Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_9.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 07:35:49 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 9 -name t1_9 -include t.h */
29
30 /*
31 * This function contains 96 FP additions, 88 FP multiplications,
32 * (or, 24 additions, 16 multiplications, 72 fused multiply/add),
33 * 72 stack variables, 10 constants, and 36 memory accesses
34 */
35 #include "t.h"
36
37 static void t1_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP954188894, +0.954188894138671133499268364187245676532219158);
40 DK(KP852868531, +0.852868531952443209628250963940074071936020296);
41 DK(KP363970234, +0.363970234266202361351047882776834043890471784);
42 DK(KP492403876, +0.492403876506104029683371512294761506835321626);
43 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
44 DK(KP777861913, +0.777861913430206160028177977318626690410586096);
45 DK(KP839099631, +0.839099631177280011763127298123181364687434283);
46 DK(KP176326980, +0.176326980708464973471090386868618986121633062);
47 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
48 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
49 {
50 INT m;
51 for (m = mb, W = W + (mb * 16); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
52 E T1K, T24, T1H, T23;
53 {
54 E T1, T1R, T1Q, T10, T1W, Te, TB, T1l, T1r, T1q, T1M, TE, T1g, Tz, T12;
55 E TC, TH, TK, T17, TR, TG, TJ, TD;
56 T1 = ri[0];
57 T1R = ii[0];
58 {
59 E T9, Tc, TY, Ta, Tb, TX, T7;
60 {
61 E T3, T6, T8, TW, T4, T2, T5;
62 T3 = ri[WS(rs, 3)];
63 T6 = ii[WS(rs, 3)];
64 T2 = W[4];
65 T9 = ri[WS(rs, 6)];
66 Tc = ii[WS(rs, 6)];
67 T8 = W[10];
68 TW = T2 * T6;
69 T4 = T2 * T3;
70 T5 = W[5];
71 TY = T8 * Tc;
72 Ta = T8 * T9;
73 Tb = W[11];
74 TX = FNMS(T5, T3, TW);
75 T7 = FMA(T5, T6, T4);
76 }
77 {
78 E Th, Tk, Ti, T1n, Tn, Tq, Tp, T1i, Tx, T1j, To, Tj, TZ, Td, Tg;
79 E TA, Tl, Ty;
80 Th = ri[WS(rs, 1)];
81 TZ = FNMS(Tb, T9, TY);
82 Td = FMA(Tb, Tc, Ta);
83 Tk = ii[WS(rs, 1)];
84 Tg = W[0];
85 T1Q = TX + TZ;
86 T10 = TX - TZ;
87 T1W = Td - T7;
88 Te = T7 + Td;
89 Ti = Tg * Th;
90 T1n = Tg * Tk;
91 {
92 E Tt, Tw, Ts, Tv, T1h, Tu, Tm;
93 Tt = ri[WS(rs, 7)];
94 Tw = ii[WS(rs, 7)];
95 Ts = W[12];
96 Tv = W[13];
97 Tn = ri[WS(rs, 4)];
98 Tq = ii[WS(rs, 4)];
99 T1h = Ts * Tw;
100 Tu = Ts * Tt;
101 Tm = W[6];
102 Tp = W[7];
103 T1i = FNMS(Tv, Tt, T1h);
104 Tx = FMA(Tv, Tw, Tu);
105 T1j = Tm * Tq;
106 To = Tm * Tn;
107 }
108 Tj = W[1];
109 TB = ri[WS(rs, 2)];
110 {
