1 /* ----------------------------------------------------------------------
2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
3 *
4 * $Date:        31. July 2014
5 * $Revision: 	V1.4.4
6 *
7 * Project: 	    CMSIS DSP Library
8 * Title:	    arm_cfft_f32.c
9 *
10 * Description:	Combined Radix Decimation in Frequency CFFT Floating point processing function
11 *
12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 *   - Redistributions of source code must retain the above copyright
18 *     notice, this list of conditions and the following disclaimer.
19 *   - Redistributions in binary form must reproduce the above copyright
20 *     notice, this list of conditions and the following disclaimer in
21 *     the documentation and/or other materials provided with the
22 *     distribution.
23 *   - Neither the name of ARM LIMITED nor the names of its contributors
24 *     may be used to endorse or promote products derived from this
25 *     software without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
38 * POSSIBILITY OF SUCH DAMAGE.
39 * -------------------------------------------------------------------- */
40 
41 #include "arm_math.h"
42 #include "arm_common_tables.h"
43 
44 extern void arm_radix8_butterfly_f32(
45     float32_t * pSrc,
46     uint16_t fftLen,
47     const float32_t * pCoef,
48     uint16_t twidCoefModifier);
49 
50 extern void arm_bitreversal_32(
51     uint32_t * pSrc,
52     const uint16_t bitRevLen,
53     const uint16_t * pBitRevTable);
54 
55 /**
56 * @ingroup groupTransforms
57 */
58 
59 /**
60 * @defgroup ComplexFFT Complex FFT Functions
61 *
62 * \par
63 * The Fast Fourier Transform (FFT) is an efficient algorithm for computing the
64 * Discrete Fourier Transform (DFT).  The FFT can be orders of magnitude faster
65 * than the DFT, especially for long lengths.
66 * The algorithms described in this section
67 * operate on complex data.  A separate set of functions is devoted to handling
68 * of real sequences.
69 * \par
70 * There are separate algorithms for handling floating-point, Q15, and Q31 data
71 * types.  The algorithms available for each data type are described next.
72 * \par
73 * The FFT functions operate in-place.  That is, the array holding the input data
74 * will also be used to hold the corresponding result.  The input data is complex
75 * and contains <code>2*fftLen</code> interleaved values as shown below.
76 * <pre> {real[0], imag[0], real[1], imag[1],..} </pre>
77 * The FFT result will be contained in the same array and the frequency domain
78 * values will have the same interleaving.
79 *
80 * \par Floating-point
81 * The floating-point complex FFT uses a mixed-radix algorithm.  Multiple radix-8
82 * stages are performed along with a single radix-2 or radix-4 stage, as needed.
83 * The algorithm supports lengths of [16, 32, 64, ..., 4096] and each length uses
84 * a different twiddle factor table.
85 * \par
86 * The function uses the standard FFT definition and output values may grow by a
87 * factor of <code>fftLen</code> when computing the forward transform.  The
88 * inverse transform includes a scale of <code>1/fftLen</code> as part of the
89 * calculation and this matches the textbook definition of the inverse FFT.
90 * \par
91 * Pre-initialized data structures containing twiddle factors and bit reversal
92 * tables are provided and defined in <code>arm_const_structs.h</code>.  Include
93 * this header in your function and then pass one of the constant structures as
94 * an argument to arm_cfft_f32.  For example:
95 * \par
96 * <code>arm_cfft_f32(arm_cfft_sR_f32_len64, pSrc, 1, 1)</code>
97 * \par
98 * computes a 64-point inverse complex FFT including bit reversal.
99 * The data structures are treated as constant data and not modified during the
100 * calculation.  The same data structure can be reused for multiple transforms
101 * including mixing forward and inverse transforms.
102 * \par
103 * Earlier releases of the library provided separate radix-2 and radix-4
104 * algorithms that operated on floating-point data.  These functions are still
105 * provided but are deprecated.  The older functions are slower and less general
106 * than the new functions.
107 * \par
108 * An example of initialization of the constants for the arm_cfft_f32 function follows:
109 * \code
110 * const static arm_cfft_instance_f32 *S;
111 * ...
