Mercurial > audlegacy-plugins

comparison src/mpg123/dct64_altivec.c @ 12:3da1b8942b8b trunk

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
[svn] - remove src/Input src/Output src/Effect src/General src/Visualization src/Container
author nenolod
date Mon, 18 Sep 2006 03:14:20 -0700
parents src/Input/mpg123/dct64_altivec.c@13389e613d67
children
comparison
equal deleted inserted replaced
11:cff1d04026ae 12:3da1b8942b8b
1
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 */
7
8 #include "mpg123.h"
9
10 #ifdef INTEGER_COMPILE
11 #error AltiVec acceleration and integer compile are mutually exclusive.
12 #endif
13
14 #ifdef HAVE_ALTIVEC
15
16 #ifndef SYS_DARWIN
17 #include <altivec.h>
18 #endif
19
20 // used to build registers permutation vectors (vcprm)
21 // the 's' are for words in the _s_econd vector
22 #define WORD_0 0x00,0x01,0x02,0x03
23 #define WORD_1 0x04,0x05,0x06,0x07
24 #define WORD_2 0x08,0x09,0x0a,0x0b
25 #define WORD_3 0x0c,0x0d,0x0e,0x0f
26 #define WORD_s0 0x10,0x11,0x12,0x13
27 #define WORD_s1 0x14,0x15,0x16,0x17
28 #define WORD_s2 0x18,0x19,0x1a,0x1b
29 #define WORD_s3 0x1c,0x1d,0x1e,0x1f
30
31 #ifdef SYS_DARWIN
32 #define vcprm(a,b,c,d) (const vector unsigned char)(WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d)
33 #else
34 #define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d}
35 #endif
36
37 // vcprmle is used to keep the same index as in the SSE version.
38 // it's the same as vcprm, with the index inversed
39 // ('le' is Little Endian)
40 #define vcprmle(a,b,c,d) vcprm(d,c,b,a)
41
42 // used to build inverse/identity vectors (vcii)
43 // n is _n_egative, p is _p_ositive
44 #define FLOAT_n -1.
45 #define FLOAT_p 1.
46
47 #ifdef SYS_DARWIN
48 #define vcii(a,b,c,d) (const vector float)(FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d)
49 #else
50 #define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d}
51 #endif
52
53 #ifdef SYS_DARWIN
54 #define FOUROF(a) (a)
55 #else
56 #define FOUROF(a) {a,a,a,a}
57 #endif
58
59 void mpgdec_dct64(mpgdec_real *a,mpgdec_real *b,mpgdec_real *c)
60 {
61 mpgdec_real __attribute__ ((aligned(16))) b1[0x20];
62 mpgdec_real __attribute__ ((aligned(16))) b2[0x20];
63
64 mpgdec_real *out0 = a;
65 mpgdec_real *out1 = b;
66 mpgdec_real *samples = c;
67
68 const vector float vczero = (const vector float)FOUROF(0.);
69 const vector unsigned char reverse = (const vector unsigned char)vcprm(3,2,1,0);
70
71
72 if (((unsigned long)b1 & 0x0000000F) ||
73 ((unsigned long)b2 & 0x0000000F))
74
75 {
76 printf("MISALIGNED:\t%p\t%p\t%p\t%p\t%p\n",
