i420_yuy2.h 23.6 KB
Newer Older
1 2 3
/*****************************************************************************
 * i420_yuy2.h : YUV to YUV conversion module for vlc
 *****************************************************************************
Jean-Baptiste Kempf's avatar
LGPL  
Jean-Baptiste Kempf committed
4
 * Copyright (C) 2000, 2001 VLC authors and VideoLAN
5
 * $Id$
6 7
 *
 * Authors: Samuel Hocevar <sam@zoy.org>
8
 *          Damien Fouilleul <damien@videolan.org>
9
 *
Jean-Baptiste Kempf's avatar
LGPL  
Jean-Baptiste Kempf committed
10 11 12
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation; either version 2.1 of the License, or
13
 * (at your option) any later version.
14
 *
15 16
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Jean-Baptiste Kempf's avatar
LGPL  
Jean-Baptiste Kempf committed
17 18
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
 * GNU Lesser General Public License for more details.
19
 *
Jean-Baptiste Kempf's avatar
LGPL  
Jean-Baptiste Kempf committed
20 21 22
 * You should have received a copy of the GNU Lesser General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
23 24 25 26
 *****************************************************************************/

#ifdef MODULE_NAME_IS_i420_yuy2_mmx

27 28 29 30 31 32
#if defined(CAN_COMPILE_MMX)

/* MMX assembly */
 
#define MMX_CALL(MMX_INSTRUCTIONS)          \
    do {                                    \
33 34
    __asm__ __volatile__(                   \
        ".p2align 3 \n\t                    \
35 36
movd       (%0), %%mm1  # Load 4 Cb           00 00 00 00 u3 u2 u1 u0     \n\
movd       (%1), %%mm2  # Load 4 Cr           00 00 00 00 v3 v2 v1 v0     \n\
37 38 39 40 41
movq       (%2), %%mm0  # Load 8 Y            y7 y6 y5 y4 y3 y2 y1 y0     \n\
movq       (%3), %%mm3  # Load 8 Y            Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
" \
        :                                   \
        : "r" (p_u), "r" (p_v),             \
42 43
          "r" (p_y1), "r" (p_y2)            \
        : "mm0", "mm1", "mm2", "mm3");      \
44 45 46 47
    __asm__ __volatile__(                   \
        ".p2align 3 \n\t"                   \
        MMX_INSTRUCTIONS                    \
        :                                   \
48 49
        : "r" (p_line1), "r" (p_line2)      \
        : "mm0", "mm1", "mm2", "mm3");      \
50 51 52
        p_line1 += 16; p_line2 += 16;       \
        p_y1 += 8; p_y2 += 8;               \
        p_u += 4; p_v += 4;                 \
53
    } while(0)
54

55 56
#define MMX_END __asm__ __volatile__ ( "emms" )

57 58 59 60 61 62 63
#define MMX_YUV420_YUYV "                                                 \n\
punpcklbw %%mm2, %%mm1  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
movq      %%mm0, %%mm2  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
punpcklbw %%mm1, %%mm2  #                     v1 y3 u1 y2 v0 y1 u0 y0     \n\
movq      %%mm2, (%0)   # Store low YUYV                                  \n\
punpckhbw %%mm1, %%mm0  #                     v3 y7 u3 y6 v2 y5 u2 y4     \n\
movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
64 65 66 67 68
movq      %%mm3, %%mm4  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
punpcklbw %%mm1, %%mm4  #                     v1 Y3 u1 Y2 v0 Y1 u0 Y0     \n\
movq      %%mm4, (%1)   # Store low YUYV                                  \n\
punpckhbw %%mm1, %%mm3  #                     v3 Y7 u3 Y6 v2 Y5 u2 Y4     \n\
movq      %%mm3, 8(%1)  # Store high YUYV                                 \n\
69 70 71
"

