?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge commit 'svn/origin' into devel
[cinelerra_cv/ct.git] / libmpeg3 / audio / dct.c
blobbf2be7ef79eaebafe76c3dafcb8b923fef87c57d
1 /*
2 *
3 * This file is part of libmpeg3
4 *
5 * libmpeg3 is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2, or (at your option)
8 * any later version.
9 *
10 * libmpeg3 is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with GNU Make; see the file COPYING. If not, write to
17 * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
18 *
19 */
20
21 /*
22 * Discrete Cosine Tansform (DCT) for subband synthesis
23 * optimized for machines with no auto-increment.
24 * The performance is highly compiler dependend. Maybe
25 * the dct64.c version for 'normal' processor may be faster
26 * even for Intel processors.
27 */
28
29 #include "mpeg3private.h"
30 #include "mpeg3protos.h"
31 #include "tables.h"
32
33 #include <math.h>
34
35 int mpeg3audio_dct64_1(float *out0, float *out1, float *b1, float *b2, float *samples)
36 {
37 register float *costab = mpeg3_pnts[0];
38
39 b1[0x00] = samples[0x00] + samples[0x1F];
40 b1[0x01] = samples[0x01] + samples[0x1E];
41 b1[0x1F] = (samples[0x00] - samples[0x1F]) * costab[0x0];
42 b1[0x1E] = (samples[0x01] - samples[0x1E]) * costab[0x1];
43
44 b1[0x02] = samples[0x02] + samples[0x1D];
45 b1[0x03] = samples[0x03] + samples[0x1C];
46 b1[0x1D] = (samples[0x02] - samples[0x1D]) * costab[0x2];
47 b1[0x1C] = (samples[0x03] - samples[0x1C]) * costab[0x3];
48
49 b1[0x04] = samples[0x04] + samples[0x1B];
50 b1[0x05] = samples[0x05] + samples[0x1A];
51 b1[0x1B] = (samples[0x04] - samples[0x1B]) * costab[0x4];
52 b1[0x1A] = (samples[0x05] - samples[0x1A]) * costab[0x5];
53
54 b1[0x06] = samples[0x06] + samples[0x19];
55 b1[0x07] = samples[0x07] + samples[0x18];
56 b1[0x19] = (samples[0x06] - samples[0x19]) * costab[0x6];
57 b1[0x18] = (samples[0x07] - samples[0x18]) * costab[0x7];
58
59 b1[0x08] = samples[0x08] + samples[0x17];
60 b1[0x09] = samples[0x09] + samples[0x16];
61 b1[0x17] = (samples[0x08] - samples[0x17]) * costab[0x8];
62 b1[0x16] = (samples[0x09] - samples[0x16]) * costab[0x9];
63
64 b1[0x0A] = samples[0x0A] + samples[0x15];
65 b1[0x0B] = samples[0x0B] + samples[0x14];
66 b1[0x15] = (samples[0x0A] - samples[0x15]) * costab[0xA];
67 b1[0x14] = (samples[0x0B] - samples[0x14]) * costab[0xB];
68