111 E T1k, Tr, T1o, T1p;
112 T1k = FNMS(Tp, Tn, T1j);
113 Tr = FMA(Tp, Tq, To);
114 T1o = FNMS(Tj, Th, T1n);
115 Tl = FMA(Tj, Tk, Ti);
116 T1p = T1k + T1i;
117 T1l = T1i - T1k;
118 Ty = Tr + Tx;
119 T1r = Tr - Tx;
120 T1q = FNMS(KP500000000, T1p, T1o);
121 T1M = T1o + T1p;
122 TE = ii[WS(rs, 2)];
123 }
124 T1g = FNMS(KP500000000, Ty, Tl);
125 Tz = Tl + Ty;
126 TA = W[2];
127 {
128 E TN, TQ, TP, T16, TO, TM;
129 TN = ri[WS(rs, 8)];
130 TQ = ii[WS(rs, 8)];
131 TM = W[14];
132 T12 = TA * TE;
133 TC = TA * TB;
134 TP = W[15];
135 T16 = TM * TQ;
136 TO = TM * TN;
137 TH = ri[WS(rs, 5)];
138 TK = ii[WS(rs, 5)];
139 T17 = FNMS(TP, TN, T16);
140 TR = FMA(TP, TQ, TO);
141 TG = W[8];
142 TJ = W[9];
143 }
144 TD = W[3];
145 }
146 }
147 {
148 E TV, Tf, T1S, T1V, T1d, T1a, T19, T1N, TT, T1c;
149 {
150 E T13, TF, T15, TL, T14, TI, TS, T18;
151 TV = FNMS(KP500000000, Te, T1);
152 Tf = T1 + Te;
153 T14 = TG * TK;
154 TI = TG * TH;
155 T13 = FNMS(TD, TB, T12);
156 TF = FMA(TD, TE, TC);
157 T15 = FNMS(TJ, TH, T14);
158 TL = FMA(TJ, TK, TI);
159 T1S = T1Q + T1R;
160 T1V = FNMS(KP500000000, T1Q, T1R);
161 T18 = T15 + T17;
162 T1d = T15 - T17;
163 TS = TL + TR;
164 T1a = TR - TL;
165 T19 = FNMS(KP500000000, T18, T13);
166 T1N = T13 + T18;
167 TT = TF + TS;
168 T1c = FNMS(KP500000000, TS, TF);
169 }
170 {
171 E T11, T1z, T1E, T1D, T21, T1X, T1I, T1C, T1Y, T1y, T20, T1u, T1U, TU;
172 T1U = TT - Tz;
173 TU = Tz + TT;
174 {
175 E T1P, T1O, T1L, T1T;
176 T1P = T1M + T1N;
177 T1O = T1M - T1N;
178 T11 = FMA(KP866025403, T10, TV);
179 T1z = FNMS(KP866025403, T10, TV);
180 T1L = FNMS(KP500000000, TU, Tf);
181 ri[0] = Tf + TU;
182 T1T = FNMS(KP500000000, T1P, T1S);
183 ii[0] = T1P + T1S;
184 ri[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
185 ri[WS(rs, 6)] = FNMS(KP866025403, T1O, T1L);
186 ii[WS(rs, 6)] = FNMS(KP866025403, T1U, T1T);
187 ii[WS(rs, 3)] = FMA(KP866025403, T1U, T1T);
188 }
189 {
190 E T1B, T1m, T1w, T1f, T1s, T1A, T1b, T1e, T1x, T1t;
191 T1E = FNMS(KP866025403, T1a, T19);
192 T1b = FMA(KP866025403, T1a, T19);
193 T1e = FMA(KP866025403, T1d, T1c);
194 T1D = FNMS(KP866025403, T1d, T1c);
195 T1B = FMA(KP866025403, T1l, T1g);
196 T1m = FNMS(KP866025403, T1l, T1g);
197 T21 = FNMS(KP866025403, T1W, T1V);
198 T1X = FMA(KP866025403, T1W, T1V);
199 T1w = FNMS(KP176326980, T1b, T1e);
200 T1f = FMA(KP176326980, T1e, T1b);
201 T1s = FNMS(KP866025403, T1r, T1q);
202 T1A = FMA(KP866025403, T1r, T1q);
203 T1x = FNMS(KP839099631, T1m, T1s);
204 T1t = FMA(KP839099631, T1s, T1m);