#define MMX_YUV420_YVYU "                                                 \n\
72 73 74 75 76
punpcklbw %%mm1, %%mm2  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
movq      %%mm0, %%mm1  #                     y7 y6 y5 y4 y3 y2 y1 y0     \n\
punpcklbw %%mm2, %%mm1  #                     u1 y3 v1 y2 u0 y1 v0 y0     \n\
movq      %%mm1, (%0)   # Store low YUYV                                  \n\
punpckhbw %%mm2, %%mm0  #                     u3 y7 v3 y6 u2 y5 v2 y4     \n\
77
movq      %%mm0, 8(%0)  # Store high YUYV                                 \n\
78
movq      %%mm3, %%mm4  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
79
punpcklbw %%mm2, %%mm4  #                     u1 Y3 v1 Y2 u0 Y1 v0 Y0     \n\
80
movq      %%mm4, (%1)   # Store low YUYV                                  \n\
81
punpckhbw %%mm2, %%mm3  #                     u3 Y7 v3 Y6 u2 Y5 v2 Y4     \n\
82
movq      %%mm3, 8(%1)  # Store high YUYV                                 \n\
83 84 85
"

#define MMX_YUV420_UYVY "                                                 \n\
86 87
punpcklbw %%mm2, %%mm1  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
movq      %%mm1, %%mm2  #                     v3 u3 v2 u2 v1 u1 v0 u0     \n\
88 89 90 91 92
punpcklbw %%mm0, %%mm2  #                     y3 v1 y2 u1 y1 v0 y0 u0     \n\
movq      %%mm2, (%0)   # Store low UYVY                                  \n\
movq      %%mm1, %%mm2  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
punpckhbw %%mm0, %%mm2  #                     y3 v1 y2 u1 y1 v0 y0 u0     \n\
movq      %%mm2, 8(%0)  # Store high UYVY                                 \n\
93 94 95
movq      %%mm1, %%mm4  #                     u3 v3 u2 v2 u1 v1 u0 v0     \n\
punpcklbw %%mm3, %%mm4  #                     Y3 v1 Y2 u1 Y1 v0 Y0 u0     \n\
movq      %%mm4, (%1)   # Store low UYVY                                  \n\
96 97 98 99
punpckhbw %%mm3, %%mm1  #                     Y7 v3 Y6 u3 Y5 v2 Y4 u2     \n\
movq      %%mm1, 8(%1)  # Store high UYVY                                 \n\
"

100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
#elif defined(HAVE_MMX_INTRINSICS)

/* MMX intrinsics */

#include <mmintrin.h>

#define MMX_CALL(MMX_INSTRUCTIONS)          \
    do {                                    \
        __m64 mm0, mm1, mm2, mm3, mm4;      \
        MMX_INSTRUCTIONS                    \
        p_line1 += 16; p_line2 += 16;       \
        p_y1 += 8; p_y2 += 8;               \
        p_u += 4; p_v += 4;                 \
    } while(0)

#define MMX_END _mm_empty()
116
 
117
#define MMX_YUV420_YUYV                     \
118 119
    mm1 = _mm_cvtsi32_si64(*(int*)p_u);     \
    mm2 = _mm_cvtsi32_si64(*(int*)p_v);     \
120 121 122 123 124
    mm0 = (__m64)*(uint64_t*)p_y1;          \
    mm3 = (__m64)*(uint64_t*)p_y2;          \
    mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
    mm2 = mm0;                              \
    mm2 = _mm_unpacklo_pi8(mm2, mm1);       \
125
    *(uint64_t*)p_line1 = (uint64_t)mm2;    \
126
    mm0 = _mm_unpackhi_pi8(mm0, mm1);       \
127
    *(uint64_t*)(p_line1+8) = (uint64_t)mm0;\
128 129
    mm4 = mm3;                              \
    mm4 = _mm_unpacklo_pi8(mm4, mm1);       \
130
    *(uint64_t*)p_line2 = (uint64_t)mm4;    \
131
    mm3 = _mm_unpackhi_pi8(mm3, mm1);       \
132
    *(uint64_t*)(p_line2+8) = (uint64_t)mm3;
133 134

