| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | Copyright 2025 Joel Svensson svenssonjoel@yahoo.se | ||
| 3 | |||
| 4 | This program is free software: you can redistribute it and/or modify | ||
| 5 | it under the terms of the GNU General Public License as published by | ||
| 6 | the Free Software Foundation, either version 3 of the License, or | ||
| 7 | (at your option) any later version. | ||
| 8 | |||
| 9 | This program 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 | ||
| 12 | GNU General Public License for more details. | ||
| 13 | |||
| 14 | You should have received a copy of the GNU General Public License | ||
| 15 | along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
| 16 | */ | ||
| 17 | |||
| 18 | #include "extensions.h" | ||
| 19 | #include "lbm_utils.h" | ||
| 20 | |||
| 21 | #include <math.h> | ||
| 22 | #include <string.h> | ||
| 23 | #include <stdio.h> | ||
| 24 | |||
| 25 | #ifdef LBM_OPT_DSP_EXTENSIONS_SIZE | ||
| 26 | #pragma GCC optimize ("-Os") | ||
| 27 | #endif | ||
| 28 | #ifdef LBM_OPT_DSP_EXTENSIONS_SIZE_AGGRESSIVE | ||
| 29 | #pragma GCC optimize ("-Oz") | ||
| 30 | #endif | ||
| 31 | |||
| 32 | #if defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | ||
| 33 | #define LBM_SYSTEM_LITTLE_ENDIAN true | ||
| 34 | #else | ||
| 35 | #define LBM_SYSTEM_LITTLE_ENDIAN false | ||
| 36 | #endif | ||
| 37 | |||
| 38 | |||
| 39 | static lbm_uint sym_inverse = 0; | ||
| 40 | static lbm_uint sym_big_endian = 0; | ||
| 41 | static lbm_uint sym_little_endian = 0; | ||
| 42 | |||
| 43 | ✗ | static inline uint32_t byte_order_swap(uint32_t w) { | |
| 44 | ✗ | uint32_t r = w; | |
| 45 | ✗ | uint8_t *bytes = (uint8_t*)&r; | |
| 46 | ✗ | uint8_t tmp = bytes[0]; | |
| 47 | ✗ | bytes[0] = bytes[3]; | |
| 48 | ✗ | bytes[3] = tmp; | |
| 49 | ✗ | tmp = bytes[1]; | |
| 50 | ✗ | bytes[1] = bytes[2]; | |
| 51 | ✗ | bytes[2] = tmp; | |
| 52 | |||
| 53 | ✗ | return r; | |
| 54 | } | ||
| 55 | |||
| 56 | |||
| 57 | ✗ | void lbm_fft(float *real, float *imag, int n, int inverse) { | |
| 58 | ✗ | int k = 0; | |
| 59 | ✗ | for (int i = 1; i < n; i++) { | |
| 60 | ✗ | int bit = n >> 1; | |
| 61 | ✗ | while (k & bit) { | |
| 62 | ✗ | k ^= bit; | |
| 63 | ✗ | bit >>= 1; | |
| 64 | } | ||
| 65 | ✗ | k ^= bit; | |
| 66 | |||
| 67 | ✗ | if (i < k) { | |
| 68 | ✗ | float temp = real[i]; | |
| 69 | ✗ | real[i] = real[k]; | |
| 70 | ✗ | real[k] = temp; | |
| 71 | |||
| 72 | ✗ | temp = imag[i]; | |
| 73 | ✗ | imag[i] = imag[k]; | |
| 74 | ✗ | imag[k] = temp; | |
| 75 | } | ||
| 76 | } | ||
| 77 | |||
| 78 | ✗ | for (int len = 2; len <= n; len <<= 1) { | |
| 79 | ✗ | float angle = (inverse ? 2.0f : -2.0f) * (float)M_PI / (float)len; | |
| 80 | ✗ | float wlen_r = cosf(angle); | |
| 81 | ✗ | float wlen_i = sinf(angle); | |
| 82 | |||
| 83 | ✗ | for (int i = 0; i < n; i += len) { | |
| 84 | ✗ | float w_r = 1.0f; | |
| 85 | ✗ | float w_i = 0.0f; | |
| 86 | |||
| 87 | ✗ | for (int j = 0; j < len / 2; j++) { | |
