Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t2_10.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 07:36:09 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 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include t.h */
29
30 /*
31 * This function contains 114 FP additions, 94 FP multiplications,
32 * (or, 48 additions, 28 multiplications, 66 fused multiply/add),
33 * 85 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "t.h"
36
37 static void t2_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 * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) {
46 E T27, T2b, T2a, T2c;
47 {
48 E T2, T3, T8, Tc, T5, T4, TX, T11, TE, T6, TB, TA;
49 T2 = W[0];
50 T3 = W[2];
51 T8 = W[4];
52 Tc = W[5];
53 T5 = W[1];
54 T4 = T2 * T3;
55 TX = T3 * T8;
56 TA = T2 * T8;
57 T11 = T3 * Tc;
58 TE = T2 * Tc;
59 T6 = W[3];
60 TB = FMA(T5, Tc, TA);
61 {
62 E T2d, T24, T1c, Tk, T1i, T28, T2l, T1a, T2f, T1I, T1R, T1Z, TL, T1v, T1d;
63 E Tz, T1S, T1r, TH, T1t;
64 {
65 E T1, TF, TY, T12, Tl, T7, T23, To, Tb, Te, Ti, Th, Td, Tw, Ts;
66 E Ta;
67 T1 = ri[0];
68 TF = FNMS(T5, T8, TE);
69 TY = FMA(T6, Tc, TX);
70 T12 = FNMS(T6, T8, T11);
71 Tl = FMA(T5, T6, T4);
72 T7 = FNMS(T5, T6, T4);
73 Ta = T2 * T6;
74 T23 = ii[0];
75 {
76 E Tg, T9, Tv, Tr;
77 Tg = T7 * Tc;
78 T9 = T7 * T8;
79 Tv = Tl * Tc;
80 Tr = Tl * T8;
81 To = FNMS(T5, T3, Ta);
82 Tb = FMA(T5, T3, Ta);
83 Te = ri[WS(rs, 5)];
84 Ti = ii[WS(rs, 5)];
85 Th = FNMS(Tb, T8, Tg);
86 Td = FMA(Tb, Tc, T9);
87 Tw = FNMS(To, T8, Tv);
88 Ts = FMA(To, Tc, Tr);
89 }
90 {
91 E T18, T1G, T1g, TW, T1P, T1C, T14, T1E;
92 {
93 E TR, T1z, TV, T1B, TZ, T13, T15, T17, T10, T1D;
94 {
95 E TO, TQ, TP, T22, Tj, T1y, T21, Tf;
96 TO = ri[WS(rs, 4)];
97 T21 = Td * Ti;
98 Tf = Td * Te;
99 TQ = ii[WS(rs, 4)];
100 TP = T7 * TO;
101 T22 = FNMS(Th, Te, T21);
102 Tj = FMA(Th, Ti, Tf);
103 T1y = T7 * TQ;
104 TR = FMA(Tb, TQ, TP);
105 T2d = T23 - T22;
106 T24 = T22 + T23;
107 T1c = T1 + Tj;
108 Tk = T1 - Tj;
109 T1z = FNMS(Tb, TO, T1y);
110 }
111 T15 = ri[WS(rs, 1)];
112 T17 = ii[WS(rs, 1)];
113 {
114 E TS, TU, T16, T1F, TT, T1A;
115 TS = ri[WS(rs, 9)];
116 TU = ii[WS(rs, 9)];
117 T16 = T2 * T15;
118 T1F = T2 * T17;
119 TT = T8 * TS;
120 T1A = T8 * TU;
121 T18 = FMA(T5, T17, T16);
122 T1G = FNMS(T5, T15, T1F);
123 TV = FMA(Tc, TU, TT);
124 T1B = FNMS(Tc, TS, T1A);
125 }
126 TZ = ri[WS(rs, 6)];
127 T13 = ii[WS(rs, 6)];
128 T1g = TR + TV;
129 TW = TR - TV;
