Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t1_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 08:04:15 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_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -n 16 -name t1_16 -include dft/scalar/t.h */
29
30 /*
31 * This function contains 174 FP additions, 100 FP multiplications,
32 * (or, 104 additions, 30 multiplications, 70 fused multiply/add),
33 * 60 stack variables, 3 constants, and 64 memory accesses
34 */
35 #include "dft/scalar/t.h"
36
37 static void t1_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT m;
44 for (m = mb, W = W + (mb * 30); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
45 E T8, T3z, T1I, T3o, T1s, T35, T2o, T2r, T1F, T36, T2p, T2w, Tl, T3A, T1N;
46 E T3k, Tz, T2V, T1T, T1U, T11, T30, T29, T2c, T1e, T31, T2a, T2h, TM, T2W;
47 E T1W, T21;
48 {
49 E T1, T3n, T3, T6, T4, T3l, T2, T7, T3m, T5;
50 T1 = ri[0];
51 T3n = ii[0];
52 T3 = ri[WS(rs, 8)];
53 T6 = ii[WS(rs, 8)];
54 T2 = W[14];
55 T4 = T2 * T3;
56 T3l = T2 * T6;
57 T5 = W[15];
58 T7 = FMA(T5, T6, T4);
59 T3m = FNMS(T5, T3, T3l);
60 T8 = T1 + T7;
61 T3z = T3n - T3m;
62 T1I = T1 - T7;
63 T3o = T3m + T3n;
64 }
65 {
66 E T1h, T1k, T1i, T2k, T1n, T1q, T1o, T2m, T1g, T1m;
67 T1h = ri[WS(rs, 15)];
68 T1k = ii[WS(rs, 15)];
69 T1g = W[28];
70 T1i = T1g * T1h;
71 T2k = T1g * T1k;
72 T1n = ri[WS(rs, 7)];
73 T1q = ii[WS(rs, 7)];
74 T1m = W[12];
75 T1o = T1m * T1n;
76 T2m = T1m * T1q;
77 {
78 E T1l, T2l, T1r, T2n, T1j, T1p;
79 T1j = W[29];
80 T1l = FMA(T1j, T1k, T1i);
81 T2l = FNMS(T1j, T1h, T2k);
82 T1p = W[13];
83 T1r = FMA(T1p, T1q, T1o);
84 T2n = FNMS(T1p, T1n, T2m);
85 T1s = T1l + T1r;
86 T35 = T2l + T2n;
87 T2o = T2l - T2n;
88 T2r = T1l - T1r;
89 }
90 }
91 {
92 E T1u, T1x, T1v, T2s, T1A, T1D, T1B, T2u, T1t, T1z;
93 T1u = ri[WS(rs, 3)];
94 T1x = ii[WS(rs, 3)];
95 T1t = W[4];
96 T1v = T1t * T1u;
97 T2s = T1t * T1x;
98 T1A = ri[WS(rs, 11)];
99 T1D = ii[WS(rs, 11)];
100 T1z = W[20];
101 T1B = T1z * T1A;
102 T2u = T1z * T1D;
103 {
104 E T1y, T2t, T1E, T2v, T1w, T1C;
105 T1w = W[5];
106 T1y = FMA(T1w, T1x, T1v);
107 T2t = FNMS(T1w, T1u, T2s);
108 T1C = W[21];
109 T1E = FMA(T1C, T1D, T1B);
110 T2v = FNMS(T1C, T1A, T2u);
111 T1F = T1y + T1E;
112 T36 = T2t + T2v;
113 T2p = T1y - T1E;
114 T2w = T2t - T2v;
115 }
116 }
117 {
118 E Ta, Td, Tb, T1J, Tg, Tj, Th, T1L, T9, Tf;
119 Ta = ri[WS(rs, 4)];
120 Td = ii[WS(rs, 4)];
