modules/video_chroma/i422_yuy2.h

   1 /*****************************************************************************
   2  * i422_yuy2.h : YUV to YUV conversion module for vlc
   3  *****************************************************************************
   4  * Copyright (C) 2002 VLC authors and VideoLAN
   5  * $Id$
   6  *
   7  * Authors: Samuel Hocevar <sam@zoy.org>
   8  *          Damien Fouilleul <damienf@videolan.org>
   9  *
  10  * This program is free software; you can redistribute it and/or modify it
  11  * under the terms of the GNU Lesser General Public License as published by
  12  * the Free Software Foundation; either version 2.1 of the License, or
  13  * (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  18  * GNU Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public License
  21  * along with this program; if not, write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  23  *****************************************************************************/
  24
  25 #ifdef MODULE_NAME_IS_i422_yuy2_mmx
  26
  27 #if defined(CAN_COMPILE_MMX)
  28
  29 /* MMX assembly */
  30
  31 #define MMX_CALL(MMX_INSTRUCTIONS)          \
  32     do {                                    \
  33     __asm__ __volatile__(                   \
  34         ".p2align 3 \n\t"                   \
  35         MMX_INSTRUCTIONS                    \
  36         :                                   \
  37         : "r" (p_line), "r" (p_y),          \
  38           "r" (p_u), "r" (p_v)              \
  39         : "mm0", "mm1", "mm2" );            \
  40         p_line += 16; p_y += 8;             \
  41         p_u += 4; p_v += 4;                 \
  42     } while(0)
  43
  44 #define MMX_END __asm__ __volatile__ ( "emms" )
  45
  46 #define MMX_YUV422_YUYV "                                                 \n\
  47 movq       (%1), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  48 movd       (%2), %%mm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
  49 movd       (%3), %%mm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
  50 punpcklbw %%mm2, %%mm1  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
  51 movq      %%mm0, %%mm2  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
  52 punpcklbw %%mm1, %%mm2  #                     v1 y3 u1 y2 v0 y1 u0 y0     \n\
  53 movq      %%mm2, (%0)   # Store low YUYV                                  \n\
  54 punpckhbw %%mm1, %%mm0  #                     v3 y7 u3 y6 v2 y5 u2 y4     \n\
  55 movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
  56 "
  57
  58 #define MMX_YUV422_YVYU "                                                 \n\
  59 movq       (%1), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  60 movd       (%2), %%mm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
  61 movd       (%3), %%mm1  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
  62 punpcklbw %%mm2, %%mm1  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
  63 movq      %%mm0, %%mm2  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
  64 punpcklbw %%mm1, %%mm2  #                     u1 y3 v1 y2 u0 y1 v0 y0     \n\
  65 movq      %%mm2, (%0)   # Store low YUYV                                  \n\
  66 punpckhbw %%mm1, %%mm0  #                     u3 y7 v3 y6 u2 y5 v2 y4     \n\
  67 movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
  68 "
  69
  70 #define MMX_YUV422_UYVY "                                                 \n\
  71 movq       (%1), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
  72 movd       (%2), %%mm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
  73 movd       (%3), %%mm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
  74 punpcklbw %%mm2, %%mm1  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
  75 movq      %%mm1, %%mm2  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
  76 punpcklbw %%mm0, %%mm2  #                     y3 v1 y2 u1 y1 v0 y0 u0     \n\
  77 movq      %%mm2, (%0)   # Store low UYVY                                  \n\
  78 punpckhbw %%mm0, %%mm1  #                     y7 v3 y6 u3 y5 v2 y4 u2     \n\
  79 movq      %%mm1, 8(%0)  # Store high UYVY                                 \n\
  80 "
  81
  82 #elif defined(HAVE_MMX_INTRINSICS)
  83
  84 /* MMX intrinsics */
  85
  86 #include <mmintrin.h>
  87
  88 #define MMX_CALL(MMX_INSTRUCTIONS)  \
  89     do {                            \
  90         __m64 mm0, mm1, mm2;        \