77 b1, b2, a, b, samples);
78 }
79
80
81 #ifdef ALTIVEC_USE_REFERENCE_C_CODE
82
83 {
84 register mpgdec_real *costab = mpgdec_pnts[0];
85
86 b1[0x00] = samples[0x00] + samples[0x1F];
87 b1[0x01] = samples[0x01] + samples[0x1E];
88 b1[0x02] = samples[0x02] + samples[0x1D];
89 b1[0x03] = samples[0x03] + samples[0x1C];
90 b1[0x04] = samples[0x04] + samples[0x1B];
91 b1[0x05] = samples[0x05] + samples[0x1A];
92 b1[0x06] = samples[0x06] + samples[0x19];
93 b1[0x07] = samples[0x07] + samples[0x18];
94 b1[0x08] = samples[0x08] + samples[0x17];
95 b1[0x09] = samples[0x09] + samples[0x16];
96 b1[0x0A] = samples[0x0A] + samples[0x15];
97 b1[0x0B] = samples[0x0B] + samples[0x14];
98 b1[0x0C] = samples[0x0C] + samples[0x13];
99 b1[0x0D] = samples[0x0D] + samples[0x12];
100 b1[0x0E] = samples[0x0E] + samples[0x11];
101 b1[0x0F] = samples[0x0F] + samples[0x10];
102 b1[0x10] = (samples[0x0F] - samples[0x10]) * costab[0xF];
103 b1[0x11] = (samples[0x0E] - samples[0x11]) * costab[0xE];
104 b1[0x12] = (samples[0x0D] - samples[0x12]) * costab[0xD];
105 b1[0x13] = (samples[0x0C] - samples[0x13]) * costab[0xC];
106 b1[0x14] = (samples[0x0B] - samples[0x14]) * costab[0xB];
107 b1[0x15] = (samples[0x0A] - samples[0x15]) * costab[0xA];
108 b1[0x16] = (samples[0x09] - samples[0x16]) * costab[0x9];
109 b1[0x17] = (samples[0x08] - samples[0x17]) * costab[0x8];
110 b1[0x18] = (samples[0x07] - samples[0x18]) * costab[0x7];
111 b1[0x19] = (samples[0x06] - samples[0x19]) * costab[0x6];
112 b1[0x1A] = (samples[0x05] - samples[0x1A]) * costab[0x5];
113 b1[0x1B] = (samples[0x04] - samples[0x1B]) * costab[0x4];
114 b1[0x1C] = (samples[0x03] - samples[0x1C]) * costab[0x3];
115 b1[0x1D] = (samples[0x02] - samples[0x1D]) * costab[0x2];
116 b1[0x1E] = (samples[0x01] - samples[0x1E]) * costab[0x1];
117 b1[0x1F] = (samples[0x00] - samples[0x1F]) * costab[0x0];
118
119 }
120 {
121 register mpgdec_real *costab = mpgdec_pnts[1];
122
123 b2[0x00] = b1[0x00] + b1[0x0F];
124 b2[0x01] = b1[0x01] + b1[0x0E];
125 b2[0x02] = b1[0x02] + b1[0x0D];
126 b2[0x03] = b1[0x03] + b1[0x0C];
127 b2[0x04] = b1[0x04] + b1[0x0B];
128 b2[0x05] = b1[0x05] + b1[0x0A];
129 b2[0x06] = b1[0x06] + b1[0x09];
130 b2[0x07] = b1[0x07] + b1[0x08];
131 b2[0x08] = (b1[0x07] - b1[0x08]) * costab[7];
132 b2[0x09] = (b1[0x06] - b1[0x09]) * costab[6];
133 b2[0x0A] = (b1[0x05] - b1[0x0A]) * costab[5];
134 b2[0x0B] = (b1[0x04] - b1[0x0B]) * costab[4];
135 b2[0x0C] = (b1[0x03] - b1[0x0C]) * costab[3];
136 b2[0x0D] = (b1[0x02] - b1[0x0D]) * costab[2];
