Move printing of ID_AID_???_LANG to ts_add_stream instead of ts_detect_streams.
[mplayer/glamo.git] / mp3lib / dct64_altivec.c
blob21a7b886995f70ccb73c868a0e3c572a03959f48
2 /*
3 * Discrete Cosine Tansform (DCT) for Altivec
4 * Copyright (c) 2004 Romain Dolbeau <romain@dolbeau.org>
5 * based upon code from "mp3lib/dct64.c"
6 * This file is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public License
8 */
10 #define real float
12 #include <stdio.h>
13 #include "mpg123.h"
15 #ifdef HAVE_ALTIVEC_H
16 #include <altivec.h>
17 #endif
19 // used to build registers permutation vectors (vcprm)
20 // the 's' are for words in the _s_econd vector
21 #define WORD_0 0x00,0x01,0x02,0x03
22 #define WORD_1 0x04,0x05,0x06,0x07
23 #define WORD_2 0x08,0x09,0x0a,0x0b
24 #define WORD_3 0x0c,0x0d,0x0e,0x0f
25 #define WORD_s0 0x10,0x11,0x12,0x13
26 #define WORD_s1 0x14,0x15,0x16,0x17
27 #define WORD_s2 0x18,0x19,0x1a,0x1b
28 #define WORD_s3 0x1c,0x1d,0x1e,0x1f
30 #define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
31 #define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
33 #define FOUROF(a) {a,a,a,a}
35 // vcprmle is used to keep the same index as in the SSE version.
36 // it's the same as vcprm, with the index inversed
37 // ('le' is Little Endian)
38 #define vcprmle(a,b,c,d) vcprm(d,c,b,a)
40 // used to build inverse/identity vectors (vcii)
41 // n is _n_egative, p is _p_ositive
42 #define FLOAT_n -1.
43 #define FLOAT_p 1.
45 void dct64_altivec(real *a,real *b,real *c)
47 real __attribute__ ((aligned(16))) b1[0x20];
48 real __attribute__ ((aligned(16))) b2[0x20];
50 real *out0 = a;
51 real *out1 = b;
52 real *samples = c;
54 const vector float vczero = (const vector float)FOUROF(0.);
55 const vector unsigned char reverse = (const vector unsigned char)vcprm(3,2,1,0);
58 if (((unsigned long)b1 & 0x0000000F) ||
59 ((unsigned long)b2 & 0x0000000F))
62 printf("MISALIGNED:\t%p\t%p\t%p\t%p\t%p\n",
63 b1, b2, a, b, samples);
67 #ifdef ALTIVEC_USE_REFERENCE_C_CODE
70 register real *costab = mp3lib_pnts[0];
72 b1[0x00] = samples[0x00] + samples[0x1F];
73 b1[0x01] = samples[0x01] + samples[0x1E];
74 b1[0x02] = samples[0x02] + samples[0x1D];
75 b1[0x03] = samples[0x03] + samples[0x1C];
76 b1[0x04] = samples[0x04] + samples[0x1B];
77 b1[0x05] = samples[0x05] + samples[0x1A];
78 b1[0x06] = samples[0x06] + samples[0x19];
79 b1[0x07] = samples[0x07] + samples[0x18];
80 b1[0x08] = samples[0x08] + samples[0x17];
81 b1[0x09] = samples[0x09] + samples[0x16];
82 b1[0x0A] = samples[0x0A] + samples[0x15];
