git.sesse.net Git - vlc/blob - plugins/chroma/i420_yuy2.h

   1 /*****************************************************************************
   2  * i420_yuy2.h : YUV to YUV conversion module for vlc
   3  *****************************************************************************
   4  * Copyright (C) 2000, 2001 VideoLAN
   5  * $Id: i420_yuy2.h,v 1.7 2002/06/01 11:08:24 sam Exp $
   6  *
   7  * Authors: Samuel Hocevar <sam@zoy.org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
  22  *****************************************************************************/
  23
  24 #ifdef MODULE_NAME_IS_chroma_i420_yuy2_mmx
  25
  26 #define UNUSED_LONGLONG(foo) \
  27     unsigned long long foo __asm__ (#foo) __attribute__((unused))
  28 UNUSED_LONGLONG(woo_00ffw) = 0x00ff00ff00ff00ff;
  29 UNUSED_LONGLONG(woo_80w)   = 0x0000000080808080;
  30
  31 #define MMX_LOAD "                                                        \n\
  32 movl         %9,%%ebx                                                     \n\
  33 "
  34
  35 #define MMX_SAVE "                                                        \n\
  36 movl      %%ebx,%9                                                        \n\
  37 "
  38
  39 #define MMX_INC "                                                         \n\
  40 addl         $16, %0                                                      \n\
  41 addl         $16, %1                                                      \n\
  42 addl         $8, %2                                                       \n\
  43 addl         $8, %3                                                       \n\
  44 addl         $4, %%eax                                                    \n\
  45 addl         $4, %%ebx                                                    \n\
  46 "
  47
  48 #define MMX_CALL(MMX_INSTRUCTIONS)                                  \
  49     __asm__ __volatile__(                                           \
  50         MMX_LOAD                                                    \
  51         ".align 8 \n\t"                                             \
  52         MMX_INSTRUCTIONS                                            \
  53         MMX_INC                                                     \
  54         MMX_SAVE                                                    \
  55         : "=c" (p_line1), "=d" (p_line2), "=D" (p_y1), "=S" (p_y2)  \
  56         :  "c" (p_line1),  "d" (p_line2),  "D" (p_y1),  "S" (p_y2), \
  57           "a" (p_u), "m" (p_v)                                      \
  58         : "ebx", "memory" );
  59
  60 #define MMX_YUV420_YUYV "                                                 \n\
  61 movq       (%2), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  62 movd    (%%eax), %%mm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
  63 movd    (%%ebx), %%mm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
  64 punpcklbw %%mm2, %%mm1  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
  65 movq      %%mm0, %%mm2  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
  66 punpcklbw %%mm1, %%mm2  #                     v1 y3 u1 y2 v0 y1 u0 y0     \n\
  67 movq      %%mm2, (%0)   # Store low YUYV                                  \n\
  68 punpckhbw %%mm1, %%mm0  #                     v3 y7 u3 y6 v2 y5 u2 y4     \n\
  69 movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
  70 movq       (%3), %%mm0  # Load 8 Y            Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
  71 movq      %%mm0, %%mm2  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
  72 punpcklbw %%mm1, %%mm2  #                     v1 Y3 u1 Y2 v0 Y1 u0 Y0     \n\
  73 movq      %%mm2, (%1)   # Store low YUYV                                  \n\
  74 punpckhbw %%mm1, %%mm0  #                     v3 Y7 u3 Y6 v2 Y5 u2 Y4     \n\
  75 movq      %%mm0, 8(%1)  # Store high YUYV                                 \n\
  76 "
  77
  78 #define MMX_YUV420_YVYU "                                                 \n\
  79 movq       (%2), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  80 movd    (%%eax), %%mm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
  81 movd    (%%ebx), %%mm1  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
  82 punpcklbw %%mm2, %%mm1  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
  83 movq      %%mm0, %%mm2  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
  84 punpcklbw %%mm1, %%mm2  #                     u1 y3 v1 y2 u0 y1 v0 y0     \n\
  85 movq      %%mm2, (%0)   # Store low YUYV                                  \n\
  86 punpckhbw %%mm1, %%mm0  #                     u3 y7 v3 y6 u2 y5 v2 y4     \n\
  87 movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
  88 movq       (%3), %%mm0  # Load 8 Y            Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
  89 movq      %%mm0, %%mm2  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
  90 punpcklbw %%mm1, %%mm2  #                     u1 Y3 v1 Y2 u0 Y1 v0 Y0     \n\
