. 

  2923.     if (isPacked(c->srcFormat)){

  2924.         pal= src[1];

  2925.         src[0]=

  2926.         src[1]=

  2927.         src[2]= src[0];

  2928.         srcStride[0]=

  2929.         srcStride[1]=

  2930.         srcStride[2]= srcStride[0];

  2931.     }

  2932.     srcStride[1]<<= c->vChrDrop;

  2933.     srcStride[2]<<= c->vChrDrop;

  2934. 

  2935.     //printf("swscale %X %X %X -> %X %X %X\n", (int)src[0], (int)src[1], (int)src[2],

  2936.     //       (int)dst[0], (int)dst[1], (int)dst[2]);

  2937. 

  2938. #if 0 //self test FIXME move to a vfilter or something

  2939.     {

  2940.     static volatile int i=0;

  2941.     i++;

  2942.     if (srcFormat==PIX_FMT_YUV420P && i==1 && srcSliceH>= c->srcH)

  2943.         selfTest(src, srcStride, c->srcW, c->srcH);

  2944.     i--;

  2945.     }

  2946. #endif

  2947. 

  2948.     //printf("sws Strides:%d %d %d -> %d %d %d\n", srcStride[0],srcStride[1],srcStride[2],

  2949.     //dstStride[0],dstStride[1],dstStride[2]);

  2950. 

  2951.     if (dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0)

  2952.     {

  2953.         static int firstTime=1; //FIXME move this into the context perhaps

  2954.         if (flags & SWS_PRINT_INFO && firstTime)

  2955.         {

  2956.             av_log(c, AV_LOG_WARNING, "Warning: dstStride is not aligned!\n"

  2957.                    "         ->cannot do aligned memory accesses anymore\n");

  2958.             firstTime=0;

  2959.         }

  2960.     }

  2961. 

  2962.     /* Note the user might start scaling the picture in the middle so this

  2963.        will not get executed. This is not really intended but works

  2964.        currently, so people might do it. */

  2965.     if (srcSliceY ==0){

  2966.         lumBufIndex=0;

  2967.         chrBufIndex=0;

  2968.         dstY=0;

  2969.         lastInLumBuf= -1;

  2970.         lastInChrBuf= -1;

  2971.     }

  2972. 

  2973.     lastDstY= dstY;

  2974. 

  2975.     for (;dstY < dstH; dstY++){

  2976.         unsigned char *dest =dst[0]+dstStride[0]*dstY;

  2977.         const int chrDstY= dstY>>c->chrDstVSubSample;

  2978.         unsigned char *uDest=dst[1]+dstStride[1]*chrDstY;

  2979.         unsigned char *vDest=dst[2]+dstStride[2]*chrDstY;

  2980. 

  2981.         const int firstLumSrcY= vLumFilterPos[dstY]; //First line needed as input

  2982.         const int firstChrSrcY= vChrFilterPos[chrDstY]; //First line needed as input

  2983.         const int lastLumSrcY= firstLumSrcY + vLumFilterSize -1; // Last line needed as input

  2984.         const int lastChrSrcY= firstChrSrcY + vChrFilterSize -1; // Last line needed as input

  2985. 

  2986.         //printf("dstY:%d dstH:%d firstLumSrcY:%d lastInLumBuf:%d vLumBufSize: %d vChrBufSize: %d slice: %d %d vLumFilterSize: %d firstChrSrcY: %d vChrFilterSize: %d c->chrSrcVSubSample: %d\n",

  2987.         // dstY, dstH, firstLumSrcY, lastInLumBuf, vLumBufSize, vChrBufSize, srcSliceY, srcSliceH, vLumFilterSize, firstChrSrcY, vChrFilterSize,  c->chrSrcVSubSample);

  2988.         //handle holes (FAST_BILINEAR & weird filters)

  2989.         if (firstLumSrcY > lastInLumBuf) lastInLumBuf= firstLumSrcY-1;

  2990.         if (firstChrSrcY > lastInChrBuf) lastInChrBuf= firstChrSrcY-1;

  2991.         //printf("%d %d %d\n", firstChrSrcY, lastInChrBuf, vChrBufSize);

  2992.         assert(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);

  2993.         assert(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);

  2994. 