69 b1[0x0C] = samples[0x0C] + samples[0x13];
70 b1[0x0D] = samples[0x0D] + samples[0x12];
71 b1[0x13] = (samples[0x0C] - samples[0x13]) * costab[0xC];
72 b1[0x12] = (samples[0x0D] - samples[0x12]) * costab[0xD];
73
74 b1[0x0E] = samples[0x0E] + samples[0x11];
75 b1[0x0F] = samples[0x0F] + samples[0x10];
76 b1[0x11] = (samples[0x0E] - samples[0x11]) * costab[0xE];
77 b1[0x10] = (samples[0x0F] - samples[0x10]) * costab[0xF];
78
79 costab = mpeg3_pnts[1];
80
81 b2[0x00] = b1[0x00] + b1[0x0F];
82 b2[0x01] = b1[0x01] + b1[0x0E];
83 b2[0x0F] = (b1[0x00] - b1[0x0F]) * costab[0];
84 b2[0x0E] = (b1[0x01] - b1[0x0E]) * costab[1];
85
86 b2[0x02] = b1[0x02] + b1[0x0D];
87 b2[0x03] = b1[0x03] + b1[0x0C];
88 b2[0x0D] = (b1[0x02] - b1[0x0D]) * costab[2];
89 b2[0x0C] = (b1[0x03] - b1[0x0C]) * costab[3];
90
91 b2[0x04] = b1[0x04] + b1[0x0B];
92 b2[0x05] = b1[0x05] + b1[0x0A];
93 b2[0x0B] = (b1[0x04] - b1[0x0B]) * costab[4];
94 b2[0x0A] = (b1[0x05] - b1[0x0A]) * costab[5];
95
96 b2[0x06] = b1[0x06] + b1[0x09];
97 b2[0x07] = b1[0x07] + b1[0x08];
98 b2[0x09] = (b1[0x06] - b1[0x09]) * costab[6];
99 b2[0x08] = (b1[0x07] - b1[0x08]) * costab[7];
100
101 /* */
102
103 b2[0x10] = b1[0x10] + b1[0x1F];
104 b2[0x11] = b1[0x11] + b1[0x1E];
105 b2[0x1F] = (b1[0x1F] - b1[0x10]) * costab[0];
106 b2[0x1E] = (b1[0x1E] - b1[0x11]) * costab[1];
107
108 b2[0x12] = b1[0x12] + b1[0x1D];
109 b2[0x13] = b1[0x13] + b1[0x1C];
110 b2[0x1D] = (b1[0x1D] - b1[0x12]) * costab[2];
111 b2[0x1C] = (b1[0x1C] - b1[0x13]) * costab[3];
112
113 b2[0x14] = b1[0x14] + b1[0x1B];
114 b2[0x15] = b1[0x15] + b1[0x1A];
115 b2[0x1B] = (b1[0x1B] - b1[0x14]) * costab[4];
116 b2[0x1A] = (b1[0x1A] - b1[0x15]) * costab[5];
117
118 b2[0x16] = b1[0x16] + b1[0x19];
119 b2[0x17] = b1[0x17] + b1[0x18];
120 b2[0x19] = (b1[0x19] - b1[0x16]) * costab[6];
121 b2[0x18] = (b1[0x18] - b1[0x17]) * costab[7];
122
123 costab = mpeg3_pnts[2];
124
125 b1[0x00] = b2[0x00] + b2[0x07];
126 b1[0x07] = (b2[0x00] - b2[0x07]) * costab[0];
127 b1[0x01] = b2[0x01] + b2[0x06];
128 b1[0x06] = (b2[0x01] - b2[0x06]) * costab[1];
129 b1[0x02] = b2[0x02] + b2[0x05];
130 b1[0x05] = (b2[0x02] - b2[0x05]) * costab[2];
131 b1[0x03] = b2[0x03] + b2[0x04];
132 b1[0x04] = (b2[0x03] - b2[0x04]) * costab[3];
133
134 b1[0x08] = b2[0x08] + b2[0x0F];
135 b1[0x0F] = (b2[0x0F] - b2[0x08]) * costab[0];
136 b1[0x09] = b2[0x09] + b2[0x0E];
