Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/n1_16.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 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 Thu May 24 08:04:11 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 16 -name n1_16 -include dft/scalar/n.h */
29
30 /*
31 * This function contains 144 FP additions, 40 FP multiplications,
32 * (or, 104 additions, 0 multiplications, 40 fused multiply/add),
33 * 50 stack variables, 3 constants, and 64 memory accesses
34 */
35 #include "dft/scalar/n.h"
36
37 static void n1_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT i;
44 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
45 E T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
46 E T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
47 E T1U, T1A;
48 {
49 E T3, TL, Ty, T1k, T6, T1j, TB, TM;
50 {
51 E T1, T2, Tw, Tx;
52 T1 = ri[0];
53 T2 = ri[WS(is, 8)];
54 T3 = T1 + T2;
55 TL = T1 - T2;
56 Tw = ii[0];
57 Tx = ii[WS(is, 8)];
58 Ty = Tw + Tx;
59 T1k = Tw - Tx;
60 }
61 {
62 E T4, T5, Tz, TA;
63 T4 = ri[WS(is, 4)];
64 T5 = ri[WS(is, 12)];
65 T6 = T4 + T5;
66 T1j = T4 - T5;
67 Tz = ii[WS(is, 4)];
68 TA = ii[WS(is, 12)];
69 TB = Tz + TA;
70 TM = Tz - TA;
71 }
72 T7 = T3 + T6;
73 T1R = T3 - T6;
74 T25 = Ty - TB;
75 TC = Ty + TB;
76 TN = TL - TM;
77 T1x = TL + TM;
78 T1H = T1k - T1j;
79 T1l = T1j + T1k;
80 }
81 {
82 E Tp, T1c, T1a, T20, Ts, T17, T1f, T21;
83 {
84 E Tn, To, T18, T19;
85 Tn = ri[WS(is, 15)];
86 To = ri[WS(is, 7)];
87 Tp = Tn + To;
88 T1c = Tn - To;
89 T18 = ii[WS(is, 15)];
90 T19 = ii[WS(is, 7)];
91 T1a = T18 - T19;
92 T20 = T18 + T19;
93 }
94 {
95 E Tq, Tr, T1d, T1e;
96 Tq = ri[WS(is, 3)];
97 Tr = ri[WS(is, 11)];
98 Ts = Tq + Tr;
99 T17 = Tq - Tr;
100 T1d = ii[WS(is, 3)];
101 T1e = ii[WS(is, 11)];
102 T1f = T1d - T1e;
103 T21 = T1d + T1e;
104 }
105 Tt = Tp + Ts;
106 T22 = T20 - T21;
107 T2h = T20 + T21;
108 T1b = T17 + T1a;
109 T1g = T1c - T1f;
110 T1E = T1a - T17;
111 T1Z = Tp - Ts;
112 T1D = T1c + T1f;
113 }
114 {
115 E Ta, TP, TF, TO, Td, TR, TI, TS;
116 {
117 E T8, T9, TD, TE;
118 T8 = ri[WS(is, 2)];
119 T9 = ri[WS(is, 10)];
120 Ta = T8 + T9;
121 TP = T8 - T9;
122 TD = ii[WS(is, 2)];
123 TE = ii[WS(is, 10)];
124 TF = TD + TE;
125 TO = TD - TE;
126 }
127 {
128 E Tb, Tc, TG, TH;
129 Tb = ri[WS(is, 14)];
130 Tc = ri[WS(is, 6)];
131 Td = Tb + Tc;
132 TR = Tb - Tc;
133 TG = ii[WS(is, 14)];
134 TH = ii[WS(is, 6)];
135 TI = TG + TH;
136 TS = TG - TH;
137 }
138 Te = Ta + Td;
139 T1S = TF - TI;
140 T26 = Td - Ta;
141 TJ = TF + TI;
142 TQ = TO - TP;
143 T1m = TR - TS;
