| 1 | /* |
| 2 | * Various fixed-point math operations |
| 3 | * |
| 4 | * Copyright (c) 2008 Vladimir Voroshilov |
| 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 | #include <inttypes.h> |
| 24 | #include <limits.h> |
| 25 | |
| 26 | #include "libavutil/avassert.h" |
| 27 | #include "avcodec.h" |
| 28 | #include "mathops.h" |
| 29 | #include "celp_math.h" |
| 30 | #include "libavutil/common.h" |
| 31 | |
| 32 | static const uint16_t exp2a[]= |
| 33 | { |
| 34 | 0, 1435, 2901, 4400, 5931, 7496, 9096, 10730, |
| 35 | 12400, 14106, 15850, 17632, 19454, 21315, 23216, 25160, |
| 36 | 27146, 29175, 31249, 33368, 35534, 37747, 40009, 42320, |
| 37 | 44682, 47095, 49562, 52082, 54657, 57289, 59979, 62727, |
| 38 | }; |
| 39 | |
| 40 | static const uint16_t exp2b[]= |
| 41 | { |
| 42 | 3, 712, 1424, 2134, 2845, 3557, 4270, 4982, |
| 43 | 5696, 6409, 7124, 7839, 8554, 9270, 9986, 10704, |
| 44 | 11421, 12138, 12857, 13576, 14295, 15014, 15734, 16455, |
| 45 | 17176, 17898, 18620, 19343, 20066, 20790, 21514, 22238, |
| 46 | }; |
| 47 | |
| 48 | int ff_exp2(uint16_t power) |
| 49 | { |
| 50 | unsigned int result= exp2a[power>>10] + 0x10000; |
| 51 | |
| 52 | av_assert2(power <= 0x7fff); |
| 53 | |
| 54 | result= (result<<3) + ((result*exp2b[(power>>5)&31])>>17); |
| 55 | return result + ((result*(power&31)*89)>>22); |
| 56 | } |
| 57 | |
| 58 | /** |
| 59 | * Table used to compute log2(x) |
| 60 | * |
| 61 | * tab_log2[i] = (1<<15) * log2(1 + i/32), i=0..32 |
| 62 | */ |
| 63 | static const uint16_t tab_log2[33] = |
| 64 | { |
| 65 | #ifdef G729_BITEXACT |
| 66 | 0, 1455, 2866, 4236, 5568, 6863, 8124, 9352, |
| 67 | 10549, 11716, 12855, 13967, 15054, 16117, 17156, 18172, |
| 68 | 19167, 20142, 21097, 22033, 22951, 23852, 24735, 25603, |
| 69 | 26455, 27291, 28113, 28922, 29716, 30497, 31266, 32023, 32767, |
| 70 | #else |
| 71 | 4, 1459, 2870, 4240, 5572, 6867, 8127, 9355, |
| 72 | 10552, 11719, 12858, 13971, 15057, 16120, 17158, 18175, |
| 73 | 19170, 20145, 21100, 22036, 22954, 23854, 24738, 25605, |
| 74 | 26457, 27294, 28116, 28924, 29719, 30500, 31269, 32025, 32769, |
| 75 | #endif |
| 76 | }; |
| 77 | |
| 78 | int ff_log2_q15(uint32_t value) |
| 79 | { |
| 80 | uint8_t power_int; |
| 81 | uint8_t frac_x0; |
| 82 | uint16_t frac_dx; |
| 83 | |
| 84 | // Stripping zeros from beginning |
| 85 | power_int = av_log2(value); |
| 86 | value <<= (31 - power_int); |
| 87 | |
| 88 | // b31 is always non-zero now |
| 89 | frac_x0 = (value & 0x7c000000) >> 26; // b26-b31 and [32..63] -> [0..31] |
| 90 | frac_dx = (value & 0x03fff800) >> 11; |
| 91 | |
| 92 | value = tab_log2[frac_x0]; |
| 93 | value += (frac_dx * (tab_log2[frac_x0+1] - tab_log2[frac_x0])) >> 15; |
| 94 | |
| 95 | return (power_int << 15) + value; |
| 96 | } |
| 97 | |
| 98 | int64_t ff_dot_product(const int16_t *a, const int16_t *b, int length) |
| 99 | { |
| 100 | int i; |
| 101 | int64_t sum = 0; |
| 102 | |
| 103 | for (i = 0; i < length; i++) |
| 104 | sum += MUL16(a[i], b[i]); |
| 105 | |
| 106 | return sum; |
| 107 | } |
| 108 | |
| 109 | float ff_dot_productf(const float* a, const float* b, int length) |
| 110 | { |
| 111 | float sum = 0; |
| 112 | int i; |
| 113 | |
| 114 | for(i=0; i<length; i++) |
| 115 | sum += a[i] * b[i]; |
| 116 | |
| 117 | return sum; |
| 118 | } |
| 119 | |
| 120 | void ff_celp_math_init(CELPMContext *c) |
| 121 | { |
| 122 | c->dot_productf = ff_dot_productf; |
| 123 | |
| 124 | if(HAVE_MIPSFPU) |
| 125 | ff_celp_math_init_mips(c); |
| 126 | } |