1 /* ----------------------------------------------------------------------
2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
4 * $Date: 17. January 2013
7 * Project: CMSIS DSP Library
8 * Title: arm_cfft_radix2_q15.c
10 * Description: Radix-2 Decimation in Frequency CFFT & CIFFT Fixed point processing function
13 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
15 * Redistribution and use in source and binary forms, with or without
16 * modification, are permitted provided that the following conditions
18 * - Redistributions of source code must retain the above copyright
19 * notice, this list of conditions and the following disclaimer.
20 * - Redistributions in binary form must reproduce the above copyright
21 * notice, this list of conditions and the following disclaimer in
22 * the documentation and/or other materials provided with the
24 * - Neither the name of ARM LIMITED nor the names of its contributors
25 * may be used to endorse or promote products derived from this
26 * software without specific prior written permission.
28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
31 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
32 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
33 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
34 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
35 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
36 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
38 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
39 * POSSIBILITY OF SUCH DAMAGE.
40 * -------------------------------------------------------------------- */
44 void arm_radix2_butterfly_q15(
48 uint16_t twidCoefModifier);
50 void arm_radix2_butterfly_inverse_q15(
54 uint16_t twidCoefModifier);
56 void arm_bitreversal_q15(
59 uint16_t bitRevFactor,
60 uint16_t * pBitRevTab);
63 * @ingroup groupTransforms
67 * @addtogroup ComplexFFT
73 * @brief Processing function for the fixed-point CFFT/CIFFT.
74 * @param[in] *S points to an instance of the fixed-point CFFT/CIFFT structure.
75 * @param[in, out] *pSrc points to the complex data buffer of size <code>2*fftLen</code>. Processing occurs in-place.
79 void arm_cfft_radix2_q15(
80 const arm_cfft_radix2_instance_q15 * S,
86 arm_radix2_butterfly_inverse_q15(pSrc, S->fftLen,
87 S->pTwiddle, S->twidCoefModifier);
91 arm_radix2_butterfly_q15(pSrc, S->fftLen,
92 S->pTwiddle, S->twidCoefModifier);
95 arm_bitreversal_q15(pSrc, S->fftLen, S->bitRevFactor, S->pBitRevTable);
99 * @} end of ComplexFFT group
102 void arm_radix2_butterfly_q15(
106 uint16_t twidCoefModifier)
108 #ifndef ARM_MATH_CM0_FAMILY
114 q31_t coeff, out1, out2;
124 for (i = 0; i < n2; i++)
126 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
128 ia = ia + twidCoefModifier;
132 T = _SIMD32_OFFSET(pSrc + (2 * i));
133 in = ((int16_t) (T & 0xFFFF)) >> 2;
134 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
136 S = _SIMD32_OFFSET(pSrc + (2 * l));
137 in = ((int16_t) (S & 0xFFFF)) >> 2;
138 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
142 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
144 #ifndef ARM_MATH_BIG_ENDIAN
146 out1 = __SMUAD(coeff, R) >> 16;
147 out2 = __SMUSDX(coeff, R);
151 out1 = __SMUSDX(R, coeff) >> 16u;
152 out2 = __SMUAD(coeff, R);
