1 /****************************************************************************** 2 * * 3 * Copyright (C) 2018 The Android Open Source Project 4 * 5 * Licensed under the Apache License, Version 2.0 (the "License"); 6 * you may not use this file except in compliance with the License. 7 * You may obtain a copy of the License at: 8 * 9 * http://www.apache.org/licenses/LICENSE-2.0 10 * 11 * Unless required by applicable law or agreed to in writing, software 12 * distributed under the License is distributed on an "AS IS" BASIS, 13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 * See the License for the specific language governing permissions and 15 * limitations under the License. 16 * 17 ***************************************************************************** 18 * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore 19 */ 20 #include <string.h> 21 #include "ixheaacd_sbr_common.h" 22 #include <ixheaacd_type_def.h> 23 24 #include "ixheaacd_constants.h" 25 #include "ixheaacd_basic_ops32.h" 26 #include "ixheaacd_basic_ops16.h" 27 #include "ixheaacd_basic_ops40.h" 28 #include "ixheaacd_basic_ops.h" 29 30 #include "ixheaacd_intrinsics.h" 31 #include "ixheaacd_common_rom.h" 32 #include "ixheaacd_bitbuffer.h" 33 #include "ixheaacd_sbrdecsettings.h" 34 #include "ixheaacd_sbr_scale.h" 35 #include "ixheaacd_lpp_tran.h" 36 #include "ixheaacd_env_extr_part.h" 37 #include "ixheaacd_sbr_rom.h" 38 #include "ixheaacd_hybrid.h" 39 #include "ixheaacd_ps_dec.h" 40 #include "ixheaacd_env_extr.h" 41 #include "ixheaacd_qmf_dec.h" 42 43 #include <ixheaacd_basic_op.h> 44 #include "ixheaacd_env_calc.h" 45 46 #include "ixheaacd_interface.h" 47 #include "ixheaacd_function_selector.h" 48 #include "ixheaacd_audioobjtypes.h" 49 50 #define mult16x16_16(a, b) ixheaacd_mult16((a), (b)) 51 #define mac16x16(a, b, c) ixheaacd_mac16x16in32((a), (b), (c)) 52 #define mpy_32x16(a, b) fixmuldiv2_32x16b((a), (b)) 53 #define mpy_16x16(a, b) ixheaacd_mult16x16in32((a), (b)) 54 #define mpy_32x32(a, b) ixheaacd_mult32((a), (b)) 55 #define mpy_32x16H_n(a, b) ixheaacd_mult32x16hin32((a), (b)) 56 #define msu16x16(a, b, c) msu16x16in32((a), (b), (c)) 57 58 #define DCT3_LEN (32) 59 #define DCT2_LEN (64) 60 61 #define LP_SHIFT_VAL 7 62 #define HQ_SHIFT_64 4 63 #define RADIXSHIFT 1 64 #define ROUNDING_SPECTRA 1 65 #define HQ_SHIFT_VAL 4 66 67 VOID ixheaacd_dct3_32(WORD32 *input, WORD32 *output, 68 const WORD16 *main_twidle_fwd, const WORD16 *post_tbl, 69 const WORD16 *w_16, const WORD32 *p_table) { 70 WORD32 n, k; 71 72 WORD32 temp1[6]; 73 WORD32 temp2[4]; 74 WORD16 twid_re, twid_im; 75 WORD32 *ptr_reverse, *ptr_forward, *p_out, *ptr_out1; 76 const WORD16 *twidle_fwd, *twidle_rev; 77 78 ptr_forward = &input[49]; 79 ptr_reverse = &input[47]; 80 81 p_out = output; 82 twidle_fwd = main_twidle_fwd; 83 twidle_fwd += 4; 84 85 *p_out++ = input[48] >> LP_SHIFT_VAL; 86 *p_out++ = 0; 87 88 for (n = 1; n < DCT3_LEN / 2; n++) { 89 temp1[0] = *ptr_forward++; 90 temp1[1] = *ptr_reverse--; 91 temp1[0] = ixheaacd_add32(ixheaacd_shr32(temp1[0], LP_SHIFT_VAL), 92 ixheaacd_shr32(temp1[1], LP_SHIFT_VAL)); 93 94 temp1[2] = *(ptr_forward - 33); 95 temp1[3] = *(ptr_reverse - 31); 96 temp1[1] = ixheaacd_sub32(ixheaacd_shr32(temp1[2], LP_SHIFT_VAL), 97 ixheaacd_shr32(temp1[3], LP_SHIFT_VAL)); 98 twid_re = *twidle_fwd++; 99 100 twid_im = *twidle_fwd; 101 twidle_fwd += 3; 102 *p_out++ = mac32x16in32_dual(temp1[0], twid_re, temp1[1], twid_im); 103 *p_out++ = msu32x16in32_dual(temp1[0], twid_im, temp1[1], twid_re); 104 } 105 twid_re = *twidle_fwd++; 106 107 twid_im = *twidle_fwd; 108 twidle_fwd += 3; 109 110 temp1[1] = *ptr_reverse--; 111 temp1[0] = *(ptr_reverse - 31); 112 temp1[1] = ixheaacd_sub32(ixheaacd_shr32(temp1[1], LP_SHIFT_VAL), 113 ixheaacd_shr32(temp1[0], LP_SHIFT_VAL)); 114 115 temp1[0] = temp1[1]; 116 117 temp2[2] = mac32x16in32_dual(temp1[0], twid_re, temp1[1], twid_im); 118 temp2[3] = msu32x16in32_dual(temp1[0], twid_im, temp1[1], twid_re); 119 120 ptr_forward = output; 121 ptr_reverse = &output[DCT3_LEN - 1]; 122 temp2[0] = *ptr_forward++; 123 temp2[1] = *ptr_forward--; 124 125 temp1[0] = -temp2[1] - temp2[3]; 126 temp1[1] = temp2[0] - temp2[2]; 127 temp2[0] = (temp2[0] + temp2[2] + temp1[0]); 128 temp2[1] = (temp2[1] - temp2[3] + temp1[1]); 129 130 temp2[0] >>= 1; 131 temp2[1] >>= 1; 132 133 *ptr_forward++ = temp2[0]; 134 *ptr_forward++ = temp2[1]; 135 136 twidle_fwd = post_tbl + 2; 137 twidle_rev = post_tbl + 14; 138 139 for (n = 1; n < DCT3_LEN / 4; n++) { 140 temp2[0] = *ptr_forward++; 141 temp2[1] = *ptr_forward--; 142 temp2[3] = *ptr_reverse--; 143 temp2[2] = *ptr_reverse++; 144 145 twid_re = *twidle_rev; 146 twidle_rev -= 2; 147 twid_im = *twidle_fwd; 148 twidle_fwd += 2; 149 150 temp1[0] = temp2[0] - temp2[2]; 