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.
10676 Commits
32 Branches
283MB
Tree: 73b46a95d2
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '73b46a95d2'
${ noResults }
FFmpeg/libswscale/rgb2rgb_template.c

2743 lines
98KB
Raw Blame History 

  1. /*

  2.  *  rgb2rgb.c, Software RGB to RGB convertor

  3.  *  pluralize by Software PAL8 to RGB convertor

  4.  *               Software YUV to YUV convertor

  5.  *               Software YUV to RGB convertor

  6.  *  Written by Nick Kurshev.

  7.  *  palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)

  8.  *  lot of big-endian byteorder fixes by Alex Beregszaszi

  9.  *

  10.  * This file is part of FFmpeg.

  11.  *

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

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

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

  15.  * (at your option) any later version.

  16.  *

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

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

  19.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

  20.  * GNU General Public License for more details.

  21.  *

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

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

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

  25.  *

  26.  * the C code (not assembly, mmx, ...) of this file can be used

  27.  * under the LGPL license too

  28.  */

  29. 

  30. #include <stddef.h>

  31. #include <inttypes.h> /* for __WORDSIZE */

  32. 

  33. #ifndef __WORDSIZE

  34. // #warning You have misconfigured system and probably will lose performance!

  35. #define __WORDSIZE MP_WORDSIZE

  36. #endif

  37. 

  38. #undef PREFETCH

  39. #undef MOVNTQ

  40. #undef EMMS

  41. #undef SFENCE

  42. #undef MMREG_SIZE

  43. #undef PREFETCHW

  44. #undef PAVGB

  45. 

  46. #ifdef HAVE_SSE2

  47. #define MMREG_SIZE 16

  48. #else

  49. #define MMREG_SIZE 8

  50. #endif

  51. 

  52. #ifdef HAVE_3DNOW

  53. #define PREFETCH  "prefetch"

  54. #define PREFETCHW "prefetchw"

  55. #define PAVGB     "pavgusb"

  56. #elif defined ( HAVE_MMX2 )

  57. #define PREFETCH "prefetchnta"

  58. #define PREFETCHW "prefetcht0"

  59. #define PAVGB     "pavgb"

  60. #else

  61. #ifdef __APPLE__

  62. #define PREFETCH "#"

  63. #define PREFETCHW "#"

  64. #else

  65. #define PREFETCH  " # nop"

  66. #define PREFETCHW " # nop"

  67. #endif

  68. #endif

  69. 

  70. #ifdef HAVE_3DNOW

  71. /* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */

  72. #define EMMS     "femms"

  73. #else

  74. #define EMMS     "emms"

  75. #endif

  76. 

  77. #ifdef HAVE_MMX2

  78. #define MOVNTQ "movntq"

  79. #define SFENCE "sfence"

  80. #else

  81. #define MOVNTQ "movq"

  82. #define SFENCE " # nop"

  83. #endif

  84. 

  85. static inline void RENAME(rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size)

  86. {

  87.     uint8_t *dest = dst;

  88.     const uint8_t *s = src;

  89.     const uint8_t *end;

  90.     #ifdef HAVE_MMX

  91.         const uint8_t *mm_end;

  92.     #endif

  93.     end = s + src_size;

  94.     #ifdef HAVE_MMX

  95.         __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  96.         mm_end = end - 23;

  97.         __asm __volatile("movq        %0, %%mm7"::"m"(mask32):"memory");

  98.         while (s < mm_end)

  99.         {

  100.             __asm __volatile(

  101.             PREFETCH"    32%1           \n\t"

  102.             "movd          %1, %%mm0    \n\t"

  103.             "punpckldq    3%1, %%mm0    \n\t"

  104.             "movd         6%1, %%mm1    \n\t"

  105.             "punpckldq    9%1, %%mm1    \n\t"

  106.             "movd        12%1, %%mm2    \n\t"

  107.             "punpckldq   15%1, %%mm2    \n\t"

  108.             "movd        18%1, %%mm3    \n\t"

  109.             "punpckldq   21%1, %%mm3    \n\t"

  110.             "pand       %%mm7, %%mm0    \n\t"

  111.             "pand       %%mm7, %%mm1    \n\t"

  112.             "pand       %%mm7, %%mm2    \n\t"

  113.             "pand       %%mm7, %%mm3    \n\t"

  114.             MOVNTQ"     %%mm0,   %0     \n\t"

  115.             MOVNTQ"     %%mm1,  8%0     \n\t"

  116.             MOVNTQ"     %%mm2, 16%0     \n\t"

  117.             MOVNTQ"     %%mm3, 24%0"

  118.             :"=m"(*dest)

  119.             :"m"(*s)

  120.             :"memory");

  121.             dest += 32;

  122.             s += 24;

  123.         }

  124.         __asm __volatile(SFENCE:::"memory");

  125.         __asm __volatile(EMMS:::"memory");

  126.     #endif

  127.     while (s < end)

  128.     {

  129.     #ifdef WORDS_BIGENDIAN

  130.         /* RGB24 (= R,G,B) -> RGB32 (= A,B,G,R) */

  131.         *dest++ = 0;

  132.         *dest++ = s[2];

  133.         *dest++ = s[1];

  134.         *dest++ = s[0];

  135.         s+=3;

  136.     #else

  137.         *dest++ = *s++;

  138.         *dest++ = *s++;

  139.         *dest++ = *s++;

  140.         *dest++ = 0;

  141.     #endif

  142.     }

  143. }

  144. 

  145. static inline void RENAME(rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size)

  146. {

  147.     uint8_t *dest = dst;

  148.     const uint8_t *s = src;

  149.     const uint8_t *end;

  150. #ifdef HAVE_MMX

  151.     const uint8_t *mm_end;

  152. #endif

  153.     end = s + src_size;

  154. #ifdef HAVE_MMX

  155.     __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  156.     mm_end = end - 31;

  157.     while (s < mm_end)

  158.     {

  159.         __asm __volatile(

  160.         PREFETCH"    32%1           \n\t"

  161.         "movq          %1, %%mm0    \n\t"

  162.         "movq         8%1, %%mm1    \n\t"

  163.         "movq        16%1, %%mm4    \n\t"

  164.         "movq        24%1, %%mm5    \n\t"

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

  166.         "movq       %%mm1, %%mm3    \n\t"

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

  168.         "movq       %%mm5, %%mm7    \n\t"

  169.         "psrlq         $8, %%mm2    \n\t"

  170.         "psrlq         $8, %%mm3    \n\t"

  171.         "psrlq         $8, %%mm6    \n\t"

  172.         "psrlq         $8, %%mm7    \n\t"

  173.         "pand          %2, %%mm0    \n\t"

  174.         "pand          %2, %%mm1    \n\t"

  175.         "pand          %2, %%mm4    \n\t"

  176.         "pand          %2, %%mm5    \n\t"

  177.         "pand          %3, %%mm2    \n\t"

  178.         "pand          %3, %%mm3    \n\t"

  179.         "pand          %3, %%mm6    \n\t"

  180.         "pand          %3, %%mm7    \n\t"

  181.         "por        %%mm2, %%mm0    \n\t"

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

  183.         "por        %%mm6, %%mm4    \n\t"

  184.         "por        %%mm7, %%mm5    \n\t"

  185. 

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

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

  188.         "psllq        $48, %%mm2    \n\t"

  189.         "psllq        $32, %%mm3    \n\t"

  190.         "pand          %4, %%mm2    \n\t"

  191.         "pand          %5, %%mm3    \n\t"

  192.         "por        %%mm2, %%mm0    \n\t"

  193.         "psrlq        $16, %%mm1    \n\t"

  194.         "psrlq        $32, %%mm4    \n\t"

  195.         "psllq        $16, %%mm5    \n\t"

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

  197.         "pand          %6, %%mm5    \n\t"

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

  199. 

  200.         MOVNTQ"     %%mm0,   %0     \n\t"

  201.         MOVNTQ"     %%mm1,  8%0     \n\t"

  202.         MOVNTQ"     %%mm4, 16%0"

  203.         :"=m"(*dest)

  204.         :"m"(*s),"m"(mask24l),

  205.          "m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh)

  206.         :"memory");

  207.         dest += 24;

  208.         s += 32;

  209.     }

  210.     __asm __volatile(SFENCE:::"memory");

  211.     __asm __volatile(EMMS:::"memory");

  212. #endif

  213.     while (s < end)

  214.     {

  215. #ifdef WORDS_BIGENDIAN

  216.         /* RGB32 (= A,B,G,R) -> RGB24 (= R,G,B) */

  217.         s++;

  218.         dest[2] = *s++;

  219.         dest[1] = *s++;

  220.         dest[0] = *s++;

  221.         dest += 3;

  222. #else

  223.         *dest++ = *s++;

  224.         *dest++ = *s++;

  225.         *dest++ = *s++;

  226.         s++;

  227. #endif

  228.     }

  229. }

  230. 

  231. /*

  232.  Original by Strepto/Astral

  233.  ported to gcc & bugfixed : A'rpi

  234.  MMX2, 3DNOW optimization by Nick Kurshev

  235.  32bit c version, and and&add trick by Michael Niedermayer

  236. */

  237. static inline void RENAME(rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size)

  238. {

  239.     register const uint8_t* s=src;

  240.     register uint8_t* d=dst;

  241.     register const uint8_t *end;

  242.     const uint8_t *mm_end;

  243.     end = s + src_size;

  244. #ifdef HAVE_MMX

  245.     __asm __volatile(PREFETCH"    %0"::"m"(*s));

  246.     __asm __volatile("movq        %0, %%mm4"::"m"(mask15s));

  247.     mm_end = end - 15;

  248.     while (s<mm_end)

  249.     {

  250.         __asm __volatile(

  251.         PREFETCH"  32%1         \n\t"

  252.         "movq        %1, %%mm0  \n\t"

  253.         "movq       8%1, %%mm2  \n\t"

  254.         "movq     %%mm0, %%mm1  \n\t"

  255.         "movq     %%mm2, %%mm3  \n\t"

  256.         "pand     %%mm4, %%mm0  \n\t"

  257.         "pand     %%mm4, %%mm2  \n\t"

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

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

  260.         MOVNTQ"   %%mm0,  %0    \n\t"

  261.         MOVNTQ"   %%mm2, 8%0"

  262.         :"=m"(*d)

  263.         :"m"(*s)

  264.         );

  265.         d+=16;

  266.         s+=16;

  267.     }

  268.     __asm __volatile(SFENCE:::"memory");

  269.     __asm __volatile(EMMS:::"memory");

  270. #endif

  271.     mm_end = end - 3;

  272.     while (s < mm_end)

  273.     {

  274.         register unsigned x= *((uint32_t *)s);

  275.         *((uint32_t *)d) = (x&0x7FFF7FFF) + (x&0x7FE07FE0);

  276.         d+=4;

  277.         s+=4;

  278.     }

  279.     if (s < end)

  280.     {

  281.         register unsigned short x= *((uint16_t *)s);

  282.         *((uint16_t *)d) = (x&0x7FFF) + (x&0x7FE0);

  283.     }

  284. }

  285. 

  286. static inline void RENAME(rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size)

  287. {

  288.     register const uint8_t* s=src;

  289.     register uint8_t* d=dst;

  290.     register const uint8_t *end;

  291.     const uint8_t *mm_end;

  292.     end = s + src_size;

  293. #ifdef HAVE_MMX

  294.     __asm __volatile(PREFETCH"    %0"::"m"(*s));

  295.     __asm __volatile("movq        %0, %%mm7"::"m"(mask15rg));

  296.     __asm __volatile("movq        %0, %%mm6"::"m"(mask15b));

  297.     mm_end = end - 15;

  298.     while (s<mm_end)

  299.     {

  300.         __asm __volatile(

  301.         PREFETCH"  32%1         \n\t"

  302.         "movq        %1, %%mm0  \n\t"

  303.         "movq       8%1, %%mm2  \n\t"

  304.         "movq     %%mm0, %%mm1  \n\t"

  305.         "movq     %%mm2, %%mm3  \n\t"

  306.         "psrlq       $1, %%mm0  \n\t"

  307.         "psrlq       $1, %%mm2  \n\t"

  308.         "pand     %%mm7, %%mm0  \n\t"

  309.         "pand     %%mm7, %%mm2  \n\t"

  310.         "pand     %%mm6, %%mm1  \n\t"

  311.         "pand     %%mm6, %%mm3  \n\t"

  312.         "por      %%mm1, %%mm0  \n\t"

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

  314.         MOVNTQ"   %%mm0,  %0    \n\t"

  315.         MOVNTQ"   %%mm2, 8%0"

  316.         :"=m"(*d)

  317.         :"m"(*s)

  318.         );

  319.         d+=16;

  320.         s+=16;

  321.     }

  322.     __asm __volatile(SFENCE:::"memory");

  323.     __asm __volatile(EMMS:::"memory");

  324. #endif

  325.     mm_end = end - 3;

  326.     while (s < mm_end)

  327.     {

  328.         register uint32_t x= *((uint32_t *)s);

  329.         *((uint32_t *)d) = ((x>>1)&0x7FE07FE0) | (x&0x001F001F);

  330.         s+=4;

  331.         d+=4;

  332.     }

  333.     if (s < end)

  334.     {

  335.         register uint16_t x= *((uint16_t *)s);

  336.         *((uint16_t *)d) = ((x>>1)&0x7FE0) | (x&0x001F);

  337.         s+=2;

  338.         d+=2;

  339.     }

  340. }

  341. 

  342. static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size)

  343. {

  344.     const uint8_t *s = src;

  345.     const uint8_t *end;

  346. #ifdef HAVE_MMX

  347.     const uint8_t *mm_end;

  348. #endif

  349.     uint16_t *d = (uint16_t *)dst;

  350.     end = s + src_size;

  351. #ifdef HAVE_MMX

  352.     mm_end = end - 15;

  353. #if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster)

  354.     asm volatile(

  355.     "movq           %3, %%mm5   \n\t"

  356.     "movq           %4, %%mm6   \n\t"

  357.     "movq           %5, %%mm7   \n\t"

  358.     "jmp 2f                     \n\t"

  359.     ASMALIGN(4)

  360.     "1:                         \n\t"

  361.     PREFETCH"   32(%1)          \n\t"

  362.     "movd         (%1), %%mm0   \n\t"

  363.     "movd        4(%1), %%mm3   \n\t"

  364.     "punpckldq   8(%1), %%mm0   \n\t"

  365.     "punpckldq  12(%1), %%mm3   \n\t"

  366.     "movq        %%mm0, %%mm1   \n\t"

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

  368.     "pand        %%mm6, %%mm0   \n\t"

  369.     "pand        %%mm6, %%mm3   \n\t"

  370.     "pmaddwd     %%mm7, %%mm0   \n\t"

  371.     "pmaddwd     %%mm7, %%mm3   \n\t"

  372.     "pand        %%mm5, %%mm1   \n\t"

  373.     "pand        %%mm5, %%mm4   \n\t"

  374.     "por         %%mm1, %%mm0   \n\t"

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

  376.     "psrld          $5, %%mm0   \n\t"

  377.     "pslld         $11, %%mm3   \n\t"

  378.     "por         %%mm3, %%mm0   \n\t"

  379.     MOVNTQ"      %%mm0, (%0)    \n\t"

  380.     "add           $16,  %1     \n\t"

  381.     "add            $8,  %0     \n\t"

  382.     "2:                         \n\t"

  383.     "cmp            %2,  %1     \n\t"

  384.     " jb            1b          \n\t"

  385.     : "+r" (d), "+r"(s)

  386.     : "r" (mm_end), "m" (mask3216g), "m" (mask3216br), "m" (mul3216)

  387.     );

  388. #else

  389.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  390.     __asm __volatile(

  391.         "movq    %0, %%mm7    \n\t"

  392.         "movq    %1, %%mm6    \n\t"

  393.         ::"m"(red_16mask),"m"(green_16mask));

  394.     while (s < mm_end)

  395.     {

  396.         __asm __volatile(

  397.         PREFETCH"    32%1           \n\t"

  398.         "movd          %1, %%mm0    \n\t"

  399.         "movd         4%1, %%mm3    \n\t"

  400.         "punpckldq    8%1, %%mm0    \n\t"

  401.         "punpckldq   12%1, %%mm3    \n\t"

  402.         "movq       %%mm0, %%mm1    \n\t"

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

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

  405.         "movq       %%mm3, %%mm5    \n\t"

  406.         "psrlq         $3, %%mm0    \n\t"

  407.         "psrlq         $3, %%mm3    \n\t"

  408.         "pand          %2, %%mm0    \n\t"

  409.         "pand          %2, %%mm3    \n\t"

  410.         "psrlq         $5, %%mm1    \n\t"

  411.         "psrlq         $5, %%mm4    \n\t"

  412.         "pand       %%mm6, %%mm1    \n\t"

  413.         "pand       %%mm6, %%mm4    \n\t"

  414.         "psrlq         $8, %%mm2    \n\t"

  415.         "psrlq         $8, %%mm5    \n\t"

  416.         "pand       %%mm7, %%mm2    \n\t"

  417.         "pand       %%mm7, %%mm5    \n\t"

  418.         "por        %%mm1, %%mm0    \n\t"

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

  420.         "por        %%mm2, %%mm0    \n\t"

  421.         "por        %%mm5, %%mm3    \n\t"

  422.         "psllq        $16, %%mm3    \n\t"

  423.         "por        %%mm3, %%mm0    \n\t"

  424.         MOVNTQ"     %%mm0, %0       \n\t"

  425.         :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");

  426.         d += 4;

  427.         s += 16;

  428.     }

  429. #endif

  430.     __asm __volatile(SFENCE:::"memory");

  431.     __asm __volatile(EMMS:::"memory");

  432. #endif

  433.     while (s < end)

  434.     {

  435.         register int rgb = *(uint32_t*)s; s += 4;

  436.         *d++ = ((rgb&0xFF)>>3) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>8);

  437.     }

  438. }

  439. 

