2 * MMX optimized DSP utils
3 * Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
5 * This file is part of FFmpeg.
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.
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.
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
22 #ifndef FFMPEG_DSPUTIL_MMX_H
23 #define FFMPEG_DSPUTIL_MMX_H
27 typedef struct { uint64_t a, b; } xmm_t;
29 extern const uint64_t ff_bone;
30 extern const uint64_t ff_wtwo;
32 extern const uint64_t ff_pdw_80000000[2];
34 extern const uint64_t ff_pw_3;
35 extern const uint64_t ff_pw_4;
36 extern const xmm_t ff_pw_5;
37 extern const uint64_t ff_pw_8;
38 extern const uint64_t ff_pw_15;
39 extern const xmm_t ff_pw_16;
40 extern const uint64_t ff_pw_20;
41 extern const xmm_t ff_pw_32;
42 extern const uint64_t ff_pw_42;
43 extern const uint64_t ff_pw_64;
44 extern const uint64_t ff_pw_96;
45 extern const uint64_t ff_pw_128;
47 extern const uint64_t ff_pb_1;
48 extern const uint64_t ff_pb_3;
49 extern const uint64_t ff_pb_7;
50 extern const uint64_t ff_pb_3F;
51 extern const uint64_t ff_pb_A1;
52 extern const uint64_t ff_pb_FC;
54 extern const double ff_pd_1[2];
55 extern const double ff_pd_2[2];
57 /* in/out: mma=mma+mmb, mmb=mmb-mma */
58 #define SUMSUB_BA( a, b ) \
59 "paddw "#b", "#a" \n\t"\
60 "paddw "#b", "#b" \n\t"\
61 "psubw "#a", "#b" \n\t"
63 #define SBUTTERFLY(a,b,t,n,m)\
64 "mov" #m " " #a ", " #t " \n\t" /* abcd */\
65 "punpckl" #n " " #b ", " #a " \n\t" /* aebf */\
66 "punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\
68 #define TRANSPOSE4(a,b,c,d,t)\
69 SBUTTERFLY(a,b,t,wd,q) /* a=aebf t=cgdh */\
70 SBUTTERFLY(c,d,b,wd,q) /* c=imjn b=kolp */\
71 SBUTTERFLY(a,c,d,dq,q) /* a=aeim d=bfjn */\
72 SBUTTERFLY(t,b,c,dq,q) /* t=cgko c=dhlp */
75 // permutes 01234567 -> 05736421
76 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
77 SBUTTERFLY(a,b,%%xmm8,wd,dqa)\
78 SBUTTERFLY(c,d,b,wd,dqa)\
79 SBUTTERFLY(e,f,d,wd,dqa)\
80 SBUTTERFLY(g,h,f,wd,dqa)\
81 SBUTTERFLY(a,c,h,dq,dqa)\
82 SBUTTERFLY(%%xmm8,b,c,dq,dqa)\
83 SBUTTERFLY(e,g,b,dq,dqa)\
84 SBUTTERFLY(d,f,g,dq,dqa)\
85 SBUTTERFLY(a,e,f,qdq,dqa)\
86 SBUTTERFLY(%%xmm8,d,e,qdq,dqa)\
87 SBUTTERFLY(h,b,d,qdq,dqa)\
88 SBUTTERFLY(c,g,b,qdq,dqa)\
89 "movdqa %%xmm8, "#g" \n\t"
91 #define TRANSPOSE8(a,b,c,d,e,f,g,h,t)\
92 "movdqa "#h", "#t" \n\t"\
93 SBUTTERFLY(a,b,h,wd,dqa)\
94 "movdqa "#h", 16"#t" \n\t"\
95 "movdqa "#t", "#h" \n\t"\
96 SBUTTERFLY(c,d,b,wd,dqa)\
97 SBUTTERFLY(e,f,d,wd,dqa)\
98 SBUTTERFLY(g,h,f,wd,dqa)\
99 SBUTTERFLY(a,c,h,dq,dqa)\
100 "movdqa "#h", "#t" \n\t"\
101 "movdqa 16"#t", "#h" \n\t"\
102 SBUTTERFLY(h,b,c,dq,dqa)\
103 SBUTTERFLY(e,g,b,dq,dqa)\
104 SBUTTERFLY(d,f,g,dq,dqa)\
105 SBUTTERFLY(a,e,f,qdq,dqa)\
106 SBUTTERFLY(h,d,e,qdq,dqa)\
107 "movdqa "#h", 16"#t" \n\t"\
108 "movdqa "#t", "#h" \n\t"\
109 SBUTTERFLY(h,b,d,qdq,dqa)\
110 SBUTTERFLY(c,g,b,qdq,dqa)\
111 "movdqa 16"#t", "#g" \n\t"
114 #endif /* FFMPEG_DSPUTIL_MMX_H */