#define MMX_YUV420_YVYU                     \
135 136
    mm2 = _mm_cvtsi32_si64(*(int*)p_u);     \
    mm1 = _mm_cvtsi32_si64(*(int*)p_v);     \
137 138 139 140 141
    mm0 = (__m64)*(uint64_t*)p_y1;          \
    mm3 = (__m64)*(uint64_t*)p_y2;          \
    mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
    mm2 = mm0;                              \
    mm2 = _mm_unpacklo_pi8(mm2, mm1);       \
142
    *(uint64_t*)p_line1 = (uint64_t)mm2;    \
143
    mm0 = _mm_unpackhi_pi8(mm0, mm1);       \
144
    *(uint64_t*)(p_line1+8) = (uint64_t)mm0;\
145 146
    mm4 = mm3;                              \
    mm4 = _mm_unpacklo_pi8(mm4, mm1);       \
147
    *(uint64_t*)p_line2 = (uint64_t)mm4;    \
148
    mm3 = _mm_unpackhi_pi8(mm3, mm1);       \
149
    *(uint64_t*)(p_line2+8) = (uint64_t)mm3;
150 151

#define MMX_YUV420_UYVY                     \
152 153
    mm1 = _mm_cvtsi32_si64(*(int*)p_u);     \
    mm2 = _mm_cvtsi32_si64(*(int*)p_v);     \
154 155 156 157 158
    mm0 = (__m64)*(uint64_t*)p_y1;          \
    mm3 = (__m64)*(uint64_t*)p_y2;          \
    mm1 = _mm_unpacklo_pi8(mm1, mm2);       \
    mm2 = mm1;                              \
    mm2 = _mm_unpacklo_pi8(mm2, mm0);       \
159
    *(uint64_t*)p_line1 = (uint64_t)mm2;    \
160 161
    mm2 = mm1;                              \
    mm2 = _mm_unpackhi_pi8(mm2, mm0);       \
162
    *(uint64_t*)(p_line1+8) = (uint64_t)mm2;\
163 164
    mm4 = mm1;                              \
    mm4 = _mm_unpacklo_pi8(mm4, mm3);       \
165
    *(uint64_t*)p_line2 = (uint64_t)mm4;    \
166
    mm1 = _mm_unpackhi_pi8(mm1, mm3);       \
167
    *(uint64_t*)(p_line2+8) = (uint64_t)mm1;
168 169

#endif
170

171 172
#elif defined( MODULE_NAME_IS_i420_yuy2_sse2 )

173 174 175 176 177 178
#if defined(CAN_COMPILE_SSE2)

/* SSE2 assembly */

#define SSE2_CALL(SSE2_INSTRUCTIONS)    \
    do {                                \
179 180 181 182 183 184
    __asm__ __volatile__(               \
        ".p2align 3 \n\t                \
movq        (%0), %%xmm1  # Load 8 Cb         u7 u6 u5 u4 u3 u2 u1 u0     \n\
movq        (%1), %%xmm2  # Load 8 Cr         v7 06 v5 v4 v3 v2 v1 v0     \n\
" \
        :                               \
185 186
        : "r" (p_u),  "r" (p_v)         \
        : "xmm1", "xmm2");              \
187 188 189 190 191
    __asm__ __volatile__(               \
        ".p2align 3 \n\t"               \
        SSE2_INSTRUCTIONS               \
        :                               \
        : "r" (p_line1), "r" (p_line2), \
192 193
          "r" (p_y1),  "r" (p_y2)       \
        : "xmm0", "xmm1", "xmm2", "xmm3", "xmm4"); \
194 195 196
        p_line1 += 32; p_line2 += 32;   \
        p_y1 += 16; p_y2 += 16;         \
        p_u += 8; p_v += 8;             \
197 198
    } while(0)