  91 movq      %%mm2, (%1)   # Store low YUYV                                  \n\
  92 punpckhbw %%mm1, %%mm0  #                     u3 Y7 v3 Y6 u2 Y5 v2 Y4     \n\
  93 movq      %%mm0, 8(%1)  # Store high YUYV                                 \n\
  94 "
  95
  96 #define MMX_YUV420_UYVY "                                                 \n\
  97 movq       (%2), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  98 movq       (%3), %%mm3  # Load 8 Y            Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
  99 movd    (%%eax), %%mm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
 100 movd    (%%ebx), %%mm1  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
 101 punpcklbw %%mm2, %%mm1  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
 102 movq      %%mm1, %%mm2  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
 103 punpcklbw %%mm0, %%mm2  #                     y3 v1 y2 u1 y1 v0 y0 u0     \n\
 104 movq      %%mm2, (%0)   # Store low UYVY                                  \n\
 105 movq      %%mm1, %%mm2  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
 106 punpckhbw %%mm0, %%mm2  #                     y3 v1 y2 u1 y1 v0 y0 u0     \n\
 107 movq      %%mm2, 8(%0)  # Store high UYVY                                 \n\
 108 movq      %%mm1, %%mm2  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
 109 punpcklbw %%mm3, %%mm2  #                     Y3 v1 Y2 u1 Y1 v0 Y0 u0     \n\
 110 movq      %%mm2, (%1)   # Store low UYVY                                  \n\
 111 punpckhbw %%mm3, %%mm1  #                     Y7 v3 Y6 u3 Y5 v2 Y4 u2     \n\
 112 movq      %%mm1, 8(%1)  # Store high UYVY                                 \n\
 113 "
 114
 115 /* FIXME: this code does not work ! Chroma seems to be wrong. */
 116 #define MMX_YUV420_Y211 "                                                 \n\
 117 movq       (%2), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
 118 movq       (%3), %%mm1  # Load 8 Y            Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
 119 movd    (%%eax), %%mm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
 120 movd    (%%ebx), %%mm3  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
 121 pand  woo_00ffw, %%mm0  # get Y even          00 Y6 00 Y4 00 Y2 00 Y0     \n\
 122 packuswb  %%mm0, %%mm0  # pack Y              y6 y4 y2 y0 y6 y4 y2 y0     \n\
 123 pand  woo_00ffw, %%mm2  # get U even          00 u6 00 u4 00 u2 00 u0     \n\
 124 packuswb  %%mm2, %%mm2  # pack U              00 00 u2 u0 00 00 u2 u0     \n\
 125 pand  woo_00ffw, %%mm3  # get V even          00 v6 00 v4 00 v2 00 v0     \n\
 126 packuswb  %%mm3, %%mm3  # pack V              00 00 v2 v0 00 00 v2 v0     \n\
 127 punpcklbw %%mm3, %%mm2  #                     00 00 00 00 v2 u2 v0 u0     \n\
 128 psubsw  woo_80w, %%mm2  # U,V -= 128                                      \n\
 129 punpcklbw %%mm2, %%mm0  #                     v2 y6 u2 y4 v0 y2 u0 y0     \n\
 130 movq      %%mm0, (%0)   # Store YUYV                                      \n\
 131 pand  woo_00ffw, %%mm1  # get Y even          00 Y6 00 Y4 00 Y2 00 Y0     \n\
 132 packuswb  %%mm1, %%mm1  # pack Y              Y6 Y4 Y2 Y0 Y6 Y4 Y2 Y0     \n\
 133 punpcklbw %%mm2, %%mm1  #                     v2 Y6 u2 Y4 v0 Y2 u0 Y0     \n\
 134 movq      %%mm1, (%1)   # Store YUYV                                      \n\
 135 "
 136
 137 #else
 138
 139 #define C_YUV420_YUYV( )                                                    \
 140     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 141     *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \
 142     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 143     *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \
 144
 145 #define C_YUV420_YVYU( )                                                    \
 146     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 147     *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \
 148     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 149     *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \
 150
 151 #define C_YUV420_UYVY( )                                                    \
 152     *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \
 153     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 154     *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \
 155     *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
 156
 157 #define C_YUV420_Y211( )                                                    \
 158     *(p_line1)++ = *(p_y1); ((u16*)p_y1)++;                                 \
 159     *(p_line2)++ = *(p_y2); ((u16*)p_y2)++;                                 \
 160     *(p_line1)++ = *(p_line2)++ = *(p_u) - 0x80; ((u16*)p_u)++;             \
 161     *(p_line1)++ = *(p_y1); ((u16*)p_y1)++;                                 \
 162     *(p_line2)++ = *(p_y2); ((u16*)p_y2)++;                                 \
 163     *(p_line1)++ = *(p_line2)++ = *(p_v) - 0x80; ((u16*)p_v)++;             \
 164
 165 #endif
 166