137 b2[0x0E] = (b1[0x01] - b1[0x0E]) * costab[1];
138 b2[0x0F] = (b1[0x00] - b1[0x0F]) * costab[0];
139 b2[0x10] = b1[0x10] + b1[0x1F];
140 b2[0x11] = b1[0x11] + b1[0x1E];
141 b2[0x12] = b1[0x12] + b1[0x1D];
142 b2[0x13] = b1[0x13] + b1[0x1C];
143 b2[0x14] = b1[0x14] + b1[0x1B];
144 b2[0x15] = b1[0x15] + b1[0x1A];
145 b2[0x16] = b1[0x16] + b1[0x19];
146 b2[0x17] = b1[0x17] + b1[0x18];
147 b2[0x18] = (b1[0x18] - b1[0x17]) * costab[7];
148 b2[0x19] = (b1[0x19] - b1[0x16]) * costab[6];
149 b2[0x1A] = (b1[0x1A] - b1[0x15]) * costab[5];
150 b2[0x1B] = (b1[0x1B] - b1[0x14]) * costab[4];
151 b2[0x1C] = (b1[0x1C] - b1[0x13]) * costab[3];
152 b2[0x1D] = (b1[0x1D] - b1[0x12]) * costab[2];
153 b2[0x1E] = (b1[0x1E] - b1[0x11]) * costab[1];
154 b2[0x1F] = (b1[0x1F] - b1[0x10]) * costab[0];
155
156 }
157
158 {
159 register mpgdec_real *costab = mpgdec_pnts[2];
160
161 b1[0x00] = b2[0x00] + b2[0x07];
162 b1[0x01] = b2[0x01] + b2[0x06];
163 b1[0x02] = b2[0x02] + b2[0x05];
164 b1[0x03] = b2[0x03] + b2[0x04];
165 b1[0x04] = (b2[0x03] - b2[0x04]) * costab[3];
166 b1[0x05] = (b2[0x02] - b2[0x05]) * costab[2];
167 b1[0x06] = (b2[0x01] - b2[0x06]) * costab[1];
168 b1[0x07] = (b2[0x00] - b2[0x07]) * costab[0];
169 b1[0x08] = b2[0x08] + b2[0x0F];
170 b1[0x09] = b2[0x09] + b2[0x0E];
171 b1[0x0A] = b2[0x0A] + b2[0x0D];
172 b1[0x0B] = b2[0x0B] + b2[0x0C];
173 b1[0x0C] = (b2[0x0C] - b2[0x0B]) * costab[3];
174 b1[0x0D] = (b2[0x0D] - b2[0x0A]) * costab[2];
175 b1[0x0E] = (b2[0x0E] - b2[0x09]) * costab[1];
176 b1[0x0F] = (b2[0x0F] - b2[0x08]) * costab[0];
177 b1[0x10] = b2[0x10] + b2[0x17];
178 b1[0x11] = b2[0x11] + b2[0x16];
179 b1[0x12] = b2[0x12] + b2[0x15];
180 b1[0x13] = b2[0x13] + b2[0x14];
181 b1[0x14] = (b2[0x13] - b2[0x14]) * costab[3];
182 b1[0x15] = (b2[0x12] - b2[0x15]) * costab[2];
183 b1[0x16] = (b2[0x11] - b2[0x16]) * costab[1];
184 b1[0x17] = (b2[0x10] - b2[0x17]) * costab[0];
185 b1[0x18] = b2[0x18] + b2[0x1F];
186 b1[0x19] = b2[0x19] + b2[0x1E];
187 b1[0x1A] = b2[0x1A] + b2[0x1D];
188 b1[0x1B] = b2[0x1B] + b2[0x1C];
189 b1[0x1C] = (b2[0x1C] - b2[0x1B]) * costab[3];
190 b1[0x1D] = (b2[0x1D] - b2[0x1A]) * costab[2];
191 b1[0x1E] = (b2[0x1E] - b2[0x19]) * costab[1];
192 b1[0x1F] = (b2[0x1F] - b2[0x18]) * costab[0];
193 }
194
195 #else /* ALTIVEC_USE_REFERENCE_C_CODE */
196
197 // How does it work ?
198 // the first three passes are reproducted in the three block below
199 // all computations are done on a 4 elements vector
200 // 'reverse' is a special perumtation vector used to reverse
201 // the order of the elements inside a vector.
