/src/aac/libFDK/src/fft_rad2.cpp
Line | Count | Source |
1 | | /* ----------------------------------------------------------------------------- |
2 | | Software License for The Fraunhofer FDK AAC Codec Library for Android |
3 | | |
4 | | © Copyright 1995 - 2018 Fraunhofer-Gesellschaft zur Förderung der angewandten |
5 | | Forschung e.V. All rights reserved. |
6 | | |
7 | | 1. INTRODUCTION |
8 | | The Fraunhofer FDK AAC Codec Library for Android ("FDK AAC Codec") is software |
9 | | that implements the MPEG Advanced Audio Coding ("AAC") encoding and decoding |
10 | | scheme for digital audio. This FDK AAC Codec software is intended to be used on |
11 | | a wide variety of Android devices. |
12 | | |
13 | | AAC's HE-AAC and HE-AAC v2 versions are regarded as today's most efficient |
14 | | general perceptual audio codecs. AAC-ELD is considered the best-performing |
15 | | full-bandwidth communications codec by independent studies and is widely |
16 | | deployed. AAC has been standardized by ISO and IEC as part of the MPEG |
17 | | specifications. |
18 | | |
19 | | Patent licenses for necessary patent claims for the FDK AAC Codec (including |
20 | | those of Fraunhofer) may be obtained through Via Licensing |
21 | | (www.vialicensing.com) or through the respective patent owners individually for |
22 | | the purpose of encoding or decoding bit streams in products that are compliant |
23 | | with the ISO/IEC MPEG audio standards. Please note that most manufacturers of |
24 | | Android devices already license these patent claims through Via Licensing or |
25 | | directly from the patent owners, and therefore FDK AAC Codec software may |
26 | | already be covered under those patent licenses when it is used for those |
27 | | licensed purposes only. |
28 | | |
29 | | Commercially-licensed AAC software libraries, including floating-point versions |
30 | | with enhanced sound quality, are also available from Fraunhofer. Users are |
31 | | encouraged to check the Fraunhofer website for additional applications |
32 | | information and documentation. |
33 | | |
34 | | 2. COPYRIGHT LICENSE |
35 | | |
36 | | Redistribution and use in source and binary forms, with or without modification, |
37 | | are permitted without payment of copyright license fees provided that you |
38 | | satisfy the following conditions: |
39 | | |
40 | | You must retain the complete text of this software license in redistributions of |
41 | | the FDK AAC Codec or your modifications thereto in source code form. |
42 | | |
43 | | You must retain the complete text of this software license in the documentation |
44 | | and/or other materials provided with redistributions of the FDK AAC Codec or |
45 | | your modifications thereto in binary form. You must make available free of |
46 | | charge copies of the complete source code of the FDK AAC Codec and your |
47 | | modifications thereto to recipients of copies in binary form. |
48 | | |
49 | | The name of Fraunhofer may not be used to endorse or promote products derived |
50 | | from this library without prior written permission. |
51 | | |
52 | | You may not charge copyright license fees for anyone to use, copy or distribute |
53 | | the FDK AAC Codec software or your modifications thereto. |