205 T1I = FNMS(KP176326980, T1A, T1B);
206 T1C = FMA(KP176326980, T1B, T1A);
207 T1Y = FNMS(KP777861913, T1x, T1w);
208 T1y = FMA(KP777861913, T1x, T1w);
209 T20 = FNMS(KP777861913, T1t, T1f);
210 T1u = FMA(KP777861913, T1t, T1f);
211 }
212 {
213 E T22, T1G, T1Z, T1F, T1J, T1v;
214 ii[WS(rs, 1)] = FNMS(KP984807753, T1Y, T1X);
215 T1v = FNMS(KP492403876, T1u, T11);
216 ri[WS(rs, 1)] = FMA(KP984807753, T1u, T11);
217 T1F = FNMS(KP363970234, T1E, T1D);
218 T1J = FMA(KP363970234, T1D, T1E);
219 ri[WS(rs, 4)] = FMA(KP852868531, T1y, T1v);
220 ri[WS(rs, 7)] = FNMS(KP852868531, T1y, T1v);
221 T1K = FNMS(KP954188894, T1J, T1I);
222 T22 = FMA(KP954188894, T1J, T1I);
223 T1G = FNMS(KP954188894, T1F, T1C);
224 T24 = FMA(KP954188894, T1F, T1C);
225 T1Z = FMA(KP492403876, T1Y, T1X);
226 ii[WS(rs, 2)] = FNMS(KP984807753, T22, T21);
227 ri[WS(rs, 2)] = FMA(KP984807753, T1G, T1z);
228 T1H = FNMS(KP492403876, T1G, T1z);
229 ii[WS(rs, 7)] = FNMS(KP852868531, T20, T1Z);
230 ii[WS(rs, 4)] = FMA(KP852868531, T20, T1Z);
231 T23 = FMA(KP492403876, T22, T21);
232 }
233 }
234 }
235 }
236 ri[WS(rs, 8)] = FMA(KP852868531, T1K, T1H);
237 ri[WS(rs, 5)] = FNMS(KP852868531, T1K, T1H);
238 ii[WS(rs, 8)] = FMA(KP852868531, T24, T23);
239 ii[WS(rs, 5)] = FNMS(KP852868531, T24, T23);
240 }
241 }
242 }
243
244 static const tw_instr twinstr[] = {
245 {TW_FULL, 0, 9},
246 {TW_NEXT, 1, 0}
247 };
248
249 static const ct_desc desc = { 9, "t1_9", twinstr, &GENUS, {24, 16, 72, 0}, 0, 0, 0 };
250
251 void X(codelet_t1_9) (planner *p) {
252 X(kdft_dit_register) (p, t1_9, &desc);
253 }
254 #else /* HAVE_FMA */
255
256 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 9 -name t1_9 -include t.h */
257
258 /*
259 * This function contains 96 FP additions, 72 FP multiplications,
260 * (or, 60 additions, 36 multiplications, 36 fused multiply/add),
261 * 41 stack variables, 8 constants, and 36 memory accesses
262 */
263 #include "t.h"
264
265 static void t1_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
266 {
267 DK(KP939692620, +0.939692620785908384054109277324731469936208134);
268 DK(KP342020143, +0.342020143325668733044099614682259580763083368);
269 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
270 DK(KP173648177, +0.173648177666930348851716626769314796000375677);
271 DK(KP642787609, +0.642787609686539326322643409907263432907559884);
272 DK(KP766044443, +0.766044443118978035202392650555416673935832457);
273 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