202 // note that all loads/stores to b1 (b2) between passes 1 and 2 (2 and 3)
203 // have been removed, all elements are stored inside b1vX (b2vX)
204 {
205 register vector float
206 b1v0, b1v1, b1v2, b1v3,
207 b1v4, b1v5, b1v6, b1v7;
208 register vector float
209 temp1, temp2;
210
211 {
212 register mpgdec_real *costab = mpgdec_pnts[0];
213
214 register vector float
215 samplesv1, samplesv2, samplesv3, samplesv4,
216 samplesv5, samplesv6, samplesv7, samplesv8,
217 samplesv9;
218 register vector unsigned char samples_perm = vec_lvsl(0, samples);
219 register vector float costabv1, costabv2, costabv3, costabv4, costabv5;
220 register vector unsigned char costab_perm = vec_lvsl(0, costab);
221
222 samplesv1 = vec_ld(0, samples);
223 samplesv2 = vec_ld(16, samples);
224 samplesv1 = vec_perm(samplesv1, samplesv2, samples_perm);
225 samplesv3 = vec_ld(32, samples);
226 samplesv2 = vec_perm(samplesv2, samplesv3, samples_perm);
227 samplesv4 = vec_ld(48, samples);
228 samplesv3 = vec_perm(samplesv3, samplesv4, samples_perm);
229 samplesv5 = vec_ld(64, samples);
230 samplesv4 = vec_perm(samplesv4, samplesv5, samples_perm);
231 samplesv6 = vec_ld(80, samples);
232 samplesv5 = vec_perm(samplesv5, samplesv6, samples_perm);
233 samplesv7 = vec_ld(96, samples);
234 samplesv6 = vec_perm(samplesv6, samplesv7, samples_perm);
235 samplesv8 = vec_ld(112, samples);
236 samplesv7 = vec_perm(samplesv7, samplesv8, samples_perm);
237 samplesv9 = vec_ld(128, samples);
238 samplesv8 = vec_perm(samplesv8, samplesv9, samples_perm);
239
240 temp1 = vec_add(samplesv1,
241 vec_perm(samplesv8, samplesv8, reverse));
242 //vec_st(temp1, 0, b1);
243 b1v0 = temp1;
244 temp1 = vec_add(samplesv2,
245 vec_perm(samplesv7, samplesv7, reverse));
246 //vec_st(temp1, 16, b1);
247 b1v1 = temp1;
248 temp1 = vec_add(samplesv3,
249 vec_perm(samplesv6, samplesv6, reverse));
250 //vec_st(temp1, 32, b1);
251 b1v2 = temp1;
252 temp1 = vec_add(samplesv4,
253 vec_perm(samplesv5, samplesv5, reverse));
254 //vec_st(temp1, 48, b1);
255 b1v3 = temp1;
256
257 costabv1 = vec_ld(0, costab);
258 costabv2 = vec_ld(16, costab);
259 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
260 costabv3 = vec_ld(32, costab);
261 costabv2 = vec_perm(costabv2, costabv3, costab_perm);
262 costabv4 = vec_ld(48, costab);
263 costabv3 = vec_perm(costabv3, costabv4, costab_perm);
264 costabv5 = vec_ld(64, costab);
265 costabv4 = vec_perm(costabv4, costabv5, costab_perm);
266
267 temp1 = vec_sub(vec_perm(samplesv4, samplesv4, reverse),
268 samplesv5);
269 temp2 = vec_madd(temp1,
270 vec_perm(costabv4, costabv4, reverse),
271 vczero);
272 //vec_st(temp2, 64, b1);
273 b1v4 = temp2;
274
275 temp1 = vec_sub(vec_perm(samplesv3, samplesv3, reverse),
276 samplesv6);
277 temp2 = vec_madd(temp1,
278 vec_perm(costabv3, costabv3, reverse),