112 *   switch (length) {
113 *     case 16:
114 *       S = &arm_cfft_sR_f32_len16;
115 *       break;
116 *     case 32:
117 *       S = &arm_cfft_sR_f32_len32;
118 *       break;
119 *     case 64:
120 *       S = &arm_cfft_sR_f32_len64;
121 *       break;
122 *     case 128:
123 *       S = &arm_cfft_sR_f32_len128;
124 *       break;
125 *     case 256:
126 *       S = &arm_cfft_sR_f32_len256;
127 *       break;
128 *     case 512:
129 *       S = &arm_cfft_sR_f32_len512;
130 *       break;
131 *     case 1024:
132 *       S = &arm_cfft_sR_f32_len1024;
133 *       break;
134 *     case 2048:
135 *       S = &arm_cfft_sR_f32_len2048;
136 *       break;
137 *     case 4096:
138 *       S = &arm_cfft_sR_f32_len4096;
139 *       break;
140 *   }
141 * \endcode
142 * \par Q15 and Q31
143 * The floating-point complex FFT uses a mixed-radix algorithm.  Multiple radix-4
144 * stages are performed along with a single radix-2 stage, as needed.
145 * The algorithm supports lengths of [16, 32, 64, ..., 4096] and each length uses
146 * a different twiddle factor table.
147 * \par
148 * The function uses the standard FFT definition and output values may grow by a
149 * factor of <code>fftLen</code> when computing the forward transform.  The
150 * inverse transform includes a scale of <code>1/fftLen</code> as part of the
151 * calculation and this matches the textbook definition of the inverse FFT.
152 * \par
153 * Pre-initialized data structures containing twiddle factors and bit reversal
154 * tables are provided and defined in <code>arm_const_structs.h</code>.  Include
155 * this header in your function and then pass one of the constant structures as
156 * an argument to arm_cfft_q31.  For example:
157 * \par
158 * <code>arm_cfft_q31(arm_cfft_sR_q31_len64, pSrc, 1, 1)</code>
159 * \par
160 * computes a 64-point inverse complex FFT including bit reversal.
161 * The data structures are treated as constant data and not modified during the
162 * calculation.  The same data structure can be reused for multiple transforms
163 * including mixing forward and inverse transforms.
164 * \par
165 * Earlier releases of the library provided separate radix-2 and radix-4
166 * algorithms that operated on floating-point data.  These functions are still
167 * provided but are deprecated.  The older functions are slower and less general
168 * than the new functions.
169 * \par
170 * An example of initialization of the constants for the arm_cfft_q31 function follows:
171 * \code
172 * const static arm_cfft_instance_q31 *S;
173 * ...
174 *   switch (length) {
175 *     case 16:
176 *       S = &arm_cfft_sR_q31_len16;
177 *       break;
178 *     case 32:
179 *       S = &arm_cfft_sR_q31_len32;
180 *       break;
181 *     case 64:
182 *       S = &arm_cfft_sR_q31_len64;
183 *       break;
184 *     case 128:
185 *       S = &arm_cfft_sR_q31_len128;
186 *       break;
187 *     case 256:
188 *       S = &arm_cfft_sR_q31_len256;
189 *       break;
190 *     case 512:
191 *       S = &arm_cfft_sR_q31_len512;
192 *       break;
193 *     case 1024:
194 *       S = &arm_cfft_sR_q31_len1024;
195 *       break;
