1 /*****************************************************************************
2 * mc.h: h264 encoder library
3 *****************************************************************************
4 * Copyright (C) 2008 x264 Project
6 * Authors: Fiona Glaser <fiona@x264.com>
7 * Loren Merritt <lorenm@u.washington.edu>
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.
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.
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., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
22 *****************************************************************************/
24 #ifndef X264_X86_UTIL_H
25 #define X264_X86_UTIL_H
28 #define x264_median_mv x264_median_mv_mmxext
29 static inline void x264_median_mv_mmxext( int16_t *dst, int16_t *a, int16_t *b, int16_t *c )
34 "movq %%mm0, %%mm3 \n"
36 "pmaxsw %%mm1, %%mm0 \n"
37 "pminsw %%mm3, %%mm1 \n"
38 "pminsw %%mm2, %%mm0 \n"
39 "pmaxsw %%mm1, %%mm0 \n"
41 :"=m"(*(uint32_t*)dst)
42 :"m"(*(uint32_t*)a), "m"(*(uint32_t*)b), "m"(*(uint32_t*)c)
45 #define x264_predictor_difference x264_predictor_difference_mmxext
46 static inline int x264_predictor_difference_mmxext( int16_t (*mvc)[2], intptr_t i_mvc )
51 "pxor %%mm4, %%mm4 \n"
54 "movd -8(%2,%1,4), %%mm0 \n"
55 "movd -4(%2,%1,4), %%mm3 \n"
56 "psubw %%mm3, %%mm0 \n"
61 "movq -8(%2,%1,4), %%mm0 \n"
62 "psubw -4(%2,%1,4), %%mm0 \n"
65 "pxor %%mm2, %%mm2 \n"
66 "psubw %%mm0, %%mm2 \n"
67 "pmaxsw %%mm2, %%mm0 \n"
68 "paddusw %%mm0, %%mm4 \n"
71 :"=m"(output), "+r"(i_mvc)
74 sum += output[0] + output[1] + output[2] + output[3];
77 #define array_non_zero_count array_non_zero_count_mmx
78 static inline int array_non_zero_count_mmx( int16_t *v )
80 static const uint64_t pw_2 = 0x0202020202020202ULL;
83 "pxor %%mm7, %%mm7 \n"
85 "movq 16(%1), %%mm1 \n"
86 "packsswb 8(%1), %%mm0 \n"
87 "packsswb 24(%1), %%mm1 \n"
88 "pcmpeqb %%mm7, %%mm0 \n"
89 "pcmpeqb %%mm7, %%mm1 \n"
90 "paddb %%mm0, %%mm1 \n"
92 "psadbw %%mm7, %%mm1 \n"
99 #undef array_non_zero_int
100 #define array_non_zero_int array_non_zero_int_mmx
101 static ALWAYS_INLINE int array_non_zero_int_mmx( void *v, int i_count )
107 "movq (%1), %%mm0 \n"
108 "por 8(%1), %%mm0 \n"
109 "por 16(%1), %%mm0 \n"
110 "por 24(%1), %%mm0 \n"
111 "por 32(%1), %%mm0 \n"
112 "por 40(%1), %%mm0 \n"
113 "por 48(%1), %%mm0 \n"
114 "por 56(%1), %%mm0 \n"
115 "por 64(%1), %%mm0 \n"
116 "por 72(%1), %%mm0 \n"
117 "por 80(%1), %%mm0 \n"
118 "por 88(%1), %%mm0 \n"
119 "por 96(%1), %%mm0 \n"
120 "por 104(%1), %%mm0 \n"
121 "por 112(%1), %%mm0 \n"
122 "por 120(%1), %%mm0 \n"
123 "packsswb %%mm0, %%mm0 \n"
130 else return array_non_zero_int_c( v, i_count );