  91         MMX_INSTRUCTIONS            \
  92         p_line += 16; p_y += 8;     \
  93         p_u += 4; p_v += 4;         \
  94     } while(0)
  95
  96 #define MMX_END _mm_empty()
  97
  98 #define MMX_YUV422_YUYV                     \
  99     mm0 = (__m64)*(uint64_t*)p_y;           \
 100     mm1 = _mm_cvtsi32_si64(*(int*)p_u);     \
 101     mm2 = _mm_cvtsi32_si64(*(int*)p_v);     \
 102     mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
 103     mm2 = mm0;                              \
 104     mm2 = _mm_unpacklo_pi8(mm2, mm1);       \
 105     *(uint64_t*)p_line = (uint64_t)mm2;     \
 106     mm0 = _mm_unpackhi_pi8(mm0, mm1);       \
 107     *(uint64_t*)(p_line+8) = (uint64_t)mm0;
 108
 109 #define MMX_YUV422_YVYU                     \
 110     mm0 = (__m64)*(uint64_t*)p_y;           \
 111     mm2 = _mm_cvtsi32_si64(*(int*)p_u);     \
 112     mm1 = _mm_cvtsi32_si64(*(int*)p_v);     \
 113     mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
 114     mm2 = mm0;                              \
 115     mm2 = _mm_unpacklo_pi8(mm2, mm1);       \
 116     *(uint64_t*)p_line = (uint64_t)mm2;     \
 117     mm0 = _mm_unpackhi_pi8(mm0, mm1);       \
 118     *(uint64_t*)(p_line+8) = (uint64_t)mm0;
 119
 120 #define MMX_YUV422_UYVY                     \
 121     mm0 = (__m64)*(uint64_t*)p_y;           \
 122     mm1 = _mm_cvtsi32_si64(*(int*)p_u);     \
 123     mm2 = _mm_cvtsi32_si64(*(int*)p_v);     \
 124     mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
 125     mm2 = mm1;                              \
 126     mm2 = _mm_unpacklo_pi8(mm2, mm0);       \
 127     *(uint64_t*)p_line = (uint64_t)mm2;     \
 128     mm1 = _mm_unpackhi_pi8(mm1, mm0);       \
 129     *(uint64_t*)(p_line+8) = (uint64_t)mm1;
 130
 131 #endif
 132
 133 #elif defined( MODULE_NAME_IS_i422_yuy2_sse2 )
 134
 135 #if defined(CAN_COMPILE_SSE2)
 136
 137 /* SSE2 assembly */
 138
 139 #define SSE2_CALL(MMX_INSTRUCTIONS)         \
 140     do {                                    \
 141     __asm__ __volatile__(                   \
 142         ".p2align 3 \n\t"                   \
 143         MMX_INSTRUCTIONS                    \
 144         :                                   \
 145         : "r" (p_line), "r" (p_y),          \
 146           "r" (p_u), "r" (p_v)              \
 147         : "xmm0", "xmm1", "xmm2" );         \
 148         p_line += 32; p_y += 16;            \
 149         p_u += 8; p_v += 8;                 \
 150     } while(0)
 151
 152 #define SSE2_END  __asm__ __volatile__ ( "sfence" ::: "memory" )
 153
 154 #define SSE2_YUV422_YUYV_ALIGNED "                                        \n\
 155 movdqa      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 156 movq        (%2), %%xmm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 157 movq        (%3), %%xmm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 158 punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 159 movdqa    %%xmm0, %%xmm2  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
 160 punpcklbw %%xmm1, %%xmm2  #                     v1 y3 u1 y2 v0 y1 u0 y0   \n\
 161 movntdq   %%xmm2, (%0)    # Store low YUYV                                \n\
 162 punpckhbw %%xmm1, %%xmm0  #                     v3 y7 u3 y6 v2 y5 u2 y4   \n\
 163 movntdq   %%xmm0, 16(%0)  # Store high YUYV                               \n\
 164 "
 165
 166 #define SSE2_YUV422_YUYV_UNALIGNED "                                      \n\
 167 movdqu      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 168 movq        (%2), %%xmm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 169 movq        (%3), %%xmm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 170 prefetchnta (%0)          # Tell CPU not to cache output YUYV data        \n\
 171 punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 172 movdqa    %%xmm0, %%xmm2  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
 173 punpcklbw %%xmm1, %%xmm2  #                     v1 y3 u1 y2 v0 y1 u0 y0   \n\
 174 movdqu    %%xmm2, (%0)    # Store low YUYV                                \n\
 175 punpckhbw %%xmm1, %%xmm0  #                     v3 y7 u3 y6 v2 y5 u2 y4   \n\
 176 movdqu    %%xmm0, 16(%0)  # Store high YUYV                               \n\
 177 "
 178
 179 #define SSE2_YUV422_YVYU_ALIGNED "                                        \n\