196 *     case 2048:
197 *       S = &arm_cfft_sR_q31_len2048;
198 *       break;
199 *     case 4096:
200 *       S = &arm_cfft_sR_q31_len4096;
201 *       break;
202 *   }
203 * \endcode
204 *
205 */
206 
arm_cfft_radix8by2_f32(arm_cfft_instance_f32 * S,float32_t * p1)207 void arm_cfft_radix8by2_f32( arm_cfft_instance_f32 * S, float32_t * p1)
208 {
209     uint32_t    L  = S->fftLen;
210     float32_t * pCol1, * pCol2, * pMid1, * pMid2;
211     float32_t * p2 = p1 + L;
212     const float32_t * tw = (float32_t *) S->pTwiddle;
213     float32_t t1[4], t2[4], t3[4], t4[4], twR, twI;
214     float32_t m0, m1, m2, m3;
215     uint32_t l;
216 
217     pCol1 = p1;
218     pCol2 = p2;
219 
220     //    Define new length
221     L >>= 1;
222     //    Initialize mid pointers
223     pMid1 = p1 + L;
224     pMid2 = p2 + L;
225 
226     // do two dot Fourier transform
227     for ( l = L >> 2; l > 0; l-- )
228     {
229         t1[0] = p1[0];
230         t1[1] = p1[1];
231         t1[2] = p1[2];
232         t1[3] = p1[3];
233 
234         t2[0] = p2[0];
235         t2[1] = p2[1];
236         t2[2] = p2[2];
237         t2[3] = p2[3];
238 
239         t3[0] = pMid1[0];
240         t3[1] = pMid1[1];
241         t3[2] = pMid1[2];
242         t3[3] = pMid1[3];
243 
244         t4[0] = pMid2[0];
245         t4[1] = pMid2[1];
246         t4[2] = pMid2[2];
247         t4[3] = pMid2[3];
248 
249         *p1++ = t1[0] + t2[0];
250         *p1++ = t1[1] + t2[1];
251         *p1++ = t1[2] + t2[2];
252         *p1++ = t1[3] + t2[3];    // col 1
253 
254         t2[0] = t1[0] - t2[0];
255         t2[1] = t1[1] - t2[1];
256         t2[2] = t1[2] - t2[2];
257         t2[3] = t1[3] - t2[3];    // for col 2
258 
259         *pMid1++ = t3[0] + t4[0];
260         *pMid1++ = t3[1] + t4[1];
261         *pMid1++ = t3[2] + t4[2];
262         *pMid1++ = t3[3] + t4[3]; // col 1
263 
264         t4[0] = t4[0] - t3[0];
265         t4[1] = t4[1] - t3[1];
266         t4[2] = t4[2] - t3[2];
267         t4[3] = t4[3] - t3[3];    // for col 2
268 
269         twR = *tw++;
270         twI = *tw++;
271 
272         // multiply by twiddle factors
273         m0 = t2[0] * twR;
274         m1 = t2[1] * twI;
275         m2 = t2[1] * twR;
276         m3 = t2[0] * twI;
277 
278         // R  =  R  *  Tr - I * Ti
279         *p2++ = m0 + m1;
280         // I  =  I  *  Tr + R * Ti
281         *p2++ = m2 - m3;
282 
283         // use vertical symmetry
284         //  0.9988 - 0.0491i <==> -0.0491 - 0.9988i
285         m0 = t4[0] * twI;
286         m1 = t4[1] * twR;
287         m2 = t4[1] * twI;
288         m3 = t4[0] * twR;
289 
290         *pMid2++ = m0 - m1;
291         *pMid2++ = m2 + m3;
292 
293         twR = *tw++;
294         twI = *tw++;
295 
296         m0 = t2[2] * twR;
297         m1 = t2[3] * twI;
298         m2 = t2[3] * twR;
299         m3 = t2[2] * twI;
300 
301         *p2++ = m0 + m1;
302         *p2++ = m2 - m3;
303 
304         m0 = t4[2] * twI;
305         m1 = t4[3] * twR;
306         m2 = t4[3] * twI;
307         m3 = t4[2] * twR;
308 
309         *pMid2++ = m0 - m1;
310         *pMid2++ = m2 + m3;
311     }
312 