| 88 | ✗ | float u_r = real[i + j]; | |
| 89 | ✗ | float u_i = imag[i + j]; | |
| 90 | ✗ | float v_r = real[i + j + len / 2]; | |
| 91 | ✗ | float v_i = imag[i + j + len / 2]; | |
| 92 | |||
| 93 | ✗ | float t_r = w_r * v_r - w_i * v_i; | |
| 94 | ✗ | float t_i = w_r * v_i + w_i * v_r; | |
| 95 | |||
| 96 | ✗ | real[i + j] = u_r + t_r; | |
| 97 | ✗ | imag[i + j] = u_i + t_i; | |
| 98 | ✗ | real[i + j + len / 2] = u_r - t_r; | |
| 99 | ✗ | imag[i + j + len / 2] = u_i - t_i; | |
| 100 | |||
| 101 | ✗ | float next_w_r = w_r * wlen_r - w_i * wlen_i; | |
| 102 | ✗ | float next_w_i = w_r * wlen_i + w_i * wlen_r; | |
| 103 | ✗ | w_r = next_w_r; | |
| 104 | ✗ | w_i = next_w_i; | |
| 105 | } | ||
| 106 | } | ||
| 107 | } | ||
| 108 | |||
| 109 | ✗ | if (inverse) { | |
| 110 | ✗ | for (int i = 0; i < n; i++) { | |
| 111 | ✗ | real[i] /= (float)n; | |
| 112 | ✗ | imag[i] /= (float)n; | |
| 113 | } | ||
| 114 | } | ||
| 115 | ✗ | } | |
| 116 | |||
| 117 | ✗ | static lbm_value ext_fft_f32(lbm_value *args, lbm_uint argn) { | |
| 118 | ✗ | lbm_value r = ENC_SYM_TERROR; | |
| 119 | ✗ | if (argn >= 2 && | |
| 120 | ✗ | lbm_is_array_r(args[0]) && | |
| 121 | ✗ | lbm_is_array_r(args[1])) { | |
| 122 | |||
| 123 | ✗ | int inverse = 0; | |
| 124 | ✗ | bool be = true; | |
| 125 | |||
| 126 | ✗ | if (argn > 2 && | |
| 127 | ✗ | lbm_is_symbol(args[2])) { | |
| 128 | ✗ | lbm_uint sym = lbm_dec_sym(args[2]); | |
| 129 | ✗ | if (sym == sym_inverse) { | |
| 130 | ✗ | inverse = 1; | |
| 131 | ✗ | } else if (sym == sym_little_endian) { | |
| 132 | ✗ | be = false; | |
| 133 | } | ||
| 134 | } | ||
| 135 | |||
| 136 | ✗ | if (argn > 3 && | |
| 137 | ✗ | lbm_is_symbol(args[3])) { | |
| 138 | ✗ | if (lbm_dec_sym(args[3]) == sym_little_endian) { | |
| 139 | ✗ | be = false; | |
| 140 | } | ||
| 141 | } | ||
| 142 | |||
| 143 | ✗ | bool swap_byte_order = | |
| 144 | (be && LBM_SYSTEM_LITTLE_ENDIAN) || | ||
| 145 | (!be && !LBM_SYSTEM_LITTLE_ENDIAN); | ||
| 146 | |||
| 147 | ✗ | lbm_array_header_t *real_arr = lbm_dec_array_r(args[0]); | |
| 148 | ✗ | lbm_array_header_t *imag_arr = lbm_dec_array_r(args[1]); | |
| 149 | |||
| 150 | ✗ | if (!real_arr || !imag_arr) { | |
| 151 | ✗ | return ENC_SYM_TERROR; | |
| 152 | } | ||
| 153 | |||
| 154 | ✗ | if (real_arr->size == imag_arr->size && | |
| 155 | ✗ | real_arr->size > sizeof(float) && // need more than 4 bytes for a float | |
| 156 | ✗ | (real_arr->size % sizeof(float) == 0)) { | |
| 157 | ✗ | lbm_uint n = real_arr->size / sizeof(float); | |
| 158 | ✗ | lbm_uint padded_size = n; | |
| 159 | ✗ | if ((n & (n - 1)) != 0) { // not a power of two. needs padding. | |
| 160 | ✗ | padded_size = 1; | |
| 161 | ✗ | while (padded_size < n) padded_size <<= 1; | |
| 162 | } | ||
| 163 | |||
| 164 | lbm_value real_copy; | ||
| 165 | lbm_value imag_copy; | ||
| 166 | ✗ | if (lbm_heap_allocate_array(&real_copy, padded_size * sizeof(float)) && | |
