dlsch_llr_computation.c 442 KB
Newer Older
1
/*
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
 * Licensed to the OpenAirInterface (OAI) Software Alliance under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The OpenAirInterface Software Alliance licenses this file to You under
 * the OAI Public License, Version 1.0  (the "License"); you may not use this file
 * except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.openairinterface.org/?page_id=698
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *-------------------------------------------------------------------------------
 * For more information about the OpenAirInterface (OAI) Software Alliance:
 *      contact@openairinterface.org
 */
21 22

/*! \file PHY/LTE_TRANSPORT/dlsch_llr_computation.c
23
 * \brief Top-level routines for LLR computation of the PDSCH physical channel from 36-211, V8.6 2009-03
24
 * \author R. Knopp, F. Kaltenberger,A. Bhamri, S. Aubert, S. Wagner, X Jiang
25 26 27 28 29 30 31
 * \date 2011
 * \version 0.1
 * \company Eurecom
 * \email: knopp@eurecom.fr,florian.kaltenberger@eurecom.fr,ankit.bhamri@eurecom.fr,sebastien.aubert@eurecom.fr, sebastian.wagner@eurecom.fr
 * \note
 * \warning
 */
32 33

#include "PHY/defs.h"
34
#include "PHY/TOOLS/defs.h"
35 36 37
#include "PHY/extern.h"
#include "defs.h"
#include "extern.h"
38
#include "PHY/sse_intrin.h"
39

40 41 42
//#define DEBUG_LLR_SIC


43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
int16_t zero[8] __attribute__ ((aligned(16))) = {0,0,0,0,0,0,0,0};
int16_t ones[8] __attribute__ ((aligned(16))) = {0xffff,0xffff,0xffff,0xffff,0xffff,0xffff,0xffff,0xffff};
#if defined(__x86_64__) || defined(__i386__)
__m128i rho_rpi __attribute__ ((aligned(16)));
__m128i rho_rmi __attribute__ ((aligned(16)));
__m128i rho_rpi_1_1 __attribute__ ((aligned(16)));
__m128i rho_rpi_1_3 __attribute__ ((aligned(16)));
__m128i rho_rpi_1_5 __attribute__ ((aligned(16)));
__m128i rho_rpi_1_7 __attribute__ ((aligned(16)));
__m128i rho_rpi_3_1 __attribute__ ((aligned(16)));
__m128i rho_rpi_3_3 __attribute__ ((aligned(16)));
__m128i rho_rpi_3_5 __attribute__ ((aligned(16)));
__m128i rho_rpi_3_7 __attribute__ ((aligned(16)));
__m128i rho_rpi_5_1 __attribute__ ((aligned(16)));
__m128i rho_rpi_5_3 __attribute__ ((aligned(16)));
__m128i rho_rpi_5_5 __attribute__ ((aligned(16)));
__m128i rho_rpi_5_7 __attribute__ ((aligned(16)));
__m128i rho_rpi_7_1 __attribute__ ((aligned(16)));
__m128i rho_rpi_7_3 __attribute__ ((aligned(16)));
__m128i rho_rpi_7_5 __attribute__ ((aligned(16)));
__m128i rho_rpi_7_7 __attribute__ ((aligned(16)));
__m128i rho_rmi_1_1 __attribute__ ((aligned(16)));
__m128i rho_rmi_1_3 __attribute__ ((aligned(16)));
__m128i rho_rmi_1_5 __attribute__ ((aligned(16)));
__m128i rho_rmi_1_7 __attribute__ ((aligned(16)));
__m128i rho_rmi_3_1 __attribute__ ((aligned(16)));
__m128i rho_rmi_3_3 __attribute__ ((aligned(16)));
__m128i rho_rmi_3_5 __attribute__ ((aligned(16)));
__m128i rho_rmi_3_7 __attribute__ ((aligned(16)));
__m128i rho_rmi_5_1 __attribute__ ((aligned(16)));
__m128i rho_rmi_5_3 __attribute__ ((aligned(16)));
__m128i rho_rmi_5_5 __attribute__ ((aligned(16)));
__m128i rho_rmi_5_7 __attribute__ ((aligned(16)));
__m128i rho_rmi_7_1 __attribute__ ((aligned(16)));
__m128i rho_rmi_7_3 __attribute__ ((aligned(16)));
__m128i rho_rmi_7_5 __attribute__ ((aligned(16)));
__m128i rho_rmi_7_7 __attribute__ ((aligned(16)));

__m128i psi_r_m7_m7 __attribute__ ((aligned(16)));
__m128i psi_r_m7_m5 __attribute__ ((aligned(16)));
__m128i psi_r_m7_m3 __attribute__ ((aligned(16)));
__m128i psi_r_m7_m1 __attribute__ ((aligned(16)));
__m128i psi_r_m7_p1 __attribute__ ((aligned(16)));
__m128i psi_r_m7_p3 __attribute__ ((aligned(16)));
__m128i psi_r_m7_p5 __attribute__ ((aligned(16)));
__m128i psi_r_m7_p7 __attribute__ ((aligned(16)));
__m128i psi_r_m5_m7 __attribute__ ((aligned(16)));
__m128i psi_r_m5_m5 __attribute__ ((aligned(16)));
__m128i psi_r_m5_m3 __attribute__ ((aligned(16)));
__m128i psi_r_m5_m1 __attribute__ ((aligned(16)));
__m128i psi_r_m5_p1 __attribute__ ((aligned(16)));
__m128i psi_r_m5_p3 __attribute__ ((aligned(16)));
__m128i psi_r_m5_p5 __attribute__ ((aligned(16)));
__m128i psi_r_m5_p7 __attribute__ ((aligned(16)));
__m128i psi_r_m3_m7 __attribute__ ((aligned(16)));
__m128i psi_r_m3_m5 __attribute__ ((aligned(16)));
__m128i psi_r_m3_m3 __attribute__ ((aligned(16)));
__m128i psi_r_m3_m1 __attribute__ ((aligned(16)));
__m128i psi_r_m3_p1 __attribute__ ((aligned(16)));
__m128i psi_r_m3_p3 __attribute__ ((aligned(16)));
__m128i psi_r_m3_p5 __attribute__ ((aligned(16)));
__m128i psi_r_m3_p7 __attribute__ ((aligned(16)));
__m128i psi_r_m1_m7 __attribute__ ((aligned(16)));
__m128i psi_r_m1_m5 __attribute__ ((aligned(16)));
__m128i psi_r_m1_m3 __attribute__ ((aligned(16)));
__m128i psi_r_m1_m1 __attribute__ ((aligned(16)));
__m128i psi_r_m1_p1 __attribute__ ((aligned(16)));
__m128i psi_r_m1_p3 __attribute__ ((aligned(16)));
__m128i psi_r_m1_p5 __attribute__ ((aligned(16)));
__m128i psi_r_m1_p7 __attribute__ ((aligned(16)));
__m128i psi_r_p1_m7 __attribute__ ((aligned(16)));
__m128i psi_r_p1_m5 __attribute__ ((aligned(16)));
__m128i psi_r_p1_m3 __attribute__ ((aligned(16)));
__m128i psi_r_p1_m1 __attribute__ ((aligned(16)));
__m128i psi_r_p1_p1 __attribute__ ((aligned(16)));
__m128i psi_r_p1_p3 __attribute__ ((aligned(16)));
__m128i psi_r_p1_p5 __attribute__ ((aligned(16)));
__m128i psi_r_p1_p7 __attribute__ ((aligned(16)));
__m128i psi_r_p3_m7 __attribute__ ((aligned(16)));
__m128i psi_r_p3_m5 __attribute__ ((aligned(16)));
__m128i psi_r_p3_m3 __attribute__ ((aligned(16)));
__m128i psi_r_p3_m1 __attribute__ ((aligned(16)));
__m128i psi_r_p3_p1 __attribute__ ((aligned(16)));
__m128i psi_r_p3_p3 __attribute__ ((aligned(16)));
__m128i psi_r_p3_p5 __attribute__ ((aligned(16)));
__m128i psi_r_p3_p7 __attribute__ ((aligned(16)));
__m128i psi_r_p5_m7 __attribute__ ((aligned(16)));
__m128i psi_r_p5_m5 __attribute__ ((aligned(16)));
__m128i psi_r_p5_m3 __attribute__ ((aligned(16)));
__m128i psi_r_p5_m1 __attribute__ ((aligned(16)));
__m128i psi_r_p5_p1 __attribute__ ((aligned(16)));
__m128i psi_r_p5_p3 __attribute__ ((aligned(16)));
__m128i psi_r_p5_p5 __attribute__ ((aligned(16)));
__m128i psi_r_p5_p7 __attribute__ ((aligned(16)));
__m128i psi_r_p7_m7 __attribute__ ((aligned(16)));
__m128i psi_r_p7_m5 __attribute__ ((aligned(16)));
__m128i psi_r_p7_m3 __attribute__ ((aligned(16)));
__m128i psi_r_p7_m1 __attribute__ ((aligned(16)));
__m128i psi_r_p7_p1 __attribute__ ((aligned(16)));
__m128i psi_r_p7_p3 __attribute__ ((aligned(16)));
__m128i psi_r_p7_p5 __attribute__ ((aligned(16)));
__m128i psi_r_p7_p7 __attribute__ ((aligned(16)));

__m128i psi_i_m7_m7 __attribute__ ((aligned(16)));
__m128i psi_i_m7_m5 __attribute__ ((aligned(16)));
__m128i psi_i_m7_m3 __attribute__ ((aligned(16)));
__m128i psi_i_m7_m1 __attribute__ ((aligned(16)));
__m128i psi_i_m7_p1 __attribute__ ((aligned(16)));
__m128i psi_i_m7_p3 __attribute__ ((aligned(16)));
__m128i psi_i_m7_p5 __attribute__ ((aligned(16)));
__m128i psi_i_m7_p7 __attribute__ ((aligned(16)));
__m128i psi_i_m5_m7 __attribute__ ((aligned(16)));
__m128i psi_i_m5_m5 __attribute__ ((aligned(16)));
__m128i psi_i_m5_m3 __attribute__ ((aligned(16)));
__m128i psi_i_m5_m1 __attribute__ ((aligned(16)));
__m128i psi_i_m5_p1 __attribute__ ((aligned(16)));
__m128i psi_i_m5_p3 __attribute__ ((aligned(16)));
__m128i psi_i_m5_p5 __attribute__ ((aligned(16)));
__m128i psi_i_m5_p7 __attribute__ ((aligned(16)));
__m128i psi_i_m3_m7 __attribute__ ((aligned(16)));
__m128i psi_i_m3_m5 __attribute__ ((aligned(16)));
__m128i psi_i_m3_m3 __attribute__ ((aligned(16)));
__m128i psi_i_m3_m1 __attribute__ ((aligned(16)));
__m128i psi_i_m3_p1 __attribute__ ((aligned(16)));
__m128i psi_i_m3_p3 __attribute__ ((aligned(16)));
__m128i psi_i_m3_p5 __attribute__ ((aligned(16)));
__m128i psi_i_m3_p7 __attribute__ ((aligned(16)));
__m128i psi_i_m1_m7 __attribute__ ((aligned(16)));
__m128i psi_i_m1_m5 __attribute__ ((aligned(16)));
__m128i psi_i_m1_m3 __attribute__ ((aligned(16)));
__m128i psi_i_m1_m1 __attribute__ ((aligned(16)));
__m128i psi_i_m1_p1 __attribute__ ((aligned(16)));
__m128i psi_i_m1_p3 __attribute__ ((aligned(16)));
__m128i psi_i_m1_p5 __attribute__ ((aligned(16)));
__m128i psi_i_m1_p7 __attribute__ ((aligned(16)));
__m128i psi_i_p1_m7 __attribute__ ((aligned(16)));
__m128i psi_i_p1_m5 __attribute__ ((aligned(16)));
__m128i psi_i_p1_m3 __attribute__ ((aligned(16)));
__m128i psi_i_p1_m1 __attribute__ ((aligned(16)));
__m128i psi_i_p1_p1 __attribute__ ((aligned(16)));
__m128i psi_i_p1_p3 __attribute__ ((aligned(16)));
__m128i psi_i_p1_p5 __attribute__ ((aligned(16)));
__m128i psi_i_p1_p7 __attribute__ ((aligned(16)));
__m128i psi_i_p3_m7 __attribute__ ((aligned(16)));
__m128i psi_i_p3_m5 __attribute__ ((aligned(16)));
__m128i psi_i_p3_m3 __attribute__ ((aligned(16)));
__m128i psi_i_p3_m1 __attribute__ ((aligned(16)));
__m128i psi_i_p3_p1 __attribute__ ((aligned(16)));
__m128i psi_i_p3_p3 __attribute__ ((aligned(16)));
__m128i psi_i_p3_p5 __attribute__ ((aligned(16)));
__m128i psi_i_p3_p7 __attribute__ ((aligned(16)));
__m128i psi_i_p5_m7 __attribute__ ((aligned(16)));
__m128i psi_i_p5_m5 __attribute__ ((aligned(16)));
__m128i psi_i_p5_m3 __attribute__ ((aligned(16)));
__m128i psi_i_p5_m1 __attribute__ ((aligned(16)));
__m128i psi_i_p5_p1 __attribute__ ((aligned(16)));
__m128i psi_i_p5_p3 __attribute__ ((aligned(16)));
__m128i psi_i_p5_p5 __attribute__ ((aligned(16)));
__m128i psi_i_p5_p7 __attribute__ ((aligned(16)));
__m128i psi_i_p7_m7 __attribute__ ((aligned(16)));
__m128i psi_i_p7_m5 __attribute__ ((aligned(16)));
__m128i psi_i_p7_m3 __attribute__ ((aligned(16)));
__m128i psi_i_p7_m1 __attribute__ ((aligned(16)));
__m128i psi_i_p7_p1 __attribute__ ((aligned(16)));
__m128i psi_i_p7_p3 __attribute__ ((aligned(16)));
__m128i psi_i_p7_p5 __attribute__ ((aligned(16)));
__m128i psi_i_p7_p7 __attribute__ ((aligned(16)));