 180 movdqa      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 181 movq        (%2), %%xmm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 182 movq        (%3), %%xmm1  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 183 punpcklbw %%xmm2, %%xmm1  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
 184 movdqa    %%xmm0, %%xmm2  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
 185 punpcklbw %%xmm1, %%xmm2  #                     u1 y3 v1 y2 u0 y1 v0 y0   \n\
 186 movntdq   %%xmm2, (%0)    # Store low YUYV                                \n\
 187 punpckhbw %%xmm1, %%xmm0  #                     u3 y7 v3 y6 u2 y5 v2 y4   \n\
 188 movntdq   %%xmm0, 16(%0)  # Store high YUYV                               \n\
 189 "
 190
 191 #define SSE2_YUV422_YVYU_UNALIGNED "                                      \n\
 192 movdqu      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 193 movq        (%2), %%xmm2  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 194 movq        (%3), %%xmm1  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 195 prefetchnta (%0)          # Tell CPU not to cache output YUYV data        \n\
 196 punpcklbw %%xmm2, %%xmm1  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
 197 movdqa    %%xmm0, %%xmm2  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
 198 punpcklbw %%xmm1, %%xmm2  #                     u1 y3 v1 y2 u0 y1 v0 y0   \n\
 199 movdqu    %%xmm2, (%0)    # Store low YUYV                                \n\
 200 punpckhbw %%xmm1, %%xmm0  #                     u3 y7 v3 y6 u2 y5 v2 y4   \n\
 201 movdqu    %%xmm0, 16(%0)  # Store high YUYV                               \n\
 202 "
 203
 204 #define SSE2_YUV422_UYVY_ALIGNED "                                        \n\
 205 movdqa      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 206 movq        (%2), %%xmm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 207 movq        (%3), %%xmm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 208 punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 209 movdqa    %%xmm1, %%xmm2  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 210 punpcklbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
 211 movntdq   %%xmm2, (%0)    # Store low UYVY                                \n\
 212 punpckhbw %%xmm0, %%xmm1  #                     y7 v3 y6 u3 y5 v2 y4 u2   \n\
 213 movntdq   %%xmm1, 16(%0)  # Store high UYVY                               \n\
 214 "
 215
 216 #define SSE2_YUV422_UYVY_UNALIGNED "                                      \n\
 217 movdqu      (%1), %%xmm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0   \n\
 218 movq        (%2), %%xmm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0   \n\
 219 movq        (%3), %%xmm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0   \n\
 220 prefetchnta (%0)          # Tell CPU not to cache output YUYV data        \n\
 221 punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 222 movdqa    %%xmm1, %%xmm2  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
 223 punpcklbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
 224 movdqu    %%xmm2, (%0)    # Store low UYVY                                \n\
 225 punpckhbw %%xmm0, %%xmm1  #                     y7 v3 y6 u3 y5 v2 y4 u2   \n\
 226 movdqu    %%xmm1, 16(%0)  # Store high UYVY                               \n\
 227 "
 228
 229 #elif defined(HAVE_SSE2_INTRINSICS)
 230
 231 /* SSE2 intrinsics */
 232
 233 #include <emmintrin.h>
 234
 235 #define SSE2_CALL(SSE2_INSTRUCTIONS)    \
 236     do {                                \
 237         __m128i xmm0, xmm1, xmm2;        \
 238         SSE2_INSTRUCTIONS               \
 239         p_line += 32; p_y += 16;        \
 240         p_u += 8; p_v += 8;             \
 241     } while(0)
 242
 243 #define SSE2_END  _mm_sfence()
 244
 245 #define SSE2_YUV422_YUYV_ALIGNED                \
 246     xmm0 = _mm_load_si128((__m128i *)p_y);      \
 247     xmm1 = _mm_loadl_epi64((__m128i *)p_u);     \
 248     xmm2 = _mm_loadl_epi64((__m128i *)p_v);     \
 249     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 250     xmm2 = xmm0;                                \
 251     xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);       \
 252     _mm_stream_si128((__m128i*)(p_line), xmm2); \
 253     xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);       \
 254     _mm_stream_si128((__m128i*)(p_line+16), xmm0);
 255
 256 #define SSE2_YUV422_YUYV_UNALIGNED              \
 257     xmm0 = _mm_loadu_si128((__m128i *)p_y);     \
 258     xmm1 = _mm_loadl_epi64((__m128i *)p_u);     \