137 b1[0x0E] = (b2[0x0E] - b2[0x09]) * costab[1];
138 b1[0x0A] = b2[0x0A] + b2[0x0D];
139 b1[0x0D] = (b2[0x0D] - b2[0x0A]) * costab[2];
140 b1[0x0B] = b2[0x0B] + b2[0x0C];
141 b1[0x0C] = (b2[0x0C] - b2[0x0B]) * costab[3];
142
143 b1[0x10] = b2[0x10] + b2[0x17];
144 b1[0x17] = (b2[0x10] - b2[0x17]) * costab[0];
145 b1[0x11] = b2[0x11] + b2[0x16];
146 b1[0x16] = (b2[0x11] - b2[0x16]) * costab[1];
147 b1[0x12] = b2[0x12] + b2[0x15];
148 b1[0x15] = (b2[0x12] - b2[0x15]) * costab[2];
149 b1[0x13] = b2[0x13] + b2[0x14];
150 b1[0x14] = (b2[0x13] - b2[0x14]) * costab[3];
151
152 b1[0x18] = b2[0x18] + b2[0x1F];
153 b1[0x1F] = (b2[0x1F] - b2[0x18]) * costab[0];
154 b1[0x19] = b2[0x19] + b2[0x1E];
155 b1[0x1E] = (b2[0x1E] - b2[0x19]) * costab[1];
156 b1[0x1A] = b2[0x1A] + b2[0x1D];
157 b1[0x1D] = (b2[0x1D] - b2[0x1A]) * costab[2];
158 b1[0x1B] = b2[0x1B] + b2[0x1C];
159 b1[0x1C] = (b2[0x1C] - b2[0x1B]) * costab[3];
160
161 {
162 register float const cos0 = mpeg3_pnts[3][0];
163 register float const cos1 = mpeg3_pnts[3][1];
164
165 b2[0x00] = b1[0x00] + b1[0x03];
166 b2[0x03] = (b1[0x00] - b1[0x03]) * cos0;
167 b2[0x01] = b1[0x01] + b1[0x02];
168 b2[0x02] = (b1[0x01] - b1[0x02]) * cos1;
169
170 b2[0x04] = b1[0x04] + b1[0x07];
171 b2[0x07] = (b1[0x07] - b1[0x04]) * cos0;
172 b2[0x05] = b1[0x05] + b1[0x06];
173 b2[0x06] = (b1[0x06] - b1[0x05]) * cos1;
174
175 b2[0x08] = b1[0x08] + b1[0x0B];
176 b2[0x0B] = (b1[0x08] - b1[0x0B]) * cos0;
177 b2[0x09] = b1[0x09] + b1[0x0A];
178 b2[0x0A] = (b1[0x09] - b1[0x0A]) * cos1;
179
180 b2[0x0C] = b1[0x0C] + b1[0x0F];
181 b2[0x0F] = (b1[0x0F] - b1[0x0C]) * cos0;
182 b2[0x0D] = b1[0x0D] + b1[0x0E];
183 b2[0x0E] = (b1[0x0E] - b1[0x0D]) * cos1;
184
185 b2[0x10] = b1[0x10] + b1[0x13];
186 b2[0x13] = (b1[0x10] - b1[0x13]) * cos0;
187 b2[0x11] = b1[0x11] + b1[0x12];
188 b2[0x12] = (b1[0x11] - b1[0x12]) * cos1;
189
190 b2[0x14] = b1[0x14] + b1[0x17];
191 b2[0x17] = (b1[0x17] - b1[0x14]) * cos0;
192 b2[0x15] = b1[0x15] + b1[0x16];
193 b2[0x16] = (b1[0x16] - b1[0x15]) * cos1;
194
195 b2[0x18] = b1[0x18] + b1[0x1B];
196 b2[0x1B] = (b1[0x18] - b1[0x1B]) * cos0;
197 b2[0x19] = b1[0x19] + b1[0x1A];
198 b2[0x1A] = (b1[0x19] - b1[0x1A]) * cos1;
199
200 b2[0x1C] = b1[0x1C] + b1[0x1F];
201 b2[0x1F] = (b1[0x1F] - b1[0x1C]) * cos0;
202 b2[0x1D] = b1[0x1D] + b1[0x1E];
203 b2[0x1E] = (b1[0x1E] - b1[0x1D]) * cos1;