274 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
275 {
276 INT m;
277 for (m = mb, W = W + (mb * 16); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
278 E T1, T1B, TQ, T1G, Tc, TN, T1A, T1H, TL, T1x, T17, T1o, T1c, T1n, Tu;
279 E T1w, TW, T1k, T11, T1l;
280 {
281 E T6, TO, Tb, TP;
282 T1 = ri[0];
283 T1B = ii[0];
284 {
285 E T3, T5, T2, T4;
286 T3 = ri[WS(rs, 3)];
287 T5 = ii[WS(rs, 3)];
288 T2 = W[4];
289 T4 = W[5];
290 T6 = FMA(T2, T3, T4 * T5);
291 TO = FNMS(T4, T3, T2 * T5);
292 }
293 {
294 E T8, Ta, T7, T9;
295 T8 = ri[WS(rs, 6)];
296 Ta = ii[WS(rs, 6)];
297 T7 = W[10];
298 T9 = W[11];
299 Tb = FMA(T7, T8, T9 * Ta);
300 TP = FNMS(T9, T8, T7 * Ta);
301 }
302 TQ = KP866025403 * (TO - TP);
303 T1G = KP866025403 * (Tb - T6);
304 Tc = T6 + Tb;
305 TN = FNMS(KP500000000, Tc, T1);
306 T1A = TO + TP;
307 T1H = FNMS(KP500000000, T1A, T1B);
308 }
309 {
310 E Tz, T19, TE, T14, TJ, T15, TK, T1a;
311 {
312 E Tw, Ty, Tv, Tx;
313 Tw = ri[WS(rs, 2)];
314 Ty = ii[WS(rs, 2)];
315 Tv = W[2];
316 Tx = W[3];
317 Tz = FMA(Tv, Tw, Tx * Ty);
318 T19 = FNMS(Tx, Tw, Tv * Ty);
319 }
320 {
321 E TB, TD, TA, TC;
322 TB = ri[WS(rs, 5)];
323 TD = ii[WS(rs, 5)];
324 TA = W[8];
325 TC = W[9];
326 TE = FMA(TA, TB, TC * TD);
327 T14 = FNMS(TC, TB, TA * TD);
328 }
329 {
330 E TG, TI, TF, TH;
331 TG = ri[WS(rs, 8)];
332 TI = ii[WS(rs, 8)];
333 TF = W[14];
334 TH = W[15];
335 TJ = FMA(TF, TG, TH * TI);
336 T15 = FNMS(TH, TG, TF * TI);
337 }
338 TK = TE + TJ;
339 T1a = T14 + T15;
340 TL = Tz + TK;
341 T1x = T19 + T1a;
342 {
343 E T13, T16, T18, T1b;
344 T13 = FNMS(KP500000000, TK, Tz);
345 T16 = KP866025403 * (T14 - T15);
346 T17 = T13 + T16;
347 T1o = T13 - T16;
348 T18 = KP866025403 * (TJ - TE);
349 T1b = FNMS(KP500000000, T1a, T19);
350 T1c = T18 + T1b;
351 T1n = T1b - T18;
352 }
353 }
354 {
355 E Ti, TY, Tn, TT, Ts, TU, Tt, TZ;
356 {
357 E Tf, Th, Te, Tg;
358 Tf = ri[WS(rs, 1)];
359 Th = ii[WS(rs, 1)];
360 Te = W[0];
361 Tg = W[1];
362 Ti = FMA(Te, Tf, Tg * Th);
363 TY = FNMS(Tg, Tf, Te * Th);
364 }
365 {
366 E Tk, Tm, Tj, Tl;
367 Tk = ri[WS(rs, 4)];
368 Tm = ii[WS(rs, 4)];
369 Tj = W[6];
370 Tl = W[7];
371 Tn = FMA(Tj, Tk, Tl * Tm);
372 TT = FNMS(Tl, Tk, Tj * Tm);
373 }
374 {
375 E Tp, Tr, To, Tq;
376 Tp = ri[WS(rs, 7)];
377 Tr = ii[WS(rs, 7)];
378 To = W[12];
379 Tq = W[13];
380 Ts = FMA(To, Tp, Tq * Tr);
381 TU = FNMS(Tq, Tp, To * Tr);
382 }
383 Tt = Tn + Ts;
384 TZ = TT + TU;
385 Tu = Ti + Tt;
386 T1w = TY + TZ;
387 {
388 E TS, TV, TX, T10;
