Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/n1_16.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
1 /*
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
3 * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
4 *
5 * This program 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 of the License, or
8 * (at your option) any later version.
9 *
10 * This program 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 this program; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 */
20
21 /* This file was automatically generated --- DO NOT EDIT */
22 /* Generated on Sun Nov 25 07:35:44 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 16 -name n1_16 -include n.h */
29
30 /*
31 * This function contains 144 FP additions, 40 FP multiplications,
32 * (or, 104 additions, 0 multiplications, 40 fused multiply/add),
33 * 82 stack variables, 3 constants, and 64 memory accesses
34 */
35 #include "n.h"
36
37 static void n1_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT i;
44 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
45 E T1z, T1L, T1M, T1N, T1P, T1J, T1K, T1G, T1O, T1Q;
46 {
47 E T1l, T1H, T1R, T7, T1x, TN, TC, T25, T1E, T1b, T1Z, Tt, T2h, T22, T1D;
48 E T1g, T1n, TQ, Te, T26, TT, T1m, TJ, T1S, Tj, T11, Ti, T1V, TZ, Tk;
49 E T12, T13;
50 {
51 E Tq, T1c, Tp, T20, T1a, Tr, T1d, T1e;
52 {
53 E T4, TL, T3, T1k, Ty, T5, Tz, TA;
54 {
55 E T1, T2, Tw, Tx;
56 T1 = ri[0];
57 T2 = ri[WS(is, 8)];
58 Tw = ii[0];
59 Tx = ii[WS(is, 8)];
60 T4 = ri[WS(is, 4)];
61 TL = T1 - T2;
62 T3 = T1 + T2;
63 T1k = Tw - Tx;
64 Ty = Tw + Tx;
65 T5 = ri[WS(is, 12)];
66 Tz = ii[WS(is, 4)];
67 TA = ii[WS(is, 12)];
68 }
69 {
70 E Tn, To, T18, T19;
71 Tn = ri[WS(is, 15)];
72 {
73 E T1j, T6, TM, TB;
74 T1j = T4 - T5;
75 T6 = T4 + T5;
76 TM = Tz - TA;
77 TB = Tz + TA;
78 T1l = T1j + T1k;
79 T1H = T1k - T1j;
80 T1R = T3 - T6;
81 T7 = T3 + T6;
82 T1x = TL + TM;
83 TN = TL - TM;
84 TC = Ty + TB;
85 T25 = Ty - TB;
86 To = ri[WS(is, 7)];
87 }
88 T18 = ii[WS(is, 15)];
89 T19 = ii[WS(is, 7)];
90 Tq = ri[WS(is, 3)];
91 T1c = Tn - To;
92 Tp = Tn + To;
93 T20 = T18 + T19;
94 T1a = T18 - T19;
95 Tr = ri[WS(is, 11)];
96 T1d = ii[WS(is, 3)];
97 T1e = ii[WS(is, 11)];
98 }
99 }
100 {
101 E Tb, TP, Ta, TO, TF, Tc, TG, TH;
102 {
103 E T8, T9, TD, TE;
104 T8 = ri[WS(is, 2)];
105 {
106 E T17, Ts, T21, T1f;
107 T17 = Tq - Tr;
108 Ts = Tq + Tr;
109 T21 = T1d + T1e;
110 T1f = T1d - T1e;
111 T1E = T1a - T17;
112 T1b = T17 + T1a;
113 T1Z = Tp - Ts;
114 Tt = Tp + Ts;
115 T2h = T20 + T21;
116 T22 = T20 - T21;
117 T1D = T1c + T1f;
118 T1g = T1c - T1f;
119 T9 = ri[WS(is, 10)];
120 }
121 TD = ii[WS(is, 2)];