130 T1P = T1z + T1B;
131 T1C = T1z - T1B;
132 T10 = TY * TZ;
133 T1D = TY * T13;
134 T14 = FMA(T12, T13, T10);
135 T1E = FNMS(T12, TZ, T1D);
136 }
137 {
138 E Tq, T1o, Ty, TC, TG, T1q, TD, T1s;
139 {
140 E TI, TK, Tt, T1p;
141 {
142 E Tm, T1n, Tp, Tn;
143 Tm = ri[WS(rs, 2)];
144 Tp = ii[WS(rs, 2)];
145 {
146 E T19, T1h, T1Q, T1H;
147 T19 = T14 - T18;
148 T1h = T14 + T18;
149 T1Q = T1E + T1G;
150 T1H = T1E - T1G;
151 Tn = Tl * Tm;
152 T1i = T1g + T1h;
153 T28 = T1g - T1h;
154 T2l = TW - T19;
155 T1a = TW + T19;
156 T2f = T1C + T1H;
157 T1I = T1C - T1H;
158 T1R = T1P - T1Q;
159 T1Z = T1P + T1Q;
160 T1n = Tl * Tp;
161 }
162 Tq = FMA(To, Tp, Tn);
163 TI = ri[WS(rs, 3)];
164 TK = ii[WS(rs, 3)];
165 T1o = FNMS(To, Tm, T1n);
166 }
167 {
168 E Tx, Tu, TJ, T1u;
169 Tt = ri[WS(rs, 7)];
170 TJ = T3 * TI;
171 T1u = T3 * TK;
172 Tx = ii[WS(rs, 7)];
173 Tu = Ts * Tt;
174 TL = FMA(T6, TK, TJ);
175 T1v = FNMS(T6, TI, T1u);
176 T1p = Ts * Tx;
177 Ty = FMA(Tw, Tx, Tu);
178 }
179 TC = ri[WS(rs, 8)];
180 TG = ii[WS(rs, 8)];
181 T1q = FNMS(Tw, Tt, T1p);
182 }
183 T1d = Tq + Ty;
184 Tz = Tq - Ty;
185 TD = TB * TC;
186 T1s = TB * TG;
187 T1S = T1o + T1q;
188 T1r = T1o - T1q;
189 TH = FMA(TF, TG, TD);
190 T1t = FNMS(TF, TC, T1s);
191 }
192 }
193 }
194 {
195 E T1f, T29, T1Y, T1U, T2j, T2n, T2m, T2o;
196 {
197 E T2k, T2e, T1l, T1L, T1J, T1k, T1b, T1e, TM;
198 T1e = TH + TL;
199 TM = TH - TL;
200 {
201 E T1w, T1T, TN, T1x;
202 T1w = T1t - T1v;
203 T1T = T1t + T1v;
204 T1f = T1d + T1e;
205 T29 = T1d - T1e;
206 T2k = Tz - TM;
207 TN = Tz + TM;
208 T1x = T1r - T1w;
209 T2e = T1r + T1w;
210 T1Y = T1S + T1T;
211 T1U = T1S - T1T;
212 T1l = TN - T1a;
213 T1b = TN + T1a;
214 T1L = FNMS(KP618033988, T1x, T1I);
215 T1J = FMA(KP618033988, T1I, T1x);
216 }
217 T1k = FNMS(KP250000000, T1b, Tk);
218 ri[WS(rs, 5)] = Tk + T1b;
219 {
220 E T2g, T2i, T2h, T1K, T1m;
221 T2g = T2e + T2f;
222 T2i = T2e - T2f;
223 T1K = FNMS(KP559016994, T1l, T1k);
224 T1m = FMA(KP559016994, T1l, T1k);
225 T2h = FNMS(KP250000000, T2g, T2d);
226 ri[WS(rs, 1)] = FMA(KP951056516, T1J, T1m);
227 ri[WS(rs, 9)] = FNMS(KP951056516, T1J, T1m);
228 ri[WS(rs, 3)] = FMA(KP951056516, T1L, T1K);
229 ri[WS(rs, 7)] = FNMS(KP951056516, T1L, T1K);
230 ii[WS(rs, 5)] = T2g + T2d;
231 T2j = FMA(KP559016994, T2i, T2h);
232 T2n = FNMS(KP559016994, T2i, T2h);
233 T2m = FMA(KP618033988, T2l, T2k);
234 T2o = FNMS(KP618033988, T2k, T2l);
235 }
236 }
237 {
