| 1 | /* |
| 2 | * Copyright (c) 2012 |
| 3 | * MIPS Technologies, Inc., California. |
| 4 | * |
| 5 | * Redistribution and use in source and binary forms, with or without |
| 6 | * modification, are permitted provided that the following conditions |
| 7 | * are met: |
| 8 | * 1. Redistributions of source code must retain the above copyright |
| 9 | * notice, this list of conditions and the following disclaimer. |
| 10 | * 2. Redistributions in binary form must reproduce the above copyright |
| 11 | * notice, this list of conditions and the following disclaimer in the |
| 12 | * documentation and/or other materials provided with the distribution. |
| 13 | * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its |
| 14 | * contributors may be used to endorse or promote products derived from |
| 15 | * this software without specific prior written permission. |
| 16 | * |
| 17 | * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND |
| 18 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 19 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 20 | * ARE DISCLAIMED. IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE |
| 21 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 22 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 23 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 24 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 25 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 26 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 27 | * SUCH DAMAGE. |
| 28 | * |
| 29 | * Authors: Djordje Pesut (djordje@mips.com) |
| 30 | * Mirjana Vulin (mvulin@mips.com) |
| 31 | * |
| 32 | * This file is part of FFmpeg. |
| 33 | * |
| 34 | * FFmpeg is free software; you can redistribute it and/or |
| 35 | * modify it under the terms of the GNU Lesser General Public |
| 36 | * License as published by the Free Software Foundation; either |
| 37 | * version 2.1 of the License, or (at your option) any later version. |
| 38 | * |
| 39 | * FFmpeg is distributed in the hope that it will be useful, |
| 40 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 41 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 42 | * Lesser General Public License for more details. |
| 43 | * |
| 44 | * You should have received a copy of the GNU Lesser General Public |
| 45 | * License along with FFmpeg; if not, write to the Free Software |
| 46 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 47 | */ |
| 48 | |
| 49 | /** |
| 50 | * @file |
| 51 | * Reference: libavcodec/aacsbr.c |
| 52 | */ |
| 53 | |
| 54 | #include "libavcodec/aac.h" |
| 55 | #include "libavcodec/aacsbr.h" |
| 56 | |
| 57 | #define ENVELOPE_ADJUSTMENT_OFFSET 2 |
| 58 | |
| 59 | #if HAVE_INLINE_ASM |
| 60 | static int sbr_lf_gen_mips(AACContext *ac, SpectralBandReplication *sbr, |
| 61 | float X_low[32][40][2], const float W[2][32][32][2], |