__m128i a_r_m7_m7 __attribute__ ((aligned(16)));
__m128i a_r_m7_m5 __attribute__ ((aligned(16)));
__m128i a_r_m7_m3 __attribute__ ((aligned(16)));
__m128i a_r_m7_m1 __attribute__ ((aligned(16)));
__m128i a_r_m7_p1 __attribute__ ((aligned(16)));
__m128i a_r_m7_p3 __attribute__ ((aligned(16)));
__m128i a_r_m7_p5 __attribute__ ((aligned(16)));
__m128i a_r_m7_p7 __attribute__ ((aligned(16)));
__m128i a_r_m5_m7 __attribute__ ((aligned(16)));
__m128i a_r_m5_m5 __attribute__ ((aligned(16)));
__m128i a_r_m5_m3 __attribute__ ((aligned(16)));
__m128i a_r_m5_m1 __attribute__ ((aligned(16)));
__m128i a_r_m5_p1 __attribute__ ((aligned(16)));
__m128i a_r_m5_p3 __attribute__ ((aligned(16)));
__m128i a_r_m5_p5 __attribute__ ((aligned(16)));
__m128i a_r_m5_p7 __attribute__ ((aligned(16)));
__m128i a_r_m3_m7 __attribute__ ((aligned(16)));
__m128i a_r_m3_m5 __attribute__ ((aligned(16)));
__m128i a_r_m3_m3 __attribute__ ((aligned(16)));
__m128i a_r_m3_m1 __attribute__ ((aligned(16)));
__m128i a_r_m3_p1 __attribute__ ((aligned(16)));
__m128i a_r_m3_p3 __attribute__ ((aligned(16)));
__m128i a_r_m3_p5 __attribute__ ((aligned(16)));
__m128i a_r_m3_p7 __attribute__ ((aligned(16)));
__m128i a_r_m1_m7 __attribute__ ((aligned(16)));
__m128i a_r_m1_m5 __attribute__ ((aligned(16)));
__m128i a_r_m1_m3 __attribute__ ((aligned(16)));
__m128i a_r_m1_m1 __attribute__ ((aligned(16)));
__m128i a_r_m1_p1 __attribute__ ((aligned(16)));
__m128i a_r_m1_p3 __attribute__ ((aligned(16)));
__m128i a_r_m1_p5 __attribute__ ((aligned(16)));
__m128i a_r_m1_p7 __attribute__ ((aligned(16)));
__m128i a_r_p1_m7 __attribute__ ((aligned(16)));
__m128i a_r_p1_m5 __attribute__ ((aligned(16)));
__m128i a_r_p1_m3 __attribute__ ((aligned(16)));
__m128i a_r_p1_m1 __attribute__ ((aligned(16)));
__m128i a_r_p1_p1 __attribute__ ((aligned(16)));
__m128i a_r_p1_p3 __attribute__ ((aligned(16)));
__m128i a_r_p1_p5 __attribute__ ((aligned(16)));
__m128i a_r_p1_p7 __attribute__ ((aligned(16)));
__m128i a_r_p3_m7 __attribute__ ((aligned(16)));
__m128i a_r_p3_m5 __attribute__ ((aligned(16)));
__m128i a_r_p3_m3 __attribute__ ((aligned(16)));
__m128i a_r_p3_m1 __attribute__ ((aligned(16)));
__m128i a_r_p3_p1 __attribute__ ((aligned(16)));
__m128i a_r_p3_p3 __attribute__ ((aligned(16)));
__m128i a_r_p3_p5 __attribute__ ((aligned(16)));
__m128i a_r_p3_p7 __attribute__ ((aligned(16)));
__m128i a_r_p5_m7 __attribute__ ((aligned(16)));
__m128i a_r_p5_m5 __attribute__ ((aligned(16)));
__m128i a_r_p5_m3 __attribute__ ((aligned(16)));
__m128i a_r_p5_m1 __attribute__ ((aligned(16)));
__m128i a_r_p5_p1 __attribute__ ((aligned(16)));
__m128i a_r_p5_p3 __attribute__ ((aligned(16)));
__m128i a_r_p5_p5 __attribute__ ((aligned(16)));
__m128i a_r_p5_p7 __attribute__ ((aligned(16)));
__m128i a_r_p7_m7 __attribute__ ((aligned(16)));
__m128i a_r_p7_m5 __attribute__ ((aligned(16)));
__m128i a_r_p7_m3 __attribute__ ((aligned(16)));
__m128i a_r_p7_m1 __attribute__ ((aligned(16)));
__m128i a_r_p7_p1 __attribute__ ((aligned(16)));
__m128i a_r_p7_p3 __attribute__ ((aligned(16)));
__m128i a_r_p7_p5 __attribute__ ((aligned(16)));
__m128i a_r_p7_p7 __attribute__ ((aligned(16)));

__m128i a_i_m7_m7 __attribute__ ((aligned(16)));
__m128i a_i_m7_m5 __attribute__ ((aligned(16)));
__m128i a_i_m7_m3 __attribute__ ((aligned(16)));
__m128i a_i_m7_m1 __attribute__ ((aligned(16)));
__m128i a_i_m7_p1 __attribute__ ((aligned(16)));
__m128i a_i_m7_p3 __attribute__ ((aligned(16)));
__m128i a_i_m7_p5 __attribute__ ((aligned(16)));
__m128i a_i_m7_p7 __attribute__ ((aligned(16)));
__m128i a_i_m5_m7 __attribute__ ((aligned(16)));
__m128i a_i_m5_m5 __attribute__ ((aligned(16)));
__m128i a_i_m5_m3 __attribute__ ((aligned(16)));
__m128i a_i_m5_m1 __attribute__ ((aligned(16)));
__m128i a_i_m5_p1 __attribute__ ((aligned(16)));
__m128i a_i_m5_p3 __attribute__ ((aligned(16)));
__m128i a_i_m5_p5 __attribute__ ((aligned(16)));
__m128i a_i_m5_p7 __attribute__ ((aligned(16)));
__m128i a_i_m3_m7 __attribute__ ((aligned(16)));
__m128i a_i_m3_m5 __attribute__ ((aligned(16)));
__m128i a_i_m3_m3 __attribute__ ((aligned(16)));
__m128i a_i_m3_m1 __attribute__ ((aligned(16)));
__m128i a_i_m3_p1 __attribute__ ((aligned(16)));
__m128i a_i_m3_p3 __attribute__ ((aligned(16)));
__m128i a_i_m3_p5 __attribute__ ((aligned(16)));
__m128i a_i_m3_p7 __attribute__ ((aligned(16)));
__m128i a_i_m1_m7 __attribute__ ((aligned(16)));
__m128i a_i_m1_m5 __attribute__ ((aligned(16)));
__m128i a_i_m1_m3 __attribute__ ((aligned(16)));
__m128i a_i_m1_m1 __attribute__ ((aligned(16)));
__m128i a_i_m1_p1 __attribute__ ((aligned(16)));
__m128i a_i_m1_p3 __attribute__ ((aligned(16)));
__m128i a_i_m1_p5 __attribute__ ((aligned(16)));
__m128i a_i_m1_p7 __attribute__ ((aligned(16)));
__m128i a_i_p1_m7 __attribute__ ((aligned(16)));
__m128i a_i_p1_m5 __attribute__ ((aligned(16)));
__m128i a_i_p1_m3 __attribute__ ((aligned(16)));
__m128i a_i_p1_m1 __attribute__ ((aligned(16)));
__m128i a_i_p1_p1 __attribute__ ((aligned(16)));
__m128i a_i_p1_p3 __attribute__ ((aligned(16)));
__m128i a_i_p1_p5 __attribute__ ((aligned(16)));
__m128i a_i_p1_p7 __attribute__ ((aligned(16)));
__m128i a_i_p3_m7 __attribute__ ((aligned(16)));
__m128i a_i_p3_m5 __attribute__ ((aligned(16)));
__m128i a_i_p3_m3 __attribute__ ((aligned(16)));
__m128i a_i_p3_m1 __attribute__ ((aligned(16)));
__m128i a_i_p3_p1 __attribute__ ((aligned(16)));
__m128i a_i_p3_p3 __attribute__ ((aligned(16)));
__m128i a_i_p3_p5 __attribute__ ((aligned(16)));
__m128i a_i_p3_p7 __attribute__ ((aligned(16)));
__m128i a_i_p5_m7 __attribute__ ((aligned(16)));
__m128i a_i_p5_m5 __attribute__ ((aligned(16)));
__m128i a_i_p5_m3 __attribute__ ((aligned(16)));
__m128i a_i_p5_m1 __attribute__ ((aligned(16)));
__m128i a_i_p5_p1 __attribute__ ((aligned(16)));
__m128i a_i_p5_p3 __attribute__ ((aligned(16)));
__m128i a_i_p5_p5 __attribute__ ((aligned(16)));
__m128i a_i_p5_p7 __attribute__ ((aligned(16)));
__m128i a_i_p7_m7 __attribute__ ((aligned(16)));
__m128i a_i_p7_m5 __attribute__ ((aligned(16)));
__m128i a_i_p7_m3 __attribute__ ((aligned(16)));
__m128i a_i_p7_m1 __attribute__ ((aligned(16)));
__m128i a_i_p7_p1 __attribute__ ((aligned(16)));
__m128i a_i_p7_p3 __attribute__ ((aligned(16)));
__m128i a_i_p7_p5 __attribute__ ((aligned(16)));
__m128i a_i_p7_p7 __attribute__ ((aligned(16)));