  2995.         // Do we have enough lines in this slice to output the dstY line

  2996.         if (lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < -((-srcSliceY - srcSliceH)>>c->chrSrcVSubSample))

  2997.         {

  2998.             //Do horizontal scaling

  2999.             while(lastInLumBuf < lastLumSrcY)

  3000.             {

  3001.                 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0];

  3002.                 lumBufIndex++;

  3003.                 //printf("%d %d %d %d\n", lumBufIndex, vLumBufSize, lastInLumBuf,  lastLumSrcY);

  3004.                 assert(lumBufIndex < 2*vLumBufSize);

  3005.                 assert(lastInLumBuf + 1 - srcSliceY < srcSliceH);

  3006.                 assert(lastInLumBuf + 1 - srcSliceY >= 0);

  3007.                 //printf("%d %d\n", lumBufIndex, vLumBufSize);

  3008.                 RENAME(hyscale)(c, lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc,

  3009.                                 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize,

  3010.                                 funnyYCode, c->srcFormat, formatConvBuffer,

  3011.                                 c->lumMmx2Filter, c->lumMmx2FilterPos, pal);

  3012.                 lastInLumBuf++;

  3013.             }

  3014.             while(lastInChrBuf < lastChrSrcY)

  3015.             {

  3016.                 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1];

  3017.                 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2];

  3018.                 chrBufIndex++;

  3019.                 assert(chrBufIndex < 2*vChrBufSize);

  3020.                 assert(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));

  3021.                 assert(lastInChrBuf + 1 - chrSrcSliceY >= 0);

  3022.                 //FIXME replace parameters through context struct (some at least)

  3023. 

  3024.                 if (!(isGray(srcFormat) || isGray(dstFormat)))

  3025.                     RENAME(hcscale)(c, chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc,

  3026.                                     flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize,

  3027.                                     funnyUVCode, c->srcFormat, formatConvBuffer,

  3028.                                     c->chrMmx2Filter, c->chrMmx2FilterPos, pal);

  3029.                 lastInChrBuf++;

  3030.             }

  3031.             //wrap buf index around to stay inside the ring buffer

  3032.             if (lumBufIndex >= vLumBufSize) lumBufIndex-= vLumBufSize;

  3033.             if (chrBufIndex >= vChrBufSize) chrBufIndex-= vChrBufSize;

  3034.         }

  3035.         else // not enough lines left in this slice -> load the rest in the buffer

  3036.         {

  3037.             /* printf("%d %d Last:%d %d LastInBuf:%d %d Index:%d %d Y:%d FSize: %d %d BSize: %d %d\n",

  3038.             firstChrSrcY,firstLumSrcY,lastChrSrcY,lastLumSrcY,

  3039.             lastInChrBuf,lastInLumBuf,chrBufIndex,lumBufIndex,dstY,vChrFilterSize,vLumFilterSize,

  3040.             vChrBufSize, vLumBufSize);*/

  3041. 

  3042.             //Do horizontal scaling

  3043.             while(lastInLumBuf+1 < srcSliceY + srcSliceH)

  3044.             {

  3045.                 uint8_t *s= src[0]+(lastInLumBuf + 1 - srcSliceY)*srcStride[0];

  3046.                 lumBufIndex++;

  3047.                 assert(lumBufIndex < 2*vLumBufSize);

  3048.                 assert(lastInLumBuf + 1 - srcSliceY < srcSliceH);

  3049.                 assert(lastInLumBuf + 1 - srcSliceY >= 0);

  3050.                 RENAME(hyscale)(c, lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc,

  3051.                                 flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize,

  3052.                                 funnyYCode, c->srcFormat, formatConvBuffer,

  3053.                                 c->lumMmx2Filter, c->lumMmx2FilterPos, pal);

  3054.                 lastInLumBuf++;

  3055.             }

  3056.             while(lastInChrBuf+1 < (chrSrcSliceY + chrSrcSliceH))

  3057.             {

  3058.                 uint8_t *src1= src[1]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[1];

  3059.                 uint8_t *src2= src[2]+(lastInChrBuf + 1 - chrSrcSliceY)*srcStride[2];

  3060.                 chrBufIndex++;

  3061.                 assert(chrBufIndex < 2*vChrBufSize);

  3062.                 assert(lastInChrBuf + 1 - chrSrcSliceY < chrSrcSliceH);

  3063.                 assert(lastInChrBuf + 1 - chrSrcSliceY >= 0);

  3064. 