199 200
#define SSE2_END  __asm__ __volatile__ ( "sfence" ::: "memory" )

201
#define SSE2_YUV420_YUYV_ALIGNED "                                        \n\
202 203
movdqa      (%2), %%xmm0  # Load 16 Y         y15 y14 y13 .. y2 y1 y0     \n\
movdqa      (%3), %%xmm3  # Load 16 Y         Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
204 205 206
punpcklbw %%xmm2, %%xmm1  #                   v7 u7 v6 u6 .. u1 v0 u0     \n\
movdqa    %%xmm0, %%xmm2  #                   y15 y14 y13 .. y2 y1 y0     \n\
punpcklbw %%xmm1, %%xmm2  #                   v3 y7 u3 .. v0 y1 u0 y0     \n\
207
movntdq   %%xmm2, (%0)    # Store low YUYV                                \n\
208
punpckhbw %%xmm1, %%xmm0  #                   v3 y7 u3 y6 v2 y5 u2 y4     \n\
209
movntdq   %%xmm0, 16(%0)  # Store high YUYV                               \n\
210 211
movdqa    %%xmm3, %%xmm4  #                   Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
punpcklbw %%xmm1, %%xmm4  #                   v1 Y3 u1 Y2 v0 Y1 u0 Y0     \n\
212
movntdq   %%xmm4, (%1)    # Store low YUYV                                \n\
213
punpckhbw %%xmm1, %%xmm3  #                   v3 Y7 u3 Y6 v2 Y5 u2 Y4     \n\
214
movntdq   %%xmm3, 16(%1)  # Store high YUYV                               \n\
215 216 217
"

#define SSE2_YUV420_YUYV_UNALIGNED "                                      \n\
218 219
movdqu      (%2), %%xmm0  # Load 16 Y         y7 y6 y5 y4 y3 y2 y1 y0     \n\
movdqu      (%3), %%xmm3  # Load 16 Y         Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
220 221
prefetchnta (%0)          # Tell CPU not to cache output YUYV data        \n\
prefetchnta (%1)          # Tell CPU not to cache output YUYV data        \n\
222 223 224 225 226 227
punpcklbw %%xmm2, %%xmm1  #                   v3 u3 v2 u2 v1 u1 v0 u0     \n\
movdqa    %%xmm0, %%xmm2  #                   y7 y6 y5 y4 y3 y2 y1 y0     \n\
punpcklbw %%xmm1, %%xmm2  #                   v1 y3 u1 y2 v0 y1 u0 y0     \n\
movdqu    %%xmm2, (%0)    # Store low YUYV                                \n\
punpckhbw %%xmm1, %%xmm0  #                   v3 y7 u3 y6 v2 y5 u2 y4     \n\
movdqu    %%xmm0, 16(%0)  # Store high YUYV                               \n\
228 229 230 231 232
movdqa    %%xmm3, %%xmm4  #                   Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0     \n\
punpcklbw %%xmm1, %%xmm4  #                   v1 Y3 u1 Y2 v0 Y1 u0 Y0     \n\
movdqu    %%xmm4, (%1)    # Store low YUYV                                \n\
punpckhbw %%xmm1, %%xmm3  #                   v3 Y7 u3 Y6 v2 Y5 u2 Y4     \n\
movdqu    %%xmm3, 16(%1)  # Store high YUYV                               \n\
233 234
"

