ppc: vsx: Implement diff_pixels and get_pixels
[libav.git] / libavutil / ppc / util_altivec.h
1 /*
2 * This file is part of Libav.
3 *
4 * Libav is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * Libav is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with Libav; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 /**
20 * @file
21 * Contains misc utility macros and inline functions
22 */
23
24 #ifndef AVUTIL_PPC_UTIL_ALTIVEC_H
25 #define AVUTIL_PPC_UTIL_ALTIVEC_H
26
27 #include <stdint.h>
28
29 #include "config.h"
30
31 #if HAVE_ALTIVEC_H
32 #include <altivec.h>
33 #endif
34
35 #include "types_altivec.h"
36
37 #if HAVE_ALTIVEC
38
39 // used to build registers permutation vectors (vcprm)
40 // the 's' are for words in the _s_econd vector
41 #define WORD_0 0x00,0x01,0x02,0x03
42 #define WORD_1 0x04,0x05,0x06,0x07
43 #define WORD_2 0x08,0x09,0x0a,0x0b
44 #define WORD_3 0x0c,0x0d,0x0e,0x0f
45 #define WORD_s0 0x10,0x11,0x12,0x13
46 #define WORD_s1 0x14,0x15,0x16,0x17
47 #define WORD_s2 0x18,0x19,0x1a,0x1b
48 #define WORD_s3 0x1c,0x1d,0x1e,0x1f
49
50 #define vcprm(a,b,c,d) (const vec_u8){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
51
52
53 // Transpose 8x8 matrix of 16-bit elements (in-place)
54 #define TRANSPOSE8(a,b,c,d,e,f,g,h) \
55 do { \
56 vec_s16 A1, B1, C1, D1, E1, F1, G1, H1; \
57 vec_s16 A2, B2, C2, D2, E2, F2, G2, H2; \
58 \
59 A1 = vec_mergeh (a, e); \
60 B1 = vec_mergel (a, e); \
61 C1 = vec_mergeh (b, f); \
62 D1 = vec_mergel (b, f); \
63 E1 = vec_mergeh (c, g); \
64 F1 = vec_mergel (c, g); \
65 G1 = vec_mergeh (d, h); \
66 H1 = vec_mergel (d, h); \
67 \
68 A2 = vec_mergeh (A1, E1); \
69 B2 = vec_mergel (A1, E1); \
70 C2 = vec_mergeh (B1, F1); \
71 D2 = vec_mergel (B1, F1); \
72 E2 = vec_mergeh (C1, G1); \
73 F2 = vec_mergel (C1, G1); \
74 G2 = vec_mergeh (D1, H1); \
75 H2 = vec_mergel (D1, H1); \
76 \
77 a = vec_mergeh (A2, E2); \
78 b = vec_mergel (A2, E2); \
79 c = vec_mergeh (B2, F2); \
80 d = vec_mergel (B2, F2); \
81 e = vec_mergeh (C2, G2); \
82 f = vec_mergel (C2, G2); \
83 g = vec_mergeh (D2, H2); \
84 h = vec_mergel (D2, H2); \
85 } while (0)
86
87
88 /** @brief loads unaligned vector @a *src with offset @a offset
89 and returns it */
90 static inline vec_u8 unaligned_load(int offset, uint8_t *src)
91 {
92 register vec_u8 first = vec_ld(offset, src);
93 register vec_u8 second = vec_ld(offset + 15, src);
94 register vec_u8 mask = vec_lvsl(offset, src);
95 return vec_perm(first, second, mask);
96 }
97
98 /**
99 * loads vector known misalignment
100 * @param perm_vec the align permute vector to combine the two loads from lvsl
101 */
102 static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec)
103 {
104 vec_u8 a = vec_ld(offset, src);
105 vec_u8 b = vec_ld(offset + 15, src);
106 return vec_perm(a, b, perm_vec);
107 }
108
109 #define vec_unaligned_load(b) \
110 vec_perm(vec_ld(0, b), vec_ld(15, b), vec_lvsl(0, b));
111
112 #endif /* HAVE_ALTIVEC */
113
114 #if HAVE_VSX
115 #if HAVE_BIGENDIAN
116 #define vsx_ld_u8_s16(off, p) \
117 ((vec_s16)vec_mergeh((vec_u8)vec_splat_u8(0), \
118 (vec_u8)vec_vsx_ld((off), (p))))
119 #else
120 #define vsx_ld_u8_s16(off, p) \
121 ((vec_s16)vec_mergeh((vec_u8)vec_vsx_ld((off), (p)), \
122 (vec_u8)vec_splat_u8(0)))
123 #endif /* HAVE_BIGENDIAN */
124 #endif /* HAVE_VSX */
125
126 #endif /* AVUTIL_PPC_UTIL_ALTIVEC_H */