Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_12.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
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 07:35:50 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -name t1_12 -include t.h */
29
30 /*
31 * This function contains 118 FP additions, 68 FP multiplications,
32 * (or, 72 additions, 22 multiplications, 46 fused multiply/add),
33 * 84 stack variables, 2 constants, and 48 memory accesses
34 */
35 #include "t.h"
36
37 static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
44 E T2B, T2C;
45 {
46 E T1, T2i, T2e, Tl, T1Y, T10, T1S, TG, T2f, T1s, T2r, Ty, T1Z, T1H, T21;
47 E T1d, TI, TL, T2h, T1l, T2o, Te, TJ, T1w, TO, TR, TN, TK, TQ;
48 {
49 E TW, TZ, TY, T1X, TX;
50 T1 = ri[0];
51 T2i = ii[0];
52 {
53 E Th, Tk, Tg, Tj, T2d, Ti, TV;
54 Th = ri[WS(rs, 6)];
55 Tk = ii[WS(rs, 6)];
56 Tg = W[10];
57 Tj = W[11];
58 TW = ri[WS(rs, 9)];
59 TZ = ii[WS(rs, 9)];
60 T2d = Tg * Tk;
61 Ti = Tg * Th;
62 TV = W[16];
63 TY = W[17];
64 T2e = FNMS(Tj, Th, T2d);
65 Tl = FMA(Tj, Tk, Ti);
66 T1X = TV * TZ;
67 TX = TV * TW;
68 }
69 {
70 E Tn, Tq, Tt, T1o, To, Tw, Ts, Tp, Tv;
71 {
72 E TC, TF, TB, TE, T1R, TD, Tm;
73 TC = ri[WS(rs, 3)];
74 TF = ii[WS(rs, 3)];
75 T1Y = FNMS(TY, TW, T1X);
76 T10 = FMA(TY, TZ, TX);
77 TB = W[4];
78 TE = W[5];
79 Tn = ri[WS(rs, 10)];
80 Tq = ii[WS(rs, 10)];
81 T1R = TB * TF;
82 TD = TB * TC;
83 Tm = W[18];
84 Tt = ri[WS(rs, 2)];
85 T1S = FNMS(TE, TC, T1R);
86 TG = FMA(TE, TF, TD);
87 T1o = Tm * Tq;
88 To = Tm * Tn;
89 Tw = ii[WS(rs, 2)];
90 Ts = W[2];
91 Tp = W[19];
92 Tv = W[3];
93 }
94 {
95 E T12, T15, T13, T1D, T18, T1b, T17, T14, T1a;
96 {
97 E T1p, Tr, T1r, Tx, T1q, Tu, T11;
98 T12 = ri[WS(rs, 1)];
99 T1q = Ts * Tw;
100 Tu = Ts * Tt;
101 T1p = FNMS(Tp, Tn, T1o);
102 Tr = FMA(Tp, Tq, To);
103 T1r = FNMS(Tv, Tt, T1q);
104 Tx = FMA(Tv, Tw, Tu);
105 T15 = ii[WS(rs, 1)];
106 T11 = W[0];
107 T2f = T1p + T1r;
108 T1s = T1p - T1r;
109 T2r = Tx - Tr;
110 Ty = Tr + Tx;
111 T13 = T11 * T12;
112 T1D = T11 * T15;
113 }
114 T18 = ri[WS(rs, 5)];
115 T1b = ii[WS(rs, 5)];
116 T17 = W[8];
117 T14 = W[1];
118 T1a = W[9];
119 {
120 E T3, T6, T4, T1h, T9, Tc, T8, T5, Tb;
121 {
122 E T1E, T16, T1G, T1c, T1F, T19, T2;
123 T3 = ri[WS(rs, 4)];
124 T1F = T17 * T1b;
125 T19 = T17 * T18;
126 T1E = FNMS(T14, T12, T1D);
127 T16 = FMA(T14, T15, T13);
128 T1G = FNMS(T1a, T18, T1F);
129 T1c = FMA(T1a, T1b, T19);
130 T6 = ii[WS(rs, 4)];
131 T2 = W[6];
132 T1Z = T1E + T1G;
133 T1H = T1E - T1G;
134 T21 = T1c - T16;
135 T1d = T16 + T1c;