122 TE = ii[WS(is, 10)];
123 Tb = ri[WS(is, 14)];
124 TP = T8 - T9;
125 Ta = T8 + T9;
126 TO = TD - TE;
127 TF = TD + TE;
128 Tc = ri[WS(is, 6)];
129 TG = ii[WS(is, 14)];
130 TH = ii[WS(is, 6)];
131 }
132 {
133 E TR, Td, TS, TI;
134 T1n = TP + TO;
135 TQ = TO - TP;
136 TR = Tb - Tc;
137 Td = Tb + Tc;
138 TS = TG - TH;
139 TI = TG + TH;
140 Te = Ta + Td;
141 T26 = Td - Ta;
142 TT = TR + TS;
143 T1m = TR - TS;
144 TJ = TF + TI;
145 T1S = TF - TI;
146 }
147 }
148 {
149 E Tg, Th, TX, TY;
150 Tg = ri[WS(is, 1)];
151 Th = ri[WS(is, 9)];
152 TX = ii[WS(is, 1)];
153 TY = ii[WS(is, 9)];
154 Tj = ri[WS(is, 5)];
155 T11 = Tg - Th;
156 Ti = Tg + Th;
157 T1V = TX + TY;
158 TZ = TX - TY;
159 Tk = ri[WS(is, 13)];
160 T12 = ii[WS(is, 5)];
161 T13 = ii[WS(is, 13)];
162 }
163 }
164 {
165 E T2f, T1B, T10, T1U, T1X, T1A, T15, Tv, TK, T2i;
166 {
167 E Tf, Tu, T2j, T2k, T2g;
168 T2f = T7 - Te;
169 Tf = T7 + Te;
170 {
171 E TW, Tl, T1W, T14, Tm;
172 TW = Tj - Tk;
173 Tl = Tj + Tk;
174 T1W = T12 + T13;
175 T14 = T12 - T13;
176 T1B = TZ - TW;
177 T10 = TW + TZ;
178 T1U = Ti - Tl;
179 Tm = Ti + Tl;
180 T2g = T1V + T1W;
181 T1X = T1V - T1W;
182 T1A = T11 + T14;
183 T15 = T11 - T14;
184 Tu = Tm + Tt;
185 Tv = Tt - Tm;
186 }
187 TK = TC - TJ;
188 T2j = TC + TJ;
189 T2k = T2g + T2h;
190 T2i = T2g - T2h;
191 ro[0] = Tf + Tu;
192 ro[WS(os, 8)] = Tf - Tu;
193 io[0] = T2j + T2k;
194 io[WS(os, 8)] = T2j - T2k;
195 }
196 {
197 E T29, T1T, T27, T2d, T2a, T2b, T28, T24, T1Y, T23;
198 T29 = T1R - T1S;
199 T1T = T1R + T1S;
200 io[WS(os, 12)] = TK - Tv;
201 io[WS(os, 4)] = Tv + TK;
202 ro[WS(os, 4)] = T2f + T2i;
203 ro[WS(os, 12)] = T2f - T2i;
204 T27 = T25 - T26;
205 T2d = T26 + T25;
206 T2a = T1X - T1U;
207 T1Y = T1U + T1X;
208 T23 = T1Z - T22;
209 T2b = T1Z + T22;
210 T28 = T23 - T1Y;
211 T24 = T1Y + T23;
212 {
213 E T1I, TV, T1v, T1y, T1t, T1s, T1r, T1p, T1q, T1i;
214 {
215 E T1o, T2e, T2c, TU, T16, T1h;
216 T1I = TQ + TT;
217 TU = TQ - TT;
218 io[WS(os, 14)] = FNMS(KP707106781, T28, T27);
219 io[WS(os, 6)] = FMA(KP707106781, T28, T27);
220 ro[WS(os, 2)] = FMA(KP707106781, T24, T1T);
221 ro[WS(os, 10)] = FNMS(KP707106781, T24, T1T);
222 T2e = T2a + T2b;
223 T2c = T2a - T2b;
224 TV = FMA(KP707106781, TU, TN);
225 T1v = FNMS(KP707106781, TU, TN);
226 io[WS(os, 10)] = FNMS(KP707106781, T2e, T2d);
227 io[WS(os, 2)] = FMA(KP707106781, T2e, T2d);
228 ro[WS(os, 6)] = FMA(KP707106781, T2c, T29);
229 ro[WS(os, 14)] = FNMS(KP707106781, T2c, T29);
230 T1o = T1m - T1n;
231 T1y = T1n + T1m;
232 T1t = FNMS(KP414213562, T10, T15);
233 T16 = FMA(KP414213562, T15, T10);
234 T1h = FNMS(KP414213562, T1g, T1b);
235 T1s = FMA(KP414213562, T1b, T1g);