83 b1[0x0B] = samples[0x0B] + samples[0x14];
84 b1[0x0C] = samples[0x0C] + samples[0x13];
85 b1[0x0D] = samples[0x0D] + samples[0x12];
86 b1[0x0E] = samples[0x0E] + samples[0x11];
87 b1[0x0F] = samples[0x0F] + samples[0x10];
88 b1[0x10] = (samples[0x0F] - samples[0x10]) * costab[0xF];
89 b1[0x11] = (samples[0x0E] - samples[0x11]) * costab[0xE];
90 b1[0x12] = (samples[0x0D] - samples[0x12]) * costab[0xD];
91 b1[0x13] = (samples[0x0C] - samples[0x13]) * costab[0xC];
92 b1[0x14] = (samples[0x0B] - samples[0x14]) * costab[0xB];
93 b1[0x15] = (samples[0x0A] - samples[0x15]) * costab[0xA];
94 b1[0x16] = (samples[0x09] - samples[0x16]) * costab[0x9];
95 b1[0x17] = (samples[0x08] - samples[0x17]) * costab[0x8];
96 b1[0x18] = (samples[0x07] - samples[0x18]) * costab[0x7];
97 b1[0x19] = (samples[0x06] - samples[0x19]) * costab[0x6];
98 b1[0x1A] = (samples[0x05] - samples[0x1A]) * costab[0x5];
99 b1[0x1B] = (samples[0x04] - samples[0x1B]) * costab[0x4];
100 b1[0x1C] = (samples[0x03] - samples[0x1C]) * costab[0x3];
101 b1[0x1D] = (samples[0x02] - samples[0x1D]) * costab[0x2];
102 b1[0x1E] = (samples[0x01] - samples[0x1E]) * costab[0x1];
103 b1[0x1F] = (samples[0x00] - samples[0x1F]) * costab[0x0];
107 register real *costab = mp3lib_pnts[1];
109 b2[0x00] = b1[0x00] + b1[0x0F];
110 b2[0x01] = b1[0x01] + b1[0x0E];
111 b2[0x02] = b1[0x02] + b1[0x0D];
112 b2[0x03] = b1[0x03] + b1[0x0C];
113 b2[0x04] = b1[0x04] + b1[0x0B];
114 b2[0x05] = b1[0x05] + b1[0x0A];
115 b2[0x06] = b1[0x06] + b1[0x09];
116 b2[0x07] = b1[0x07] + b1[0x08];
117 b2[0x08] = (b1[0x07] - b1[0x08]) * costab[7];
118 b2[0x09] = (b1[0x06] - b1[0x09]) * costab[6];
119 b2[0x0A] = (b1[0x05] - b1[0x0A]) * costab[5];
120 b2[0x0B] = (b1[0x04] - b1[0x0B]) * costab[4];
121 b2[0x0C] = (b1[0x03] - b1[0x0C]) * costab[3];
122 b2[0x0D] = (b1[0x02] - b1[0x0D]) * costab[2];
123 b2[0x0E] = (b1[0x01] - b1[0x0E]) * costab[1];
124 b2[0x0F] = (b1[0x00] - b1[0x0F]) * costab[0];
125 b2[0x10] = b1[0x10] + b1[0x1F];
126 b2[0x11] = b1[0x11] + b1[0x1E];
127 b2[0x12] = b1[0x12] + b1[0x1D];
128 b2[0x13] = b1[0x13] + b1[0x1C];
129 b2[0x14] = b1[0x14] + b1[0x1B];
130 b2[0x15] = b1[0x15] + b1[0x1A];
131 b2[0x16] = b1[0x16] + b1[0x19];
132 b2[0x17] = b1[0x17] + b1[0x18];
133 b2[0x18] = (b1[0x18] - b1[0x17]) * costab[7];
134 b2[0x19] = (b1[0x19] - b1[0x16]) * costab[6];
135 b2[0x1A] = (b1[0x1A] - b1[0x15]) * costab[5];
136 b2[0x1B] = (b1[0x1B] - b1[0x14]) * costab[4];
137 b2[0x1C] = (b1[0x1C] - b1[0x13]) * costab[3];
138 b2[0x1D] = (b1[0x1D] - b1[0x12]) * costab[2];
139 b2[0x1E] = (b1[0x1E] - b1[0x11]) * costab[1];