  3065.                 if (!(isGray(srcFormat) || isGray(dstFormat)))

  3066.                     RENAME(hcscale)(c, chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, chrSrcW, chrXInc,

  3067.                             flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize,

  3068.                             funnyUVCode, c->srcFormat, formatConvBuffer,

  3069.                             c->chrMmx2Filter, c->chrMmx2FilterPos, pal);

  3070.                 lastInChrBuf++;

  3071.             }

  3072.             //wrap buf index around to stay inside the ring buffer

  3073.             if (lumBufIndex >= vLumBufSize) lumBufIndex-= vLumBufSize;

  3074.             if (chrBufIndex >= vChrBufSize) chrBufIndex-= vChrBufSize;

  3075.             break; //we can't output a dstY line so let's try with the next slice

  3076.         }

  3077. 

  3078. #ifdef HAVE_MMX

  3079.         b5Dither= ff_dither8[dstY&1];

  3080.         g6Dither= ff_dither4[dstY&1];

  3081.         g5Dither= ff_dither8[dstY&1];

  3082.         r5Dither= ff_dither8[(dstY+1)&1];

  3083. #endif

  3084.         if (dstY < dstH-2)

  3085.         {

  3086.             int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;

  3087.             int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;

  3088. #ifdef HAVE_MMX

  3089.             int i;

  3090.         if (flags & SWS_ACCURATE_RND){

  3091.             int s= APCK_SIZE / 8;

  3092.             for (i=0; i<vLumFilterSize; i+=2){

  3093.                 *(void**)&lumMmxFilter[s*i              ]= lumSrcPtr[i  ];

  3094.                 *(void**)&lumMmxFilter[s*i+APCK_PTR2/4  ]= lumSrcPtr[i+(vLumFilterSize>1)];

  3095.                           lumMmxFilter[s*i+APCK_COEF/4  ]=

  3096.                           lumMmxFilter[s*i+APCK_COEF/4+1]= vLumFilter[dstY*vLumFilterSize + i    ]

  3097.                     + (vLumFilterSize>1 ? vLumFilter[dstY*vLumFilterSize + i + 1]<<16 : 0);

  3098.             }

  3099.             for (i=0; i<vChrFilterSize; i+=2){

  3100.                 *(void**)&chrMmxFilter[s*i              ]= chrSrcPtr[i  ];

  3101.                 *(void**)&chrMmxFilter[s*i+APCK_PTR2/4  ]= chrSrcPtr[i+(vChrFilterSize>1)];

  3102.                           chrMmxFilter[s*i+APCK_COEF/4  ]=

  3103.                           chrMmxFilter[s*i+APCK_COEF/4+1]= vChrFilter[chrDstY*vChrFilterSize + i    ]

  3104.                     + (vChrFilterSize>1 ? vChrFilter[chrDstY*vChrFilterSize + i + 1]<<16 : 0);

  3105.             }

  3106.         }else{

  3107.             for (i=0; i<vLumFilterSize; i++)

  3108.             {

  3109.                 lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i];

  3110.                 lumMmxFilter[4*i+1]= (uint64_t)lumSrcPtr[i] >> 32;

  3111.                 lumMmxFilter[4*i+2]=

  3112.                 lumMmxFilter[4*i+3]=

  3113.                     ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;

  3114.             }

  3115.             for (i=0; i<vChrFilterSize; i++)

  3116.             {

  3117.                 chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i];

  3118.                 chrMmxFilter[4*i+1]= (uint64_t)chrSrcPtr[i] >> 32;

  3119.                 chrMmxFilter[4*i+2]=

  3120.                 chrMmxFilter[4*i+3]=

  3121.                     ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;

  3122.             }

  3123.         }

  3124. #endif

  3125.             if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21){

  3126.                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;

  3127.                 if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi

  3128.                 RENAME(yuv2nv12X)(c,

  3129.                     vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,

  3130.                     vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3131.                     dest, uDest, dstW, chrDstW, dstFormat);

  3132.             }

  3133.             else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) //YV12 like

  3134.             {

  3135.                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;

  3136.                 if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi

  3137.                 if (vLumFilterSize == 1 && vChrFilterSize == 1) // unscaled YV12

  3138.                 {

  3139.                     int16_t *lumBuf = lumPixBuf[0];

  3140.                     int16_t *chrBuf= chrPixBuf[0];

  3141.                     RENAME(yuv2yuv1)(c, lumBuf, chrBuf, dest, uDest, vDest, dstW, chrDstW);

  3142.                 }

  3143.                 else //General YV12

  3144.                 {

  3145.                     RENAME(yuv2yuvX)(c,

  3146.                         vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,