236 T1r = FMA(KP707106781, T1o, T1l);
237 T1p = FNMS(KP707106781, T1o, T1l);
238 T1q = T16 + T1h;
239 T1i = T16 - T1h;
240 }
241 {
242 E T1w, T1u, T1C, T1F;
243 io[WS(os, 15)] = FMA(KP923879532, T1q, T1p);
244 io[WS(os, 7)] = FNMS(KP923879532, T1q, T1p);
245 ro[WS(os, 3)] = FMA(KP923879532, T1i, TV);
246 ro[WS(os, 11)] = FNMS(KP923879532, T1i, TV);
247 T1w = T1t + T1s;
248 T1u = T1s - T1t;
249 T1z = FMA(KP707106781, T1y, T1x);
250 T1L = FNMS(KP707106781, T1y, T1x);
251 ro[WS(os, 15)] = FMA(KP923879532, T1w, T1v);
252 ro[WS(os, 7)] = FNMS(KP923879532, T1w, T1v);
253 io[WS(os, 3)] = FMA(KP923879532, T1u, T1r);
254 io[WS(os, 11)] = FNMS(KP923879532, T1u, T1r);
255 T1M = FNMS(KP414213562, T1A, T1B);
256 T1C = FMA(KP414213562, T1B, T1A);
257 T1F = FNMS(KP414213562, T1E, T1D);
258 T1N = FMA(KP414213562, T1D, T1E);
259 T1P = FMA(KP707106781, T1I, T1H);
260 T1J = FNMS(KP707106781, T1I, T1H);
261 T1K = T1F - T1C;
262 T1G = T1C + T1F;
263 }
264 }
265 }
266 }
267 }
268 io[WS(os, 5)] = FMA(KP923879532, T1K, T1J);
269 io[WS(os, 13)] = FNMS(KP923879532, T1K, T1J);
270 ro[WS(os, 1)] = FMA(KP923879532, T1G, T1z);
271 ro[WS(os, 9)] = FNMS(KP923879532, T1G, T1z);
272 T1O = T1M - T1N;
273 T1Q = T1M + T1N;
274 io[WS(os, 1)] = FMA(KP923879532, T1Q, T1P);
275 io[WS(os, 9)] = FNMS(KP923879532, T1Q, T1P);
276 ro[WS(os, 5)] = FMA(KP923879532, T1O, T1L);
277 ro[WS(os, 13)] = FNMS(KP923879532, T1O, T1L);
278 }
279 }
280 }
281
282 static const kdft_desc desc = { 16, "n1_16", {104, 0, 40, 0}, &GENUS, 0, 0, 0, 0 };
283
284 void X(codelet_n1_16) (planner *p) {
285 X(kdft_register) (p, n1_16, &desc);
286 }
287
288 #else /* HAVE_FMA */
289
290 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 16 -name n1_16 -include n.h */
291
292 /*
293 * This function contains 144 FP additions, 24 FP multiplications,
294 * (or, 136 additions, 16 multiplications, 8 fused multiply/add),
295 * 50 stack variables, 3 constants, and 64 memory accesses
296 */
297 #include "n.h"
298
299 static void n1_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
300 {
301 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
302 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
303 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
304 {
305 INT i;
306 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
307 E T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
308 E T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
309 E T1U, T1A;
310 {
311 E T3, TL, Ty, T1k, T6, T1j, TB, TM;
312 {
313 E T1, T2, Tw, Tx;
314 T1 = ri[0];
315 T2 = ri[WS(is, 8)];
316 T3 = T1 + T2;
317 TL = T1 - T2;
318 Tw = ii[0];
319 Tx = ii[WS(is, 8)];
