| 1 | /* |
| 2 | * This file is part of FFmpeg. |
| 3 | * |
| 4 | * FFmpeg 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 | * FFmpeg 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 FFmpeg; 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 | #define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d} |
| 50 | |
| 51 | #define SWP_W2S0 0x02,0x03,0x00,0x01 |
| 52 | #define SWP_W2S1 0x06,0x07,0x04,0x05 |
| 53 | #define SWP_W2S2 0x0a,0x0b,0x08,0x09 |
| 54 | #define SWP_W2S3 0x0e,0x0f,0x0c,0x0d |
| 55 | #define SWP_W2Ss0 0x12,0x13,0x10,0x11 |
| 56 | #define SWP_W2Ss1 0x16,0x17,0x14,0x15 |
| 57 | #define SWP_W2Ss2 0x1a,0x1b,0x18,0x19 |
| 58 | #define SWP_W2Ss3 0x1e,0x1f,0x1c,0x1d |
| 59 | #define vcswapi2s(a,b,c,d) (const vector unsigned char){SWP_W2S ## a, SWP_W2S ## b, SWP_W2S ## c, SWP_W2S ## d} |
| 60 | |
| 61 | #define vcswapc() \ |
| 62 | (const vector unsigned char){0x0f,0x0e,0x0d,0x0c,0x0b,0x0a,0x09,0x08,0x07,0x06,0x05,0x04,0x03,0x02,0x01,0x00} |
| 63 | |
| 64 | |
| 65 | // Transpose 8x8 matrix of 16-bit elements (in-place) |
| 66 | #define TRANSPOSE8(a,b,c,d,e,f,g,h) \ |
| 67 | do { \ |
| 68 | vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \ |
| 69 | vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \ |
| 70 | \ |
| 71 | A1 = vec_mergeh (a, e); \ |
| 72 | B1 = vec_mergel (a, e); \ |
| 73 | C1 = vec_mergeh (b, f); \ |
| 74 | D1 = vec_mergel (b, f); \ |
| 75 | E1 = vec_mergeh (c, g); \ |
| 76 | F1 = vec_mergel (c, g); \ |
| 77 | G1 = vec_mergeh (d, h); \ |
| 78 | H1 = vec_mergel (d, h); \ |
| 79 | \ |
| 80 | A2 = vec_mergeh (A1, E1); \ |
| 81 | B2 = vec_mergel (A1, E1); \ |
| 82 | C2 = vec_mergeh (B1, F1); \ |
| 83 | D2 = vec_mergel (B1, F1); \ |
| 84 | E2 = vec_mergeh (C1, G1); \ |
| 85 | F2 = vec_mergel (C1, G1); \ |
| 86 | G2 = vec_mergeh (D1, H1); \ |
| 87 | H2 = vec_mergel (D1, H1); \ |
| 88 | \ |
| 89 | a = vec_mergeh (A2, E2); \ |
| 90 | b = vec_mergel (A2, E2); \ |
| 91 | c = vec_mergeh (B2, F2); \ |
| 92 | d = vec_mergel (B2, F2); \ |
| 93 | e = vec_mergeh (C2, G2); \ |
| 94 | f = vec_mergel (C2, G2); \ |
| 95 | g = vec_mergeh (D2, H2); \ |
| 96 | h = vec_mergel (D2, H2); \ |
| 97 | } while (0) |
| 98 | |
| 99 | |
| 100 | #if HAVE_BIGENDIAN |
| 101 | #define VEC_LD(offset,b) \ |
| 102 | vec_perm(vec_ld(offset, b), vec_ld((offset)+15, b), vec_lvsl(offset, b)) |
| 103 | #else |
| 104 | #define VEC_LD(offset,b) \ |
| 105 | vec_vsx_ld(offset, b) |
| 106 | #endif |
| 107 | |
| 108 | /** @brief loads unaligned vector @a *src with offset @a offset |
| 109 | and returns it */ |
| 110 | #if HAVE_BIGENDIAN |
| 111 | static inline vector unsigned char unaligned_load(int offset, const uint8_t *src) |
| 112 | { |
| 113 | register vector unsigned char first = vec_ld(offset, src); |
| 114 | register vector unsigned char second = vec_ld(offset+15, src); |
| 115 | register vector unsigned char mask = vec_lvsl(offset, src); |
| 116 | return vec_perm(first, second, mask); |
| 117 | } |
| 118 | static inline vec_u8 load_with_perm_vec(int offset, const uint8_t *src, vec_u8 perm_vec) |
| 119 | { |
| 120 | vec_u8 a = vec_ld(offset, src); |
| 121 | vec_u8 b = vec_ld(offset+15, src); |
| 122 | return vec_perm(a, b, perm_vec); |
| 123 | } |
| 124 | #else |
| 125 | #define unaligned_load(a,b) VEC_LD(a,b) |
| 126 | #define load_with_perm_vec(a,b,c) VEC_LD(a,b) |
| 127 | #endif |
| 128 | |
| 129 | |
| 130 | /** |
| 131 | * loads vector known misalignment |
| 132 | * @param perm_vec the align permute vector to combine the two loads from lvsl |
| 133 | */ |
| 134 | |
| 135 | #define vec_unaligned_load(b) VEC_LD(0, b) |
| 136 | |
| 137 | #if HAVE_BIGENDIAN |
| 138 | #define VEC_MERGEH(a, b) vec_mergeh(a, b) |
| 139 | #define VEC_MERGEL(a, b) vec_mergel(a, b) |
| 140 | #else |
| 141 | #define VEC_MERGEH(a, b) vec_mergeh(b, a) |
| 142 | #define VEC_MERGEL(a, b) vec_mergel(b, a) |
| 143 | #endif |
| 144 | |
| 145 | #if HAVE_BIGENDIAN |
| 146 | #define VEC_ST(a,b,c) vec_st(a,b,c) |
| 147 | #else |
| 148 | #define VEC_ST(a,b,c) vec_vsx_st(a,b,c) |
| 149 | #endif |
| 150 | |
| 151 | #if HAVE_BIGENDIAN |
| 152 | #define VEC_SPLAT16(a,b) vec_splat((vec_s16)(a), b) |
| 153 | #else |
| 154 | #define VEC_SPLAT16(a,b) vec_splat((vec_s16)(vec_perm(a, a, vcswapi2s(0,1,2,3))), b) |
| 155 | #endif |
| 156 | |
| 157 | #if HAVE_BIGENDIAN |
| 158 | #define VEC_SLD16(a,b,c) vec_sld(a, b, c) |
| 159 | #else |
| 160 | #define VEC_SLD16(a,b,c) vec_sld(b, a, c) |
| 161 | #endif |
| 162 | |
| 163 | #endif /* HAVE_ALTIVEC */ |
| 164 | |
| 165 | #endif /* AVUTIL_PPC_UTIL_ALTIVEC_H */ |