389 TS = FNMS(KP500000000, Tt, Ti);
390 TV = KP866025403 * (TT - TU);
391 TW = TS + TV;
392 T1k = TS - TV;
393 TX = KP866025403 * (Ts - Tn);
394 T10 = FNMS(KP500000000, TZ, TY);
395 T11 = TX + T10;
396 T1l = T10 - TX;
397 }
398 }
399 {
400 E T1y, Td, TM, T1v;
401 T1y = KP866025403 * (T1w - T1x);
402 Td = T1 + Tc;
403 TM = Tu + TL;
404 T1v = FNMS(KP500000000, TM, Td);
405 ri[0] = Td + TM;
406 ri[WS(rs, 3)] = T1v + T1y;
407 ri[WS(rs, 6)] = T1v - T1y;
408 }
409 {
410 E T1D, T1z, T1C, T1E;
411 T1D = KP866025403 * (TL - Tu);
412 T1z = T1w + T1x;
413 T1C = T1A + T1B;
414 T1E = FNMS(KP500000000, T1z, T1C);
415 ii[0] = T1z + T1C;
416 ii[WS(rs, 6)] = T1E - T1D;
417 ii[WS(rs, 3)] = T1D + T1E;
418 }
419 {
420 E TR, T1I, T1e, T1J, T1i, T1F, T1f, T1K;
421 TR = TN + TQ;
422 T1I = T1G + T1H;
423 {
424 E T12, T1d, T1g, T1h;
425 T12 = FMA(KP766044443, TW, KP642787609 * T11);
426 T1d = FMA(KP173648177, T17, KP984807753 * T1c);
427 T1e = T12 + T1d;
428 T1J = KP866025403 * (T1d - T12);
429 T1g = FNMS(KP642787609, TW, KP766044443 * T11);
430 T1h = FNMS(KP984807753, T17, KP173648177 * T1c);
431 T1i = KP866025403 * (T1g - T1h);
432 T1F = T1g + T1h;
433 }
434 ri[WS(rs, 1)] = TR + T1e;
435 ii[WS(rs, 1)] = T1F + T1I;
436 T1f = FNMS(KP500000000, T1e, TR);
437 ri[WS(rs, 7)] = T1f - T1i;
438 ri[WS(rs, 4)] = T1f + T1i;
439 T1K = FNMS(KP500000000, T1F, T1I);
440 ii[WS(rs, 4)] = T1J + T1K;
441 ii[WS(rs, 7)] = T1K - T1J;
442 }
443 {
444 E T1j, T1M, T1q, T1N, T1u, T1L, T1r, T1O;
445 T1j = TN - TQ;
446 T1M = T1H - T1G;
447 {
448 E T1m, T1p, T1s, T1t;
449 T1m = FMA(KP173648177, T1k, KP984807753 * T1l);
450 T1p = FNMS(KP939692620, T1o, KP342020143 * T1n);
451 T1q = T1m + T1p;
452 T1N = KP866025403 * (T1p - T1m);
453 T1s = FNMS(KP984807753, T1k, KP173648177 * T1l);
454 T1t = FMA(KP342020143, T1o, KP939692620 * T1n);
455 T1u = KP866025403 * (T1s + T1t);
456 T1L = T1s - T1t;
457 }
458 ri[WS(rs, 2)] = T1j + T1q;
459 ii[WS(rs, 2)] = T1L + T1M;
460 T1r = FNMS(KP500000000, T1q, T1j);
461 ri[WS(rs, 8)] = T1r - T1u;
462 ri[WS(rs, 5)] = T1r + T1u;
463 T1O = FNMS(KP500000000, T1L, T1M);
464 ii[WS(rs, 5)] = T1N + T1O;
465 ii[WS(rs, 8)] = T1O - T1N;
466 }
467 }
468 }
469 }
470
471 static const tw_instr twinstr[] = {
472 {TW_FULL, 0, 9},
473 {TW_NEXT, 1, 0}
474 };
475
476 static const ct_desc desc = { 9, "t1_9", twinstr, &GENUS, {60, 36, 36, 0}, 0, 0, 0 };
477
478 void X(codelet_t1_9) (planner *p) {
479 X(kdft_dit_register) (p, t1_9, &desc);
480 }
481 #endif /* HAVE_FMA */