  440. static inline void RENAME(rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size)

  441. {

  442.     const uint8_t *s = src;

  443.     const uint8_t *end;

  444. #ifdef HAVE_MMX

  445.     const uint8_t *mm_end;

  446. #endif

  447.     uint16_t *d = (uint16_t *)dst;

  448.     end = s + src_size;

  449. #ifdef HAVE_MMX

  450.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  451.     __asm __volatile(

  452.         "movq          %0, %%mm7    \n\t"

  453.         "movq          %1, %%mm6    \n\t"

  454.         ::"m"(red_16mask),"m"(green_16mask));

  455.     mm_end = end - 15;

  456.     while (s < mm_end)

  457.     {

  458.         __asm __volatile(

  459.         PREFETCH"    32%1           \n\t"

  460.         "movd          %1, %%mm0    \n\t"

  461.         "movd         4%1, %%mm3    \n\t"

  462.         "punpckldq    8%1, %%mm0    \n\t"

  463.         "punpckldq   12%1, %%mm3    \n\t"

  464.         "movq       %%mm0, %%mm1    \n\t"

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

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

  467.         "movq       %%mm3, %%mm5    \n\t"

  468.         "psllq         $8, %%mm0    \n\t"

  469.         "psllq         $8, %%mm3    \n\t"

  470.         "pand       %%mm7, %%mm0    \n\t"

  471.         "pand       %%mm7, %%mm3    \n\t"

  472.         "psrlq         $5, %%mm1    \n\t"

  473.         "psrlq         $5, %%mm4    \n\t"

  474.         "pand       %%mm6, %%mm1    \n\t"

  475.         "pand       %%mm6, %%mm4    \n\t"

  476.         "psrlq        $19, %%mm2    \n\t"

  477.         "psrlq        $19, %%mm5    \n\t"

  478.         "pand          %2, %%mm2    \n\t"

  479.         "pand          %2, %%mm5    \n\t"

  480.         "por        %%mm1, %%mm0    \n\t"

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

  482.         "por        %%mm2, %%mm0    \n\t"

  483.         "por        %%mm5, %%mm3    \n\t"

  484.         "psllq        $16, %%mm3    \n\t"

  485.         "por        %%mm3, %%mm0    \n\t"

  486.         MOVNTQ"     %%mm0, %0       \n\t"

  487.         :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");

  488.         d += 4;

  489.         s += 16;

  490.     }

  491.     __asm __volatile(SFENCE:::"memory");

  492.     __asm __volatile(EMMS:::"memory");

  493. #endif

  494.     while (s < end)

  495.     {

  496.         register int rgb = *(uint32_t*)s; s += 4;

  497.         *d++ = ((rgb&0xF8)<<8) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>19);

  498.     }

  499. }

  500. 

  501. static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size)

  502. {

  503.     const uint8_t *s = src;

  504.     const uint8_t *end;

  505. #ifdef HAVE_MMX

  506.     const uint8_t *mm_end;

  507. #endif

  508.     uint16_t *d = (uint16_t *)dst;

  509.     end = s + src_size;

  510. #ifdef HAVE_MMX

  511.     mm_end = end - 15;

  512. #if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster)

  513.     asm volatile(

  514.     "movq           %3, %%mm5   \n\t"

  515.     "movq           %4, %%mm6   \n\t"

  516.     "movq           %5, %%mm7   \n\t"

  517.     "jmp            2f          \n\t"

  518.     ASMALIGN(4)

  519.     "1:                         \n\t"

  520.     PREFETCH"   32(%1)          \n\t"

  521.     "movd         (%1), %%mm0   \n\t"

  522.     "movd        4(%1), %%mm3   \n\t"

  523.     "punpckldq   8(%1), %%mm0   \n\t"

  524.     "punpckldq  12(%1), %%mm3   \n\t"

  525.     "movq        %%mm0, %%mm1   \n\t"

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

  527.     "pand        %%mm6, %%mm0   \n\t"

  528.     "pand        %%mm6, %%mm3   \n\t"

  529.     "pmaddwd     %%mm7, %%mm0   \n\t"

  530.     "pmaddwd     %%mm7, %%mm3   \n\t"

  531.     "pand        %%mm5, %%mm1   \n\t"

  532.     "pand        %%mm5, %%mm4   \n\t"

  533.     "por         %%mm1, %%mm0   \n\t"

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

  535.     "psrld          $6, %%mm0   \n\t"

  536.     "pslld         $10, %%mm3   \n\t"

  537.     "por         %%mm3, %%mm0   \n\t"

  538.     MOVNTQ"      %%mm0, (%0)    \n\t"

  539.     "add           $16,  %1     \n\t"

  540.     "add            $8,  %0     \n\t"

  541.     "2:                         \n\t"

  542.     "cmp            %2,  %1     \n\t"

  543.     " jb            1b          \n\t"

  544.     : "+r" (d), "+r"(s)

  545.     : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215)

  546.     );

  547. #else

  548.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  549.     __asm __volatile(

  550.         "movq          %0, %%mm7    \n\t"

  551.         "movq          %1, %%mm6    \n\t"

  552.         ::"m"(red_15mask),"m"(green_15mask));

  553.     while (s < mm_end)

  554.     {

  555.         __asm __volatile(

  556.         PREFETCH"    32%1           \n\t"

  557.         "movd          %1, %%mm0    \n\t"

  558.         "movd         4%1, %%mm3    \n\t"

  559.         "punpckldq    8%1, %%mm0    \n\t"

  560.         "punpckldq   12%1, %%mm3    \n\t"

  561.         "movq       %%mm0, %%mm1    \n\t"

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

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

  564.         "movq       %%mm3, %%mm5    \n\t"

  565.         "psrlq         $3, %%mm0    \n\t"

  566.         "psrlq         $3, %%mm3    \n\t"

  567.         "pand          %2, %%mm0    \n\t"

  568.         "pand          %2, %%mm3    \n\t"

  569.         "psrlq         $6, %%mm1    \n\t"

  570.         "psrlq         $6, %%mm4    \n\t"

  571.         "pand       %%mm6, %%mm1    \n\t"

  572.         "pand       %%mm6, %%mm4    \n\t"

  573.         "psrlq         $9, %%mm2    \n\t"

  574.         "psrlq         $9, %%mm5    \n\t"

  575.         "pand       %%mm7, %%mm2    \n\t"

  576.         "pand       %%mm7, %%mm5    \n\t"

  577.         "por        %%mm1, %%mm0    \n\t"

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

  579.         "por        %%mm2, %%mm0    \n\t"

  580.         "por        %%mm5, %%mm3    \n\t"

  581.         "psllq        $16, %%mm3    \n\t"

  582.         "por        %%mm3, %%mm0    \n\t"

  583.         MOVNTQ"     %%mm0, %0       \n\t"

  584.         :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");

  585.         d += 4;

  586.         s += 16;

  587.     }

  588. #endif

  589.     __asm __volatile(SFENCE:::"memory");

  590.     __asm __volatile(EMMS:::"memory");

  591. #endif

  592.     while (s < end)

  593.     {

  594.         register int rgb = *(uint32_t*)s; s += 4;

  595.         *d++ = ((rgb&0xFF)>>3) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>9);

  596.     }

  597. }

  598. 

  599. static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size)

  600. {

  601.     const uint8_t *s = src;

  602.     const uint8_t *end;

  603. #ifdef HAVE_MMX

  604.     const uint8_t *mm_end;

  605. #endif

  606.     uint16_t *d = (uint16_t *)dst;

  607.     end = s + src_size;

  608. #ifdef HAVE_MMX

  609.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  610.     __asm __volatile(

  611.         "movq          %0, %%mm7    \n\t"

  612.         "movq          %1, %%mm6    \n\t"

  613.         ::"m"(red_15mask),"m"(green_15mask));

  614.     mm_end = end - 15;

  615.     while (s < mm_end)

  616.     {

  617.         __asm __volatile(

  618.         PREFETCH"    32%1           \n\t"

  619.         "movd          %1, %%mm0    \n\t"

  620.         "movd         4%1, %%mm3    \n\t"

  621.         "punpckldq    8%1, %%mm0    \n\t"

  622.         "punpckldq   12%1, %%mm3    \n\t"

  623.         "movq       %%mm0, %%mm1    \n\t"

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

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

  626.         "movq       %%mm3, %%mm5    \n\t"

  627.         "psllq         $7, %%mm0    \n\t"

  628.         "psllq         $7, %%mm3    \n\t"

  629.         "pand       %%mm7, %%mm0    \n\t"

  630.         "pand       %%mm7, %%mm3    \n\t"

  631.         "psrlq         $6, %%mm1    \n\t"

  632.         "psrlq         $6, %%mm4    \n\t"

  633.         "pand       %%mm6, %%mm1    \n\t"

  634.         "pand       %%mm6, %%mm4    \n\t"

  635.         "psrlq        $19, %%mm2    \n\t"

  636.         "psrlq        $19, %%mm5    \n\t"

  637.         "pand          %2, %%mm2    \n\t"

  638.         "pand          %2, %%mm5    \n\t"

  639.         "por        %%mm1, %%mm0    \n\t"

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

  641.         "por        %%mm2, %%mm0    \n\t"

  642.         "por        %%mm5, %%mm3    \n\t"

  643.         "psllq        $16, %%mm3    \n\t"

  644.         "por        %%mm3, %%mm0    \n\t"

  645.         MOVNTQ"     %%mm0, %0       \n\t"

  646.         :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");

  647.         d += 4;

  648.         s += 16;

  649.     }

  650.     __asm __volatile(SFENCE:::"memory");

  651.     __asm __volatile(EMMS:::"memory");

  652. #endif

  653.     while (s < end)

  654.     {

  655.         register int rgb = *(uint32_t*)s; s += 4;

  656.         *d++ = ((rgb&0xF8)<<7) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>19);

  657.     }

  658. }

  659. 

  660. static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size)

  661. {

  662.     const uint8_t *s = src;

  663.     const uint8_t *end;

  664. #ifdef HAVE_MMX

  665.     const uint8_t *mm_end;

  666. #endif

  667.     uint16_t *d = (uint16_t *)dst;

  668.     end = s + src_size;

  669. #ifdef HAVE_MMX

  670.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  671.     __asm __volatile(

  672.         "movq         %0, %%mm7     \n\t"

  673.         "movq         %1, %%mm6     \n\t"

  674.         ::"m"(red_16mask),"m"(green_16mask));

  675.     mm_end = end - 11;

  676.     while (s < mm_end)

  677.     {

  678.         __asm __volatile(

  679.         PREFETCH"    32%1           \n\t"

  680.         "movd          %1, %%mm0    \n\t"

  681.         "movd         3%1, %%mm3    \n\t"

  682.         "punpckldq    6%1, %%mm0    \n\t"

  683.         "punpckldq    9%1, %%mm3    \n\t"

  684.         "movq       %%mm0, %%mm1    \n\t"

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

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

  687.         "movq       %%mm3, %%mm5    \n\t"

  688.         "psrlq         $3, %%mm0    \n\t"

  689.         "psrlq         $3, %%mm3    \n\t"

  690.         "pand          %2, %%mm0    \n\t"

  691.         "pand          %2, %%mm3    \n\t"

  692.         "psrlq         $5, %%mm1    \n\t"

  693.         "psrlq         $5, %%mm4    \n\t"

  694.         "pand       %%mm6, %%mm1    \n\t"

  695.         "pand       %%mm6, %%mm4    \n\t"

  696.         "psrlq         $8, %%mm2    \n\t"

  697.         "psrlq         $8, %%mm5    \n\t"

  698.         "pand       %%mm7, %%mm2    \n\t"

  699.         "pand       %%mm7, %%mm5    \n\t"

  700.         "por        %%mm1, %%mm0    \n\t"

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

  702.         "por        %%mm2, %%mm0    \n\t"

  703.         "por        %%mm5, %%mm3    \n\t"

  704.         "psllq        $16, %%mm3    \n\t"

  705.         "por        %%mm3, %%mm0    \n\t"

  706.         MOVNTQ"     %%mm0, %0       \n\t"

  707.         :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");

  708.         d += 4;

  709.         s += 12;

  710.     }

  711.     __asm __volatile(SFENCE:::"memory");

  712.     __asm __volatile(EMMS:::"memory");

  713. #endif

  714.     while (s < end)

  715.     {

  716.         const int b = *s++;

  717.         const int g = *s++;

  718.         const int r = *s++;

  719.         *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);

  720.     }

  721. }

  722. 

  723. static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size)

  724. {

  725.     const uint8_t *s = src;

  726.     const uint8_t *end;

  727. #ifdef HAVE_MMX

  728.     const uint8_t *mm_end;

  729. #endif

  730.     uint16_t *d = (uint16_t *)dst;

  731.     end = s + src_size;

  732. #ifdef HAVE_MMX

  733.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  734.     __asm __volatile(

  735.         "movq         %0, %%mm7     \n\t"

  736.         "movq         %1, %%mm6     \n\t"

  737.         ::"m"(red_16mask),"m"(green_16mask));

  738.     mm_end = end - 15;

  739.     while (s < mm_end)

  740.     {

  741.         __asm __volatile(

  742.         PREFETCH"    32%1           \n\t"

  743.         "movd          %1, %%mm0    \n\t"

  744.         "movd         3%1, %%mm3    \n\t"

  745.         "punpckldq    6%1, %%mm0    \n\t"

  746.         "punpckldq    9%1, %%mm3    \n\t"

  747.         "movq       %%mm0, %%mm1    \n\t"

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

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

  750.         "movq       %%mm3, %%mm5    \n\t"

  751.         "psllq         $8, %%mm0    \n\t"

  752.         "psllq         $8, %%mm3    \n\t"

  753.         "pand       %%mm7, %%mm0    \n\t"

  754.         "pand       %%mm7, %%mm3    \n\t"

  755.         "psrlq         $5, %%mm1    \n\t"

  756.         "psrlq         $5, %%mm4    \n\t"

  757.         "pand       %%mm6, %%mm1    \n\t"

  758.         "pand       %%mm6, %%mm4    \n\t"

  759.         "psrlq        $19, %%mm2    \n\t"

  760.         "psrlq        $19, %%mm5    \n\t"

  761.         "pand          %2, %%mm2    \n\t"

  762.         "pand          %2, %%mm5    \n\t"

  763.         "por        %%mm1, %%mm0    \n\t"

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

  765.         "por        %%mm2, %%mm0    \n\t"

  766.         "por        %%mm5, %%mm3    \n\t"

  767.         "psllq        $16, %%mm3    \n\t"

  768.         "por        %%mm3, %%mm0    \n\t"

  769.         MOVNTQ"     %%mm0, %0       \n\t"

  770.         :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");

  771.         d += 4;

  772.         s += 12;

  773.     }

  774.     __asm __volatile(SFENCE:::"memory");

  775.     __asm __volatile(EMMS:::"memory");

  776. #endif

  777.     while (s < end)

  778.     {

  779.         const int r = *s++;

  780.         const int g = *s++;

  781.         const int b = *s++;

  782.         *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);

  783.     }

  784. }

  785. 

  786. static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size)

  787. {

  788.     const uint8_t *s = src;

  789.     const uint8_t *end;

  790. #ifdef HAVE_MMX

  791.     const uint8_t *mm_end;

  792. #endif

  793.     uint16_t *d = (uint16_t *)dst;

  794.     end = s + src_size;

  795. #ifdef HAVE_MMX

  796.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  797.     __asm __volatile(

  798.         "movq          %0, %%mm7    \n\t"

  799.         "movq          %1, %%mm6    \n\t"

  800.         ::"m"(red_15mask),"m"(green_15mask));

  801.     mm_end = end - 11;

  802.     while (s < mm_end)

  803.     {

  804.         __asm __volatile(

  805.         PREFETCH"    32%1           \n\t"

  806.         "movd          %1, %%mm0    \n\t"

  807.         "movd         3%1, %%mm3    \n\t"

  808.         "punpckldq    6%1, %%mm0    \n\t"

  809.         "punpckldq    9%1, %%mm3    \n\t"

  810.         "movq       %%mm0, %%mm1    \n\t"

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

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

  813.         "movq       %%mm3, %%mm5    \n\t"

  814.         "psrlq         $3, %%mm0    \n\t"

  815.         "psrlq         $3, %%mm3    \n\t"

  816.         "pand          %2, %%mm0    \n\t"

  817.         "pand          %2, %%mm3    \n\t"

  818.         "psrlq         $6, %%mm1    \n\t"

  819.         "psrlq         $6, %%mm4    \n\t"

  820.         "pand       %%mm6, %%mm1    \n\t"

  821.         "pand       %%mm6, %%mm4    \n\t"

  822.         "psrlq         $9, %%mm2    \n\t"

  823.         "psrlq         $9, %%mm5    \n\t"

  824.         "pand       %%mm7, %%mm2    \n\t"

  825.         "pand       %%mm7, %%mm5    \n\t"

  826.         "por        %%mm1, %%mm0    \n\t"

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

  828.         "por        %%mm2, %%mm0    \n\t"

  829.         "por        %%mm5, %%mm3    \n\t"

  830.         "psllq        $16, %%mm3    \n\t"

  831.         "por        %%mm3, %%mm0    \n\t"

  832.         MOVNTQ"     %%mm0, %0       \n\t"

  833.         :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");

  834.         d += 4;

  835.         s += 12;

  836.     }

  837.     __asm __volatile(SFENCE:::"memory");

  838.     __asm __volatile(EMMS:::"memory");

  839. #endif

  840.     while (s < end)

  841.     {

  842.         const int b = *s++;

  843.         const int g = *s++;

  844.         const int r = *s++;

  845.         *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);

  846.     }

  847. }

  848. 