320 Ty = Tw + Tx;
321 T1k = Tw - Tx;
322 }
323 {
324 E T4, T5, Tz, TA;
325 T4 = ri[WS(is, 4)];
326 T5 = ri[WS(is, 12)];
327 T6 = T4 + T5;
328 T1j = T4 - T5;
329 Tz = ii[WS(is, 4)];
330 TA = ii[WS(is, 12)];
331 TB = Tz + TA;
332 TM = Tz - TA;
333 }
334 T7 = T3 + T6;
335 T1R = T3 - T6;
336 T25 = Ty - TB;
337 TC = Ty + TB;
338 TN = TL - TM;
339 T1x = TL + TM;
340 T1H = T1k - T1j;
341 T1l = T1j + T1k;
342 }
343 {
344 E Tp, T17, T1f, T20, Ts, T1c, T1a, T21;
345 {
346 E Tn, To, T1d, T1e;
347 Tn = ri[WS(is, 15)];
348 To = ri[WS(is, 7)];
349 Tp = Tn + To;
350 T17 = Tn - To;
351 T1d = ii[WS(is, 15)];
352 T1e = ii[WS(is, 7)];
353 T1f = T1d - T1e;
354 T20 = T1d + T1e;
355 }
356 {
357 E Tq, Tr, T18, T19;
358 Tq = ri[WS(is, 3)];
359 Tr = ri[WS(is, 11)];
360 Ts = Tq + Tr;
361 T1c = Tq - Tr;
362 T18 = ii[WS(is, 3)];
363 T19 = ii[WS(is, 11)];
364 T1a = T18 - T19;
365 T21 = T18 + T19;
366 }
367 Tt = Tp + Ts;
368 T22 = T20 - T21;
369 T2h = T20 + T21;
370 T1b = T17 - T1a;
371 T1g = T1c + T1f;
372 T1E = T1f - T1c;
373 T1Z = Tp - Ts;
374 T1D = T17 + T1a;
375 }
376 {
377 E Ta, TP, TF, TO, Td, TR, TI, TS;
378 {
379 E T8, T9, TD, TE;
380 T8 = ri[WS(is, 2)];
381 T9 = ri[WS(is, 10)];
382 Ta = T8 + T9;
383 TP = T8 - T9;
384 TD = ii[WS(is, 2)];
385 TE = ii[WS(is, 10)];
386 TF = TD + TE;
387 TO = TD - TE;
388 }
389 {
390 E Tb, Tc, TG, TH;
391 Tb = ri[WS(is, 14)];
392 Tc = ri[WS(is, 6)];
393 Td = Tb + Tc;
394 TR = Tb - Tc;
395 TG = ii[WS(is, 14)];
396 TH = ii[WS(is, 6)];
397 TI = TG + TH;
398 TS = TG - TH;
399 }
400 Te = Ta + Td;
401 T1S = TF - TI;
402 T26 = Td - Ta;
403 TJ = TF + TI;
404 TQ = TO - TP;
405 T1m = TR - TS;
406 T1n = TP + TO;
407 TT = TR + TS;
408 }
409 {
410 E Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
411 {
412 E Tg, Th, TX, TY;
413 Tg = ri[WS(is, 1)];
414 Th = ri[WS(is, 9)];
415 Ti = Tg + Th;
416 T11 = Tg - Th;
417 TX = ii[WS(is, 1)];
418 TY = ii[WS(is, 9)];
419 TZ = TX - TY;
420 T1V = TX + TY;
421 }
422 {
423 E Tj, Tk, T12, T13;
424 Tj = ri[WS(is, 5)];
425 Tk = ri[WS(is, 13)];
426 Tl = Tj + Tk;
427 TW = Tj - Tk;
428 T12 = ii[WS(is, 5)];
429 T13 = ii[WS(is, 13)];
430 T14 = T12 - T13;
431 T1W = T12 + T13;
432 }
433 Tm = Ti + Tl;
434 T1X = T1V - T1W;
435 T2g = T1V + T1W;
436 T10 = TW + TZ;
437 T15 = T11 - T14;
438 T1B = T11 + T14;
439 T1U = Ti - Tl;
440 T1A = TZ - TW;
441 }
442 {
443 E Tf, Tu, T2j, T2k;
444 Tf = T7 + Te;
445 Tu = Tm + Tt;
446 ro[WS(os, 8)] = Tf - Tu;
447 ro[0] = Tf + Tu;
448 T2j = TC + TJ;
449 T2k = T2g + T2h;
450 io[WS(os, 8)] = T2j - T2k;
451 io[0] = T2j + T2k;
452 }
453 {
454 E Tv, TK, T2f, T2i;
455 Tv = Tt - Tm;
456 TK = TC - TJ;
457 io[WS(os, 4)] = Tv + TK;