238 E T1O, T1W, T1V, T1X, T1j, T1N, T1M, T20, T26, T25;
239 T1j = T1f + T1i;
240 T1N = T1f - T1i;
241 ii[WS(rs, 7)] = FMA(KP951056516, T2o, T2n);
242 ii[WS(rs, 3)] = FNMS(KP951056516, T2o, T2n);
243 ii[WS(rs, 9)] = FMA(KP951056516, T2m, T2j);
244 ii[WS(rs, 1)] = FNMS(KP951056516, T2m, T2j);
245 T1M = FNMS(KP250000000, T1j, T1c);
246 ri[0] = T1c + T1j;
247 T1O = FNMS(KP559016994, T1N, T1M);
248 T1W = FMA(KP559016994, T1N, T1M);
249 T1V = FNMS(KP618033988, T1U, T1R);
250 T1X = FMA(KP618033988, T1R, T1U);
251 T20 = T1Y + T1Z;
252 T26 = T1Y - T1Z;
253 ri[WS(rs, 6)] = FMA(KP951056516, T1X, T1W);
254 ri[WS(rs, 4)] = FNMS(KP951056516, T1X, T1W);
255 ri[WS(rs, 8)] = FMA(KP951056516, T1V, T1O);
256 ri[WS(rs, 2)] = FNMS(KP951056516, T1V, T1O);
257 T25 = FNMS(KP250000000, T20, T24);
258 ii[0] = T20 + T24;
259 T27 = FNMS(KP559016994, T26, T25);
260 T2b = FMA(KP559016994, T26, T25);
261 T2a = FNMS(KP618033988, T29, T28);
262 T2c = FMA(KP618033988, T28, T29);
263 }
264 }
265 }
266 }
267 ii[WS(rs, 6)] = FNMS(KP951056516, T2c, T2b);
268 ii[WS(rs, 4)] = FMA(KP951056516, T2c, T2b);
269 ii[WS(rs, 8)] = FNMS(KP951056516, T2a, T27);
270 ii[WS(rs, 2)] = FMA(KP951056516, T2a, T27);
271 }
272 }
273 }
274
275 static const tw_instr twinstr[] = {
276 {TW_CEXP, 0, 1},
277 {TW_CEXP, 0, 3},
278 {TW_CEXP, 0, 9},
279 {TW_NEXT, 1, 0}
280 };
281
282 static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {48, 28, 66, 0}, 0, 0, 0 };
283
284 void X(codelet_t2_10) (planner *p) {
285 X(kdft_dit_register) (p, t2_10, &desc);
286 }
287 #else /* HAVE_FMA */
288
289 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include t.h */
290
291 /*
292 * This function contains 114 FP additions, 80 FP multiplications,
293 * (or, 76 additions, 42 multiplications, 38 fused multiply/add),
294 * 63 stack variables, 4 constants, and 40 memory accesses
295 */
296 #include "t.h"
297
298 static void t2_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
299 {
300 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
301 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
302 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
303 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
304 {
305 INT m;
306 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) {
307 E T2, T5, T3, T6, T8, Tm, Tc, Tk, T9, Td, Te, TM, TO, Tg, Tp;
308 E Tv, Tx, Tr;
309 {
310 E T4, Tb, T7, Ta;
311 T2 = W[0];
312 T5 = W[1];
313 T3 = W[2];