  849. static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size)

  850. {

  851.     const uint8_t *s = src;

  852.     const uint8_t *end;

  853. #ifdef HAVE_MMX

  854.     const uint8_t *mm_end;

  855. #endif

  856.     uint16_t *d = (uint16_t *)dst;

  857.     end = s + src_size;

  858. #ifdef HAVE_MMX

  859.     __asm __volatile(PREFETCH"    %0"::"m"(*src):"memory");

  860.     __asm __volatile(

  861.         "movq         %0, %%mm7     \n\t"

  862.         "movq         %1, %%mm6     \n\t"

  863.         ::"m"(red_15mask),"m"(green_15mask));

  864.     mm_end = end - 15;

  865.     while (s < mm_end)

  866.     {

  867.         __asm __volatile(

  868.         PREFETCH"   32%1            \n\t"

  869.         "movd         %1, %%mm0     \n\t"

  870.         "movd        3%1, %%mm3     \n\t"

  871.         "punpckldq   6%1, %%mm0     \n\t"

  872.         "punpckldq   9%1, %%mm3     \n\t"

  873.         "movq      %%mm0, %%mm1     \n\t"

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

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

  876.         "movq      %%mm3, %%mm5     \n\t"

  877.         "psllq        $7, %%mm0     \n\t"

  878.         "psllq        $7, %%mm3     \n\t"

  879.         "pand      %%mm7, %%mm0     \n\t"

  880.         "pand      %%mm7, %%mm3     \n\t"

  881.         "psrlq        $6, %%mm1     \n\t"

  882.         "psrlq        $6, %%mm4     \n\t"

  883.         "pand      %%mm6, %%mm1     \n\t"

  884.         "pand      %%mm6, %%mm4     \n\t"

  885.         "psrlq       $19, %%mm2     \n\t"

  886.         "psrlq       $19, %%mm5     \n\t"

  887.         "pand         %2, %%mm2     \n\t"

  888.         "pand         %2, %%mm5     \n\t"

  889.         "por       %%mm1, %%mm0     \n\t"

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

  891.         "por       %%mm2, %%mm0     \n\t"

  892.         "por       %%mm5, %%mm3     \n\t"

  893.         "psllq       $16, %%mm3     \n\t"

  894.         "por       %%mm3, %%mm0     \n\t"

  895.         MOVNTQ"    %%mm0, %0        \n\t"

  896.         :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");

  897.         d += 4;

  898.         s += 12;

  899.     }

  900.     __asm __volatile(SFENCE:::"memory");

  901.     __asm __volatile(EMMS:::"memory");

  902. #endif

  903.     while (s < end)

  904.     {

  905.         const int r = *s++;

  906.         const int g = *s++;

  907.         const int b = *s++;

  908.         *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);

  909.     }

  910. }

  911. 

  912. /*

  913.   I use here less accurate approximation by simply

  914.  left-shifting the input

  915.   value and filling the low order bits with

  916.  zeroes. This method improves png's

  917.   compression but this scheme cannot reproduce white exactly, since it does not

  918.   generate an all-ones maximum value; the net effect is to darken the

  919.   image slightly.

  920. 

  921.   The better method should be "left bit replication":

  922. 

  923.    4 3 2 1 0

  924.    ---------

  925.    1 1 0 1 1

  926. 

  927.    7 6 5 4 3  2 1 0

  928.    ----------------

  929.    1 1 0 1 1  1 1 0

  930.    |=======|  |===|

  931.        |      Leftmost Bits Repeated to Fill Open Bits

  932.        |

  933.    Original Bits

  934. */

  935. static inline void RENAME(rgb15to24)(const uint8_t *src, uint8_t *dst, long src_size)

  936. {

  937.     const uint16_t *end;

  938. #ifdef HAVE_MMX

  939.     const uint16_t *mm_end;

  940. #endif

  941.     uint8_t *d = (uint8_t *)dst;

  942.     const uint16_t *s = (uint16_t *)src;

  943.     end = s + src_size/2;

  944. #ifdef HAVE_MMX

  945.     __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  946.     mm_end = end - 7;

  947.     while (s < mm_end)

  948.     {

  949.         __asm __volatile(

  950.         PREFETCH"    32%1           \n\t"

  951.         "movq          %1, %%mm0    \n\t"

  952.         "movq          %1, %%mm1    \n\t"

  953.         "movq          %1, %%mm2    \n\t"

  954.         "pand          %2, %%mm0    \n\t"

  955.         "pand          %3, %%mm1    \n\t"

  956.         "pand          %4, %%mm2    \n\t"

  957.         "psllq         $3, %%mm0    \n\t"

  958.         "psrlq         $2, %%mm1    \n\t"

  959.         "psrlq         $7, %%mm2    \n\t"

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

  961.         "movq       %%mm1, %%mm4    \n\t"

  962.         "movq       %%mm2, %%mm5    \n\t"

  963.         "punpcklwd     %5, %%mm0    \n\t"

  964.         "punpcklwd     %5, %%mm1    \n\t"

  965.         "punpcklwd     %5, %%mm2    \n\t"

  966.         "punpckhwd     %5, %%mm3    \n\t"

  967.         "punpckhwd     %5, %%mm4    \n\t"

  968.         "punpckhwd     %5, %%mm5    \n\t"

  969.         "psllq         $8, %%mm1    \n\t"

  970.         "psllq        $16, %%mm2    \n\t"

  971.         "por        %%mm1, %%mm0    \n\t"

  972.         "por        %%mm2, %%mm0    \n\t"

  973.         "psllq         $8, %%mm4    \n\t"

  974.         "psllq        $16, %%mm5    \n\t"

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

  976.         "por        %%mm5, %%mm3    \n\t"

  977. 

  978.         "movq       %%mm0, %%mm6    \n\t"

  979.         "movq       %%mm3, %%mm7    \n\t"

  980. 

  981.         "movq         8%1, %%mm0    \n\t"

  982.         "movq         8%1, %%mm1    \n\t"

  983.         "movq         8%1, %%mm2    \n\t"

  984.         "pand          %2, %%mm0    \n\t"

  985.         "pand          %3, %%mm1    \n\t"

  986.         "pand          %4, %%mm2    \n\t"

  987.         "psllq         $3, %%mm0    \n\t"

  988.         "psrlq         $2, %%mm1    \n\t"

  989.         "psrlq         $7, %%mm2    \n\t"

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

  991.         "movq       %%mm1, %%mm4    \n\t"

  992.         "movq       %%mm2, %%mm5    \n\t"

  993.         "punpcklwd     %5, %%mm0    \n\t"

  994.         "punpcklwd     %5, %%mm1    \n\t"

  995.         "punpcklwd     %5, %%mm2    \n\t"

  996.         "punpckhwd     %5, %%mm3    \n\t"

  997.         "punpckhwd     %5, %%mm4    \n\t"

  998.         "punpckhwd     %5, %%mm5    \n\t"

  999.         "psllq         $8, %%mm1    \n\t"

  1000.         "psllq        $16, %%mm2    \n\t"

  1001.         "por        %%mm1, %%mm0    \n\t"

  1002.         "por        %%mm2, %%mm0    \n\t"

  1003.         "psllq         $8, %%mm4    \n\t"

  1004.         "psllq        $16, %%mm5    \n\t"

  1005.         "por        %%mm4, %%mm3    \n\t"

  1006.         "por        %%mm5, %%mm3    \n\t"

  1007. 

  1008.         :"=m"(*d)

  1009.         :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null)

  1010.         :"memory");

  1011.         /* Borrowed 32 to 24 */

  1012.         __asm __volatile(

  1013.         "movq       %%mm0, %%mm4    \n\t"

  1014.         "movq       %%mm3, %%mm5    \n\t"

  1015.         "movq       %%mm6, %%mm0    \n\t"

  1016.         "movq       %%mm7, %%mm1    \n\t"

  1017. 

  1018.         "movq       %%mm4, %%mm6    \n\t"

  1019.         "movq       %%mm5, %%mm7    \n\t"

  1020.         "movq       %%mm0, %%mm2    \n\t"

  1021.         "movq       %%mm1, %%mm3    \n\t"

  1022. 

  1023.         "psrlq         $8, %%mm2    \n\t"

  1024.         "psrlq         $8, %%mm3    \n\t"

  1025.         "psrlq         $8, %%mm6    \n\t"

  1026.         "psrlq         $8, %%mm7    \n\t"

  1027.         "pand          %2, %%mm0    \n\t"

  1028.         "pand          %2, %%mm1    \n\t"

  1029.         "pand          %2, %%mm4    \n\t"

  1030.         "pand          %2, %%mm5    \n\t"

  1031.         "pand          %3, %%mm2    \n\t"

  1032.         "pand          %3, %%mm3    \n\t"

  1033.         "pand          %3, %%mm6    \n\t"

  1034.         "pand          %3, %%mm7    \n\t"

  1035.         "por        %%mm2, %%mm0    \n\t"

  1036.         "por        %%mm3, %%mm1    \n\t"

  1037.         "por        %%mm6, %%mm4    \n\t"

  1038.         "por        %%mm7, %%mm5    \n\t"

  1039. 

  1040.         "movq       %%mm1, %%mm2    \n\t"

  1041.         "movq       %%mm4, %%mm3    \n\t"

  1042.         "psllq        $48, %%mm2    \n\t"

  1043.         "psllq        $32, %%mm3    \n\t"

  1044.         "pand          %4, %%mm2    \n\t"

  1045.         "pand          %5, %%mm3    \n\t"

  1046.         "por        %%mm2, %%mm0    \n\t"

  1047.         "psrlq        $16, %%mm1    \n\t"

  1048.         "psrlq        $32, %%mm4    \n\t"

  1049.         "psllq        $16, %%mm5    \n\t"

  1050.         "por        %%mm3, %%mm1    \n\t"

  1051.         "pand          %6, %%mm5    \n\t"

  1052.         "por        %%mm5, %%mm4    \n\t"

  1053. 

  1054.         MOVNTQ"     %%mm0,   %0     \n\t"

  1055.         MOVNTQ"     %%mm1,  8%0     \n\t"

  1056.         MOVNTQ"     %%mm4, 16%0"

  1057. 

  1058.         :"=m"(*d)

  1059.         :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh)

  1060.         :"memory");

  1061.         d += 24;

  1062.         s += 8;

  1063.     }

  1064.     __asm __volatile(SFENCE:::"memory");

  1065.     __asm __volatile(EMMS:::"memory");

  1066. #endif

  1067.     while (s < end)

  1068.     {

  1069.         register uint16_t bgr;

  1070.         bgr = *s++;

  1071.         *d++ = (bgr&0x1F)<<3;

  1072.         *d++ = (bgr&0x3E0)>>2;

  1073.         *d++ = (bgr&0x7C00)>>7;

  1074.     }

  1075. }

  1076. 

  1077. static inline void RENAME(rgb16to24)(const uint8_t *src, uint8_t *dst, long src_size)

  1078. {

  1079.     const uint16_t *end;

  1080. #ifdef HAVE_MMX

  1081.     const uint16_t *mm_end;

  1082. #endif

  1083.     uint8_t *d = (uint8_t *)dst;

  1084.     const uint16_t *s = (const uint16_t *)src;

  1085.     end = s + src_size/2;

  1086. #ifdef HAVE_MMX

  1087.     __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  1088.     mm_end = end - 7;

  1089.     while (s < mm_end)

  1090.     {

  1091.         __asm __volatile(

  1092.         PREFETCH"    32%1           \n\t"

  1093.         "movq          %1, %%mm0    \n\t"

  1094.         "movq          %1, %%mm1    \n\t"

  1095.         "movq          %1, %%mm2    \n\t"

  1096.         "pand          %2, %%mm0    \n\t"

  1097.         "pand          %3, %%mm1    \n\t"

  1098.         "pand          %4, %%mm2    \n\t"

  1099.         "psllq         $3, %%mm0    \n\t"

  1100.         "psrlq         $3, %%mm1    \n\t"

  1101.         "psrlq         $8, %%mm2    \n\t"

  1102.         "movq       %%mm0, %%mm3    \n\t"

  1103.         "movq       %%mm1, %%mm4    \n\t"

  1104.         "movq       %%mm2, %%mm5    \n\t"

  1105.         "punpcklwd     %5, %%mm0    \n\t"

  1106.         "punpcklwd     %5, %%mm1    \n\t"

  1107.         "punpcklwd     %5, %%mm2    \n\t"

  1108.         "punpckhwd     %5, %%mm3    \n\t"

  1109.         "punpckhwd     %5, %%mm4    \n\t"

  1110.         "punpckhwd     %5, %%mm5    \n\t"

  1111.         "psllq         $8, %%mm1    \n\t"

  1112.         "psllq        $16, %%mm2    \n\t"

  1113.         "por        %%mm1, %%mm0    \n\t"

  1114.         "por        %%mm2, %%mm0    \n\t"

  1115.         "psllq         $8, %%mm4    \n\t"

  1116.         "psllq        $16, %%mm5    \n\t"

  1117.         "por        %%mm4, %%mm3    \n\t"

  1118.         "por        %%mm5, %%mm3    \n\t"

  1119. 

  1120.         "movq       %%mm0, %%mm6    \n\t"

  1121.         "movq       %%mm3, %%mm7    \n\t"

  1122. 

  1123.         "movq         8%1, %%mm0    \n\t"

  1124.         "movq         8%1, %%mm1    \n\t"

  1125.         "movq         8%1, %%mm2    \n\t"

  1126.         "pand          %2, %%mm0    \n\t"

  1127.         "pand          %3, %%mm1    \n\t"

  1128.         "pand          %4, %%mm2    \n\t"

  1129.         "psllq         $3, %%mm0    \n\t"

  1130.         "psrlq         $3, %%mm1    \n\t"

  1131.         "psrlq         $8, %%mm2    \n\t"

  1132.         "movq       %%mm0, %%mm3    \n\t"

  1133.         "movq       %%mm1, %%mm4    \n\t"

  1134.         "movq       %%mm2, %%mm5    \n\t"

  1135.         "punpcklwd     %5, %%mm0    \n\t"

  1136.         "punpcklwd     %5, %%mm1    \n\t"

  1137.         "punpcklwd     %5, %%mm2    \n\t"

  1138.         "punpckhwd     %5, %%mm3    \n\t"

  1139.         "punpckhwd     %5, %%mm4    \n\t"

  1140.         "punpckhwd     %5, %%mm5    \n\t"

  1141.         "psllq         $8, %%mm1    \n\t"

  1142.         "psllq        $16, %%mm2    \n\t"

  1143.         "por        %%mm1, %%mm0    \n\t"

  1144.         "por        %%mm2, %%mm0    \n\t"

  1145.         "psllq         $8, %%mm4    \n\t"

  1146.         "psllq        $16, %%mm5    \n\t"

  1147.         "por        %%mm4, %%mm3    \n\t"

  1148.         "por        %%mm5, %%mm3    \n\t"

  1149.         :"=m"(*d)

  1150.         :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null)

  1151.         :"memory");

  1152.         /* Borrowed 32 to 24 */

  1153.         __asm __volatile(

  1154.         "movq       %%mm0, %%mm4    \n\t"

  1155.         "movq       %%mm3, %%mm5    \n\t"

  1156.         "movq       %%mm6, %%mm0    \n\t"

  1157.         "movq       %%mm7, %%mm1    \n\t"

  1158. 

  1159.         "movq       %%mm4, %%mm6    \n\t"

  1160.         "movq       %%mm5, %%mm7    \n\t"

  1161.         "movq       %%mm0, %%mm2    \n\t"

  1162.         "movq       %%mm1, %%mm3    \n\t"

  1163. 

  1164.         "psrlq         $8, %%mm2    \n\t"

  1165.         "psrlq         $8, %%mm3    \n\t"

  1166.         "psrlq         $8, %%mm6    \n\t"

  1167.         "psrlq         $8, %%mm7    \n\t"

  1168.         "pand          %2, %%mm0    \n\t"

  1169.         "pand          %2, %%mm1    \n\t"

  1170.         "pand          %2, %%mm4    \n\t"

  1171.         "pand          %2, %%mm5    \n\t"

  1172.         "pand          %3, %%mm2    \n\t"

  1173.         "pand          %3, %%mm3    \n\t"

  1174.         "pand          %3, %%mm6    \n\t"

  1175.         "pand          %3, %%mm7    \n\t"

  1176.         "por        %%mm2, %%mm0    \n\t"

  1177.         "por        %%mm3, %%mm1    \n\t"

  1178.         "por        %%mm6, %%mm4    \n\t"

  1179.         "por        %%mm7, %%mm5    \n\t"

  1180. 