235 236 237
#define SSE2_YUV420_YVYU_ALIGNED "                                        \n\
movdqa      (%2), %%xmm0  # Load 16 Y           y7 y6 y5 y4 y3 y2 y1 y0   \n\
movdqa      (%3), %%xmm3  # Load 16 Y           Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
238 239 240 241 242
punpcklbw %%xmm1, %%xmm2  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
movdqa    %%xmm0, %%xmm1  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
punpcklbw %%xmm2, %%xmm1  #                     u1 y3 v1 y2 u0 y1 v0 y0   \n\
movntdq   %%xmm1, (%0)    # Store low YUYV                                \n\
punpckhbw %%xmm2, %%xmm0  #                     u3 y7 v3 y6 u2 y5 v2 y4   \n\
243 244
movntdq   %%xmm0, 16(%0)  # Store high YUYV                               \n\
movdqa    %%xmm3, %%xmm4  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
245
punpcklbw %%xmm2, %%xmm4  #                     u1 Y3 v1 Y2 u0 Y1 v0 Y0   \n\
246
movntdq   %%xmm4, (%1)    # Store low YUYV                                \n\
247
punpckhbw %%xmm2, %%xmm3  #                     u3 Y7 v3 Y6 u2 Y5 v2 Y4   \n\
248
movntdq   %%xmm3, 16(%1)  # Store high YUYV                               \n\
249 250
"

251 252 253 254 255
#define SSE2_YUV420_YVYU_UNALIGNED "                                      \n\
movdqu      (%2), %%xmm0  # Load 16 Y           y7 y6 y5 y4 y3 y2 y1 y0   \n\
movdqu      (%3), %%xmm3  # Load 16 Y           Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
prefetchnta (%0)          # Tell CPU not to cache output YVYU data        \n\
prefetchnta (%1)          # Tell CPU not to cache output YVYU data        \n\
256 257 258 259 260
punpcklbw %%xmm1, %%xmm2  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
movdqu    %%xmm0, %%xmm1  #                     y7 y6 y5 y4 y3 y2 y1 y0   \n\
punpcklbw %%xmm2, %%xmm1  #                     u1 y3 v1 y2 u0 y1 v0 y0   \n\
movdqu    %%xmm1, (%0)    # Store low YUYV                                \n\
punpckhbw %%xmm2, %%xmm0  #                     u3 y7 v3 y6 u2 y5 v2 y4   \n\
261 262
movdqu    %%xmm0, 16(%0)  # Store high YUYV                               \n\
movdqu    %%xmm3, %%xmm4  #                     Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
263
punpcklbw %%xmm2, %%xmm4  #                     u1 Y3 v1 Y2 u0 Y1 v0 Y0   \n\
264
movdqu    %%xmm4, (%1)    # Store low YUYV                                \n\
265
punpckhbw %%xmm2, %%xmm3  #                     u3 Y7 v3 Y6 u2 Y5 v2 Y4   \n\
266
movdqu    %%xmm3, 16(%1)  # Store high YUYV                               \n\
267 268
"

269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
#define SSE2_YUV420_UYVY_ALIGNED "                                        \n\
movdqa      (%2), %%xmm0  # Load 16 Y           y7 y6 y5 y4 y3 y2 y1 y0   \n\
movdqa      (%3), %%xmm3  # Load 16 Y           Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
movdqa    %%xmm1, %%xmm2  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
punpcklbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
movntdq   %%xmm2, (%0)    # Store low UYVY                                \n\
movdqa    %%xmm1, %%xmm2  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
punpckhbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
movntdq   %%xmm2, 16(%0)  # Store high UYVY                               \n\
movdqa    %%xmm1, %%xmm4  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
punpcklbw %%xmm3, %%xmm4  #                     Y3 v1 Y2 u1 Y1 v0 Y0 u0   \n\
movntdq   %%xmm4, (%1)    # Store low UYVY                                \n\
punpckhbw %%xmm3, %%xmm1  #                     Y7 v3 Y6 u3 Y5 v2 Y4 u2   \n\
movntdq   %%xmm1, 16(%1)  # Store high UYVY                               \n\
284 285
"