__m128i psi_a_m7_m7 __attribute__ ((aligned(16)));
__m128i psi_a_m7_m5 __attribute__ ((aligned(16)));
__m128i psi_a_m7_m3 __attribute__ ((aligned(16)));
__m128i psi_a_m7_m1 __attribute__ ((aligned(16)));
__m128i psi_a_m7_p1 __attribute__ ((aligned(16)));
__m128i psi_a_m7_p3 __attribute__ ((aligned(16)));
__m128i psi_a_m7_p5 __attribute__ ((aligned(16)));
__m128i psi_a_m7_p7 __attribute__ ((aligned(16)));
__m128i psi_a_m5_m7 __attribute__ ((aligned(16)));
__m128i psi_a_m5_m5 __attribute__ ((aligned(16)));
__m128i psi_a_m5_m3 __attribute__ ((aligned(16)));
__m128i psi_a_m5_m1 __attribute__ ((aligned(16)));
__m128i psi_a_m5_p1 __attribute__ ((aligned(16)));
__m128i psi_a_m5_p3 __attribute__ ((aligned(16)));
__m128i psi_a_m5_p5 __attribute__ ((aligned(16)));
__m128i psi_a_m5_p7 __attribute__ ((aligned(16)));
__m128i psi_a_m3_m7 __attribute__ ((aligned(16)));
__m128i psi_a_m3_m5 __attribute__ ((aligned(16)));
__m128i psi_a_m3_m3 __attribute__ ((aligned(16)));
__m128i psi_a_m3_m1 __attribute__ ((aligned(16)));
__m128i psi_a_m3_p1 __attribute__ ((aligned(16)));
__m128i psi_a_m3_p3 __attribute__ ((aligned(16)));
__m128i psi_a_m3_p5 __attribute__ ((aligned(16)));
__m128i psi_a_m3_p7 __attribute__ ((aligned(16)));
__m128i psi_a_m1_m7 __attribute__ ((aligned(16)));
__m128i psi_a_m1_m5 __attribute__ ((aligned(16)));
__m128i psi_a_m1_m3 __attribute__ ((aligned(16)));
__m128i psi_a_m1_m1 __attribute__ ((aligned(16)));
__m128i psi_a_m1_p1 __attribute__ ((aligned(16)));
__m128i psi_a_m1_p3 __attribute__ ((aligned(16)));
__m128i psi_a_m1_p5 __attribute__ ((aligned(16)));
__m128i psi_a_m1_p7 __attribute__ ((aligned(16)));
__m128i psi_a_p1_m7 __attribute__ ((aligned(16)));
__m128i psi_a_p1_m5 __attribute__ ((aligned(16)));
__m128i psi_a_p1_m3 __attribute__ ((aligned(16)));
__m128i psi_a_p1_m1 __attribute__ ((aligned(16)));
__m128i psi_a_p1_p1 __attribute__ ((aligned(16)));
__m128i psi_a_p1_p3 __attribute__ ((aligned(16)));
__m128i psi_a_p1_p5 __attribute__ ((aligned(16)));
__m128i psi_a_p1_p7 __attribute__ ((aligned(16)));
__m128i psi_a_p3_m7 __attribute__ ((aligned(16)));
__m128i psi_a_p3_m5 __attribute__ ((aligned(16)));
__m128i psi_a_p3_m3 __attribute__ ((aligned(16)));
__m128i psi_a_p3_m1 __attribute__ ((aligned(16)));
__m128i psi_a_p3_p1 __attribute__ ((aligned(16)));
__m128i psi_a_p3_p3 __attribute__ ((aligned(16)));
__m128i psi_a_p3_p5 __attribute__ ((aligned(16)));
__m128i psi_a_p3_p7 __attribute__ ((aligned(16)));
__m128i psi_a_p5_m7 __attribute__ ((aligned(16)));
__m128i psi_a_p5_m5 __attribute__ ((aligned(16)));
__m128i psi_a_p5_m3 __attribute__ ((aligned(16)));
__m128i psi_a_p5_m1 __attribute__ ((aligned(16)));
__m128i psi_a_p5_p1 __attribute__ ((aligned(16)));
__m128i psi_a_p5_p3 __attribute__ ((aligned(16)));
__m128i psi_a_p5_p5 __attribute__ ((aligned(16)));
__m128i psi_a_p5_p7 __attribute__ ((aligned(16)));
__m128i psi_a_p7_m7 __attribute__ ((aligned(16)));
__m128i psi_a_p7_m5 __attribute__ ((aligned(16)));
__m128i psi_a_p7_m3 __attribute__ ((aligned(16)));
__m128i psi_a_p7_m1 __attribute__ ((aligned(16)));
__m128i psi_a_p7_p1 __attribute__ ((aligned(16)));
__m128i psi_a_p7_p3 __attribute__ ((aligned(16)));
__m128i psi_a_p7_p5 __attribute__ ((aligned(16)));
__m128i psi_a_p7_p7 __attribute__ ((aligned(16)));

__m128i a_sq_m7_m7 __attribute__ ((aligned(16)));
__m128i a_sq_m7_m5 __attribute__ ((aligned(16)));
__m128i a_sq_m7_m3 __attribute__ ((aligned(16)));
__m128i a_sq_m7_m1 __attribute__ ((aligned(16)));
__m128i a_sq_m7_p1 __attribute__ ((aligned(16)));
__m128i a_sq_m7_p3 __attribute__ ((aligned(16)));
__m128i a_sq_m7_p5 __attribute__ ((aligned(16)));
__m128i a_sq_m7_p7 __attribute__ ((aligned(16)));
__m128i a_sq_m5_m7 __attribute__ ((aligned(16)));
__m128i a_sq_m5_m5 __attribute__ ((aligned(16)));
__m128i a_sq_m5_m3 __attribute__ ((aligned(16)));
__m128i a_sq_m5_m1 __attribute__ ((aligned(16)));
__m128i a_sq_m5_p1 __attribute__ ((aligned(16)));
__m128i a_sq_m5_p3 __attribute__ ((aligned(16)));
__m128i a_sq_m5_p5 __attribute__ ((aligned(16)));
__m128i a_sq_m5_p7 __attribute__ ((aligned(16)));
__m128i a_sq_m3_m7 __attribute__ ((aligned(16)));
__m128i a_sq_m3_m5 __attribute__ ((aligned(16)));
__m128i a_sq_m3_m3 __attribute__ ((aligned(16)));
__m128i a_sq_m3_m1 __attribute__ ((aligned(16)));
__m128i a_sq_m3_p1 __attribute__ ((aligned(16)));
__m128i a_sq_m3_p3 __attribute__ ((aligned(16)));
__m128i a_sq_m3_p5 __attribute__ ((aligned(16)));
__m128i a_sq_m3_p7 __attribute__ ((aligned(16)));
__m128i a_sq_m1_m7 __attribute__ ((aligned(16)));
__m128i a_sq_m1_m5 __attribute__ ((aligned(16)));
__m128i a_sq_m1_m3 __attribute__ ((aligned(16)));
__m128i a_sq_m1_m1 __attribute__ ((aligned(16)));
__m128i a_sq_m1_p1 __attribute__ ((aligned(16)));
__m128i a_sq_m1_p3 __attribute__ ((aligned(16)));
__m128i a_sq_m1_p5 __attribute__ ((aligned(16)));
__m128i a_sq_m1_p7 __attribute__ ((aligned(16)));
__m128i a_sq_p1_m7 __attribute__ ((aligned(16)));
__m128i a_sq_p1_m5 __attribute__ ((aligned(16)));
__m128i a_sq_p1_m3 __attribute__ ((aligned(16)));
__m128i a_sq_p1_m1 __attribute__ ((aligned(16)));
__m128i a_sq_p1_p1 __attribute__ ((aligned(16)));
__m128i a_sq_p1_p3 __attribute__ ((aligned(16)));
__m128i a_sq_p1_p5 __attribute__ ((aligned(16)));
__m128i a_sq_p1_p7 __attribute__ ((aligned(16)));
__m128i a_sq_p3_m7 __attribute__ ((aligned(16)));
__m128i a_sq_p3_m5 __attribute__ ((aligned(16)));
__m128i a_sq_p3_m3 __attribute__ ((aligned(16)));
__m128i a_sq_p3_m1 __attribute__ ((aligned(16)));
__m128i a_sq_p3_p1 __attribute__ ((aligned(16)));
__m128i a_sq_p3_p3 __attribute__ ((aligned(16)));
__m128i a_sq_p3_p5 __attribute__ ((aligned(16)));
__m128i a_sq_p3_p7 __attribute__ ((aligned(16)));
__m128i a_sq_p5_m7 __attribute__ ((aligned(16)));
__m128i a_sq_p5_m5 __attribute__ ((aligned(16)));
__m128i a_sq_p5_m3 __attribute__ ((aligned(16)));
__m128i a_sq_p5_m1 __attribute__ ((aligned(16)));
__m128i a_sq_p5_p1 __attribute__ ((aligned(16)));
__m128i a_sq_p5_p3 __attribute__ ((aligned(16)));
__m128i a_sq_p5_p5 __attribute__ ((aligned(16)));
__m128i a_sq_p5_p7 __attribute__ ((aligned(16)));
__m128i a_sq_p7_m7 __attribute__ ((aligned(16)));
__m128i a_sq_p7_m5 __attribute__ ((aligned(16)));
__m128i a_sq_p7_m3 __attribute__ ((aligned(16)));
__m128i a_sq_p7_m1 __attribute__ ((aligned(16)));
__m128i a_sq_p7_p1 __attribute__ ((aligned(16)));
__m128i a_sq_p7_p3 __attribute__ ((aligned(16)));
__m128i a_sq_p7_p5 __attribute__ ((aligned(16)));
__m128i a_sq_p7_p7 __attribute__ ((aligned(16)));