154 #endif // #ifndef ARM_MATH_BIG_ENDIAN
156 _SIMD32_OFFSET(pSrc + (2u * l)) =
157 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
159 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
161 ia = ia + twidCoefModifier;
163 // loop for butterfly
167 T = _SIMD32_OFFSET(pSrc + (2 * i));
168 in = ((int16_t) (T & 0xFFFF)) >> 2;
169 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
171 S = _SIMD32_OFFSET(pSrc + (2 * l));
172 in = ((int16_t) (S & 0xFFFF)) >> 2;
173 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
177 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
179 #ifndef ARM_MATH_BIG_ENDIAN
181 out1 = __SMUAD(coeff, R) >> 16;
182 out2 = __SMUSDX(coeff, R);
186 out1 = __SMUSDX(R, coeff) >> 16u;
187 out2 = __SMUAD(coeff, R);
189 #endif // #ifndef ARM_MATH_BIG_ENDIAN
191 _SIMD32_OFFSET(pSrc + (2u * l)) =
192 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
196 twidCoefModifier = twidCoefModifier << 1u;
199 for (k = fftLen / 2; k > 2; k = k >> 1)
206 for (j = 0; j < n2; j++)
208 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
210 ia = ia + twidCoefModifier;
212 // loop for butterfly
213 for (i = j; i < fftLen; i += n1)
217 T = _SIMD32_OFFSET(pSrc + (2 * i));
219 S = _SIMD32_OFFSET(pSrc + (2 * l));
223 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
225 #ifndef ARM_MATH_BIG_ENDIAN
227 out1 = __SMUAD(coeff, R) >> 16;
228 out2 = __SMUSDX(coeff, R);
232 out1 = __SMUSDX(R, coeff) >> 16u;
233 out2 = __SMUAD(coeff, R);
235 #endif // #ifndef ARM_MATH_BIG_ENDIAN
237 _SIMD32_OFFSET(pSrc + (2u * l)) =
238 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
244 T = _SIMD32_OFFSET(pSrc + (2 * i));
246 S = _SIMD32_OFFSET(pSrc + (2 * l));
250 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
252 #ifndef ARM_MATH_BIG_ENDIAN
254 out1 = __SMUAD(coeff, R) >> 16;
255 out2 = __SMUSDX(coeff, R);
259 out1 = __SMUSDX(R, coeff) >> 16u;
260 out2 = __SMUAD(coeff, R);
262 #endif // #ifndef ARM_MATH_BIG_ENDIAN
264 _SIMD32_OFFSET(pSrc + (2u * l)) =
265 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
267 } // butterfly loop end
271 twidCoefModifier = twidCoefModifier << 1u;
278 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
280 ia = ia + twidCoefModifier;
282 // loop for butterfly
283 for (i = 0; i < fftLen; i += n1)
287 T = _SIMD32_OFFSET(pSrc + (2 * i));
289 S = _SIMD32_OFFSET(pSrc + (2 * l));
293 _SIMD32_OFFSET(pSrc + (2 * i)) = __QADD16(T, S);
295 _SIMD32_OFFSET(pSrc + (2u * l)) = R;
300 T = _SIMD32_OFFSET(pSrc + (2 * i));
302 S = _SIMD32_OFFSET(pSrc + (2 * l));
306 _SIMD32_OFFSET(pSrc + (2 * i)) = __QADD16(T, S);
308 _SIMD32_OFFSET(pSrc + (2u * l)) = R;
317 q15_t xt, yt, cosVal, sinVal;
328 for (j = 0; j < n2; j++)
330 cosVal = pCoef[ia * 2];
331 sinVal = pCoef[(ia * 2) + 1];
332 ia = ia + twidCoefModifier;
334 // loop for butterfly
335 for (i = j; i < fftLen; i += n1)
338 xt = (pSrc[2 * i] >> 2u) - (pSrc[2 * l] >> 2u);
339 pSrc[2 * i] = ((pSrc[2 * i] >> 2u) + (pSrc[2 * l] >> 2u)) >> 1u;
341 yt = (pSrc[2 * i + 1] >> 2u) - (pSrc[2 * l + 1] >> 2u);
343 ((pSrc[2 * l + 1] >> 2u) + (pSrc[2 * i + 1] >> 2u)) >> 1u;
345 pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) +
346 ((int16_t) (((q31_t) yt * sinVal) >> 16)));
348 pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) -
349 ((int16_t) (((q31_t) xt * sinVal) >> 16)));