286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
#define SSE2_YUV420_UYVY_UNALIGNED "                                      \n\
movdqu      (%2), %%xmm0  # Load 16 Y           y7 y6 y5 y4 y3 y2 y1 y0   \n\
movdqu      (%3), %%xmm3  # Load 16 Y           Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0   \n\
prefetchnta (%0)          # Tell CPU not to cache output UYVY data        \n\
prefetchnta (%1)          # Tell CPU not to cache output UYVY data        \n\
punpcklbw %%xmm2, %%xmm1  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
movdqu    %%xmm1, %%xmm2  #                     v3 u3 v2 u2 v1 u1 v0 u0   \n\
punpcklbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
movdqu    %%xmm2, (%0)    # Store low UYVY                                \n\
movdqu    %%xmm1, %%xmm2  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
punpckhbw %%xmm0, %%xmm2  #                     y3 v1 y2 u1 y1 v0 y0 u0   \n\
movdqu    %%xmm2, 16(%0)  # Store high UYVY                               \n\
movdqu    %%xmm1, %%xmm4  #                     u3 v3 u2 v2 u1 v1 u0 v0   \n\
punpcklbw %%xmm3, %%xmm4  #                     Y3 v1 Y2 u1 Y1 v0 Y0 u0   \n\
movdqu    %%xmm4, (%1)    # Store low UYVY                                \n\
punpckhbw %%xmm3, %%xmm1  #                     Y7 v3 Y6 u3 Y5 v2 Y4 u2   \n\
movdqu    %%xmm1, 16(%1)  # Store high UYVY                               \n\
303 304
"

305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
#elif defined(HAVE_SSE2_INTRINSICS)

/* SSE2 intrinsics */

#include <emmintrin.h>

#define SSE2_CALL(SSE2_INSTRUCTIONS)            \
    do {                                        \
        __m128i xmm0, xmm1, xmm2, xmm3, xmm4;   \
        SSE2_INSTRUCTIONS                       \
        p_line1 += 32; p_line2 += 32;           \
        p_y1 += 16; p_y2 += 16;                 \
        p_u += 8; p_v += 8;                     \
    } while(0)

#define SSE2_END  _mm_sfence()

#define SSE2_YUV420_YUYV_ALIGNED                    \
    xmm1 = _mm_loadl_epi64((__m128i *)p_u);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_v);         \
    xmm0 = _mm_load_si128((__m128i *)p_y1);         \
    xmm3 = _mm_load_si128((__m128i *)p_y2);         \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm0;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1), xmm2);    \
    xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm0); \
    xmm4 = xmm3;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line2), xmm4);    \
    xmm3 = _mm_unpackhi_epi8(xmm3, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm3);

#define SSE2_YUV420_YUYV_UNALIGNED                  \
    xmm1 = _mm_loadl_epi64((__m128i *)p_u);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_v);         \
342 343
    xmm0 = _mm_loadu_si128((__m128i *)p_y1);        \
    xmm3 = _mm_loadu_si128((__m128i *)p_y2);        \
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
    _mm_prefetch(p_line1, _MM_HINT_NTA);            \
    _mm_prefetch(p_line2, _MM_HINT_NTA);            \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm0;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1), xmm2);    \
    xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm0); \
    xmm4 = xmm3;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line2), xmm4);    \
    xmm3 = _mm_unpackhi_epi8(xmm3, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm3);

#define SSE2_YUV420_YVYU_ALIGNED                    \
    xmm1 = _mm_loadl_epi64((__m128i *)p_v);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_u);         \
    xmm0 = _mm_load_si128((__m128i *)p_y1);         \
    xmm3 = _mm_load_si128((__m128i *)p_y2);         \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm0;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1), xmm2);    \
    xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm0); \
    xmm4 = xmm3;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line2), xmm4);    \
    xmm3 = _mm_unpackhi_epi8(xmm3, xmm1);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm3);