| 62 | int buf_idx) |
| 63 | { |
| 64 | int i, k; |
| 65 | int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; |
| 66 | float *p_x_low = &X_low[0][8][0]; |
| 67 | float *p_w = (float*)&W[buf_idx][0][0][0]; |
| 68 | float *p_x1_low = &X_low[0][0][0]; |
| 69 | float *p_w1 = (float*)&W[1-buf_idx][24][0][0]; |
| 70 | |
| 71 | float *loop_end=p_x1_low + 2560; |
| 72 | |
| 73 | /* loop unrolled 8 times */ |
| 74 | __asm__ volatile ( |
| 75 | "1: \n\t" |
| 76 | "sw $0, 0(%[p_x1_low]) \n\t" |
| 77 | "sw $0, 4(%[p_x1_low]) \n\t" |
| 78 | "sw $0, 8(%[p_x1_low]) \n\t" |
| 79 | "sw $0, 12(%[p_x1_low]) \n\t" |
| 80 | "sw $0, 16(%[p_x1_low]) \n\t" |
| 81 | "sw $0, 20(%[p_x1_low]) \n\t" |
| 82 | "sw $0, 24(%[p_x1_low]) \n\t" |
| 83 | "sw $0, 28(%[p_x1_low]) \n\t" |
| 84 | "addiu %[p_x1_low], %[p_x1_low], 32 \n\t" |
| 85 | "bne %[p_x1_low], %[loop_end], 1b \n\t" |
| 86 | "addiu %[p_x1_low], %[p_x1_low], -10240 \n\t" |
| 87 | |
| 88 | : [p_x1_low]"+r"(p_x1_low) |
| 89 | : [loop_end]"r"(loop_end) |
| 90 | : "memory" |
| 91 | ); |
| 92 | |
| 93 | for (k = 0; k < sbr->kx[1]; k++) { |
| 94 | for (i = 0; i < 32; i+=4) { |
| 95 | /* loop unrolled 4 times */ |
| 96 | __asm__ volatile ( |
| 97 | "lw %[temp0], 0(%[p_w]) \n\t" |
| 98 | "lw %[temp1], 4(%[p_w]) \n\t" |
| 99 | "lw %[temp2], 256(%[p_w]) \n\t" |
| 100 | "lw %[temp3], 260(%[p_w]) \n\t" |
| 101 | "lw %[temp4], 512(%[p_w]) \n\t" |
| 102 | "lw %[temp5], 516(%[p_w]) \n\t" |
| 103 | "lw %[temp6], 768(%[p_w]) \n\t" |
| 104 | "lw %[temp7], 772(%[p_w]) \n\t" |
| 105 | "sw %[temp0], 0(%[p_x_low]) \n\t" |
| 106 | "sw %[temp1], 4(%[p_x_low]) \n\t" |
| 107 | "sw %[temp2], 8(%[p_x_low]) \n\t" |
| 108 | "sw %[temp3], 12(%[p_x_low]) \n\t" |
| 109 | "sw %[temp4], 16(%[p_x_low]) \n\t" |
| 110 | "sw %[temp5], 20(%[p_x_low]) \n\t" |
| 111 | "sw %[temp6], 24(%[p_x_low]) \n\t" |
| 112 | "sw %[temp7], 28(%[p_x_low]) \n\t" |
| 113 | "addiu %[p_x_low], %[p_x_low], 32 \n\t" |
| 114 | "addiu %[p_w], %[p_w], 1024 \n\t" |
| 115 | |
| 116 | : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 117 | [temp2]"=&r"(temp2), [temp3]"=&r"(temp3), |
| 118 | [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 119 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7), |
| 120 | [p_w]"+r"(p_w), [p_x_low]"+r"(p_x_low) |
| 121 | : |
| 122 | : "memory" |
| 123 | ); |
| 124 | } |
| 125 | p_x_low += 16; |
| 126 | p_w -= 2046; |
| 127 | } |
| 128 | |
| 129 | for (k = 0; k < sbr->kx[0]; k++) { |
| 130 | for (i = 0; i < 2; i++) { |
| 131 | |
| 132 | /* loop unrolled 4 times */ |
| 133 | __asm__ volatile ( |
| 134 | "lw %[temp0], 0(%[p_w1]) \n\t" |
| 135 | "lw %[temp1], 4(%[p_w1]) \n\t" |
| 136 | "lw %[temp2], 256(%[p_w1]) \n\t" |
| 137 | "lw %[temp3], 260(%[p_w1]) \n\t" |
| 138 | "lw %[temp4], 512(%[p_w1]) \n\t" |
| 139 | "lw %[temp5], 516(%[p_w1]) \n\t" |
| 140 | "lw %[temp6], 768(%[p_w1]) \n\t" |
| 141 | "lw %[temp7], 772(%[p_w1]) \n\t" |