54 | | |
55 | | Your modified versions of the FDK AAC Codec must carry prominent notices stating |
56 | | that you changed the software and the date of any change. For modified versions |
57 | | of the FDK AAC Codec, the term "Fraunhofer FDK AAC Codec Library for Android" |
58 | | must be replaced by the term "Third-Party Modified Version of the Fraunhofer FDK |
59 | | AAC Codec Library for Android." |
60 | | |
61 | | 3. NO PATENT LICENSE |
62 | | |
63 | | NO EXPRESS OR IMPLIED LICENSES TO ANY PATENT CLAIMS, including without |
64 | | limitation the patents of Fraunhofer, ARE GRANTED BY THIS SOFTWARE LICENSE. |
65 | | Fraunhofer provides no warranty of patent non-infringement with respect to this |
66 | | software. |
67 | | |
68 | | You may use this FDK AAC Codec software or modifications thereto only for |
69 | | purposes that are authorized by appropriate patent licenses. |
70 | | |
71 | | 4. DISCLAIMER |
72 | | |
73 | | This FDK AAC Codec software is provided by Fraunhofer on behalf of the copyright |
74 | | holders and contributors "AS IS" and WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES, |
75 | | including but not limited to the implied warranties of merchantability and |
76 | | fitness for a particular purpose. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR |
77 | | CONTRIBUTORS BE LIABLE for any direct, indirect, incidental, special, exemplary, |
78 | | or consequential damages, including but not limited to procurement of substitute |
79 | | goods or services; loss of use, data, or profits, or business interruption, |
80 | | however caused and on any theory of liability, whether in contract, strict |
81 | | liability, or tort (including negligence), arising in any way out of the use of |
82 | | this software, even if advised of the possibility of such damage. |
83 | | |
84 | | 5. CONTACT INFORMATION |
85 | | |
86 | | Fraunhofer Institute for Integrated Circuits IIS |
87 | | Attention: Audio and Multimedia Departments - FDK AAC LL |
88 | | Am Wolfsmantel 33 |
89 | | 91058 Erlangen, Germany |
90 | | |
91 | | www.iis.fraunhofer.de/amm |
92 | | amm-info@iis.fraunhofer.de |
93 | | ----------------------------------------------------------------------------- */ |
94 | | |
95 | | /******************* Library for basic calculation routines ******************** |
96 | | |
97 | | Author(s): M. Lohwasser, M. Gayer |
98 | | |
99 | | Description: |
100 | | |
101 | | *******************************************************************************/ |
102 | | |
103 | | #include "fft_rad2.h" |
104 | | |
105 | | #include "scramble.h" |
106 | | |
107 | | #define __FFT_RAD2_CPP__ |
108 | | |
109 | | #if defined(__arm__) |
110 | | #include "arm/fft_rad2_arm.cpp" |
111 | | |
112 | | #elif defined(__GNUC__) && defined(__mips__) && defined(__mips_dsp) |
113 | | #include "mips/fft_rad2_mips.cpp" |
114 | | |
115 | | #endif |
116 | | |
117 | | /***************************************************************************** |
118 | | |
119 | | functionname: dit_fft (analysis) |
120 | | description: dit-tukey-algorithm |
121 | | scrambles data at entry |
122 | | i.e. loop is made with scrambled data |
123 | | returns: |
124 | | input: |
125 | | output: |
126 | | |
