Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_10.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 07:35:49 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 10 -name t1_10 -include t.h */
29
30 /*
31 * This function contains 102 FP additions, 72 FP multiplications,
32 * (or, 48 additions, 18 multiplications, 54 fused multiply/add),
33 * 70 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "t.h"
36
37 static void t1_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT m;
45 for (m = mb, W = W + (mb * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
46 E T1X, T21, T20, T22;
47 {
48 E T23, T1U, T8, T12, T1y, T25, T1P, T1H, T1Y, T18, T10, T2b, T1K, T1O, T15;
49 E T1Z, T2a, Tz, T24, T1n;
50 {
51 E T1, T1T, T3, T6, T2, T5;
52 T1 = ri[0];
53 T1T = ii[0];
54 T3 = ri[WS(rs, 5)];
55 T6 = ii[WS(rs, 5)];
56 T2 = W[8];
57 T5 = W[9];
58 {
59 E T1w, TY, T1s, T1F, TM, T16, T1u, TS;
60 {
61 E TF, T1p, TO, TR, T1r, TL, TN, TQ, T1t, TP;
62 {
63 E TU, TX, TT, TW;
64 {
65 E TB, TE, T1R, T4, TA, TD;
66 TB = ri[WS(rs, 4)];
67 TE = ii[WS(rs, 4)];
68 T1R = T2 * T6;
69 T4 = T2 * T3;
70 TA = W[6];
71 TD = W[7];
72 {
73 E T1S, T7, T1o, TC;
74 T1S = FNMS(T5, T3, T1R);
75 T7 = FMA(T5, T6, T4);
76 T1o = TA * TE;
77 TC = TA * TB;
78 T23 = T1T - T1S;
79 T1U = T1S + T1T;
80 T8 = T1 - T7;
81 T12 = T1 + T7;
82 TF = FMA(TD, TE, TC);
83 T1p = FNMS(TD, TB, T1o);
84 }
85 }
86 TU = ri[WS(rs, 1)];
87 TX = ii[WS(rs, 1)];
88 TT = W[0];
89 TW = W[1];
90 {
91 E TH, TK, TJ, T1q, TI, T1v, TV, TG;
92 TH = ri[WS(rs, 9)];
93 TK = ii[WS(rs, 9)];
94 T1v = TT * TX;
95 TV = TT * TU;
96 TG = W[16];
97 TJ = W[17];
98 T1w = FNMS(TW, TU, T1v);
99 TY = FMA(TW, TX, TV);
100 T1q = TG * TK;
101 TI = TG * TH;
102 TO = ri[WS(rs, 6)];
103 TR = ii[WS(rs, 6)];
104 T1r = FNMS(TJ, TH, T1q);
105 TL = FMA(TJ, TK, TI);
106 TN = W[10];
107 TQ = W[11];
108 }
109 }
110 T1s = T1p - T1r;
111 T1F = T1p + T1r;
112 TM = TF - TL;
113 T16 = TF + TL;
114 T1t = TN * TR;
115 TP = TN * TO;
116 T1u = FNMS(TQ, TO, T1t);
117 TS = FMA(TQ, TR, TP);
118 }
119 {
120 E T1e, Te, T1l, Tx, Tn, Tq, Tp, T1g, Tk, T1i, To;
121 {
122 E Tt, Tw, Tv, T1k, Tu;
123 {
124 E Ta, Td, T9, Tc, T1d, Tb, Ts;
125 Ta = ri[WS(rs, 2)];
126 Td = ii[WS(rs, 2)];
127 {
128 E T1G, T1x, TZ, T17;
129 T1G = T1u + T1w;
130 T1x = T1u - T1w;
131 TZ = TS - TY;
132 T17 = TS + TY;
133 T1y = T1s - T1x;
134 T25 = T1s + T1x;
135 T1P = T1F + T1G;
136 T1H = T1F - T1G;
137 T1Y = T16 - T17;
138 T18 = T16 + T17;
139 T10 = TM + TZ;
140 T2b = TM - TZ;
141 T9 = W[2];
142 }
143 Tc = W[3];
144 Tt = ri[WS(rs, 3)];
145 Tw = ii[WS(rs, 3)];
146 T1d = T9 * Td;
147 Tb = T9 * Ta;
148 Ts = W[4];
149 Tv = W[5];
150 T1e = FNMS(Tc, Ta, T1d);
151 Te = FMA(Tc, Td, Tb);
152 T1k = Ts * Tw;
153 Tu = Ts * Tt;
154 }
155 {
156 E Tg, Tj, Tf, Ti, T1f, Th, Tm;
157 Tg = ri[WS(rs, 7)];
158 Tj = ii[WS(rs, 7)];
159 T1l = FNMS(Tv, Tt, T1k);
