arm_rfft_f32.c
Go to the documentation of this file.
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2011 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. December 2011 00005 * $Revision: V2.0.0 00006 * 00007 * Project: Cortex-R DSP Library 00008 * Title: arm_rfft_f32.c 00009 * 00010 * Description: RFFT & RIFFT Floating point process function 00011 * 00012 * Target Processor: Cortex-R4/R5 00013 * 00014 * Version 1.0.0 2011/03/08 00015 * Alpha release. 00016 * 00017 * Version 1.0.1 2011/09/30 00018 * Beta release. 00019 * 00020 * Version 2.0.0 2011/12/15 00021 * Final release. 00022 * 00023 * -------------------------------------------------------------------- */ 00024 #include "arm_math.h" 00025 00109 /*-------------------------------------------------------------------- 00110 * Internal functions prototypes 00111 *--------------------------------------------------------------------*/ 00112 00113 void arm_split_rfft_f32( 00114 float32_t * pSrc, 00115 uint32_t fftLen, 00116 float32_t * pATable, 00117 float32_t * pBTable, 00118 float32_t * pDst, 00119 uint32_t modifier); 00120 void arm_split_rifft_f32( 00121 float32_t * pSrc, 00122 uint32_t fftLen, 00123 float32_t * pATable, 00124 float32_t * pBTable, 00125 float32_t * pDst, 00126 uint32_t modifier); 00127 00141 void arm_rfft_f32( 00142 const arm_rfft_instance_f32 * S, 00143 float32_t * pSrc, 00144 float32_t * pDst) 00145 { 00146 const arm_cfft_radix4_instance_f32 *S_CFFT = S->pCfft; 00147 00148 00149 /* Calculation of Real IFFT of input */ 00150 if(S->ifftFlagR == 1u) 00151 { 00152 /* Real IFFT core process */ 00153 arm_split_rifft_f32(pSrc, S->fftLenBy2, S->pTwiddleAReal, 00154 S->pTwiddleBReal, pDst, S->twidCoefRModifier); 00155 00156 00157 /* Complex radix-4 IFFT process */ 00158 arm_radix4_butterfly_inverse_f32(pDst, S_CFFT->fftLen, 00159 S_CFFT->pTwiddle, 00160 S_CFFT->twidCoefModifier, 00161 S_CFFT->onebyfftLen); 00162 00163 /* Bit reversal process */ 00164 if(S->bitReverseFlagR == 1u) 00165 { 00166 arm_bitreversal_f32(pDst, S_CFFT->fftLen, 00167 S_CFFT->bitRevFactor, S_CFFT->pBitRevTable); 00168 } 00169 } 00170 else 00171 { 00172 00173 /* Calculation of RFFT of input */ 00174 00175 /* Complex radix-4 FFT process */ 00176 arm_radix4_butterfly_f32(pSrc, S_CFFT->fftLen, 00177 S_CFFT->pTwiddle, S_CFFT->twidCoefModifier); 00178 00179 /* Bit reversal process */ 00180 if(S->bitReverseFlagR == 1u) 00181 { 00182 arm_bitreversal_f32(pSrc, S_CFFT->fftLen, 00183 S_CFFT->bitRevFactor, S_CFFT->pBitRevTable); 00184 } 00185 00186 00187 /* Real FFT core process */ 00188 arm_split_rfft_f32(pSrc, S->fftLenBy2, S->pTwiddleAReal, 00189 S->pTwiddleBReal, pDst, S->twidCoefRModifier); 00190 } 00191 00192 } 00193 00209 void arm_split_rfft_f32( 00210 float32_t * pSrc, 00211 uint32_t fftLen, 00212 float32_t * pATable, 00213 float32_t * pBTable, 00214 float32_t * pDst, 00215 uint32_t modifier) 00216 { 00217 uint32_t i, blkCnt; /* Loop Counter */ 00218 float32_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */ 00219 float32_t *pDst1 = &pDst[2], *pDst2 = &pDst[(4u * fftLen) - 1u]; /* temp pointers for output buffer */ 00220 float32_t *pSrc1 = &pSrc[2], *pSrc2 = &pSrc[(2u * fftLen) - 1u]; /* temp pointers for input buffer */ 00221 float32_t inR1, inR2, inR4, inI1, inI2, inI4; 00222 float32_t CoefA1, CoefA2, CoefA3, CoefA4, CoefB1, CoefB2; 00223 float32_t outR1, outR2, outI1, outI2; 00224 00225 /* Init coefficient pointers */ 00226 pCoefA = &pATable[modifier * 2u]; 00227 pCoefB = &pBTable[modifier * 2u]; 00228 00229 blkCnt = fftLen - 1u; 00230 00231 /* Unroll by 2 */ 00232 i = blkCnt >> 1u; 