279 vczero);
280 //vec_st(temp2, 80, b1);
281 b1v5 = temp2;
282 temp1 = vec_sub(vec_perm(samplesv2, samplesv2, reverse),
283 samplesv7);
284 temp2 = vec_madd(temp1,
285 vec_perm(costabv2, costabv2, reverse),
286 vczero);
287 //vec_st(temp2, 96, b1);
288 b1v6 = temp2;
289
290 temp1 = vec_sub(vec_perm(samplesv1, samplesv1, reverse),
291 samplesv8);
292 temp2 = vec_madd(temp1,
293 vec_perm(costabv1, costabv1, reverse),
294 vczero);
295 //vec_st(temp2, 112, b1);
296 b1v7 = temp2;
297
298 }
299
300 {
301 register vector float
302 b2v0, b2v1, b2v2, b2v3,
303 b2v4, b2v5, b2v6, b2v7;
304 {
305 register mpgdec_real *costab = mpgdec_pnts[1];
306 register vector float costabv1r, costabv2r, costabv1, costabv2, costabv3;
307 register vector unsigned char costab_perm = vec_lvsl(0, costab);
308
309 costabv1 = vec_ld(0, costab);
310 costabv2 = vec_ld(16, costab);
311 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
312 costabv3 = vec_ld(32, costab);
313 costabv2 = vec_perm(costabv2, costabv3 , costab_perm);
314 costabv1r = vec_perm(costabv1, costabv1, reverse);
315 costabv2r = vec_perm(costabv2, costabv2, reverse);
316
317 temp1 = vec_add(b1v0, vec_perm(b1v3, b1v3, reverse));
318 //vec_st(temp1, 0, b2);
319 b2v0 = temp1;
320 temp1 = vec_add(b1v1, vec_perm(b1v2, b1v2, reverse));
321 //vec_st(temp1, 16, b2);
322 b2v1 = temp1;
323 temp2 = vec_sub(vec_perm(b1v1, b1v1, reverse), b1v2);
324 temp1 = vec_madd(temp2, costabv2r, vczero);
325 //vec_st(temp1, 32, b2);
326 b2v2 = temp1;
327 temp2 = vec_sub(vec_perm(b1v0, b1v0, reverse), b1v3);
328 temp1 = vec_madd(temp2, costabv1r, vczero);
329 //vec_st(temp1, 48, b2);
330 b2v3 = temp1;
331 temp1 = vec_add(b1v4, vec_perm(b1v7, b1v7, reverse));
332 //vec_st(temp1, 64, b2);
333 b2v4 = temp1;
334 temp1 = vec_add(b1v5, vec_perm(b1v6, b1v6, reverse));
335 //vec_st(temp1, 80, b2);
336 b2v5 = temp1;
337 temp2 = vec_sub(b1v6, vec_perm(b1v5, b1v5, reverse));
338 temp1 = vec_madd(temp2, costabv2r, vczero);
339 //vec_st(temp1, 96, b2);
340 b2v6 = temp1;
341 temp2 = vec_sub(b1v7, vec_perm(b1v4, b1v4, reverse));
342 temp1 = vec_madd(temp2, costabv1r, vczero);
343 //vec_st(temp1, 112, b2);
344 b2v7 = temp1;
345 }
346
347 {
348 register mpgdec_real *costab = mpgdec_pnts[2];
349
350
351 vector float costabv1r, costabv1, costabv2;
352 vector unsigned char costab_perm = vec_lvsl(0, costab);
353
354 costabv1 = vec_ld(0, costab);
355 costabv2 = vec_ld(16, costab);
356 costabv1 = vec_perm(costabv1, costabv2, costab_perm);
357 costabv1r = vec_perm(costabv1, costabv1, reverse);
358
359 temp1 = vec_add(b2v0, vec_perm(b2v1, b2v1, reverse));
360 vec_st(temp1, 0, b1);
361 temp2 = vec_sub(vec_perm(b2v0, b2v0, reverse), b2v1);
362 temp1 = vec_madd(temp2, costabv1r, vczero);
363 vec_st(temp1, 16, b1);
364
365 temp1 = vec_add(b2v2, vec_perm(b2v3, b2v3, reverse));