204 }
205
206 {
207 register float const cos0 = mpeg3_pnts[4][0];
208
209 b1[0x00] = b2[0x00] + b2[0x01];
210 b1[0x01] = (b2[0x00] - b2[0x01]) * cos0;
211 b1[0x02] = b2[0x02] + b2[0x03];
212 b1[0x03] = (b2[0x03] - b2[0x02]) * cos0;
213 b1[0x02] += b1[0x03];
214
215 b1[0x04] = b2[0x04] + b2[0x05];
216 b1[0x05] = (b2[0x04] - b2[0x05]) * cos0;
217 b1[0x06] = b2[0x06] + b2[0x07];
218 b1[0x07] = (b2[0x07] - b2[0x06]) * cos0;
219 b1[0x06] += b1[0x07];
220 b1[0x04] += b1[0x06];
221 b1[0x06] += b1[0x05];
222 b1[0x05] += b1[0x07];
223
224 b1[0x08] = b2[0x08] + b2[0x09];
225 b1[0x09] = (b2[0x08] - b2[0x09]) * cos0;
226 b1[0x0A] = b2[0x0A] + b2[0x0B];
227 b1[0x0B] = (b2[0x0B] - b2[0x0A]) * cos0;
228 b1[0x0A] += b1[0x0B];
229
230 b1[0x0C] = b2[0x0C] + b2[0x0D];
231 b1[0x0D] = (b2[0x0C] - b2[0x0D]) * cos0;
232 b1[0x0E] = b2[0x0E] + b2[0x0F];
233 b1[0x0F] = (b2[0x0F] - b2[0x0E]) * cos0;
234 b1[0x0E] += b1[0x0F];
235 b1[0x0C] += b1[0x0E];
236 b1[0x0E] += b1[0x0D];
237 b1[0x0D] += b1[0x0F];
238
239 b1[0x10] = b2[0x10] + b2[0x11];
240 b1[0x11] = (b2[0x10] - b2[0x11]) * cos0;
241 b1[0x12] = b2[0x12] + b2[0x13];
242 b1[0x13] = (b2[0x13] - b2[0x12]) * cos0;
243 b1[0x12] += b1[0x13];
244
245 b1[0x14] = b2[0x14] + b2[0x15];
246 b1[0x15] = (b2[0x14] - b2[0x15]) * cos0;
247 b1[0x16] = b2[0x16] + b2[0x17];
248 b1[0x17] = (b2[0x17] - b2[0x16]) * cos0;
249 b1[0x16] += b1[0x17];
250 b1[0x14] += b1[0x16];
251 b1[0x16] += b1[0x15];
252 b1[0x15] += b1[0x17];
253
254 b1[0x18] = b2[0x18] + b2[0x19];
255 b1[0x19] = (b2[0x18] - b2[0x19]) * cos0;
256 b1[0x1A] = b2[0x1A] + b2[0x1B];
257 b1[0x1B] = (b2[0x1B] - b2[0x1A]) * cos0;
258 b1[0x1A] += b1[0x1B];
259
260 b1[0x1C] = b2[0x1C] + b2[0x1D];
261 b1[0x1D] = (b2[0x1C] - b2[0x1D]) * cos0;
262 b1[0x1E] = b2[0x1E] + b2[0x1F];
263 b1[0x1F] = (b2[0x1F] - b2[0x1E]) * cos0;
264 b1[0x1E] += b1[0x1F];
265 b1[0x1C] += b1[0x1E];
266 b1[0x1E] += b1[0x1D];
267 b1[0x1D] += b1[0x1F];
268 }
269
270 out0[0x10*16] = b1[0x00];
271 out0[0x10*12] = b1[0x04];
272 out0[0x10* 8] = b1[0x02];
273 out0[0x10* 4] = b1[0x06];
274 out0[0x10* 0] = b1[0x01];
275 out1[0x10* 0] = b1[0x01];
276 out1[0x10* 4] = b1[0x05];
277 out1[0x10* 8] = b1[0x03];
278 out1[0x10*12] = b1[0x07];
279
280 out0[0x10*14] = b1[0x08] + b1[0x0C];
281 out0[0x10*10] = b1[0x0C] + b1[0x0a];
282 out0[0x10* 6] = b1[0x0A] + b1[0x0E];
283 out0[0x10* 2] = b1[0x0E] + b1[0x09];
