Statistics
| Branch: | Revision:

ffmpeg / libavcodec / ppc / dsputil_altivec.h @ 2722c981

History | View | Annotate | Download (3.67 KB)

1
/*
2
 * Copyright (c) 2002 Brian Foley
3
 * Copyright (c) 2002 Dieter Shirley
4
 * Copyright (c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
5
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg is free software; you can redistribute it and/or
9
 * modify it under the terms of the GNU Lesser General Public
10
 * License as published by the Free Software Foundation; either
11
 * version 2.1 of the License, or (at your option) any later version.
12
 *
13
 * FFmpeg is distributed in the hope that it will be useful,
14
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16
 * Lesser General Public License for more details.
17
 *
18
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21
 */
22

    
23
#ifndef DSPUTIL_ALTIVEC_H
24
#define DSPUTIL_ALTIVEC_H
25

    
26
#include "dsputil_ppc.h"
27

    
28
extern int has_altivec(void);
29

    
30
void put_pixels16_altivec(uint8_t *block, const uint8_t *pixels, int line_size, int h);
31

    
32
void avg_pixels16_altivec(uint8_t *block, const uint8_t *pixels, int line_size, int h);
33

    
34
// used to build registers permutation vectors (vcprm)
35
// the 's' are for words in the _s_econd vector
36
#define WORD_0 0x00,0x01,0x02,0x03
37
#define WORD_1 0x04,0x05,0x06,0x07
38
#define WORD_2 0x08,0x09,0x0a,0x0b
39
#define WORD_3 0x0c,0x0d,0x0e,0x0f
40
#define WORD_s0 0x10,0x11,0x12,0x13
41
#define WORD_s1 0x14,0x15,0x16,0x17
42
#define WORD_s2 0x18,0x19,0x1a,0x1b
43
#define WORD_s3 0x1c,0x1d,0x1e,0x1f
44

    
45
#ifdef __APPLE_CC__
46
#define vcprm(a,b,c,d) (const vector unsigned char)(WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d)
47
#else
48
#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
49
#endif
50

    
51
// vcprmle is used to keep the same index as in the SSE version.
52
// it's the same as vcprm, with the index inversed
53
// ('le' is Little Endian)
54
#define vcprmle(a,b,c,d) vcprm(d,c,b,a)
55

    
56
// used to build inverse/identity vectors (vcii)
57
// n is _n_egative, p is _p_ositive
58
#define FLOAT_n -1.
59
#define FLOAT_p 1.
60

    
61

    
62
#ifdef __APPLE_CC__
63
#define vcii(a,b,c,d) (const vector float)(FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d)
64
#else
65
#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
66
#endif
67

    
68
// Transpose 8x8 matrix of 16-bit elements (in-place)
69
#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
70
do { \
71
    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
72
    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
73
 \
74
    A1 = vec_mergeh (a, e); \
75
    B1 = vec_mergel (a, e); \
76
    C1 = vec_mergeh (b, f); \
77
    D1 = vec_mergel (b, f); \
78
    E1 = vec_mergeh (c, g); \
79
    F1 = vec_mergel (c, g); \
80
    G1 = vec_mergeh (d, h); \
81
    H1 = vec_mergel (d, h); \
82
 \
83
    A2 = vec_mergeh (A1, E1); \
84
    B2 = vec_mergel (A1, E1); \
85
    C2 = vec_mergeh (B1, F1); \
86
    D2 = vec_mergel (B1, F1); \
87
    E2 = vec_mergeh (C1, G1); \
88
    F2 = vec_mergel (C1, G1); \
89
    G2 = vec_mergeh (D1, H1); \
90
    H2 = vec_mergel (D1, H1); \
91
 \
92
    a = vec_mergeh (A2, E2); \
93
    b = vec_mergel (A2, E2); \
94
    c = vec_mergeh (B2, F2); \
95
    d = vec_mergel (B2, F2); \
96
    e = vec_mergeh (C2, G2); \
97
    f = vec_mergel (C2, G2); \
98
    g = vec_mergeh (D2, H2); \
99
    h = vec_mergel (D2, H2); \
100
} while (0)
101

    
102

    
103
/** \brief loads unaligned vector \a *src with offset \a offset
104
    and returns it */
105
static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
106
{
107
    register vector unsigned char first = vec_ld(offset, src);
108
    register vector unsigned char second = vec_ld(offset+15, src);
109
    register vector unsigned char mask = vec_lvsl(offset, src);
110
    return vec_perm(first, second, mask);
111
}
112

    
113
#endif /* DSPUTIL_ALTIVEC_H */