160 Tx = FMA(Tv, Tw, Tu);
161 Tf = W[12];
162 Ti = W[13];
163 Tn = ri[WS(rs, 8)];
164 Tq = ii[WS(rs, 8)];
165 T1f = Tf * Tj;
166 Th = Tf * Tg;
167 Tm = W[14];
168 Tp = W[15];
169 T1g = FNMS(Ti, Tg, T1f);
170 Tk = FMA(Ti, Tj, Th);
171 T1i = Tm * Tq;
172 To = Tm * Tn;
173 }
174 }
175 {
176 E T1h, T1I, Tl, T13, T1j, Tr;
177 T1h = T1e - T1g;
178 T1I = T1e + T1g;
179 Tl = Te - Tk;
180 T13 = Te + Tk;
181 T1j = FNMS(Tp, Tn, T1i);
182 Tr = FMA(Tp, Tq, To);
183 {
184 E T1m, T1J, T14, Ty;
185 T1m = T1j - T1l;
186 T1J = T1j + T1l;
187 T14 = Tr + Tx;
188 Ty = Tr - Tx;
189 T1K = T1I - T1J;
190 T1O = T1I + T1J;
191 T15 = T13 + T14;
192 T1Z = T13 - T14;
193 T2a = Tl - Ty;
194 Tz = Tl + Ty;
195 T24 = T1h + T1m;
196 T1n = T1h - T1m;
197 }
198 }
199 }
200 }
201 }
202 {
203 E T2c, T2e, T29, T2d;
204 {
205 E T1b, T11, T26, T28, T27;
206 T1b = Tz - T10;
207 T11 = Tz + T10;
208 T26 = T24 + T25;
209 T28 = T24 - T25;
210 {
211 E T1B, T1z, T1a, T1A, T1c;
212 T1B = FNMS(KP618033988, T1n, T1y);
213 T1z = FMA(KP618033988, T1y, T1n);
214 ri[WS(rs, 5)] = T8 + T11;
215 T1a = FNMS(KP250000000, T11, T8);
216 T1A = FNMS(KP559016994, T1b, T1a);
217 T1c = FMA(KP559016994, T1b, T1a);
218 T27 = FNMS(KP250000000, T26, T23);
219 T2c = FMA(KP618033988, T2b, T2a);
220 T2e = FNMS(KP618033988, T2a, T2b);
221 ri[WS(rs, 1)] = FMA(KP951056516, T1z, T1c);
222 ri[WS(rs, 9)] = FNMS(KP951056516, T1z, T1c);
223 ri[WS(rs, 3)] = FMA(KP951056516, T1B, T1A);
224 ri[WS(rs, 7)] = FNMS(KP951056516, T1B, T1A);
225 }
226 ii[WS(rs, 5)] = T26 + T23;
227 T29 = FMA(KP559016994, T28, T27);
228 T2d = FNMS(KP559016994, T28, T27);
229 }
230 {
231 E T1E, T1M, T1L, T1N, T19, T1D, T1C, T1Q, T1W, T1V;
232 T19 = T15 + T18;
233 T1D = T15 - T18;
234 ii[WS(rs, 7)] = FMA(KP951056516, T2e, T2d);
235 ii[WS(rs, 3)] = FNMS(KP951056516, T2e, T2d);
236 ii[WS(rs, 9)] = FMA(KP951056516, T2c, T29);
237 ii[WS(rs, 1)] = FNMS(KP951056516, T2c, T29);
238 T1C = FNMS(KP250000000, T19, T12);
239 ri[0] = T12 + T19;
240 T1E = FNMS(KP559016994, T1D, T1C);
241 T1M = FMA(KP559016994, T1D, T1C);
242 T1L = FNMS(KP618033988, T1K, T1H);
243 T1N = FMA(KP618033988, T1H, T1K);
244 T1Q = T1O + T1P;
245 T1W = T1O - T1P;
246 ri[WS(rs, 6)] = FMA(KP951056516, T1N, T1M);
247 ri[WS(rs, 4)] = FNMS(KP951056516, T1N, T1M);
248 ri[WS(rs, 8)] = FMA(KP951056516, T1L, T1E);
249 ri[WS(rs, 2)] = FNMS(KP951056516, T1L, T1E);
250 T1V = FNMS(KP250000000, T1Q, T1U);
251 ii[0] = T1Q + T1U;
252 T1X = FNMS(KP559016994, T1W, T1V);
253 T21 = FMA(KP559016994, T1W, T1V);
254 T20 = FNMS(KP618033988, T1Z, T1Y);
255 T22 = FMA(KP618033988, T1Y, T1Z);
256 }
257 }
258 }
259 ii[WS(rs, 6)] = FNMS(KP951056516, T22, T21);
260 ii[WS(rs, 4)] = FMA(KP951056516, T22, T21);
261 ii[WS(rs, 8)] = FNMS(KP951056516, T20, T1X);
262 ii[WS(rs, 2)] = FMA(KP951056516, T20, T1X);
263 }
264 }
265 }
266
267 static const tw_instr twinstr[] = {
268 {TW_FULL, 0, 10},
269 {TW_NEXT, 1, 0}