  3147.                         vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3148.                         dest, uDest, vDest, dstW, chrDstW);

  3149.                 }

  3150.             }

  3151.             else

  3152.             {

  3153.                 assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);

  3154.                 assert(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2);

  3155.                 if (vLumFilterSize == 1 && vChrFilterSize == 2) //unscaled RGB

  3156.                 {

  3157.                     int chrAlpha= vChrFilter[2*dstY+1];

  3158.                     if(flags & SWS_FULL_CHR_H_INT){

  3159.                         yuv2rgbXinC_full(c, //FIXME write a packed1_full function

  3160.                             vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3161.                             vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3162.                             dest, dstW, dstY);

  3163.                     }else{

  3164.                         RENAME(yuv2packed1)(c, *lumSrcPtr, *chrSrcPtr, *(chrSrcPtr+1),

  3165.                             dest, dstW, chrAlpha, dstFormat, flags, dstY);

  3166.                     }

  3167.                 }

  3168.                 else if (vLumFilterSize == 2 && vChrFilterSize == 2) //bilinear upscale RGB

  3169.                 {

  3170.                     int lumAlpha= vLumFilter[2*dstY+1];

  3171.                     int chrAlpha= vChrFilter[2*dstY+1];

  3172.                     lumMmxFilter[2]=

  3173.                     lumMmxFilter[3]= vLumFilter[2*dstY   ]*0x10001;

  3174.                     chrMmxFilter[2]=

  3175.                     chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001;

  3176.                     if(flags & SWS_FULL_CHR_H_INT){

  3177.                         yuv2rgbXinC_full(c, //FIXME write a packed2_full function

  3178.                             vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3179.                             vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3180.                             dest, dstW, dstY);

  3181.                     }else{

  3182.                         RENAME(yuv2packed2)(c, *lumSrcPtr, *(lumSrcPtr+1), *chrSrcPtr, *(chrSrcPtr+1),

  3183.                             dest, dstW, lumAlpha, chrAlpha, dstY);

  3184.                     }

  3185.                 }

  3186.                 else //general RGB

  3187.                 {

  3188.                     if(flags & SWS_FULL_CHR_H_INT){

  3189.                         yuv2rgbXinC_full(c,

  3190.                             vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3191.                             vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3192.                             dest, dstW, dstY);

  3193.                     }else{

  3194.                         RENAME(yuv2packedX)(c,

  3195.                             vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3196.                             vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3197.                             dest, dstW, dstY);

  3198.                     }

  3199.                 }

  3200.             }

  3201.         }

  3202.         else // hmm looks like we can't use MMX here without overwriting this array's tail

  3203.         {

  3204.             int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;

  3205.             int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;

  3206.             if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21){

  3207.                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;

  3208.                 if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi

  3209.                 yuv2nv12XinC(

  3210.                     vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,

  3211.                     vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3212.                     dest, uDest, dstW, chrDstW, dstFormat);

  3213.             }

  3214.             else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) //YV12

  3215.             {

  3216.                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;

  3217.                 if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi

  3218.                 yuv2yuvXinC(

  3219.                     vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,

  3220.                     vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3221.                     dest, uDest, vDest, dstW, chrDstW);

  3222.             }

  3223.             else

  3224.             {

  3225.                 assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);

  3226.                 assert(chrSrcPtr + vChrFilterSize - 1 < chrPixBuf + vChrBufSize*2);

  3227.                 if(flags & SWS_FULL_CHR_H_INT){

  3228.                     yuv2rgbXinC_full(c,

  3229.                         vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3230.                         vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3231.                         dest, dstW, dstY);

  3232.                 }else{

  3233.                     yuv2packedXinC(c,

  3234.                         vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,

  3235.                         vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,

  3236.                         dest, dstW, dstY);

  3237.                 }

  3238.             }

  3239.         }

  3240.     }

  3241. 

  3242. #ifdef HAVE_MMX

  3243.     asm volatile(SFENCE:::"memory");

  3244.     asm volatile(EMMS:::"memory");

  3245. #endif

  3246.     /* store changed local vars back in the context */

  3247.     c->dstY= dstY;

  3248.     c->lumBufIndex= lumBufIndex;

  3249.     c->chrBufIndex= chrBufIndex;

  3250.     c->lastInLumBuf= lastInLumBuf;

  3251.     c->lastInChrBuf= lastInChrBuf;

  3252. 

  3253.     return dstY - lastDstY;

  3254. }