136 T4 = T2 * T3;
137 T1h = T2 * T6;
138 }
139 T9 = ri[WS(rs, 8)];
140 Tc = ii[WS(rs, 8)];
141 T8 = W[14];
142 T5 = W[7];
143 Tb = W[15];
144 {
145 E T1i, T7, T1k, Td, T1j, Ta, TH;
146 TI = ri[WS(rs, 7)];
147 T1j = T8 * Tc;
148 Ta = T8 * T9;
149 T1i = FNMS(T5, T3, T1h);
150 T7 = FMA(T5, T6, T4);
151 T1k = FNMS(Tb, T9, T1j);
152 Td = FMA(Tb, Tc, Ta);
153 TL = ii[WS(rs, 7)];
154 TH = W[12];
155 T2h = T1i + T1k;
156 T1l = T1i - T1k;
157 T2o = Td - T7;
158 Te = T7 + Td;
159 TJ = TH * TI;
160 T1w = TH * TL;
161 }
162 TO = ri[WS(rs, 11)];
163 TR = ii[WS(rs, 11)];
164 TN = W[20];
165 TK = W[13];
166 TQ = W[21];
167 }
168 }
169 }
170 }
171 {
172 E T1g, T1n, T2q, T1A, T1V, T28, TA, T2n, T1v, T1C, T1U, T29, T2m, T2k, T2l;
173 E T1f, T2a, T20;
174 {
175 E T2g, T1T, TT, T2j, TU, T1e;
176 {
177 E Tf, T1x, TM, T1z, TS, Tz, T1y, TP;
178 T1g = FNMS(KP500000000, Te, T1);
179 Tf = T1 + Te;
180 T1y = TN * TR;
181 TP = TN * TO;
182 T1x = FNMS(TK, TI, T1w);
183 TM = FMA(TK, TL, TJ);
184 T1z = FNMS(TQ, TO, T1y);
185 TS = FMA(TQ, TR, TP);
186 Tz = Tl + Ty;
187 T1n = FNMS(KP500000000, Ty, Tl);
188 T2q = FNMS(KP500000000, T2f, T2e);
189 T2g = T2e + T2f;
190 T1T = T1x + T1z;
191 T1A = T1x - T1z;
192 T1V = TS - TM;
193 TT = TM + TS;
194 T28 = Tf - Tz;
195 TA = Tf + Tz;
196 T2j = T2h + T2i;
197 T2n = FNMS(KP500000000, T2h, T2i);
198 }
199 T1v = FNMS(KP500000000, TT, TG);
200 TU = TG + TT;
201 T1e = T10 + T1d;
202 T1C = FNMS(KP500000000, T1d, T10);
203 T1U = FNMS(KP500000000, T1T, T1S);
204 T29 = T1S + T1T;
205 T2m = T2j - T2g;
206 T2k = T2g + T2j;
207 T2l = TU - T1e;
208 T1f = TU + T1e;
209 T2a = T1Y + T1Z;
210 T20 = FNMS(KP500000000, T1Z, T1Y);
211 }
212 {
213 E T1m, T1K, T2y, T2p, T2x, T2s, T1L, T1t, T1B, T1N, T2c, T2b;
214 ii[WS(rs, 9)] = T2m - T2l;
215 ii[WS(rs, 3)] = T2l + T2m;
216 ri[0] = TA + T1f;
217 ri[WS(rs, 6)] = TA - T1f;
218 T2c = T29 + T2a;
219 T2b = T29 - T2a;
220 T1m = FNMS(KP866025403, T1l, T1g);
221 T1K = FMA(KP866025403, T1l, T1g);
222 ii[0] = T2c + T2k;
223 ii[WS(rs, 6)] = T2k - T2c;
224 ri[WS(rs, 9)] = T28 + T2b;
225 ri[WS(rs, 3)] = T28 - T2b;
226 T2y = FNMS(KP866025403, T2o, T2n);
227 T2p = FMA(KP866025403, T2o, T2n);
228 T2x = FNMS(KP866025403, T2r, T2q);
229 T2s = FMA(KP866025403, T2r, T2q);
230 T1L = FMA(KP866025403, T1s, T1n);
231 T1t = FNMS(KP866025403, T1s, T1n);
232 T1B = FNMS(KP866025403, T1A, T1v);
233 T1N = FMA(KP866025403, T1A, T1v);
234 {
235 E T24, T27, T1Q, T2u, T23, T2v, T2w, T2t;
236 {
237 E T1u, T1W, T22, T1O, T1I, T2z, T2A, T25, T26, T1M, T1J, T1P;
238 T24 = T1m - T1t;