127 | | *****************************************************************************/ |
128 | | |
129 | | #ifndef FUNCTION_dit_fft |
130 | | |
131 | | void dit_fft(FIXP_DBL *x, const INT ldn, const FIXP_STP *trigdata, |
132 | 752k | const INT trigDataSize) { |
133 | 752k | const INT n = 1 << ldn; |
134 | 752k | INT trigstep, i, ldm; |
135 | | |
136 | 752k | C_ALLOC_ALIGNED_CHECK(x); |
137 | | |
138 | 752k | scramble(x, n); |
139 | | /* |
140 | | * 1+2 stage radix 4 |
141 | | */ |
142 | | |
143 | 37.2M | for (i = 0; i < n * 2; i += 8) { |
144 | 36.4M | FIXP_DBL a00, a10, a20, a30; |
145 | 36.4M | a00 = (x[i + 0] + x[i + 2]) >> 1; /* Re A + Re B */ |
146 | 36.4M | a10 = (x[i + 4] + x[i + 6]) >> 1; /* Re C + Re D */ |
147 | 36.4M | a20 = (x[i + 1] + x[i + 3]) >> 1; /* Im A + Im B */ |
148 | 36.4M | a30 = (x[i + 5] + x[i + 7]) >> 1; /* Im C + Im D */ |
149 | | |
150 | 36.4M | x[i + 0] = a00 + a10; /* Re A' = Re A + Re B + Re C + Re D */ |
151 | 36.4M | x[i + 4] = a00 - a10; /* Re C' = Re A + Re B - Re C - Re D */ |
152 | 36.4M | x[i + 1] = a20 + a30; /* Im A' = Im A + Im B + Im C + Im D */ |
153 | 36.4M | x[i + 5] = a20 - a30; /* Im C' = Im A + Im B - Im C - Im D */ |
154 | | |
155 | 36.4M | a00 = a00 - x[i + 2]; /* Re A - Re B */ |
156 | 36.4M | a10 = a10 - x[i + 6]; /* Re C - Re D */ |
157 | 36.4M | a20 = a20 - x[i + 3]; /* Im A - Im B */ |
158 | 36.4M | a30 = a30 - x[i + 7]; /* Im C - Im D */ |
159 | | |
160 | 36.4M | x[i + 2] = a00 + a30; /* Re B' = Re A - Re B + Im C - Im D */ |
161 | 36.4M | x[i + 6] = a00 - a30; /* Re D' = Re A - Re B - Im C + Im D */ |
162 | 36.4M | x[i + 3] = a20 - a10; /* Im B' = Im A - Im B - Re C + Re D */ |
163 | 36.4M | x[i + 7] = a20 + a10; /* Im D' = Im A - Im B + Re C - Re D */ |
164 | 36.4M | } |
165 | | |
166 | 4.51M | for (ldm = 3; ldm <= ldn; ++ldm) { |
167 | 3.76M | INT m = (1 << ldm); |
168 | 3.76M | INT mh = (m >> 1); |
169 | 3.76M | INT j, r; |
170 | | |
171 | 3.76M | trigstep = ((trigDataSize << 2) >> ldm); |
172 | | |
173 | 3.76M | FDK_ASSERT(trigstep > 0); |
174 | | |
175 | | /* Do first iteration with c=1.0 and s=0.0 separately to avoid loosing to |
176 | | much precision. Beware: The impact on the overal FFT precision is rather |
177 | | large. */ |
178 | 3.76M | { /* block 1 */ |
179 | | |
180 | 3.76M | j = 0; |
181 | | |
182 | 39.4M | for (r = 0; r < n; r += m) { |
183 | 35.7M | INT t1 = (r + j) << 1; |
184 | 35.7M | INT t2 = t1 + (mh << 1); |
185 | 35.7M | FIXP_DBL vr, vi, ur, ui; |
186 | | |
187 | | // cplxMultDiv2(&vi, &vr, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); |
188 | 35.7M | vi = x[t2 + 1] >> 1; |
189 | 35.7M | vr = x[t2] >> 1; |
190 | | |
191 | 35.7M | ur = x[t1] >> 1; |
192 | 35.7M | ui = x[t1 + 1] >> 1; |
193 | | |
194 | 35.7M | x[t1] = ur + vr; |
195 | 35.7M | x[t1 + 1] = ui + vi; |
196 | | |
197 | 35.7M | x[t2] = ur - vr; |
198 | 35.7M | x[t2 + 1] = ui - vi; |
199 | | |
200 | 35.7M | t1 += mh; |
201 | 35.7M | t2 = t1 + (mh << 1); |
202 | | |
203 | | // cplxMultDiv2(&vr, &vi, x[t2+1], x[t2], (FIXP_SGL)1.0, (FIXP_SGL)0.0); |
204 | 35.7M | vr = x[t2 + 1] >> 1; |
205 | 35.7M | vi = x[t2] >> 1; |
206 | | |
207 | 35.7M | ur = x[t1] >> 1; |
208 | 35.7M | ui = x[t1 + 1] >> 1; |
209 | | |
210 | 35.7M | x[t1] = ur + vr; |
