1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_cfft_radix8_f32.c
4 * Description: Radix-8 Decimation in Frequency CFFT & CIFFT Floating point processing function
5 *
6 * $Date: 23 April 2021
7 * $Revision: V1.9.0
8 *
9 * Target Processor: Cortex-M and Cortex-A cores
10 * -------------------------------------------------------------------- */
11 /*
12 * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
13 *
14 * SPDX-License-Identifier: Apache-2.0
15 *
16 * Licensed under the Apache License, Version 2.0 (the License); you may
17 * not use this file except in compliance with the License.
18 * You may obtain a copy of the License at
19 *
20 * www.apache.org/licenses/LICENSE-2.0
21 *
22 * Unless required by applicable law or agreed to in writing, software
23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
25 * See the License for the specific language governing permissions and
26 * limitations under the License.
27 */
28
29 #include "dsp/transform_functions.h"
30
31
32 /* ----------------------------------------------------------------------
33 * Internal helper function used by the FFTs
34 * -------------------------------------------------------------------- */
35
36 void arm_radix8_butterfly_f32(
37 float32_t * pSrc,
38 uint16_t fftLen,
39 const float32_t * pCoef,
40 uint16_t twidCoefModifier);
41
42 /**
43 brief Core function for the floating-point CFFT butterfly process.
44 param[in,out] pSrc points to the in-place buffer of floating-point data type.
45 param[in] fftLen length of the FFT.
46 param[in] pCoef points to the twiddle coefficient buffer.
47 param[in] twidCoefModifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table.
48 return none
49 */
50
arm_radix8_butterfly_f32(float32_t * pSrc,uint16_t fftLen,const float32_t * pCoef,uint16_t twidCoefModifier)51 ARM_DSP_ATTRIBUTE void arm_radix8_butterfly_f32(
52 float32_t * pSrc,
53 uint16_t fftLen,
54 const float32_t * pCoef,
55 uint16_t twidCoefModifier)
56 {
57 uint32_t ia1, ia2, ia3, ia4, ia5, ia6, ia7;
58 uint32_t i1, i2, i3, i4, i5, i6, i7, i8;
59 uint32_t id;
60 uint32_t n1, n2, j;
61
62 float32_t r1, r2, r3, r4, r5, r6, r7, r8;
63 float32_t t1, t2;
64 float32_t s1, s2, s3, s4, s5, s6, s7, s8;
65 float32_t p1, p2, p3, p4;
66 float32_t co2, co3, co4, co5, co6, co7, co8;
67 float32_t si2, si3, si4, si5, si6, si7, si8;
68 const float32_t C81 = 0.70710678118f;
69
70 n2 = fftLen;
71
72 do
73 {
74 n1 = n2;
75 n2 = n2 >> 3;
76 i1 = 0;
77
78 do
79 {
80 i2 = i1 + n2;
81 i3 = i2 + n2;
82 i4 = i3 + n2;