__m128i bit_met_m7_m7 __attribute__ ((aligned(16)));
__m128i bit_met_m7_m5 __attribute__ ((aligned(16)));
__m128i bit_met_m7_m3 __attribute__ ((aligned(16)));
__m128i bit_met_m7_m1 __attribute__ ((aligned(16)));
__m128i bit_met_m7_p1 __attribute__ ((aligned(16)));
__m128i bit_met_m7_p3 __attribute__ ((aligned(16)));
__m128i bit_met_m7_p5 __attribute__ ((aligned(16)));
__m128i bit_met_m7_p7 __attribute__ ((aligned(16)));
__m128i bit_met_m5_m7 __attribute__ ((aligned(16)));
__m128i bit_met_m5_m5 __attribute__ ((aligned(16)));
__m128i bit_met_m5_m3 __attribute__ ((aligned(16)));
__m128i bit_met_m5_m1 __attribute__ ((aligned(16)));
__m128i bit_met_m5_p1 __attribute__ ((aligned(16)));
__m128i bit_met_m5_p3 __attribute__ ((aligned(16)));
__m128i bit_met_m5_p5 __attribute__ ((aligned(16)));
__m128i bit_met_m5_p7 __attribute__ ((aligned(16)));
__m128i bit_met_m3_m7 __attribute__ ((aligned(16)));
__m128i bit_met_m3_m5 __attribute__ ((aligned(16)));
__m128i bit_met_m3_m3 __attribute__ ((aligned(16)));
__m128i bit_met_m3_m1 __attribute__ ((aligned(16)));
__m128i bit_met_m3_p1 __attribute__ ((aligned(16)));
__m128i bit_met_m3_p3 __attribute__ ((aligned(16)));
__m128i bit_met_m3_p5 __attribute__ ((aligned(16)));
__m128i bit_met_m3_p7 __attribute__ ((aligned(16)));
__m128i bit_met_m1_m7 __attribute__ ((aligned(16)));
__m128i bit_met_m1_m5 __attribute__ ((aligned(16)));
__m128i bit_met_m1_m3 __attribute__ ((aligned(16)));
__m128i bit_met_m1_m1 __attribute__ ((aligned(16)));
__m128i bit_met_m1_p1 __attribute__ ((aligned(16)));
__m128i bit_met_m1_p3 __attribute__ ((aligned(16)));
__m128i bit_met_m1_p5 __attribute__ ((aligned(16)));
__m128i bit_met_m1_p7 __attribute__ ((aligned(16)));
__m128i bit_met_p1_m7 __attribute__ ((aligned(16)));
__m128i bit_met_p1_m5 __attribute__ ((aligned(16)));
__m128i bit_met_p1_m3 __attribute__ ((aligned(16)));
__m128i bit_met_p1_m1 __attribute__ ((aligned(16)));
__m128i bit_met_p1_p1 __attribute__ ((aligned(16)));
__m128i bit_met_p1_p3 __attribute__ ((aligned(16)));
__m128i bit_met_p1_p5 __attribute__ ((aligned(16)));
__m128i bit_met_p1_p7 __attribute__ ((aligned(16)));
__m128i bit_met_p3_m7 __attribute__ ((aligned(16)));
__m128i bit_met_p3_m5 __attribute__ ((aligned(16)));
__m128i bit_met_p3_m3 __attribute__ ((aligned(16)));
__m128i bit_met_p3_m1 __attribute__ ((aligned(16)));
__m128i bit_met_p3_p1 __attribute__ ((aligned(16)));
__m128i bit_met_p3_p3 __attribute__ ((aligned(16)));
__m128i bit_met_p3_p5 __attribute__ ((aligned(16)));
__m128i bit_met_p3_p7 __attribute__ ((aligned(16)));
__m128i bit_met_p5_m7 __attribute__ ((aligned(16)));
__m128i bit_met_p5_m5 __attribute__ ((aligned(16)));
__m128i bit_met_p5_m3 __attribute__ ((aligned(16)));
__m128i bit_met_p5_m1 __attribute__ ((aligned(16)));
__m128i bit_met_p5_p1 __attribute__ ((aligned(16)));
__m128i bit_met_p5_p3 __attribute__ ((aligned(16)));
__m128i bit_met_p5_p5 __attribute__ ((aligned(16)));
__m128i bit_met_p5_p7 __attribute__ ((aligned(16)));
__m128i bit_met_p7_m7 __attribute__ ((aligned(16)));
__m128i bit_met_p7_m5 __attribute__ ((aligned(16)));
__m128i bit_met_p7_m3 __attribute__ ((aligned(16)));
__m128i bit_met_p7_m1 __attribute__ ((aligned(16)));
__m128i bit_met_p7_p1 __attribute__ ((aligned(16)));
__m128i bit_met_p7_p3 __attribute__ ((aligned(16)));
__m128i bit_met_p7_p5 __attribute__ ((aligned(16)));
__m128i bit_met_p7_p7 __attribute__ ((aligned(16)));

__m128i  y0_p_1_1 __attribute__ ((aligned(16)));
__m128i  y0_p_1_3 __attribute__ ((aligned(16)));
__m128i  y0_p_1_5 __attribute__ ((aligned(16)));
__m128i  y0_p_1_7 __attribute__ ((aligned(16)));
__m128i  y0_p_3_1 __attribute__ ((aligned(16)));
__m128i  y0_p_3_3 __attribute__ ((aligned(16)));
__m128i  y0_p_3_5 __attribute__ ((aligned(16)));
__m128i  y0_p_3_7 __attribute__ ((aligned(16)));
__m128i  y0_p_5_1 __attribute__ ((aligned(16)));
__m128i  y0_p_5_3 __attribute__ ((aligned(16)));
__m128i  y0_p_5_5 __attribute__ ((aligned(16)));
__m128i  y0_p_5_7 __attribute__ ((aligned(16)));
__m128i  y0_p_7_1 __attribute__ ((aligned(16)));
__m128i  y0_p_7_3 __attribute__ ((aligned(16)));
__m128i  y0_p_7_5 __attribute__ ((aligned(16)));
__m128i  y0_p_7_7 __attribute__ ((aligned(16)));
__m128i  y0_m_1_1 __attribute__ ((aligned(16)));
__m128i  y0_m_1_3 __attribute__ ((aligned(16)));
__m128i  y0_m_1_5 __attribute__ ((aligned(16)));
__m128i  y0_m_1_7 __attribute__ ((aligned(16)));
__m128i  y0_m_3_1 __attribute__ ((aligned(16)));
__m128i  y0_m_3_3 __attribute__ ((aligned(16)));
__m128i  y0_m_3_5 __attribute__ ((aligned(16)));
__m128i  y0_m_3_7 __attribute__ ((aligned(16)));
__m128i  y0_m_5_1 __attribute__ ((aligned(16)));
__m128i  y0_m_5_3 __attribute__ ((aligned(16)));
__m128i  y0_m_5_5 __attribute__ ((aligned(16)));
__m128i  y0_m_5_7 __attribute__ ((aligned(16)));
__m128i  y0_m_7_1 __attribute__ ((aligned(16)));
__m128i  y0_m_7_3 __attribute__ ((aligned(16)));
__m128i  y0_m_7_5 __attribute__ ((aligned(16)));
__m128i  y0_m_7_7 __attribute__ ((aligned(16)));

__m128i  xmm0 __attribute__ ((aligned(16)));
__m128i  xmm1 __attribute__ ((aligned(16)));
__m128i  xmm2 __attribute__ ((aligned(16)));
__m128i  xmm3 __attribute__ ((aligned(16)));
__m128i  xmm4 __attribute__ ((aligned(16)));
__m128i  xmm5 __attribute__ ((aligned(16)));
__m128i  xmm6 __attribute__ ((aligned(16)));
__m128i  xmm7 __attribute__ ((aligned(16)));
__m128i  xmm8 __attribute__ ((aligned(16)));

__m128i  y0r __attribute__ ((aligned(16)));
__m128i  y0i __attribute__ ((aligned(16)));
__m128i  y1r __attribute__ ((aligned(16)));
__m128i  y1i __attribute__ ((aligned(16)));
__m128i  y2r __attribute__ ((aligned(16)));
__m128i  y2i __attribute__ ((aligned(16)));

__m128i  logmax_num_re0 __attribute__ ((aligned(16)));
__m128i  logmax_num_im0 __attribute__ ((aligned(16)));
__m128i  logmax_den_re0 __attribute__ ((aligned(16)));
__m128i  logmax_den_im0 __attribute__ ((aligned(16)));
__m128i  logmax_num_re1 __attribute__ ((aligned(16)));
__m128i  logmax_num_im1 __attribute__ ((aligned(16)));
__m128i  logmax_den_re1 __attribute__ ((aligned(16)));
__m128i  logmax_den_im1 __attribute__ ((aligned(16)));

__m128i tmp_result  __attribute__ ((aligned(16)));
__m128i tmp_result2 __attribute__ ((aligned(16)));
__m128i tmp_result3 __attribute__ ((aligned(16)));
__m128i tmp_result4 __attribute__ ((aligned(16)));
599 600 601 602 603


//==============================================================================================
// Auxiliary Makros

604
// calculates psi_a = psi_r*a_r + psi_i*a_i
605 606 607
#define prodsum_psi_a_epi16(psi_r,a_r,psi_i,a_i,psi_a) tmp_result = _mm_mulhi_epi16(psi_r,a_r); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result2 = _mm_mulhi_epi16(psi_i,a_i); tmp_result2 = _mm_slli_epi16(tmp_result2,1); psi_a = _mm_adds_epi16(tmp_result,tmp_result2);

// calculate interference magnitude
608
#define interference_abs_epi16(psi,int_ch_mag,int_mag,c1,c2) tmp_result = _mm_cmplt_epi16(psi,int_ch_mag); tmp_result2 = _mm_xor_si128(tmp_result,(*(__m128i*)&ones[0])); tmp_result = _mm_and_si128(tmp_result,c1); tmp_result2 = _mm_and_si128(tmp_result2,c2); int_mag = _mm_or_si128(tmp_result,tmp_result2);
609 610

// calculate interference magnitude
611 612
// tmp_result = ones in shorts corr. to interval 2<=x<=4, tmp_result2 interval < 2, tmp_result3 interval 4<x<6 and tmp_result4 interval x>6
#define interference_abs_64qam_epi16(psi,int_ch_mag,int_two_ch_mag,int_three_ch_mag,a,c1,c3,c5,c7) tmp_result = _mm_cmplt_epi16(psi,int_two_ch_mag); tmp_result3 = _mm_xor_si128(tmp_result,(*(__m128i*)&ones[0])); tmp_result2 = _mm_cmplt_epi16(psi,int_ch_mag); tmp_result = _mm_xor_si128(tmp_result,tmp_result2); tmp_result4 = _mm_cmpgt_epi16(psi,int_three_ch_mag); tmp_result3 = _mm_xor_si128(tmp_result3,tmp_result4); tmp_result = _mm_and_si128(tmp_result,c3); tmp_result2 = _mm_and_si128(tmp_result2,c1); tmp_result3 = _mm_and_si128(tmp_result3,c5); tmp_result4 = _mm_and_si128(tmp_result4,c7); tmp_result = _mm_or_si128(tmp_result,tmp_result2); tmp_result3 = _mm_or_si128(tmp_result3,tmp_result4); a = _mm_or_si128(tmp_result,tmp_result3);
613

614
// calculates a_sq = int_ch_mag*(a_r^2 + a_i^2)*scale_factor
615 616 617 618 619
#define square_a_epi16(a_r,a_i,int_ch_mag,scale_factor,a_sq) tmp_result = _mm_mulhi_epi16(a_r,a_r); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result = _mm_mulhi_epi16(tmp_result,scale_factor); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result = _mm_mulhi_epi16(tmp_result,int_ch_mag); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result2 = _mm_mulhi_epi16(a_i,a_i); tmp_result2 = _mm_slli_epi16(tmp_result2,1); tmp_result2 = _mm_mulhi_epi16(tmp_result2,scale_factor); tmp_result2 = _mm_slli_epi16(tmp_result2,1); tmp_result2 = _mm_mulhi_epi16(tmp_result2,int_ch_mag); tmp_result2 = _mm_slli_epi16(tmp_result2,1); a_sq = _mm_adds_epi16(tmp_result,tmp_result2);

// calculates a_sq = int_ch_mag*(a_r^2 + a_i^2)*scale_factor for 64-QAM
#define square_a_64qam_epi16(a_r,a_i,int_ch_mag,scale_factor,a_sq)  tmp_result = _mm_mulhi_epi16(a_r,a_r); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result = _mm_mulhi_epi16(tmp_result,scale_factor); tmp_result = _mm_slli_epi16(tmp_result,3); tmp_result = _mm_mulhi_epi16(tmp_result,int_ch_mag); tmp_result = _mm_slli_epi16(tmp_result,1); tmp_result2 = _mm_mulhi_epi16(a_i,a_i); tmp_result2 = _mm_slli_epi16(tmp_result2,1); tmp_result2 = _mm_mulhi_epi16(tmp_result2,scale_factor); tmp_result2 = _mm_slli_epi16(tmp_result2,3); tmp_result2 = _mm_mulhi_epi16(tmp_result2,int_ch_mag); tmp_result2 = _mm_slli_epi16(tmp_result2,1); a_sq = _mm_adds_epi16(tmp_result,tmp_result2);

620 621 622 623
#elif defined(__arm__)

#endif

624 625 626 627 628 629 630 631 632
//==============================================================================================
// SINGLE-STREAM
//==============================================================================================

