Statistics
| Branch: | Revision:

ffmpeg / libavcodec / i386 / dsputil_mmx.h @ 28748a91

History | View | Annotate | Download (3.49 KB)

1
/*
2
 * MMX optimized DSP utils
3
 * Copyright (c) 2007  Aurelien Jacobs <aurel@gnuage.org>
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21

    
22
#ifndef FFMPEG_DSPUTIL_MMX_H
23
#define FFMPEG_DSPUTIL_MMX_H
24

    
25
#include <stdint.h>
26

    
27
extern const uint64_t ff_bone;
28
extern const uint64_t ff_wtwo;
29

    
30
extern const uint64_t ff_pdw_80000000[2];
31

    
32
extern const uint64_t ff_pw_3;
33
extern const uint64_t ff_pw_4;
34
extern const uint64_t ff_pw_5;
35
extern const uint64_t ff_pw_8;
36
extern const uint64_t ff_pw_15;
37
extern const uint64_t ff_pw_16;
38
extern const uint64_t ff_pw_20;
39
extern const uint64_t ff_pw_32;
40
extern const uint64_t ff_pw_42;
41
extern const uint64_t ff_pw_64;
42
extern const uint64_t ff_pw_96;
43
extern const uint64_t ff_pw_128;
44

    
45
extern const uint64_t ff_pb_1;
46
extern const uint64_t ff_pb_3;
47
extern const uint64_t ff_pb_7;
48
extern const uint64_t ff_pb_3F;
49
extern const uint64_t ff_pb_A1;
50
extern const uint64_t ff_pb_FC;
51

    
52
extern const double ff_pd_1[2];
53
extern const double ff_pd_2[2];
54

    
55
/* in/out: mma=mma+mmb, mmb=mmb-mma */
56
#define SUMSUB_BA( a, b ) \
57
    "paddw "#b", "#a" \n\t"\
58
    "paddw "#b", "#b" \n\t"\
59
    "psubw "#a", "#b" \n\t"
60

    
61
#define SBUTTERFLY(a,b,t,n,m)\
62
    "mov" #m " " #a ", " #t "         \n\t" /* abcd */\
63
    "punpckl" #n " " #b ", " #a "     \n\t" /* aebf */\
64
    "punpckh" #n " " #b ", " #t "     \n\t" /* cgdh */\
65

    
66
#define TRANSPOSE4(a,b,c,d,t)\
67
    SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
68
    SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
69
    SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
70
    SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
71

    
72
#ifdef ARCH_X86_64
73
// permutes 01234567 -> 05736421
74
#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
75
    SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
76
    SBUTTERFLY(c,d,b,wd,dqa)\
77
    SBUTTERFLY(e,f,d,wd,dqa)\
78
    SBUTTERFLY(g,h,f,wd,dqa)\
79
    SBUTTERFLY(a,c,h,dq,dqa)\
80
    SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
81
    SBUTTERFLY(e,g,b,dq,dqa)\
82
    SBUTTERFLY(d,f,g,dq,dqa)\
83
    SBUTTERFLY(a,e,f,qdq,dqa)\
84
    SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
85
    SBUTTERFLY(h,b,d,qdq,dqa)\
86
    SBUTTERFLY(c,g,b,qdq,dqa)\
87
    "movdqa %%xmm8, "#g"              \n\t"
88
#else
89
#define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
90
    "movdqa "#h", "#t"                \n\t"\
91
    SBUTTERFLY(a,b,h,wd,dqa)\
92
    "movdqa "#h", 16"#t"              \n\t"\
93
    "movdqa "#t", "#h"                \n\t"\
94
    SBUTTERFLY(c,d,b,wd,dqa)\
95
    SBUTTERFLY(e,f,d,wd,dqa)\
96
    SBUTTERFLY(g,h,f,wd,dqa)\
97
    SBUTTERFLY(a,c,h,dq,dqa)\
98
    "movdqa "#h", "#t"                \n\t"\
99
    "movdqa 16"#t", "#h"              \n\t"\
100
    SBUTTERFLY(h,b,c,dq,dqa)\
101
    SBUTTERFLY(e,g,b,dq,dqa)\
102
    SBUTTERFLY(d,f,g,dq,dqa)\
103
    SBUTTERFLY(a,e,f,qdq,dqa)\
104
    SBUTTERFLY(h,d,e,qdq,dqa)\
105
    "movdqa "#h", 16"#t"              \n\t"\
106
    "movdqa "#t", "#h"                \n\t"\
107
    SBUTTERFLY(h,b,d,qdq,dqa)\
108
    SBUTTERFLY(c,g,b,qdq,dqa)\
109
    "movdqa 16"#t", "#g"              \n\t"
110
#endif
111

    
112
#endif /* FFMPEG_DSPUTIL_MMX_H */