366 vec_st(temp1, 32, b1);
367 temp2 = vec_sub(b2v3, vec_perm(b2v2, b2v2, reverse));
368 temp1 = vec_madd(temp2, costabv1r, vczero);
369 vec_st(temp1, 48, b1);
370
371 temp1 = vec_add(b2v4, vec_perm(b2v5, b2v5, reverse));
372 vec_st(temp1, 64, b1);
373 temp2 = vec_sub(vec_perm(b2v4, b2v4, reverse), b2v5);
374 temp1 = vec_madd(temp2, costabv1r, vczero);
375 vec_st(temp1, 80, b1);
376
377 temp1 = vec_add(b2v6, vec_perm(b2v7, b2v7, reverse));
378 vec_st(temp1, 96, b1);
379 temp2 = vec_sub(b2v7, vec_perm(b2v6, b2v6, reverse));
380 temp1 = vec_madd(temp2, costabv1r, vczero);
381 vec_st(temp1, 112, b1);
382
383 }
384 }
385 }
386
387 #endif /* ALTIVEC_USE_REFERENCE_C_CODE */
388
389 {
390 register mpgdec_real const cos0 = mpgdec_pnts[3][0];
391 register mpgdec_real const cos1 = mpgdec_pnts[3][1];
392
393 b2[0x00] = b1[0x00] + b1[0x03];
394 b2[0x01] = b1[0x01] + b1[0x02];
395 b2[0x02] = (b1[0x01] - b1[0x02]) * cos1;
396 b2[0x03] = (b1[0x00] - b1[0x03]) * cos0;
397 b2[0x04] = b1[0x04] + b1[0x07];
398 b2[0x05] = b1[0x05] + b1[0x06];
399 b2[0x06] = (b1[0x06] - b1[0x05]) * cos1;
400 b2[0x07] = (b1[0x07] - b1[0x04]) * cos0;
401 b2[0x08] = b1[0x08] + b1[0x0B];
402 b2[0x09] = b1[0x09] + b1[0x0A];
403 b2[0x0A] = (b1[0x09] - b1[0x0A]) * cos1;
404 b2[0x0B] = (b1[0x08] - b1[0x0B]) * cos0;
405 b2[0x0C] = b1[0x0C] + b1[0x0F];
406 b2[0x0D] = b1[0x0D] + b1[0x0E];
407 b2[0x0E] = (b1[0x0E] - b1[0x0D]) * cos1;
408 b2[0x0F] = (b1[0x0F] - b1[0x0C]) * cos0;
409 b2[0x10] = b1[0x10] + b1[0x13];
410 b2[0x11] = b1[0x11] + b1[0x12];
411 b2[0x12] = (b1[0x11] - b1[0x12]) * cos1;
412 b2[0x13] = (b1[0x10] - b1[0x13]) * cos0;
413 b2[0x14] = b1[0x14] + b1[0x17];
414 b2[0x15] = b1[0x15] + b1[0x16];
415 b2[0x16] = (b1[0x16] - b1[0x15]) * cos1;
416 b2[0x17] = (b1[0x17] - b1[0x14]) * cos0;
417 b2[0x18] = b1[0x18] + b1[0x1B];
418 b2[0x19] = b1[0x19] + b1[0x1A];
419 b2[0x1A] = (b1[0x19] - b1[0x1A]) * cos1;
420 b2[0x1B] = (b1[0x18] - b1[0x1B]) * cos0;
421 b2[0x1C] = b1[0x1C] + b1[0x1F];
422 b2[0x1D] = b1[0x1D] + b1[0x1E];
423 b2[0x1E] = (b1[0x1E] - b1[0x1D]) * cos1;
424 b2[0x1F] = (b1[0x1F] - b1[0x1C]) * cos0;
425 }
426
427 {
428 register mpgdec_real const cos0 = mpgdec_pnts[4][0];
429
430 b1[0x00] = b2[0x00] + b2[0x01];
431 b1[0x01] = (b2[0x00] - b2[0x01]) * cos0;
432 b1[0x02] = b2[0x02] + b2[0x03];
433 b1[0x03] = (b2[0x03] - b2[0x02]) * cos0;
434 b1[0x02] += b1[0x03];
435
436 b1[0x04] = b2[0x04] + b2[0x05];
437 b1[0x05] = (b2[0x04] - b2[0x05]) * cos0;
438 b1[0x06] = b2[0x06] + b2[0x07];
439 b1[0x07] = (b2[0x07] - b2[0x06]) * cos0;
440 b1[0x06] += b1[0x07];
441 b1[0x04] += b1[0x06];
442 b1[0x06] += b1[0x05];
443 b1[0x05] += b1[0x07];
444
445 b1[0x08] = b2[0x08] + b2[0x09];
446 b1[0x09] = (b2[0x08] - b2[0x09]) * cos0;
447 b1[0x0A] = b2[0x0A] + b2[0x0B];
448 b1[0x0B] = (b2[0x0B] - b2[0x0A]) * cos0;