314 T6 = W[3];
315 T4 = T2 * T3;
316 Tb = T5 * T3;
317 T7 = T5 * T6;
318 Ta = T2 * T6;
319 T8 = T4 - T7;
320 Tm = Ta - Tb;
321 Tc = Ta + Tb;
322 Tk = T4 + T7;
323 T9 = W[4];
324 Td = W[5];
325 Te = FMA(T8, T9, Tc * Td);
326 TM = FMA(T3, T9, T6 * Td);
327 TO = FNMS(T6, T9, T3 * Td);
328 Tg = FNMS(Tc, T9, T8 * Td);
329 Tp = FMA(Tk, T9, Tm * Td);
330 Tv = FMA(T2, T9, T5 * Td);
331 Tx = FNMS(T5, T9, T2 * Td);
332 Tr = FNMS(Tm, T9, Tk * Td);
333 }
334 {
335 E Tj, T1S, TX, T1G, TL, TU, TV, T1s, T1t, T1C, T11, T12, T13, T1h, T1k;
336 E T1Q, Tu, TD, TE, T1v, T1w, T1B, TY, TZ, T10, T1a, T1d, T1P;
337 {
338 E T1, T1F, Ti, T1E, Tf, Th;
339 T1 = ri[0];
340 T1F = ii[0];
341 Tf = ri[WS(rs, 5)];
342 Th = ii[WS(rs, 5)];
343 Ti = FMA(Te, Tf, Tg * Th);
344 T1E = FNMS(Tg, Tf, Te * Th);
345 Tj = T1 - Ti;
346 T1S = T1F - T1E;
347 TX = T1 + Ti;
348 T1G = T1E + T1F;
349 }
350 {
351 E TH, T1f, TT, T1j, TK, T1g, TQ, T1i;
352 {
353 E TF, TG, TR, TS;
354 TF = ri[WS(rs, 4)];
355 TG = ii[WS(rs, 4)];
356 TH = FMA(T8, TF, Tc * TG);
357 T1f = FNMS(Tc, TF, T8 * TG);
358 TR = ri[WS(rs, 1)];
359 TS = ii[WS(rs, 1)];
360 TT = FMA(T2, TR, T5 * TS);
361 T1j = FNMS(T5, TR, T2 * TS);
362 }
363 {
364 E TI, TJ, TN, TP;
365 TI = ri[WS(rs, 9)];
366 TJ = ii[WS(rs, 9)];
367 TK = FMA(T9, TI, Td * TJ);
368 T1g = FNMS(Td, TI, T9 * TJ);
369 TN = ri[WS(rs, 6)];
370 TP = ii[WS(rs, 6)];
371 TQ = FMA(TM, TN, TO * TP);
372 T1i = FNMS(TO, TN, TM * TP);
373 }
374 TL = TH - TK;
375 TU = TQ - TT;
376 TV = TL + TU;
377 T1s = T1f + T1g;
378 T1t = T1i + T1j;
379 T1C = T1s + T1t;
380 T11 = TH + TK;
381 T12 = TQ + TT;
382 T13 = T11 + T12;
383 T1h = T1f - T1g;
384 T1k = T1i - T1j;
385 T1Q = T1h + T1k;
386 }
387 {
388 E To, T18, TC, T1c, Tt, T19, Tz, T1b;
389 {
390 E Tl, Tn, TA, TB;
391 Tl = ri[WS(rs, 2)];
392 Tn = ii[WS(rs, 2)];
393 To = FMA(Tk, Tl, Tm * Tn);
394 T18 = FNMS(Tm, Tl, Tk * Tn);
395 TA = ri[WS(rs, 3)];
396 TB = ii[WS(rs, 3)];
397 TC = FMA(T3, TA, T6 * TB);
398 T1c = FNMS(T6, TA, T3 * TB);
399 }
400 {
401 E Tq, Ts, Tw, Ty;
402 Tq = ri[WS(rs, 7)];
403 Ts = ii[WS(rs, 7)];
404 Tt = FMA(Tp, Tq, Tr * Ts);
405 T19 = FNMS(Tr, Tq, Tp * Ts);
406 Tw = ri[WS(rs, 8)];
407 Ty = ii[WS(rs, 8)];
408 Tz = FMA(Tv, Tw, Tx * Ty);
409 T1b = FNMS(Tx, Tw, Tv * Ty);
410 }
411 Tu = To - Tt;
412 TD = Tz - TC;
413 TE = Tu + TD;
414 T1v = T18 + T19;
415 T1w = T1b + T1c;
416 T1B = T1v + T1w;
417 TY = To + Tt;
418 TZ = Tz + TC;
419 T10 = TY + TZ;
