Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cf/hf_9.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 07:39:51 EST 2012 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include hf.h */
29
30 /*
31 * This function contains 96 FP additions, 88 FP multiplications,
32 * (or, 24 additions, 16 multiplications, 72 fused multiply/add),
33 * 69 stack variables, 10 constants, and 36 memory accesses
34 */
35 #include "hf.h"
36
37 static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP777861913, +0.777861913430206160028177977318626690410586096);
40 DK(KP852868531, +0.852868531952443209628250963940074071936020296);
41 DK(KP839099631, +0.839099631177280011763127298123181364687434283);
42 DK(KP492403876, +0.492403876506104029683371512294761506835321626);
43 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
44 DK(KP954188894, +0.954188894138671133499268364187245676532219158);
45 DK(KP363970234, +0.363970234266202361351047882776834043890471784);
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 - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
52 E T20, T1Z;
53 {
54 E T1, T1P, T1Q, T10, T1S, Te, TB, T1d, T1a, T19, T1M, TE, T1c, Tz, T1n;
55 E TC, TH, TK, T1k, TR, TG, TJ, TD;
56 T1 = cr[0];
57 T1P = ci[0];
58 {
59 E T9, Tc, TY, Ta, Tb, TX, T7;
60 {
61 E T3, T6, T8, TW, T4, T2, T5;
62 T3 = cr[WS(rs, 3)];
63 T6 = ci[WS(rs, 3)];
64 T2 = W[4];
65 T9 = cr[WS(rs, 6)];
66 Tc = ci[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, T12, Tn, Tq, Tp, T17, Tx, T14, To, Tj, TZ, Td, Tg;
79 E TA, Tl, Ty;
80 Th = cr[WS(rs, 1)];
81 TZ = FNMS(Tb, T9, TY);
82 Td = FMA(Tb, Tc, Ta);
83 Tk = ci[WS(rs, 1)];
84 Tg = W[0];
85 T1Q = TX + TZ;
86 T10 = TX - TZ;
87 T1S = Td - T7;
88 Te = T7 + Td;
89 Ti = Tg * Th;
90 T12 = Tg * Tk;
91 {
92 E Tt, Tw, Ts, Tv, T16, Tu, Tm;
93 Tt = cr[WS(rs, 7)];
94 Tw = ci[WS(rs, 7)];
95 Ts = W[12];
96 Tv = W[13];
97 Tn = cr[WS(rs, 4)];
98 Tq = ci[WS(rs, 4)];
99 T16 = Ts * Tw;
100 Tu = Ts * Tt;
101 Tm = W[6];
102 Tp = W[7];
103 T17 = FNMS(Tv, Tt, T16);
104 Tx = FMA(Tv, Tw, Tu);
105 T14 = Tm * Tq;
106 To = Tm * Tn;
107 }
108 Tj = W[1];
109 TB = cr[WS(rs, 2)];
110 {
111 E T15, Tr, T13, T18;
112 T15 = FNMS(Tp, Tn, T14);
113 Tr = FMA(Tp, Tq, To);
114 T13 = FNMS(Tj, Th, T12);
115 Tl = FMA(Tj, Tk, Ti);
116 T18 = T15 + T17;
117 T1d = T15 - T17;
118 Ty = Tr + Tx;
119 T1a = Tr - Tx;
120 T19 = FNMS(KP500000000, T18, T13);
121 T1M = T13 + T18;
122 TE = ci[WS(rs, 2)];
123 }
124 T1c = FNMS(KP500000000, Ty, Tl);
125 Tz = Tl + Ty;
126 TA = W[2];
127 {
128 E TN, TQ, TP, T1j, TO, TM;
129 TN = cr[WS(rs, 8)];
130 TQ = ci[WS(rs, 8)];
131 TM = W[14];
132 T1n = TA * TE;
133 TC = TA * TB;
134 TP = W[15];
135 T1j = TM * TQ;
136 TO = TM * TN;
137 TH = cr[WS(rs, 5)];
138 TK = ci[WS(rs, 5)];
139 T1k = FNMS(TP, TN, T1j);
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, T21, T1R, T1l, T1r, T1q, T1N, TT, T1g;
149 {
150 E T1o, TF, T1i, TL, T1h, TI, TS, T1p;
151 TV = FNMS(KP500000000, Te, T1);