121 T9 = W[6];
122 Tb = T9 * Ta;
123 T1J = T9 * Td;
124 Tg = ri[WS(rs, 12)];
125 Tj = ii[WS(rs, 12)];
126 Tf = W[22];
127 Th = Tf * Tg;
128 T1L = Tf * Tj;
129 {
130 E Te, T1K, Tk, T1M, Tc, Ti;
131 Tc = W[7];
132 Te = FMA(Tc, Td, Tb);
133 T1K = FNMS(Tc, Ta, T1J);
134 Ti = W[23];
135 Tk = FMA(Ti, Tj, Th);
136 T1M = FNMS(Ti, Tg, T1L);
137 Tl = Te + Tk;
138 T3A = Te - Tk;
139 T1N = T1K - T1M;
140 T3k = T1K + T1M;
141 }
142 }
143 {
144 E To, Tr, Tp, T1P, Tu, Tx, Tv, T1R, Tn, Tt;
145 To = ri[WS(rs, 2)];
146 Tr = ii[WS(rs, 2)];
147 Tn = W[2];
148 Tp = Tn * To;
149 T1P = Tn * Tr;
150 Tu = ri[WS(rs, 10)];
151 Tx = ii[WS(rs, 10)];
152 Tt = W[18];
153 Tv = Tt * Tu;
154 T1R = Tt * Tx;
155 {
156 E Ts, T1Q, Ty, T1S, Tq, Tw;
157 Tq = W[3];
158 Ts = FMA(Tq, Tr, Tp);
159 T1Q = FNMS(Tq, To, T1P);
160 Tw = W[19];
161 Ty = FMA(Tw, Tx, Tv);
162 T1S = FNMS(Tw, Tu, T1R);
163 Tz = Ts + Ty;
164 T2V = T1Q + T1S;
165 T1T = T1Q - T1S;
166 T1U = Ts - Ty;
167 }
168 }
169 {
170 E TQ, TT, TR, T25, TW, TZ, TX, T27, TP, TV;
171 TQ = ri[WS(rs, 1)];
172 TT = ii[WS(rs, 1)];
173 TP = W[0];
174 TR = TP * TQ;
175 T25 = TP * TT;
176 TW = ri[WS(rs, 9)];
177 TZ = ii[WS(rs, 9)];
178 TV = W[16];
179 TX = TV * TW;
180 T27 = TV * TZ;
181 {
182 E TU, T26, T10, T28, TS, TY;
183 TS = W[1];
184 TU = FMA(TS, TT, TR);
185 T26 = FNMS(TS, TQ, T25);
186 TY = W[17];
187 T10 = FMA(TY, TZ, TX);
188 T28 = FNMS(TY, TW, T27);
189 T11 = TU + T10;
190 T30 = T26 + T28;
191 T29 = T26 - T28;
192 T2c = TU - T10;
193 }
194 }
195 {
196 E T13, T16, T14, T2d, T19, T1c, T1a, T2f, T12, T18;
197 T13 = ri[WS(rs, 5)];
198 T16 = ii[WS(rs, 5)];
199 T12 = W[8];
200 T14 = T12 * T13;
201 T2d = T12 * T16;
202 T19 = ri[WS(rs, 13)];
203 T1c = ii[WS(rs, 13)];
204 T18 = W[24];
205 T1a = T18 * T19;
206 T2f = T18 * T1c;
207 {
208 E T17, T2e, T1d, T2g, T15, T1b;
209 T15 = W[9];
210 T17 = FMA(T15, T16, T14);
211 T2e = FNMS(T15, T13, T2d);
212 T1b = W[25];
213 T1d = FMA(T1b, T1c, T1a);
214 T2g = FNMS(T1b, T19, T2f);
215 T1e = T17 + T1d;
216 T31 = T2e + T2g;
217 T2a = T17 - T1d;
218 T2h = T2e - T2g;
219 }
220 }
221 {
222 E TB, TE, TC, T1X, TH, TK, TI, T1Z, TA, TG;
223 TB = ri[WS(rs, 14)];
224 TE = ii[WS(rs, 14)];
225 TA = W[26];
226 TC = TA * TB;
227 T1X = TA * TE;
228 TH = ri[WS(rs, 6)];
229 TK = ii[WS(rs, 6)];
230 TG = W[10];
231 TI = TG * TH;
232 T1Z = TG * TK;
233 {
234 E TF, T1Y, TL, T20, TD, TJ;
235 TD = W[27];
236 TF = FMA(TD, TE, TC);