| 142 | "sw %[temp0], 0(%[p_x1_low]) \n\t" |
| 143 | "sw %[temp1], 4(%[p_x1_low]) \n\t" |
| 144 | "sw %[temp2], 8(%[p_x1_low]) \n\t" |
| 145 | "sw %[temp3], 12(%[p_x1_low]) \n\t" |
| 146 | "sw %[temp4], 16(%[p_x1_low]) \n\t" |
| 147 | "sw %[temp5], 20(%[p_x1_low]) \n\t" |
| 148 | "sw %[temp6], 24(%[p_x1_low]) \n\t" |
| 149 | "sw %[temp7], 28(%[p_x1_low]) \n\t" |
| 150 | "addiu %[p_x1_low], %[p_x1_low], 32 \n\t" |
| 151 | "addiu %[p_w1], %[p_w1], 1024 \n\t" |
| 152 | |
| 153 | : [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 154 | [temp2]"=&r"(temp2), [temp3]"=&r"(temp3), |
| 155 | [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), |
| 156 | [temp6]"=&r"(temp6), [temp7]"=&r"(temp7), |
| 157 | [p_w1]"+r"(p_w1), [p_x1_low]"+r"(p_x1_low) |
| 158 | : |
| 159 | : "memory" |
| 160 | ); |
| 161 | } |
| 162 | p_x1_low += 64; |
| 163 | p_w1 -= 510; |
| 164 | } |
| 165 | return 0; |
| 166 | } |
| 167 | |
| 168 | static int sbr_x_gen_mips(SpectralBandReplication *sbr, float X[2][38][64], |
| 169 | const float Y0[38][64][2], const float Y1[38][64][2], |
| 170 | const float X_low[32][40][2], int ch) |
| 171 | { |
| 172 | int k, i; |
| 173 | const int i_f = 32; |
| 174 | int temp0, temp1, temp2, temp3; |
| 175 | const float *X_low1, *Y01, *Y11; |
| 176 | float *x1=&X[0][0][0]; |
| 177 | float *j=x1+4864; |
| 178 | const int i_Temp = FFMAX(2*sbr->data[ch].t_env_num_env_old - i_f, 0); |
| 179 | |
| 180 | /* loop unrolled 8 times */ |
| 181 | __asm__ volatile ( |
| 182 | "1: \n\t" |
| 183 | "sw $0, 0(%[x1]) \n\t" |
| 184 | "sw $0, 4(%[x1]) \n\t" |
| 185 | "sw $0, 8(%[x1]) \n\t" |
| 186 | "sw $0, 12(%[x1]) \n\t" |
| 187 | "sw $0, 16(%[x1]) \n\t" |
| 188 | "sw $0, 20(%[x1]) \n\t" |
| 189 | "sw $0, 24(%[x1]) \n\t" |
| 190 | "sw $0, 28(%[x1]) \n\t" |
| 191 | "addiu %[x1], %[x1], 32 \n\t" |
| 192 | "bne %[x1], %[j], 1b \n\t" |
| 193 | "addiu %[x1], %[x1], -19456 \n\t" |
| 194 | |
| 195 | : [x1]"+r"(x1) |
| 196 | : [j]"r"(j) |
| 197 | : "memory" |
| 198 | ); |
| 199 | |
| 200 | if (i_Temp != 0) { |
| 201 | |
| 202 | X_low1=&X_low[0][2][0]; |
| 203 | |
| 204 | for (k = 0; k < sbr->kx[0]; k++) { |
| 205 | |
| 206 | __asm__ volatile ( |
| 207 | "move %[i], $zero \n\t" |
| 208 | "2: \n\t" |
| 209 | "lw %[temp0], 0(%[X_low1]) \n\t" |
| 210 | "lw %[temp1], 4(%[X_low1]) \n\t" |
| 211 | "sw %[temp0], 0(%[x1]) \n\t" |
| 212 | "sw %[temp1], 9728(%[x1]) \n\t" |
| 213 | "addiu %[x1], %[x1], 256 \n\t" |
| 214 | "addiu %[X_low1], %[X_low1], 8 \n\t" |
| 215 | "addiu %[i], %[i], 1 \n\t" |
| 216 | "bne %[i], %[i_Temp], 2b \n\t" |
| 217 | |
| 218 | : [x1]"+r"(x1), [X_low1]"+r"(X_low1), [i]"=&r"(i), |
| 219 | [temp0]"=&r"(temp0), [temp1]"=&r"(temp1) |
| 220 | : [i_Temp]"r"(i_Temp) |
| 221 | : "memory" |
| 222 | ); |
| 223 | x1-=(i_Temp<<6)-1; |
| 224 | X_low1-=(i_Temp<<1)-80; |
| 225 | } |
| 226 | |
| 227 | x1=&X[0][0][k]; |