152 Tf = T1 + Te;
153 T1h = TG * TK;
154 TI = TG * TH;
155 T1o = FNMS(TD, TB, T1n);
156 TF = FMA(TD, TE, TC);
157 T1i = FNMS(TJ, TH, T1h);
158 TL = FMA(TJ, TK, TI);
159 T21 = T1Q + T1P;
160 T1R = FNMS(KP500000000, T1Q, T1P);
161 T1p = T1i + T1k;
162 T1l = T1i - T1k;
163 TS = TL + TR;
164 T1r = TR - TL;
165 T1q = FNMS(KP500000000, T1p, T1o);
166 T1N = T1o + T1p;
167 TT = TF + TS;
168 T1g = FNMS(KP500000000, TS, TF);
169 }
170 {
171 E T11, T1z, T1E, T1D, T1X, T1T, T1I, T1C, T1Y, T1y, T1u, T24, TU;
172 T24 = TT - Tz;
173 TU = Tz + TT;
174 {
175 E T22, T1O, T1L, T23;
176 T22 = T1M + T1N;
177 T1O = T1M - T1N;
178 T11 = FNMS(KP866025403, T10, TV);
179 T1z = FMA(KP866025403, T10, TV);
180 T1L = FNMS(KP500000000, TU, Tf);
181 cr[0] = Tf + TU;
182 T23 = FNMS(KP500000000, T22, T21);
183 ci[WS(rs, 8)] = T22 + T21;
184 cr[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
185 ci[WS(rs, 2)] = FNMS(KP866025403, T1O, T1L);
186 ci[WS(rs, 5)] = FMA(KP866025403, T24, T23);
187 cr[WS(rs, 6)] = FMS(KP866025403, T24, T23);
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 = FNMS(KP866025403, T1d, T1c);
194 T1D = FMA(KP866025403, T1d, T1c);
195 T1B = FMA(KP866025403, T1l, T1g);
196 T1m = FNMS(KP866025403, T1l, T1g);
197 T1X = FNMS(KP866025403, T1S, T1R);
198 T1T = FMA(KP866025403, T1S, T1R);
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 = FMA(KP363970234, T1m, T1s);
204 T1t = FNMS(KP363970234, T1s, T1m);
205 T1I = FNMS(KP176326980, T1A, T1B);
206 T1C = FMA(KP176326980, T1B, T1A);
207 T1Y = FMA(KP954188894, T1x, T1w);
208 T1y = FNMS(KP954188894, T1x, T1w);
209 T20 = FMA(KP954188894, T1t, T1f);
210 T1u = FNMS(KP954188894, T1t, T1f);
211 }
212 {
213 E T1F, T1J, T1v, T1U, T1K;
214 ci[WS(rs, 6)] = FNMS(KP984807753, T1Y, T1X);
215 T1v = FNMS(KP492403876, T1u, T11);
216 cr[WS(rs, 2)] = FMA(KP984807753, T1u, T11);
217 T1F = FMA(KP839099631, T1E, T1D);
218 T1J = FNMS(KP839099631, T1D, T1E);
219 ci[WS(rs, 3)] = FNMS(KP852868531, T1y, T1v);
220 ci[0] = FMA(KP852868531, T1y, T1v);
221 T1U = FNMS(KP777861913, T1J, T1I);
222 T1K = FMA(KP777861913, T1J, T1I);
223 {
224 E T1G, T1W, T1V, T1H;
225 T1G = FMA(KP777861913, T1F, T1C);
226 T1W = FNMS(KP777861913, T1F, T1C);
227 T1Z = FMA(KP492403876, T1Y, T1X);
228 T1V = FMA(KP492403876, T1U, T1T);
229 ci[WS(rs, 7)] = FNMS(KP984807753, T1U, T1T);
230 T1H = FNMS(KP492403876, T1G, T1z);
231 cr[WS(rs, 1)] = FMA(KP984807753, T1G, T1z);
232 ci[WS(rs, 4)] = FMA(KP852868531, T1W, T1V);
233 cr[WS(rs, 7)] = FMS(KP852868531, T1W, T1V);
234 cr[WS(rs, 4)] = FMA(KP852868531, T1K, T1H);
235 ci[WS(rs, 1)] = FNMS(KP852868531, T1K, T1H);
236 }
237 }
238 }
239 }
240 }
241 cr[WS(rs, 8)] = -(FMA(KP852868531, T20, T1Z));
242 cr[WS(rs, 5)] = FMS(KP852868531, T20, T1Z);
243 }
244 }
245 }
246
247 static const tw_instr twinstr[] = {
248 {TW_FULL, 1, 9},
249 {TW_NEXT, 1, 0}
250 };
251
252 static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {24, 16, 72, 0} };
253