83 i5 = i4 + n2;
84 i6 = i5 + n2;
85 i7 = i6 + n2;
86 i8 = i7 + n2;
87 r1 = pSrc[2 * i1] + pSrc[2 * i5];
88 r5 = pSrc[2 * i1] - pSrc[2 * i5];
89 r2 = pSrc[2 * i2] + pSrc[2 * i6];
90 r6 = pSrc[2 * i2] - pSrc[2 * i6];
91 r3 = pSrc[2 * i3] + pSrc[2 * i7];
92 r7 = pSrc[2 * i3] - pSrc[2 * i7];
93 r4 = pSrc[2 * i4] + pSrc[2 * i8];
94 r8 = pSrc[2 * i4] - pSrc[2 * i8];
95 t1 = r1 - r3;
96 r1 = r1 + r3;
97 r3 = r2 - r4;
98 r2 = r2 + r4;
99 pSrc[2 * i1] = r1 + r2;
100 pSrc[2 * i5] = r1 - r2;
101 r1 = pSrc[2 * i1 + 1] + pSrc[2 * i5 + 1];
102 s5 = pSrc[2 * i1 + 1] - pSrc[2 * i5 + 1];
103 r2 = pSrc[2 * i2 + 1] + pSrc[2 * i6 + 1];
104 s6 = pSrc[2 * i2 + 1] - pSrc[2 * i6 + 1];
105 s3 = pSrc[2 * i3 + 1] + pSrc[2 * i7 + 1];
106 s7 = pSrc[2 * i3 + 1] - pSrc[2 * i7 + 1];
107 r4 = pSrc[2 * i4 + 1] + pSrc[2 * i8 + 1];
108 s8 = pSrc[2 * i4 + 1] - pSrc[2 * i8 + 1];
109 t2 = r1 - s3;
110 r1 = r1 + s3;
111 s3 = r2 - r4;
112 r2 = r2 + r4;
113 pSrc[2 * i1 + 1] = r1 + r2;
114 pSrc[2 * i5 + 1] = r1 - r2;
115 pSrc[2 * i3] = t1 + s3;
116 pSrc[2 * i7] = t1 - s3;
117 pSrc[2 * i3 + 1] = t2 - r3;
118 pSrc[2 * i7 + 1] = t2 + r3;
119 r1 = (r6 - r8) * C81;
120 r6 = (r6 + r8) * C81;
121 r2 = (s6 - s8) * C81;
122 s6 = (s6 + s8) * C81;
123 t1 = r5 - r1;
124 r5 = r5 + r1;
125 r8 = r7 - r6;
126 r7 = r7 + r6;
127 t2 = s5 - r2;
128 s5 = s5 + r2;
129 s8 = s7 - s6;
130 s7 = s7 + s6;
131 pSrc[2 * i2] = r5 + s7;
132 pSrc[2 * i8] = r5 - s7;
133 pSrc[2 * i6] = t1 + s8;
134 pSrc[2 * i4] = t1 - s8;
135 pSrc[2 * i2 + 1] = s5 - r7;
136 pSrc[2 * i8 + 1] = s5 + r7;
137 pSrc[2 * i6 + 1] = t2 - r8;
138 pSrc[2 * i4 + 1] = t2 + r8;
139
140 i1 += n1;
141 } while (i1 < fftLen);
142
143 if (n2 < 8)
144 break;
145
146 ia1 = 0;
147 j = 1;
148
149 do
150 {
151 /* index calculation for the coefficients */
152 id = ia1 + twidCoefModifier;
153 ia1 = id;
154 ia2 = ia1 + id;
155 ia3 = ia2 + id;
156 ia4 = ia3 + id;
157 ia5 = ia4 + id;
158 ia6 = ia5 + id;
159 ia7 = ia6 + id;
160
161 co2 = pCoef[2 * ia1];
162 co3 = pCoef[2 * ia2];
163 co4 = pCoef[2 * ia3];
164 co5 = pCoef[2 * ia4];
165 co6 = pCoef[2 * ia5];
166 co7 = pCoef[2 * ia6];
167 co8 = pCoef[2 * ia7];
168 si2 = pCoef[2 * ia1 + 1];
169 si3 = pCoef[2 * ia2 + 1];
170 si4 = pCoef[2 * ia3 + 1];
171 si5 = pCoef[2 * ia4 + 1];
172 si6 = pCoef[2 * ia5 + 1];
173 si7 = pCoef[2 * ia6 + 1];
174 si8 = pCoef[2 * ia7 + 1];
175
176 i1 = j;
177
178 do
179 {
180 /* index calculation for the input */
181 i2 = i1 + n2;
182 i3 = i2 + n2;
183 i4 = i3 + n2;
184 i5 = i4 + n2;
185 i6 = i5 + n2;