211 | 35.7M | x[t1 + 1] = ui - vi; |
212 | | |
213 | 35.7M | x[t2] = ur - vr; |
214 | 35.7M | x[t2 + 1] = ui + vi; |
215 | 35.7M | } |
216 | | |
217 | 3.76M | } /* end of block 1 */ |
218 | | |
219 | 35.7M | for (j = 1; j < mh / 4; ++j) { |
220 | 31.9M | FIXP_STP cs; |
221 | | |
222 | 31.9M | cs = trigdata[j * trigstep]; |
223 | | |
224 | 108M | for (r = 0; r < n; r += m) { |
225 | 76.8M | INT t1 = (r + j) << 1; |
226 | 76.8M | INT t2 = t1 + (mh << 1); |
227 | 76.8M | FIXP_DBL vr, vi, ur, ui; |
228 | | |
229 | 76.8M | cplxMultDiv2(&vi, &vr, x[t2 + 1], x[t2], cs); |
230 | | |
231 | 76.8M | ur = x[t1] >> 1; |
232 | 76.8M | ui = x[t1 + 1] >> 1; |
233 | | |
234 | 76.8M | x[t1] = ur + vr; |
235 | 76.8M | x[t1 + 1] = ui + vi; |
236 | | |
237 | 76.8M | x[t2] = ur - vr; |
238 | 76.8M | x[t2 + 1] = ui - vi; |
239 | | |
240 | 76.8M | t1 += mh; |
241 | 76.8M | t2 = t1 + (mh << 1); |
242 | | |
243 | 76.8M | cplxMultDiv2(&vr, &vi, x[t2 + 1], x[t2], cs); |
244 | | |
245 | 76.8M | ur = x[t1] >> 1; |
246 | 76.8M | ui = x[t1 + 1] >> 1; |
247 | | |
248 | 76.8M | x[t1] = ur + vr; |
249 | 76.8M | x[t1 + 1] = ui - vi; |
250 | | |
251 | 76.8M | x[t2] = ur - vr; |
252 | 76.8M | x[t2 + 1] = ui + vi; |
253 | | |
254 | | /* Same as above but for t1,t2 with j>mh/4 and thus cs swapped */ |
255 | 76.8M | t1 = (r + mh / 2 - j) << 1; |
256 | 76.8M | t2 = t1 + (mh << 1); |
257 | | |
258 | 76.8M | cplxMultDiv2(&vi, &vr, x[t2], x[t2 + 1], cs); |
259 | | |
260 | 76.8M | ur = x[t1] >> 1; |
261 | 76.8M | ui = x[t1 + 1] >> 1; |
262 | | |
263 | 76.8M | x[t1] = ur + vr; |
264 | 76.8M | x[t1 + 1] = ui - vi; |
265 | | |
266 | 76.8M | x[t2] = ur - vr; |
267 | 76.8M | x[t2 + 1] = ui + vi; |
268 | | |
269 | 76.8M | t1 += mh; |
270 | 76.8M | t2 = t1 + (mh << 1); |
271 | | |
272 | 76.8M | cplxMultDiv2(&vr, &vi, x[t2], x[t2 + 1], cs); |
273 | | |
274 | 76.8M | ur = x[t1] >> 1; |
275 | 76.8M | ui = x[t1 + 1] >> 1; |
276 | | |
277 | 76.8M | x[t1] = ur - vr; |
278 | 76.8M | x[t1 + 1] = ui - vi; |
279 | | |
280 | 76.8M | x[t2] = ur + vr; |
281 | 76.8M | x[t2 + 1] = ui + vi; |
282 | 76.8M | } |
283 | 31.9M | } |
284 | | |
285 | 3.76M | { /* block 2 */ |
286 | 3.76M | j = mh / 4; |
287 | | |
288 | 39.4M | for (r = 0; r < n; r += m) { |
289 | 35.7M | INT t1 = (r + j) << 1; |
290 | 35.7M | INT t2 = t1 + (mh << 1); |
291 | 35.7M | FIXP_DBL vr, vi, ur, ui; |
292 | | |
293 | 35.7M | cplxMultDiv2(&vi, &vr, x[t2 + 1], x[t2], STC(0x5a82799a), |
294 | 35.7M | STC(0x5a82799a)); |
295 | | |
296 | 35.7M | ur = x[t1] >> 1; |
297 | 35.7M | ui = x[t1 + 1] >> 1; |
298 | | |
299 | 35.7M | x[t1] = ur + vr; |
300 | 35.7M | x[t1 + 1] = ui + vi; |
301 | | |
302 | 35.7M | x[t2] = ur - vr; |
303 | 35.7M | x[t2 + 1] = ui - vi; |
304 | | |
305 | 35.7M | t1 += mh; |
306 | 35.7M | t2 = t1 + (mh << 1); |
307 | | |
308 | 35.7M | cplxMultDiv2(&vr, &vi, x[t2 + 1], x[t2], STC(0x5a82799a), |
309 | 35.7M | STC(0x5a82799a)); |
310 | | |
311 | 35.7M | ur = x[t1] >> 1; |
312 | 35.7M | ui = x[t1 + 1] >> 1; |
313 | | |
314 | 35.7M | x[t1] = ur + vr; |
315 | 35.7M | x[t1 + 1] = ui - vi; |
316 | | |
317 | 35.7M | x[t2] = ur - vr; |
318 | 35.7M | x[t2 + 1] = ui + vi; |
319 | 35.7M | } |
320 | 3.76M | } /* end of block 2 */ |
321 | 3.76M | } |
322 | 752k | } |
323 | | |
324 | | #endif |