313     // first col
314     arm_radix8_butterfly_f32( pCol1, L, (float32_t *) S->pTwiddle, 2u);
315     // second col
316     arm_radix8_butterfly_f32( pCol2, L, (float32_t *) S->pTwiddle, 2u);
317 }
318 
arm_cfft_radix8by4_f32(arm_cfft_instance_f32 * S,float32_t * p1)319 void arm_cfft_radix8by4_f32( arm_cfft_instance_f32 * S, float32_t * p1)
320 {
321     uint32_t    L  = S->fftLen >> 1;
322     float32_t * pCol1, *pCol2, *pCol3, *pCol4, *pEnd1, *pEnd2, *pEnd3, *pEnd4;
323     const float32_t *tw2, *tw3, *tw4;
324     float32_t * p2 = p1 + L;
325     float32_t * p3 = p2 + L;
326     float32_t * p4 = p3 + L;
327     float32_t t2[4], t3[4], t4[4], twR, twI;
328     float32_t p1ap3_0, p1sp3_0, p1ap3_1, p1sp3_1;
329     float32_t m0, m1, m2, m3;
330     uint32_t l, twMod2, twMod3, twMod4;
331 
332     pCol1 = p1;         // points to real values by default
333     pCol2 = p2;
334     pCol3 = p3;
335     pCol4 = p4;
336     pEnd1 = p2 - 1;     // points to imaginary values by default
337     pEnd2 = p3 - 1;
338     pEnd3 = p4 - 1;
339     pEnd4 = pEnd3 + L;
340 
341     tw2 = tw3 = tw4 = (float32_t *) S->pTwiddle;
342 
343     L >>= 1;
344 
345     // do four dot Fourier transform
346 
347     twMod2 = 2;
348     twMod3 = 4;
349     twMod4 = 6;
350 
351     // TOP
352     p1ap3_0 = p1[0] + p3[0];
353     p1sp3_0 = p1[0] - p3[0];
354     p1ap3_1 = p1[1] + p3[1];
355     p1sp3_1 = p1[1] - p3[1];
356 
357     // col 2
358     t2[0] = p1sp3_0 + p2[1] - p4[1];
359     t2[1] = p1sp3_1 - p2[0] + p4[0];
360     // col 3
361     t3[0] = p1ap3_0 - p2[0] - p4[0];
362     t3[1] = p1ap3_1 - p2[1] - p4[1];
363     // col 4
364     t4[0] = p1sp3_0 - p2[1] + p4[1];
365     t4[1] = p1sp3_1 + p2[0] - p4[0];
366     // col 1
367     *p1++ = p1ap3_0 + p2[0] + p4[0];
368     *p1++ = p1ap3_1 + p2[1] + p4[1];
369 
370     // Twiddle factors are ones
371     *p2++ = t2[0];
372     *p2++ = t2[1];
373     *p3++ = t3[0];
374     *p3++ = t3[1];
375     *p4++ = t4[0];
376     *p4++ = t4[1];
377 
378     tw2 += twMod2;
379     tw3 += twMod3;
380     tw4 += twMod4;
381 
382     for (l = (L - 2) >> 1; l > 0; l-- )
383     {
384         // TOP
385         p1ap3_0 = p1[0] + p3[0];
386         p1sp3_0 = p1[0] - p3[0];
387         p1ap3_1 = p1[1] + p3[1];
388         p1sp3_1 = p1[1] - p3[1];
389         // col 2
390         t2[0] = p1sp3_0 + p2[1] - p4[1];
391         t2[1] = p1sp3_1 - p2[0] + p4[0];
392         // col 3
393         t3[0] = p1ap3_0 - p2[0] - p4[0];
394         t3[1] = p1ap3_1 - p2[1] - p4[1];
395         // col 4
396         t4[0] = p1sp3_0 - p2[1] + p4[1];
397         t4[1] = p1sp3_1 + p2[0] - p4[0];
398         // col 1 - top
399         *p1++ = p1ap3_0 + p2[0] + p4[0];
400         *p1++ = p1ap3_1 + p2[1] + p4[1];
401 
402         // BOTTOM
403         p1ap3_1 = pEnd1[-1] + pEnd3[-1];
404         p1sp3_1 = pEnd1[-1] - pEnd3[-1];
405         p1ap3_0 = pEnd1[0] + pEnd3[0];
406         p1sp3_0 = pEnd1[0] - pEnd3[0];
407         // col 2
408         t2[2] = pEnd2[0]  - pEnd4[0] + p1sp3_1;
409         t2[3] = pEnd1[0] - pEnd3[0] - pEnd2[-1] + pEnd4[-1];
410         // col 3
411         t3[2] = p1ap3_1 - pEnd2[-1] - pEnd4[-1];