140 b2[0x1F] = (b1[0x1F] - b1[0x10]) * costab[0];
145 register real *costab = mp3lib_pnts[2];
147 b1[0x00] = b2[0x00] + b2[0x07];
148 b1[0x01] = b2[0x01] + b2[0x06];
149 b1[0x02] = b2[0x02] + b2[0x05];
150 b1[0x03] = b2[0x03] + b2[0x04];
151 b1[0x04] = (b2[0x03] - b2[0x04]) * costab[3];
152 b1[0x05] = (b2[0x02] - b2[0x05]) * costab[2];
153 b1[0x06] = (b2[0x01] - b2[0x06]) * costab[1];
154 b1[0x07] = (b2[0x00] - b2[0x07]) * costab[0];
155 b1[0x08] = b2[0x08] + b2[0x0F];
156 b1[0x09] = b2[0x09] + b2[0x0E];
157 b1[0x0A] = b2[0x0A] + b2[0x0D];
158 b1[0x0B] = b2[0x0B] + b2[0x0C];
159 b1[0x0C] = (b2[0x0C] - b2[0x0B]) * costab[3];
160 b1[0x0D] = (b2[0x0D] - b2[0x0A]) * costab[2];
161 b1[0x0E] = (b2[0x0E] - b2[0x09]) * costab[1];
162 b1[0x0F] = (b2[0x0F] - b2[0x08]) * costab[0];
163 b1[0x10] = b2[0x10] + b2[0x17];
164 b1[0x11] = b2[0x11] + b2[0x16];
165 b1[0x12] = b2[0x12] + b2[0x15];
166 b1[0x13] = b2[0x13] + b2[0x14];
167 b1[0x14] = (b2[0x13] - b2[0x14]) * costab[3];
168 b1[0x15] = (b2[0x12] - b2[0x15]) * costab[2];
169 b1[0x16] = (b2[0x11] - b2[0x16]) * costab[1];
170 b1[0x17] = (b2[0x10] - b2[0x17]) * costab[0];
171 b1[0x18] = b2[0x18] + b2[0x1F];
172 b1[0x19] = b2[0x19] + b2[0x1E];
173 b1[0x1A] = b2[0x1A] + b2[0x1D];
174 b1[0x1B] = b2[0x1B] + b2[0x1C];
175 b1[0x1C] = (b2[0x1C] - b2[0x1B]) * costab[3];
176 b1[0x1D] = (b2[0x1D] - b2[0x1A]) * costab[2];
177 b1[0x1E] = (b2[0x1E] - b2[0x19]) * costab[1];
178 b1[0x1F] = (b2[0x1F] - b2[0x18]) * costab[0];
181 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
183 // How does it work ?
184 // the first three passes are reproducted in the three block below
185 // all computations are done on a 4 elements vector
186 // 'reverse' is a special perumtation vector used to reverse
187 // the order of the elements inside a vector.
188 // note that all loads/stores to b1 (b2) between passes 1 and 2 (2 and 3)
189 // have been removed, all elements are stored inside b1vX (b2vX)
191 register vector float
192 b1v0, b1v1, b1v2, b1v3,
193 b1v4, b1v5, b1v6, b1v7;
194 register vector float
195 temp1, temp2;
198 register real *costab = mp3lib_pnts[0];
200 register vector float
201 samplesv1, samplesv2, samplesv3, samplesv4,
202 samplesv5, samplesv6, samplesv7, samplesv8,
203 samplesv9;
204 register vector unsigned char samples_perm = vec_lvsl(0, samples);
205 register vector float costabv1, costabv2, costabv3, costabv4, costabv5;
206 register vector unsigned char costab_perm = vec_lvsl(0, costab);
208 samplesv1 = vec_ld(0, samples);
209 samplesv2 = vec_ld(16, samples);