351 } // butterfly loop end
355 twidCoefModifier = twidCoefModifier << 1u;
358 for (k = fftLen / 2; k > 2; k = k >> 1)
365 for (j = 0; j < n2; j++)
367 cosVal = pCoef[ia * 2];
368 sinVal = pCoef[(ia * 2) + 1];
369 ia = ia + twidCoefModifier;
371 // loop for butterfly
372 for (i = j; i < fftLen; i += n1)
375 xt = pSrc[2 * i] - pSrc[2 * l];
376 pSrc[2 * i] = (pSrc[2 * i] + pSrc[2 * l]) >> 1u;
378 yt = pSrc[2 * i + 1] - pSrc[2 * l + 1];
379 pSrc[2 * i + 1] = (pSrc[2 * l + 1] + pSrc[2 * i + 1]) >> 1u;
381 pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) +
382 ((int16_t) (((q31_t) yt * sinVal) >> 16)));
384 pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) -
385 ((int16_t) (((q31_t) xt * sinVal) >> 16)));
387 } // butterfly loop end
391 twidCoefModifier = twidCoefModifier << 1u;
399 for (j = 0; j < n2; j++)
401 cosVal = pCoef[ia * 2];
402 sinVal = pCoef[(ia * 2) + 1];
404 ia = ia + twidCoefModifier;
406 // loop for butterfly
407 for (i = j; i < fftLen; i += n1)
410 xt = pSrc[2 * i] - pSrc[2 * l];
411 pSrc[2 * i] = (pSrc[2 * i] + pSrc[2 * l]);
413 yt = pSrc[2 * i + 1] - pSrc[2 * l + 1];
414 pSrc[2 * i + 1] = (pSrc[2 * l + 1] + pSrc[2 * i + 1]);
418 pSrc[2u * l + 1u] = yt;
420 } // butterfly loop end
424 twidCoefModifier = twidCoefModifier << 1u;
426 #endif // #ifndef ARM_MATH_CM0_FAMILY
431 void arm_radix2_butterfly_inverse_q15(
435 uint16_t twidCoefModifier)
437 #ifndef ARM_MATH_CM0_FAMILY
443 q31_t coeff, out1, out2;
453 for (i = 0; i < n2; i++)
455 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
457 ia = ia + twidCoefModifier;
461 T = _SIMD32_OFFSET(pSrc + (2 * i));
462 in = ((int16_t) (T & 0xFFFF)) >> 2;
463 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
465 S = _SIMD32_OFFSET(pSrc + (2 * l));
466 in = ((int16_t) (S & 0xFFFF)) >> 2;
467 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
471 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
473 #ifndef ARM_MATH_BIG_ENDIAN
475 out1 = __SMUSD(coeff, R) >> 16;
476 out2 = __SMUADX(coeff, R);
479 out1 = __SMUADX(R, coeff) >> 16u;
480 out2 = __SMUSD(__QSUB(0, coeff), R);
482 #endif // #ifndef ARM_MATH_BIG_ENDIAN
484 _SIMD32_OFFSET(pSrc + (2u * l)) =
485 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
487 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
489 ia = ia + twidCoefModifier;
491 // loop for butterfly
495 T = _SIMD32_OFFSET(pSrc + (2 * i));
496 in = ((int16_t) (T & 0xFFFF)) >> 2;
497 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
499 S = _SIMD32_OFFSET(pSrc + (2 * l));
500 in = ((int16_t) (S & 0xFFFF)) >> 2;
501 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
505 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
507 #ifndef ARM_MATH_BIG_ENDIAN
509 out1 = __SMUSD(coeff, R) >> 16;
510 out2 = __SMUADX(coeff, R);
513 out1 = __SMUADX(R, coeff) >> 16u;
514 out2 = __SMUSD(__QSUB(0, coeff), R);
516 #endif // #ifndef ARM_MATH_BIG_ENDIAN
518 _SIMD32_OFFSET(pSrc + (2u * l)) =
519 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
523 twidCoefModifier = twidCoefModifier << 1u;
526 for (k = fftLen / 2; k > 2; k = k >> 1)
533 for (j = 0; j < n2; j++)
535 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
537 ia = ia + twidCoefModifier;
539 // loop for butterfly
540 for (i = j; i < fftLen; i += n1)
