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 Dec 10 07:04:08 EST 2020 */
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 14 -name n1_14 -include dft/scalar/n.h */
29
30 /*
31 * This function contains 148 FP additions, 84 FP multiplications,
32 * (or, 64 additions, 0 multiplications, 84 fused multiply/add),
33 * 67 stack variables, 6 constants, and 56 memory accesses
34 */
35 #include "dft/scalar/n.h"
36
n1_14(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)37 static void n1_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP974927912, +0.974927912181823607018131682993931217232785801);
40 DK(KP801937735, +0.801937735804838252472204639014890102331838324);
41 DK(KP554958132, +0.554958132087371191422194871006410481067288862);
42 DK(KP900968867, +0.900968867902419126236102319507445051165919162);
43 DK(KP692021471, +0.692021471630095869627814897002069140197260599);
44 DK(KP356895867, +0.356895867892209443894399510021300583399127187);
45 {
46 INT i;
47 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(56, is), MAKE_VOLATILE_STRIDE(56, os)) {
48 E T3, Tp, T1b, T1x, T1i, T1L, T1M, T1j, T1k, T1K, Ta, To, Th, Tz, T14;
49 E TZ, Ts, Ty, Tv, T1Z, T2c, T27, TI, T23, T24, TP, TW, T22, T1c, T1e;
50 E T1d, T1f, T1s, T1n, T1A, T1G, T1D, T1H, T1U, T1P;
51 {
52 E T1, T2, T19, T1a;
53 T1 = ri[0];
54 T2 = ri[WS(is, 7)];
55 T3 = T1 - T2;
56 Tp = T1 + T2;
57 T19 = ii[0];
58 T1a = ii[WS(is, 7)];
59 T1b = T19 - T1a;
60 T1x = T19 + T1a;
61 }
62 {
63 E T6, Tq, T9, Tr, Tn, Tx, Tk, Tw, Tg, Tu, Td, Tt;
64 {
65 E T4, T5, Ti, Tj;
66 T4 = ri[WS(is, 2)];
67 T5 = ri[WS(is, 9)];
68 T6 = T4 - T5;
69 Tq = T4 + T5;
70 {
71 E T7, T8, Tl, Tm;
72 T7 = ri[WS(is, 12)];
73 T8 = ri[WS(is, 5)];
74 T9 = T7 - T8;
75 Tr = T7 + T8;
76 Tl = ri[WS(is, 8)];
77 Tm = ri[WS(is, 1)];
78 Tn = Tl - Tm;
79 Tx = Tl + Tm;
80 }
81 Ti = ri[WS(is, 6)];
82 Tj = ri[WS(is, 13)];
83 Tk = Ti - Tj;
84 Tw = Ti + Tj;
85 {
86 E Te, Tf, Tb, Tc;
87 Te = ri[WS(is, 10)];
88 Tf = ri[WS(is, 3)];
89 Tg = Te - Tf;
90 Tu = Te + Tf;
91 Tb = ri[WS(is, 4)];
92 Tc = ri[WS(is, 11)];
93 Td = Tb - Tc;
94 Tt = Tb + Tc;
95 }
96 }
97 T1i = Tn - Tk;
98 T1L = Tt - Tu;
99 T1M = Tr - Tq;
100 T1j = Tg - Td;
101 T1k = T9 - T6;
102 T1K = Tw - Tx;
103 Ta = T6 + T9;
104 To = Tk + Tn;
105 Th = Td + Tg;
106 Tz = FNMS(KP356895867, Th, Ta);
107 T14 = FNMS(KP356895867, To, Th);
108 TZ = FNMS(KP356895867, Ta, To);
109 Ts = Tq + Tr;
110 Ty = Tw + Tx;
111 Tv = Tt + Tu;
112 T1Z = FNMS(KP356895867, Ts, Ty);
113 T2c = FNMS(KP356895867, Ty, Tv);
114 T27 = FNMS(KP356895867, Tv, Ts);
115 }
116 {
117 E TE, T1B, TH, T1C, TV, T1F, TS, T1E, TO, T1z, TL, T1y;
118 {
119 E TC, TD, TQ, TR;
120 TC = ii[WS(is, 4)];
121 TD = ii[WS(is, 11)];
122 TE = TC - TD;
123 T1B = TC + TD;
124 {
125 E TF, TG, TT, TU;
126 TF = ii[WS(is, 10)];
127 TG = ii[WS(is, 3)];