//----------------------------------------------------------------------------------------------
// QPSK
//----------------------------------------------------------------------------------------------

int dlsch_qpsk_llr(LTE_DL_FRAME_PARMS *frame_parms,
633 634 635
                   int32_t **rxdataF_comp,
                   int16_t *dlsch_llr,
                   uint8_t symbol,
636 637 638
                   uint8_t first_symbol_flag,
                   uint16_t nb_rb,
                   uint16_t pbch_pss_sss_adjust,
639
                   uint8_t beamforming_mode)
640
{
641

642
  uint32_t *rxF = (uint32_t*)&rxdataF_comp[0][((int32_t)symbol*frame_parms->N_RB_DL*12)];
643
  uint32_t *llr32;
644
  int i,len;
645
  uint8_t symbol_mod = (symbol >= (7-frame_parms->Ncp))? (symbol-(7-frame_parms->Ncp)) : symbol;
646

647
  /*
648
  if (first_symbol_flag==1) {
649 650 651
    llr32 = (uint32_t*)dlsch_llr;
  } else {
    llr32 = (uint32_t*)(*llr32p);
652
  }*/
653

654
  llr32 = (uint32_t*)dlsch_llr;
655
  if (!llr32) {
656 657 658
    msg("dlsch_qpsk_llr: llr is null, symbol %d, llr32=%p\n",symbol, llr32);
    return(-1);
  }
659 660 661


  if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
662
    if (frame_parms->mode1_flag==0)
663
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
664
    else
665
      len = (nb_rb*10) - (5*pbch_pss_sss_adjust/6);
666
  } else if((beamforming_mode==7) && (frame_parms->Ncp==0) && (symbol==3 || symbol==6 || symbol==9 || symbol==12)){
667
      len = (nb_rb*9) - (3*pbch_pss_sss_adjust/4);
668
  } else if((beamforming_mode==7) && (frame_parms->Ncp==1) && (symbol==4 || symbol==7 || symbol==10)){
669
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
670
  } else {
671
    len = (nb_rb*12) - pbch_pss_sss_adjust;
672 673
  }

674

675
  //printf("dlsch_qpsk_llr: symbol %d,nb_rb %d, len %d,pbch_pss_sss_adjust %d\n",symbol,nb_rb,len,pbch_pss_sss_adjust);
676 677 678 679 680 681 682
  /*LOG_I(PHY,"dlsch_qpsk_llr: [symb %d / FirstSym %d / Length %d]: @LLR Buff %x, @LLR Buff(symb) %x \n",
             symbol,
             first_symbol_flag,
             len,
             dlsch_llr,
             llr32);*/

683
  //printf("ll32p=%p , dlsch_llr=%p, symbol=%d, flag=%d \n", llr32, dlsch_llr, symbol, first_symbol_flag);
684 685
  for (i=0; i<len; i++) {
    *llr32 = *rxF;
686
     //printf("llr %d : (%d,%d)\n",i,((int16_t*)llr32)[0],((int16_t*)llr32)[1]);
687 688
    rxF++;
    llr32++;
689 690
  }

691
  //*llr32p = (int16_t *)llr32;
692 693 694 695

  return(0);
}

696
int32_t dlsch_qpsk_llr_SIC(LTE_DL_FRAME_PARMS *frame_parms,
697
                           int32_t **rxdataF_comp,
Elena Lukashova's avatar
Elena Lukashova committed
698 699
                           int32_t **sic_buffer,  //Q15
                           int32_t **rho_i,
700
                           short *dlsch_llr,
Elena Lukashova's avatar
Elena Lukashova committed
701
                           uint8_t num_pdcch_symbols,
702
                           uint16_t nb_rb,
Elena Lukashova's avatar
Elena Lukashova committed
703
                           uint8_t subframe,
704
                           uint16_t mod_order_0,
705
                           uint32_t rb_alloc)
706 707 708 709 710 711 712 713 714 715
{

  int16_t rho_amp_x0[2*frame_parms->N_RB_DL*12];
  int16_t rho_rho_amp_x0[2*frame_parms->N_RB_DL*12];
  uint16_t amp_tmp;
  uint16_t *llr16=(uint16_t*)dlsch_llr;
  int i, len,  nsymb;
  uint8_t symbol, symbol_mod;
  int len_acc=0;
  uint16_t *sic_data;
716
  uint16_t pbch_pss_sss_adjust;
717 718 719 720

  nsymb = (frame_parms->Ncp==0) ? 14:12;

  for (symbol=num_pdcch_symbols; symbol<nsymb; symbol++) {
721 722
    uint16_t *rxF = (uint16_t*)(&rxdataF_comp[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    int16_t *rho_1=(int16_t*)(&rho_i[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
723
    sic_data = (uint16_t*)&sic_buffer[0][((int16_t)len_acc)];
724 725 726

    symbol_mod = (symbol>=(7-frame_parms->Ncp)) ? symbol-(7-frame_parms->Ncp) : symbol;

727

728

729

730 731 732 733
    if ((symbol_mod == 0) || (symbol_mod == (4-frame_parms->Ncp))) //pilots=1
      amp_tmp=0x1fff;//dlsch0->sqrt_rho_b; already taken into account
    else //pilots=0
      amp_tmp=0x1fff;//1.5*dlsch0->sqrt_rho_a; already taken into account
734

735
    if (mod_order_0==6)
Elena Lukashova's avatar
Elena Lukashova committed
736
      amp_tmp=amp_tmp<<1; // to compensate for >> 1 shift in modulation
737 738


739
    pbch_pss_sss_adjust=adjust_G2(frame_parms,&rb_alloc,2,subframe,symbol);
740 741 742 743 744 745 746 747 748

    if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
      if (frame_parms->mode1_flag==0)
        len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
      else
        len = (nb_rb*10) - (5*pbch_pss_sss_adjust/6);
    } else {
      len = (nb_rb*12) - pbch_pss_sss_adjust;
    }
749

750
   //  printf("dlsch_qpsk_llr: symbol %d,nb_rb %d, len %d,pbch_pss_sss_adjust %d\n",symbol,nb_rb,len,pbch_pss_sss_adjust);
751 752

    len_acc+=len; //accumulated length; this is done because in sic_buffer we have only data symbols
753

754
    multadd_complex_vector_real_scalar((int16_t *)sic_data,
Elena Lukashova's avatar
Elena Lukashova committed
755 756 757 758
                                       amp_tmp,
                                       (int16_t *)rho_amp_x0, //this is in Q13
                                       1,
                                       len);
759

760 761
    mult_cpx_vector((int16_t *)rho_1, //Q15
                    (int16_t *)rho_amp_x0, //Q13
762 763
                    (int16_t*)rho_rho_amp_x0,
                    len,
764
                    13);
765 766 767

#ifdef DEBUG_LLR_SIC
    write_output("rho_for_multipl.m","rho_for_m", rho_1,len,1,
Elena Lukashova's avatar
Elena Lukashova committed
768
     symbol==num_pdcch_symbols ? 15 :
769
     symbol==nsymb-1 ? 14 : 13);
770 771

    write_output("rho_rho_in_llr.m","rho2", rho_rho_amp_x0,len,1,
Elena Lukashova's avatar
Elena Lukashova committed
772 773
     symbol==num_pdcch_symbols ? 15 :
     symbol==nsymb-1 ? 14 : 13);
774
#endif
775

776
    sub_cpx_vector16((int16_t *)rxF,
Elena Lukashova's avatar
Elena Lukashova committed
777 778 779
                     (int16_t *)rho_rho_amp_x0,
                     //(int16_t *)clean_x1,
                     (int16_t *)rxF,
780
                     len*2);
781

782 783 784 785 786 787
#ifdef DEBUG_LLR_SIC
    write_output("rxFdata_comp1_after.m","rxF_a", rxF,len,1,1);
    write_output("rxF_comp1.m","rxF_1_comp", rxF,len,1,
                 symbol==num_pdcch_symbols ? 15 :
                 symbol==nsymb-1 ? 14 : 13);
#endif
788

789 790
    //this is for QPSK only!!!
    for (i=0; i<len*2; i++) {
791
      *llr16 =rxF[i];
792 793 794
      //printf("llr %d : (%d,%d)\n",i,((int16_t*)llr32)[0],((int16_t*)llr32)[1]);
      llr16++;
    }
795

796 797
  }

798
 // printf("dlsch_qpsk_llr_SIC: acc_len=%d\n",len_acc);
799 800 801 802 803

  return(0);
}


804 805 806 807 808
//----------------------------------------------------------------------------------------------
// 16-QAM
//----------------------------------------------------------------------------------------------

void dlsch_16qam_llr(LTE_DL_FRAME_PARMS *frame_parms,
809 810 811 812
                     int32_t **rxdataF_comp,
                     int16_t *dlsch_llr,
                     int32_t **dl_ch_mag,
                     uint8_t symbol,
813
                     uint8_t first_symbol_flag,
814
                     uint16_t nb_rb,
815
                     uint16_t pbch_pss_sss_adjust,
816 817
                     int16_t **llr32p,
                     uint8_t beamforming_mode)
818 819
{

820
#if defined(__x86_64__) || defined(__i386__)
821 822 823
  __m128i *rxF = (__m128i*)&rxdataF_comp[0][(symbol*frame_parms->N_RB_DL*12)];
  __m128i *ch_mag;
  __m128i llr128[2];
824 825 826 827 828 829 830 831 832
  uint32_t *llr32;
#elif defined(__arm__)
  int16x8_t *rxF = (int16x8_t*)&rxdataF_comp[0][(symbol*frame_parms->N_RB_DL*12)];
  int16x8_t *ch_mag;
  int16x8_t xmm0;
  int16_t *llr16;
#endif


833 834 835 836
  int i,len;
  unsigned char symbol_mod,len_mod4=0;


837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859
#if defined(__x86_64__) || defined(__i386__)
  if (first_symbol_flag==1) {
    llr32 = (uint32_t*)dlsch_llr;
  } else {
    llr32 = (uint32_t*)*llr32p;
  }
#elif defined(__arm__)
  if (first_symbol_flag==1) {
    llr16 = (int16_t*)dlsch_llr;
  } else {
    llr16 = (int16_t*)*llr32p;
  }
#endif

  symbol_mod = (symbol>=(7-frame_parms->Ncp)) ? symbol-(7-frame_parms->Ncp) : symbol;

#if defined(__x86_64__) || defined(__i386__)
  ch_mag = (__m128i*)&dl_ch_mag[0][(symbol*frame_parms->N_RB_DL*12)];
#elif defined(__arm__)
  ch_mag = (int16x8_t*)&dl_ch_mag[0][(symbol*frame_parms->N_RB_DL*12)];
#endif
  if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
    if (frame_parms->mode1_flag==0)
860
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
861
    else
862 863 864 865 866
      len = (nb_rb*10) - (5*pbch_pss_sss_adjust/6);
  } else if((beamforming_mode==7) && (frame_parms->Ncp==0) && (symbol==3 || symbol==6 || symbol==9 || symbol==12)){
      len = (nb_rb*9) - (3*pbch_pss_sss_adjust/4);
  } else if((beamforming_mode==7) && (frame_parms->Ncp==1) && (symbol==4 || symbol==7 || symbol==10)){
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
867
  } else {
868
    len = (nb_rb*12) - pbch_pss_sss_adjust;
869 870 871 872 873 874 875 876
  }

  // update output pointer according to number of REs in this symbol (<<2 because 4 bits per RE)
  if (first_symbol_flag == 1)
    *llr32p = dlsch_llr + (len<<2);
  else
    *llr32p += (len<<2);

877
 // printf("len=%d\n", len);
878
  len_mod4 = len&3;
879
 // printf("len_mod4=%d\n", len_mod4);
880
  len>>=2;  // length in quad words (4 REs)
881
 // printf("len>>=2=%d\n", len);
882
  len+=(len_mod4==0 ? 0 : 1);
883
 // printf("len+=%d\n", len);
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
  for (i=0; i<len; i++) {