151 temp1[1] = (temp2[0] + temp2[2]); 152 153 temp1[2] = temp2[1] + temp2[3]; 154 temp1[3] = (temp2[1] - temp2[3]); 155 temp1[4] = mac32x16in32_dual(temp1[0], twid_re, temp1[2], twid_im); 156 temp1[5] = msu32x16in32_dual(temp1[0], twid_im, temp1[2], twid_re); 157 158 temp1[1] >>= 1; 159 temp1[3] >>= 1; 160 161 *ptr_forward++ = temp1[1] - temp1[4]; 162 *ptr_forward++ = temp1[3] + temp1[5]; 163 164 *ptr_reverse-- = -temp1[3] + temp1[5]; 165 *ptr_reverse-- = temp1[1] + temp1[4]; 166 } 167 temp2[0] = *ptr_forward++; 168 temp2[1] = *ptr_forward--; 169 temp2[3] = *ptr_reverse--; 170 temp2[2] = *ptr_reverse++; 171 172 twid_re = *twidle_rev; 173 twidle_rev -= 2; 174 twid_im = *twidle_fwd; 175 twidle_fwd += 2; 176 177 temp1[0] = temp2[0] - temp2[2]; 178 temp1[1] = (temp2[0] + temp2[2]); 179 180 temp1[2] = temp2[1] + temp2[3]; 181 temp1[3] = (temp2[1] - temp2[3]); 182 183 temp1[4] = -mac32x16in32_dual(temp1[0], twid_re, temp1[2], twid_im); 184 temp1[5] = msu32x16in32_dual(temp1[0], twid_im, temp1[2], twid_re); 185 186 temp1[1] >>= 1; 187 temp1[3] >>= 1; 188 *ptr_forward++ = temp1[1] + temp1[4]; 189 *ptr_forward++ = temp1[3] + temp1[5]; 190 191 ixheaacd_radix4bfly(w_16, output, 1, 4); 192 ixheaacd_postradixcompute4(input, output, p_table, 16); 193 194 output[0] = input[0]; 195 output[2] = input[1]; 196 197 p_out = input + 2; 198 ptr_forward = output + 1; 199 ptr_reverse = output + 30; 200 ptr_out1 = input + 18; 201 202 for (k = (DCT3_LEN / 4) - 1; k != 0; k--) { 203 WORD32 tempre, tempim; 204 205 tempre = *p_out++; 206 tempim = *p_out++; 207 *ptr_forward = (tempim); 208 ptr_forward += 2; 209 *ptr_forward = (tempre); 210 ptr_forward += 2; 211 212 tempre = *ptr_out1++; 213 tempim = *ptr_out1++; 214 *ptr_reverse = (tempim); 215 ptr_reverse -= 2; 216 *ptr_reverse = (tempre); 217 ptr_reverse -= 2; 218 } 219 220 { 221 WORD32 tempre, tempim; 222 tempre = *p_out++; 223 tempim = *p_out++; 224 *ptr_forward = (tempim); 225 ptr_forward += 2; 226 *ptr_forward = (tempre); 227 ptr_forward += 2; 228 } 229 230 return; 231 } 232 VOID ixheaacd_dct2_64(WORD32 *x, WORD32 *X, 233 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr, 234 WORD16 *filter_states) { 235 ixheaacd_pretwdct2(x, X); 236 237 ixheaacd_sbr_imdct_using_fft(qmf_dec_tables_ptr->w1024, 32, X, x, 238 qmf_dec_tables_ptr->dig_rev_table2_128, 239 qmf_dec_tables_ptr->dig_rev_table2_128, 240 qmf_dec_tables_ptr->dig_rev_table2_128, 241 qmf_dec_tables_ptr->dig_rev_table2_128); 242 243 ixheaacd_fftposttw(x, qmf_dec_tables_ptr); 244 245 ixheaacd_posttwdct2(x, filter_states, qmf_dec_tables_ptr); 246 247 return; 248 } 249 250 VOID ixheaacd_cos_sin_mod(WORD32 *subband, 251 ia_sbr_qmf_filter_bank_struct *qmf_bank, 252 WORD16 *p_twiddle, WORD32 *p_dig_rev_tbl) { 253 WORD32 M = ixheaacd_shr32(qmf_bank->no_channels, 1); 254 255 const WORD16 *p_sin; 256 const WORD16 *p_sin_cos = &qmf_bank->cos_twiddle[0]; 257 WORD32 subband_tmp[128]; 258 259 ixheaacd_cos_sin_mod_loop1(subband, M, p_sin_cos, subband_tmp); 260 261 if (M == 32) { 262 ixheaacd_sbr_imdct_using_fft( 263 (const WORD32 *)p_twiddle, 32, subband_tmp, subband, 264 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl, 265 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl); 266 267 ixheaacd_sbr_imdct_using_fft( 268 (const WORD32 *)p_twiddle, 32, &subband_tmp[64], &subband[64], 269 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl, 270 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl); 271 272 } else { 273 ixheaacd_sbr_imdct_using_fft( 274 (const WORD32 *)p_twiddle, 16, subband_tmp, subband, 275 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl, 276 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl); 277 278 ixheaacd_sbr_imdct_using_fft( 279 (const WORD32 *)p_twiddle, 16, &subband_tmp[64], &subband[64], 280 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl, 281 (UWORD8 *)p_dig_rev_tbl, (UWORD8 *)p_dig_rev_tbl); 282 } 283 284 p_sin = &qmf_bank->alt_sin_twiddle[0]; 285 ixheaacd_cos_sin_mod_loop2(subband, p_sin, M); 286 } 287 288 VOID ixheaacd_fwd_modulation(const WORD32 *p_time_in1, WORD32 *real_subband, 289 WORD32 *imag_subband, 290 ia_sbr_qmf_filter_bank_struct *qmf_bank, 291 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr) { 292 WORD32 i; 293 const WORD32 *p_time_in2 = &p_time_in1[2 * NO_ANALYSIS_CHANNELS - 1]; 294 WORD32 temp1, temp2; 295 WORD32 *t_real_subband = real_subband; 296 WORD32 *t_imag_subband = imag_subband; 297 const WORD16 *tcos; 298 299 for (i = NO_ANALYSIS_CHANNELS - 1; i >= 0; i--) { 300 temp1 = ixheaacd_shr32(*p_time_in1++, HQ_SHIFT_VAL); 301 temp2 = ixheaacd_shr32(*p_time_in2--, HQ_SHIFT_VAL); 302 303 *t_real_subband++ = ixheaacd_sub32_sat(temp1, temp2); 304 ; 305 *t_imag_subband++ = ixheaacd_add32(temp1, temp2); 306 ; 307 } 308 309 ixheaacd_cos_sin_mod(real_subband, qmf_bank, 310 (WORD16 *)qmf_dec_tables_ptr->w1024, 311 (WORD32 *)qmf_dec_tables_ptr->dig_rev_table2_128); 312 313 tcos = qmf_bank->t_cos; 314 315 for (i = (qmf_bank->usb - qmf_bank->lsb - 1); i >= 0; i--) { 316 WORD16 cosh, sinh; 317 WORD32 re, im; 318 319 re = *real_subband; 320 im = *imag_subband; 321 cosh = *tcos++; 322 sinh = *tcos++; 323 *real_subband++ = ixheaacd_add32(ixheaacd_mult32x16in32_shl(re, cosh), 324 ixheaacd_mult32x16in32_shl(im, sinh)); 325 *imag_subband++ = ixheaacd_sub32_sat(ixheaacd_mult32x16in32_shl(im, cosh), 326 ixheaacd_mult32x16in32_shl(re, sinh)); 327 } 328 } 329 330 VOID ixheaacd_cplx_anal_qmffilt(const WORD16 *time_sample_buf, 331 ia_sbr_scale_fact_struct *sbr_scale_factor, 332 WORD32 **qmf_real, WORD32 **qmf_imag, 333 ia_sbr_qmf_filter_bank_struct *qmf_bank, 334 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr, 335 WORD32 ch_fac, WORD32 low_pow_flag, 336 WORD audio_object_type) { 337 WORD32 i, k; 338 WORD32 num_time_slots = qmf_bank->num_time_slots; 339 340 WORD32 analysis_buffer[4 * NO_ANALYSIS_CHANNELS]; 341 WORD16 *filter_states = qmf_bank->core_samples_buffer; 342 343 WORD16 *fp1, *fp2, *tmp; 344 345 WORD16 *filter_1; 346 WORD16 *filter_2; 347 WORD16 *filt_ptr; 348 if (audio_object_type != AOT_ER_AAC_ELD && 349 audio_object_type != AOT_ER_AAC_LD) { 350 qmf_bank->filter_pos += 351 (qmf_dec_tables_ptr->qmf_c - qmf_bank->analy_win_coeff); 352 qmf_bank->analy_win_coeff = qmf_dec_tables_ptr->qmf_c; 353 } else { 354 qmf_bank->filter_pos += 355 (qmf_dec_tables_ptr->qmf_c_eld3 - qmf_bank->analy_win_coeff); 356 qmf_bank->analy_win_coeff = qmf_dec_tables_ptr->qmf_c_eld3; 357 } 358 359 filter_1 = qmf_bank->filter_pos; 360 361 if (audio_object_type != AOT_ER_AAC_ELD && 362 audio_object_type != AOT_ER_AAC_LD) { 363 filter_2 = filter_1 + 64; 364 } else { 365 filter_2 = filter_1 + 32; 366 } 367 368 sbr_scale_factor->st_lb_scale = 0; 369 sbr_scale_factor->lb_scale = -10; 370 if (!low_pow_flag) { 371 if (audio_object_type != AOT_ER_AAC_ELD && 372 audio_object_type != AOT_ER_AAC_LD) { 373 sbr_scale_factor->lb_scale = -8; 374 } else { 375 sbr_scale_factor->lb_scale = -9; 376 } 377 qmf_bank->cos_twiddle = 378 (WORD16 *)qmf_dec_tables_ptr->sbr_sin_cos_twiddle_l32; 379 qmf_bank->alt_sin_twiddle = 380 (WORD16 *)qmf_dec_tables_ptr->sbr_alt_sin_twiddle_l32; 381 if (audio_object_type != AOT_ER_AAC_ELD && 382 audio_object_type != AOT_ER_AAC_LD) { 383 qmf_bank->t_cos = (WORD16 *)qmf_dec_tables_ptr->sbr_t_cos_sin_l32; 384 } else { 385 qmf_bank->t_cos = 386 (WORD16 *)qmf_dec_tables_ptr->ixheaacd_sbr_t_cos_sin_l32_eld; 387 } 388 } 389 390 fp1 = qmf_bank->anal_filter_states; 391 fp2 = qmf_bank->anal_filter_states + NO_ANALYSIS_CHANNELS; 392 393 if (audio_object_type == AOT_ER_AAC_ELD || 394 audio_object_type == AOT_ER_AAC_LD) { 395 filter_2 = qmf_bank->filter_2; 396 fp1 = qmf_bank->fp1_anal; 397 fp2 = qmf_bank->fp2_anal; 398 } 399 400 for (i = 0; i < num_time_slots; i++) { 401 for (k = 0; k < NO_ANALYSIS_CHANNELS; k++) 402 filter_states[NO_ANALYSIS_CHANNELS - 1 - k] = time_sample_buf[ch_fac * k]; 403 404 if (audio_object_type != AOT_ER_AAC_ELD && 405 audio_object_type != AOT_ER_AAC_LD) { 406 ixheaacd_sbr_qmfanal32_winadds(fp1, fp2, filter_1, filter_2, 407 analysis_buffer, filter_states, 408 time_sample_buf, ch_fac); 409 } 410 411 else { 412 ixheaacd_sbr_qmfanal32_winadd_eld(fp1, fp2, filter_1, filter_2, 413 analysis_buffer); 414 } 415 416 time_sample_buf += NO_ANALYSIS_CHANNELS * ch_fac; 417 418 filter_states -= NO_ANALYSIS_CHANNELS; 419 if (filter_states < qmf_bank->anal_filter_states) { 420 filter_states = qmf_bank->anal_filter_states + 288; 421 } 422 423 tmp = fp1; 424 fp1 = fp2; 425 fp2 = tmp; 426 if (audio_object_type != AOT_ER_AAC_ELD && 427 audio_object_type != AOT_ER_AAC_LD) { 428 filter_1 += 64; 429 filter_2 += 64; 430 } else { 431 filter_1 += 32; 432 filter_2 += 32; 433 } 434 435 filt_ptr = filter_1; 436 filter_1 = filter_2; 437 filter_2 = filt_ptr; 438 if (audio_object_type != AOT_ER_AAC_ELD && 439 audio_object_type != AOT_ER_AAC_LD) { 440 if (filter_2 > (qmf_bank->analy_win_coeff + 640)) { 441 filter_1 = (WORD16 *)qmf_bank->analy_win_coeff; 442 filter_2 = (WORD16 *)qmf_bank->analy_win_coeff + 64; 443 } 444 } else { 445 if (filter_2 > (qmf_bank->analy_win_coeff + 320)) { 446 filter_1 = (WORD16 *)qmf_bank->analy_win_coeff; 447 filter_2 = (WORD16 *)qmf_bank->analy_win_coeff + 32; 448 } 449 } 450 451 if (!low_pow_flag) { 452 ixheaacd_fwd_modulation(analysis_buffer, qmf_real[i], qmf_imag[i], 453 qmf_bank, qmf_dec_tables_ptr); 454 } else { 455 ixheaacd_dct3_32( 456 (WORD32 *)analysis_buffer, qmf_real[i], qmf_dec_tables_ptr->dct23_tw, 457 qmf_dec_tables_ptr->post_fft_tbl, qmf_dec_tables_ptr->w_16, 