128 TH = TF - TG;
129 T1C = TF + TG;
130 TT = ii[WS(is, 8)];
131 TU = ii[WS(is, 1)];
132 TV = TT - TU;
133 T1F = TT + TU;
134 }
135 TQ = ii[WS(is, 6)];
136 TR = ii[WS(is, 13)];
137 TS = TQ - TR;
138 T1E = TQ + TR;
139 {
140 E TM, TN, TJ, TK;
141 TM = ii[WS(is, 12)];
142 TN = ii[WS(is, 5)];
143 TO = TM - TN;
144 T1z = TM + TN;
145 TJ = ii[WS(is, 2)];
146 TK = ii[WS(is, 9)];
147 TL = TJ - TK;
148 T1y = TJ + TK;
149 }
150 }
151 TI = TE - TH;
152 T23 = T1F - T1E;
153 T24 = T1C - T1B;
154 TP = TL - TO;
155 TW = TS - TV;
156 T22 = T1y - T1z;
157 T1c = TL + TO;
158 T1e = TS + TV;
159 T1d = TE + TH;
160 T1f = FNMS(KP356895867, T1e, T1d);
161 T1s = FNMS(KP356895867, T1d, T1c);
162 T1n = FNMS(KP356895867, T1c, T1e);
163 T1A = T1y + T1z;
164 T1G = T1E + T1F;
165 T1D = T1B + T1C;
166 T1H = FNMS(KP356895867, T1G, T1D);
167 T1U = FNMS(KP356895867, T1D, T1A);
168 T1P = FNMS(KP356895867, T1A, T1G);
169 }
170 ro[WS(os, 7)] = T3 + Ta + Th + To;
171 io[WS(os, 7)] = T1b + T1c + T1d + T1e;
172 ro[0] = Tp + Ts + Tv + Ty;
173 io[0] = T1x + T1A + T1D + T1G;
174 {
175 E TB, TY, TA, TX;
176 TA = FNMS(KP692021471, Tz, To);
177 TB = FNMS(KP900968867, TA, T3);
178 TX = FMA(KP554958132, TW, TP);
179 TY = FMA(KP801937735, TX, TI);
180 ro[WS(os, 13)] = FNMS(KP974927912, TY, TB);
181 ro[WS(os, 1)] = FMA(KP974927912, TY, TB);
182 }
183 {
184 E T1u, T1w, T1t, T1v;
185 T1t = FNMS(KP692021471, T1s, T1e);
186 T1u = FNMS(KP900968867, T1t, T1b);
187 T1v = FMA(KP554958132, T1i, T1k);
188 T1w = FMA(KP801937735, T1v, T1j);
189 io[WS(os, 1)] = FMA(KP974927912, T1w, T1u);
190 io[WS(os, 13)] = FNMS(KP974927912, T1w, T1u);
191 }
192 {
193 E T11, T13, T10, T12;
194 T10 = FNMS(KP692021471, TZ, Th);
195 T11 = FNMS(KP900968867, T10, T3);
196 T12 = FMA(KP554958132, TI, TW);
197 T13 = FNMS(KP801937735, T12, TP);
198 ro[WS(os, 5)] = FNMS(KP974927912, T13, T11);
199 ro[WS(os, 9)] = FMA(KP974927912, T13, T11);
200 }
201 {
202 E T1p, T1r, T1o, T1q;
203 T1o = FNMS(KP692021471, T1n, T1d);
204 T1p = FNMS(KP900968867, T1o, T1b);
205 T1q = FMA(KP554958132, T1j, T1i);
206 T1r = FNMS(KP801937735, T1q, T1k);
207 io[WS(os, 5)] = FNMS(KP974927912, T1r, T1p);
208 io[WS(os, 9)] = FMA(KP974927912, T1r, T1p);
209 }
210 {
211 E T16, T18, T15, T17;
212 T15 = FNMS(KP692021471, T14, Ta);
213 T16 = FNMS(KP900968867, T15, T3);
214 T17 = FNMS(KP554958132, TP, TI);
215 T18 = FNMS(KP801937735, T17, TW);
216 ro[WS(os, 11)] = FNMS(KP974927912, T18, T16);
217 ro[WS(os, 3)] = FMA(KP974927912, T18, T16);
218 }
219 {
220 E T1h, T1m, T1g, T1l;
221 T1g = FNMS(KP692021471, T1f, T1c);
222 T1h = FNMS(KP900968867, T1g, T1b);
223 T1l = FNMS(KP554958132, T1k, T1j);
224 T1m = FNMS(KP801937735, T1l, T1i);
225 io[WS(os, 3)] = FMA(KP974927912, T1m, T1h);
226 io[WS(os, 11)] = FNMS(KP974927912, T1m, T1h);
227 }
228 {
229 E T1J, T1O, T1I, T1N;