#if defined(__x86_64__) || defined(__i386)
    xmm0 = _mm_abs_epi16(rxF[i]);
    xmm0 = _mm_subs_epi16(ch_mag[i],xmm0);

    // lambda_1=y_R, lambda_2=|y_R|-|h|^2, lamda_3=y_I, lambda_4=|y_I|-|h|^2
    llr128[0] = _mm_unpacklo_epi32(rxF[i],xmm0);
    llr128[1] = _mm_unpackhi_epi32(rxF[i],xmm0);
    llr32[0] = _mm_extract_epi32(llr128[0],0); //((uint32_t *)&llr128[0])[0];
    llr32[1] = _mm_extract_epi32(llr128[0],1); //((uint32_t *)&llr128[0])[1];
    llr32[2] = _mm_extract_epi32(llr128[0],2); //((uint32_t *)&llr128[0])[2];
    llr32[3] = _mm_extract_epi32(llr128[0],3); //((uint32_t *)&llr128[0])[3];
    llr32[4] = _mm_extract_epi32(llr128[1],0); //((uint32_t *)&llr128[1])[0];
    llr32[5] = _mm_extract_epi32(llr128[1],1); //((uint32_t *)&llr128[1])[1];
    llr32[6] = _mm_extract_epi32(llr128[1],2); //((uint32_t *)&llr128[1])[2];
    llr32[7] = _mm_extract_epi32(llr128[1],3); //((uint32_t *)&llr128[1])[3];
    llr32+=8;
#elif defined(__arm__)
    xmm0 = vabsq_s16(rxF[i]);
    xmm0 = vqsubq_s16(ch_mag[i],xmm0);
    // lambda_1=y_R, lambda_2=|y_R|-|h|^2, lamda_3=y_I, lambda_4=|y_I|-|h|^2

    llr16[0] = vgetq_lane_s16(rxF[i],0);
    llr16[1] = vgetq_lane_s16(rxF[i],1);
    llr16[2] = vgetq_lane_s16(xmm0,0);
    llr16[3] = vgetq_lane_s16(xmm0,1);
    llr16[4] = vgetq_lane_s16(rxF[i],2);
    llr16[5] = vgetq_lane_s16(rxF[i],3);
    llr16[6] = vgetq_lane_s16(xmm0,2);
    llr16[7] = vgetq_lane_s16(xmm0,3);
    llr16[8] = vgetq_lane_s16(rxF[i],4);
    llr16[9] = vgetq_lane_s16(rxF[i],5);
    llr16[10] = vgetq_lane_s16(xmm0,4);
    llr16[11] = vgetq_lane_s16(xmm0,5);
    llr16[12] = vgetq_lane_s16(rxF[i],6);
    llr16[13] = vgetq_lane_s16(rxF[i],6);
    llr16[14] = vgetq_lane_s16(xmm0,7);
    llr16[15] = vgetq_lane_s16(xmm0,7);
    llr16+=16;
#endif

  }

#if defined(__x86_64__) || defined(__i386__)
  _mm_empty();
  _m_empty();
#endif
}

934
void dlsch_16qam_llr_SIC (LTE_DL_FRAME_PARMS *frame_parms,
935
                          int32_t **rxdataF_comp,
Elena Lukashova's avatar
Elena Lukashova committed
936 937 938 939 940 941 942
                          int32_t **sic_buffer,  //Q15
                          int32_t **rho_i,
                          int16_t *dlsch_llr,
                          uint8_t num_pdcch_symbols,
                          int32_t **dl_ch_mag,
                          uint16_t nb_rb,
                          uint8_t subframe,
943
                          uint16_t mod_order_0,
944
                          uint32_t rb_alloc)
945
{
946 947 948 949 950 951 952 953 954 955
  int16_t rho_amp_x0[2*frame_parms->N_RB_DL*12];
  int16_t rho_rho_amp_x0[2*frame_parms->N_RB_DL*12];
  uint16_t amp_tmp;
  uint32_t *llr32=(uint32_t*)dlsch_llr;
  int i, len,  nsymb;
  uint8_t symbol, symbol_mod;
  int len_acc=0;
  uint16_t *sic_data;
  uint16_t pbch_pss_sss_adjust;
  unsigned char len_mod4=0;
956 957
  __m128i llr128[2];
  __m128i *ch_mag;
958
  nsymb = (frame_parms->Ncp==0) ? 14:12;
959

960 961 962 963
    for (symbol=num_pdcch_symbols; symbol<nsymb; symbol++) {
    uint16_t *rxF = (uint16_t*)(&rxdataF_comp[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    int16_t *rho_1=(int16_t*)(&rho_i[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    ch_mag = (__m128i*)(&dl_ch_mag[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
964 965
    sic_data = (uint16_t*)(&sic_buffer[0][((int16_t)len_acc)]);

966
    symbol_mod = (symbol>=(7-frame_parms->Ncp)) ? symbol-(7-frame_parms->Ncp) : symbol;
967

968 969
    pbch_pss_sss_adjust=adjust_G2(frame_parms,&rb_alloc,4,subframe,symbol);

970 971 972 973 974 975 976 977 978 979
    if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
      amp_tmp=0x1fff;//dlsch0->sqrt_rho_b; already taken into account
      if (frame_parms->mode1_flag==0)
        len = nb_rb*8 - (2*pbch_pss_sss_adjust/3);
      else
        len = nb_rb*10 - (5*pbch_pss_sss_adjust/6);
    } else {
      amp_tmp=0x1fff;;//dlsch0->sqrt_rho_a; already taken into account
      len = nb_rb*12 - pbch_pss_sss_adjust;
    }
980

981 982
    if (mod_order_0==6)
      amp_tmp=amp_tmp<<1; // to compensate for >> 1 shift in modulation
983

984
    len_acc+=len;
985

986
    multadd_complex_vector_real_scalar((int16_t *)sic_data,
Elena Lukashova's avatar
Elena Lukashova committed
987 988 989 990
                                       amp_tmp,
                                       (int16_t *)rho_amp_x0, //this is in Q13
                                       1,
                                       len);
991

992 993 994 995 996
     mult_cpx_vector((int16_t *)rho_1, //Q15
                    (int16_t *)rho_amp_x0, //Q13
                    (int16_t*)rho_rho_amp_x0,
                    len,
                    13);
997

998
     sub_cpx_vector16((int16_t *)rxF,
Elena Lukashova's avatar
Elena Lukashova committed
999 1000 1001
                      (int16_t *)rho_rho_amp_x0,
                      //(int16_t *)clean_x1,
                      (int16_t *)rxF,
1002
                      len*2);
1003

1004 1005 1006 1007 1008
    len_mod4 = len&3;
    len>>=2;  // length in quad words (4 REs)
    len+=(len_mod4==0 ? 0 : 1);

    for (i=0; i<len; i++) {
1009

1010 1011 1012 1013 1014 1015

    __m128i *x1 = (__m128i*)rxF;//clean_x1;
//printf("%p %p %p\n", clean_x1, &clean_x1, &clean_x1[0]);
//int *a = malloc(10*sizeof(int));
//printf("%p %p\n", a, &a);
//exit(0);
1016
    xmm0 = _mm_abs_epi16(x1[i]);
1017 1018 1019
    xmm0 = _mm_subs_epi16(ch_mag[i],xmm0);

    // lambda_1=y_R, lambda_2=|y_R|-|h|^2, lamda_3=y_I, lambda_4=|y_I|-|h|^2
1020 1021
    llr128[0] = _mm_unpacklo_epi32(x1[i],xmm0);
    llr128[1] = _mm_unpackhi_epi32(x1[i],xmm0);
1022 1023 1024 1025 1026 1027 1028 1029 1030
    llr32[0] = _mm_extract_epi32(llr128[0],0); //((uint32_t *)&llr128[0])[0];
    llr32[1] = _mm_extract_epi32(llr128[0],1); //((uint32_t *)&llr128[0])[1];
    llr32[2] = _mm_extract_epi32(llr128[0],2); //((uint32_t *)&llr128[0])[2];
    llr32[3] = _mm_extract_epi32(llr128[0],3); //((uint32_t *)&llr128[0])[3];
    llr32[4] = _mm_extract_epi32(llr128[1],0); //((uint32_t *)&llr128[1])[0];
    llr32[5] = _mm_extract_epi32(llr128[1],1); //((uint32_t *)&llr128[1])[1];
    llr32[6] = _mm_extract_epi32(llr128[1],2); //((uint32_t *)&llr128[1])[2];
    llr32[7] = _mm_extract_epi32(llr128[1],3); //((uint32_t *)&llr128[1])[3];
    llr32+=8;
1031

1032 1033 1034 1035
  }
  _mm_empty();
  _m_empty();
}
1036
}
1037 1038 1039 1040 1041 1042

//----------------------------------------------------------------------------------------------
// 64-QAM
//----------------------------------------------------------------------------------------------

void dlsch_64qam_llr(LTE_DL_FRAME_PARMS *frame_parms,
1043 1044 1045 1046 1047
                     int32_t **rxdataF_comp,
                     int16_t *dlsch_llr,
                     int32_t **dl_ch_mag,
                     int32_t **dl_ch_magb,
                     uint8_t symbol,
1048
                     uint8_t first_symbol_flag,
1049
                     uint16_t nb_rb,
1050
                     uint16_t pbch_pss_sss_adjust,
1051 1052
                     //int16_t **llr_save,
                     uint32_t llr_offset,
1053
                     uint8_t beamforming_mode)
1054
{
1055
#if defined(__x86_64__) || defined(__i386__)
1056 1057
  __m128i *rxF = (__m128i*)&rxdataF_comp[0][(symbol*frame_parms->N_RB_DL*12)];
  __m128i *ch_mag,*ch_magb;
1058 1059 1060 1061
#elif defined(__arm__)
  int16x8_t *rxF = (int16x8_t*)&rxdataF_comp[0][(symbol*frame_parms->N_RB_DL*12)];
  int16x8_t *ch_mag,*ch_magb,xmm1,xmm2;
#endif
1062 1063 1064 1065
  int i,len,len2;
  unsigned char symbol_mod,len_mod4;
  short *llr;
  int16_t *llr2;
1066
  int8_t *pllr_symbol;
1067

1068
  /*
1069 1070 1071 1072
  if (first_symbol_flag==1)
    llr = dlsch_llr;
  else
    llr = *llr_save;
1073 1074 1075 1076 1077
  */
  llr = dlsch_llr;

  pllr_symbol = (int8_t*)dlsch_llr;
  pllr_symbol += llr_offset;
1078 1079 1080

  symbol_mod = (symbol>=(7-frame_parms->Ncp)) ? symbol-(7-frame_parms->Ncp) : symbol;

1081
#if defined(__x86_64__) || defined(__i386__)
1082 1083
  ch_mag = (__m128i*)&dl_ch_mag[0][(symbol*frame_parms->N_RB_DL*12)];
  ch_magb = (__m128i*)&dl_ch_magb[0][(symbol*frame_parms->N_RB_DL*12)];
1084 1085 1086 1087
#elif defined(__arm__)
  ch_mag = (int16x8_t*)&dl_ch_mag[0][(symbol*frame_parms->N_RB_DL*12)];
  ch_magb = (int16x8_t*)&dl_ch_magb[0][(symbol*frame_parms->N_RB_DL*12)];
#endif
1088 1089
  if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
    if (frame_parms->mode1_flag==0)
1090
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
1091
    else
1092 1093 1094 1095 1096
      len = (nb_rb*10) - (5*pbch_pss_sss_adjust/6);
  } else if((beamforming_mode==7) && (frame_parms->Ncp==0) && (symbol==3 || symbol==6 || symbol==9 || symbol==12)){
      len = (nb_rb*9) - (3*pbch_pss_sss_adjust/4);
  } else if((beamforming_mode==7) && (frame_parms->Ncp==1) && (symbol==4 || symbol==7 || symbol==10)){
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
1097
  } else {
1098
    len = (nb_rb*12) - pbch_pss_sss_adjust;
1099 1100
  }