  1181.         "movq       %%mm1, %%mm2    \n\t"

  1182.         "movq       %%mm4, %%mm3    \n\t"

  1183.         "psllq        $48, %%mm2    \n\t"

  1184.         "psllq        $32, %%mm3    \n\t"

  1185.         "pand          %4, %%mm2    \n\t"

  1186.         "pand          %5, %%mm3    \n\t"

  1187.         "por        %%mm2, %%mm0    \n\t"

  1188.         "psrlq        $16, %%mm1    \n\t"

  1189.         "psrlq        $32, %%mm4    \n\t"

  1190.         "psllq        $16, %%mm5    \n\t"

  1191.         "por        %%mm3, %%mm1    \n\t"

  1192.         "pand          %6, %%mm5    \n\t"

  1193.         "por        %%mm5, %%mm4    \n\t"

  1194. 

  1195.         MOVNTQ"     %%mm0,   %0     \n\t"

  1196.         MOVNTQ"     %%mm1,  8%0     \n\t"

  1197.         MOVNTQ"     %%mm4, 16%0"

  1198. 

  1199.         :"=m"(*d)

  1200.         :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh)

  1201.         :"memory");

  1202.         d += 24;

  1203.         s += 8;

  1204.     }

  1205.     __asm __volatile(SFENCE:::"memory");

  1206.     __asm __volatile(EMMS:::"memory");

  1207. #endif

  1208.     while (s < end)

  1209.     {

  1210.         register uint16_t bgr;

  1211.         bgr = *s++;

  1212.         *d++ = (bgr&0x1F)<<3;

  1213.         *d++ = (bgr&0x7E0)>>3;

  1214.         *d++ = (bgr&0xF800)>>8;

  1215.     }

  1216. }

  1217. 

  1218. static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size)

  1219. {

  1220.     const uint16_t *end;

  1221. #ifdef HAVE_MMX

  1222.     const uint16_t *mm_end;

  1223. #endif

  1224.     uint8_t *d = (uint8_t *)dst;

  1225.     const uint16_t *s = (const uint16_t *)src;

  1226.     end = s + src_size/2;

  1227. #ifdef HAVE_MMX

  1228.     __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  1229.     __asm __volatile("pxor    %%mm7,%%mm7    \n\t":::"memory");

  1230.     mm_end = end - 3;

  1231.     while (s < mm_end)

  1232.     {

  1233.         __asm __volatile(

  1234.         PREFETCH"    32%1           \n\t"

  1235.         "movq          %1, %%mm0    \n\t"

  1236.         "movq          %1, %%mm1    \n\t"

  1237.         "movq          %1, %%mm2    \n\t"

  1238.         "pand          %2, %%mm0    \n\t"

  1239.         "pand          %3, %%mm1    \n\t"

  1240.         "pand          %4, %%mm2    \n\t"

  1241.         "psllq         $3, %%mm0    \n\t"

  1242.         "psrlq         $2, %%mm1    \n\t"

  1243.         "psrlq         $7, %%mm2    \n\t"

  1244.         "movq       %%mm0, %%mm3    \n\t"

  1245.         "movq       %%mm1, %%mm4    \n\t"

  1246.         "movq       %%mm2, %%mm5    \n\t"

  1247.         "punpcklwd  %%mm7, %%mm0    \n\t"

  1248.         "punpcklwd  %%mm7, %%mm1    \n\t"

  1249.         "punpcklwd  %%mm7, %%mm2    \n\t"

  1250.         "punpckhwd  %%mm7, %%mm3    \n\t"

  1251.         "punpckhwd  %%mm7, %%mm4    \n\t"

  1252.         "punpckhwd  %%mm7, %%mm5    \n\t"

  1253.         "psllq         $8, %%mm1    \n\t"

  1254.         "psllq        $16, %%mm2    \n\t"

  1255.         "por        %%mm1, %%mm0    \n\t"

  1256.         "por        %%mm2, %%mm0    \n\t"

  1257.         "psllq         $8, %%mm4    \n\t"

  1258.         "psllq        $16, %%mm5    \n\t"

  1259.         "por        %%mm4, %%mm3    \n\t"

  1260.         "por        %%mm5, %%mm3    \n\t"

  1261.         MOVNTQ"     %%mm0,  %0      \n\t"

  1262.         MOVNTQ"     %%mm3, 8%0      \n\t"

  1263.         :"=m"(*d)

  1264.         :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r)

  1265.         :"memory");

  1266.         d += 16;

  1267.         s += 4;

  1268.     }

  1269.     __asm __volatile(SFENCE:::"memory");

  1270.     __asm __volatile(EMMS:::"memory");

  1271. #endif

  1272.     while (s < end)

  1273.     {

  1274. #if 0 //slightly slower on athlon

  1275.         int bgr= *s++;

  1276.         *((uint32_t*)d)++ = ((bgr&0x1F)<<3) + ((bgr&0x3E0)<<6) + ((bgr&0x7C00)<<9);

  1277. #else

  1278.         register uint16_t bgr;

  1279.         bgr = *s++;

  1280. #ifdef WORDS_BIGENDIAN

  1281.         *d++ = 0;

  1282.         *d++ = (bgr&0x7C00)>>7;

  1283.         *d++ = (bgr&0x3E0)>>2;

  1284.         *d++ = (bgr&0x1F)<<3;

  1285. #else

  1286.         *d++ = (bgr&0x1F)<<3;

  1287.         *d++ = (bgr&0x3E0)>>2;

  1288.         *d++ = (bgr&0x7C00)>>7;

  1289.         *d++ = 0;

  1290. #endif

  1291. 

  1292. #endif

  1293.     }

  1294. }

  1295. 

  1296. static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size)

  1297. {

  1298.     const uint16_t *end;

  1299. #ifdef HAVE_MMX

  1300.     const uint16_t *mm_end;

  1301. #endif

  1302.     uint8_t *d = (uint8_t *)dst;

  1303.     const uint16_t *s = (uint16_t *)src;

  1304.     end = s + src_size/2;

  1305. #ifdef HAVE_MMX

  1306.     __asm __volatile(PREFETCH"    %0"::"m"(*s):"memory");

  1307.     __asm __volatile("pxor    %%mm7,%%mm7    \n\t":::"memory");

  1308.     mm_end = end - 3;

  1309.     while (s < mm_end)

  1310.     {

  1311.         __asm __volatile(

  1312.         PREFETCH"    32%1           \n\t"

  1313.         "movq          %1, %%mm0    \n\t"

  1314.         "movq          %1, %%mm1    \n\t"

  1315.         "movq          %1, %%mm2    \n\t"

  1316.         "pand          %2, %%mm0    \n\t"

  1317.         "pand          %3, %%mm1    \n\t"

  1318.         "pand          %4, %%mm2    \n\t"

  1319.         "psllq         $3, %%mm0    \n\t"

  1320.         "psrlq         $3, %%mm1    \n\t"

  1321.         "psrlq         $8, %%mm2    \n\t"

  1322.         "movq       %%mm0, %%mm3    \n\t"

  1323.         "movq       %%mm1, %%mm4    \n\t"

  1324.         "movq       %%mm2, %%mm5    \n\t"

  1325.         "punpcklwd  %%mm7, %%mm0    \n\t"

  1326.         "punpcklwd  %%mm7, %%mm1    \n\t"

  1327.         "punpcklwd  %%mm7, %%mm2    \n\t"

  1328.         "punpckhwd  %%mm7, %%mm3    \n\t"

  1329.         "punpckhwd  %%mm7, %%mm4    \n\t"

  1330.         "punpckhwd  %%mm7, %%mm5    \n\t"

  1331.         "psllq         $8, %%mm1    \n\t"

  1332.         "psllq        $16, %%mm2    \n\t"

  1333.         "por        %%mm1, %%mm0    \n\t"

  1334.         "por        %%mm2, %%mm0    \n\t"

  1335.         "psllq         $8, %%mm4    \n\t"

  1336.         "psllq        $16, %%mm5    \n\t"

  1337.         "por        %%mm4, %%mm3    \n\t"

  1338.         "por        %%mm5, %%mm3    \n\t"

  1339.         MOVNTQ"     %%mm0, %0       \n\t"

  1340.         MOVNTQ"     %%mm3, 8%0      \n\t"

  1341.         :"=m"(*d)

  1342.         :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r)

  1343.         :"memory");

  1344.         d += 16;

  1345.         s += 4;

  1346.     }

  1347.     __asm __volatile(SFENCE:::"memory");

  1348.     __asm __volatile(EMMS:::"memory");

  1349. #endif

  1350.     while (s < end)

  1351.     {

  1352.         register uint16_t bgr;

  1353.         bgr = *s++;

  1354. #ifdef WORDS_BIGENDIAN

  1355.         *d++ = 0;

  1356.         *d++ = (bgr&0xF800)>>8;

  1357.         *d++ = (bgr&0x7E0)>>3;

  1358.         *d++ = (bgr&0x1F)<<3;

  1359. #else

  1360.         *d++ = (bgr&0x1F)<<3;

  1361.         *d++ = (bgr&0x7E0)>>3;

  1362.         *d++ = (bgr&0xF800)>>8;

  1363.         *d++ = 0;

  1364. #endif

  1365.     }

  1366. }

  1367. 

  1368. static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size)

  1369. {

  1370.     long idx = 15 - src_size;

  1371.     uint8_t *s = (uint8_t *) src-idx, *d = dst-idx;

  1372. #ifdef HAVE_MMX

  1373.     __asm __volatile(

  1374.     "test          %0, %0           \n\t"

  1375.     "jns           2f               \n\t"

  1376.     PREFETCH"       (%1, %0)        \n\t"

  1377.     "movq          %3, %%mm7        \n\t"

  1378.     "pxor          %4, %%mm7        \n\t"

  1379.     "movq       %%mm7, %%mm6        \n\t"

  1380.     "pxor          %5, %%mm7        \n\t"

  1381.     ASMALIGN(4)

  1382.     "1:                             \n\t"

  1383.     PREFETCH"     32(%1, %0)        \n\t"

  1384.     "movq           (%1, %0), %%mm0 \n\t"

  1385.     "movq          8(%1, %0), %%mm1 \n\t"

  1386. # ifdef HAVE_MMX2

  1387.     "pshufw      $177, %%mm0, %%mm3 \n\t"

  1388.     "pshufw      $177, %%mm1, %%mm5 \n\t"

  1389.     "pand       %%mm7, %%mm0        \n\t"

  1390.     "pand       %%mm6, %%mm3        \n\t"

  1391.     "pand       %%mm7, %%mm1        \n\t"

  1392.     "pand       %%mm6, %%mm5        \n\t"

  1393.     "por        %%mm3, %%mm0        \n\t"

  1394.     "por        %%mm5, %%mm1        \n\t"

  1395. # else

  1396.     "movq       %%mm0, %%mm2        \n\t"

  1397.     "movq       %%mm1, %%mm4        \n\t"

  1398.     "pand       %%mm7, %%mm0        \n\t"

  1399.     "pand       %%mm6, %%mm2        \n\t"

  1400.     "pand       %%mm7, %%mm1        \n\t"

  1401.     "pand       %%mm6, %%mm4        \n\t"

  1402.     "movq       %%mm2, %%mm3        \n\t"

  1403.     "movq       %%mm4, %%mm5        \n\t"

  1404.     "pslld        $16, %%mm2        \n\t"

  1405.     "psrld        $16, %%mm3        \n\t"

  1406.     "pslld        $16, %%mm4        \n\t"

  1407.     "psrld        $16, %%mm5        \n\t"

  1408.     "por        %%mm2, %%mm0        \n\t"

  1409.     "por        %%mm4, %%mm1        \n\t"

  1410.     "por        %%mm3, %%mm0        \n\t"

  1411.     "por        %%mm5, %%mm1        \n\t"

  1412. # endif

  1413.     MOVNTQ"     %%mm0,  (%2, %0)    \n\t"

  1414.     MOVNTQ"     %%mm1, 8(%2, %0)    \n\t"

  1415.     "add          $16, %0           \n\t"

  1416.     "js            1b               \n\t"

  1417.     SFENCE"                         \n\t"

  1418.     EMMS"                           \n\t"

  1419.     "2:                             \n\t"

  1420.     : "+&r"(idx)

  1421.     : "r" (s), "r" (d), "m" (mask32b), "m" (mask32r), "m" (mmx_one)

  1422.     : "memory");

  1423. #endif

  1424.     for (; idx<15; idx+=4) {

  1425.         register int v = *(uint32_t *)&s[idx], g = v & 0xff00ff00;

  1426.         v &= 0xff00ff;

  1427.         *(uint32_t *)&d[idx] = (v>>16) + g + (v<<16);

  1428.     }

  1429. }

  1430. 

  1431. static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size)

  1432. {

  1433.     unsigned i;

  1434. #ifdef HAVE_MMX

  1435.     long mmx_size= 23 - src_size;

  1436.     asm volatile (

  1437.     "test             %%"REG_a", %%"REG_a"          \n\t"

  1438.     "jns                     2f                     \n\t"

  1439.     "movq     "MANGLE(mask24r)", %%mm5              \n\t"

  1440.     "movq     "MANGLE(mask24g)", %%mm6              \n\t"

  1441.     "movq     "MANGLE(mask24b)", %%mm7              \n\t"

  1442.     ASMALIGN(4)

  1443.     "1:                                             \n\t"

  1444.     PREFETCH" 32(%1, %%"REG_a")                     \n\t"

  1445.     "movq       (%1, %%"REG_a"), %%mm0              \n\t" // BGR BGR BG

  1446.     "movq       (%1, %%"REG_a"), %%mm1              \n\t" // BGR BGR BG

  1447.     "movq      2(%1, %%"REG_a"), %%mm2              \n\t" // R BGR BGR B

  1448.     "psllq                  $16, %%mm0              \n\t" // 00 BGR BGR

  1449.     "pand                 %%mm5, %%mm0              \n\t"

  1450.     "pand                 %%mm6, %%mm1              \n\t"

  1451.     "pand                 %%mm7, %%mm2              \n\t"

  1452.     "por                  %%mm0, %%mm1              \n\t"

  1453.     "por                  %%mm2, %%mm1              \n\t"

  1454.     "movq      6(%1, %%"REG_a"), %%mm0              \n\t" // BGR BGR BG

  1455.     MOVNTQ"               %%mm1,   (%2, %%"REG_a")  \n\t" // RGB RGB RG

  1456.     "movq      8(%1, %%"REG_a"), %%mm1              \n\t" // R BGR BGR B

  1457.     "movq     10(%1, %%"REG_a"), %%mm2              \n\t" // GR BGR BGR

  1458.     "pand                 %%mm7, %%mm0              \n\t"

  1459.     "pand                 %%mm5, %%mm1              \n\t"

  1460.     "pand                 %%mm6, %%mm2              \n\t"

  1461.     "por                  %%mm0, %%mm1              \n\t"

  1462.     "por                  %%mm2, %%mm1              \n\t"

  1463.     "movq     14(%1, %%"REG_a"), %%mm0              \n\t" // R BGR BGR B

  1464.     MOVNTQ"               %%mm1,  8(%2, %%"REG_a")  \n\t" // B RGB RGB R

  1465.     "movq     16(%1, %%"REG_a"), %%mm1              \n\t" // GR BGR BGR

  1466.     "movq     18(%1, %%"REG_a"), %%mm2              \n\t" // BGR BGR BG

  1467.     "pand                 %%mm6, %%mm0              \n\t"

  1468.     "pand                 %%mm7, %%mm1              \n\t"

  1469.     "pand                 %%mm5, %%mm2              \n\t"

  1470.     "por                  %%mm0, %%mm1              \n\t"

  1471.     "por                  %%mm2, %%mm1              \n\t"

  1472.     MOVNTQ"               %%mm1, 16(%2, %%"REG_a")  \n\t"

  1473.     "add                    $24, %%"REG_a"          \n\t"

  1474.     " js                     1b                     \n\t"

  1475.     "2:                                             \n\t"

  1476.     : "+a" (mmx_size)

  1477.     : "r" (src-mmx_size), "r"(dst-mmx_size)

  1478.     );

  1479. 

  1480.     __asm __volatile(SFENCE:::"memory");

  1481.     __asm __volatile(EMMS:::"memory");

  1482. 

  1483.     if (mmx_size==23) return; //finihsed, was multiple of 8

  1484. 

  1485.     src+= src_size;

  1486.     dst+= src_size;

  1487.     src_size= 23-mmx_size;

  1488.     src-= src_size;

  1489.     dst-= src_size;

  1490. #endif

  1491.     for (i=0; i<src_size; i+=3)

  1492.     {

  1493.         register uint8_t x;

  1494.         x          = src[i + 2];

  1495.         dst[i + 1] = src[i + 1];

  1496.         dst[i + 2] = src[i + 0];

  1497.         dst[i + 0] = x;

  1498.     }

  1499. }

  1500. 

  1501. static inline void RENAME(yuvPlanartoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,

  1502.                                            long width, long height,

  1503.                                            long lumStride, long chromStride, long dstStride, long vertLumPerChroma)

  1504. {

  1505.     long y;

  1506.     const long chromWidth= width>>1;

  1507.     for (y=0; y<height; y++)

  1508.     {

  1509. #ifdef HAVE_MMX

  1510. //FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)

  1511.         asm volatile(

  1512.         "xor                 %%"REG_a", %%"REG_a"   \n\t"

  1513.         ASMALIGN(4)

  1514.         "1:                                         \n\t"

  1515.         PREFETCH"    32(%1, %%"REG_a", 2)           \n\t"

  1516.         PREFETCH"    32(%2, %%"REG_a")              \n\t"

  1517.         PREFETCH"    32(%3, %%"REG_a")              \n\t"

  1518.         "movq          (%2, %%"REG_a"), %%mm0       \n\t" // U(0)

  1519.         "movq                    %%mm0, %%mm2       \n\t" // U(0)

  1520.         "movq          (%3, %%"REG_a"), %%mm1       \n\t" // V(0)

  1521.         "punpcklbw               %%mm1, %%mm0       \n\t" // UVUV UVUV(0)

  1522.         "punpckhbw               %%mm1, %%mm2       \n\t" // UVUV UVUV(8)

  1523. 