237 T1Y = FNMS(TD, TB, T1X);
238 TJ = W[11];
239 TL = FMA(TJ, TK, TI);
240 T20 = FNMS(TJ, TH, T1Z);
241 TM = TF + TL;
242 T2W = T1Y + T20;
243 T1W = TF - TL;
244 T21 = T1Y - T20;
245 }
246 }
247 {
248 E TO, T3e, T3q, T3s, T1H, T3r, T3h, T3i;
249 {
250 E Tm, TN, T3j, T3p;
251 Tm = T8 + Tl;
252 TN = Tz + TM;
253 TO = Tm + TN;
254 T3e = Tm - TN;
255 T3j = T2V + T2W;
256 T3p = T3k + T3o;
257 T3q = T3j + T3p;
258 T3s = T3p - T3j;
259 }
260 {
261 E T1f, T1G, T3f, T3g;
262 T1f = T11 + T1e;
263 T1G = T1s + T1F;
264 T1H = T1f + T1G;
265 T3r = T1G - T1f;
266 T3f = T30 + T31;
267 T3g = T35 + T36;
268 T3h = T3f - T3g;
269 T3i = T3f + T3g;
270 }
271 ri[WS(rs, 8)] = TO - T1H;
272 ii[WS(rs, 8)] = T3q - T3i;
273 ri[0] = TO + T1H;
274 ii[0] = T3i + T3q;
275 ri[WS(rs, 12)] = T3e - T3h;
276 ii[WS(rs, 12)] = T3s - T3r;
277 ri[WS(rs, 4)] = T3e + T3h;
278 ii[WS(rs, 4)] = T3r + T3s;
279 }
280 {
281 E T2Y, T3a, T3v, T3x, T33, T3b, T38, T3c;
282 {
283 E T2U, T2X, T3t, T3u;
284 T2U = T8 - Tl;
285 T2X = T2V - T2W;
286 T2Y = T2U + T2X;
287 T3a = T2U - T2X;
288 T3t = TM - Tz;
289 T3u = T3o - T3k;
290 T3v = T3t + T3u;
291 T3x = T3u - T3t;
292 }
293 {
294 E T2Z, T32, T34, T37;
295 T2Z = T11 - T1e;
296 T32 = T30 - T31;
297 T33 = T2Z + T32;
298 T3b = T32 - T2Z;
299 T34 = T1s - T1F;
300 T37 = T35 - T36;
301 T38 = T34 - T37;
302 T3c = T34 + T37;
303 }
304 {
305 E T39, T3w, T3d, T3y;
306 T39 = T33 + T38;
307 ri[WS(rs, 10)] = FNMS(KP707106781, T39, T2Y);
308 ri[WS(rs, 2)] = FMA(KP707106781, T39, T2Y);
309 T3w = T3b + T3c;
310 ii[WS(rs, 2)] = FMA(KP707106781, T3w, T3v);
311 ii[WS(rs, 10)] = FNMS(KP707106781, T3w, T3v);
312 T3d = T3b - T3c;
313 ri[WS(rs, 14)] = FNMS(KP707106781, T3d, T3a);
314 ri[WS(rs, 6)] = FMA(KP707106781, T3d, T3a);
315 T3y = T38 - T33;
316 ii[WS(rs, 6)] = FMA(KP707106781, T3y, T3x);
317 ii[WS(rs, 14)] = FNMS(KP707106781, T3y, T3x);
318 }
319 }
320 {
321 E T1O, T3B, T3H, T2E, T23, T3C, T2O, T2S, T2H, T3I, T2j, T2B, T2L, T2R, T2y;
322 E T2C;
323 {
324 E T1V, T22, T2b, T2i;
325 T1O = T1I - T1N;
326 T3B = T3z - T3A;
327 T3H = T3A + T3z;
328 T2E = T1I + T1N;
329 T1V = T1T - T1U;
330 T22 = T1W + T21;
331 T23 = T1V - T22;
332 T3C = T1V + T22;
333 {
334 E T2M, T2N, T2F, T2G;
335 T2M = T2r + T2w;
336 T2N = T2o - T2p;
337 T2O = FNMS(KP414213562, T2N, T2M);
338 T2S = FMA(KP414213562, T2M, T2N);
339 T2F = T1U + T1T;
340 T2G = T1W - T21;
341 T2H = T2F + T2G;
342 T3I = T2G - T2F;