270 };
271
272 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {48, 18, 54, 0}, 0, 0, 0 };
273
274 void X(codelet_t1_10) (planner *p) {
275 X(kdft_dit_register) (p, t1_10, &desc);
276 }
277 #else /* HAVE_FMA */
278
279 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 10 -name t1_10 -include t.h */
280
281 /*
282 * This function contains 102 FP additions, 60 FP multiplications,
283 * (or, 72 additions, 30 multiplications, 30 fused multiply/add),
284 * 45 stack variables, 4 constants, and 40 memory accesses
285 */
286 #include "t.h"
287
288 static void t1_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
289 {
290 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
291 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
292 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
293 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
294 {
295 INT m;
296 for (m = mb, W = W + (mb * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
297 E T7, T1O, TT, T1C, TF, TQ, TR, T1o, T1p, T1y, TX, TY, TZ, T1d, T1g;
298 E T1M, Ti, Tt, Tu, T1r, T1s, T1x, TU, TV, TW, T16, T19, T1L;
299 {
300 E T1, T1B, T6, T1A;
301 T1 = ri[0];
302 T1B = ii[0];
303 {
304 E T3, T5, T2, T4;
305 T3 = ri[WS(rs, 5)];
306 T5 = ii[WS(rs, 5)];
307 T2 = W[8];
308 T4 = W[9];
309 T6 = FMA(T2, T3, T4 * T5);
310 T1A = FNMS(T4, T3, T2 * T5);
311 }
312 T7 = T1 - T6;
313 T1O = T1B - T1A;
314 TT = T1 + T6;
315 T1C = T1A + T1B;
316 }
317 {
318 E Tz, T1b, TP, T1f, TE, T1c, TK, T1e;
319 {
320 E Tw, Ty, Tv, Tx;
321 Tw = ri[WS(rs, 4)];
322 Ty = ii[WS(rs, 4)];
323 Tv = W[6];
324 Tx = W[7];
325 Tz = FMA(Tv, Tw, Tx * Ty);
326 T1b = FNMS(Tx, Tw, Tv * Ty);
327 }
328 {
329 E TM, TO, TL, TN;
330 TM = ri[WS(rs, 1)];
331 TO = ii[WS(rs, 1)];
332 TL = W[0];
333 TN = W[1];
334 TP = FMA(TL, TM, TN * TO);
335 T1f = FNMS(TN, TM, TL * TO);
336 }
337 {
338 E TB, TD, TA, TC;
339 TB = ri[WS(rs, 9)];
340 TD = ii[WS(rs, 9)];
341 TA = W[16];
342 TC = W[17];
343 TE = FMA(TA, TB, TC * TD);
344 T1c = FNMS(TC, TB, TA * TD);
345 }
346 {
347 E TH, TJ, TG, TI;
348 TH = ri[WS(rs, 6)];
349 TJ = ii[WS(rs, 6)];
350 TG = W[10];
351 TI = W[11];
352 TK = FMA(TG, TH, TI * TJ);
353 T1e = FNMS(TI, TH, TG * TJ);
354 }
355 TF = Tz - TE;
356 TQ = TK - TP;
357 TR = TF + TQ;
358 T1o = T1b + T1c;
359 T1p = T1e + T1f;
360 T1y = T1o + T1p;
361 TX = Tz + TE;
362 TY = TK + TP;
363 TZ = TX + TY;
364 T1d = T1b - T1c;
365 T1g = T1e - T1f;
366 T1M = T1d + T1g;
367 }
368 {
369 E Tc, T14, Ts, T18, Th, T15, Tn, T17;
370 {
371 E T9, Tb, T8, Ta;
372 T9 = ri[WS(rs, 2)];
373 Tb = ii[WS(rs, 2)];
374 T8 = W[2];
375 Ta = W[3];
376 Tc = FMA(T8, T9, Ta * Tb);
377 T14 = FNMS(Ta, T9, T8 * Tb);
378 }
379 {
380 E Tp, Tr, To, Tq;
381 Tp = ri[WS(rs, 3)];
382 Tr = ii[WS(rs, 3)];
383 To = W[4];
384 Tq = W[5];
385 Ts = FMA(To, Tp, Tq * Tr);
386 T18 = FNMS(Tq, Tp, To * Tr);
387 }
388 {
389 E Te, Tg, Td, Tf;