#define SSE2_YUV420_YVYU_UNALIGNED                  \
    xmm1 = _mm_loadl_epi64((__m128i *)p_v);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_u);         \
378 379
    xmm0 = _mm_loadu_si128((__m128i *)p_y1);        \
    xmm3 = _mm_loadu_si128((__m128i *)p_y2);        \
380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
    _mm_prefetch(p_line1, _MM_HINT_NTA);            \
    _mm_prefetch(p_line2, _MM_HINT_NTA);            \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm0;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1), xmm2);    \
    xmm0 = _mm_unpackhi_epi8(xmm0, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm0); \
    xmm4 = xmm3;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line2), xmm4);    \
    xmm3 = _mm_unpackhi_epi8(xmm3, xmm1);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm3);

#define SSE2_YUV420_UYVY_ALIGNED                    \
    xmm1 = _mm_loadl_epi64((__m128i *)p_u);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_v);         \
    xmm0 = _mm_load_si128((__m128i *)p_y1);         \
    xmm3 = _mm_load_si128((__m128i *)p_y2);         \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm1;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm0);           \
    _mm_stream_si128((__m128i*)(p_line1), xmm2);    \
    xmm2 = xmm1;                                    \
    xmm2 = _mm_unpackhi_epi8(xmm2, xmm0);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm2); \
    xmm4 = xmm1;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm3);           \
    _mm_stream_si128((__m128i*)(p_line2), xmm4);    \
    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3);           \
    _mm_stream_si128((__m128i*)(p_line1+16), xmm1);

#define SSE2_YUV420_UYVY_UNALIGNED                  \
    xmm1 = _mm_loadl_epi64((__m128i *)p_u);         \
    xmm2 = _mm_loadl_epi64((__m128i *)p_v);         \
415 416
    xmm0 = _mm_loadu_si128((__m128i *)p_y1);        \
    xmm3 = _mm_loadu_si128((__m128i *)p_y2);        \
417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
    _mm_prefetch(p_line1, _MM_HINT_NTA);            \
    _mm_prefetch(p_line2, _MM_HINT_NTA);            \
    xmm1 = _mm_unpacklo_epi8(xmm1, xmm2);           \
    xmm2 = xmm1;                                    \
    xmm2 = _mm_unpacklo_epi8(xmm2, xmm0);           \
    _mm_storeu_si128((__m128i*)(p_line1), xmm2);    \
    xmm2 = xmm1;                                    \
    xmm2 = _mm_unpackhi_epi8(xmm2, xmm0);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm2); \
    xmm4 = xmm1;                                    \
    xmm4 = _mm_unpacklo_epi8(xmm4, xmm3);           \
    _mm_storeu_si128((__m128i*)(p_line2), xmm4);    \
    xmm1 = _mm_unpackhi_epi8(xmm1, xmm3);           \
    _mm_storeu_si128((__m128i*)(p_line1+16), xmm1);

#endif

434 435 436
#endif

/* Used in both accelerated and C modules */
437 438 439 440 441 442 443 444

#define C_YUV420_YVYU( )                                                    \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
    *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
    *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \

#define C_YUV420_Y211( )                                                    \
445 446 447 448 449 450
    *(p_line1)++ = *(p_y1); p_y1 += 2;                                      \
    *(p_line2)++ = *(p_y2); p_y2 += 2;                                      \
    *(p_line1)++ = *(p_line2)++ = *(p_u) - 0x80; p_u += 2;                  \
    *(p_line1)++ = *(p_y1); p_y1 += 2;                                      \
    *(p_line2)++ = *(p_y2); p_y2 += 2;                                      \
    *(p_line1)++ = *(p_line2)++ = *(p_v) - 0x80; p_v += 2;                  \
451 452


453 454 455 456 457 458
#define C_YUV420_YUYV( )                                                    \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
    *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
    *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \

459 460 461 462 463 464
#define C_YUV420_UYVY( )                                                    \
    *(p_line1)++ =            *(p_line2)++ = *(p_u)++;                      \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \
    *(p_line1)++ =            *(p_line2)++ = *(p_v)++;                      \
    *(p_line1)++ = *(p_y1)++; *(p_line2)++ = *(p_y2)++;                     \