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 AVCODEC_PPC_UTIL_ALTIVEC_H
25#define AVCODEC_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// used to build registers permutation vectors (vcprm)
38// the 's' are for words in the _s_econd vector
39#define WORD_0 0x00,0x01,0x02,0x03
40#define WORD_1 0x04,0x05,0x06,0x07
41#define WORD_2 0x08,0x09,0x0a,0x0b
42#define WORD_3 0x0c,0x0d,0x0e,0x0f
43#define WORD_s0 0x10,0x11,0x12,0x13
44#define WORD_s1 0x14,0x15,0x16,0x17
45#define WORD_s2 0x18,0x19,0x1a,0x1b
46#define WORD_s3 0x1c,0x1d,0x1e,0x1f
47
48#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
49#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
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// Transpose 8x8 matrix of 16-bit elements (in-place)
63#define TRANSPOSE8(a,b,c,d,e,f,g,h) \
64do { \
65    vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \
66    vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \
67 \
68    A1 = vec_mergeh (a, e); \
69    B1 = vec_mergel (a, e); \
70    C1 = vec_mergeh (b, f); \
71    D1 = vec_mergel (b, f); \
72    E1 = vec_mergeh (c, g); \
73    F1 = vec_mergel (c, g); \
74    G1 = vec_mergeh (d, h); \
75    H1 = vec_mergel (d, h); \
76 \
77    A2 = vec_mergeh (A1, E1); \
78    B2 = vec_mergel (A1, E1); \
79    C2 = vec_mergeh (B1, F1); \
80    D2 = vec_mergel (B1, F1); \
81    E2 = vec_mergeh (C1, G1); \
82    F2 = vec_mergel (C1, G1); \
83    G2 = vec_mergeh (D1, H1); \
84    H2 = vec_mergel (D1, H1); \
85 \
86    a = vec_mergeh (A2, E2); \
87    b = vec_mergel (A2, E2); \
88    c = vec_mergeh (B2, F2); \
89    d = vec_mergel (B2, F2); \
90    e = vec_mergeh (C2, G2); \
91    f = vec_mergel (C2, G2); \
92    g = vec_mergeh (D2, H2); \
93    h = vec_mergel (D2, H2); \
94} while (0)
95
96
97/** @brief loads unaligned vector @a *src with offset @a offset
98    and returns it */
99static inline vector unsigned char unaligned_load(int offset, uint8_t *src)
100{
101    register vector unsigned char first = vec_ld(offset, src);
102    register vector unsigned char second = vec_ld(offset+15, src);
103    register vector unsigned char mask = vec_lvsl(offset, src);
104    return vec_perm(first, second, mask);
105}
106
107/**
108 * loads vector known misalignment
109 * @param perm_vec the align permute vector to combine the two loads from lvsl
110 */
111static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec)
112{
113    vec_u8 a = vec_ld(offset, src);
114    vec_u8 b = vec_ld(offset+15, src);
115    return vec_perm(a, b, perm_vec);
116}
117
118#endif /* AVCODEC_PPC_UTIL_ALTIVEC_H */
119