| 228 | Y01=(float*)&Y0[32][k][0]; |
| 229 | |
| 230 | for (; k < sbr->kx[0] + sbr->m[0]; k++) { |
| 231 | __asm__ volatile ( |
| 232 | "move %[i], $zero \n\t" |
| 233 | "3: \n\t" |
| 234 | "lw %[temp0], 0(%[Y01]) \n\t" |
| 235 | "lw %[temp1], 4(%[Y01]) \n\t" |
| 236 | "sw %[temp0], 0(%[x1]) \n\t" |
| 237 | "sw %[temp1], 9728(%[x1]) \n\t" |
| 238 | "addiu %[x1], %[x1], 256 \n\t" |
| 239 | "addiu %[Y01], %[Y01], 512 \n\t" |
| 240 | "addiu %[i], %[i], 1 \n\t" |
| 241 | "bne %[i], %[i_Temp], 3b \n\t" |
| 242 | |
| 243 | : [x1]"+r"(x1), [Y01]"+r"(Y01), [i]"=&r"(i), |
| 244 | [temp0]"=&r"(temp0), [temp1]"=&r"(temp1) |
| 245 | : [i_Temp]"r"(i_Temp) |
| 246 | : "memory" |
| 247 | ); |
| 248 | x1 -=(i_Temp<<6)-1; |
| 249 | Y01 -=(i_Temp<<7)-2; |
| 250 | } |
| 251 | } |
| 252 | |
| 253 | x1=&X[0][i_Temp][0]; |
| 254 | X_low1=&X_low[0][i_Temp+2][0]; |
| 255 | temp3=38; |
| 256 | |
| 257 | for (k = 0; k < sbr->kx[1]; k++) { |
| 258 | |
| 259 | __asm__ volatile ( |
| 260 | "move %[i], %[i_Temp] \n\t" |
| 261 | "4: \n\t" |
| 262 | "lw %[temp0], 0(%[X_low1]) \n\t" |
| 263 | "lw %[temp1], 4(%[X_low1]) \n\t" |
| 264 | "sw %[temp0], 0(%[x1]) \n\t" |
| 265 | "sw %[temp1], 9728(%[x1]) \n\t" |
| 266 | "addiu %[x1], %[x1], 256 \n\t" |
| 267 | "addiu %[X_low1], %[X_low1], 8 \n\t" |
| 268 | "addiu %[i], %[i], 1 \n\t" |
| 269 | "bne %[i], %[temp3], 4b \n\t" |
| 270 | |
| 271 | : [x1]"+r"(x1), [X_low1]"+r"(X_low1), [i]"=&r"(i), |
| 272 | [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), |
| 273 | [temp2]"=&r"(temp2) |
| 274 | : [i_Temp]"r"(i_Temp), [temp3]"r"(temp3) |
| 275 | : "memory" |
| 276 | ); |
| 277 | x1 -= ((38-i_Temp)<<6)-1; |
| 278 | X_low1 -= ((38-i_Temp)<<1)- 80; |
| 279 | } |
| 280 | |
| 281 | x1=&X[0][i_Temp][k]; |
| 282 | Y11=&Y1[i_Temp][k][0]; |
| 283 | temp2=32; |
| 284 | |
| 285 | for (; k < sbr->kx[1] + sbr->m[1]; k++) { |
| 286 | |
| 287 | __asm__ volatile ( |
| 288 | "move %[i], %[i_Temp] \n\t" |
| 289 | "5: \n\t" |
| 290 | "lw %[temp0], 0(%[Y11]) \n\t" |
| 291 | "lw %[temp1], 4(%[Y11]) \n\t" |
| 292 | "sw %[temp0], 0(%[x1]) \n\t" |
| 293 | "sw %[temp1], 9728(%[x1]) \n\t" |
| 294 | "addiu %[x1], %[x1], 256 \n\t" |
| 295 | "addiu %[Y11], %[Y11], 512 \n\t" |
| 296 | "addiu %[i], %[i], 1 \n\t" |
| 297 | "bne %[i], %[temp2], 5b \n\t" |
| 298 | |
| 299 | : [x1]"+r"(x1), [Y11]"+r"(Y11), [i]"=&r"(i), |
| 300 | [temp0]"=&r"(temp0), [temp1]"=&r"(temp1) |
| 301 | : [i_Temp]"r"(i_Temp), [temp3]"r"(temp3), |
| 302 | [temp2]"r"(temp2) |
| 303 | : "memory" |
| 304 | ); |
| 305 | |
| 306 | x1 -= ((32-i_Temp)<<6)-1; |
| 307 | Y11 -= ((32-i_Temp)<<7)-2; |
| 308 | } |
| 309 | return 0; |
| 310 | } |
| 311 | |
| 312 | #if HAVE_MIPSFPU |
| 313 | static void sbr_hf_assemble_mips(float Y1[38][64][2], |
| 314 | const float X_high[64][40][2], |
| 315 | SpectralBandReplication *sbr, SBRData *ch_data, |
| 316 | const int e_a[2]) |
| 317 | { |