230 T1I = FNMS(KP692021471, T1H, T1A);
231 T1J = FNMS(KP900968867, T1I, T1x);
232 T1N = FMA(KP554958132, T1M, T1L);
233 T1O = FNMS(KP801937735, T1N, T1K);
234 io[WS(os, 4)] = FMA(KP974927912, T1O, T1J);
235 io[WS(os, 10)] = FNMS(KP974927912, T1O, T1J);
236 }
237 {
238 E T2e, T2g, T2d, T2f;
239 T2d = FNMS(KP692021471, T2c, Ts);
240 T2e = FNMS(KP900968867, T2d, Tp);
241 T2f = FMA(KP554958132, T22, T24);
242 T2g = FNMS(KP801937735, T2f, T23);
243 ro[WS(os, 10)] = FNMS(KP974927912, T2g, T2e);
244 ro[WS(os, 4)] = FMA(KP974927912, T2g, T2e);
245 }
246 {
247 E T1R, T1T, T1Q, T1S;
248 T1Q = FNMS(KP692021471, T1P, T1D);
249 T1R = FNMS(KP900968867, T1Q, T1x);
250 T1S = FMA(KP554958132, T1L, T1K);
251 T1T = FMA(KP801937735, T1S, T1M);
252 io[WS(os, 2)] = FMA(KP974927912, T1T, T1R);
253 io[WS(os, 12)] = FNMS(KP974927912, T1T, T1R);
254 }
255 {
256 E T21, T26, T20, T25;
257 T20 = FNMS(KP692021471, T1Z, Tv);
258 T21 = FNMS(KP900968867, T20, Tp);
259 T25 = FMA(KP554958132, T24, T23);
260 T26 = FMA(KP801937735, T25, T22);
261 ro[WS(os, 12)] = FNMS(KP974927912, T26, T21);
262 ro[WS(os, 2)] = FMA(KP974927912, T26, T21);
263 }
264 {
265 E T1W, T1Y, T1V, T1X;
266 T1V = FNMS(KP692021471, T1U, T1G);
267 T1W = FNMS(KP900968867, T1V, T1x);
268 T1X = FNMS(KP554958132, T1K, T1M);
269 T1Y = FNMS(KP801937735, T1X, T1L);
270 io[WS(os, 6)] = FMA(KP974927912, T1Y, T1W);
271 io[WS(os, 8)] = FNMS(KP974927912, T1Y, T1W);
272 }
273 {
274 E T29, T2b, T28, T2a;
275 T28 = FNMS(KP692021471, T27, Ty);
276 T29 = FNMS(KP900968867, T28, Tp);
277 T2a = FNMS(KP554958132, T23, T22);
278 T2b = FNMS(KP801937735, T2a, T24);
279 ro[WS(os, 8)] = FNMS(KP974927912, T2b, T29);
280 ro[WS(os, 6)] = FMA(KP974927912, T2b, T29);
281 }
282 }
283 }
284 }
285
286 static const kdft_desc desc = { 14, "n1_14", { 64, 0, 84, 0 }, &GENUS, 0, 0, 0, 0 };
287
X(codelet_n1_14)288 void X(codelet_n1_14) (planner *p) { X(kdft_register) (p, n1_14, &desc);
289 }
290
291 #else
292
293 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 14 -name n1_14 -include dft/scalar/n.h */
294
295 /*
296 * This function contains 148 FP additions, 72 FP multiplications,
297 * (or, 100 additions, 24 multiplications, 48 fused multiply/add),
298 * 43 stack variables, 6 constants, and 56 memory accesses
299 */
300 #include "dft/scalar/n.h"
301
n1_14(const R * ri,const R * ii,R * ro,R * io,stride is,stride os,INT v,INT ivs,INT ovs)302 static void n1_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
303 {
304 DK(KP222520933, +0.222520933956314404288902564496794759466355569);
305 DK(KP900968867, +0.900968867902419126236102319507445051165919162);
306 DK(KP623489801, +0.623489801858733530525004884004239810632274731);
307 DK(KP433883739, +0.433883739117558120475768332848358754609990728);
308 DK(KP781831482, +0.781831482468029808708444526674057750232334519);