144 T1n = TP + TO;
145 TT = TR + TS;
146 }
147 {
148 E Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
149 {
150 E Tg, Th, TX, TY;
151 Tg = ri[WS(is, 1)];
152 Th = ri[WS(is, 9)];
153 Ti = Tg + Th;
154 T11 = Tg - Th;
155 TX = ii[WS(is, 1)];
156 TY = ii[WS(is, 9)];
157 TZ = TX - TY;
158 T1V = TX + TY;
159 }
160 {
161 E Tj, Tk, T12, T13;
162 Tj = ri[WS(is, 5)];
163 Tk = ri[WS(is, 13)];
164 Tl = Tj + Tk;
165 TW = Tj - Tk;
166 T12 = ii[WS(is, 5)];
167 T13 = ii[WS(is, 13)];
168 T14 = T12 - T13;
169 T1W = T12 + T13;
170 }
171 Tm = Ti + Tl;
172 T1X = T1V - T1W;
173 T2g = T1V + T1W;
174 T10 = TW + TZ;
175 T15 = T11 - T14;
176 T1B = TZ - TW;
177 T1U = Ti - Tl;
178 T1A = T11 + T14;
179 }
180 {
181 E Tf, Tu, T2j, T2k;
182 Tf = T7 + Te;
183 Tu = Tm + Tt;
184 ro[WS(os, 8)] = Tf - Tu;
185 ro[0] = Tf + Tu;
186 T2j = TC + TJ;
187 T2k = T2g + T2h;
188 io[WS(os, 8)] = T2j - T2k;
189 io[0] = T2j + T2k;
190 }
191 {
192 E Tv, TK, T2f, T2i;
193 Tv = Tt - Tm;
194 TK = TC - TJ;
195 io[WS(os, 4)] = Tv + TK;
196 io[WS(os, 12)] = TK - Tv;
197 T2f = T7 - Te;
198 T2i = T2g - T2h;
199 ro[WS(os, 12)] = T2f - T2i;
200 ro[WS(os, 4)] = T2f + T2i;
201 }
202 {
203 E T1T, T27, T24, T28, T1Y, T23;
204 T1T = T1R + T1S;
205 T27 = T25 - T26;
206 T1Y = T1U + T1X;
207 T23 = T1Z - T22;
208 T24 = T1Y + T23;
209 T28 = T23 - T1Y;
210 ro[WS(os, 10)] = FNMS(KP707106781, T24, T1T);
211 io[WS(os, 6)] = FMA(KP707106781, T28, T27);
212 ro[WS(os, 2)] = FMA(KP707106781, T24, T1T);
213 io[WS(os, 14)] = FNMS(KP707106781, T28, T27);
214 }
215 {
216 E T29, T2d, T2c, T2e, T2a, T2b;
217 T29 = T1R - T1S;
218 T2d = T26 + T25;
219 T2a = T1X - T1U;
220 T2b = T1Z + T22;
221 T2c = T2a - T2b;
222 T2e = T2a + T2b;
223 ro[WS(os, 14)] = FNMS(KP707106781, T2c, T29);
224 io[WS(os, 2)] = FMA(KP707106781, T2e, T2d);
225 ro[WS(os, 6)] = FMA(KP707106781, T2c, T29);
226 io[WS(os, 10)] = FNMS(KP707106781, T2e, T2d);
227 }
228 {
229 E TV, T1v, T1p, T1r, T1i, T1q, T1u, T1w, TU, T1o;
230 TU = TQ - TT;
231 TV = FMA(KP707106781, TU, TN);
232 T1v = FNMS(KP707106781, TU, TN);
233 T1o = T1m - T1n;
234 T1p = FNMS(KP707106781, T1o, T1l);
235 T1r = FMA(KP707106781, T1o, T1l);
236 {
237 E T16, T1h, T1s, T1t;
238 T16 = FMA(KP414213562, T15, T10);
239 T1h = FNMS(KP414213562, T1g, T1b);
240 T1i = T16 - T1h;
241 T1q = T16 + T1h;
242 T1s = FMA(KP414213562, T1b, T1g);
243 T1t = FNMS(KP414213562, T10, T15);
244 T1u = T1s - T1t;
245 T1w = T1t + T1s;
246 }
247 ro[WS(os, 11)] = FNMS(KP923879532, T1i, TV);
248 io[WS(os, 11)] = FNMS(KP923879532, T1u, T1r);
249 ro[WS(os, 3)] = FMA(KP923879532, T1i, TV);