284 out1[0x10* 2] = b1[0x09] + b1[0x0D];
285 out1[0x10* 6] = b1[0x0D] + b1[0x0B];
286 out1[0x10*10] = b1[0x0B] + b1[0x0F];
287 out1[0x10*14] = b1[0x0F];
288
289 {
290 register float tmp;
291 tmp = b1[0x18] + b1[0x1C];
292 out0[0x10*15] = tmp + b1[0x10];
293 out0[0x10*13] = tmp + b1[0x14];
294 tmp = b1[0x1C] + b1[0x1A];
295 out0[0x10*11] = tmp + b1[0x14];
296 out0[0x10* 9] = tmp + b1[0x12];
297 tmp = b1[0x1A] + b1[0x1E];
298 out0[0x10* 7] = tmp + b1[0x12];
299 out0[0x10* 5] = tmp + b1[0x16];
300 tmp = b1[0x1E] + b1[0x19];
301 out0[0x10* 3] = tmp + b1[0x16];
302 out0[0x10* 1] = tmp + b1[0x11];
303 tmp = b1[0x19] + b1[0x1D];
304 out1[0x10* 1] = tmp + b1[0x11];
305 out1[0x10* 3] = tmp + b1[0x15];
306 tmp = b1[0x1D] + b1[0x1B];
307 out1[0x10* 5] = tmp + b1[0x15];
308 out1[0x10* 7] = tmp + b1[0x13];
309 tmp = b1[0x1B] + b1[0x1F];
310 out1[0x10* 9] = tmp + b1[0x13];
311 out1[0x10*11] = tmp + b1[0x17];
312 out1[0x10*13] = b1[0x17] + b1[0x1F];
313 out1[0x10*15] = b1[0x1F];
314 }
315 return 0;
316 }
317
318 /*
319 * the call via dct64 is a trick to force GCC to use
320 * (new) registers for the b1,b2 pointer to the bufs[xx] field
321 */
322 int mpeg3audio_dct64(float *a, float *b, float *c)
323 {
324 float bufs[0x40];
325 return mpeg3audio_dct64_1(a, b, bufs, bufs + 0x20, c);
326 }
327
328 /*//////////////////////////////////////////////////////////////// */
329 /* */
330 /* 9 Point Inverse Discrete Cosine Transform */
331 /* */
332 /* This piece of code is Copyright 1997 Mikko Tommila and is freely usable */
333 /* by anybody. The algorithm itself is of course in the public domain. */
334 /* */
335 /* Again derived heuristically from the 9-point WFTA. */
336 /* */
337 /* The algorithm is optimized (?) for speed, not for small rounding errors or */
338 /* good readability. */
339 /* */
340 /* 36 additions, 11 multiplications */
341 /* */
342 /* Again this is very likely sub-optimal. */
343 /* */
344 /* The code is optimized to use a minimum number of temporary variables, */
345 /* so it should compile quite well even on 8-register Intel x86 processors. */
346 /* This makes the code quite obfuscated and very difficult to understand. */
347 /* */
348 /* References: */
349 /* [1] S. Winograd: "On Computing the Discrete Fourier Transform", */
350 /* Mathematics of Computation, Volume 32, Number 141, January 1978, */