309 DK(KP974927912, +0.974927912181823607018131682993931217232785801);
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(56, is), MAKE_VOLATILE_STRIDE(56, os)) {
313 E T3, Tp, T16, T1f, Ta, T1q, Ts, T10, TG, T1z, T19, T1i, Th, T1s, Tv;
314 E T12, TU, T1B, T17, T1o, To, T1r, Ty, T11, TN, T1A, T18, T1l;
315 {
316 E T1, T2, T14, T15;
317 T1 = ri[0];
318 T2 = ri[WS(is, 7)];
319 T3 = T1 - T2;
320 Tp = T1 + T2;
321 T14 = ii[0];
322 T15 = ii[WS(is, 7)];
323 T16 = T14 - T15;
324 T1f = T14 + T15;
325 }
326 {
327 E T6, Tq, T9, Tr;
328 {
329 E T4, T5, T7, T8;
330 T4 = ri[WS(is, 2)];
331 T5 = ri[WS(is, 9)];
332 T6 = T4 - T5;
333 Tq = T4 + T5;
334 T7 = ri[WS(is, 12)];
335 T8 = ri[WS(is, 5)];
336 T9 = T7 - T8;
337 Tr = T7 + T8;
338 }
339 Ta = T6 + T9;
340 T1q = Tr - Tq;
341 Ts = Tq + Tr;
342 T10 = T9 - T6;
343 }
344 {
345 E TC, T1g, TF, T1h;
346 {
347 E TA, TB, TD, TE;
348 TA = ii[WS(is, 2)];
349 TB = ii[WS(is, 9)];
350 TC = TA - TB;
351 T1g = TA + TB;
352 TD = ii[WS(is, 12)];
353 TE = ii[WS(is, 5)];
354 TF = TD - TE;
355 T1h = TD + TE;
356 }
357 TG = TC - TF;
358 T1z = T1g - T1h;
359 T19 = TC + TF;
360 T1i = T1g + T1h;
361 }
362 {
363 E Td, Tt, Tg, Tu;
364 {
365 E Tb, Tc, Te, Tf;
366 Tb = ri[WS(is, 4)];
367 Tc = ri[WS(is, 11)];
368 Td = Tb - Tc;
369 Tt = Tb + Tc;
370 Te = ri[WS(is, 10)];
371 Tf = ri[WS(is, 3)];
372 Tg = Te - Tf;
373 Tu = Te + Tf;
374 }
375 Th = Td + Tg;
376 T1s = Tt - Tu;
377 Tv = Tt + Tu;
378 T12 = Tg - Td;
379 }
380 {
381 E TQ, T1m, TT, T1n;
382 {
383 E TO, TP, TR, TS;
384 TO = ii[WS(is, 4)];
385 TP = ii[WS(is, 11)];
386 TQ = TO - TP;
387 T1m = TO + TP;
388 TR = ii[WS(is, 10)];
389 TS = ii[WS(is, 3)];
390 TT = TR - TS;
391 T1n = TR + TS;
392 }
393 TU = TQ - TT;
394 T1B = T1n - T1m;
395 T17 = TQ + TT;
396 T1o = T1m + T1n;
397 }
398 {
399 E Tk, Tw, Tn, Tx;
400 {
401 E Ti, Tj, Tl, Tm;
402 Ti = ri[WS(is, 6)];
403 Tj = ri[WS(is, 13)];
404 Tk = Ti - Tj;
405 Tw = Ti + Tj;
406 Tl = ri[WS(is, 8)];
407 Tm = ri[WS(is, 1)];
408 Tn = Tl - Tm;
409 Tx = Tl + Tm;
410 }
411 To = Tk + Tn;
412 T1r = Tw - Tx;
413 Ty = Tw + Tx;
414 T11 = Tn - Tk;
415 }
416 {
417 E TJ, T1j, TM, T1k;
418 {
419 E TH, TI, TK, TL;
420 TH = ii[WS(is, 6)];
421 TI = ii[WS(is, 13)];
422 TJ = TH - TI;
423 T1j = TH + TI;
424 TK = ii[WS(is, 8)];
425 TL = ii[WS(is, 1)];
426 TM = TK - TL;
427 T1k = TK + TL;
428 }
429 TN = TJ - TM;
430 T1A = T1k - T1j;
431 T18 = TJ + TM;
432 T1l = T1j + T1k;
433 }
434 ro[WS(os, 7)] = T3 + Ta + Th + To;
435 io[WS(os, 7)] = T16 + T19 + T17 + T18;
436 ro[0] = Tp + Ts + Tv + Ty;
437 io[0] = T1f + T1i + T1o + T1l;
438 {
439 E TV, Tz, T1e, T1d;
440 TV = FNMS(KP781831482, TN, KP974927912 * TG) - (KP433883739 * TU);
441 Tz = FMA(KP623489801, To, T3) + FNMA(KP900968867, Th, KP222520933 * Ta);
442 ro[WS(os, 5)] = Tz - TV;
443 ro[WS(os, 9)] = Tz + TV;