239 T1u = T1m + T1t;
240 T25 = FNMS(KP866025403, T1V, T1U);
241 T1W = FMA(KP866025403, T1V, T1U);
242 T26 = FNMS(KP866025403, T21, T20);
243 T22 = FMA(KP866025403, T21, T20);
244 T1O = FMA(KP866025403, T1H, T1C);
245 T1I = FNMS(KP866025403, T1H, T1C);
246 T2z = T2x + T2y;
247 T2B = T2y - T2x;
248 T27 = T25 - T26;
249 T2A = T25 + T26;
250 T1M = T1K + T1L;
251 T1Q = T1K - T1L;
252 T2C = T1B - T1I;
253 T1J = T1B + T1I;
254 T1P = T1N + T1O;
255 T2u = T1N - T1O;
256 ii[WS(rs, 8)] = T2A + T2z;
257 ii[WS(rs, 2)] = T2z - T2A;
258 ri[WS(rs, 8)] = T1u + T1J;
259 ri[WS(rs, 2)] = T1u - T1J;
260 ri[WS(rs, 10)] = T1M - T1P;
261 ri[WS(rs, 4)] = T1M + T1P;
262 T23 = T1W - T22;
263 T2v = T1W + T22;
264 T2w = T2s + T2p;
265 T2t = T2p - T2s;
266 }
267 ii[WS(rs, 10)] = T2w - T2v;
268 ii[WS(rs, 4)] = T2v + T2w;
269 ri[WS(rs, 1)] = T1Q + T23;
270 ri[WS(rs, 7)] = T1Q - T23;
271 ii[WS(rs, 7)] = T2u + T2t;
272 ii[WS(rs, 1)] = T2t - T2u;
273 ri[WS(rs, 5)] = T24 + T27;
274 ri[WS(rs, 11)] = T24 - T27;
275 }
276 }
277 }
278 }
279 ii[WS(rs, 11)] = T2C + T2B;
280 ii[WS(rs, 5)] = T2B - T2C;
281 }
282 }
283 }
284
285 static const tw_instr twinstr[] = {
286 {TW_FULL, 0, 12},
287 {TW_NEXT, 1, 0}
288 };
289
290 static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {72, 22, 46, 0}, 0, 0, 0 };
291
292 void X(codelet_t1_12) (planner *p) {
293 X(kdft_dit_register) (p, t1_12, &desc);
294 }
295 #else /* HAVE_FMA */
296
297 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 12 -name t1_12 -include t.h */
298
299 /*
300 * This function contains 118 FP additions, 60 FP multiplications,
301 * (or, 88 additions, 30 multiplications, 30 fused multiply/add),
302 * 47 stack variables, 2 constants, and 48 memory accesses
303 */
304 #include "t.h"
305
306 static void t1_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
307 {
308 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
309 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
310 {
311 INT m;
312 for (m = mb, W = W + (mb * 22); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 22, MAKE_VOLATILE_STRIDE(24, rs)) {
313 E T1, T1W, T18, T21, Tc, T15, T1V, T22, TR, T1E, T1o, T1D, T12, T1l, T1F;
314 E T1G, Ti, T1S, T1d, T24, Tt, T1a, T1T, T25, TA, T1z, T1j, T1y, TL, T1g;
315 E T1A, T1B;
316 {
317 E T6, T16, Tb, T17;
318 T1 = ri[0];
319 T1W = ii[0];
320 {
321 E T3, T5, T2, T4;
322 T3 = ri[WS(rs, 4)];
323 T5 = ii[WS(rs, 4)];
324 T2 = W[6];
325 T4 = W[7];
326 T6 = FMA(T2, T3, T4 * T5);
327 T16 = FNMS(T4, T3, T2 * T5);
328 }
329 {
330 E T8, Ta, T7, T9;
331 T8 = ri[WS(rs, 8)];