412         t3[3] = p1ap3_0 - pEnd2[0]  - pEnd4[0];
413         // col 4
414         t4[2] = pEnd2[0]  - pEnd4[0]  - p1sp3_1;
415         t4[3] = pEnd4[-1] - pEnd2[-1] - p1sp3_0;
416         // col 1 - Bottom
417         *pEnd1-- = p1ap3_0 + pEnd2[0] + pEnd4[0];
418         *pEnd1-- = p1ap3_1 + pEnd2[-1] + pEnd4[-1];
419 
420         // COL 2
421         // read twiddle factors
422         twR = *tw2++;
423         twI = *tw2++;
424         // multiply by twiddle factors
425         //  let    Z1 = a + i(b),   Z2 = c + i(d)
426         //   =>  Z1 * Z2  =  (a*c - b*d) + i(b*c + a*d)
427 
428         // Top
429         m0 = t2[0] * twR;
430         m1 = t2[1] * twI;
431         m2 = t2[1] * twR;
432         m3 = t2[0] * twI;
433 
434         *p2++ = m0 + m1;
435         *p2++ = m2 - m3;
436         // use vertical symmetry col 2
437         // 0.9997 - 0.0245i  <==>  0.0245 - 0.9997i
438         // Bottom
439         m0 = t2[3] * twI;
440         m1 = t2[2] * twR;
441         m2 = t2[2] * twI;
442         m3 = t2[3] * twR;
443 
444         *pEnd2-- = m0 - m1;
445         *pEnd2-- = m2 + m3;
446 
447         // COL 3
448         twR = tw3[0];
449         twI = tw3[1];
450         tw3 += twMod3;
451         // Top
452         m0 = t3[0] * twR;
453         m1 = t3[1] * twI;
454         m2 = t3[1] * twR;
455         m3 = t3[0] * twI;
456 
457         *p3++ = m0 + m1;
458         *p3++ = m2 - m3;
459         // use vertical symmetry col 3
460         // 0.9988 - 0.0491i  <==>  -0.9988 - 0.0491i
461         // Bottom
462         m0 = -t3[3] * twR;
463         m1 = t3[2] * twI;
464         m2 = t3[2] * twR;
465         m3 = t3[3] * twI;
466 
467         *pEnd3-- = m0 - m1;
468         *pEnd3-- = m3 - m2;
469 
470         // COL 4
471         twR = tw4[0];
472         twI = tw4[1];
473         tw4 += twMod4;
474         // Top
475         m0 = t4[0] * twR;
476         m1 = t4[1] * twI;
477         m2 = t4[1] * twR;
478         m3 = t4[0] * twI;
479 
480         *p4++ = m0 + m1;
481         *p4++ = m2 - m3;
482         // use vertical symmetry col 4
483         // 0.9973 - 0.0736i  <==>  -0.0736 + 0.9973i
484         // Bottom
485         m0 = t4[3] * twI;
486         m1 = t4[2] * twR;
487         m2 = t4[2] * twI;
488         m3 = t4[3] * twR;
489 
490         *pEnd4-- = m0 - m1;
491         *pEnd4-- = m2 + m3;
492     }
493 
494     //MIDDLE
495     // Twiddle factors are
496     //  1.0000  0.7071-0.7071i  -1.0000i  -0.7071-0.7071i
497     p1ap3_0 = p1[0] + p3[0];
498     p1sp3_0 = p1[0] - p3[0];
499     p1ap3_1 = p1[1] + p3[1];
500     p1sp3_1 = p1[1] - p3[1];
501 
502     // col 2
503     t2[0] = p1sp3_0 + p2[1] - p4[1];
504     t2[1] = p1sp3_1 - p2[0] + p4[0];
505     // col 3
506     t3[0] = p1ap3_0 - p2[0] - p4[0];
507     t3[1] = p1ap3_1 - p2[1] - p4[1];
508     // col 4
509     t4[0] = p1sp3_0 - p2[1] + p4[1];
510     t4[1] = p1sp3_1 + p2[0] - p4[0];
511     // col 1 - Top
512     *p1++ = p1ap3_0 + p2[0] + p4[0];
513     *p1++ = p1ap3_1 + p2[1] + p4[1];
514 
515     // COL 2
516     twR = tw2[0];
517     twI = tw2[1];
518 
519     m0 = t2[0] * twR;
520     m1 = t2[1] * twI;