1101 1102 1103 1104 1105 1106 1107 1108 1109
//  printf("dlsch_64qam_llr: symbol %d,nb_rb %d, len %d,pbch_pss_sss_adjust %d\n",symbol,nb_rb,len,pbch_pss_sss_adjust);

/*  LOG_I(PHY,"dlsch_64qam_llr [symb %d / FirstSym %d / Length %d]: @LLR Buff %x \n",
             symbol,
             first_symbol_flag,
             len,
             dlsch_llr,
             pllr_symbol);*/

1110 1111 1112 1113 1114
  llr2 = llr;
  llr += (len*6);

  len_mod4 =len&3;
  len2=len>>2;  // length in quad words (4 REs)
knopp's avatar
knopp committed
1115
  len2+=((len_mod4==0)?0:1);
1116 1117 1118

  for (i=0; i<len2; i++) {

1119
#if defined(__x86_64__) || defined(__i386__)
1120 1121 1122 1123
    xmm1 = _mm_abs_epi16(rxF[i]);
    xmm1 = _mm_subs_epi16(ch_mag[i],xmm1);
    xmm2 = _mm_abs_epi16(xmm1);
    xmm2 = _mm_subs_epi16(ch_magb[i],xmm2);
1124 1125 1126 1127 1128 1129
#elif defined(__arm__)
    xmm1 = vabsq_s16(rxF[i]);
    xmm1 = vsubq_s16(ch_mag[i],xmm1);
    xmm2 = vabsq_s16(xmm1);
    xmm2 = vsubq_s16(ch_magb[i],xmm2);
#endif
1130 1131
    // loop over all LLRs in quad word (24 coded bits)
    /*
1132 1133 1134 1135 1136 1137 1138 1139
      for (j=0;j<8;j+=2) {
      llr2[0] = ((short *)&rxF[i])[j];
      llr2[1] = ((short *)&rxF[i])[j+1];
      llr2[2] = ((short *)&xmm1)[j];
      llr2[3] = ((short *)&xmm1)[j+1];
      llr2[4] = ((short *)&xmm2)[j];
      llr2[5] = ((short *)&xmm2)[j+1];

Elena Lukashova's avatar
Elena Lukashova committed
1140
     llr2+=6;
1141
      }
1142 1143 1144
    */
    llr2[0] = ((short *)&rxF[i])[0];
    llr2[1] = ((short *)&rxF[i])[1];
1145
#if defined(__x86_64__) || defined(__i386__)
1146 1147 1148 1149
    llr2[2] = _mm_extract_epi16(xmm1,0);
    llr2[3] = _mm_extract_epi16(xmm1,1);//((short *)&xmm1)[j+1];
    llr2[4] = _mm_extract_epi16(xmm2,0);//((short *)&xmm2)[j];
    llr2[5] = _mm_extract_epi16(xmm2,1);//((short *)&xmm2)[j+1];
1150 1151 1152 1153 1154 1155
#elif defined(__arm__)
    llr2[2] = vgetq_lane_s16(xmm1,0);
    llr2[3] = vgetq_lane_s16(xmm1,1);//((short *)&xmm1)[j+1];
    llr2[4] = vgetq_lane_s16(xmm2,0);//((short *)&xmm2)[j];
    llr2[5] = vgetq_lane_s16(xmm2,1);//((short *)&xmm2)[j+1];
#endif
1156 1157 1158 1159

    llr2+=6;
    llr2[0] = ((short *)&rxF[i])[2];
    llr2[1] = ((short *)&rxF[i])[3];
1160
#if defined(__x86_64__) || defined(__i386__)
1161 1162 1163 1164
    llr2[2] = _mm_extract_epi16(xmm1,2);
    llr2[3] = _mm_extract_epi16(xmm1,3);//((short *)&xmm1)[j+1];
    llr2[4] = _mm_extract_epi16(xmm2,2);//((short *)&xmm2)[j];
    llr2[5] = _mm_extract_epi16(xmm2,3);//((short *)&xmm2)[j+1];
1165 1166 1167 1168 1169 1170
#elif defined(__arm__)
    llr2[2] = vgetq_lane_s16(xmm1,2);
    llr2[3] = vgetq_lane_s16(xmm1,3);//((short *)&xmm1)[j+1];
    llr2[4] = vgetq_lane_s16(xmm2,2);//((short *)&xmm2)[j];
    llr2[5] = vgetq_lane_s16(xmm2,3);//((short *)&xmm2)[j+1];
#endif
1171 1172 1173 1174

    llr2+=6;
    llr2[0] = ((short *)&rxF[i])[4];
    llr2[1] = ((short *)&rxF[i])[5];
1175
#if defined(__x86_64__) || defined(__i386__)
1176 1177 1178 1179
    llr2[2] = _mm_extract_epi16(xmm1,4);
    llr2[3] = _mm_extract_epi16(xmm1,5);//((short *)&xmm1)[j+1];
    llr2[4] = _mm_extract_epi16(xmm2,4);//((short *)&xmm2)[j];
    llr2[5] = _mm_extract_epi16(xmm2,5);//((short *)&xmm2)[j+1];
1180 1181 1182 1183 1184 1185
#elif defined(__arm__)
    llr2[2] = vgetq_lane_s16(xmm1,4);
    llr2[3] = vgetq_lane_s16(xmm1,5);//((short *)&xmm1)[j+1];
    llr2[4] = vgetq_lane_s16(xmm2,4);//((short *)&xmm2)[j];
    llr2[5] = vgetq_lane_s16(xmm2,5);//((short *)&xmm2)[j+1];
#endif
1186 1187 1188
    llr2+=6;
    llr2[0] = ((short *)&rxF[i])[6];
    llr2[1] = ((short *)&rxF[i])[7];
1189
#if defined(__x86_64__) || defined(__i386__)
1190 1191 1192 1193
    llr2[2] = _mm_extract_epi16(xmm1,6);
    llr2[3] = _mm_extract_epi16(xmm1,7);//((short *)&xmm1)[j+1];
    llr2[4] = _mm_extract_epi16(xmm2,6);//((short *)&xmm2)[j];
    llr2[5] = _mm_extract_epi16(xmm2,7);//((short *)&xmm2)[j+1];
1194 1195 1196 1197 1198 1199
#elif defined(__arm__)
    llr2[2] = vgetq_lane_s16(xmm1,6);
    llr2[3] = vgetq_lane_s16(xmm1,7);//((short *)&xmm1)[j+1];
    llr2[4] = vgetq_lane_s16(xmm2,6);//((short *)&xmm2)[j];
    llr2[5] = vgetq_lane_s16(xmm2,7);//((short *)&xmm2)[j+1];
#endif
1200 1201 1202 1203
    llr2+=6;

  }

1204
#if defined(__x86_64__) || defined(__i386__)
1205 1206
  _mm_empty();
  _m_empty();
1207
#endif
1208 1209
}

1210
//#if 0
1211
void dlsch_64qam_llr_SIC(LTE_DL_FRAME_PARMS *frame_parms,
1212
                         int32_t **rxdataF_comp,
Elena Lukashova's avatar
Elena Lukashova committed
1213 1214 1215 1216 1217 1218 1219 1220
                         int32_t **sic_buffer,  //Q15
                         int32_t **rho_i,
                         int16_t *dlsch_llr,
                         uint8_t num_pdcch_symbols,
                         int32_t **dl_ch_mag,
                         int32_t **dl_ch_magb,
                         uint16_t nb_rb,
                         uint8_t subframe,
1221
                         uint16_t mod_order_0,
1222
                         uint32_t rb_alloc)
1223
{
1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
  int16_t rho_amp_x0[2*frame_parms->N_RB_DL*12];
  int16_t rho_rho_amp_x0[2*frame_parms->N_RB_DL*12];
  uint16_t amp_tmp;
  uint16_t *llr32=(uint16_t*)dlsch_llr;
  int i, len,  nsymb, len2;
  uint8_t symbol, symbol_mod;
  int len_acc=0;
  uint16_t *sic_data;
  uint16_t pbch_pss_sss_adjust;
  unsigned char len_mod4=0;
1234
  uint16_t *llr2;
1235
  __m128i *ch_mag,*ch_magb;
1236

Elena Lukashova's avatar
Elena Lukashova committed
1237
  nsymb = (frame_parms->Ncp==0) ? 14:12;
1238

1239 1240 1241 1242 1243
  for (symbol=num_pdcch_symbols; symbol<nsymb; symbol++) {
    uint16_t *rxF = (uint16_t*)(&rxdataF_comp[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    int16_t *rho_1=(int16_t*)(&rho_i[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    ch_mag = (__m128i*)(&dl_ch_mag[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
    ch_magb = (__m128i*)(&dl_ch_magb[0][((int16_t)symbol*frame_parms->N_RB_DL*12)]);
1244 1245
    sic_data = (uint16_t*)(&sic_buffer[0][((int16_t)len_acc)]);

1246
    symbol_mod = (symbol>=(7-frame_parms->Ncp)) ? symbol-(7-frame_parms->Ncp) : symbol;
1247

1248 1249
    pbch_pss_sss_adjust=adjust_G2(frame_parms,&rb_alloc,6,subframe,symbol);

Elena Lukashova's avatar
Elena Lukashova committed
1250
    if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
1251 1252 1253 1254 1255
      amp_tmp = 0x1fff;//dlsch0->sqrt_rho_b; already taken into account
      if (frame_parms->mode1_flag==0)
        len = nb_rb*8 - (2*pbch_pss_sss_adjust/3);
      else
        len = nb_rb*10 - (5*pbch_pss_sss_adjust/6);
1256
      } else {
1257 1258
        amp_tmp = 0x1fff; //dlsch0->sqrt_rho_a; already taken into account
        len = nb_rb*12 - pbch_pss_sss_adjust;
1259
      }
1260

1261 1262
    if (mod_order_0==6)
      amp_tmp=amp_tmp<<1; // to compensate for >> 1 shift in modulation
1263

1264
    len_acc+=len;
1265

1266
    multadd_complex_vector_real_scalar((int16_t *)sic_data,
Elena Lukashova's avatar
Elena Lukashova committed
1267 1268 1269 1270
                                        amp_tmp,
                                        (int16_t *)rho_amp_x0, //this is in Q13
                                        1,
                                        len);
1271

1272
    mult_cpx_vector((int16_t *)rho_1, //Q15
1273 1274 1275 1276
                    (int16_t *)rho_amp_x0, //Q13
                    (int16_t*)rho_rho_amp_x0,
                    len,
                    13);
1277

1278
    sub_cpx_vector16((int16_t *)rxF,
Elena Lukashova's avatar
Elena Lukashova committed
1279 1280 1281 1282
                      (int16_t *)rho_rho_amp_x0,
                      //(int16_t *)clean_x1,
                      (int16_t *)rxF,
                      len*2);
1283

1284 1285
    llr2 = llr32;
    llr32 += (len*6);
1286

1287 1288 1289
    len_mod4 =len&3;
    len2=len>>2;  // length in quad words (4 REs)
    len2+=(len_mod4?0:1);
1290 1291


1292

1293
    for (i=0; i<len2; i++) {
1294

1295 1296 1297 1298 1299
      __m128i *x1 = (__m128i*)rxF;
      xmm1 = _mm_abs_epi16(x1[i]);
      xmm1 = _mm_subs_epi16(ch_mag[i],xmm1);
      xmm2 = _mm_abs_epi16(xmm1);
      xmm2 = _mm_subs_epi16(ch_magb[i],xmm2);
1300