  1524.         "movq        (%1, %%"REG_a",2), %%mm3       \n\t" // Y(0)

  1525.         "movq       8(%1, %%"REG_a",2), %%mm5       \n\t" // Y(8)

  1526.         "movq                    %%mm3, %%mm4       \n\t" // Y(0)

  1527.         "movq                    %%mm5, %%mm6       \n\t" // Y(8)

  1528.         "punpcklbw               %%mm0, %%mm3       \n\t" // YUYV YUYV(0)

  1529.         "punpckhbw               %%mm0, %%mm4       \n\t" // YUYV YUYV(4)

  1530.         "punpcklbw               %%mm2, %%mm5       \n\t" // YUYV YUYV(8)

  1531.         "punpckhbw               %%mm2, %%mm6       \n\t" // YUYV YUYV(12)

  1532. 

  1533.         MOVNTQ"                  %%mm3,   (%0, %%"REG_a", 4)    \n\t"

  1534.         MOVNTQ"                  %%mm4,  8(%0, %%"REG_a", 4)    \n\t"

  1535.         MOVNTQ"                  %%mm5, 16(%0, %%"REG_a", 4)    \n\t"

  1536.         MOVNTQ"                  %%mm6, 24(%0, %%"REG_a", 4)    \n\t"

  1537. 

  1538.         "add                        $8, %%"REG_a"   \n\t"

  1539.         "cmp                        %4, %%"REG_a"   \n\t"

  1540.         " jb                        1b              \n\t"

  1541.         ::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (chromWidth)

  1542.         : "%"REG_a

  1543.         );

  1544. #else

  1545. 

  1546. #if defined ARCH_ALPHA && defined HAVE_MVI

  1547. #define pl2yuy2(n)                  \

  1548.     y1 = yc[n];                     \

  1549.     y2 = yc2[n];                    \

  1550.     u = uc[n];                      \

  1551.     v = vc[n];                      \

  1552.     asm("unpkbw %1, %0" : "=r"(y1) : "r"(y1));  \

  1553.     asm("unpkbw %1, %0" : "=r"(y2) : "r"(y2));  \

  1554.     asm("unpkbl %1, %0" : "=r"(u) : "r"(u));    \

  1555.     asm("unpkbl %1, %0" : "=r"(v) : "r"(v));    \

  1556.     yuv1 = (u << 8) + (v << 24);                \

  1557.     yuv2 = yuv1 + y2;               \

  1558.     yuv1 += y1;                     \

  1559.     qdst[n]  = yuv1;                \

  1560.     qdst2[n] = yuv2;

  1561. 

  1562.         int i;

  1563.         uint64_t *qdst = (uint64_t *) dst;

  1564.         uint64_t *qdst2 = (uint64_t *) (dst + dstStride);

  1565.         const uint32_t *yc = (uint32_t *) ysrc;

  1566.         const uint32_t *yc2 = (uint32_t *) (ysrc + lumStride);

  1567.         const uint16_t *uc = (uint16_t*) usrc, *vc = (uint16_t*) vsrc;

  1568.         for (i = 0; i < chromWidth; i += 8){

  1569.             uint64_t y1, y2, yuv1, yuv2;

  1570.             uint64_t u, v;

  1571.             /* Prefetch */

  1572.             asm("ldq $31,64(%0)" :: "r"(yc));

  1573.             asm("ldq $31,64(%0)" :: "r"(yc2));

  1574.             asm("ldq $31,64(%0)" :: "r"(uc));

  1575.             asm("ldq $31,64(%0)" :: "r"(vc));

  1576. 

  1577.             pl2yuy2(0);

  1578.             pl2yuy2(1);

  1579.             pl2yuy2(2);

  1580.             pl2yuy2(3);

  1581. 

  1582.             yc    += 4;

  1583.             yc2   += 4;

  1584.             uc    += 4;

  1585.             vc    += 4;

  1586.             qdst  += 4;

  1587.             qdst2 += 4;

  1588.         }

  1589.         y++;

  1590.         ysrc += lumStride;

  1591.         dst += dstStride;

  1592. 

  1593. #elif __WORDSIZE >= 64

  1594.         int i;

  1595.         uint64_t *ldst = (uint64_t *) dst;

  1596.         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;

  1597.         for (i = 0; i < chromWidth; i += 2){

  1598.             uint64_t k, l;

  1599.             k = yc[0] + (uc[0] << 8) +

  1600.                 (yc[1] << 16) + (vc[0] << 24);

  1601.             l = yc[2] + (uc[1] << 8) +

  1602.                 (yc[3] << 16) + (vc[1] << 24);

  1603.             *ldst++ = k + (l << 32);

  1604.             yc += 4;

  1605.             uc += 2;

  1606.             vc += 2;

  1607.         }

  1608. 

  1609. #else

  1610.         int i, *idst = (int32_t *) dst;

  1611.         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;

  1612.         for (i = 0; i < chromWidth; i++){

  1613. #ifdef WORDS_BIGENDIAN

  1614.             *idst++ = (yc[0] << 24)+ (uc[0] << 16) +

  1615.                 (yc[1] << 8) + (vc[0] << 0);

  1616. #else

  1617.             *idst++ = yc[0] + (uc[0] << 8) +

  1618.                 (yc[1] << 16) + (vc[0] << 24);

  1619. #endif

  1620.             yc += 2;

  1621.             uc++;

  1622.             vc++;

  1623.         }

  1624. #endif

  1625. #endif

  1626.         if ((y&(vertLumPerChroma-1))==(vertLumPerChroma-1) )

  1627.         {

  1628.             usrc += chromStride;

  1629.             vsrc += chromStride;

  1630.         }

  1631.         ysrc += lumStride;

  1632.         dst  += dstStride;

  1633.     }

  1634. #ifdef HAVE_MMX

  1635. asm(    EMMS"       \n\t"

  1636.         SFENCE"     \n\t"

  1637.         :::"memory");

  1638. #endif

  1639. }

  1640. 

  1641. /**

  1642.  *

  1643.  * height should be a multiple of 2 and width should be a multiple of 16 (if this is a

  1644.  * problem for anyone then tell me, and ill fix it)

  1645.  */

  1646. static inline void RENAME(yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,

  1647.                                       long width, long height,

  1648.                                       long lumStride, long chromStride, long dstStride)

  1649. {

  1650.     //FIXME interpolate chroma

  1651.     RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2);

  1652. }

  1653. 

  1654. static inline void RENAME(yuvPlanartouyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,

  1655.                                            long width, long height,

  1656.                                            long lumStride, long chromStride, long dstStride, long vertLumPerChroma)

  1657. {

  1658.     long y;

  1659.     const long chromWidth= width>>1;

  1660.     for (y=0; y<height; y++)

  1661.     {

  1662. #ifdef HAVE_MMX

  1663. //FIXME handle 2 lines a once (fewer prefetch, reuse some chrom, but very likely limited by mem anyway)

  1664.         asm volatile(

  1665.         "xor                %%"REG_a", %%"REG_a"    \n\t"

  1666.         ASMALIGN(4)

  1667.         "1:                                         \n\t"

  1668.         PREFETCH"   32(%1, %%"REG_a", 2)            \n\t"

  1669.         PREFETCH"   32(%2, %%"REG_a")               \n\t"

  1670.         PREFETCH"   32(%3, %%"REG_a")               \n\t"

  1671.         "movq         (%2, %%"REG_a"), %%mm0        \n\t" // U(0)

  1672.         "movq                   %%mm0, %%mm2        \n\t" // U(0)

  1673.         "movq         (%3, %%"REG_a"), %%mm1        \n\t" // V(0)

  1674.         "punpcklbw              %%mm1, %%mm0        \n\t" // UVUV UVUV(0)

  1675.         "punpckhbw              %%mm1, %%mm2        \n\t" // UVUV UVUV(8)

  1676. 

  1677.         "movq       (%1, %%"REG_a",2), %%mm3        \n\t" // Y(0)

  1678.         "movq      8(%1, %%"REG_a",2), %%mm5        \n\t" // Y(8)

  1679.         "movq                   %%mm0, %%mm4        \n\t" // Y(0)

  1680.         "movq                   %%mm2, %%mm6        \n\t" // Y(8)

  1681.         "punpcklbw              %%mm3, %%mm0        \n\t" // YUYV YUYV(0)

  1682.         "punpckhbw              %%mm3, %%mm4        \n\t" // YUYV YUYV(4)

  1683.         "punpcklbw              %%mm5, %%mm2        \n\t" // YUYV YUYV(8)

  1684.         "punpckhbw              %%mm5, %%mm6        \n\t" // YUYV YUYV(12)

  1685. 

  1686.         MOVNTQ"                 %%mm0,   (%0, %%"REG_a", 4)     \n\t"

  1687.         MOVNTQ"                 %%mm4,  8(%0, %%"REG_a", 4)     \n\t"

  1688.         MOVNTQ"                 %%mm2, 16(%0, %%"REG_a", 4)     \n\t"

  1689.         MOVNTQ"                 %%mm6, 24(%0, %%"REG_a", 4)     \n\t"

  1690. 

  1691.         "add                       $8, %%"REG_a"    \n\t"

  1692.         "cmp                       %4, %%"REG_a"    \n\t"

  1693.         " jb                       1b               \n\t"

  1694.         ::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (chromWidth)

  1695.         : "%"REG_a

  1696.         );

  1697. #else

  1698. //FIXME adapt the alpha asm code from yv12->yuy2

  1699. 

  1700. #if __WORDSIZE >= 64

  1701.         int i;

  1702.         uint64_t *ldst = (uint64_t *) dst;

  1703.         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;

  1704.         for (i = 0; i < chromWidth; i += 2){

  1705.             uint64_t k, l;

  1706.             k = uc[0] + (yc[0] << 8) +

  1707.                 (vc[0] << 16) + (yc[1] << 24);

  1708.             l = uc[1] + (yc[2] << 8) +

  1709.                 (vc[1] << 16) + (yc[3] << 24);

  1710.             *ldst++ = k + (l << 32);

  1711.             yc += 4;

  1712.             uc += 2;

  1713.             vc += 2;

  1714.         }

  1715. 

  1716. #else

  1717.         int i, *idst = (int32_t *) dst;

  1718.         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;

  1719.         for (i = 0; i < chromWidth; i++){

  1720. #ifdef WORDS_BIGENDIAN

  1721.             *idst++ = (uc[0] << 24)+ (yc[0] << 16) +

  1722.                 (vc[0] << 8) + (yc[1] << 0);

  1723. #else

  1724.             *idst++ = uc[0] + (yc[0] << 8) +

  1725.                 (vc[0] << 16) + (yc[1] << 24);

  1726. #endif

  1727.             yc += 2;

  1728.             uc++;

  1729.             vc++;

  1730.         }

  1731. #endif

  1732. #endif

  1733.         if ((y&(vertLumPerChroma-1))==(vertLumPerChroma-1) )

  1734.         {

  1735.             usrc += chromStride;

  1736.             vsrc += chromStride;

  1737.         }

  1738.         ysrc += lumStride;

  1739.         dst += dstStride;

  1740.     }

  1741. #ifdef HAVE_MMX

  1742. asm(    EMMS"       \n\t"

  1743.         SFENCE"     \n\t"

  1744.         :::"memory");

  1745. #endif

  1746. }

  1747. 

  1748. /**

  1749.  *

  1750.  * height should be a multiple of 2 and width should be a multiple of 16 (if this is a

  1751.  * problem for anyone then tell me, and ill fix it)

  1752.  */

  1753. static inline void RENAME(yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,

  1754.                                       long width, long height,

  1755.                                       long lumStride, long chromStride, long dstStride)

  1756. {

  1757.     //FIXME interpolate chroma

  1758.     RENAME(yuvPlanartouyvy)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2);

  1759. }

  1760. 

  1761. /**

  1762.  *

  1763.  * width should be a multiple of 16

  1764.  */

  1765. static inline void RENAME(yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,

  1766.                                          long width, long height,

  1767.                                          long lumStride, long chromStride, long dstStride)

  1768. {

  1769.     RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 1);

  1770. }

  1771. 

  1772. /**

  1773.  *

  1774.  * height should be a multiple of 2 and width should be a multiple of 16 (if this is a

  1775.  * problem for anyone then tell me, and ill fix it)

  1776.  */

  1777. static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,

  1778.                                       long width, long height,

  1779.                                       long lumStride, long chromStride, long srcStride)

  1780. {

  1781.     long y;

  1782.     const long chromWidth= width>>1;

  1783.     for (y=0; y<height; y+=2)

  1784.     {

  1785. #ifdef HAVE_MMX

  1786.         asm volatile(

  1787.         "xor                 %%"REG_a", %%"REG_a"   \n\t"

  1788.         "pcmpeqw                 %%mm7, %%mm7       \n\t"

  1789.         "psrlw                      $8, %%mm7       \n\t" // FF,00,FF,00...

  1790.         ASMALIGN(4)

  1791.         "1:                \n\t"

  1792.         PREFETCH" 64(%0, %%"REG_a", 4)              \n\t"

  1793.         "movq       (%0, %%"REG_a", 4), %%mm0       \n\t" // YUYV YUYV(0)

  1794.         "movq      8(%0, %%"REG_a", 4), %%mm1       \n\t" // YUYV YUYV(4)

  1795.         "movq                    %%mm0, %%mm2       \n\t" // YUYV YUYV(0)

  1796.         "movq                    %%mm1, %%mm3       \n\t" // YUYV YUYV(4)

  1797.         "psrlw                      $8, %%mm0       \n\t" // U0V0 U0V0(0)

  1798.         "psrlw                      $8, %%mm1       \n\t" // U0V0 U0V0(4)

  1799.         "pand                    %%mm7, %%mm2       \n\t" // Y0Y0 Y0Y0(0)

  1800.         "pand                    %%mm7, %%mm3       \n\t" // Y0Y0 Y0Y0(4)

  1801.         "packuswb                %%mm1, %%mm0       \n\t" // UVUV UVUV(0)

  1802.         "packuswb                %%mm3, %%mm2       \n\t" // YYYY YYYY(0)

  1803. 

  1804.         MOVNTQ"                  %%mm2, (%1, %%"REG_a", 2)  \n\t"

  1805. 

  1806.         "movq     16(%0, %%"REG_a", 4), %%mm1       \n\t" // YUYV YUYV(8)

  1807.         "movq     24(%0, %%"REG_a", 4), %%mm2       \n\t" // YUYV YUYV(12)

  1808.         "movq                    %%mm1, %%mm3       \n\t" // YUYV YUYV(8)

  1809.         "movq                    %%mm2, %%mm4       \n\t" // YUYV YUYV(12)

  1810.         "psrlw                      $8, %%mm1       \n\t" // U0V0 U0V0(8)

  1811.         "psrlw                      $8, %%mm2       \n\t" // U0V0 U0V0(12)

  1812.         "pand                    %%mm7, %%mm3       \n\t" // Y0Y0 Y0Y0(8)

  1813.         "pand                    %%mm7, %%mm4       \n\t" // Y0Y0 Y0Y0(12)

  1814.         "packuswb                %%mm2, %%mm1       \n\t" // UVUV UVUV(8)

  1815.         "packuswb                %%mm4, %%mm3       \n\t" // YYYY YYYY(8)

  1816. 

  1817.         MOVNTQ"                  %%mm3, 8(%1, %%"REG_a", 2) \n\t"

  1818. 

  1819.         "movq                    %%mm0, %%mm2       \n\t" // UVUV UVUV(0)

  1820.         "movq                    %%mm1, %%mm3       \n\t" // UVUV UVUV(8)

  1821.         "psrlw                      $8, %%mm0       \n\t" // V0V0 V0V0(0)

  1822.         "psrlw                      $8, %%mm1       \n\t" // V0V0 V0V0(8)

  1823.         "pand                    %%mm7, %%mm2       \n\t" // U0U0 U0U0(0)

  1824.         "pand                    %%mm7, %%mm3       \n\t" // U0U0 U0U0(8)

  1825.         "packuswb                %%mm1, %%mm0       \n\t" // VVVV VVVV(0)

  1826.         "packuswb                %%mm3, %%mm2       \n\t" // UUUU UUUU(0)

  1827. 

  1828.         MOVNTQ"                  %%mm0, (%3, %%"REG_a")     \n\t"

  1829.         MOVNTQ"                  %%mm2, (%2, %%"REG_a")     \n\t"

  1830. 

  1831.         "add                        $8, %%"REG_a"   \n\t"

  1832.         "cmp                        %4, %%"REG_a"   \n\t"

  1833.         " jb                        1b              \n\t"

  1834.         ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)

  1835.         : "memory", "%"REG_a

  1836.         );

  1837. 

  1838.         ydst += lumStride;

  1839.         src  += srcStride;

  1840. 