210 samplesv1 = vec_perm(samplesv1, samplesv2, samples_perm);
211 samplesv3 = vec_ld(32, samples);
212 samplesv2 = vec_perm(samplesv2, samplesv3, samples_perm);
213 samplesv4 = vec_ld(48, samples);
214 samplesv3 = vec_perm(samplesv3, samplesv4, samples_perm);
215 samplesv5 = vec_ld(64, samples);
216 samplesv4 = vec_perm(samplesv4, samplesv5, samples_perm);
217 samplesv6 = vec_ld(80, samples);
218 samplesv5 = vec_perm(samplesv5, samplesv6, samples_perm);
219 samplesv7 = vec_ld(96, samples);
220 samplesv6 = vec_perm(samplesv6, samplesv7, samples_perm);
221 samplesv8 = vec_ld(112, samples);
222 samplesv7 = vec_perm(samplesv7, samplesv8, samples_perm);
223 samplesv9 = vec_ld(128, samples);
224 samplesv8 = vec_perm(samplesv8, samplesv9, samples_perm);
226 temp1 = vec_add(samplesv1,
227 vec_perm(samplesv8, samplesv8, reverse));
228 //vec_st(temp1, 0, b1);
229 b1v0 = temp1;
230 temp1 = vec_add(samplesv2,
231 vec_perm(samplesv7, samplesv7, reverse));
232 //vec_st(temp1, 16, b1);
233 b1v1 = temp1;
234 temp1 = vec_add(samplesv3,
235 vec_perm(samplesv6, samplesv6, reverse));
236 //vec_st(temp1, 32, b1);
237 b1v2 = temp1;
238 temp1 = vec_add(samplesv4,
239 vec_perm(samplesv5, samplesv5, reverse));
240 //vec_st(temp1, 48, b1);
241 b1v3 = temp1;
243 costabv1 = vec_ld(0, costab);
244 costabv2 = vec_ld(16, costab);
245 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
246 costabv3 = vec_ld(32, costab);
247 costabv2 = vec_perm(costabv2, costabv3, costab_perm);
248 costabv4 = vec_ld(48, costab);
249 costabv3 = vec_perm(costabv3, costabv4, costab_perm);
250 costabv5 = vec_ld(64, costab);
251 costabv4 = vec_perm(costabv4, costabv5, costab_perm);
253 temp1 = vec_sub(vec_perm(samplesv4, samplesv4, reverse),
254 samplesv5);
255 temp2 = vec_madd(temp1,
256 vec_perm(costabv4, costabv4, reverse),
257 vczero);
258 //vec_st(temp2, 64, b1);
259 b1v4 = temp2;
261 temp1 = vec_sub(vec_perm(samplesv3, samplesv3, reverse),
262 samplesv6);
263 temp2 = vec_madd(temp1,
264 vec_perm(costabv3, costabv3, reverse),
265 vczero);
266 //vec_st(temp2, 80, b1);
267 b1v5 = temp2;
268 temp1 = vec_sub(vec_perm(samplesv2, samplesv2, reverse),
269 samplesv7);
270 temp2 = vec_madd(temp1,
271 vec_perm(costabv2, costabv2, reverse),
272 vczero);
273 //vec_st(temp2, 96, b1);
274 b1v6 = temp2;
276 temp1 = vec_sub(vec_perm(samplesv1, samplesv1, reverse),
277 samplesv8);
278 temp2 = vec_madd(temp1,
279 vec_perm(costabv1, costabv1, reverse),
280 vczero);
281 //vec_st(temp2, 112, b1);
282 b1v7 = temp2;
287 register vector float