390 Te = ri[WS(rs, 7)];
391 Tg = ii[WS(rs, 7)];
392 Td = W[12];
393 Tf = W[13];
394 Th = FMA(Td, Te, Tf * Tg);
395 T15 = FNMS(Tf, Te, Td * Tg);
396 }
397 {
398 E Tk, Tm, Tj, Tl;
399 Tk = ri[WS(rs, 8)];
400 Tm = ii[WS(rs, 8)];
401 Tj = W[14];
402 Tl = W[15];
403 Tn = FMA(Tj, Tk, Tl * Tm);
404 T17 = FNMS(Tl, Tk, Tj * Tm);
405 }
406 Ti = Tc - Th;
407 Tt = Tn - Ts;
408 Tu = Ti + Tt;
409 T1r = T14 + T15;
410 T1s = T17 + T18;
411 T1x = T1r + T1s;
412 TU = Tc + Th;
413 TV = Tn + Ts;
414 TW = TU + TV;
415 T16 = T14 - T15;
416 T19 = T17 - T18;
417 T1L = T16 + T19;
418 }
419 {
420 E T11, TS, T12, T1i, T1k, T1a, T1h, T1j, T13;
421 T11 = KP559016994 * (Tu - TR);
422 TS = Tu + TR;
423 T12 = FNMS(KP250000000, TS, T7);
424 T1a = T16 - T19;
425 T1h = T1d - T1g;
426 T1i = FMA(KP951056516, T1a, KP587785252 * T1h);
427 T1k = FNMS(KP587785252, T1a, KP951056516 * T1h);
428 ri[WS(rs, 5)] = T7 + TS;
429 T1j = T12 - T11;
430 ri[WS(rs, 7)] = T1j - T1k;
431 ri[WS(rs, 3)] = T1j + T1k;
432 T13 = T11 + T12;
433 ri[WS(rs, 9)] = T13 - T1i;
434 ri[WS(rs, 1)] = T13 + T1i;
435 }
436 {
437 E T1N, T1P, T1Q, T1U, T1W, T1S, T1T, T1V, T1R;
438 T1N = KP559016994 * (T1L - T1M);
439 T1P = T1L + T1M;
440 T1Q = FNMS(KP250000000, T1P, T1O);
441 T1S = Ti - Tt;
442 T1T = TF - TQ;
443 T1U = FMA(KP951056516, T1S, KP587785252 * T1T);
444 T1W = FNMS(KP587785252, T1S, KP951056516 * T1T);
445 ii[WS(rs, 5)] = T1P + T1O;
446 T1V = T1Q - T1N;
447 ii[WS(rs, 3)] = T1V - T1W;
448 ii[WS(rs, 7)] = T1W + T1V;
449 T1R = T1N + T1Q;
450 ii[WS(rs, 1)] = T1R - T1U;
451 ii[WS(rs, 9)] = T1U + T1R;
452 }
453 {
454 E T1m, T10, T1l, T1u, T1w, T1q, T1t, T1v, T1n;
455 T1m = KP559016994 * (TW - TZ);
456 T10 = TW + TZ;
457 T1l = FNMS(KP250000000, T10, TT);
458 T1q = T1o - T1p;
459 T1t = T1r - T1s;
460 T1u = FNMS(KP587785252, T1t, KP951056516 * T1q);
461 T1w = FMA(KP951056516, T1t, KP587785252 * T1q);
462 ri[0] = TT + T10;
463 T1v = T1m + T1l;
464 ri[WS(rs, 4)] = T1v - T1w;
465 ri[WS(rs, 6)] = T1v + T1w;
466 T1n = T1l - T1m;
467 ri[WS(rs, 2)] = T1n - T1u;
468 ri[WS(rs, 8)] = T1n + T1u;
469 }
470 {
471 E T1H, T1z, T1G, T1F, T1J, T1D, T1E, T1K, T1I;
472 T1H = KP559016994 * (T1x - T1y);
473 T1z = T1x + T1y;
474 T1G = FNMS(KP250000000, T1z, T1C);
475 T1D = TX - TY;
476 T1E = TU - TV;
477 T1F = FNMS(KP587785252, T1E, KP951056516 * T1D);
478 T1J = FMA(KP951056516, T1E, KP587785252 * T1D);
479 ii[0] = T1z + T1C;
480 T1K = T1H + T1G;
481 ii[WS(rs, 4)] = T1J + T1K;
482 ii[WS(rs, 6)] = T1K - T1J;
483 T1I = T1G - T1H;
484 ii[WS(rs, 2)] = T1F + T1I;
485 ii[WS(rs, 8)] = T1I - T1F;
486 }
487 }
488 }
489 }
490
491 static const tw_instr twinstr[] = {
492 {TW_FULL, 0, 10},
493 {TW_NEXT, 1, 0}
494 };
495
496 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {72, 30, 30, 0}, 0, 0, 0 };
497
498 void X(codelet_t1_10) (planner *p) {
499 X(kdft_dit_register) (p, t1_10, &desc);
500 }
501 #endif /* HAVE_FMA */