  1841.         asm volatile(

  1842.         "xor                 %%"REG_a", %%"REG_a"   \n\t"

  1843.         ASMALIGN(4)

  1844.         "1:                                         \n\t"

  1845.         PREFETCH" 64(%0, %%"REG_a", 4)              \n\t"

  1846.         "movq       (%0, %%"REG_a", 4), %%mm0       \n\t" // YUYV YUYV(0)

  1847.         "movq      8(%0, %%"REG_a", 4), %%mm1       \n\t" // YUYV YUYV(4)

  1848.         "movq     16(%0, %%"REG_a", 4), %%mm2       \n\t" // YUYV YUYV(8)

  1849.         "movq     24(%0, %%"REG_a", 4), %%mm3       \n\t" // YUYV YUYV(12)

  1850.         "pand                    %%mm7, %%mm0       \n\t" // Y0Y0 Y0Y0(0)

  1851.         "pand                    %%mm7, %%mm1       \n\t" // Y0Y0 Y0Y0(4)

  1852.         "pand                    %%mm7, %%mm2       \n\t" // Y0Y0 Y0Y0(8)

  1853.         "pand                    %%mm7, %%mm3       \n\t" // Y0Y0 Y0Y0(12)

  1854.         "packuswb                %%mm1, %%mm0       \n\t" // YYYY YYYY(0)

  1855.         "packuswb                %%mm3, %%mm2       \n\t" // YYYY YYYY(8)

  1856. 

  1857.         MOVNTQ"                  %%mm0,  (%1, %%"REG_a", 2) \n\t"

  1858.         MOVNTQ"                  %%mm2, 8(%1, %%"REG_a", 2) \n\t"

  1859. 

  1860.         "add                        $8, %%"REG_a"   \n\t"

  1861.         "cmp                        %4, %%"REG_a"   \n\t"

  1862.         " jb                        1b              \n\t"

  1863. 

  1864.         ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)

  1865.         : "memory", "%"REG_a

  1866.         );

  1867. #else

  1868.         long i;

  1869.         for (i=0; i<chromWidth; i++)

  1870.         {

  1871.             ydst[2*i+0]     = src[4*i+0];

  1872.             udst[i]     = src[4*i+1];

  1873.             ydst[2*i+1]     = src[4*i+2];

  1874.             vdst[i]     = src[4*i+3];

  1875.         }

  1876.         ydst += lumStride;

  1877.         src  += srcStride;

  1878. 

  1879.         for (i=0; i<chromWidth; i++)

  1880.         {

  1881.             ydst[2*i+0]     = src[4*i+0];

  1882.             ydst[2*i+1]     = src[4*i+2];

  1883.         }

  1884. #endif

  1885.         udst += chromStride;

  1886.         vdst += chromStride;

  1887.         ydst += lumStride;

  1888.         src  += srcStride;

  1889.     }

  1890. #ifdef HAVE_MMX

  1891. asm volatile(   EMMS"       \n\t"

  1892.                 SFENCE"     \n\t"

  1893.                 :::"memory");

  1894. #endif

  1895. }

  1896. 

  1897. static inline void RENAME(yvu9toyv12)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc,

  1898.                                       uint8_t *ydst, uint8_t *udst, uint8_t *vdst,

  1899.                                       long width, long height, long lumStride, long chromStride)

  1900. {

  1901.     /* Y Plane */

  1902.     memcpy(ydst, ysrc, width*height);

  1903. 

  1904.     /* XXX: implement upscaling for U,V */

  1905. }

  1906. 

  1907. static inline void RENAME(planar2x)(const uint8_t *src, uint8_t *dst, long srcWidth, long srcHeight, long srcStride, long dstStride)

  1908. {

  1909.     long x,y;

  1910. 

  1911.     dst[0]= src[0];

  1912. 

  1913.     // first line

  1914.     for (x=0; x<srcWidth-1; x++){

  1915.         dst[2*x+1]= (3*src[x] +   src[x+1])>>2;

  1916.         dst[2*x+2]= (  src[x] + 3*src[x+1])>>2;

  1917.     }

  1918.     dst[2*srcWidth-1]= src[srcWidth-1];

  1919. 

  1920.         dst+= dstStride;

  1921. 

  1922.     for (y=1; y<srcHeight; y++){

  1923. #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)

  1924.         const long mmxSize= srcWidth&~15;

  1925.         asm volatile(

  1926.         "mov           %4, %%"REG_a"            \n\t"

  1927.         "1:                                     \n\t"

  1928.         "movq         (%0, %%"REG_a"), %%mm0    \n\t"

  1929.         "movq         (%1, %%"REG_a"), %%mm1    \n\t"

  1930.         "movq        1(%0, %%"REG_a"), %%mm2    \n\t"

  1931.         "movq        1(%1, %%"REG_a"), %%mm3    \n\t"

  1932.         "movq       -1(%0, %%"REG_a"), %%mm4    \n\t"

  1933.         "movq       -1(%1, %%"REG_a"), %%mm5    \n\t"

  1934.         PAVGB"                  %%mm0, %%mm5    \n\t"

  1935.         PAVGB"                  %%mm0, %%mm3    \n\t"

  1936.         PAVGB"                  %%mm0, %%mm5    \n\t"

  1937.         PAVGB"                  %%mm0, %%mm3    \n\t"

  1938.         PAVGB"                  %%mm1, %%mm4    \n\t"

  1939.         PAVGB"                  %%mm1, %%mm2    \n\t"

  1940.         PAVGB"                  %%mm1, %%mm4    \n\t"

  1941.         PAVGB"                  %%mm1, %%mm2    \n\t"

  1942.         "movq                   %%mm5, %%mm7    \n\t"

  1943.         "movq                   %%mm4, %%mm6    \n\t"

  1944.         "punpcklbw              %%mm3, %%mm5    \n\t"

  1945.         "punpckhbw              %%mm3, %%mm7    \n\t"

  1946.         "punpcklbw              %%mm2, %%mm4    \n\t"

  1947.         "punpckhbw              %%mm2, %%mm6    \n\t"

  1948. #if 1

  1949.         MOVNTQ"                 %%mm5,  (%2, %%"REG_a", 2)  \n\t"

  1950.         MOVNTQ"                 %%mm7, 8(%2, %%"REG_a", 2)  \n\t"

  1951.         MOVNTQ"                 %%mm4,  (%3, %%"REG_a", 2)  \n\t"

  1952.         MOVNTQ"                 %%mm6, 8(%3, %%"REG_a", 2)  \n\t"

  1953. #else

  1954.         "movq                   %%mm5,  (%2, %%"REG_a", 2)  \n\t"

  1955.         "movq                   %%mm7, 8(%2, %%"REG_a", 2)  \n\t"

  1956.         "movq                   %%mm4,  (%3, %%"REG_a", 2)  \n\t"

  1957.         "movq                   %%mm6, 8(%3, %%"REG_a", 2)  \n\t"

  1958. #endif

  1959.         "add                       $8, %%"REG_a"            \n\t"

  1960.         " js                       1b                       \n\t"

  1961.         :: "r" (src + mmxSize  ), "r" (src + srcStride + mmxSize  ),

  1962.            "r" (dst + mmxSize*2), "r" (dst + dstStride + mmxSize*2),

  1963.            "g" (-mmxSize)

  1964.         : "%"REG_a

  1965. 

  1966.         );

  1967. #else

  1968.         const long mmxSize=1;

  1969. #endif

  1970.         dst[0        ]= (3*src[0] +   src[srcStride])>>2;

  1971.         dst[dstStride]= (  src[0] + 3*src[srcStride])>>2;

  1972. 

  1973.         for (x=mmxSize-1; x<srcWidth-1; x++){

  1974.             dst[2*x          +1]= (3*src[x+0] +   src[x+srcStride+1])>>2;

  1975.             dst[2*x+dstStride+2]= (  src[x+0] + 3*src[x+srcStride+1])>>2;

  1976.             dst[2*x+dstStride+1]= (  src[x+1] + 3*src[x+srcStride  ])>>2;

  1977.             dst[2*x          +2]= (3*src[x+1] +   src[x+srcStride  ])>>2;

  1978.         }

  1979.         dst[srcWidth*2 -1            ]= (3*src[srcWidth-1] +   src[srcWidth-1 + srcStride])>>2;

  1980.         dst[srcWidth*2 -1 + dstStride]= (  src[srcWidth-1] + 3*src[srcWidth-1 + srcStride])>>2;

  1981. 

  1982.         dst+=dstStride*2;

  1983.         src+=srcStride;

  1984.     }

  1985. 

  1986.     // last line

  1987. #if 1

  1988.     dst[0]= src[0];

  1989. 

  1990.     for (x=0; x<srcWidth-1; x++){

  1991.         dst[2*x+1]= (3*src[x] +   src[x+1])>>2;

  1992.         dst[2*x+2]= (  src[x] + 3*src[x+1])>>2;

  1993.     }

  1994.     dst[2*srcWidth-1]= src[srcWidth-1];

  1995. #else

  1996.     for (x=0; x<srcWidth; x++){

  1997.         dst[2*x+0]=

  1998.         dst[2*x+1]= src[x];

  1999.     }

  2000. #endif

  2001. 

  2002. #ifdef HAVE_MMX

  2003. asm volatile(   EMMS"       \n\t"

  2004.                 SFENCE"     \n\t"

  2005.                 :::"memory");

  2006. #endif

  2007. }

  2008. 

  2009. /**

  2010.  *

  2011.  * height should be a multiple of 2 and width should be a multiple of 16 (if this is a

  2012.  * problem for anyone then tell me, and ill fix it)

  2013.  * chrominance data is only taken from every secound line others are ignored FIXME write HQ version

  2014.  */

  2015. static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,

  2016.                                       long width, long height,

  2017.                                       long lumStride, long chromStride, long srcStride)

  2018. {

  2019.     long y;

  2020.     const long chromWidth= width>>1;

  2021.     for (y=0; y<height; y+=2)

  2022.     {

  2023. #ifdef HAVE_MMX

  2024.         asm volatile(

  2025.         "xorl                %%eax, %%eax   \n\t"

  2026.         "pcmpeqw             %%mm7, %%mm7   \n\t"

  2027.         "psrlw                  $8, %%mm7   \n\t" // FF,00,FF,00...

  2028.         ASMALIGN(4)

  2029.         "1:                                 \n\t"

  2030.         PREFETCH" 64(%0, %%eax, 4)          \n\t"

  2031.         "movq       (%0, %%eax, 4), %%mm0   \n\t" // UYVY UYVY(0)

  2032.         "movq      8(%0, %%eax, 4), %%mm1   \n\t" // UYVY UYVY(4)

  2033.         "movq                %%mm0, %%mm2   \n\t" // UYVY UYVY(0)

  2034.         "movq                %%mm1, %%mm3   \n\t" // UYVY UYVY(4)

  2035.         "pand                %%mm7, %%mm0   \n\t" // U0V0 U0V0(0)

  2036.         "pand                %%mm7, %%mm1   \n\t" // U0V0 U0V0(4)

  2037.         "psrlw                  $8, %%mm2   \n\t" // Y0Y0 Y0Y0(0)

  2038.         "psrlw                  $8, %%mm3   \n\t" // Y0Y0 Y0Y0(4)

  2039.         "packuswb            %%mm1, %%mm0   \n\t" // UVUV UVUV(0)

  2040.         "packuswb            %%mm3, %%mm2   \n\t" // YYYY YYYY(0)

  2041. 

  2042.         MOVNTQ"              %%mm2,  (%1, %%eax, 2) \n\t"

  2043. 

  2044.         "movq     16(%0, %%eax, 4), %%mm1   \n\t" // UYVY UYVY(8)

  2045.         "movq     24(%0, %%eax, 4), %%mm2   \n\t" // UYVY UYVY(12)

  2046.         "movq                %%mm1, %%mm3   \n\t" // UYVY UYVY(8)

  2047.         "movq                %%mm2, %%mm4   \n\t" // UYVY UYVY(12)

  2048.         "pand                %%mm7, %%mm1   \n\t" // U0V0 U0V0(8)

  2049.         "pand                %%mm7, %%mm2   \n\t" // U0V0 U0V0(12)

  2050.         "psrlw                  $8, %%mm3   \n\t" // Y0Y0 Y0Y0(8)

  2051.         "psrlw                  $8, %%mm4   \n\t" // Y0Y0 Y0Y0(12)

  2052.         "packuswb            %%mm2, %%mm1   \n\t" // UVUV UVUV(8)

  2053.         "packuswb            %%mm4, %%mm3   \n\t" // YYYY YYYY(8)

  2054. 

  2055.         MOVNTQ"              %%mm3, 8(%1, %%eax, 2) \n\t"

  2056. 

  2057.         "movq                %%mm0, %%mm2   \n\t" // UVUV UVUV(0)

  2058.         "movq                %%mm1, %%mm3   \n\t" // UVUV UVUV(8)

  2059.         "psrlw                  $8, %%mm0   \n\t" // V0V0 V0V0(0)

  2060.         "psrlw                  $8, %%mm1   \n\t" // V0V0 V0V0(8)

  2061.         "pand                %%mm7, %%mm2   \n\t" // U0U0 U0U0(0)

  2062.         "pand                %%mm7, %%mm3   \n\t" // U0U0 U0U0(8)

  2063.         "packuswb            %%mm1, %%mm0   \n\t" // VVVV VVVV(0)

  2064.         "packuswb            %%mm3, %%mm2   \n\t" // UUUU UUUU(0)

  2065. 

  2066.         MOVNTQ"              %%mm0, (%3, %%eax) \n\t"

  2067.         MOVNTQ"              %%mm2, (%2, %%eax) \n\t"

  2068. 

  2069.         "addl                   $8, %%eax   \n\t"

  2070.         "cmpl                   %4, %%eax   \n\t"

  2071.         " jb                    1b          \n\t"

  2072.         ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)

  2073.         : "memory", "%eax"

  2074.         );

  2075. 

  2076.         ydst += lumStride;

  2077.         src  += srcStride;

  2078. 

  2079.         asm volatile(

  2080.         "xorl                %%eax, %%eax   \n\t"

  2081.         ASMALIGN(4)

  2082.         "1:                                 \n\t"

  2083.         PREFETCH" 64(%0, %%eax, 4)          \n\t"

  2084.         "movq       (%0, %%eax, 4), %%mm0   \n\t" // YUYV YUYV(0)

  2085.         "movq      8(%0, %%eax, 4), %%mm1   \n\t" // YUYV YUYV(4)

  2086.         "movq     16(%0, %%eax, 4), %%mm2   \n\t" // YUYV YUYV(8)

  2087.         "movq     24(%0, %%eax, 4), %%mm3   \n\t" // YUYV YUYV(12)

  2088.         "psrlw                  $8, %%mm0   \n\t" // Y0Y0 Y0Y0(0)

  2089.         "psrlw                  $8, %%mm1   \n\t" // Y0Y0 Y0Y0(4)

  2090.         "psrlw                  $8, %%mm2   \n\t" // Y0Y0 Y0Y0(8)

  2091.         "psrlw                  $8, %%mm3   \n\t" // Y0Y0 Y0Y0(12)

  2092.         "packuswb            %%mm1, %%mm0   \n\t" // YYYY YYYY(0)

  2093.         "packuswb            %%mm3, %%mm2   \n\t" // YYYY YYYY(8)

  2094. 

  2095.         MOVNTQ"              %%mm0,  (%1, %%eax, 2) \n\t"

  2096.         MOVNTQ"              %%mm2, 8(%1, %%eax, 2) \n\t"

  2097. 

  2098.         "addl                   $8, %%eax   \n\t"

  2099.         "cmpl                   %4, %%eax   \n\t"

  2100.         " jb                    1b          \n\t"

  2101. 

  2102.         ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth)

  2103.         : "memory", "%eax"

  2104.         );

  2105. #else

  2106.         long i;

  2107.         for (i=0; i<chromWidth; i++)

  2108.         {

  2109.             udst[i]     = src[4*i+0];

  2110.             ydst[2*i+0] = src[4*i+1];

  2111.             vdst[i]     = src[4*i+2];

  2112.             ydst[2*i+1] = src[4*i+3];

  2113.         }

  2114.         ydst += lumStride;

  2115.         src  += srcStride;

  2116. 

  2117.         for (i=0; i<chromWidth; i++)

  2118.         {

  2119.             ydst[2*i+0] = src[4*i+1];

  2120.             ydst[2*i+1] = src[4*i+3];

  2121.         }

  2122. #endif

  2123.         udst += chromStride;

  2124.         vdst += chromStride;

  2125.         ydst += lumStride;

  2126.         src  += srcStride;

  2127.     }

  2128. #ifdef HAVE_MMX

  2129. asm volatile(   EMMS"       \n\t"

  2130.                 SFENCE"     \n\t"

  2131.                 :::"memory");

  2132. #endif

  2133. }

  2134. 