250 io[WS(os, 3)] = FMA(KP923879532, T1u, T1r);
251 io[WS(os, 7)] = FNMS(KP923879532, T1q, T1p);
252 ro[WS(os, 7)] = FNMS(KP923879532, T1w, T1v);
253 io[WS(os, 15)] = FMA(KP923879532, T1q, T1p);
254 ro[WS(os, 15)] = FMA(KP923879532, T1w, T1v);
255 }
256 {
257 E T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
258 T1y = T1n + T1m;
259 T1z = FMA(KP707106781, T1y, T1x);
260 T1L = FNMS(KP707106781, T1y, T1x);
261 T1I = TQ + TT;
262 T1J = FNMS(KP707106781, T1I, T1H);
263 T1P = FMA(KP707106781, T1I, T1H);
264 {
265 E T1C, T1F, T1M, T1N;
266 T1C = FMA(KP414213562, T1B, T1A);
267 T1F = FNMS(KP414213562, T1E, T1D);
268 T1G = T1C + T1F;
269 T1K = T1F - T1C;
270 T1M = FNMS(KP414213562, T1A, T1B);
271 T1N = FMA(KP414213562, T1D, T1E);
272 T1O = T1M - T1N;
273 T1Q = T1M + T1N;
274 }
275 ro[WS(os, 9)] = FNMS(KP923879532, T1G, T1z);
276 io[WS(os, 9)] = FNMS(KP923879532, T1Q, T1P);
277 ro[WS(os, 1)] = FMA(KP923879532, T1G, T1z);
278 io[WS(os, 1)] = FMA(KP923879532, T1Q, T1P);
279 io[WS(os, 13)] = FNMS(KP923879532, T1K, T1J);
280 ro[WS(os, 13)] = FNMS(KP923879532, T1O, T1L);
281 io[WS(os, 5)] = FMA(KP923879532, T1K, T1J);
282 ro[WS(os, 5)] = FMA(KP923879532, T1O, T1L);
283 }
284 }
285 }
286 }
287
288 static const kdft_desc desc = { 16, "n1_16", {104, 0, 40, 0}, &GENUS, 0, 0, 0, 0 };
289
290 void X(codelet_n1_16) (planner *p) {
291 X(kdft_register) (p, n1_16, &desc);
292 }
293
294 #else
295
296 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 16 -name n1_16 -include dft/scalar/n.h */
297
298 /*
299 * This function contains 144 FP additions, 24 FP multiplications,
300 * (or, 136 additions, 16 multiplications, 8 fused multiply/add),
301 * 50 stack variables, 3 constants, and 64 memory accesses
302 */
303 #include "dft/scalar/n.h"
304
305 static void n1_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
306 {
307 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
308 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
309 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
310 {
311 INT i;
312 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
313 E T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
314 E T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
315 E T1U, T1A;
316 {
317 E T3, TL, Ty, T1k, T6, T1j, TB, TM;
318 {
319 E T1, T2, Tw, Tx;
320 T1 = ri[0];
321 T2 = ri[WS(is, 8)];
322 T3 = T1 + T2;
323 TL = T1 - T2;
324 Tw = ii[0];
325 Tx = ii[WS(is, 8)];
326 Ty = Tw + Tx;
327 T1k = Tw - Tx;
328 }
329 {
330 E T4, T5, Tz, TA;
331 T4 = ri[WS(is, 4)];
332 T5 = ri[WS(is, 12)];
333 T6 = T4 + T5;