| 318 | int e, i, j, m; |
| 319 | const int h_SL = 4 * !sbr->bs_smoothing_mode; |
| 320 | const int kx = sbr->kx[1]; |
| 321 | const int m_max = sbr->m[1]; |
| 322 | static const float h_smooth[5] = { |
| 323 | 0.33333333333333, |
| 324 | 0.30150283239582, |
| 325 | 0.21816949906249, |
| 326 | 0.11516383427084, |
| 327 | 0.03183050093751, |
| 328 | }; |
| 329 | |
| 330 | float (*g_temp)[48] = ch_data->g_temp, (*q_temp)[48] = ch_data->q_temp; |
| 331 | int indexnoise = ch_data->f_indexnoise; |
| 332 | int indexsine = ch_data->f_indexsine; |
| 333 | float *g_temp1, *q_temp1, *pok, *pok1; |
| 334 | float temp1, temp2, temp3, temp4; |
| 335 | int size = m_max; |
| 336 | |
| 337 | if (sbr->reset) { |
| 338 | for (i = 0; i < h_SL; i++) { |
| 339 | memcpy(g_temp[i + 2*ch_data->t_env[0]], sbr->gain[0], m_max * sizeof(sbr->gain[0][0])); |
| 340 | memcpy(q_temp[i + 2*ch_data->t_env[0]], sbr->q_m[0], m_max * sizeof(sbr->q_m[0][0])); |
| 341 | } |
| 342 | } else if (h_SL) { |
| 343 | memcpy(g_temp[2*ch_data->t_env[0]], g_temp[2*ch_data->t_env_num_env_old], 4*sizeof(g_temp[0])); |
| 344 | memcpy(q_temp[2*ch_data->t_env[0]], q_temp[2*ch_data->t_env_num_env_old], 4*sizeof(q_temp[0])); |
| 345 | } |
| 346 | |
| 347 | for (e = 0; e < ch_data->bs_num_env; e++) { |
| 348 | for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) { |
| 349 | g_temp1 = g_temp[h_SL + i]; |
| 350 | pok = sbr->gain[e]; |
| 351 | q_temp1 = q_temp[h_SL + i]; |
| 352 | pok1 = sbr->q_m[e]; |
| 353 | |
| 354 | /* loop unrolled 4 times */ |
| 355 | for (j=0; j<(size>>2); j++) { |
| 356 | __asm__ volatile ( |
| 357 | "lw %[temp1], 0(%[pok]) \n\t" |
| 358 | "lw %[temp2], 4(%[pok]) \n\t" |
| 359 | "lw %[temp3], 8(%[pok]) \n\t" |
| 360 | "lw %[temp4], 12(%[pok]) \n\t" |
| 361 | "sw %[temp1], 0(%[g_temp1]) \n\t" |
| 362 | "sw %[temp2], 4(%[g_temp1]) \n\t" |
| 363 | "sw %[temp3], 8(%[g_temp1]) \n\t" |
| 364 | "sw %[temp4], 12(%[g_temp1]) \n\t" |
| 365 | "lw %[temp1], 0(%[pok1]) \n\t" |
| 366 | "lw %[temp2], 4(%[pok1]) \n\t" |
| 367 | "lw %[temp3], 8(%[pok1]) \n\t" |
| 368 | "lw %[temp4], 12(%[pok1]) \n\t" |
| 369 | "sw %[temp1], 0(%[q_temp1]) \n\t" |
| 370 | "sw %[temp2], 4(%[q_temp1]) \n\t" |
| 371 | "sw %[temp3], 8(%[q_temp1]) \n\t" |
| 372 | "sw %[temp4], 12(%[q_temp1]) \n\t" |
| 373 | "addiu %[pok], %[pok], 16 \n\t" |
| 374 | "addiu %[g_temp1], %[g_temp1], 16 \n\t" |
| 375 | "addiu %[pok1], %[pok1], 16 \n\t" |
| 376 | "addiu %[q_temp1], %[q_temp1], 16 \n\t" |
| 377 | |
| 378 | : [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), |
| 379 | [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), |
| 380 | [pok]"+r"(pok), [g_temp1]"+r"(g_temp1), |
| 381 | [pok1]"+r"(pok1), [q_temp1]"+r"(q_temp1) |
| 382 | : |
| 383 | : "memory" |
| 384 | ); |
| 385 | } |
| 386 | |
| 387 | for (j=0; j<(size&3); j++) { |
| 388 | __asm__ volatile ( |
| 389 | "lw %[temp1], 0(%[pok]) \n\t" |