186 i7 = i6 + n2;
187 i8 = i7 + n2;
188 r1 = pSrc[2 * i1] + pSrc[2 * i5];
189 r5 = pSrc[2 * i1] - pSrc[2 * i5];
190 r2 = pSrc[2 * i2] + pSrc[2 * i6];
191 r6 = pSrc[2 * i2] - pSrc[2 * i6];
192 r3 = pSrc[2 * i3] + pSrc[2 * i7];
193 r7 = pSrc[2 * i3] - pSrc[2 * i7];
194 r4 = pSrc[2 * i4] + pSrc[2 * i8];
195 r8 = pSrc[2 * i4] - pSrc[2 * i8];
196 t1 = r1 - r3;
197 r1 = r1 + r3;
198 r3 = r2 - r4;
199 r2 = r2 + r4;
200 pSrc[2 * i1] = r1 + r2;
201 r2 = r1 - r2;
202 s1 = pSrc[2 * i1 + 1] + pSrc[2 * i5 + 1];
203 s5 = pSrc[2 * i1 + 1] - pSrc[2 * i5 + 1];
204 s2 = pSrc[2 * i2 + 1] + pSrc[2 * i6 + 1];
205 s6 = pSrc[2 * i2 + 1] - pSrc[2 * i6 + 1];
206 s3 = pSrc[2 * i3 + 1] + pSrc[2 * i7 + 1];
207 s7 = pSrc[2 * i3 + 1] - pSrc[2 * i7 + 1];
208 s4 = pSrc[2 * i4 + 1] + pSrc[2 * i8 + 1];
209 s8 = pSrc[2 * i4 + 1] - pSrc[2 * i8 + 1];
210 t2 = s1 - s3;
211 s1 = s1 + s3;
212 s3 = s2 - s4;
213 s2 = s2 + s4;
214 r1 = t1 + s3;
215 t1 = t1 - s3;
216 pSrc[2 * i1 + 1] = s1 + s2;
217 s2 = s1 - s2;
218 s1 = t2 - r3;
219 t2 = t2 + r3;
220 p1 = co5 * r2;
221 p2 = si5 * s2;
222 p3 = co5 * s2;
223 p4 = si5 * r2;
224 pSrc[2 * i5] = p1 + p2;
225 pSrc[2 * i5 + 1] = p3 - p4;
226 p1 = co3 * r1;
227 p2 = si3 * s1;
228 p3 = co3 * s1;
229 p4 = si3 * r1;
230 pSrc[2 * i3] = p1 + p2;
231 pSrc[2 * i3 + 1] = p3 - p4;
232 p1 = co7 * t1;
233 p2 = si7 * t2;
234 p3 = co7 * t2;
235 p4 = si7 * t1;
236 pSrc[2 * i7] = p1 + p2;
237 pSrc[2 * i7 + 1] = p3 - p4;
238 r1 = (r6 - r8) * C81;
239 r6 = (r6 + r8) * C81;
240 s1 = (s6 - s8) * C81;
241 s6 = (s6 + s8) * C81;
242 t1 = r5 - r1;
243 r5 = r5 + r1;
244 r8 = r7 - r6;
245 r7 = r7 + r6;
246 t2 = s5 - s1;
247 s5 = s5 + s1;
248 s8 = s7 - s6;
249 s7 = s7 + s6;
250 r1 = r5 + s7;
251 r5 = r5 - s7;
252 r6 = t1 + s8;
253 t1 = t1 - s8;
254 s1 = s5 - r7;
255 s5 = s5 + r7;
256 s6 = t2 - r8;
257 t2 = t2 + r8;
258 p1 = co2 * r1;
259 p2 = si2 * s1;
260 p3 = co2 * s1;
261 p4 = si2 * r1;
262 pSrc[2 * i2] = p1 + p2;
263 pSrc[2 * i2 + 1] = p3 - p4;
264 p1 = co8 * r5;
265 p2 = si8 * s5;
266 p3 = co8 * s5;
267 p4 = si8 * r5;
268 pSrc[2 * i8] = p1 + p2;
269 pSrc[2 * i8 + 1] = p3 - p4;
270 p1 = co6 * r6;
271 p2 = si6 * s6;
272 p3 = co6 * s6;
273 p4 = si6 * r6;
274 pSrc[2 * i6] = p1 + p2;
275 pSrc[2 * i6 + 1] = p3 - p4;
276 p1 = co4 * t1;
277 p2 = si4 * t2;
278 p3 = co4 * t2;
279 p4 = si4 * t1;
280 pSrc[2 * i4] = p1 + p2;
281 pSrc[2 * i4 + 1] = p3 - p4;
282
283 i1 += n1;
284 } while (i1 < fftLen);
285
286 j++;
287 } while (j < n2);
288
289 twidCoefModifier <<= 3;
290 } while (n2 > 7);
291 }
292