458 io[WS(os, 12)] = TK - Tv;
459 T2f = T7 - Te;
460 T2i = T2g - T2h;
461 ro[WS(os, 12)] = T2f - T2i;
462 ro[WS(os, 4)] = T2f + T2i;
463 }
464 {
465 E T1T, T27, T24, T28, T1Y, T23;
466 T1T = T1R + T1S;
467 T27 = T25 - T26;
468 T1Y = T1U + T1X;
469 T23 = T1Z - T22;
470 T24 = KP707106781 * (T1Y + T23);
471 T28 = KP707106781 * (T23 - T1Y);
472 ro[WS(os, 10)] = T1T - T24;
473 io[WS(os, 6)] = T27 + T28;
474 ro[WS(os, 2)] = T1T + T24;
475 io[WS(os, 14)] = T27 - T28;
476 }
477 {
478 E T29, T2d, T2c, T2e, T2a, T2b;
479 T29 = T1R - T1S;
480 T2d = T26 + T25;
481 T2a = T1X - T1U;
482 T2b = T1Z + T22;
483 T2c = KP707106781 * (T2a - T2b);
484 T2e = KP707106781 * (T2a + T2b);
485 ro[WS(os, 14)] = T29 - T2c;
486 io[WS(os, 2)] = T2d + T2e;
487 ro[WS(os, 6)] = T29 + T2c;
488 io[WS(os, 10)] = T2d - T2e;
489 }
490 {
491 E TV, T1r, T1p, T1v, T1i, T1q, T1u, T1w, TU, T1o;
492 TU = KP707106781 * (TQ - TT);
493 TV = TN + TU;
494 T1r = TN - TU;
495 T1o = KP707106781 * (T1m - T1n);
496 T1p = T1l - T1o;
497 T1v = T1l + T1o;
498 {
499 E T16, T1h, T1s, T1t;
500 T16 = FMA(KP923879532, T10, KP382683432 * T15);
501 T1h = FNMS(KP923879532, T1g, KP382683432 * T1b);
502 T1i = T16 + T1h;
503 T1q = T1h - T16;
504 T1s = FNMS(KP923879532, T15, KP382683432 * T10);
505 T1t = FMA(KP382683432, T1g, KP923879532 * T1b);
506 T1u = T1s - T1t;
507 T1w = T1s + T1t;
508 }
509 ro[WS(os, 11)] = TV - T1i;
510 io[WS(os, 11)] = T1v - T1w;
511 ro[WS(os, 3)] = TV + T1i;
512 io[WS(os, 3)] = T1v + T1w;
513 io[WS(os, 15)] = T1p - T1q;
514 ro[WS(os, 15)] = T1r - T1u;
515 io[WS(os, 7)] = T1p + T1q;
516 ro[WS(os, 7)] = T1r + T1u;
517 }
518 {
519 E T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
520 T1y = KP707106781 * (T1n + T1m);
521 T1z = T1x + T1y;
522 T1L = T1x - T1y;
523 T1I = KP707106781 * (TQ + TT);
524 T1J = T1H - T1I;
525 T1P = T1H + T1I;
526 {
527 E T1C, T1F, T1M, T1N;
528 T1C = FMA(KP382683432, T1A, KP923879532 * T1B);
529 T1F = FNMS(KP382683432, T1E, KP923879532 * T1D);
530 T1G = T1C + T1F;
531 T1K = T1F - T1C;
532 T1M = FNMS(KP382683432, T1B, KP923879532 * T1A);
533 T1N = FMA(KP923879532, T1E, KP382683432 * T1D);
534 T1O = T1M - T1N;
535 T1Q = T1M + T1N;
536 }
537 ro[WS(os, 9)] = T1z - T1G;
538 io[WS(os, 9)] = T1P - T1Q;
539 ro[WS(os, 1)] = T1z + T1G;
540 io[WS(os, 1)] = T1P + T1Q;
541 io[WS(os, 13)] = T1J - T1K;
542 ro[WS(os, 13)] = T1L - T1O;
543 io[WS(os, 5)] = T1J + T1K;
544 ro[WS(os, 5)] = T1L + T1O;
545 }
546 }
547 }
548 }
549
550 static const kdft_desc desc = { 16, "n1_16", {136, 16, 8, 0}, &GENUS, 0, 0, 0, 0 };
551
552 void X(codelet_n1_16) (planner *p) {
553 X(kdft_register) (p, n1_16, &desc);
554 }
555
556 #endif /* HAVE_FMA */