| 390 | "lw %[temp2], 0(%[pok1]) \n\t" |
| 391 | "sw %[temp1], 0(%[g_temp1]) \n\t" |
| 392 | "sw %[temp2], 0(%[q_temp1]) \n\t" |
| 393 | "addiu %[pok], %[pok], 4 \n\t" |
| 394 | "addiu %[g_temp1], %[g_temp1], 4 \n\t" |
| 395 | "addiu %[pok1], %[pok1], 4 \n\t" |
| 396 | "addiu %[q_temp1], %[q_temp1], 4 \n\t" |
| 397 | |
| 398 | : [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), |
| 399 | [temp3]"=&r"(temp3), [temp4]"=&r"(temp4), |
| 400 | [pok]"+r"(pok), [g_temp1]"+r"(g_temp1), |
| 401 | [pok1]"+r"(pok1), [q_temp1]"+r"(q_temp1) |
| 402 | : |
| 403 | : "memory" |
| 404 | ); |
| 405 | } |
| 406 | } |
| 407 | } |
| 408 | |
| 409 | for (e = 0; e < ch_data->bs_num_env; e++) { |
| 410 | for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) { |
| 411 | LOCAL_ALIGNED_16(float, g_filt_tab, [48]); |
| 412 | LOCAL_ALIGNED_16(float, q_filt_tab, [48]); |
| 413 | float *g_filt, *q_filt; |
| 414 | |
| 415 | if (h_SL && e != e_a[0] && e != e_a[1]) { |
| 416 | g_filt = g_filt_tab; |
| 417 | q_filt = q_filt_tab; |
| 418 | |
| 419 | for (m = 0; m < m_max; m++) { |
| 420 | const int idx1 = i + h_SL; |
| 421 | g_filt[m] = 0.0f; |
| 422 | q_filt[m] = 0.0f; |
| 423 | |
| 424 | for (j = 0; j <= h_SL; j++) { |
| 425 | g_filt[m] += g_temp[idx1 - j][m] * h_smooth[j]; |
| 426 | q_filt[m] += q_temp[idx1 - j][m] * h_smooth[j]; |
| 427 | } |
| 428 | } |
| 429 | } else { |
| 430 | g_filt = g_temp[i + h_SL]; |
| 431 | q_filt = q_temp[i]; |
| 432 | } |
| 433 | |
| 434 | sbr->dsp.hf_g_filt(Y1[i] + kx, X_high + kx, g_filt, m_max, |
| 435 | i + ENVELOPE_ADJUSTMENT_OFFSET); |
| 436 | |
| 437 | if (e != e_a[0] && e != e_a[1]) { |
| 438 | sbr->dsp.hf_apply_noise[indexsine](Y1[i] + kx, sbr->s_m[e], |
| 439 | q_filt, indexnoise, |
| 440 | kx, m_max); |
| 441 | } else { |
| 442 | int idx = indexsine&1; |
| 443 | int A = (1-((indexsine+(kx & 1))&2)); |
| 444 | int B = (A^(-idx)) + idx; |
| 445 | float *out = &Y1[i][kx][idx]; |
| 446 | float *in = sbr->s_m[e]; |
| 447 | float temp0, temp1, temp2, temp3, temp4, temp5; |
| 448 | float A_f = (float)A; |
| 449 | float B_f = (float)B; |
| 450 | |
| 451 | for (m = 0; m+1 < m_max; m+=2) { |
| 452 | |
| 453 | temp2 = out[0]; |
| 454 | temp3 = out[2]; |
| 455 | |
| 456 | __asm__ volatile( |
| 457 | "lwc1 %[temp0], 0(%[in]) \n\t" |
| 458 | "lwc1 %[temp1], 4(%[in]) \n\t" |
| 459 | "madd.s %[temp4], %[temp2], %[temp0], %[A_f] \n\t" |
| 460 | "madd.s %[temp5], %[temp3], %[temp1], %[B_f] \n\t" |
| 461 | "swc1 %[temp4], 0(%[out]) \n\t" |
| 462 | "swc1 %[temp5], 8(%[out]) \n\t" |
| 463 | "addiu %[in], %[in], 8 \n\t" |
| 464 | "addiu %[out], %[out], 16 \n\t" |
| 465 | |
| 466 | : [temp0]"=&f" (temp0), [temp1]"=&f"(temp1), |
| 467 | [temp4]"=&f" (temp4), [temp5]"=&f"(temp5), |
| 468 | [in]"+r"(in), [out]"+r"(out) |
| 469 | : [A_f]"f"(A_f), [B_f]"f"(B_f), [temp2]"f"(temp2), |
| 470 | [temp3]"f"(temp3) |
| 471 | : "memory" |