  2135. /**

  2136.  *

  2137.  * height should be a multiple of 2 and width should be a multiple of 2 (if this is a

  2138.  * problem for anyone then tell me, and ill fix it)

  2139.  * chrominance data is only taken from every secound line others are ignored in the C version FIXME write HQ version

  2140.  */

  2141. static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,

  2142.                                        long width, long height,

  2143.                                        long lumStride, long chromStride, long srcStride)

  2144. {

  2145.     long y;

  2146.     const long chromWidth= width>>1;

  2147. #ifdef HAVE_MMX

  2148.     for (y=0; y<height-2; y+=2)

  2149.     {

  2150.         long i;

  2151.         for (i=0; i<2; i++)

  2152.         {

  2153.             asm volatile(

  2154.             "mov                        %2, %%"REG_a"   \n\t"

  2155.             "movq     "MANGLE(bgr2YCoeff)", %%mm6       \n\t"

  2156.             "movq          "MANGLE(w1111)", %%mm5       \n\t"

  2157.             "pxor                    %%mm7, %%mm7       \n\t"

  2158.             "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d"   \n\t"

  2159.             ASMALIGN(4)

  2160.             "1:                                         \n\t"

  2161.             PREFETCH"    64(%0, %%"REG_d")              \n\t"

  2162.             "movd          (%0, %%"REG_d"), %%mm0       \n\t"

  2163.             "movd         3(%0, %%"REG_d"), %%mm1       \n\t"

  2164.             "punpcklbw               %%mm7, %%mm0       \n\t"

  2165.             "punpcklbw               %%mm7, %%mm1       \n\t"

  2166.             "movd         6(%0, %%"REG_d"), %%mm2       \n\t"

  2167.             "movd         9(%0, %%"REG_d"), %%mm3       \n\t"

  2168.             "punpcklbw               %%mm7, %%mm2       \n\t"

  2169.             "punpcklbw               %%mm7, %%mm3       \n\t"

  2170.             "pmaddwd                 %%mm6, %%mm0       \n\t"

  2171.             "pmaddwd                 %%mm6, %%mm1       \n\t"

  2172.             "pmaddwd                 %%mm6, %%mm2       \n\t"

  2173.             "pmaddwd                 %%mm6, %%mm3       \n\t"

  2174. #ifndef FAST_BGR2YV12

  2175.             "psrad                      $8, %%mm0       \n\t"

  2176.             "psrad                      $8, %%mm1       \n\t"

  2177.             "psrad                      $8, %%mm2       \n\t"

  2178.             "psrad                      $8, %%mm3       \n\t"

  2179. #endif

  2180.             "packssdw                %%mm1, %%mm0       \n\t"

  2181.             "packssdw                %%mm3, %%mm2       \n\t"

  2182.             "pmaddwd                 %%mm5, %%mm0       \n\t"

  2183.             "pmaddwd                 %%mm5, %%mm2       \n\t"

  2184.             "packssdw                %%mm2, %%mm0       \n\t"

  2185.             "psraw                      $7, %%mm0       \n\t"

  2186. 

  2187.             "movd        12(%0, %%"REG_d"), %%mm4       \n\t"

  2188.             "movd        15(%0, %%"REG_d"), %%mm1       \n\t"

  2189.             "punpcklbw               %%mm7, %%mm4       \n\t"

  2190.             "punpcklbw               %%mm7, %%mm1       \n\t"

  2191.             "movd        18(%0, %%"REG_d"), %%mm2       \n\t"

  2192.             "movd        21(%0, %%"REG_d"), %%mm3       \n\t"

  2193.             "punpcklbw               %%mm7, %%mm2       \n\t"

  2194.             "punpcklbw               %%mm7, %%mm3       \n\t"

  2195.             "pmaddwd                 %%mm6, %%mm4       \n\t"

  2196.             "pmaddwd                 %%mm6, %%mm1       \n\t"

  2197.             "pmaddwd                 %%mm6, %%mm2       \n\t"

  2198.             "pmaddwd                 %%mm6, %%mm3       \n\t"

  2199. #ifndef FAST_BGR2YV12

  2200.             "psrad                      $8, %%mm4       \n\t"

  2201.             "psrad                      $8, %%mm1       \n\t"

  2202.             "psrad                      $8, %%mm2       \n\t"

  2203.             "psrad                      $8, %%mm3       \n\t"

  2204. #endif

  2205.             "packssdw                %%mm1, %%mm4       \n\t"

  2206.             "packssdw                %%mm3, %%mm2       \n\t"

  2207.             "pmaddwd                 %%mm5, %%mm4       \n\t"

  2208.             "pmaddwd                 %%mm5, %%mm2       \n\t"

  2209.             "add                       $24, %%"REG_d"   \n\t"

  2210.             "packssdw                %%mm2, %%mm4       \n\t"

  2211.             "psraw                      $7, %%mm4       \n\t"

  2212. 

  2213.             "packuswb                %%mm4, %%mm0       \n\t"

  2214.             "paddusb "MANGLE(bgr2YOffset)", %%mm0       \n\t"

  2215. 

  2216.             MOVNTQ"                  %%mm0, (%1, %%"REG_a") \n\t"

  2217.             "add                        $8,      %%"REG_a"  \n\t"

  2218.             " js                        1b                  \n\t"

  2219.             : : "r" (src+width*3), "r" (ydst+width), "g" (-width)

  2220.             : "%"REG_a, "%"REG_d

  2221.             );

  2222.             ydst += lumStride;

  2223.             src  += srcStride;

  2224.         }

  2225.         src -= srcStride*2;

  2226.         asm volatile(

  2227.         "mov                        %4, %%"REG_a"   \n\t"

  2228.         "movq          "MANGLE(w1111)", %%mm5       \n\t"

  2229.         "movq     "MANGLE(bgr2UCoeff)", %%mm6       \n\t"

  2230.         "pxor                    %%mm7, %%mm7       \n\t"

  2231.         "lea (%%"REG_a", %%"REG_a", 2), %%"REG_d"   \n\t"

  2232.         "add                 %%"REG_d", %%"REG_d"   \n\t"

  2233.         ASMALIGN(4)

  2234.         "1:                                         \n\t"

  2235.         PREFETCH"    64(%0, %%"REG_d")              \n\t"

  2236.         PREFETCH"    64(%1, %%"REG_d")              \n\t"

  2237. #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)

  2238.         "movq          (%0, %%"REG_d"), %%mm0       \n\t"

  2239.         "movq          (%1, %%"REG_d"), %%mm1       \n\t"

  2240.         "movq         6(%0, %%"REG_d"), %%mm2       \n\t"

  2241.         "movq         6(%1, %%"REG_d"), %%mm3       \n\t"

  2242.         PAVGB"                   %%mm1, %%mm0       \n\t"

  2243.         PAVGB"                   %%mm3, %%mm2       \n\t"

  2244.         "movq                    %%mm0, %%mm1       \n\t"

  2245.         "movq                    %%mm2, %%mm3       \n\t"

  2246.         "psrlq                     $24, %%mm0       \n\t"

  2247.         "psrlq                     $24, %%mm2       \n\t"

  2248.         PAVGB"                   %%mm1, %%mm0       \n\t"

  2249.         PAVGB"                   %%mm3, %%mm2       \n\t"

  2250.         "punpcklbw               %%mm7, %%mm0       \n\t"

  2251.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2252. #else

  2253.         "movd          (%0, %%"REG_d"), %%mm0       \n\t"

  2254.         "movd          (%1, %%"REG_d"), %%mm1       \n\t"

  2255.         "movd         3(%0, %%"REG_d"), %%mm2       \n\t"

  2256.         "movd         3(%1, %%"REG_d"), %%mm3       \n\t"

  2257.         "punpcklbw               %%mm7, %%mm0       \n\t"

  2258.         "punpcklbw               %%mm7, %%mm1       \n\t"

  2259.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2260.         "punpcklbw               %%mm7, %%mm3       \n\t"

  2261.         "paddw                   %%mm1, %%mm0       \n\t"

  2262.         "paddw                   %%mm3, %%mm2       \n\t"

  2263.         "paddw                   %%mm2, %%mm0       \n\t"

  2264.         "movd         6(%0, %%"REG_d"), %%mm4       \n\t"

  2265.         "movd         6(%1, %%"REG_d"), %%mm1       \n\t"

  2266.         "movd         9(%0, %%"REG_d"), %%mm2       \n\t"

  2267.         "movd         9(%1, %%"REG_d"), %%mm3       \n\t"

  2268.         "punpcklbw               %%mm7, %%mm4       \n\t"

  2269.         "punpcklbw               %%mm7, %%mm1       \n\t"

  2270.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2271.         "punpcklbw               %%mm7, %%mm3       \n\t"

  2272.         "paddw                   %%mm1, %%mm4       \n\t"

  2273.         "paddw                   %%mm3, %%mm2       \n\t"

  2274.         "paddw                   %%mm4, %%mm2       \n\t"

  2275.         "psrlw                      $2, %%mm0       \n\t"

  2276.         "psrlw                      $2, %%mm2       \n\t"

  2277. #endif

  2278.         "movq     "MANGLE(bgr2VCoeff)", %%mm1       \n\t"

  2279.         "movq     "MANGLE(bgr2VCoeff)", %%mm3       \n\t"

  2280. 

  2281.         "pmaddwd                 %%mm0, %%mm1       \n\t"

  2282.         "pmaddwd                 %%mm2, %%mm3       \n\t"

  2283.         "pmaddwd                 %%mm6, %%mm0       \n\t"

  2284.         "pmaddwd                 %%mm6, %%mm2       \n\t"

  2285. #ifndef FAST_BGR2YV12

  2286.         "psrad                      $8, %%mm0       \n\t"

  2287.         "psrad                      $8, %%mm1       \n\t"

  2288.         "psrad                      $8, %%mm2       \n\t"

  2289.         "psrad                      $8, %%mm3       \n\t"

  2290. #endif

  2291.         "packssdw                %%mm2, %%mm0       \n\t"

  2292.         "packssdw                %%mm3, %%mm1       \n\t"

  2293.         "pmaddwd                 %%mm5, %%mm0       \n\t"

  2294.         "pmaddwd                 %%mm5, %%mm1       \n\t"

  2295.         "packssdw                %%mm1, %%mm0       \n\t" // V1 V0 U1 U0

  2296.         "psraw                      $7, %%mm0       \n\t"

  2297. 

  2298. #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)

  2299.         "movq        12(%0, %%"REG_d"), %%mm4       \n\t"

  2300.         "movq        12(%1, %%"REG_d"), %%mm1       \n\t"

  2301.         "movq        18(%0, %%"REG_d"), %%mm2       \n\t"

  2302.         "movq        18(%1, %%"REG_d"), %%mm3       \n\t"

  2303.         PAVGB"                   %%mm1, %%mm4       \n\t"

  2304.         PAVGB"                   %%mm3, %%mm2       \n\t"

  2305.         "movq                    %%mm4, %%mm1       \n\t"

  2306.         "movq                    %%mm2, %%mm3       \n\t"

  2307.         "psrlq                     $24, %%mm4       \n\t"

  2308.         "psrlq                     $24, %%mm2       \n\t"

  2309.         PAVGB"                   %%mm1, %%mm4       \n\t"

  2310.         PAVGB"                   %%mm3, %%mm2       \n\t"

  2311.         "punpcklbw               %%mm7, %%mm4       \n\t"

  2312.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2313. #else

  2314.         "movd        12(%0, %%"REG_d"), %%mm4       \n\t"

  2315.         "movd        12(%1, %%"REG_d"), %%mm1       \n\t"

  2316.         "movd        15(%0, %%"REG_d"), %%mm2       \n\t"

  2317.         "movd        15(%1, %%"REG_d"), %%mm3       \n\t"

  2318.         "punpcklbw               %%mm7, %%mm4       \n\t"

  2319.         "punpcklbw               %%mm7, %%mm1       \n\t"

  2320.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2321.         "punpcklbw               %%mm7, %%mm3       \n\t"

  2322.         "paddw                   %%mm1, %%mm4       \n\t"

  2323.         "paddw                   %%mm3, %%mm2       \n\t"

  2324.         "paddw                   %%mm2, %%mm4       \n\t"

  2325.         "movd        18(%0, %%"REG_d"), %%mm5       \n\t"

  2326.         "movd        18(%1, %%"REG_d"), %%mm1       \n\t"

  2327.         "movd        21(%0, %%"REG_d"), %%mm2       \n\t"

  2328.         "movd        21(%1, %%"REG_d"), %%mm3       \n\t"

  2329.         "punpcklbw               %%mm7, %%mm5       \n\t"

  2330.         "punpcklbw               %%mm7, %%mm1       \n\t"

  2331.         "punpcklbw               %%mm7, %%mm2       \n\t"

  2332.         "punpcklbw               %%mm7, %%mm3       \n\t"

  2333.         "paddw                   %%mm1, %%mm5       \n\t"

  2334.         "paddw                   %%mm3, %%mm2       \n\t"

  2335.         "paddw                   %%mm5, %%mm2       \n\t"

  2336.         "movq          "MANGLE(w1111)", %%mm5       \n\t"

  2337.         "psrlw                      $2, %%mm4       \n\t"

  2338.         "psrlw                      $2, %%mm2       \n\t"

  2339. #endif

  2340.         "movq     "MANGLE(bgr2VCoeff)", %%mm1       \n\t"

  2341.         "movq     "MANGLE(bgr2VCoeff)", %%mm3       \n\t"

  2342. 

  2343.         "pmaddwd                 %%mm4, %%mm1       \n\t"

  2344.         "pmaddwd                 %%mm2, %%mm3       \n\t"

  2345.         "pmaddwd                 %%mm6, %%mm4       \n\t"

  2346.         "pmaddwd                 %%mm6, %%mm2       \n\t"

  2347. #ifndef FAST_BGR2YV12

  2348.         "psrad                      $8, %%mm4       \n\t"

  2349.         "psrad                      $8, %%mm1       \n\t"

  2350.         "psrad                      $8, %%mm2       \n\t"

  2351.         "psrad                      $8, %%mm3       \n\t"

  2352. #endif

  2353.         "packssdw                %%mm2, %%mm4       \n\t"

  2354.         "packssdw                %%mm3, %%mm1       \n\t"

  2355.         "pmaddwd                 %%mm5, %%mm4       \n\t"

  2356.         "pmaddwd                 %%mm5, %%mm1       \n\t"

  2357.         "add                       $24, %%"REG_d"   \n\t"

  2358.         "packssdw                %%mm1, %%mm4       \n\t" // V3 V2 U3 U2

  2359.         "psraw                      $7, %%mm4       \n\t"

  2360. 

  2361.         "movq                    %%mm0, %%mm1           \n\t"

  2362.         "punpckldq               %%mm4, %%mm0           \n\t"

  2363.         "punpckhdq               %%mm4, %%mm1           \n\t"

  2364.         "packsswb                %%mm1, %%mm0           \n\t"

  2365.         "paddb  "MANGLE(bgr2UVOffset)", %%mm0           \n\t"

  2366.         "movd                    %%mm0, (%2, %%"REG_a") \n\t"

  2367.         "punpckhdq               %%mm0, %%mm0           \n\t"

  2368.         "movd                    %%mm0, (%3, %%"REG_a") \n\t"

  2369.         "add                        $4, %%"REG_a"       \n\t"

  2370.         " js                        1b                  \n\t"

  2371.         : : "r" (src+chromWidth*6), "r" (src+srcStride+chromWidth*6), "r" (udst+chromWidth), "r" (vdst+chromWidth), "g" (-chromWidth)

  2372.         : "%"REG_a, "%"REG_d

  2373.         );

  2374. 

  2375.         udst += chromStride;

  2376.         vdst += chromStride;

  2377.         src  += srcStride*2;

  2378.     }

  2379. 

  2380.     asm volatile(   EMMS"       \n\t"

  2381.                     SFENCE"     \n\t"

  2382.                     :::"memory");

  2383. #else

  2384.     y=0;

  2385. #endif

  2386.     for (; y<height; y+=2)

  2387.     {

  2388.         long i;

  2389.         for (i=0; i<chromWidth; i++)

  2390.         {

  2391.             unsigned int b = src[6*i+0];

  2392.             unsigned int g = src[6*i+1];

  2393.             unsigned int r = src[6*i+2];

  2394. 

  2395.             unsigned int Y  =  ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;

  2396.             unsigned int V  =  ((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128;

  2397.             unsigned int U  =  ((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128;

  2398. 

  2399.             udst[i]     = U;

  2400.             vdst[i]     = V;

  2401.             ydst[2*i]   = Y;

  2402. 

  2403.             b = src[6*i+3];

  2404.             g = src[6*i+4];

  2405.             r = src[6*i+5];

  2406. 

  2407.             Y  =  ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;

  2408.             ydst[2*i+1]     = Y;

  2409.         }

  2410.         ydst += lumStride;

  2411.         src  += srcStride;

  2412. 

  2413.         for (i=0; i<chromWidth; i++)

  2414.         {

  2415.             unsigned int b = src[6*i+0];

  2416.             unsigned int g = src[6*i+1];

  2417.             unsigned int r = src[6*i+2];

  2418. 

  2419.             unsigned int Y  =  ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;

  2420. 

  2421.             ydst[2*i]     = Y;

  2422. 

  2423.             b = src[6*i+3];

  2424.             g = src[6*i+4];

  2425.             r = src[6*i+5];

  2426. 

  2427.             Y  =  ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;

  2428.             ydst[2*i+1]     = Y;

  2429.         }

  2430.         udst += chromStride;

  2431.         vdst += chromStride;

  2432.         ydst += lumStride;

  2433.         src  += srcStride;

  2434.     }

  2435. }

  2436. 

  2437. void RENAME(interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dest,

  2438.                              long width, long height, long src1Stride,

  2439.                              long src2Stride, long dstStride){

  2440.     long h;

  2441. 

  2442.     for (h=0; h < height; h++)

  2443.     {

  2444.         long w;

  2445. 