334 T1j = T4 - T5;
335 Tz = ii[WS(is, 4)];
336 TA = ii[WS(is, 12)];
337 TB = Tz + TA;
338 TM = Tz - TA;
339 }
340 T7 = T3 + T6;
341 T1R = T3 - T6;
342 T25 = Ty - TB;
343 TC = Ty + TB;
344 TN = TL - TM;
345 T1x = TL + TM;
346 T1H = T1k - T1j;
347 T1l = T1j + T1k;
348 }
349 {
350 E Tp, T17, T1f, T20, Ts, T1c, T1a, T21;
351 {
352 E Tn, To, T1d, T1e;
353 Tn = ri[WS(is, 15)];
354 To = ri[WS(is, 7)];
355 Tp = Tn + To;
356 T17 = Tn - To;
357 T1d = ii[WS(is, 15)];
358 T1e = ii[WS(is, 7)];
359 T1f = T1d - T1e;
360 T20 = T1d + T1e;
361 }
362 {
363 E Tq, Tr, T18, T19;
364 Tq = ri[WS(is, 3)];
365 Tr = ri[WS(is, 11)];
366 Ts = Tq + Tr;
367 T1c = Tq - Tr;
368 T18 = ii[WS(is, 3)];
369 T19 = ii[WS(is, 11)];
370 T1a = T18 - T19;
371 T21 = T18 + T19;
372 }
373 Tt = Tp + Ts;
374 T22 = T20 - T21;
375 T2h = T20 + T21;
376 T1b = T17 - T1a;
377 T1g = T1c + T1f;
378 T1E = T1f - T1c;
379 T1Z = Tp - Ts;
380 T1D = T17 + T1a;
381 }
382 {
383 E Ta, TP, TF, TO, Td, TR, TI, TS;
384 {
385 E T8, T9, TD, TE;
386 T8 = ri[WS(is, 2)];
387 T9 = ri[WS(is, 10)];
388 Ta = T8 + T9;
389 TP = T8 - T9;
390 TD = ii[WS(is, 2)];
391 TE = ii[WS(is, 10)];
392 TF = TD + TE;
393 TO = TD - TE;
394 }
395 {
396 E Tb, Tc, TG, TH;
397 Tb = ri[WS(is, 14)];
398 Tc = ri[WS(is, 6)];
399 Td = Tb + Tc;
400 TR = Tb - Tc;
401 TG = ii[WS(is, 14)];
402 TH = ii[WS(is, 6)];
403 TI = TG + TH;
404 TS = TG - TH;
405 }
406 Te = Ta + Td;
407 T1S = TF - TI;
408 T26 = Td - Ta;
409 TJ = TF + TI;
410 TQ = TO - TP;
411 T1m = TR - TS;
412 T1n = TP + TO;
413 TT = TR + TS;
414 }
415 {
416 E Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
417 {
418 E Tg, Th, TX, TY;
419 Tg = ri[WS(is, 1)];
420 Th = ri[WS(is, 9)];
421 Ti = Tg + Th;
422 T11 = Tg - Th;
423 TX = ii[WS(is, 1)];
424 TY = ii[WS(is, 9)];
425 TZ = TX - TY;
426 T1V = TX + TY;
427 }
428 {
429 E Tj, Tk, T12, T13;
430 Tj = ri[WS(is, 5)];
431 Tk = ri[WS(is, 13)];
432 Tl = Tj + Tk;
433 TW = Tj - Tk;
434 T12 = ii[WS(is, 5)];
435 T13 = ii[WS(is, 13)];
436 T14 = T12 - T13;
437 T1W = T12 + T13;
438 }
439 Tm = Ti + Tl;
440 T1X = T1V - T1W;
441 T2g = T1V + T1W;
442 T10 = TW + TZ;
443 T15 = T11 - T14;
444 T1B = T11 + T14;
445 T1U = Ti - Tl;
446 T1A = TZ - TW;
447 }
448 {
449 E Tf, Tu, T2j, T2k;
450 Tf = T7 + Te;
451 Tu = Tm + Tt;
452 ro[WS(os, 8)] = Tf - Tu;
453 ro[0] = Tf + Tu;
454 T2j = TC + TJ;
455 T2k = T2g + T2h;
456 io[WS(os, 8)] = T2j - T2k;
457 io[0] = T2j + T2k;
458 }
459 {
460 E Tv, TK, T2f, T2i;
461 Tv = Tt - Tm;
462 TK = TC - TJ;
463 io[WS(os, 4)] = Tv + TK;
464 io[WS(os, 12)] = TK - Tv;
465 T2f = T7 - Te;
466 T2i = T2g - T2h;