458 qmf_dec_tables_ptr->dig_rev_table4_16); 459 } 460 } 461 462 qmf_bank->filter_pos = filter_1; 463 qmf_bank->core_samples_buffer = filter_states; 464 465 if (audio_object_type == AOT_ER_AAC_ELD || audio_object_type == AOT_ER_AAC_LD) 466 467 { 468 qmf_bank->fp1_anal = fp1; 469 qmf_bank->fp2_anal = fp2; 470 qmf_bank->filter_2 = filter_2; 471 } 472 } 473 474 VOID ixheaacd_inv_modulation_lp(WORD32 *qmf_real, WORD16 *filter_states, 475 ia_sbr_qmf_filter_bank_struct *syn_qmf, 476 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr) { 477 WORD32 L = syn_qmf->no_channels; 478 const WORD32 M = (L >> 1); 479 WORD32 *dct_in = qmf_real; 480 WORD32 time_out[2 * NO_SYNTHESIS_CHANNELS]; 481 482 WORD32 ui_rem = ((WORD64)(&time_out[0]) % 8); 483 WORD32 *ptime_out = (pVOID)((WORD8 *)&time_out[0] + 8 - ui_rem); 484 485 if (L == 64) 486 ixheaacd_dct2_64(dct_in, ptime_out, qmf_dec_tables_ptr, filter_states + M); 487 else 488 ixheaacd_dct2_32(dct_in, time_out, qmf_dec_tables_ptr, filter_states); 489 490 filter_states[3 * M] = 0; 491 } 492 493 VOID ixheaacd_inv_emodulation(WORD32 *qmf_real, 494 ia_sbr_qmf_filter_bank_struct *syn_qmf, 495 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr) { 496 ixheaacd_cos_sin_mod(qmf_real, syn_qmf, (WORD16 *)qmf_dec_tables_ptr->w1024, 497 (WORD32 *)qmf_dec_tables_ptr->dig_rev_table2_128); 498 } 499 500 VOID ixheaacd_esbr_radix4bfly(const WORD32 *w, WORD32 *x, WORD32 index1, 501 WORD32 index) { 502 int i; 503 WORD32 l1, l2, h2, fft_jmp; 504 WORD32 xt0_0, yt0_0, xt1_0, yt1_0, xt2_0, yt2_0; 505 WORD32 xh0_0, xh1_0, xh20_0, xh21_0, xl0_0, xl1_0, xl20_0, xl21_0; 506 WORD32 x_0, x_1, x_l1_0, x_l1_1, x_l2_0, x_l2_1; 507 WORD32 x_h2_0, x_h2_1; 508 WORD32 si10, si20, si30, co10, co20, co30; 509 510 WORD64 mul_1, mul_2, mul_3, mul_4, mul_5, mul_6; 511 WORD64 mul_7, mul_8, mul_9, mul_10, mul_11, mul_12; 512 WORD32 *x_l1; 513 WORD32 *x_l2; 514 WORD32 *x_h2; 515 const WORD32 *w_ptr = w; 516 WORD32 i1; 517 518 h2 = index << 1; 519 l1 = index << 2; 520 l2 = (index << 2) + (index << 1); 521 522 x_l1 = &(x[l1]); 523 x_l2 = &(x[l2]); 524 x_h2 = &(x[h2]); 525 526 fft_jmp = 6 * (index); 527 528 for (i1 = 0; i1 < index1; i1++) { 529 for (i = 0; i < index; i++) { 530 si10 = (*w_ptr++); 531 co10 = (*w_ptr++); 532 si20 = (*w_ptr++); 533 co20 = (*w_ptr++); 534 si30 = (*w_ptr++); 535 co30 = (*w_ptr++); 536 537 x_0 = x[0]; 538 x_h2_0 = x[h2]; 539 x_l1_0 = x[l1]; 540 x_l2_0 = x[l2]; 541 542 xh0_0 = x_0 + x_l1_0; 543 xl0_0 = x_0 - x_l1_0; 544 545 xh20_0 = x_h2_0 + x_l2_0; 546 xl20_0 = x_h2_0 - x_l2_0; 547 548 x[0] = xh0_0 + xh20_0; 549 xt0_0 = xh0_0 - xh20_0; 550 551 x_1 = x[1]; 552 x_h2_1 = x[h2 + 1]; 553 x_l1_1 = x[l1 + 1]; 554 x_l2_1 = x[l2 + 1]; 555 556 xh1_0 = x_1 + x_l1_1; 557 xl1_0 = x_1 - x_l1_1; 558 559 xh21_0 = x_h2_1 + x_l2_1; 560 xl21_0 = x_h2_1 - x_l2_1; 561 562 x[1] = xh1_0 + xh21_0; 563 yt0_0 = xh1_0 - xh21_0; 564 565 xt1_0 = xl0_0 + xl21_0; 566 xt2_0 = xl0_0 - xl21_0; 567 568 yt2_0 = xl1_0 + xl20_0; 569 yt1_0 = xl1_0 - xl20_0; 570 571 mul_11 = ixheaacd_mult64(xt2_0, co30); 572 mul_3 = ixheaacd_mult64(yt2_0, si30); 573 x[l2] = (WORD32)((mul_3 + mul_11) >> 32) << RADIXSHIFT; 574 575 mul_5 = ixheaacd_mult64(xt2_0, si30); 576 mul_9 = ixheaacd_mult64(yt2_0, co30); 577 x[l2 + 1] = (WORD32)((mul_9 - mul_5) >> 32) << RADIXSHIFT; 578 579 mul_12 = ixheaacd_mult64(xt0_0, co20); 580 mul_2 = ixheaacd_mult64(yt0_0, si20); 581 x[l1] = (WORD32)((mul_2 + mul_12) >> 32) << RADIXSHIFT; 582 583 mul_6 = ixheaacd_mult64(xt0_0, si20); 584 mul_8 = ixheaacd_mult64(yt0_0, co20); 585 x[l1 + 1] = (WORD32)((mul_8 - mul_6) >> 32) << RADIXSHIFT; 586 587 mul_4 = ixheaacd_mult64(xt1_0, co10); 588 mul_1 = ixheaacd_mult64(yt1_0, si10); 589 x[h2] = (WORD32)((mul_1 + mul_4) >> 32) << RADIXSHIFT; 590 591 mul_10 = ixheaacd_mult64(xt1_0, si10); 592 mul_7 = ixheaacd_mult64(yt1_0, co10); 593 x[h2 + 1] = (WORD32)((mul_7 - mul_10) >> 32) << RADIXSHIFT; 594 595 x += 2; 596 } 597 x += fft_jmp; 598 w_ptr = w_ptr - fft_jmp; 599 } 600 } 601 602 VOID ixheaacd_esbr_postradixcompute2(WORD32 *ptr_y, WORD32 *ptr_x, 603 const WORD32 *pdig_rev_tbl, 604 WORD32 npoints) { 605 WORD32 i, k; 606 WORD32 h2; 607 WORD32 x_0, x_1, x_2, x_3; 608 WORD32 x_4, x_5, x_6, x_7; 609 WORD32 x_8, x_9, x_a, x_b, x_c, x_d, x_e, x_f; 610 WORD32 n00, n10, n20, n30, n01, n11, n21, n31; 611 WORD32 n02, n12, n22, n32, n03, n13, n23, n33; 612 WORD32 n0, j0; 613 WORD32 *x2, *x0; 614 WORD32 *y0, *y1, *y2, *y3; 615 616 y0 = ptr_y; 617 y2 = ptr_y + (WORD32)npoints; 618 x0 = ptr_x; 619 x2 = ptr_x + (WORD32)(npoints >> 1); 620 621 y1 = y0 + (WORD32)(npoints >> 2); 622 y3 = y2 + (WORD32)(npoints >> 2); 623 j0 = 8; 624 n0 = npoints >> 1; 625 626 for (k = 0; k < 2; k++) { 627 for (i = 0; i<npoints>> 1; i += 8) { 628 h2 = *pdig_rev_tbl++ >> 2; 629 630 x_0 = *x0++; 631 x_1 = *x0++; 632 x_2 = *x0++; 633 x_3 = *x0++; 634 x_4 = *x0++; 635 x_5 = *x0++; 636 x_6 = *x0++; 637 x_7 = *x0++; 638 639 n00 = x_0 + x_2; 640 n01 = x_1 + x_3; 641 n20 = x_0 - x_2; 642 n21 = x_1 - x_3; 643 n10 = x_4 + x_6; 644 n11 = x_5 + x_7; 645 n30 = x_4 - x_6; 646 n31 = x_5 - x_7; 647 648 y0[h2] = n00; 649 y0[h2 + 1] = n01; 650 y1[h2] = n10; 651 y1[h2 + 1] = n11; 652 y2[h2] = n20; 653 y2[h2 + 1] = n21; 654 y3[h2] = n30; 655 y3[h2 + 1] = n31; 656 657 x_8 = *x2++; 658 x_9 = *x2++; 659 x_a = *x2++; 660 x_b = *x2++; 661 x_c = *x2++; 662 x_d = *x2++; 663 x_e = *x2++; 664 x_f = *x2++; 665 666 n02 = x_8 + x_a; 667 n03 = x_9 + x_b; 668 n22 = x_8 - x_a; 669 n23 = x_9 - x_b; 670 n12 = x_c + x_e; 671 n13 = x_d + x_f; 672 n32 = x_c - x_e; 673 n33 = x_d - x_f; 674 675 y0[h2 + 2] = n02; 676 y0[h2 + 3] = n03; 677 y1[h2 + 2] = n12; 678 y1[h2 + 3] = n13; 679 y2[h2 + 2] = n22; 680 y2[h2 + 3] = n23; 681 y3[h2 + 2] = n32; 682 y3[h2 + 3] = n33; 683 } 684 x0 += (WORD32)npoints >> 1; 685 x2 += (WORD32)npoints >> 1; 686 } 687 } 688 689 VOID ixheaacd_esbr_postradixcompute4(WORD32 *ptr_y, WORD32 *ptr_x, 690 const WORD32 *p_dig_rev_tbl, 691 WORD32 npoints) { 692 WORD32 i, k; 693 WORD32 h2; 694 WORD32 xh0_0, xh1_0, xl0_0, xl1_0; 695 WORD32 xh0_1, xh1_1, xl0_1, xl1_1; 696 WORD32 x_0, x_1, x_2, x_3; 697 WORD32 xh0_2, xh1_2, xl0_2, xl1_2, xh0_3, xh1_3, xl0_3, xl1_3; 698 WORD32 x_4, x_5, x_6, x_7; 699 WORD32 x_8, x_9, x_a, x_b, x_c, x_d, x_e, x_f; 700 WORD32 n00, n10, n20, n30, n01, n11, n21, n31; 701 WORD32 n02, n12, n22, n32, n03, n13, n23, n33; 702 WORD32 n0, j0; 703 WORD32 *x2, *x0; 704 WORD32 *y0, *y1, *y2, *y3; 705 706 y0 = ptr_y; 707 y2 = ptr_y + (WORD32)npoints; 708 x0 = ptr_x; 709 x2 = ptr_x + (WORD32)(npoints >> 1); 710 711 y1 = y0 + (WORD32)(npoints >> 1); 712 y3 = y2 + (WORD32)(npoints >> 1); 713 714 j0 = 4; 715 n0 = npoints >> 2; 716 717 for (k = 0; k < 2; k++) { 718 for (i = 0; i<npoints>> 1; i += 8) { 719 h2 = *p_dig_rev_tbl++ >> 2; 720 x_0 = *x0++; 721 x_1 = *x0++; 722 x_2 = *x0++; 723 x_3 = *x0++; 724 x_4 = *x0++; 725 x_5 = *x0++; 726 x_6 = *x0++; 727 x_7 = *x0++; 728 729 xh0_0 = x_0 + x_4; 730 xh1_0 = x_1 + x_5; 731 xl0_0 = x_0 - x_4; 732 xl1_0 = x_1 - x_5; 733 xh0_1 = x_2 + x_6; 734 xh1_1 = x_3 + x_7; 735 xl0_1 = x_2 - x_6; 736 xl1_1 = x_3 - x_7; 737 738 n00 = xh0_0 + xh0_1; 739 n01 = xh1_0 + xh1_1; 740 n10 = xl0_0 + xl1_1; 741 n11 = xl1_0 - xl0_1; 742 n20 = xh0_0 - xh0_1; 743 n21 = xh1_0 - xh1_1; 744 n30 = xl0_0 - xl1_1; 745 n31 = xl1_0 + xl0_1; 746 747 y0[h2] = n00; 748 y0[h2 + 1] = n01; 749 y1[h2] = n10; 750 y1[h2 + 1] = n11; 751 y2[h2] = n20; 752 y2[h2 + 1] = n21; 753 y3[h2] = n30; 754 y3[h2 + 1] = n31; 755 756 x_8 = *x2++; 757 x_9 = *x2++; 758 x_a = *x2++; 759 x_b = *x2++; 760 x_c = *x2++; 761 x_d = *x2++; 762 x_e = *x2++; 763 x_f = *x2++; 764 765 xh0_2 = x_8 + x_c; 766 xh1_2 = x_9 + x_d; 767 xl0_2 = x_8 - x_c; 768 xl1_2 = x_9 - x_d; 769 xh0_3 = x_a + x_e; 770 xh1_3 = x_b + x_f; 771 xl0_3 = x_a - x_e; 772 xl1_3 = x_b - x_f; 773 774 n02 = xh0_2 + xh0_3; 775 n03 = xh1_2 + xh1_3; 776 n12 = xl0_2 + xl1_3; 777 n13 = xl1_2 - xl0_3; 778 n22 = xh0_2 - xh0_3; 779 n23 = xh1_2 - xh1_3; 780 n32 = xl0_2 - xl1_3; 781 n33 = xl1_2 + xl0_3; 782 783 y0[h2 + 2] = n02; 784 y0[h2 + 3] = n03; 785 y1[h2 + 2] = n12; 786 y1[h2 + 3] = n13; 787 y2[h2 + 2] = n22; 788 y2[h2 + 3] = n23; 789 y3[h2 + 2] = n32; 790 y3[h2 + 3] = n33; 791 } 792 x0 += (WORD32)npoints >> 1; 793 x2 += (WORD32)npoints >> 1; 794 } 795 } 796 797 VOID ixheaacd_esbr_cos_sin_mod(WORD32 *subband, 798 ia_sbr_qmf_filter_bank_struct *qmf_bank, 799 WORD32 *p_twiddle, WORD32 *p_dig_rev_tbl) { 800 WORD32 z; 801 WORD32 temp[128]; 802 WORD32 scaleshift = 0; 803 804 WORD32 re2, re3; 805 WORD32 wim, wre; 806 807 WORD32 i, M_2; 808 WORD32 M = ixheaacd_shr32(qmf_bank->no_channels, 1); 809 810 const WORD32 *p_sin; 811 const WORD32 *p_sin_cos; 812 813 WORD32 subband_tmp[128]; 814 WORD32 re; 815 WORD32 im; 816 WORD32 *psubband, *psubband1; 817 WORD32 *psubband_t, *psubband1_t; 818 WORD32 *psubband2, *psubband12; 819 WORD32 *psubband_t2, *psubband1_t2; 820 821 M_2 = ixheaacd_shr32(M, 1); 822 823 p_sin_cos = qmf_bank->esbr_cos_twiddle; 824 825 psubband = &subband[0]; 826 psubband1 = &subband[2 * M - 1]; 827 psubband_t = subband_tmp; 828 psubband1_t = &subband_tmp[2 * M - 1]; 829 830 psubband2 = &subband[64]; 831 psubband12 = &subband[2 * M - 1 + 64]; 832 psubband_t2 = &subband_tmp[64]; 833 psubband1_t2 = &subband_tmp[2 * M - 1 + 64]; 834 835 for (i = (M_2 >> 1) - 1; i >= 0; i--) { 836 re = *psubband++; 837 im = *psubband1--; 838 839 wim = *p_sin_cos++; 840 wre = *p_sin_cos++; 841 842 *psubband_t++ = (WORD32)( 843 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 844 32); 845 *psubband_t++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 846 ixheaacd_mult64(re, wim))) >> 847 32); 848 849 re = *psubband2++; 850 im = *psubband12--; 851 852 *psubband_t2++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wim), 853 ixheaacd_mult64(re, wre))) >> 854 32); 855 *psubband_t2++ = (WORD32)( 856 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 857 32); 858 859 re = *psubband1--; 860 im = *psubband++; 861 862 wim = *p_sin_cos++; 863 wre = *p_sin_cos++; 864 865 *psubband1_t-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 866 ixheaacd_mult64(re, wim))) >> 867 32); 868 *psubband1_t-- = (WORD32)( 869 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 870 32); 871 872 re = *psubband12--; 873 im = *psubband2++; 874 875 *psubband1_t2-- = (WORD32)( 876 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 877 32); 878 *psubband1_t2-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wim), 879 ixheaacd_mult64(re, wre))) >> 880 32); 881 882 re = *psubband++; 883 im = *psubband1--; 884 885 wim = *p_sin_cos++; 886 wre = *p_sin_cos++; 887 888 *psubband_t++ = (WORD32)( 889 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 890 32); 891 *psubband_t++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 892 ixheaacd_mult64(re, wim))) >> 893 32); 894 895 re = *psubband2++; 896 im = *psubband12--; 897 898 *psubband_t2++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wim), 899 ixheaacd_mult64(re, wre))) >> 900 32); 901 *psubband_t2++ = (WORD32)( 902 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 903 32); 904 905 re = *psubband1--; 906 im = *psubband++; 907 ; 908 909 wim = *p_sin_cos++; 910 wre = *p_sin_cos++; 911 912 *psubband1_t-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 913 ixheaacd_mult64(re, wim))) >> 914 32); 915 *psubband1_t-- = (WORD32)( 916 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 917 32); 918 919 re = *psubband12--; 920 im = *psubband2++; 921 ; 922 923 *psubband1_t2-- = (WORD32)( 924 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 925 32); 926 *psubband1_t2-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wim), 927 ixheaacd_mult64(re, wre))) >> 928 32); 929 } 930 931 if (M == 32) { 932 ixheaacd_esbr_radix4bfly(p_twiddle, subband_tmp, 1, 8); 933 ixheaacd_esbr_radix4bfly(p_twiddle + 48, subband_tmp, 4, 2); 934 ixheaacd_esbr_postradixcompute2(subband, subband_tmp, p_dig_rev_tbl, 32); 935 936 ixheaacd_esbr_radix4bfly(p_twiddle, &subband_tmp[64], 1, 8); 937 ixheaacd_esbr_radix4bfly(p_twiddle + 48, &subband_tmp[64], 4, 2); 938 ixheaacd_esbr_postradixcompute2(&subband[64], &subband_tmp[64], 939 p_dig_rev_tbl, 32); 940 941 } 942 943 else if (M == 16) { 944 ixheaacd_esbr_radix4bfly(p_twiddle, subband_tmp, 1, 4); 945 ixheaacd_esbr_postradixcompute4(subband, subband_tmp, p_dig_rev_tbl, 16); 946 947 ixheaacd_esbr_radix4bfly(p_twiddle, &subband_tmp[64], 1, 4); 948 ixheaacd_esbr_postradixcompute4(&subband[64], &subband_tmp[64], 949 p_dig_rev_tbl, 16); 950 951 } 952 953 else if (M == 12) { 954 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 955 temp[z] = subband_tmp[2 * z]; 956 temp[12 + z] = subband_tmp[2 * z + 1]; 957 } 958 959 ixheaacd_complex_fft_p3(temp, &temp[12], 12, -1, &scaleshift); 960 961 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 962 subband[2 * z] = temp[z]; 963 subband[2 * z + 1] = temp[z + 12]; 964 } 965 scaleshift = 0; 966 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 967 temp[z] = subband_tmp[64 + 2 * z]; 968 temp[12 + z] = subband_tmp[64 + 2 * z + 1]; 969 } 970 971 ixheaacd_complex_fft_p3(temp, &temp[12], 12, -1, &scaleshift); 972 973 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 974 subband[64 + 2 * z] = temp[z]; 975 subband[64 + 2 * z + 1] = temp[z + 12]; 976 } 977 978 } 979 980 else { 981 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 982 temp[z] = subband_tmp[2 * z]; 983 temp[8 + z] = subband_tmp[2 * z + 1]; 984 } 985 986 (*ixheaacd_complex_fft_p2)(temp, &temp[8], 8, -1, &scaleshift); 987 988 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 989 subband[2 * z] = temp[z] << scaleshift; 990 subband[2 * z + 1] = temp[z + 8] << scaleshift; 991 } 992 scaleshift = 0; 993 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 994 temp[z] = subband_tmp[64 + 2 * z]; 995 temp[8 + z] = subband_tmp[64 + 2 * z + 1]; 996 } 997 998 (*ixheaacd_complex_fft_p2)(temp, &temp[8], 8, -1, &scaleshift); 999 1000 for (z = 0; z < (qmf_bank->no_channels >> 1); z++) { 1001 subband[64 + 2 * z] = temp[z] << scaleshift; 1002 subband[64 + 2 * z + 1] = temp[8 + z] << scaleshift; 1003 } 1004 } 1005 1006 psubband = &subband[0]; 1007 psubband1 = &subband[2 * M - 1]; 1008 1009 re = *psubband1; 1010 1011 *psubband = *psubband >> 1; 1012 psubband++; 1013 *psubband1 = ixheaacd_negate32(*psubband >> 1); 1014 psubband1--; 1015 1016 p_sin = qmf_bank->esbr_alt_sin_twiddle; 1017 1018 wim = *p_sin++; 1019 wre = *p_sin++; 1020 1021 im = *psubband1; 1022 ; 1023 1024 *psubband1-- = (WORD32)( 1025 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 1026 32); 1027 *psubband++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 1028 ixheaacd_mult64(re, wim))) >> 1029 32); 1030 1031 psubband2 = &subband[64]; 1032 psubband12 = &subband[2 * M - 1 + 64]; 1033 1034 re = *psubband12; 1035 ; 1036 1037 *psubband12-- = ixheaacd_negate32_sat(*psubband2 >> 1); 1038 ; 1039 *psubband2 = psubband2[1] >> 1; 1040 ; 1041 psubband2++; 1042 1043 im = *psubband12; 1044 ; 1045 1046 *psubband2++ = ixheaacd_negate32_sat((WORD32)( 1047 (ixheaacd_add64(ixheaacd_mult64(re, wre), ixheaacd_mult64(im, wim))) >> 1048 32)); 1049 *psubband12-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(re, wim), 1050 ixheaacd_mult64(im, wre))) >> 1051 32); 1052 1053 for (i = (M_2 - 2); i >= 0; i--) { 1054 im = psubband[0]; 1055 ; 1056 re = psubband[1]; 1057 ; 1058 re2 = *psubband1; 1059 ; 1060 1061 *psubband++ = (WORD32)( 1062 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 1063 32); 1064 *psubband1-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wim), 1065 ixheaacd_mult64(re, wre))) >> 1066 32); 1067 1068 im = psubband2[0]; 1069 ; 1070 re = psubband2[1]; 1071 ; 1072 re3 = *psubband12; 1073 ; 1074 1075 *psubband12-- = ixheaacd_negate32_sat((WORD32)( 1076 (ixheaacd_add64(ixheaacd_mult64(re, wim), ixheaacd_mult64(im, wre))) >> 1077 32)); 1078 *psubband2++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(re, wre), 1079 ixheaacd_mult64(im, wim))) >> 1080 32); 1081 1082 wim = *p_sin++; 1083 wre = *p_sin++; 1084 im = psubband1[0]; 1085 ; 1086 1087 *psubband1-- = (WORD32)( 1088 (ixheaacd_add64(ixheaacd_mult64(re2, wre), ixheaacd_mult64(im, wim))) >> 1089 32); 1090 *psubband++ = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, wre), 1091 ixheaacd_mult64(re2, wim))) >> 1092 32); 1093 1094 im = psubband12[0]; 1095 ; 1096 1097 *psubband2++ = ixheaacd_negate32_sat((WORD32)( 1098 (ixheaacd_add64(ixheaacd_mult64(re3, wre), ixheaacd_mult64(im, wim))) >> 1099 32)); 1100 *psubband12-- = (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(re3, wim), 1101 ixheaacd_mult64(im, wre))) >> 1102 32); 1103 } 1104 } 1105 1106 VOID ixheaacd_esbr_fwd_modulation( 1107 const WORD32 *time_sample_buf, WORD32 *real_subband, WORD32 *imag_subband, 1108 ia_sbr_qmf_filter_bank_struct *qmf_bank, 1109 ia_qmf_dec_tables_struct *qmf_dec_tables_ptr) { 1110 WORD32 i; 1111 const WORD32 *time_sample_buf1 = 1112 &time_sample_buf[2 * qmf_bank->no_channels - 1]; 1113 WORD32 temp1, temp2; 1114 WORD32 *t_real_subband = real_subband; 1115 WORD32 *t_imag_subband = imag_subband; 1116 const WORD32 *tcos; 1117 1118 for (i = qmf_bank->no_channels - 1; i >= 0; i--) { 1119 temp1 = ixheaacd_shr32(*time_sample_buf++, HQ_SHIFT_64); 1120 temp2 = ixheaacd_shr32(*time_sample_buf1--, HQ_SHIFT_64); 1121 1122 *t_real_subband++ = ixheaacd_sub32_sat(temp1, temp2); 1123 ; 1124 *t_imag_subband++ = ixheaacd_add32(temp1, temp2); 1125 ; 1126 } 1127 1128 ixheaacd_esbr_cos_sin_mod(real_subband, qmf_bank, 1129 qmf_dec_tables_ptr->esbr_w_16, 1130 qmf_dec_tables_ptr->dig_rev_table4_16); 1131 1132 tcos = qmf_bank->esbr_t_cos; 1133 1134 for (i = (qmf_bank->usb - qmf_bank->lsb - 1); i >= 0; i--) { 1135 WORD32 cosh, sinh; 1136 WORD32 re, im; 1137 1138 re = *real_subband; 1139 im = *imag_subband; 1140 cosh = *tcos++; 1141 sinh = *tcos++; 1142 *real_subband++ = (WORD32)((ixheaacd_add64(ixheaacd_mult64(re, cosh), 1143 ixheaacd_mult64(im, sinh))) >> 1144 31); 1145 *imag_subband++ = 1146 (WORD32)((ixheaacd_sub64_sat(ixheaacd_mult64(im, cosh), 1147 ixheaacd_mult64(re, sinh))) >> 1148 31); 1149 } 1150 } 1151 1152 VOID ixheaacd_esbr_qmfsyn64_winadd(WORD32 *tmp1, WORD32 *tmp2, WORD32 *inp1, 1153 WORD32 *sample_buffer, WORD32 ch_fac) { 1154 WORD32 k; 1155 1156 for (k = 0; k < 64; k++) { 1157 WORD64 syn_out = 0; 1158 1159 syn_out = 1160 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp1[0 + k], inp1[k + 0])); 1161 syn_out = 1162 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp1[256 + k], inp1[k + 128])); 1163 syn_out = 1164 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp1[512 + k], inp1[k + 256])); 1165 syn_out = 1166 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp1[768 + k], inp1[k + 384])); 1167 syn_out = 1168 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp1[1024 + k], inp1[k + 512])); 1169 1170 syn_out = 1171 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp2[128 + k], inp1[k + 64])); 1172 syn_out = 1173 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp2[384 + k], inp1[k + 192])); 1174 syn_out = 1175 