| 167 | ✗ | lbm_heap_allocate_array(&imag_copy, padded_size * sizeof(float))) { | |
| 168 | // cppcheck-suppress invalidPointerCast | ||
| 169 | ✗ | float *real_data = (float*)real_arr->data; | |
| 170 | // cppcheck-suppress invalidPointerCast | ||
| 171 | ✗ | float *imag_data = (float*)imag_arr->data; | |
| 172 | |||
| 173 | ✗ | lbm_array_header_t *real_copy_arr = lbm_dec_array_r(real_copy); | |
| 174 | ✗ | lbm_array_header_t *imag_copy_arr = lbm_dec_array_r(imag_copy); | |
| 175 | // cppcheck-suppress invalidPointerCast | ||
| 176 | ✗ | float *real_copy_data = (float*)real_copy_arr->data; | |
| 177 | // cppcheck-suppress invalidPointerCast | ||
| 178 | ✗ | float *imag_copy_data = (float*)imag_copy_arr->data; | |
| 179 | ✗ | memset(real_copy_data, 0, padded_size * sizeof(float)); | |
| 180 | ✗ | memset(imag_copy_data, 0, padded_size * sizeof(float)); | |
| 181 | |||
| 182 | ✗ | if (swap_byte_order) { | |
| 183 | ✗ | for (lbm_uint w = 0; w < n; w ++) { | |
| 184 | // cppcheck-suppress invalidPointerCast | ||
| 185 | ✗ | uint32_t swapped = byte_order_swap(((uint32_t*)real_data)[w]); | |
| 186 | // cppcheck-suppress invalidPointerCast | ||
| 187 | ✗ | ((uint32_t*)real_copy_data)[w] = swapped; | |
| 188 | // cppcheck-suppress invalidPointerCast | ||
| 189 | ✗ | swapped = byte_order_swap(((uint32_t*)imag_data)[w]); | |
| 190 | // cppcheck-suppress invalidPointerCast | ||
| 191 | ✗ | ((uint32_t*)imag_copy_data)[w] = swapped; | |
| 192 | } | ||
| 193 | } else { | ||
| 194 | ✗ | memcpy(real_copy_data, real_data, real_arr->size); | |
| 195 | ✗ | memcpy(imag_copy_data, imag_data, real_arr->size); | |
| 196 | } | ||
| 197 | |||
| 198 | ✗ | lbm_fft(real_copy_data, imag_copy_data, (int)padded_size, inverse); | |
| 199 | |||
| 200 | ✗ | if (swap_byte_order) { | |
| 201 | ✗ | for (lbm_uint w = 0; w < padded_size; w ++) { | |
| 202 | // cppcheck-suppress invalidPointerCast | ||
| 203 | ✗ | uint32_t swapped = byte_order_swap(((uint32_t*)real_copy_data)[w]); | |
| 204 | // cppcheck-suppress invalidPointerCast | ||
| 205 | ✗ | ((uint32_t*)real_copy_data)[w] = swapped; | |
| 206 | // cppcheck-suppress invalidPointerCast | ||
| 207 | ✗ | swapped = byte_order_swap(((uint32_t*)imag_copy_data)[w]); | |
| 208 | // cppcheck-suppress invalidPointerCast | ||
| 209 | ✗ | ((uint32_t*)imag_copy_data)[w] = swapped; | |
| 210 | } | ||
| 211 | } | ||
| 212 | ✗ | r = lbm_cons(real_copy, imag_copy); | |
| 213 | } else { | ||
| 214 | ✗ | r = ENC_SYM_MERROR; | |
| 215 | } | ||
| 216 | } else { | ||
| 217 | ✗ | r = ENC_SYM_NIL; | |
| 218 | } | ||
| 219 | } | ||
| 220 | ✗ | return r; | |
| 221 | } | ||
| 222 | |||
| 223 | 281 | void lbm_dsp_extensions_init(void) { | |
| 224 | 281 | lbm_add_symbol("inverse", &sym_inverse); | |
| 225 | 281 | lbm_add_symbol("little-endian", &sym_little_endian); | |
| 226 | 281 | lbm_add_symbol("big-endian", &sym_big_endian); | |
| 227 | |||
| 228 | 281 | lbm_add_extension("fft", ext_fft_f32); | |
| 229 | 281 | } | |
| 230 |