| 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 | #ifndef AVUTIL_FLOAT_DSP_H |
| 20 | #define AVUTIL_FLOAT_DSP_H |
| 21 | |
| 22 | #include "config.h" |
| 23 | |
| 24 | typedef struct AVFloatDSPContext { |
| 25 | /** |
| 26 | * Calculate the product of two vectors of floats and store the result in |
| 27 | * a vector of floats. |
| 28 | * |
| 29 | * @param dst output vector |
| 30 | * constraints: 32-byte aligned |
| 31 | * @param src0 first input vector |
| 32 | * constraints: 32-byte aligned |
| 33 | * @param src1 second input vector |
| 34 | * constraints: 32-byte aligned |
| 35 | * @param len number of elements in the input |
| 36 | * constraints: multiple of 16 |
| 37 | */ |
| 38 | void (*vector_fmul)(float *dst, const float *src0, const float *src1, |
| 39 | int len); |
| 40 | |
| 41 | /** |
| 42 | * Multiply a vector of floats by a scalar float and add to |
| 43 | * destination vector. Source and destination vectors must |
| 44 | * overlap exactly or not at all. |
| 45 | * |
| 46 | * @param dst result vector |
| 47 | * constraints: 32-byte aligned |
| 48 | * @param src input vector |
| 49 | * constraints: 32-byte aligned |
| 50 | * @param mul scalar value |
| 51 | * @param len length of vector |
| 52 | * constraints: multiple of 16 |
| 53 | */ |
| 54 | void (*vector_fmac_scalar)(float *dst, const float *src, float mul, |
| 55 | int len); |
| 56 | |
| 57 | /** |
| 58 | * Multiply a vector of floats by a scalar float. Source and |
| 59 | * destination vectors must overlap exactly or not at all. |
| 60 | * |
| 61 | * @param dst result vector |
| 62 | * constraints: 16-byte aligned |
| 63 | * @param src input vector |
| 64 | * constraints: 16-byte aligned |
| 65 | * @param mul scalar value |
| 66 | * @param len length of vector |
| 67 | * constraints: multiple of 4 |
| 68 | */ |
| 69 | void (*vector_fmul_scalar)(float *dst, const float *src, float mul, |
| 70 | int len); |
| 71 | |
| 72 | /** |
| 73 | * Multiply a vector of double by a scalar double. Source and |
| 74 | * destination vectors must overlap exactly or not at all. |
| 75 | * |
| 76 | * @param dst result vector |
| 77 | * constraints: 32-byte aligned |
| 78 | * @param src input vector |
| 79 | * constraints: 32-byte aligned |
| 80 | * @param mul scalar value |
| 81 | * @param len length of vector |
| 82 | * constraints: multiple of 8 |
| 83 | */ |
| 84 | void (*vector_dmul_scalar)(double *dst, const double *src, double mul, |
| 85 | int len); |
| 86 | |
| 87 | /** |
| 88 | * Overlap/add with window function. |
| 89 | * Used primarily by MDCT-based audio codecs. |
| 90 | * Source and destination vectors must overlap exactly or not at all. |
| 91 | * |
| 92 | * @param dst result vector |
| 93 | * constraints: 16-byte aligned |
| 94 | * @param src0 first source vector |
| 95 | * constraints: 16-byte aligned |
| 96 | * @param src1 second source vector |
| 97 | * constraints: 16-byte aligned |
| 98 | * @param win half-window vector |
| 99 | * constraints: 16-byte aligned |
| 100 | * @param len length of vector |
| 101 | * constraints: multiple of 4 |
| 102 | */ |
| 103 | void (*vector_fmul_window)(float *dst, const float *src0, |
| 104 | const float *src1, const float *win, int len); |
| 105 | |
| 106 | /** |
| 107 | * Calculate the product of two vectors of floats, add a third vector of |
| 108 | * floats and store the result in a vector of floats. |
| 109 | * |
| 110 | * @param dst output vector |
| 111 | * constraints: 32-byte aligned |
| 112 | * @param src0 first input vector |
| 113 | * constraints: 32-byte aligned |
| 114 | * @param src1 second input vector |
| 115 | * constraints: 32-byte aligned |
| 116 | * @param src2 third input vector |
| 117 | * constraints: 32-byte aligned |
| 118 | * @param len number of elements in the input |
| 119 | * constraints: multiple of 16 |
| 120 | */ |
| 121 | void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, |
| 122 | const float *src2, int len); |
| 123 | |
| 124 | /** |
| 125 | * Calculate the product of two vectors of floats, and store the result |
| 126 | * in a vector of floats. The second vector of floats is iterated over |
| 127 | * in reverse order. |
| 128 | * |
| 129 | * @param dst output vector |
| 130 | * constraints: 32-byte aligned |
| 131 | * @param src0 first input vector |
| 132 | * constraints: 32-byte aligned |
| 133 | * @param src1 second input vector |
| 134 | * constraints: 32-byte aligned |
| 135 | * @param len number of elements in the input |
| 136 | * constraints: multiple of 16 |
| 137 | */ |
| 138 | void (*vector_fmul_reverse)(float *dst, const float *src0, |
| 139 | const float *src1, int len); |
| 140 | |
| 141 | /** |
| 142 | * Calculate the sum and difference of two vectors of floats. |
| 143 | * |
| 144 | * @param v1 first input vector, sum output, 16-byte aligned |
| 145 | * @param v2 second input vector, difference output, 16-byte aligned |
| 146 | * @param len length of vectors, multiple of 4 |
| 147 | */ |
| 148 | void (*butterflies_float)(float *av_restrict v1, float *av_restrict v2, int len); |
| 149 | |
| 150 | /** |
| 151 | * Calculate the scalar product of two vectors of floats. |
| 152 | * |
| 153 | * @param v1 first vector, 16-byte aligned |
| 154 | * @param v2 second vector, 16-byte aligned |
| 155 | * @param len length of vectors, multiple of 4 |
| 156 | * |
| 157 | * @return sum of elementwise products |
| 158 | */ |
| 159 | float (*scalarproduct_float)(const float *v1, const float *v2, int len); |
| 160 | } AVFloatDSPContext; |
| 161 | |
| 162 | /** |
| 163 | * Return the scalar product of two vectors. |
| 164 | * |
| 165 | * @param v1 first input vector |
| 166 | * @param v2 first input vector |
| 167 | * @param len number of elements |
| 168 | * |
| 169 | * @return sum of elementwise products |
| 170 | */ |
| 171 | float avpriv_scalarproduct_float_c(const float *v1, const float *v2, int len); |
| 172 | |
| 173 | /** |
| 174 | * Initialize a float DSP context. |
| 175 | * |
| 176 | * @param fdsp float DSP context |
| 177 | * @param strict setting to non-zero avoids using functions which may not be IEEE-754 compliant |
| 178 | */ |
| 179 | void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int strict); |
| 180 | |
| 181 | |
| 182 | void ff_float_dsp_init_aarch64(AVFloatDSPContext *fdsp); |
| 183 | void ff_float_dsp_init_arm(AVFloatDSPContext *fdsp); |
| 184 | void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int strict); |
| 185 | void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp); |
| 186 | void ff_float_dsp_init_mips(AVFloatDSPContext *fdsp); |
| 187 | |
| 188 | /** |
| 189 | * Allocate a float DSP context. |
| 190 | * |
| 191 | * @param strict setting to non-zero avoids using functions which may not be IEEE-754 compliant |
| 192 | */ |
| 193 | AVFloatDSPContext *avpriv_float_dsp_alloc(int strict); |
| 194 | |
| 195 | #endif /* AVUTIL_FLOAT_DSP_H */ |