449 b1[0x0A] += b1[0x0B];
450
451 b1[0x0C] = b2[0x0C] + b2[0x0D];
452 b1[0x0D] = (b2[0x0C] - b2[0x0D]) * cos0;
453 b1[0x0E] = b2[0x0E] + b2[0x0F];
454 b1[0x0F] = (b2[0x0F] - b2[0x0E]) * cos0;
455 b1[0x0E] += b1[0x0F];
456 b1[0x0C] += b1[0x0E];
457 b1[0x0E] += b1[0x0D];
458 b1[0x0D] += b1[0x0F];
459
460 b1[0x10] = b2[0x10] + b2[0x11];
461 b1[0x11] = (b2[0x10] - b2[0x11]) * cos0;
462 b1[0x12] = b2[0x12] + b2[0x13];
463 b1[0x13] = (b2[0x13] - b2[0x12]) * cos0;
464 b1[0x12] += b1[0x13];
465
466 b1[0x14] = b2[0x14] + b2[0x15];
467 b1[0x15] = (b2[0x14] - b2[0x15]) * cos0;
468 b1[0x16] = b2[0x16] + b2[0x17];
469 b1[0x17] = (b2[0x17] - b2[0x16]) * cos0;
470 b1[0x16] += b1[0x17];
471 b1[0x14] += b1[0x16];
472 b1[0x16] += b1[0x15];
473 b1[0x15] += b1[0x17];
474
475 b1[0x18] = b2[0x18] + b2[0x19];
476 b1[0x19] = (b2[0x18] - b2[0x19]) * cos0;
477 b1[0x1A] = b2[0x1A] + b2[0x1B];
478 b1[0x1B] = (b2[0x1B] - b2[0x1A]) * cos0;
479 b1[0x1A] += b1[0x1B];
480
481 b1[0x1C] = b2[0x1C] + b2[0x1D];
482 b1[0x1D] = (b2[0x1C] - b2[0x1D]) * cos0;
483 b1[0x1E] = b2[0x1E] + b2[0x1F];
484 b1[0x1F] = (b2[0x1F] - b2[0x1E]) * cos0;
485 b1[0x1E] += b1[0x1F];
486 b1[0x1C] += b1[0x1E];
487 b1[0x1E] += b1[0x1D];
488 b1[0x1D] += b1[0x1F];
489 }
490
491 out0[0x10*16] = b1[0x00];
492 out0[0x10*12] = b1[0x04];
493 out0[0x10* 8] = b1[0x02];
494 out0[0x10* 4] = b1[0x06];
495 out0[0x10* 0] = b1[0x01];
496 out1[0x10* 0] = b1[0x01];
497 out1[0x10* 4] = b1[0x05];
498 out1[0x10* 8] = b1[0x03];
499 out1[0x10*12] = b1[0x07];
500
501 b1[0x08] += b1[0x0C];
502 out0[0x10*14] = b1[0x08];
503 b1[0x0C] += b1[0x0a];
504 out0[0x10*10] = b1[0x0C];
505 b1[0x0A] += b1[0x0E];
506 out0[0x10* 6] = b1[0x0A];
507 b1[0x0E] += b1[0x09];
508 out0[0x10* 2] = b1[0x0E];
509 b1[0x09] += b1[0x0D];
510 out1[0x10* 2] = b1[0x09];
511 b1[0x0D] += b1[0x0B];
512 out1[0x10* 6] = b1[0x0D];
513 b1[0x0B] += b1[0x0F];
514 out1[0x10*10] = b1[0x0B];
515 out1[0x10*14] = b1[0x0F];
516
517 b1[0x18] += b1[0x1C];
518 out0[0x10*15] = b1[0x10] + b1[0x18];
519 out0[0x10*13] = b1[0x18] + b1[0x14];
520 b1[0x1C] += b1[0x1a];
521 out0[0x10*11] = b1[0x14] + b1[0x1C];
522 out0[0x10* 9] = b1[0x1C] + b1[0x12];
523 b1[0x1A] += b1[0x1E];
524 out0[0x10* 7] = b1[0x12] + b1[0x1A];
525 out0[0x10* 5] = b1[0x1A] + b1[0x16];
526 b1[0x1E] += b1[0x19];
527 out0[0x10* 3] = b1[0x16] + b1[0x1E];
528 out0[0x10* 1] = b1[0x1E] + b1[0x11];
529 b1[0x19] += b1[0x1D];
530 out1[0x10* 1] = b1[0x11] + b1[0x19];
531 out1[0x10* 3] = b1[0x19] + b1[0x15];
532 b1[0x1D] += b1[0x1B];
533 out1[0x10* 5] = b1[0x15] + b1[0x1D];
534 out1[0x10* 7] = b1[0x1D] + b1[0x13];
535 b1[0x1B] += b1[0x1F];
536 out1[0x10* 9] = b1[0x13] + b1[0x1B];
537 out1[0x10*11] = b1[0x1B] + b1[0x17];
538 out1[0x10*13] = b1[0x17] + b1[0x1F];
539 out1[0x10*15] = b1[0x1F];
540 }
541
542 #endif /* HAVE_ALTIVEC */
543