254 void X(codelet_hf_9) (planner *p) {
255 X(khc2hc_register) (p, hf_9, &desc);
256 }
257 #else /* HAVE_FMA */
258
259 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include hf.h */
260
261 /*
262 * This function contains 96 FP additions, 72 FP multiplications,
263 * (or, 60 additions, 36 multiplications, 36 fused multiply/add),
264 * 41 stack variables, 8 constants, and 36 memory accesses
265 */
266 #include "hf.h"
267
268 static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
269 {
270 DK(KP642787609, +0.642787609686539326322643409907263432907559884);
271 DK(KP766044443, +0.766044443118978035202392650555416673935832457);
272 DK(KP939692620, +0.939692620785908384054109277324731469936208134);
273 DK(KP342020143, +0.342020143325668733044099614682259580763083368);
274 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
275 DK(KP173648177, +0.173648177666930348851716626769314796000375677);
276 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
277 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
278 {
279 INT m;
280 for (m = mb, W = W + ((mb - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
281 E T1, T1B, TQ, T1A, Tc, TN, T1C, T1D, TL, T1x, T19, T1o, T1c, T1n, Tu;
282 E T1w, TW, T1k, T11, T1l;
283 {
284 E T6, TO, Tb, TP;
285 T1 = cr[0];
286 T1B = ci[0];
287 {
288 E T3, T5, T2, T4;
289 T3 = cr[WS(rs, 3)];
290 T5 = ci[WS(rs, 3)];
291 T2 = W[4];
292 T4 = W[5];
293 T6 = FMA(T2, T3, T4 * T5);
294 TO = FNMS(T4, T3, T2 * T5);
295 }
296 {
297 E T8, Ta, T7, T9;
298 T8 = cr[WS(rs, 6)];
299 Ta = ci[WS(rs, 6)];
300 T7 = W[10];
301 T9 = W[11];
302 Tb = FMA(T7, T8, T9 * Ta);
303 TP = FNMS(T9, T8, T7 * Ta);
304 }
305 TQ = KP866025403 * (TO - TP);
306 T1A = KP866025403 * (Tb - T6);
307 Tc = T6 + Tb;
308 TN = FNMS(KP500000000, Tc, T1);
309 T1C = TO + TP;
310 T1D = FNMS(KP500000000, T1C, T1B);
311 }
312 {
313 E Tz, T13, TE, T14, TJ, T15, TK, T16;
314 {
315 E Tw, Ty, Tv, Tx;
316 Tw = cr[WS(rs, 2)];
317 Ty = ci[WS(rs, 2)];
318 Tv = W[2];
319 Tx = W[3];
320 Tz = FMA(Tv, Tw, Tx * Ty);
321 T13 = FNMS(Tx, Tw, Tv * Ty);
322 }
323 {
324 E TB, TD, TA, TC;
325 TB = cr[WS(rs, 5)];
326 TD = ci[WS(rs, 5)];
327 TA = W[8];
328 TC = W[9];
329 TE = FMA(TA, TB, TC * TD);
330 T14 = FNMS(TC, TB, TA * TD);
331 }
332 {
333 E TG, TI, TF, TH;
334 TG = cr[WS(rs, 8)];
335 TI = ci[WS(rs, 8)];
336 TF = W[14];
337 TH = W[15];
338 TJ = FMA(TF, TG, TH * TI);
339 T15 = FNMS(TH, TG, TF * TI);
340 }
341 TK = TE + TJ;
342 T16 = T14 + T15;
343 TL = Tz + TK;
344 T1x = T13 + T16;
345 {
346 E T17, T18, T1a, T1b;
347 T17 = FNMS(KP500000000, T16, T13);
348 T18 = KP866025403 * (TJ - TE);
349 T19 = T17 - T18;
350 T1o = T18 + T17;
351 T1a = FNMS(KP500000000, TK, Tz);
352 T1b = KP866025403 * (T14 - T15);
353 T1c = T1a - T1b;
354 T1n = T1a + T1b;
355 }
356 }
357 {
358 E Ti, TX, Tn, TT, Ts, TU, Tt, TY;
359 {
360 E Tf, Th, Te, Tg;
361 Tf = cr[WS(rs, 1)];
362 Th = ci[WS(rs, 1)];
363 Te = W[0];
364 Tg = W[1];
365 Ti = FMA(Te, Tf, Tg * Th);