332 Ta = ii[WS(rs, 8)];
333 T7 = W[14];
334 T9 = W[15];
335 Tb = FMA(T7, T8, T9 * Ta);
336 T17 = FNMS(T9, T8, T7 * Ta);
337 }
338 T18 = KP866025403 * (T16 - T17);
339 T21 = KP866025403 * (Tb - T6);
340 Tc = T6 + Tb;
341 T15 = FNMS(KP500000000, Tc, T1);
342 T1V = T16 + T17;
343 T22 = FNMS(KP500000000, T1V, T1W);
344 }
345 {
346 E T11, T1n, TW, T1m;
347 {
348 E TO, TQ, TN, TP;
349 TO = ri[WS(rs, 9)];
350 TQ = ii[WS(rs, 9)];
351 TN = W[16];
352 TP = W[17];
353 TR = FMA(TN, TO, TP * TQ);
354 T1E = FNMS(TP, TO, TN * TQ);
355 }
356 {
357 E TY, T10, TX, TZ;
358 TY = ri[WS(rs, 5)];
359 T10 = ii[WS(rs, 5)];
360 TX = W[8];
361 TZ = W[9];
362 T11 = FMA(TX, TY, TZ * T10);
363 T1n = FNMS(TZ, TY, TX * T10);
364 }
365 {
366 E TT, TV, TS, TU;
367 TT = ri[WS(rs, 1)];
368 TV = ii[WS(rs, 1)];
369 TS = W[0];
370 TU = W[1];
371 TW = FMA(TS, TT, TU * TV);
372 T1m = FNMS(TU, TT, TS * TV);
373 }
374 T1o = KP866025403 * (T1m - T1n);
375 T1D = KP866025403 * (T11 - TW);
376 T12 = TW + T11;
377 T1l = FNMS(KP500000000, T12, TR);
378 T1F = T1m + T1n;
379 T1G = FNMS(KP500000000, T1F, T1E);
380 }
381 {
382 E Ts, T1c, Tn, T1b;
383 {
384 E Tf, Th, Te, Tg;
385 Tf = ri[WS(rs, 6)];
386 Th = ii[WS(rs, 6)];
387 Te = W[10];
388 Tg = W[11];
389 Ti = FMA(Te, Tf, Tg * Th);
390 T1S = FNMS(Tg, Tf, Te * Th);
391 }
392 {
393 E Tp, Tr, To, Tq;
394 Tp = ri[WS(rs, 2)];
395 Tr = ii[WS(rs, 2)];
396 To = W[2];
397 Tq = W[3];
398 Ts = FMA(To, Tp, Tq * Tr);
399 T1c = FNMS(Tq, Tp, To * Tr);
400 }
401 {
402 E Tk, Tm, Tj, Tl;
403 Tk = ri[WS(rs, 10)];
404 Tm = ii[WS(rs, 10)];
405 Tj = W[18];
406 Tl = W[19];
407 Tn = FMA(Tj, Tk, Tl * Tm);
408 T1b = FNMS(Tl, Tk, Tj * Tm);
409 }
410 T1d = KP866025403 * (T1b - T1c);
411 T24 = KP866025403 * (Ts - Tn);
412 Tt = Tn + Ts;
413 T1a = FNMS(KP500000000, Tt, Ti);
414 T1T = T1b + T1c;
415 T25 = FNMS(KP500000000, T1T, T1S);
416 }
417 {
418 E TK, T1i, TF, T1h;
419 {
420 E Tx, Tz, Tw, Ty;
421 Tx = ri[WS(rs, 3)];
422 Tz = ii[WS(rs, 3)];
423 Tw = W[4];
424 Ty = W[5];
425 TA = FMA(Tw, Tx, Ty * Tz);
426 T1z = FNMS(Ty, Tx, Tw * Tz);
427 }
428 {
429 E TH, TJ, TG, TI;
430 TH = ri[WS(rs, 11)];
431 TJ = ii[WS(rs, 11)];
432 TG = W[20];
433 TI = W[21];
434 TK = FMA(TG, TH, TI * TJ);
435 T1i = FNMS(TI, TH, TG * TJ);
436 }
437 {
438 E TC, TE, TB, TD;
439 TC = ri[WS(rs, 7)];
440 TE = ii[WS(rs, 7)];
441 TB = W[12];
442 TD = W[13];
443 TF = FMA(TB, TC, TD * TE);
444 T1h = FNMS(TD, TC, TB * TE);
445 }
446 T1j = KP866025403 * (T1h - T1i);
447 T1y = KP866025403 * (TK - TF);
448 TL = TF + TK;
449 T1g = FNMS(KP500000000, TL, TA);