521     m2 = t2[1] * twR;
522     m3 = t2[0] * twI;
523 
524     *p2++ = m0 + m1;
525     *p2++ = m2 - m3;
526     // COL 3
527     twR = tw3[0];
528     twI = tw3[1];
529 
530     m0 = t3[0] * twR;
531     m1 = t3[1] * twI;
532     m2 = t3[1] * twR;
533     m3 = t3[0] * twI;
534 
535     *p3++ = m0 + m1;
536     *p3++ = m2 - m3;
537     // COL 4
538     twR = tw4[0];
539     twI = tw4[1];
540 
541     m0 = t4[0] * twR;
542     m1 = t4[1] * twI;
543     m2 = t4[1] * twR;
544     m3 = t4[0] * twI;
545 
546     *p4++ = m0 + m1;
547     *p4++ = m2 - m3;
548 
549     // first col
550     arm_radix8_butterfly_f32( pCol1, L, (float32_t *) S->pTwiddle, 4u);
551     // second col
552     arm_radix8_butterfly_f32( pCol2, L, (float32_t *) S->pTwiddle, 4u);
553     // third col
554     arm_radix8_butterfly_f32( pCol3, L, (float32_t *) S->pTwiddle, 4u);
555     // fourth col
556     arm_radix8_butterfly_f32( pCol4, L, (float32_t *) S->pTwiddle, 4u);
557 }
558 
559 /**
560 * @addtogroup ComplexFFT
561 * @{
562 */
563 
564 /**
565 * @details
566 * @brief       Processing function for the floating-point complex FFT.
567 * @param[in]      *S    points to an instance of the floating-point CFFT structure.
568 * @param[in, out] *p1   points to the complex data buffer of size <code>2*fftLen</code>. Processing occurs in-place.
569 * @param[in]     ifftFlag       flag that selects forward (ifftFlag=0) or inverse (ifftFlag=1) transform.
570 * @param[in]     bitReverseFlag flag that enables (bitReverseFlag=1) or disables (bitReverseFlag=0) bit reversal of output.
571 * @return none.
572 */
573 
arm_cfft_f32(const arm_cfft_instance_f32 * S,float32_t * p1,uint8_t ifftFlag,uint8_t bitReverseFlag)574 void arm_cfft_f32(
575     const arm_cfft_instance_f32 * S,
576     float32_t * p1,
577     uint8_t ifftFlag,
578     uint8_t bitReverseFlag)
579 {
580     uint32_t  L = S->fftLen, l;
581     float32_t invL, * pSrc;
582 
583     if(ifftFlag == 1u)
584     {
585         /*  Conjugate input data  */
586         pSrc = p1 + 1;
587         for(l=0; l<L; l++)
588         {
589             *pSrc = -*pSrc;
590             pSrc += 2;
591         }
592     }
593 
594     switch (L)
595     {
596     case 16:
597     case 128:
598     case 1024:
599         arm_cfft_radix8by2_f32  ( (arm_cfft_instance_f32 *) S, p1);
600         break;
601     case 32:
602     case 256:
603     case 2048:
604         arm_cfft_radix8by4_f32  ( (arm_cfft_instance_f32 *) S, p1);
605         break;
606     case 64:
607     case 512:
608     case 4096:
609         arm_radix8_butterfly_f32( p1, L, (float32_t *) S->pTwiddle, 1);
610         break;
611     }
612 
613     if( bitReverseFlag )
614         arm_bitreversal_32((uint32_t*)p1,S->bitRevLength,S->pBitRevTable);
615 
616     if(ifftFlag == 1u)
617     {
618         invL = 1.0f/(float32_t)L;
619         /*  Conjugate and scale output data */
620         pSrc = p1;
621         for(l=0; l<L; l++)
622         {
623             *pSrc++ *=   invL ;
624             *pSrc  = -(*pSrc) * invL;
625             pSrc++;
626         }
627     }
628 }
629 
630 /**
631 * @} end of ComplexFFT group
632 */
633