351 /* Pages 175-199 */
352
353
354 /*------------------------------------------------------------------*/
355 /* */
356 /* Function: Calculation of the inverse MDCT */
357 /* */
358 /*------------------------------------------------------------------*/
359
360 int mpeg3audio_dct36(float *inbuf, float *o1, float *o2, float *wintab, float *tsbuf)
361 {
362 float tmp[18];
363
364 {
365 register float *in = inbuf;
366
367 in[17]+=in[16]; in[16]+=in[15]; in[15]+=in[14];
368 in[14]+=in[13]; in[13]+=in[12]; in[12]+=in[11];
369 in[11]+=in[10]; in[10]+=in[9]; in[9] +=in[8];
370 in[8] +=in[7]; in[7] +=in[6]; in[6] +=in[5];
371 in[5] +=in[4]; in[4] +=in[3]; in[3] +=in[2];
372 in[2] +=in[1]; in[1] +=in[0];
373
374 in[17]+=in[15]; in[15]+=in[13]; in[13]+=in[11]; in[11]+=in[9];
375 in[9] +=in[7]; in[7] +=in[5]; in[5] +=in[3]; in[3] +=in[1];
376
377
378 {
379 float t3;
380 {
381 float t0, t1, t2;
382
383 t0 = mpeg3_COS6_2 * (in[8] + in[16] - in[4]);
384 t1 = mpeg3_COS6_2 * in[12];
385
386 t3 = in[0];
387 t2 = t3 - t1 - t1;
388 tmp[1] = tmp[7] = t2 - t0;
389 tmp[4] = t2 + t0 + t0;
390 t3 += t1;
391
392 t2 = mpeg3_COS6_1 * (in[10] + in[14] - in[2]);
393 tmp[1] -= t2;
394 tmp[7] += t2;
395 }
396 {
397 float t0, t1, t2;
398
399 t0 = mpeg3_cos9[0] * (in[4] + in[8] );
400 t1 = mpeg3_cos9[1] * (in[8] - in[16]);
401 t2 = mpeg3_cos9[2] * (in[4] + in[16]);
402
403 tmp[2] = tmp[6] = t3 - t0 - t2;
404 tmp[0] = tmp[8] = t3 + t0 + t1;
405 tmp[3] = tmp[5] = t3 - t1 + t2;
406 }
407 }
408 {
409 float t1, t2, t3;
410
411 t1 = mpeg3_cos18[0] * (in[2] + in[10]);
412 t2 = mpeg3_cos18[1] * (in[10] - in[14]);
413 t3 = mpeg3_COS6_1 * in[6];
414
415 {
416 float t0 = t1 + t2 + t3;
417 tmp[0] += t0;
418 tmp[8] -= t0;
419 }
420
421 t2 -= t3;
422 t1 -= t3;
423
424 t3 = mpeg3_cos18[2] * (in[2] + in[14]);
425
426 t1 += t3;
427 tmp[3] += t1;
428 tmp[5] -= t1;
429
430 t2 -= t3;
431 tmp[2] += t2;
432 tmp[6] -= t2;
433 }
434
435
436 {
437 float t0, t1, t2, t3, t4, t5, t6, t7;
438
439 t1 = mpeg3_COS6_2 * in[13];
440 t2 = mpeg3_COS6_2 * (in[9] + in[17] - in[5]);
441
442 t3 = in[1] + t1;
443 t4 = in[1] - t1 - t1;
444 t5 = t4 - t2;
445
446 t0 = mpeg3_cos9[0] * (in[5] + in[9]);
447 t1 = mpeg3_cos9[1] * (in[9] - in[17]);
448
449 tmp[13] = (t4 + t2 + t2) * mpeg3_tfcos36[17-13];
450 t2 = mpeg3_cos9[2] * (in[5] + in[17]);
451
452 t6 = t3 - t0 - t2;
453 t0 += t3 + t1;
454 t3 += t2 - t1;