288 b2v0, b2v1, b2v2, b2v3,
289 b2v4, b2v5, b2v6, b2v7;
291 register real *costab = mp3lib_pnts[1];
292 register vector float costabv1r, costabv2r, costabv1, costabv2, costabv3;
293 register vector unsigned char costab_perm = vec_lvsl(0, costab);
295 costabv1 = vec_ld(0, costab);
296 costabv2 = vec_ld(16, costab);
297 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
298 costabv3 = vec_ld(32, costab);
299 costabv2 = vec_perm(costabv2, costabv3 , costab_perm);
300 costabv1r = vec_perm(costabv1, costabv1, reverse);
301 costabv2r = vec_perm(costabv2, costabv2, reverse);
303 temp1 = vec_add(b1v0, vec_perm(b1v3, b1v3, reverse));
304 //vec_st(temp1, 0, b2);
305 b2v0 = temp1;
306 temp1 = vec_add(b1v1, vec_perm(b1v2, b1v2, reverse));
307 //vec_st(temp1, 16, b2);
308 b2v1 = temp1;
309 temp2 = vec_sub(vec_perm(b1v1, b1v1, reverse), b1v2);
310 temp1 = vec_madd(temp2, costabv2r, vczero);
311 //vec_st(temp1, 32, b2);
312 b2v2 = temp1;
313 temp2 = vec_sub(vec_perm(b1v0, b1v0, reverse), b1v3);
314 temp1 = vec_madd(temp2, costabv1r, vczero);
315 //vec_st(temp1, 48, b2);
316 b2v3 = temp1;
317 temp1 = vec_add(b1v4, vec_perm(b1v7, b1v7, reverse));
318 //vec_st(temp1, 64, b2);
319 b2v4 = temp1;
320 temp1 = vec_add(b1v5, vec_perm(b1v6, b1v6, reverse));
321 //vec_st(temp1, 80, b2);
322 b2v5 = temp1;
323 temp2 = vec_sub(b1v6, vec_perm(b1v5, b1v5, reverse));
324 temp1 = vec_madd(temp2, costabv2r, vczero);
325 //vec_st(temp1, 96, b2);
326 b2v6 = temp1;
327 temp2 = vec_sub(b1v7, vec_perm(b1v4, b1v4, reverse));
328 temp1 = vec_madd(temp2, costabv1r, vczero);
329 //vec_st(temp1, 112, b2);
330 b2v7 = temp1;
334 register real *costab = mp3lib_pnts[2];
337 vector float costabv1r, costabv1, costabv2;
338 vector unsigned char costab_perm = vec_lvsl(0, costab);
340 costabv1 = vec_ld(0, costab);
341 costabv2 = vec_ld(16, costab);
342 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
343 costabv1r = vec_perm(costabv1, costabv1, reverse);
345 temp1 = vec_add(b2v0, vec_perm(b2v1, b2v1, reverse));
346 vec_st(temp1, 0, b1);
347 temp2 = vec_sub(vec_perm(b2v0, b2v0, reverse), b2v1);
348 temp1 = vec_madd(temp2, costabv1r, vczero);
349 vec_st(temp1, 16, b1);
351 temp1 = vec_add(b2v2, vec_perm(b2v3, b2v3, reverse));
352 vec_st(temp1, 32, b1);
353 temp2 = vec_sub(b2v3, vec_perm(b2v2, b2v2, reverse));
354 temp1 = vec_madd(temp2, costabv1r, vczero);
355 vec_st(temp1, 48, b1);
357 temp1 = vec_add(b2v4, vec_perm(b2v5, b2v5, reverse));
358 vec_st(temp1, 64, b1);
359 temp2 = vec_sub(vec_perm(b2v4, b2v4, reverse), b2v5);
360 temp1 = vec_madd(temp2, costabv1r, vczero);