 259     xmm2 = _mm_loadl_epi64((__m128i *)p_v);     \
 260     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 261     xmm2 = xmm0;                                \
 262     xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);       \
 263     _mm_storeu_si128((__m128i*)(p_line), xmm2); \
 264     xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);       \
 265     _mm_storeu_si128((__m128i*)(p_line+16), xmm0);
 266
 267 #define SSE2_YUV422_YVYU_ALIGNED                \
 268     xmm0 = _mm_load_si128((__m128i *)p_y);      \
 269     xmm2 = _mm_loadl_epi64((__m128i *)p_u);     \
 270     xmm1 = _mm_loadl_epi64((__m128i *)p_v);     \
 271     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 272     xmm2 = xmm0;                                \
 273     xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);       \
 274     _mm_stream_si128((__m128i*)(p_line), xmm2); \
 275     xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);       \
 276     _mm_stream_si128((__m128i*)(p_line+16), xmm0);
 277
 278 #define SSE2_YUV422_YVYU_UNALIGNED              \
 279     xmm0 = _mm_loadu_si128((__m128i *)p_y);     \
 280     xmm2 = _mm_loadl_epi64((__m128i *)p_u);     \
 281     xmm1 = _mm_loadl_epi64((__m128i *)p_v);     \
 282     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 283     xmm2 = xmm0;                                \
 284     xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);       \
 285     _mm_storeu_si128((__m128i*)(p_line), xmm2); \
 286     xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);       \
 287     _mm_storeu_si128((__m128i*)(p_line+16), xmm0);
 288
 289 #define SSE2_YUV422_UYVY_ALIGNED                \
 290     xmm0 = _mm_load_si128((__m128i *)p_y);      \
 291     xmm1 = _mm_loadl_epi64((__m128i *)p_u);     \
 292     xmm2 = _mm_loadl_epi64((__m128i *)p_v);     \
 293     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 294     xmm2 = xmm1;                                \
 295     xmm2 = _mm_unpacklo_epi8(xmm2, xmm0);       \
 296     _mm_stream_si128((__m128i*)(p_line), xmm2); \
 297     xmm1 = _mm_unpackhi_epi8(xmm1, xmm0);       \
 298     _mm_stream_si128((__m128i*)(p_line+16), xmm1);
 299
 300 #define SSE2_YUV422_UYVY_UNALIGNED              \
 301     xmm0 = _mm_loadu_si128((__m128i *)p_y);     \
 302     xmm1 = _mm_loadl_epi64((__m128i *)p_u);     \
 303     xmm2 = _mm_loadl_epi64((__m128i *)p_v);     \
 304     xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);       \
 305     xmm2 = xmm1;                                \
 306     xmm2 = _mm_unpacklo_epi8(xmm2, xmm0);       \
 307     _mm_storeu_si128((__m128i*)(p_line), xmm2); \
 308     xmm1 = _mm_unpackhi_epi8(xmm1, xmm0);       \
 309     _mm_storeu_si128((__m128i*)(p_line+16), xmm1);
 310
 311 #endif
 312
 313 #endif
 314
 315 #define C_YUV422_YUYV( p_line, p_y, p_u, p_v )                              \
 316     *(p_line)++ = *(p_y)++;                                                 \
 317     *(p_line)++ = *(p_u)++;                                                 \
 318     *(p_line)++ = *(p_y)++;                                                 \
 319     *(p_line)++ = *(p_v)++;                                                 \
 320
 321 #define C_YUV422_YVYU( p_line, p_y, p_u, p_v )                              \
 322     *(p_line)++ = *(p_y)++;                                                 \
 323     *(p_line)++ = *(p_v)++;                                                 \
 324     *(p_line)++ = *(p_y)++;                                                 \
 325     *(p_line)++ = *(p_u)++;                                                 \
 326
 327 #define C_YUV422_UYVY( p_line, p_y, p_u, p_v )                              \
 328     *(p_line)++ = *(p_u)++;                                                 \
 329     *(p_line)++ = *(p_y)++;                                                 \
 330     *(p_line)++ = *(p_v)++;                                                 \
 331     *(p_line)++ = *(p_y)++;                                                 \
 332
 333 #define C_YUV422_Y211( p_line, p_y, p_u, p_v )                              \
 334     *(p_line)++ = *(p_y); p_y += 2;                                         \
 335     *(p_line)++ = *(p_u) - 0x80; p_u += 2;                                  \
 336     *(p_line)++ = *(p_y); p_y += 2;                                         \
 337     *(p_line)++ = *(p_v) - 0x80; p_v += 2;                                  \
 338
 339