  2446. #ifdef HAVE_MMX

  2447. #ifdef HAVE_SSE2

  2448.         asm(

  2449.         "xor              %%"REG_a", %%"REG_a"  \n\t"

  2450.         "1:                                     \n\t"

  2451.         PREFETCH" 64(%1, %%"REG_a")             \n\t"

  2452.         PREFETCH" 64(%2, %%"REG_a")             \n\t"

  2453.         "movdqa     (%1, %%"REG_a"), %%xmm0     \n\t"

  2454.         "movdqa     (%1, %%"REG_a"), %%xmm1     \n\t"

  2455.         "movdqa     (%2, %%"REG_a"), %%xmm2     \n\t"

  2456.         "punpcklbw           %%xmm2, %%xmm0     \n\t"

  2457.         "punpckhbw           %%xmm2, %%xmm1     \n\t"

  2458.         "movntdq             %%xmm0,   (%0, %%"REG_a", 2)   \n\t"

  2459.         "movntdq             %%xmm1, 16(%0, %%"REG_a", 2)   \n\t"

  2460.         "add                    $16, %%"REG_a"  \n\t"

  2461.         "cmp                     %3, %%"REG_a"  \n\t"

  2462.         " jb                     1b             \n\t"

  2463.         ::"r"(dest), "r"(src1), "r"(src2), "r" (width-15)

  2464.         : "memory", "%"REG_a""

  2465.         );

  2466. #else

  2467.         asm(

  2468.         "xor %%"REG_a", %%"REG_a"               \n\t"

  2469.         "1:                                     \n\t"

  2470.         PREFETCH" 64(%1, %%"REG_a")             \n\t"

  2471.         PREFETCH" 64(%2, %%"REG_a")             \n\t"

  2472.         "movq       (%1, %%"REG_a"), %%mm0      \n\t"

  2473.         "movq      8(%1, %%"REG_a"), %%mm2      \n\t"

  2474.         "movq                 %%mm0, %%mm1      \n\t"

  2475.         "movq                 %%mm2, %%mm3      \n\t"

  2476.         "movq       (%2, %%"REG_a"), %%mm4      \n\t"

  2477.         "movq      8(%2, %%"REG_a"), %%mm5      \n\t"

  2478.         "punpcklbw            %%mm4, %%mm0      \n\t"

  2479.         "punpckhbw            %%mm4, %%mm1      \n\t"

  2480.         "punpcklbw            %%mm5, %%mm2      \n\t"

  2481.         "punpckhbw            %%mm5, %%mm3      \n\t"

  2482.         MOVNTQ"               %%mm0,   (%0, %%"REG_a", 2)   \n\t"

  2483.         MOVNTQ"               %%mm1,  8(%0, %%"REG_a", 2)   \n\t"

  2484.         MOVNTQ"               %%mm2, 16(%0, %%"REG_a", 2)   \n\t"

  2485.         MOVNTQ"               %%mm3, 24(%0, %%"REG_a", 2)   \n\t"

  2486.         "add                    $16, %%"REG_a"  \n\t"

  2487.         "cmp                     %3, %%"REG_a"  \n\t"

  2488.         " jb                     1b             \n\t"

  2489.         ::"r"(dest), "r"(src1), "r"(src2), "r" (width-15)

  2490.         : "memory", "%"REG_a

  2491.         );

  2492. #endif

  2493.         for (w= (width&(~15)); w < width; w++)

  2494.         {

  2495.             dest[2*w+0] = src1[w];

  2496.             dest[2*w+1] = src2[w];

  2497.         }

  2498. #else

  2499.         for (w=0; w < width; w++)

  2500.         {

  2501.             dest[2*w+0] = src1[w];

  2502.             dest[2*w+1] = src2[w];

  2503.         }

  2504. #endif

  2505.         dest += dstStride;

  2506.                 src1 += src1Stride;

  2507.                 src2 += src2Stride;

  2508.     }

  2509. #ifdef HAVE_MMX

  2510.     asm(

  2511.         EMMS"       \n\t"

  2512.         SFENCE"     \n\t"

  2513.         ::: "memory"

  2514.         );

  2515. #endif

  2516. }

  2517. 

  2518. static inline void RENAME(vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2,

  2519.                                        uint8_t *dst1, uint8_t *dst2,

  2520.                                        long width, long height,

  2521.                                        long srcStride1, long srcStride2,

  2522.                                        long dstStride1, long dstStride2)

  2523. {

  2524.     long y,x,w,h;

  2525.     w=width/2; h=height/2;

  2526. #ifdef HAVE_MMX

  2527.     asm volatile(

  2528.     PREFETCH" %0    \n\t"

  2529.     PREFETCH" %1    \n\t"

  2530.     ::"m"(*(src1+srcStride1)),"m"(*(src2+srcStride2)):"memory");

  2531. #endif

  2532.     for (y=0;y<h;y++){

  2533.     const uint8_t* s1=src1+srcStride1*(y>>1);

  2534.     uint8_t* d=dst1+dstStride1*y;

  2535.     x=0;

  2536. #ifdef HAVE_MMX

  2537.     for (;x<w-31;x+=32)

  2538.     {

  2539.         asm volatile(

  2540.         PREFETCH"   32%1        \n\t"

  2541.         "movq         %1, %%mm0 \n\t"

  2542.         "movq        8%1, %%mm2 \n\t"

  2543.         "movq       16%1, %%mm4 \n\t"

  2544.         "movq       24%1, %%mm6 \n\t"

  2545.         "movq      %%mm0, %%mm1 \n\t"

  2546.         "movq      %%mm2, %%mm3 \n\t"

  2547.         "movq      %%mm4, %%mm5 \n\t"

  2548.         "movq      %%mm6, %%mm7 \n\t"

  2549.         "punpcklbw %%mm0, %%mm0 \n\t"

  2550.         "punpckhbw %%mm1, %%mm1 \n\t"

  2551.         "punpcklbw %%mm2, %%mm2 \n\t"

  2552.         "punpckhbw %%mm3, %%mm3 \n\t"

  2553.         "punpcklbw %%mm4, %%mm4 \n\t"

  2554.         "punpckhbw %%mm5, %%mm5 \n\t"

  2555.         "punpcklbw %%mm6, %%mm6 \n\t"

  2556.         "punpckhbw %%mm7, %%mm7 \n\t"

  2557.         MOVNTQ"    %%mm0,   %0  \n\t"

  2558.         MOVNTQ"    %%mm1,  8%0  \n\t"

  2559.         MOVNTQ"    %%mm2, 16%0  \n\t"

  2560.         MOVNTQ"    %%mm3, 24%0  \n\t"

  2561.         MOVNTQ"    %%mm4, 32%0  \n\t"

  2562.         MOVNTQ"    %%mm5, 40%0  \n\t"

  2563.         MOVNTQ"    %%mm6, 48%0  \n\t"

  2564.         MOVNTQ"    %%mm7, 56%0"

  2565.         :"=m"(d[2*x])

  2566.         :"m"(s1[x])

  2567.         :"memory");

  2568.     }

  2569. #endif

  2570.     for (;x<w;x++) d[2*x]=d[2*x+1]=s1[x];

  2571.     }

  2572.     for (y=0;y<h;y++){

  2573.     const uint8_t* s2=src2+srcStride2*(y>>1);

  2574.     uint8_t* d=dst2+dstStride2*y;

  2575.     x=0;

  2576. #ifdef HAVE_MMX

  2577.     for (;x<w-31;x+=32)

  2578.     {

  2579.         asm volatile(

  2580.         PREFETCH"   32%1        \n\t"

  2581.         "movq         %1, %%mm0 \n\t"

  2582.         "movq        8%1, %%mm2 \n\t"

  2583.         "movq       16%1, %%mm4 \n\t"

  2584.         "movq       24%1, %%mm6 \n\t"

  2585.         "movq      %%mm0, %%mm1 \n\t"

  2586.         "movq      %%mm2, %%mm3 \n\t"

  2587.         "movq      %%mm4, %%mm5 \n\t"

  2588.         "movq      %%mm6, %%mm7 \n\t"

  2589.         "punpcklbw %%mm0, %%mm0 \n\t"

  2590.         "punpckhbw %%mm1, %%mm1 \n\t"

  2591.         "punpcklbw %%mm2, %%mm2 \n\t"

  2592.         "punpckhbw %%mm3, %%mm3 \n\t"

  2593.         "punpcklbw %%mm4, %%mm4 \n\t"

  2594.         "punpckhbw %%mm5, %%mm5 \n\t"

  2595.         "punpcklbw %%mm6, %%mm6 \n\t"

  2596.         "punpckhbw %%mm7, %%mm7 \n\t"

  2597.         MOVNTQ"    %%mm0,   %0  \n\t"

  2598.         MOVNTQ"    %%mm1,  8%0  \n\t"

  2599.         MOVNTQ"    %%mm2, 16%0  \n\t"

  2600.         MOVNTQ"    %%mm3, 24%0  \n\t"

  2601.         MOVNTQ"    %%mm4, 32%0  \n\t"

  2602.         MOVNTQ"    %%mm5, 40%0  \n\t"

  2603.         MOVNTQ"    %%mm6, 48%0  \n\t"

  2604.         MOVNTQ"    %%mm7, 56%0"

  2605.         :"=m"(d[2*x])

  2606.         :"m"(s2[x])

  2607.         :"memory");

  2608.     }

  2609. #endif

  2610.     for (;x<w;x++) d[2*x]=d[2*x+1]=s2[x];

  2611.     }

  2612. #ifdef HAVE_MMX

  2613.     asm(

  2614.         EMMS"       \n\t"

  2615.         SFENCE"     \n\t"

  2616.         ::: "memory"

  2617.         );

  2618. #endif

  2619. }

  2620. 

  2621. static inline void RENAME(yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,

  2622.                                         uint8_t *dst,

  2623.                                         long width, long height,

  2624.                                         long srcStride1, long srcStride2,

  2625.                                         long srcStride3, long dstStride)

  2626. {

  2627.     long y,x,w,h;

  2628.     w=width/2; h=height;

  2629.     for (y=0;y<h;y++){

  2630.     const uint8_t* yp=src1+srcStride1*y;

  2631.     const uint8_t* up=src2+srcStride2*(y>>2);

  2632.     const uint8_t* vp=src3+srcStride3*(y>>2);

  2633.     uint8_t* d=dst+dstStride*y;

  2634.     x=0;

  2635. #ifdef HAVE_MMX

  2636.     for (;x<w-7;x+=8)

  2637.     {

  2638.         asm volatile(

  2639.         PREFETCH"   32(%1, %0)          \n\t"

  2640.         PREFETCH"   32(%2, %0)          \n\t"

  2641.         PREFETCH"   32(%3, %0)          \n\t"

  2642.         "movq      (%1, %0, 4), %%mm0   \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */

  2643.         "movq         (%2, %0), %%mm1   \n\t" /* U0U1U2U3U4U5U6U7 */

  2644.         "movq         (%3, %0), %%mm2   \n\t" /* V0V1V2V3V4V5V6V7 */

  2645.         "movq            %%mm0, %%mm3   \n\t" /* Y0Y1Y2Y3Y4Y5Y6Y7 */

  2646.         "movq            %%mm1, %%mm4   \n\t" /* U0U1U2U3U4U5U6U7 */

  2647.         "movq            %%mm2, %%mm5   \n\t" /* V0V1V2V3V4V5V6V7 */

  2648.         "punpcklbw       %%mm1, %%mm1   \n\t" /* U0U0 U1U1 U2U2 U3U3 */

  2649.         "punpcklbw       %%mm2, %%mm2   \n\t" /* V0V0 V1V1 V2V2 V3V3 */

  2650.         "punpckhbw       %%mm4, %%mm4   \n\t" /* U4U4 U5U5 U6U6 U7U7 */

  2651.         "punpckhbw       %%mm5, %%mm5   \n\t" /* V4V4 V5V5 V6V6 V7V7 */

  2652. 

  2653.         "movq            %%mm1, %%mm6   \n\t"

  2654.         "punpcklbw       %%mm2, %%mm1   \n\t" /* U0V0 U0V0 U1V1 U1V1*/

  2655.         "punpcklbw       %%mm1, %%mm0   \n\t" /* Y0U0 Y1V0 Y2U0 Y3V0*/

  2656.         "punpckhbw       %%mm1, %%mm3   \n\t" /* Y4U1 Y5V1 Y6U1 Y7V1*/

  2657.         MOVNTQ"          %%mm0,  (%4, %0, 8)    \n\t"

  2658.         MOVNTQ"          %%mm3, 8(%4, %0, 8)    \n\t"

  2659. 

  2660.         "punpckhbw       %%mm2, %%mm6   \n\t" /* U2V2 U2V2 U3V3 U3V3*/

  2661.         "movq     8(%1, %0, 4), %%mm0   \n\t"

  2662.         "movq            %%mm0, %%mm3   \n\t"

  2663.         "punpcklbw       %%mm6, %%mm0   \n\t" /* Y U2 Y V2 Y U2 Y V2*/

  2664.         "punpckhbw       %%mm6, %%mm3   \n\t" /* Y U3 Y V3 Y U3 Y V3*/

  2665.         MOVNTQ"          %%mm0, 16(%4, %0, 8)   \n\t"

  2666.         MOVNTQ"          %%mm3, 24(%4, %0, 8)   \n\t"

  2667. 

  2668.         "movq            %%mm4, %%mm6   \n\t"

  2669.         "movq    16(%1, %0, 4), %%mm0   \n\t"

  2670.         "movq            %%mm0, %%mm3   \n\t"

  2671.         "punpcklbw       %%mm5, %%mm4   \n\t"

  2672.         "punpcklbw       %%mm4, %%mm0   \n\t" /* Y U4 Y V4 Y U4 Y V4*/

  2673.         "punpckhbw       %%mm4, %%mm3   \n\t" /* Y U5 Y V5 Y U5 Y V5*/

  2674.         MOVNTQ"          %%mm0, 32(%4, %0, 8)   \n\t"

  2675.         MOVNTQ"          %%mm3, 40(%4, %0, 8)   \n\t"

  2676. 

  2677.         "punpckhbw       %%mm5, %%mm6   \n\t"

  2678.         "movq    24(%1, %0, 4), %%mm0   \n\t"

  2679.         "movq            %%mm0, %%mm3   \n\t"

  2680.         "punpcklbw       %%mm6, %%mm0   \n\t" /* Y U6 Y V6 Y U6 Y V6*/

  2681.         "punpckhbw       %%mm6, %%mm3   \n\t" /* Y U7 Y V7 Y U7 Y V7*/

  2682.         MOVNTQ"          %%mm0, 48(%4, %0, 8)   \n\t"

  2683.         MOVNTQ"          %%mm3, 56(%4, %0, 8)   \n\t"

  2684. 

  2685.         : "+r" (x)

  2686.         : "r"(yp), "r" (up), "r"(vp), "r"(d)

  2687.         :"memory");

  2688.     }

  2689. #endif

  2690.     for (; x<w; x++)

  2691.     {

  2692.         const long x2 = x<<2;

  2693.         d[8*x+0] = yp[x2];

  2694.         d[8*x+1] = up[x];

  2695.         d[8*x+2] = yp[x2+1];

  2696.         d[8*x+3] = vp[x];

  2697.         d[8*x+4] = yp[x2+2];

  2698.         d[8*x+5] = up[x];

  2699.         d[8*x+6] = yp[x2+3];

  2700.         d[8*x+7] = vp[x];

  2701.     }

  2702.     }

  2703. #ifdef HAVE_MMX

  2704.     asm(

  2705.         EMMS"       \n\t"

  2706.         SFENCE"     \n\t"

  2707.         ::: "memory"

  2708.         );

  2709. #endif

  2710. }

  2711. 

  2712. static inline void RENAME(rgb2rgb_init)(void){

  2713.     rgb15to16       = RENAME(rgb15to16);

  2714.     rgb15to24       = RENAME(rgb15to24);

  2715.     rgb15to32       = RENAME(rgb15to32);

  2716.     rgb16to24       = RENAME(rgb16to24);

  2717.     rgb16to32       = RENAME(rgb16to32);

  2718.     rgb16to15       = RENAME(rgb16to15);

  2719.     rgb24to16       = RENAME(rgb24to16);

  2720.     rgb24to15       = RENAME(rgb24to15);

  2721.     rgb24to32       = RENAME(rgb24to32);

  2722.     rgb32to16       = RENAME(rgb32to16);

  2723.     rgb32to15       = RENAME(rgb32to15);

  2724.     rgb32to24       = RENAME(rgb32to24);

  2725.     rgb24tobgr15    = RENAME(rgb24tobgr15);

  2726.     rgb24tobgr16    = RENAME(rgb24tobgr16);

  2727.     rgb24tobgr24    = RENAME(rgb24tobgr24);

  2728.     rgb32tobgr32    = RENAME(rgb32tobgr32);

  2729.     rgb32tobgr16    = RENAME(rgb32tobgr16);

  2730.     rgb32tobgr15    = RENAME(rgb32tobgr15);

  2731.     yv12toyuy2      = RENAME(yv12toyuy2);

  2732.     yv12touyvy      = RENAME(yv12touyvy);

  2733.     yuv422ptoyuy2   = RENAME(yuv422ptoyuy2);

  2734.     yuy2toyv12      = RENAME(yuy2toyv12);

  2735. //    uyvytoyv12      = RENAME(uyvytoyv12);

  2736. //    yvu9toyv12      = RENAME(yvu9toyv12);

  2737.     planar2x        = RENAME(planar2x);

  2738.     rgb24toyv12     = RENAME(rgb24toyv12);

  2739.     interleaveBytes = RENAME(interleaveBytes);

  2740.     vu9_to_vu12     = RENAME(vu9_to_vu12);

  2741.     yvu9_to_yuy2    = RENAME(yvu9_to_yuy2);

  2742. }