361 vec_st(temp1, 80, b1);
363 temp1 = vec_add(b2v6, vec_perm(b2v7, b2v7, reverse));
364 vec_st(temp1, 96, b1);
365 temp2 = vec_sub(b2v7, vec_perm(b2v6, b2v6, reverse));
366 temp1 = vec_madd(temp2, costabv1r, vczero);
367 vec_st(temp1, 112, b1);
373 #endif /* ALTIVEC_USE_REFERENCE_C_CODE */
376 register real const cos0 = mp3lib_pnts[3][0];
377 register real const cos1 = mp3lib_pnts[3][1];
379 b2[0x00] = b1[0x00] + b1[0x03];
380 b2[0x01] = b1[0x01] + b1[0x02];
381 b2[0x02] = (b1[0x01] - b1[0x02]) * cos1;
382 b2[0x03] = (b1[0x00] - b1[0x03]) * cos0;
383 b2[0x04] = b1[0x04] + b1[0x07];
384 b2[0x05] = b1[0x05] + b1[0x06];
385 b2[0x06] = (b1[0x06] - b1[0x05]) * cos1;
386 b2[0x07] = (b1[0x07] - b1[0x04]) * cos0;
387 b2[0x08] = b1[0x08] + b1[0x0B];
388 b2[0x09] = b1[0x09] + b1[0x0A];
389 b2[0x0A] = (b1[0x09] - b1[0x0A]) * cos1;
390 b2[0x0B] = (b1[0x08] - b1[0x0B]) * cos0;
391 b2[0x0C] = b1[0x0C] + b1[0x0F];
392 b2[0x0D] = b1[0x0D] + b1[0x0E];
393 b2[0x0E] = (b1[0x0E] - b1[0x0D]) * cos1;
394 b2[0x0F] = (b1[0x0F] - b1[0x0C]) * cos0;
395 b2[0x10] = b1[0x10] + b1[0x13];
396 b2[0x11] = b1[0x11] + b1[0x12];
397 b2[0x12] = (b1[0x11] - b1[0x12]) * cos1;
398 b2[0x13] = (b1[0x10] - b1[0x13]) * cos0;
399 b2[0x14] = b1[0x14] + b1[0x17];
400 b2[0x15] = b1[0x15] + b1[0x16];
401 b2[0x16] = (b1[0x16] - b1[0x15]) * cos1;
402 b2[0x17] = (b1[0x17] - b1[0x14]) * cos0;
403 b2[0x18] = b1[0x18] + b1[0x1B];
404 b2[0x19] = b1[0x19] + b1[0x1A];
405 b2[0x1A] = (b1[0x19] - b1[0x1A]) * cos1;
406 b2[0x1B] = (b1[0x18] - b1[0x1B]) * cos0;
407 b2[0x1C] = b1[0x1C] + b1[0x1F];
408 b2[0x1D] = b1[0x1D] + b1[0x1E];
409 b2[0x1E] = (b1[0x1E] - b1[0x1D]) * cos1;
410 b2[0x1F] = (b1[0x1F] - b1[0x1C]) * cos0;
414 register real const cos0 = mp3lib_pnts[4][0];
416 b1[0x00] = b2[0x00] + b2[0x01];
417 b1[0x01] = (b2[0x00] - b2[0x01]) * cos0;
418 b1[0x02] = b2[0x02] + b2[0x03];
419 b1[0x03] = (b2[0x03] - b2[0x02]) * cos0;
420 b1[0x02] += b1[0x03];
422 b1[0x04] = b2[0x04] + b2[0x05];
423 b1[0x05] = (b2[0x04] - b2[0x05]) * cos0;
424 b1[0x06] = b2[0x06] + b2[0x07];
425 b1[0x07] = (b2[0x07] - b2[0x06]) * cos0;
426 b1[0x06] += b1[0x07];
427 b1[0x04] += b1[0x06];
428 b1[0x06] += b1[0x05];
429 b1[0x05] += b1[0x07];
431 b1[0x08] = b2[0x08] + b2[0x09];
432 b1[0x09] = (b2[0x08] - b2[0x09]) * cos0;
433 b1[0x0A] = b2[0x0A] + b2[0x0B];
434 b1[0x0B] = (b2[0x0B] - b2[0x0A]) * cos0;
435 b1[0x0A] += b1[0x0B];
437 b1[0x0C] = b2[0x0C] + b2[0x0D];
438 b1[0x0D] = (b2[0x0C] - b2[0x0D]) * cos0;
439 b1[0x0E] = b2[0x0E] + b2[0x0F];
440 b1[0x0F] = (b2[0x0F] - b2[0x0E]) * cos0;