467 ro[WS(os, 12)] = T2f - T2i;
468 ro[WS(os, 4)] = T2f + T2i;
469 }
470 {
471 E T1T, T27, T24, T28, T1Y, T23;
472 T1T = T1R + T1S;
473 T27 = T25 - T26;
474 T1Y = T1U + T1X;
475 T23 = T1Z - T22;
476 T24 = KP707106781 * (T1Y + T23);
477 T28 = KP707106781 * (T23 - T1Y);
478 ro[WS(os, 10)] = T1T - T24;
479 io[WS(os, 6)] = T27 + T28;
480 ro[WS(os, 2)] = T1T + T24;
481 io[WS(os, 14)] = T27 - T28;
482 }
483 {
484 E T29, T2d, T2c, T2e, T2a, T2b;
485 T29 = T1R - T1S;
486 T2d = T26 + T25;
487 T2a = T1X - T1U;
488 T2b = T1Z + T22;
489 T2c = KP707106781 * (T2a - T2b);
490 T2e = KP707106781 * (T2a + T2b);
491 ro[WS(os, 14)] = T29 - T2c;
492 io[WS(os, 2)] = T2d + T2e;
493 ro[WS(os, 6)] = T29 + T2c;
494 io[WS(os, 10)] = T2d - T2e;
495 }
496 {
497 E TV, T1r, T1p, T1v, T1i, T1q, T1u, T1w, TU, T1o;
498 TU = KP707106781 * (TQ - TT);
499 TV = TN + TU;
500 T1r = TN - TU;
501 T1o = KP707106781 * (T1m - T1n);
502 T1p = T1l - T1o;
503 T1v = T1l + T1o;
504 {
505 E T16, T1h, T1s, T1t;
506 T16 = FMA(KP923879532, T10, KP382683432 * T15);
507 T1h = FNMS(KP923879532, T1g, KP382683432 * T1b);
508 T1i = T16 + T1h;
509 T1q = T1h - T16;
510 T1s = FNMS(KP923879532, T15, KP382683432 * T10);
511 T1t = FMA(KP382683432, T1g, KP923879532 * T1b);
512 T1u = T1s - T1t;
513 T1w = T1s + T1t;
514 }
515 ro[WS(os, 11)] = TV - T1i;
516 io[WS(os, 11)] = T1v - T1w;
517 ro[WS(os, 3)] = TV + T1i;
518 io[WS(os, 3)] = T1v + T1w;
519 io[WS(os, 15)] = T1p - T1q;
520 ro[WS(os, 15)] = T1r - T1u;
521 io[WS(os, 7)] = T1p + T1q;
522 ro[WS(os, 7)] = T1r + T1u;
523 }
524 {
525 E T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
526 T1y = KP707106781 * (T1n + T1m);
527 T1z = T1x + T1y;
528 T1L = T1x - T1y;
529 T1I = KP707106781 * (TQ + TT);
530 T1J = T1H - T1I;
531 T1P = T1H + T1I;
532 {
533 E T1C, T1F, T1M, T1N;
534 T1C = FMA(KP382683432, T1A, KP923879532 * T1B);
535 T1F = FNMS(KP382683432, T1E, KP923879532 * T1D);
536 T1G = T1C + T1F;
537 T1K = T1F - T1C;
538 T1M = FNMS(KP382683432, T1B, KP923879532 * T1A);
539 T1N = FMA(KP923879532, T1E, KP382683432 * T1D);
540 T1O = T1M - T1N;
541 T1Q = T1M + T1N;
542 }
543 ro[WS(os, 9)] = T1z - T1G;
544 io[WS(os, 9)] = T1P - T1Q;
545 ro[WS(os, 1)] = T1z + T1G;
546 io[WS(os, 1)] = T1P + T1Q;
547 io[WS(os, 13)] = T1J - T1K;
548 ro[WS(os, 13)] = T1L - T1O;
549 io[WS(os, 5)] = T1J + T1K;
550 ro[WS(os, 5)] = T1L + T1O;
551 }
552 }
553 }
554 }
555
556 static const kdft_desc desc = { 16, "n1_16", {136, 16, 8, 0}, &GENUS, 0, 0, 0, 0 };
557
558 void X(codelet_n1_16) (planner *p) {
559 X(kdft_register) (p, n1_16, &desc);
560 }
561
562 #endif