455
456 t2 = mpeg3_cos18[0] * (in[3] + in[11]);
457 t4 = mpeg3_cos18[1] * (in[11] - in[15]);
458 t7 = mpeg3_COS6_1 * in[7];
459
460 t1 = t2 + t4 + t7;
461 tmp[17] = (t0 + t1) * mpeg3_tfcos36[17-17];
462 tmp[9] = (t0 - t1) * mpeg3_tfcos36[17-9];
463 t1 = mpeg3_cos18[2] * (in[3] + in[15]);
464 t2 += t1 - t7;
465
466 tmp[14] = (t3 + t2) * mpeg3_tfcos36[17-14];
467 t0 = mpeg3_COS6_1 * (in[11] + in[15] - in[3]);
468 tmp[12] = (t3 - t2) * mpeg3_tfcos36[17-12];
469
470 t4 -= t1 + t7;
471
472 tmp[16] = (t5 - t0) * mpeg3_tfcos36[17-16];
473 tmp[10] = (t5 + t0) * mpeg3_tfcos36[17-10];
474 tmp[15] = (t6 + t4) * mpeg3_tfcos36[17-15];
475 tmp[11] = (t6 - t4) * mpeg3_tfcos36[17-11];
476 }
477
478 #define MACRO(v) \
479 { \
480 float tmpval; \
481 tmpval = tmp[(v)] + tmp[17-(v)]; \
482 out2[9+(v)] = tmpval * w[27+(v)]; \
483 out2[8-(v)] = tmpval * w[26-(v)]; \
484 tmpval = tmp[(v)] - tmp[17-(v)]; \
485 ts[SBLIMIT*(8-(v))] = out1[8-(v)] + tmpval * w[8-(v)]; \
486 ts[SBLIMIT*(9+(v))] = out1[9+(v)] + tmpval * w[9+(v)]; \
487 }
488
489 {
490 register float *out2 = o2;
491 register float *w = wintab;
492 register float *out1 = o1;
493 register float *ts = tsbuf;
494
495 MACRO(0);
496 MACRO(1);
497 MACRO(2);
498 MACRO(3);
499 MACRO(4);
500 MACRO(5);
501 MACRO(6);
502 MACRO(7);
503 MACRO(8);
504 }
505 }
506 return 0;
507 }
508
509 /*
510 * new DCT12
511 */
512 int mpeg3audio_dct12(float *in,float *rawout1,float *rawout2,register float *wi,register float *ts)
513 {
514 #define DCT12_PART1 \
515 in5 = in[5*3]; \
516 in5 += (in4 = in[4*3]); \
517 in4 += (in3 = in[3*3]); \
518 in3 += (in2 = in[2*3]); \
519 in2 += (in1 = in[1*3]); \
520 in1 += (in0 = in[0*3]); \
521 \
522 in5 += in3; in3 += in1; \
523 \
524 in2 *= mpeg3_COS6_1; \
525 in3 *= mpeg3_COS6_1; \
526
527 #define DCT12_PART2 \
528 in0 += in4 * mpeg3_COS6_2; \
529 \
530 in4 = in0 + in2; \
531 in0 -= in2; \
532 \
533 in1 += in5 * mpeg3_COS6_2; \
534 \
535 in5 = (in1 + in3) * mpeg3_tfcos12[0]; \
536 in1 = (in1 - in3) * mpeg3_tfcos12[2]; \
537 \
538 in3 = in4 + in5; \
539 in4 -= in5; \
540 \
541 in2 = in0 + in1; \
542 in0 -= in1;
543
544
545 {
546 float in0,in1,in2,in3,in4,in5;
547 register float *out1 = rawout1;
548 ts[SBLIMIT*0] = out1[0]; ts[SBLIMIT*1] = out1[1]; ts[SBLIMIT*2] = out1[2];
549 ts[SBLIMIT*3] = out1[3]; ts[SBLIMIT*4] = out1[4]; ts[SBLIMIT*5] = out1[5];
550
551 DCT12_PART1