| 472 | ); |
| 473 | } |
| 474 | if(m_max&1) |
| 475 | out[2*m ] += in[m ] * A; |
| 476 | } |
| 477 | indexnoise = (indexnoise + m_max) & 0x1ff; |
| 478 | indexsine = (indexsine + 1) & 3; |
| 479 | } |
| 480 | } |
| 481 | ch_data->f_indexnoise = indexnoise; |
| 482 | ch_data->f_indexsine = indexsine; |
| 483 | } |
| 484 | |
| 485 | static void sbr_hf_inverse_filter_mips(SBRDSPContext *dsp, |
| 486 | float (*alpha0)[2], float (*alpha1)[2], |
| 487 | const float X_low[32][40][2], int k0) |
| 488 | { |
| 489 | int k; |
| 490 | float temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7, c; |
| 491 | float *phi1, *alpha_1, *alpha_0, res1, res2, temp_real, temp_im; |
| 492 | |
| 493 | c = 1.000001f; |
| 494 | |
| 495 | for (k = 0; k < k0; k++) { |
| 496 | LOCAL_ALIGNED_16(float, phi, [3], [2][2]); |
| 497 | float dk; |
| 498 | phi1 = &phi[0][0][0]; |
| 499 | alpha_1 = &alpha1[k][0]; |
| 500 | alpha_0 = &alpha0[k][0]; |
| 501 | dsp->autocorrelate(X_low[k], phi); |
| 502 | |
| 503 | __asm__ volatile ( |
| 504 | "lwc1 %[temp0], 40(%[phi1]) \n\t" |
| 505 | "lwc1 %[temp1], 16(%[phi1]) \n\t" |
| 506 | "lwc1 %[temp2], 24(%[phi1]) \n\t" |
| 507 | "lwc1 %[temp3], 28(%[phi1]) \n\t" |
| 508 | "mul.s %[dk], %[temp0], %[temp1] \n\t" |
| 509 | "lwc1 %[temp4], 0(%[phi1]) \n\t" |
| 510 | "mul.s %[res2], %[temp2], %[temp2] \n\t" |
| 511 | "lwc1 %[temp5], 4(%[phi1]) \n\t" |
| 512 | "madd.s %[res2], %[res2], %[temp3], %[temp3] \n\t" |
| 513 | "lwc1 %[temp6], 8(%[phi1]) \n\t" |
| 514 | "div.s %[res2], %[res2], %[c] \n\t" |
| 515 | "lwc1 %[temp0], 12(%[phi1]) \n\t" |
| 516 | "sub.s %[dk], %[dk], %[res2] \n\t" |
| 517 | |
| 518 | : [temp0]"=&f"(temp0), [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 519 | [temp3]"=&f"(temp3), [temp4]"=&f"(temp4), [temp5]"=&f"(temp5), |
| 520 | [temp6]"=&f"(temp6), [res2]"=&f"(res2), [dk]"=&f"(dk) |
| 521 | : [phi1]"r"(phi1), [c]"f"(c) |
| 522 | : "memory" |
| 523 | ); |
| 524 | |
| 525 | if (!dk) { |
| 526 | alpha_1[0] = 0; |
| 527 | alpha_1[1] = 0; |
| 528 | } else { |
| 529 | __asm__ volatile ( |
| 530 | "mul.s %[temp_real], %[temp4], %[temp2] \n\t" |
| 531 | "nmsub.s %[temp_real], %[temp_real], %[temp5], %[temp3] \n\t" |
| 532 | "nmsub.s %[temp_real], %[temp_real], %[temp6], %[temp1] \n\t" |
| 533 | "mul.s %[temp_im], %[temp4], %[temp3] \n\t" |
| 534 | "madd.s %[temp_im], %[temp_im], %[temp5], %[temp2] \n\t" |
| 535 | "nmsub.s %[temp_im], %[temp_im], %[temp0], %[temp1] \n\t" |
| 536 | "div.s %[temp_real], %[temp_real], %[dk] \n\t" |
| 537 | "div.s %[temp_im], %[temp_im], %[dk] \n\t" |
| 538 | "swc1 %[temp_real], 0(%[alpha_1]) \n\t" |
| 539 | "swc1 %[temp_im], 4(%[alpha_1]) \n\t" |
| 540 | |
| 541 | : [temp_real]"=&f" (temp_real), [temp_im]"=&f"(temp_im) |
| 542 | : [phi1]"r"(phi1), [temp0]"f"(temp0), [temp1]"f"(temp1), |
| 543 | [temp2]"f"(temp2), [temp3]"f"(temp3), [temp4]"f"(temp4), |