444 T1e = FNMS(KP781831482, T11, KP974927912 * T10) - (KP433883739 * T12);
445 T1d = FMA(KP623489801, T18, T16) + FNMA(KP900968867, T17, KP222520933 * T19);
446 io[WS(os, 5)] = T1d - T1e;
447 io[WS(os, 9)] = T1e + T1d;
448 }
449 {
450 E TX, TW, T1b, T1c;
451 TX = FMA(KP781831482, TG, KP974927912 * TU) + (KP433883739 * TN);
452 TW = FMA(KP623489801, Ta, T3) + FNMA(KP900968867, To, KP222520933 * Th);
453 ro[WS(os, 13)] = TW - TX;
454 ro[WS(os, 1)] = TW + TX;
455 T1b = FMA(KP781831482, T10, KP974927912 * T12) + (KP433883739 * T11);
456 T1c = FMA(KP623489801, T19, T16) + FNMA(KP900968867, T18, KP222520933 * T17);
457 io[WS(os, 1)] = T1b + T1c;
458 io[WS(os, 13)] = T1c - T1b;
459 }
460 {
461 E TZ, TY, T13, T1a;
462 TZ = FMA(KP433883739, TG, KP974927912 * TN) - (KP781831482 * TU);
463 TY = FMA(KP623489801, Th, T3) + FNMA(KP222520933, To, KP900968867 * Ta);
464 ro[WS(os, 11)] = TY - TZ;
465 ro[WS(os, 3)] = TY + TZ;
466 T13 = FMA(KP433883739, T10, KP974927912 * T11) - (KP781831482 * T12);
467 T1a = FMA(KP623489801, T17, T16) + FNMA(KP222520933, T18, KP900968867 * T19);
468 io[WS(os, 3)] = T13 + T1a;
469 io[WS(os, 11)] = T1a - T13;
470 }
471 {
472 E T1t, T1p, T1C, T1y;
473 T1t = FNMS(KP433883739, T1r, KP781831482 * T1q) - (KP974927912 * T1s);
474 T1p = FMA(KP623489801, T1i, T1f) + FNMA(KP900968867, T1l, KP222520933 * T1o);
475 io[WS(os, 6)] = T1p - T1t;
476 io[WS(os, 8)] = T1t + T1p;
477 T1C = FNMS(KP433883739, T1A, KP781831482 * T1z) - (KP974927912 * T1B);
478 T1y = FMA(KP623489801, Ts, Tp) + FNMA(KP900968867, Ty, KP222520933 * Tv);
479 ro[WS(os, 6)] = T1y - T1C;
480 ro[WS(os, 8)] = T1y + T1C;
481 }
482 {
483 E T1v, T1u, T1E, T1D;
484 T1v = FMA(KP433883739, T1q, KP781831482 * T1s) - (KP974927912 * T1r);
485 T1u = FMA(KP623489801, T1o, T1f) + FNMA(KP222520933, T1l, KP900968867 * T1i);
486 io[WS(os, 4)] = T1u - T1v;
487 io[WS(os, 10)] = T1v + T1u;
488 T1E = FMA(KP433883739, T1z, KP781831482 * T1B) - (KP974927912 * T1A);
489 T1D = FMA(KP623489801, Tv, Tp) + FNMA(KP222520933, Ty, KP900968867 * Ts);
490 ro[WS(os, 4)] = T1D - T1E;
491 ro[WS(os, 10)] = T1D + T1E;
492 }
493 {
494 E T1w, T1x, T1G, T1F;
495 T1w = FMA(KP974927912, T1q, KP433883739 * T1s) + (KP781831482 * T1r);
496 T1x = FMA(KP623489801, T1l, T1f) + FNMA(KP900968867, T1o, KP222520933 * T1i);
497 io[WS(os, 2)] = T1w + T1x;
498 io[WS(os, 12)] = T1x - T1w;
499 T1G = FMA(KP974927912, T1z, KP433883739 * T1B) + (KP781831482 * T1A);
500 T1F = FMA(KP623489801, Ty, Tp) + FNMA(KP900968867, Tv, KP222520933 * Ts);
501 ro[WS(os, 12)] = T1F - T1G;
502 ro[WS(os, 2)] = T1F + T1G;
503 }
504 }
505 }
506 }
507
508 static const kdft_desc desc = { 14, "n1_14", { 100, 24, 48, 0 }, &GENUS, 0, 0, 0, 0 };
509
X(codelet_n1_14)510 void X(codelet_n1_14) (planner *p) { X(kdft_register) (p, n1_14, &desc);
511 }
512
513 #endif
514