552
553 {
554 float tmp0,tmp1 = (in0 - in4);
555 {
556 float tmp2 = (in1 - in5) * mpeg3_tfcos12[1];
557 tmp0 = tmp1 + tmp2;
558 tmp1 -= tmp2;
559 }
560 ts[(17-1)*SBLIMIT] = out1[17-1] + tmp0 * wi[11-1];
561 ts[(12+1)*SBLIMIT] = out1[12+1] + tmp0 * wi[6+1];
562 ts[(6 +1)*SBLIMIT] = out1[6 +1] + tmp1 * wi[1];
563 ts[(11-1)*SBLIMIT] = out1[11-1] + tmp1 * wi[5-1];
564 }
565
566 DCT12_PART2
567
568 ts[(17-0)*SBLIMIT] = out1[17-0] + in2 * wi[11-0];
569 ts[(12+0)*SBLIMIT] = out1[12+0] + in2 * wi[6+0];
570 ts[(12+2)*SBLIMIT] = out1[12+2] + in3 * wi[6+2];
571 ts[(17-2)*SBLIMIT] = out1[17-2] + in3 * wi[11-2];
572
573 ts[(6+0)*SBLIMIT] = out1[6+0] + in0 * wi[0];
574 ts[(11-0)*SBLIMIT] = out1[11-0] + in0 * wi[5-0];
575 ts[(6+2)*SBLIMIT] = out1[6+2] + in4 * wi[2];
576 ts[(11-2)*SBLIMIT] = out1[11-2] + in4 * wi[5-2];
577 }
578
579 in++;
580
581 {
582 float in0,in1,in2,in3,in4,in5;
583 register float *out2 = rawout2;
584
585 DCT12_PART1
586
587 {
588 float tmp0,tmp1 = (in0 - in4);
589 {
590 float tmp2 = (in1 - in5) * mpeg3_tfcos12[1];
591 tmp0 = tmp1 + tmp2;
592 tmp1 -= tmp2;
593 }
594 out2[5-1] = tmp0 * wi[11-1];
595 out2[0+1] = tmp0 * wi[6+1];
596 ts[(12+1)*SBLIMIT] += tmp1 * wi[1];
597 ts[(17-1)*SBLIMIT] += tmp1 * wi[5-1];
598 }
599
600 DCT12_PART2
601
602 out2[5-0] = in2 * wi[11-0];
603 out2[0+0] = in2 * wi[6+0];
604 out2[0+2] = in3 * wi[6+2];
605 out2[5-2] = in3 * wi[11-2];
606
607 ts[(12+0)*SBLIMIT] += in0 * wi[0];
608 ts[(17-0)*SBLIMIT] += in0 * wi[5-0];
609 ts[(12+2)*SBLIMIT] += in4 * wi[2];
610 ts[(17-2)*SBLIMIT] += in4 * wi[5-2];
611 }
612
613 in++;
614
615 {
616 float in0,in1,in2,in3,in4,in5;
617 register float *out2 = rawout2;
618 out2[12]=out2[13]=out2[14]=out2[15]=out2[16]=out2[17]=0.0;
619
620 DCT12_PART1
621
622 {
623 float tmp0,tmp1 = (in0 - in4);
624 {
625 float tmp2 = (in1 - in5) * mpeg3_tfcos12[1];
626 tmp0 = tmp1 + tmp2;
627 tmp1 -= tmp2;
628 }
629 out2[11-1] = tmp0 * wi[11-1];
630 out2[6 +1] = tmp0 * wi[6+1];
631 out2[0+1] += tmp1 * wi[1];
632 out2[5-1] += tmp1 * wi[5-1];
633 }
634
635 DCT12_PART2
636
637 out2[11-0] = in2 * wi[11-0];
638 out2[6 +0] = in2 * wi[6+0];
639 out2[6 +2] = in3 * wi[6+2];
640 out2[11-2] = in3 * wi[11-2];
641
642 out2[0+0] += in0 * wi[0];
643 out2[5-0] += in0 * wi[5-0];
644 out2[0+2] += in4 * wi[2];
645 out2[5-2] += in4 * wi[5-2];
646 }
647 return 0;
648 }
cehtehs branch of CinelerraCV