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp2[640 + k], inp1[k + 320])); 1176 syn_out = 1177 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp2[896 + k], inp1[k + 448])); 1178 syn_out = 1179 ixheaacd_add64(syn_out, ixheaacd_mult64(tmp2[1152 + k], inp1[k + 576])); 1180 1181 sample_buffer[ch_fac * k] = (WORD32)(syn_out >> 31); 1182 } 1183 } 1184 1185 VOID ixheaacd_shiftrountine(WORD32 *qmf_real, WORD32 *qmf_imag, WORD32 len, 1186 WORD32 common_shift) { 1187 WORD32 treal, timag; 1188 WORD32 j; 1189 1190 if (common_shift < 0) { 1191 WORD32 cshift = -common_shift; 1192 cshift = ixheaacd_min32(cshift, 31); 1193 for (j = len - 1; j >= 0; j--) { 1194 treal = *qmf_real; 1195 timag = *qmf_imag; 1196 1197 treal = (ixheaacd_shr32(treal, cshift)); 1198 timag = (ixheaacd_shr32(timag, cshift)); 1199 1200 *qmf_real++ = treal; 1201 *qmf_imag++ = timag; 1202 } 1203 } else { 1204 for (j = len - 1; j >= 0; j--) { 1205 treal = (ixheaacd_shl32_sat(*qmf_real, common_shift)); 1206 timag = (ixheaacd_shl32_sat(*qmf_imag, common_shift)); 1207 *qmf_real++ = treal; 1208 *qmf_imag++ = timag; 1209 } 1210 } 1211 } 1212 1213 VOID ixheaacd_shiftrountine_with_rnd_hq(WORD32 *qmf_real, WORD32 *qmf_imag, 1214 WORD32 *filter_states, WORD32 len, 1215 WORD32 shift) { 1216 WORD32 *filter_states_rev = filter_states + len; 1217 WORD32 treal, timag; 1218 WORD32 j; 1219 1220 for (j = (len - 1); j >= 0; j -= 2) { 1221 WORD32 r1, r2, i1, i2; 1222 i2 = qmf_imag[j]; 1223 r2 = qmf_real[j]; 1224 r1 = *qmf_real++; 1225 i1 = *qmf_imag++; 1226 1227 timag = ixheaacd_add32(i1, r1); 1228 timag = (ixheaacd_shl32_sat(timag, shift)); 1229 filter_states_rev[j] = timag; 1230 1231 treal = ixheaacd_sub32(i2, r2); 1232 treal = (ixheaacd_shl32_sat(treal, shift)); 1233 filter_states[j] = treal; 1234 1235 treal = ixheaacd_sub32(i1, r1); 1236 treal = (ixheaacd_shl32_sat(treal, shift)); 1237 *filter_states++ = treal; 1238 1239 timag = ixheaacd_add32(i2, r2); 1240 timag = (ixheaacd_shl32_sat(timag, shift)); 1241 *filter_states_rev++ = timag; 1242 } 1243 } 1244 1245 VOID ixheaacd_radix4bfly(const WORD16 *w, WORD32 *x, WORD32 index1, 1246 WORD32 index) { 1247 int i; 1248 WORD32 l1, l2, h2, fft_jmp; 1249 WORD32 xt0_0, yt0_0, xt1_0, yt1_0, xt2_0, yt2_0; 1250 WORD32 xh0_0, xh1_0, xh20_0, xh21_0, xl0_0, xl1_0, xl20_0, xl21_0; 1251 WORD32 x_0, x_1, x_l1_0, x_l1_1, x_l2_0, x_l2_1; 1252 WORD32 x_h2_0, x_h2_1; 1253 WORD16 si10, si20, si30, co10, co20, co30; 1254 1255 WORD32 mul_1, mul_2, mul_3, mul_4, mul_5, mul_6; 1256 WORD32 mul_7, mul_8, mul_9, mul_10, mul_11, mul_12; 1257 WORD32 *x_l1; 1258 WORD32 *x_l2; 1259 WORD32 *x_h2; 1260 const WORD16 *w_ptr = w; 1261 WORD32 i1; 1262 1263 h2 = index << 1; 1264 l1 = index << 2; 1265 l2 = (index << 2) + (index << 1); 1266 1267 x_l1 = &(x[l1]); 1268 x_l2 = &(x[l2]); 1269 x_h2 = &(x[h2]); 1270 1271 fft_jmp = 6 * (index); 1272 1273 for (i1 = 0; i1 < index1; i1++) { 1274 for (i = 0; i < index; i++) { 1275 si10 = (*w_ptr++); 1276 co10 = (*w_ptr++); 1277 si20 = (*w_ptr++); 1278 co20 = (*w_ptr++); 1279 si30 = (*w_ptr++); 1280 co30 = (*w_ptr++); 1281 1282 x_0 = x[0]; 1283 x_h2_0 = x[h2]; 1284 x_l1_0 = x[l1]; 1285 x_l2_0 = x[l2]; 1286 1287 xh0_0 = x_0 + x_l1_0; 1288 xl0_0 = x_0 - x_l1_0; 1289 1290 xh20_0 = x_h2_0 + x_l2_0; 1291 xl20_0 = x_h2_0 - x_l2_0; 1292 1293 x[0] = xh0_0 + xh20_0; 1294 xt0_0 = xh0_0 - xh20_0; 1295 1296 x_1 = x[1]; 1297 x_h2_1 = x[h2 + 1]; 1298 x_l1_1 = x[l1 + 1]; 1299 x_l2_1 = x[l2 + 1]; 1300 1301 xh1_0 = x_1 + x_l1_1; 1302 xl1_0 = x_1 - x_l1_1; 1303 1304 xh21_0 = x_h2_1 + x_l2_1; 1305 xl21_0 = x_h2_1 - x_l2_1; 1306 1307 x[1] = xh1_0 + xh21_0; 1308 yt0_0 = xh1_0 - xh21_0; 1309 1310 xt1_0 = xl0_0 + xl21_0; 1311 xt2_0 = xl0_0 - xl21_0; 1312 1313 yt2_0 = xl1_0 + xl20_0; 1314 yt1_0 = xl1_0 - xl20_0; 1315 1316 mul_11 = ixheaacd_mult32x16in32(xt2_0, co30); 1317 mul_3 = ixheaacd_mult32x16in32(yt2_0, si30); 1318 x[l2] = (mul_3 + mul_11) << RADIXSHIFT; 1319 1320 mul_5 = ixheaacd_mult32x16in32(xt2_0, si30); 1321 mul_9 = ixheaacd_mult32x16in32(yt2_0, co30); 1322 x[l2 + 1] = (mul_9 - mul_5) << RADIXSHIFT; 1323 1324 mul_12 = ixheaacd_mult32x16in32(xt0_0, co20); 1325 mul_2 = ixheaacd_mult32x16in32(yt0_0, si20); 1326 x[l1] = (mul_2 + mul_12) << RADIXSHIFT; 1327 1328 mul_6 = ixheaacd_mult32x16in32(xt0_0, si20); 1329 mul_8 = ixheaacd_mult32x16in32(yt0_0, co20); 1330 x[l1 + 1] = (mul_8 - mul_6) << RADIXSHIFT; 1331 1332 mul_4 = ixheaacd_mult32x16in32(xt1_0, co10); 1333 mul_1 = ixheaacd_mult32x16in32(yt1_0, si10); 1334 x[h2] = (mul_1 + mul_4) << RADIXSHIFT; 1335 1336 mul_10 = ixheaacd_mult32x16in32(xt1_0, si10); 1337 mul_7 = ixheaacd_mult32x16in32(yt1_0, co10); 1338 x[h2 + 1] = (mul_7 - mul_10) << RADIXSHIFT; 1339 1340 x += 2; 1341 } 1342 x += fft_jmp; 1343 w_ptr = w_ptr - fft_jmp; 1344 } 1345 } 1346