1301 1302 1303 1304 1305 1306 1307 1308 1309
      // loop over all LLRs in quad word (24 coded bits)
      /*
        for (j=0;j<8;j+=2) {
        llr2[0] = ((short *)&rxF[i])[j];
        llr2[1] = ((short *)&rxF[i])[j+1];
        llr2[2] = ((short *)&xmm1)[j];
        llr2[3] = ((short *)&xmm1)[j+1];
        llr2[4] = ((short *)&xmm2)[j];
        llr2[5] = ((short *)&xmm2)[j+1];
1310

1311 1312 1313 1314 1315 1316 1317 1318 1319
       llr2+=6;
        }
      */
      llr2[0] = ((short *)&x1[i])[0];
      llr2[1] = ((short *)&x1[i])[1];
      llr2[2] = _mm_extract_epi16(xmm1,0);
      llr2[3] = _mm_extract_epi16(xmm1,1);//((short *)&xmm1)[j+1];
      llr2[4] = _mm_extract_epi16(xmm2,0);//((short *)&xmm2)[j];
      llr2[5] = _mm_extract_epi16(xmm2,1);//((short *)&xmm2)[j+1];
1320

1321

1322 1323 1324
      llr2+=6;
      llr2[0] = ((short *)&x1[i])[2];
      llr2[1] = ((short *)&x1[i])[3];
1325

1326 1327 1328 1329
      llr2[2] = _mm_extract_epi16(xmm1,2);
      llr2[3] = _mm_extract_epi16(xmm1,3);//((short *)&xmm1)[j+1];
      llr2[4] = _mm_extract_epi16(xmm2,2);//((short *)&xmm2)[j];
      llr2[5] = _mm_extract_epi16(xmm2,3);//((short *)&xmm2)[j+1];
1330

1331 1332 1333
      llr2+=6;
      llr2[0] = ((short *)&x1[i])[4];
      llr2[1] = ((short *)&x1[i])[5];
1334

1335 1336 1337 1338
      llr2[2] = _mm_extract_epi16(xmm1,4);
      llr2[3] = _mm_extract_epi16(xmm1,5);//((short *)&xmm1)[j+1];
      llr2[4] = _mm_extract_epi16(xmm2,4);//((short *)&xmm2)[j];
      llr2[5] = _mm_extract_epi16(xmm2,5);//((short *)&xmm2)[j+1];
1339

1340 1341 1342
      llr2+=6;
      llr2[0] = ((short *)&x1[i])[6];
      llr2[1] = ((short *)&x1[i])[7];
1343

1344 1345 1346 1347
      llr2[2] = _mm_extract_epi16(xmm1,6);
      llr2[3] = _mm_extract_epi16(xmm1,7);//((short *)&xmm1)[j+1];
      llr2[4] = _mm_extract_epi16(xmm2,6);//((short *)&xmm2)[j];
      llr2[5] = _mm_extract_epi16(xmm2,7);//((short *)&xmm2)[j+1];
1348

1349
      llr2+=6;
1350

1351
    }
1352

1353 1354
 // *llr_save = llr;

1355 1356 1357
  _mm_empty();
  _m_empty();

1358
  }
1359 1360
}
//#endif
1361 1362 1363 1364 1365 1366 1367 1368
//==============================================================================================
// DUAL-STREAM
//==============================================================================================

//----------------------------------------------------------------------------------------------
// QPSK
//----------------------------------------------------------------------------------------------

1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
#if defined(__x86_64__) || defined(__i386)
__m128i  y0r_over2 __attribute__ ((aligned(16)));
__m128i  y0i_over2 __attribute__ ((aligned(16)));
__m128i  y1r_over2 __attribute__ ((aligned(16)));
__m128i  y1i_over2 __attribute__ ((aligned(16)));

__m128i  A __attribute__ ((aligned(16)));
__m128i  B __attribute__ ((aligned(16)));
__m128i  C __attribute__ ((aligned(16)));
__m128i  D __attribute__ ((aligned(16)));
__m128i  E __attribute__ ((aligned(16)));
__m128i  F __attribute__ ((aligned(16)));
__m128i  G __attribute__ ((aligned(16)));
__m128i  H __attribute__ ((aligned(16)));
1383

1384
#endif
1385 1386 1387 1388 1389 1390 1391 1392 1393

int dlsch_qpsk_qpsk_llr(LTE_DL_FRAME_PARMS *frame_parms,
                        int **rxdataF_comp,
                        int **rxdataF_comp_i,
                        int **rho_i,
                        short *dlsch_llr,
                        unsigned char symbol,
                        unsigned char first_symbol_flag,
                        unsigned short nb_rb,
1394
                        uint16_t pbch_pss_sss_adjust,
1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
                        short **llr16p)
{

  int16_t *rxF=(int16_t*)&rxdataF_comp[0][(symbol*frame_parms->N_RB_DL*12)];
  int16_t *rxF_i=(int16_t*)&rxdataF_comp_i[0][(symbol*frame_parms->N_RB_DL*12)];
  int16_t *rho=(int16_t*)&rho_i[0][(symbol*frame_parms->N_RB_DL*12)];
  int16_t *llr16;
  int len;
  uint8_t symbol_mod = (symbol >= (7-frame_parms->Ncp))? (symbol-(7-frame_parms->Ncp)) : symbol;

  if (first_symbol_flag == 1) {
    llr16 = (int16_t*)dlsch_llr;
  } else {
    llr16 = (int16_t*)(*llr16p);
  }

  if (!llr16) {
    msg("dlsch_qpsk_qpsk_llr: llr is null, symbol %d\n",symbol);
    return -1;
  }

  if ((symbol_mod==0) || (symbol_mod==(4-frame_parms->Ncp))) {
    // if symbol has pilots
    if (frame_parms->mode1_flag==0)
      // in 2 antenna ports we have 8 REs per symbol per RB
      len = (nb_rb*8) - (2*pbch_pss_sss_adjust/3);
    else
      // for 1 antenna port we have 10 REs per symbol per RB
      len = (nb_rb*10) - (5*pbch_pss_sss_adjust/6);
  } else {
    // symbol has no pilots
    len = (nb_rb*12) - pbch_pss_sss_adjust;
  }

1429
  // printf("dlsch_qpsk_qpsk_llr: symbol %d,nb_rb %d, len %d,pbch_pss_sss_adjust %d\n",symbol,nb_rb,len,pbch_pss_sss_adjust);
1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440
  //    printf("qpsk_qpsk: len %d, llr16 %p\n",len,llr16);
  qpsk_qpsk((short *)rxF,
            (short *)rxF_i,
            (short *)llr16,
            (short *)rho,
            len);

  llr16 += (len<<1);
  *llr16p = (short *)llr16;

  return(0);
1441 1442
}

1443
//__m128i ONE_OVER_SQRT_8 __attribute__((aligned(16)));
1444 1445

void qpsk_qpsk(short *stream0_in,
1446
               short *stream1_in,
1447 1448 1449
               short *stream0_out,
               short *rho01,
               int length
Elena Lukashova's avatar
Elena Lukashova committed
1450
         )
1451
{
1452

1453
  /*
1454 1455 1456 1457 1458 1459 1460 1461
    This function computes the LLRs of stream 0 (s_0) in presence of the interfering stream 1 (s_1) assuming that both symbols are QPSK. It can be used for both MU-MIMO interference-aware receiver or for SU-MIMO receivers.

    Parameters:
    stream0_in = Matched filter output y0' = (h0*g0)*y0
    stream1_in = Matched filter output y1' = (h0*g1)*y0
    stream0_out = LLRs
    rho01 = Correlation between the two effective channels \rho_{10} = (h1*g1)*(h0*g0)
    length = number of resource elements
1462
  */
1463

1464
#if defined(__x86_64__) || defined(__i386__)
1465 1466 1467 1468
  __m128i *rho01_128i = (__m128i *)rho01;
  __m128i *stream0_128i_in = (__m128i *)stream0_in;
  __m128i *stream1_128i_in = (__m128i *)stream1_in;
  __m128i *stream0_128i_out = (__m128i *)stream0_out;
1469 1470 1471 1472 1473 1474 1475
  __m128i ONE_OVER_SQRT_8 = _mm_set1_epi16(23170); //round(2^16/sqrt(8))
#elif defined(__arm__)
  int16x8_t *rho01_128i = (int16x8_t *)rho01;
  int16x8_t *stream0_128i_in = (int16x8_t *)stream0_in;
  int16x8_t *stream1_128i_in = (int16x8_t *)stream1_in;
  int16x8_t *stream0_128i_out = (int16x8_t *)stream0_out;
  int16x8_t ONE_OVER_SQRT_8 = vdupq_n_s16(23170); //round(2^16/sqrt(8))
1476 1477
#endif

1478
  int i;
1479 1480


1481 1482
  for (i=0; i<length>>2; i+=2) {
    // in each iteration, we take 8 complex samples
1483
#if defined(__x86_64__) || defined(__i386__)
1484 1485 1486 1487 1488
    xmm0 = rho01_128i[i]; // 4 symbols
    xmm1 = rho01_128i[i+1];

    // put (rho_r + rho_i)/2sqrt2 in rho_rpi
    // put (rho_r - rho_i)/2sqrt2 in rho_rmi
1489

1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
    xmm0 = _mm_shufflelo_epi16(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm0 = _mm_shufflehi_epi16(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm0 = _mm_shuffle_epi32(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shufflelo_epi16(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shufflehi_epi16(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shuffle_epi32(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    //xmm0 = [Re(0,1) Re(2,3) Im(0,1) Im(2,3)]
    //xmm1 = [Re(4,5) Re(6,7) Im(4,5) Im(6,7)]
    xmm2 = _mm_unpacklo_epi64(xmm0,xmm1); // Re(rho)
    xmm3 = _mm_unpackhi_epi64(xmm0,xmm1); // Im(rho)
    rho_rpi = _mm_adds_epi16(xmm2,xmm3); // rho = Re(rho) + Im(rho)
    rho_rmi = _mm_subs_epi16(xmm2,xmm3); // rho* = Re(rho) - Im(rho)

    // divide by sqrt(8), no shift needed ONE_OVER_SQRT_8 = Q1.16
    rho_rpi = _mm_mulhi_epi16(rho_rpi,ONE_OVER_SQRT_8);
    rho_rmi = _mm_mulhi_epi16(rho_rmi,ONE_OVER_SQRT_8);
1506
#elif defined(__arm__)
1507

1508 1509

#endif
1510 1511 1512
    // Compute LLR for first bit of stream 0

    // Compute real and imaginary parts of MF output for stream 0
1513
#if defined(__x86_64__) || defined(__i386__)
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529
    xmm0 = stream0_128i_in[i];
    xmm1 = stream0_128i_in[i+1];

    xmm0 = _mm_shufflelo_epi16(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm0 = _mm_shufflehi_epi16(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm0 = _mm_shuffle_epi32(xmm0,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shufflelo_epi16(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shufflehi_epi16(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    xmm1 = _mm_shuffle_epi32(xmm1,0xd8); //_MM_SHUFFLE(0,2,1,3));
    //xmm0 = [Re(0,1) Re(2,3) Im(0,1) Im(2,3)]
    //xmm1 = [Re(4,5) Re(6,7) Im(4,5) Im(6,7)]
    y0r = _mm_unpacklo_epi64(xmm0,xmm1); // = [y0r(1),y0r(2),y0r(3),y0r(4)]
    y0i = _mm_unpackhi_epi64(xmm0,xmm1);

    y0r_over2  = _mm_srai_epi16(y0r,1);   // divide by 2
    y0i_over2  = _mm_srai_epi16(y0i,1);   // divide by 2
1530 1531
#elif defined(__arm__)

1532 1533

#endif
1534
    // Compute real and imaginary parts of MF output for stream 1
1535
#if defined(__x86_64__) || defined(__i386__)
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578