450 T1A = T1h + T1i;
451 T1B = FNMS(KP500000000, T1A, T1z);
452 }
453 {
454 E Tv, T1N, T1Y, T20, T14, T1Z, T1Q, T1R;
455 {
456 E Td, Tu, T1U, T1X;
457 Td = T1 + Tc;
458 Tu = Ti + Tt;
459 Tv = Td + Tu;
460 T1N = Td - Tu;
461 T1U = T1S + T1T;
462 T1X = T1V + T1W;
463 T1Y = T1U + T1X;
464 T20 = T1X - T1U;
465 }
466 {
467 E TM, T13, T1O, T1P;
468 TM = TA + TL;
469 T13 = TR + T12;
470 T14 = TM + T13;
471 T1Z = TM - T13;
472 T1O = T1z + T1A;
473 T1P = T1E + T1F;
474 T1Q = T1O - T1P;
475 T1R = T1O + T1P;
476 }
477 ri[WS(rs, 6)] = Tv - T14;
478 ii[WS(rs, 6)] = T1Y - T1R;
479 ri[0] = Tv + T14;
480 ii[0] = T1R + T1Y;
481 ri[WS(rs, 3)] = T1N - T1Q;
482 ii[WS(rs, 3)] = T1Z + T20;
483 ri[WS(rs, 9)] = T1N + T1Q;
484 ii[WS(rs, 9)] = T20 - T1Z;
485 }
486 {
487 E T1t, T1x, T27, T2a, T1w, T28, T1I, T29;
488 {
489 E T1r, T1s, T23, T26;
490 T1r = T15 + T18;
491 T1s = T1a + T1d;
492 T1t = T1r + T1s;
493 T1x = T1r - T1s;
494 T23 = T21 + T22;
495 T26 = T24 + T25;
496 T27 = T23 - T26;
497 T2a = T26 + T23;
498 }
499 {
500 E T1u, T1v, T1C, T1H;
501 T1u = T1g + T1j;
502 T1v = T1l + T1o;
503 T1w = T1u + T1v;
504 T28 = T1u - T1v;
505 T1C = T1y + T1B;
506 T1H = T1D + T1G;
507 T1I = T1C - T1H;
508 T29 = T1C + T1H;
509 }
510 ri[WS(rs, 10)] = T1t - T1w;
511 ii[WS(rs, 10)] = T2a - T29;
512 ri[WS(rs, 4)] = T1t + T1w;
513 ii[WS(rs, 4)] = T29 + T2a;
514 ri[WS(rs, 7)] = T1x - T1I;
515 ii[WS(rs, 7)] = T28 + T27;
516 ri[WS(rs, 1)] = T1x + T1I;
517 ii[WS(rs, 1)] = T27 - T28;
518 }
519 {
520 E T1f, T1J, T2d, T2f, T1q, T2g, T1M, T2e;
521 {
522 E T19, T1e, T2b, T2c;
523 T19 = T15 - T18;
524 T1e = T1a - T1d;
525 T1f = T19 + T1e;
526 T1J = T19 - T1e;
527 T2b = T25 - T24;
528 T2c = T22 - T21;
529 T2d = T2b + T2c;
530 T2f = T2c - T2b;
531 }
532 {
533 E T1k, T1p, T1K, T1L;
534 T1k = T1g - T1j;
535 T1p = T1l - T1o;
536 T1q = T1k + T1p;
537 T2g = T1k - T1p;
538 T1K = T1B - T1y;
539 T1L = T1G - T1D;
540 T1M = T1K - T1L;
541 T2e = T1K + T1L;
542 }
543 ri[WS(rs, 2)] = T1f - T1q;
544 ii[WS(rs, 2)] = T2d - T2e;
545 ri[WS(rs, 8)] = T1f + T1q;
546 ii[WS(rs, 8)] = T2e + T2d;
547 ri[WS(rs, 11)] = T1J - T1M;
548 ii[WS(rs, 11)] = T2g + T2f;
549 ri[WS(rs, 5)] = T1J + T1M;
550 ii[WS(rs, 5)] = T2f - T2g;
551 }
552 }
553 }
554 }
555
556 static const tw_instr twinstr[] = {
557 {TW_FULL, 0, 12},
558 {TW_NEXT, 1, 0}
559 };
560
561 static const ct_desc desc = { 12, "t1_12", twinstr, &GENUS, {88, 30, 30, 0}, 0, 0, 0 };
562
563 void X(codelet_t1_12) (planner *p) {
564 X(kdft_dit_register) (p, t1_12, &desc);
565 }
566 #endif /* HAVE_FMA */