366 TX = FNMS(Tg, Tf, Te * Th);
367 }
368 {
369 E Tk, Tm, Tj, Tl;
370 Tk = cr[WS(rs, 4)];
371 Tm = ci[WS(rs, 4)];
372 Tj = W[6];
373 Tl = W[7];
374 Tn = FMA(Tj, Tk, Tl * Tm);
375 TT = FNMS(Tl, Tk, Tj * Tm);
376 }
377 {
378 E Tp, Tr, To, Tq;
379 Tp = cr[WS(rs, 7)];
380 Tr = ci[WS(rs, 7)];
381 To = W[12];
382 Tq = W[13];
383 Ts = FMA(To, Tp, Tq * Tr);
384 TU = FNMS(Tq, Tp, To * Tr);
385 }
386 Tt = Tn + Ts;
387 TY = TT + TU;
388 Tu = Ti + Tt;
389 T1w = TX + TY;
390 {
391 E TS, TV, TZ, T10;
392 TS = FNMS(KP500000000, Tt, Ti);
393 TV = KP866025403 * (TT - TU);
394 TW = TS - TV;
395 T1k = TS + TV;
396 TZ = FNMS(KP500000000, TY, TX);
397 T10 = KP866025403 * (Ts - Tn);
398 T11 = TZ - T10;
399 T1l = T10 + TZ;
400 }
401 }
402 {
403 E T1y, Td, TM, T1v;
404 T1y = KP866025403 * (T1w - T1x);
405 Td = T1 + Tc;
406 TM = Tu + TL;
407 T1v = FNMS(KP500000000, TM, Td);
408 cr[0] = Td + TM;
409 cr[WS(rs, 3)] = T1v + T1y;
410 ci[WS(rs, 2)] = T1v - T1y;
411 }
412 {
413 E TR, T1I, T1e, T1K, T1i, T1H, T1f, T1J;
414 TR = TN - TQ;
415 T1I = T1D - T1A;
416 {
417 E T12, T1d, T1g, T1h;
418 T12 = FMA(KP173648177, TW, KP984807753 * T11);
419 T1d = FNMS(KP939692620, T1c, KP342020143 * T19);
420 T1e = T12 + T1d;
421 T1K = KP866025403 * (T1d - T12);
422 T1g = FNMS(KP984807753, TW, KP173648177 * T11);
423 T1h = FMA(KP342020143, T1c, KP939692620 * T19);
424 T1i = KP866025403 * (T1g + T1h);
425 T1H = T1g - T1h;
426 }
427 cr[WS(rs, 2)] = TR + T1e;
428 ci[WS(rs, 6)] = T1H + T1I;
429 T1f = FNMS(KP500000000, T1e, TR);
430 ci[0] = T1f - T1i;
431 ci[WS(rs, 3)] = T1f + T1i;
432 T1J = FMS(KP500000000, T1H, T1I);
433 cr[WS(rs, 5)] = T1J - T1K;
434 cr[WS(rs, 8)] = T1K + T1J;
435 }
436 {
437 E T1L, T1M, T1N, T1O;
438 T1L = KP866025403 * (TL - Tu);
439 T1M = T1C + T1B;
440 T1N = T1w + T1x;
441 T1O = FNMS(KP500000000, T1N, T1M);
442 cr[WS(rs, 6)] = T1L - T1O;
443 ci[WS(rs, 8)] = T1N + T1M;
444 ci[WS(rs, 5)] = T1L + T1O;
445 }
446 {
447 E T1j, T1E, T1q, T1z, T1u, T1F, T1r, T1G;
448 T1j = TN + TQ;
449 T1E = T1A + T1D;
450 {
451 E T1m, T1p, T1s, T1t;
452 T1m = FMA(KP766044443, T1k, KP642787609 * T1l);
453 T1p = FMA(KP173648177, T1n, KP984807753 * T1o);
454 T1q = T1m + T1p;
455 T1z = KP866025403 * (T1p - T1m);
456 T1s = FNMS(KP642787609, T1k, KP766044443 * T1l);
457 T1t = FNMS(KP984807753, T1n, KP173648177 * T1o);
458 T1u = KP866025403 * (T1s - T1t);
459 T1F = T1s + T1t;
460 }
461 cr[WS(rs, 1)] = T1j + T1q;
462 T1r = FNMS(KP500000000, T1q, T1j);
463 ci[WS(rs, 1)] = T1r - T1u;
464 cr[WS(rs, 4)] = T1r + T1u;
465 ci[WS(rs, 7)] = T1F + T1E;
466 T1G = FNMS(KP500000000, T1F, T1E);
467 cr[WS(rs, 7)] = T1z - T1G;
468 ci[WS(rs, 4)] = T1z + T1G;
469 }
470 }
471 }
472 }
473
474 static const tw_instr twinstr[] = {
475 {TW_FULL, 1, 9},
476 {TW_NEXT, 1, 0}
477 };
478
479 static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {60, 36, 36, 0} };
480
481 void X(codelet_hf_9) (planner *p) {
482 X(khc2hc_register) (p, hf_9, &desc);
483 }
484 #endif /* HAVE_FMA */