420 T1a = T18 - T19;
421 T1d = T1b - T1c;
422 T1P = T1a + T1d;
423 }
424 {
425 E T15, TW, T16, T1m, T1o, T1e, T1l, T1n, T17;
426 T15 = KP559016994 * (TE - TV);
427 TW = TE + TV;
428 T16 = FNMS(KP250000000, TW, Tj);
429 T1e = T1a - T1d;
430 T1l = T1h - T1k;
431 T1m = FMA(KP951056516, T1e, KP587785252 * T1l);
432 T1o = FNMS(KP587785252, T1e, KP951056516 * T1l);
433 ri[WS(rs, 5)] = Tj + TW;
434 T1n = T16 - T15;
435 ri[WS(rs, 7)] = T1n - T1o;
436 ri[WS(rs, 3)] = T1n + T1o;
437 T17 = T15 + T16;
438 ri[WS(rs, 9)] = T17 - T1m;
439 ri[WS(rs, 1)] = T17 + T1m;
440 }
441 {
442 E T1R, T1T, T1U, T1Y, T20, T1W, T1X, T1Z, T1V;
443 T1R = KP559016994 * (T1P - T1Q);
444 T1T = T1P + T1Q;
445 T1U = FNMS(KP250000000, T1T, T1S);
446 T1W = Tu - TD;
447 T1X = TL - TU;
448 T1Y = FMA(KP951056516, T1W, KP587785252 * T1X);
449 T20 = FNMS(KP587785252, T1W, KP951056516 * T1X);
450 ii[WS(rs, 5)] = T1T + T1S;
451 T1Z = T1U - T1R;
452 ii[WS(rs, 3)] = T1Z - T20;
453 ii[WS(rs, 7)] = T20 + T1Z;
454 T1V = T1R + T1U;
455 ii[WS(rs, 1)] = T1V - T1Y;
456 ii[WS(rs, 9)] = T1Y + T1V;
457 }
458 {
459 E T1q, T14, T1p, T1y, T1A, T1u, T1x, T1z, T1r;
460 T1q = KP559016994 * (T10 - T13);
461 T14 = T10 + T13;
462 T1p = FNMS(KP250000000, T14, TX);
463 T1u = T1s - T1t;
464 T1x = T1v - T1w;
465 T1y = FNMS(KP587785252, T1x, KP951056516 * T1u);
466 T1A = FMA(KP951056516, T1x, KP587785252 * T1u);
467 ri[0] = TX + T14;
468 T1z = T1q + T1p;
469 ri[WS(rs, 4)] = T1z - T1A;
470 ri[WS(rs, 6)] = T1z + T1A;
471 T1r = T1p - T1q;
472 ri[WS(rs, 2)] = T1r - T1y;
473 ri[WS(rs, 8)] = T1r + T1y;
474 }
475 {
476 E T1L, T1D, T1K, T1J, T1N, T1H, T1I, T1O, T1M;
477 T1L = KP559016994 * (T1B - T1C);
478 T1D = T1B + T1C;
479 T1K = FNMS(KP250000000, T1D, T1G);
480 T1H = T11 - T12;
481 T1I = TY - TZ;
482 T1J = FNMS(KP587785252, T1I, KP951056516 * T1H);
483 T1N = FMA(KP951056516, T1I, KP587785252 * T1H);
484 ii[0] = T1D + T1G;
485 T1O = T1L + T1K;
486 ii[WS(rs, 4)] = T1N + T1O;
487 ii[WS(rs, 6)] = T1O - T1N;
488 T1M = T1K - T1L;
489 ii[WS(rs, 2)] = T1J + T1M;
490 ii[WS(rs, 8)] = T1M - T1J;
491 }
492 }
493 }
494 }
495 }
496
497 static const tw_instr twinstr[] = {
498 {TW_CEXP, 0, 1},
499 {TW_CEXP, 0, 3},
500 {TW_CEXP, 0, 9},
501 {TW_NEXT, 1, 0}
502 };
503
504 static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {76, 42, 38, 0}, 0, 0, 0 };
505
506 void X(codelet_t2_10) (planner *p) {
507 X(kdft_dit_register) (p, t2_10, &desc);
508 }
509 #endif /* HAVE_FMA */