544 T = _SIMD32_OFFSET(pSrc + (2 * i));
546 S = _SIMD32_OFFSET(pSrc + (2 * l));
550 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
552 #ifndef ARM_MATH_BIG_ENDIAN
554 out1 = __SMUSD(coeff, R) >> 16;
555 out2 = __SMUADX(coeff, R);
559 out1 = __SMUADX(R, coeff) >> 16u;
560 out2 = __SMUSD(__QSUB(0, coeff), R);
562 #endif // #ifndef ARM_MATH_BIG_ENDIAN
564 _SIMD32_OFFSET(pSrc + (2u * l)) =
565 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
571 T = _SIMD32_OFFSET(pSrc + (2 * i));
573 S = _SIMD32_OFFSET(pSrc + (2 * l));
577 _SIMD32_OFFSET(pSrc + (2 * i)) = __SHADD16(T, S);
579 #ifndef ARM_MATH_BIG_ENDIAN
581 out1 = __SMUSD(coeff, R) >> 16;
582 out2 = __SMUADX(coeff, R);
585 out1 = __SMUADX(R, coeff) >> 16u;
586 out2 = __SMUSD(__QSUB(0, coeff), R);
588 #endif // #ifndef ARM_MATH_BIG_ENDIAN
590 _SIMD32_OFFSET(pSrc + (2u * l)) =
591 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
593 } // butterfly loop end
597 twidCoefModifier = twidCoefModifier << 1u;
605 for (j = 0; j < n2; j++)
607 coeff = _SIMD32_OFFSET(pCoef + (ia * 2u));
609 ia = ia + twidCoefModifier;
611 // loop for butterfly
612 for (i = j; i < fftLen; i += n1)
616 T = _SIMD32_OFFSET(pSrc + (2 * i));
618 S = _SIMD32_OFFSET(pSrc + (2 * l));
622 _SIMD32_OFFSET(pSrc + (2 * i)) = __QADD16(T, S);
624 _SIMD32_OFFSET(pSrc + (2u * l)) = R;
626 } // butterfly loop end
630 twidCoefModifier = twidCoefModifier << 1u;
637 q15_t xt, yt, cosVal, sinVal;
647 for (j = 0; j < n2; j++)
649 cosVal = pCoef[ia * 2];
650 sinVal = pCoef[(ia * 2) + 1];
651 ia = ia + twidCoefModifier;
653 // loop for butterfly
654 for (i = j; i < fftLen; i += n1)
657 xt = (pSrc[2 * i] >> 2u) - (pSrc[2 * l] >> 2u);
658 pSrc[2 * i] = ((pSrc[2 * i] >> 2u) + (pSrc[2 * l] >> 2u)) >> 1u;
660 yt = (pSrc[2 * i + 1] >> 2u) - (pSrc[2 * l + 1] >> 2u);
662 ((pSrc[2 * l + 1] >> 2u) + (pSrc[2 * i + 1] >> 2u)) >> 1u;
664 pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) -
665 ((int16_t) (((q31_t) yt * sinVal) >> 16)));
667 pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) +
668 ((int16_t) (((q31_t) xt * sinVal) >> 16)));
670 } // butterfly loop end
674 twidCoefModifier = twidCoefModifier << 1u;
677 for (k = fftLen / 2; k > 2; k = k >> 1)
684 for (j = 0; j < n2; j++)
686 cosVal = pCoef[ia * 2];
687 sinVal = pCoef[(ia * 2) + 1];
688 ia = ia + twidCoefModifier;
690 // loop for butterfly
691 for (i = j; i < fftLen; i += n1)
694 xt = pSrc[2 * i] - pSrc[2 * l];
695 pSrc[2 * i] = (pSrc[2 * i] + pSrc[2 * l]) >> 1u;
697 yt = pSrc[2 * i + 1] - pSrc[2 * l + 1];
698 pSrc[2 * i + 1] = (pSrc[2 * l + 1] + pSrc[2 * i + 1]) >> 1u;
700 pSrc[2u * l] = (((int16_t) (((q31_t) xt * cosVal) >> 16)) -
701 ((int16_t) (((q31_t) yt * sinVal) >> 16)));
703 pSrc[2u * l + 1u] = (((int16_t) (((q31_t) yt * cosVal) >> 16)) +
704 ((int16_t) (((q31_t) xt * sinVal) >> 16)));
706 } // butterfly loop end
710 twidCoefModifier = twidCoefModifier << 1u;
717 cosVal = pCoef[ia * 2];
718 sinVal = pCoef[(ia * 2) + 1];
720 ia = ia + twidCoefModifier;
722 // loop for butterfly
723 for (i = 0; i < fftLen; i += n1)
726 xt = pSrc[2 * i] - pSrc[2 * l];
727 pSrc[2 * i] = (pSrc[2 * i] + pSrc[2 * l]);
729 yt = pSrc[2 * i + 1] - pSrc[2 * l + 1];
730 pSrc[2 * i + 1] = (pSrc[2 * l + 1] + pSrc[2 * i + 1]);
734 pSrc[2u * l + 1u] = yt;
739 #endif // #ifndef ARM_MATH_CM0_FAMILY