00233 00234 while(i > 0u) 00235 { 00236 /* 00237 outR = (pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] 00238 + pSrc[2 * n - 2 * i] * pBTable[2 * i] + 00239 pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00240 */ 00241 00242 /* outI = (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] + 00243 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00244 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); */ 00245 /* 00246 outR = (pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] 00247 + pSrc[2 * n - 2 * i] * pBTable[2 * i] + 00248 pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00249 */ 00250 00251 /* outI = (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] + 00252 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00253 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); */ 00254 00255 inR1 = *pSrc1; 00256 inI1 = pSrc1[1]; 00257 00258 /* read pATable[2 * i] */ 00259 CoefA1 = pCoefA[0]; 00260 /* pATable[2 * i + 1] */ 00261 CoefA2 = pCoefA[1]; 00262 00263 /* read pABTable[2 * i] */ 00264 CoefB1 = *pCoefB; 00265 00266 /* update coefficient pointer */ 00267 pCoefA = pCoefA + (modifier * 2u); 00268 pCoefB = pCoefB + (modifier * 2u); 00269 00270 pSrc2 -= 4u; 00271 00272 /* read pATable[2 * i] */ 00273 CoefA3 = pCoefA[0]; 00274 /* pATable[2 * i + 1] */ 00275 CoefA4 = pCoefA[1]; 00276 00277 CoefB2 = *pCoefB; 00278 00279 /* update coefficient pointer */ 00280 pCoefA = pCoefA + (modifier * 2u); 00281 pCoefB = pCoefB + (modifier * 2u); 00282 00283 /* pSrc[2 * i] * pATable[2 * i] */ 00284 outR1 = inR1 * CoefA1; 00285 00286 inR2 = pSrc2[4]; 00287 inI2 = pSrc2[3]; 00288 00289 /* pSrc[2 * i] * CoefA2 */ 00290 outI1 = inR1 * CoefA2; 00291 00292 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00293 outR1 -= (inI1 + inR2) * CoefA2; 00294 00295 /* pSrc[2 * i + 1] * CoefA1 */ 00296 outI1 += inI1 * CoefA1; 00297 00298 /* pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00299 outI1 -= inR2 * CoefB1; 00300 00301 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00302 outI1 -= inI2 * CoefA2; 00303 00304 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00305 outR1 += inI2 * CoefB1; 00306 00307 pDst2 -= 4u; 00308 00309 inR1 = pSrc1[2]; 00310 inI1 = pSrc1[3]; 00311 00312 pSrc1 += 4u; 00313 00314 /* pSrc[2 * i] * pATable[2 * i] */ 00315 outR2 = inR1 * CoefA3; 00316 00317 inR4 = pSrc2[2]; 00318 00319 /* pSrc[2 * i] * CoefA2 */ 00320 outI2 = inR1 * CoefA4; 00321 00322 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00323 outR2 -= (inI1 + inR4) * CoefA4; 00324 00325 /* pSrc[2 * i + 1] * CoefA1 */ 00326 outI2 += inI1 * CoefA3; 00327 00328 inI4 = pSrc2[1]; 00329 00330 /* pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00331 outI2 -= inR4 * CoefB2; 00332 00333 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00334 outI2 -= inI4 * CoefA4; 00335 00336 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00337 outR2 += inI4 * CoefB2; 00338 00339 /* write output */ 00340 pDst1[0] = outR1; 00341 pDst1[1] = outI1; 00342 pDst1[2] = outR2; 00343 pDst1[3] = outI2; 00344 00345 /* write complex conjugate output */ 00346 pDst2[1] = outR2; 00347 pDst2[2] = -outI2; 00348 pDst2[3] = outR1; 00349 pDst2[4] = -outI1; 00350 00351 00352 pDst1 += 4u; 00353 00354 /* Decrement loop count */ 00355 i--; 00356 00357 } 00358 00359 i = blkCnt & 1u; 00360 00361 while(i > 0u) 00362 { 00363 /* 00364 outR = (pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] 00365 + pSrc[2 * n - 2 * i] * pBTable[2 * i] + 00366 pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00367 */ 00368 00369 /* outI = (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] + 00370 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00371 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); */ 00372 00373 inR1 = *pSrc1; 00374 00375 /* read pATable[2 * i] */ 00376 CoefA1 = *pCoefA++; 00377 /* pATable[2 * i + 1] */ 00378 CoefA2 = *pCoefA; 00379 00380 pSrc2 -= 2u; 00381 00382 /* pSrc[2 * i] * pATable[2 * i] */ 00383 outR1 = inR1 * CoefA1; 00384 00385 inI1 = pSrc1[1]; 00386 00387 inR2 = pSrc2[2]; 00388 00389 /* pSrc[2 * i] * CoefA2 */ 00390 outI1 = inR1 * CoefA2; 00391 00392 pSrc1 += 2u; 00393 00394 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00395 outR1 -= (inI1 + inR2) * CoefA2; 00396 00397 /* pSrc[2 * i + 1] * CoefA1 */ 00398 outI1 += inI1 * CoefA1; 00399 00400 CoefB1 = *pCoefB; 00401 00402 inI2 = pSrc2[1]; 00403 00404 /* pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00405 outI1 -= inR2 * CoefB1; 00406 00407 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00408 outI1 -= inI2 * CoefA2; 00409 00410 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00411 outR1 += inI2 * CoefB1; 00412 00413 pDst2 -= 2u; 00414 00415 /* write output */ 00416 pDst1[0] = outR1; 00417 pDst1[1] = outI1; 00418 00419 /* write complex conjugate output */ 00420 pDst2[2] = -outI1; 00421 pDst2[1] = outR1; 00422 00423 pDst1 += 2u; 00424 00425 /* update coefficient pointer */ 00426 pCoefB = pCoefB + (modifier * 2u); 00427 pCoefA = pCoefA + ((modifier * 2u) - 1u); 00428 00429 /* Decrement loop count */ 00430 i--; 00431 00432 } 00433 00434 pDst[2u * fftLen] = pSrc[0] - pSrc[1]; 00435 pDst[(2u * fftLen) + 1u] = 0.0f; 00436 00437 pDst[0] = pSrc[0] + pSrc[1]; 00438 pDst[1] = 0.0f; 00439 00440 } 00441 00442 00454 void arm_split_rifft_f32( 00455 float32_t * pSrc, 00456 uint32_t fftLen, 00457 float32_t * pATable, 00458 float32_t * pBTable, 00459 float32_t * pDst, 00460 uint32_t modifier) 00461 { 00462 uint32_t i, blkCnt; /* Loop Counter */ 00463 float32_t inVal1, inVal2, inVal3; 00464 float32_t outR, outI; /* Temporary variables for output */ 00465 float32_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */ 00466 float32_t *pSrc1 = &pSrc[0], *pSrc2 = &pSrc[(2u * fftLen) + 1u]; 00467 float32_t CoefA1, CoefA2, CoefB1; /* Temporary variables for twiddle coefficients */ 00468 00469 pCoefA = &pATable[0]; 00470 pCoefB = &pBTable[0]; 00471 00472 blkCnt = fftLen - 1u; 00473 00474 /* Unroll by 2 */ 00475 i = blkCnt >> 1u; 00476 00477 while(i > 0u) 00478 { 00479 /* 00480 outR = (pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] 00481 + pSrc[2 * n - 2 * i] * pBTable[2 * i] + 00482 pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00483 */ 00484 00485 /* outI = (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] + 00486 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00487 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); */ 00488 00489 /* 00490 outR = (pIn[2 * i] * pATable[2 * i] + pIn[2 * i + 1] * pATable[2 * i + 1] + 00491 pIn[2 * n - 2 * i] * pBTable[2 * i] - 00492 pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00493 00494 outI = (pIn[2 * i + 1] * pATable[2 * i] - pIn[2 * i] * pATable[2 * i + 1] - 00495 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00496 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); 00497 00498 */ 00499 00500 inVal1 = pSrc1[0]; 00501 00502 CoefA1 = pCoefA[0]; 00503 CoefA2 = pCoefA[1]; 00504 00505 CoefB1 = *pCoefB; 00506 00507 /* update coefficient pointer */ 00508 pCoefB = pCoefB + (modifier * 2u); 00509 pCoefA = pCoefA + (modifier * 2u); 00510 00511 pSrc2 -= 2u; 00512 /* outR = (pSrc[2 * i] * CoefA1 */ 00513 outR = inVal1 * CoefA1; 00514 00515 inVal2 = pSrc1[1]; 00516 00517 inVal3 = pSrc2[2]; 00518 00519 /* - pSrc[2 * i] * CoefA2 */ 00520 outI = -(inVal1) * CoefA2; 00521 00522 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00523 outR += (inVal2 + inVal3) * CoefA2; 00524 00525 pSrc1 += 2u; 00526 00527 /* pSrc[2 * i + 1] * CoefA1 */ 00528 outI += (inVal2) * CoefA1; 00529 00530 inVal1 = pSrc2[1]; 00531 00532 /* - pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00533 outI -= inVal3 * CoefB1; 00534 00535 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00536 outR += inVal1 * CoefB1; 00537 00538 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00539 outI += inVal1 * CoefA2; 00540 00541 /* write output */ 00542 pDst[0] = outR; 00543 pDst[1] = outI; 00544 00545 inVal1 = pSrc1[0]; 00546 00547 CoefA1 = pCoefA[0]; 00548 CoefA2 = pCoefA[1]; 00549 00550 CoefB1 = *pCoefB; 00551 00552 /* update coefficient pointer */ 00553 pCoefB = pCoefB + (modifier * 2u); 00554 pCoefA = pCoefA + (modifier * 2u); 00555 00556 pSrc2 -= 2u; 00557 /* outR = (pSrc[2 * i] * CoefA1 */ 00558 outR = inVal1 * CoefA1; 00559 00560 inVal2 = pSrc1[1]; 00561 00562 inVal3 = pSrc2[2]; 00563 00564 /* - pSrc[2 * i] * CoefA2 */ 00565 outI = -(inVal1) * CoefA2; 00566 00567 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00568 outR += (inVal2 + inVal3) * CoefA2; 00569 00570 pSrc1 += 2u; 00571 00572 /* pSrc[2 * i + 1] * CoefA1 */ 00573 outI += (inVal2) * CoefA1; 00574 00575 inVal1 = pSrc2[1]; 00576 00577 /* - pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00578 outI -= inVal3 * CoefB1; 00579 00580 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00581 outR += inVal1 * CoefB1; 00582 00583 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00584 outI += inVal1 * CoefA2; 00585 00586 /* write output */ 00587 pDst[2] = outR; 00588 pDst[3] = outI; 00589 00590 pDst += 4u; 00591 00592 /* decrement loop counter */ 00593 i--; 00594 00595 } 00596 00597 i = blkCnt & 1u; 00598 00599 while(i > 0u) 00600 { 00601 /* 00602 outR = (pIn[2 * i] * pATable[2 * i] + pIn[2 * i + 1] * pATable[2 * i + 1] + 00603 pIn[2 * n - 2 * i] * pBTable[2 * i] - 00604 pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); 00605 00606 outI = (pIn[2 * i + 1] * pATable[2 * i] - pIn[2 * i] * pATable[2 * i + 1] - 00607 pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - 00608 pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); 00609 00610 */ 00611 00612 inVal1 = pSrc1[0]; 00613 00614 CoefA1 = *pCoefA++; 00615 CoefA2 = *pCoefA; 00616 00617 pSrc2 -= 2u; 00618 /* outR = (pSrc[2 * i] * CoefA1 */ 00619 outR = inVal1 * CoefA1; 00620 00621 inVal2 = pSrc1[1]; 00622 00623 inVal3 = pSrc2[2]; 00624 00625 /* - pSrc[2 * i] * CoefA2 */ 00626 outI = -(inVal1) * CoefA2; 00627 00628 /* (pSrc[2 * i + 1] + pSrc[2 * fftLen - 2 * i + 1]) * CoefA2 */ 00629 outR += (inVal2 + inVal3) * CoefA2; 00630 00631 pSrc1 += 2u; 00632 00633 /* pSrc[2 * i + 1] * CoefA1 */ 00634 outI += (inVal2) * CoefA1; 00635 00636 00637 CoefB1 = *pCoefB; 00638 00639 inVal1 = pSrc2[1]; 00640 00641 /* - pSrc[2 * fftLen - 2 * i + 1] * CoefB1 */ 00642 outI -= inVal3 * CoefB1; 00643 00644 /* pSrc[2 * fftLen - 2 * i] * CoefB1 */ 00645 outR += inVal1 * CoefB1; 00646 00647 /* pSrc[2 * fftLen - 2 * i] * CoefA2 */ 00648 outI += inVal1 * CoefA2; 00649 00650 /* write output */ 00651 pDst[0] = outR; 00652 pDst[1] = outI; 00653 00654 pDst += 2u; 00655 00656 /* update coefficient pointer */ 00657 pCoefB = pCoefB + (modifier * 2u); 00658 pCoefA = pCoefA + ((modifier * 2u) - 1u); 00659 00660 /* Decrement loop count */ 00661 i--; 00662 } 00663 00664 } 00665