441 b1[0x0E] += b1[0x0F];
442 b1[0x0C] += b1[0x0E];
443 b1[0x0E] += b1[0x0D];
444 b1[0x0D] += b1[0x0F];
446 b1[0x10] = b2[0x10] + b2[0x11];
447 b1[0x11] = (b2[0x10] - b2[0x11]) * cos0;
448 b1[0x12] = b2[0x12] + b2[0x13];
449 b1[0x13] = (b2[0x13] - b2[0x12]) * cos0;
450 b1[0x12] += b1[0x13];
452 b1[0x14] = b2[0x14] + b2[0x15];
453 b1[0x15] = (b2[0x14] - b2[0x15]) * cos0;
454 b1[0x16] = b2[0x16] + b2[0x17];
455 b1[0x17] = (b2[0x17] - b2[0x16]) * cos0;
456 b1[0x16] += b1[0x17];
457 b1[0x14] += b1[0x16];
458 b1[0x16] += b1[0x15];
459 b1[0x15] += b1[0x17];
461 b1[0x18] = b2[0x18] + b2[0x19];
462 b1[0x19] = (b2[0x18] - b2[0x19]) * cos0;
463 b1[0x1A] = b2[0x1A] + b2[0x1B];
464 b1[0x1B] = (b2[0x1B] - b2[0x1A]) * cos0;
465 b1[0x1A] += b1[0x1B];
467 b1[0x1C] = b2[0x1C] + b2[0x1D];
468 b1[0x1D] = (b2[0x1C] - b2[0x1D]) * cos0;
469 b1[0x1E] = b2[0x1E] + b2[0x1F];
470 b1[0x1F] = (b2[0x1F] - b2[0x1E]) * cos0;
471 b1[0x1E] += b1[0x1F];
472 b1[0x1C] += b1[0x1E];
473 b1[0x1E] += b1[0x1D];
474 b1[0x1D] += b1[0x1F];
477 out0[0x10*16] = b1[0x00];
478 out0[0x10*12] = b1[0x04];
479 out0[0x10* 8] = b1[0x02];
480 out0[0x10* 4] = b1[0x06];
481 out0[0x10* 0] = b1[0x01];
482 out1[0x10* 0] = b1[0x01];
483 out1[0x10* 4] = b1[0x05];
484 out1[0x10* 8] = b1[0x03];
485 out1[0x10*12] = b1[0x07];
487 b1[0x08] += b1[0x0C];
488 out0[0x10*14] = b1[0x08];
489 b1[0x0C] += b1[0x0a];
490 out0[0x10*10] = b1[0x0C];
491 b1[0x0A] += b1[0x0E];
492 out0[0x10* 6] = b1[0x0A];
493 b1[0x0E] += b1[0x09];
494 out0[0x10* 2] = b1[0x0E];
495 b1[0x09] += b1[0x0D];
496 out1[0x10* 2] = b1[0x09];
497 b1[0x0D] += b1[0x0B];
498 out1[0x10* 6] = b1[0x0D];
499 b1[0x0B] += b1[0x0F];
500 out1[0x10*10] = b1[0x0B];
501 out1[0x10*14] = b1[0x0F];
503 b1[0x18] += b1[0x1C];
504 out0[0x10*15] = b1[0x10] + b1[0x18];
505 out0[0x10*13] = b1[0x18] + b1[0x14];
506 b1[0x1C] += b1[0x1a];
507 out0[0x10*11] = b1[0x14] + b1[0x1C];
508 out0[0x10* 9] = b1[0x1C] + b1[0x12];
509 b1[0x1A] += b1[0x1E];
510 out0[0x10* 7] = b1[0x12] + b1[0x1A];
511 out0[0x10* 5] = b1[0x1A] + b1[0x16];
512 b1[0x1E] += b1[0x19];
513 out0[0x10* 3] = b1[0x16] + b1[0x1E];
514 out0[0x10* 1] = b1[0x1E] + b1[0x11];
515 b1[0x19] += b1[0x1D];
516 out1[0x10* 1] = b1[0x11] + b1[0x19];
517 out1[0x10* 3] = b1[0x19] + b1[0x15];
518 b1[0x1D] += b1[0x1B];
519 out1[0x10* 5] = b1[0x15] + b1[0x1D];
520 out1[0x10* 7] = b1[0x1D] + b1[0x13];
521 b1[0x1B] += b1[0x1F];
522 out1[0x10* 9] = b1[0x13] + b1[0x1B];
523 out1[0x10*11] = b1[0x1B] + b1[0x17];
524 out1[0x10*13] = b1[0x17] + b1[0x1F];
525 out1[0x10*15] = b1[0x1F];