| 544 | [temp5]"f"(temp5), [temp6]"f"(temp6), |
| 545 | [alpha_1]"r"(alpha_1), [dk]"f"(dk) |
| 546 | : "memory" |
| 547 | ); |
| 548 | } |
| 549 | |
| 550 | if (!phi1[4]) { |
| 551 | alpha_0[0] = 0; |
| 552 | alpha_0[1] = 0; |
| 553 | } else { |
| 554 | __asm__ volatile ( |
| 555 | "lwc1 %[temp6], 0(%[alpha_1]) \n\t" |
| 556 | "lwc1 %[temp7], 4(%[alpha_1]) \n\t" |
| 557 | "mul.s %[temp_real], %[temp6], %[temp2] \n\t" |
| 558 | "add.s %[temp_real], %[temp_real], %[temp4] \n\t" |
| 559 | "madd.s %[temp_real], %[temp_real], %[temp7], %[temp3] \n\t" |
| 560 | "mul.s %[temp_im], %[temp7], %[temp2] \n\t" |
| 561 | "add.s %[temp_im], %[temp_im], %[temp5] \n\t" |
| 562 | "nmsub.s %[temp_im], %[temp_im], %[temp6], %[temp3] \n\t" |
| 563 | "div.s %[temp_real], %[temp_real], %[temp1] \n\t" |
| 564 | "div.s %[temp_im], %[temp_im], %[temp1] \n\t" |
| 565 | "neg.s %[temp_real], %[temp_real] \n\t" |
| 566 | "neg.s %[temp_im], %[temp_im] \n\t" |
| 567 | "swc1 %[temp_real], 0(%[alpha_0]) \n\t" |
| 568 | "swc1 %[temp_im], 4(%[alpha_0]) \n\t" |
| 569 | |
| 570 | : [temp_real]"=&f"(temp_real), [temp_im]"=&f"(temp_im), |
| 571 | [temp6]"=&f"(temp6), [temp7]"=&f"(temp7), |
| 572 | [res1]"=&f"(res1), [res2]"=&f"(res2) |
| 573 | : [alpha_1]"r"(alpha_1), [alpha_0]"r"(alpha_0), |
| 574 | [temp0]"f"(temp0), [temp1]"f"(temp1), [temp2]"f"(temp2), |
| 575 | [temp3]"f"(temp3), [temp4]"f"(temp4), [temp5]"f"(temp5) |
| 576 | : "memory" |
| 577 | ); |
| 578 | } |
| 579 | |
| 580 | __asm__ volatile ( |
| 581 | "lwc1 %[temp1], 0(%[alpha_1]) \n\t" |
| 582 | "lwc1 %[temp2], 4(%[alpha_1]) \n\t" |
| 583 | "lwc1 %[temp_real], 0(%[alpha_0]) \n\t" |
| 584 | "lwc1 %[temp_im], 4(%[alpha_0]) \n\t" |
| 585 | "mul.s %[res1], %[temp1], %[temp1] \n\t" |
| 586 | "madd.s %[res1], %[res1], %[temp2], %[temp2] \n\t" |
| 587 | "mul.s %[res2], %[temp_real], %[temp_real] \n\t" |
| 588 | "madd.s %[res2], %[res2], %[temp_im], %[temp_im] \n\t" |
| 589 | |
| 590 | : [temp_real]"=&f"(temp_real), [temp_im]"=&f"(temp_im), |
| 591 | [temp1]"=&f"(temp1), [temp2]"=&f"(temp2), |
| 592 | [res1]"=&f"(res1), [res2]"=&f"(res2) |
| 593 | : [alpha_1]"r"(alpha_1), [alpha_0]"r"(alpha_0) |
| 594 | : "memory" |
| 595 | ); |
| 596 | |
| 597 | if (res1 >= 16.0f || res2 >= 16.0f) { |
| 598 | alpha_1[0] = 0; |
| 599 | alpha_1[1] = 0; |
| 600 | alpha_0[0] = 0; |
| 601 | alpha_0[1] = 0; |
| 602 | } |
| 603 | } |
| 604 | } |
| 605 | #endif /* HAVE_MIPSFPU */ |
| 606 | #endif /* HAVE_INLINE_ASM */ |
| 607 | |
| 608 | void ff_aacsbr_func_ptr_init_mips(AACSBRContext *c) |
| 609 | { |
| 610 | #if HAVE_INLINE_ASM |
| 611 | c->sbr_lf_gen = sbr_lf_gen_mips; |
| 612 | c->sbr_x_gen = sbr_x_gen_mips; |
| 613 | #if HAVE_MIPSFPU |
| 614 | c->sbr_hf_inverse_filter = sbr_hf_inverse_filter_mips; |
| 615 | c->sbr_hf_assemble = sbr_hf_assemble_mips; |
| 616 | #endif /* HAVE_MIPSFPU */ |
| 617 | #endif /* HAVE_INLINE_ASM */ |
| 618 | } |