343 }
344 T2b = T29 + T2a;
345 T2i = T2c - T2h;
346 T2j = FMA(KP414213562, T2i, T2b);
347 T2B = FNMS(KP414213562, T2b, T2i);
348 {
349 E T2J, T2K, T2q, T2x;
350 T2J = T2c + T2h;
351 T2K = T29 - T2a;
352 T2L = FMA(KP414213562, T2K, T2J);
353 T2R = FNMS(KP414213562, T2J, T2K);
354 T2q = T2o + T2p;
355 T2x = T2r - T2w;
356 T2y = FNMS(KP414213562, T2x, T2q);
357 T2C = FMA(KP414213562, T2q, T2x);
358 }
359 }
360 {
361 E T24, T2z, T3J, T3K;
362 T24 = FMA(KP707106781, T23, T1O);
363 T2z = T2j - T2y;
364 ri[WS(rs, 11)] = FNMS(KP923879532, T2z, T24);
365 ri[WS(rs, 3)] = FMA(KP923879532, T2z, T24);
366 T3J = FMA(KP707106781, T3I, T3H);
367 T3K = T2C - T2B;
368 ii[WS(rs, 3)] = FMA(KP923879532, T3K, T3J);
369 ii[WS(rs, 11)] = FNMS(KP923879532, T3K, T3J);
370 }
371 {
372 E T2A, T2D, T3L, T3M;
373 T2A = FNMS(KP707106781, T23, T1O);
374 T2D = T2B + T2C;
375 ri[WS(rs, 7)] = FNMS(KP923879532, T2D, T2A);
376 ri[WS(rs, 15)] = FMA(KP923879532, T2D, T2A);
377 T3L = FNMS(KP707106781, T3I, T3H);
378 T3M = T2j + T2y;
379 ii[WS(rs, 7)] = FNMS(KP923879532, T3M, T3L);
380 ii[WS(rs, 15)] = FMA(KP923879532, T3M, T3L);
381 }
382 {
383 E T2I, T2P, T3D, T3E;
384 T2I = FMA(KP707106781, T2H, T2E);
385 T2P = T2L + T2O;
386 ri[WS(rs, 9)] = FNMS(KP923879532, T2P, T2I);
387 ri[WS(rs, 1)] = FMA(KP923879532, T2P, T2I);
388 T3D = FMA(KP707106781, T3C, T3B);
389 T3E = T2R + T2S;
390 ii[WS(rs, 1)] = FMA(KP923879532, T3E, T3D);
391 ii[WS(rs, 9)] = FNMS(KP923879532, T3E, T3D);
392 }
393 {
394 E T2Q, T2T, T3F, T3G;
395 T2Q = FNMS(KP707106781, T2H, T2E);
396 T2T = T2R - T2S;
397 ri[WS(rs, 13)] = FNMS(KP923879532, T2T, T2Q);
398 ri[WS(rs, 5)] = FMA(KP923879532, T2T, T2Q);
399 T3F = FNMS(KP707106781, T3C, T3B);
400 T3G = T2O - T2L;
401 ii[WS(rs, 5)] = FMA(KP923879532, T3G, T3F);
402 ii[WS(rs, 13)] = FNMS(KP923879532, T3G, T3F);
403 }
404 }
405 }
406 }
407 }
408
409 static const tw_instr twinstr[] = {
410 {TW_FULL, 0, 16},
411 {TW_NEXT, 1, 0}
412 };
413
414 static const ct_desc desc = { 16, "t1_16", twinstr, &GENUS, {104, 30, 70, 0}, 0, 0, 0 };
415
416 void X(codelet_t1_16) (planner *p) {
417 X(kdft_dit_register) (p, t1_16, &desc);
418 }
419 #else
420
421 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 16 -name t1_16 -include dft/scalar/t.h */
422
423 /*
424 * This function contains 174 FP additions, 84 FP multiplications,
425 * (or, 136 additions, 46 multiplications, 38 fused multiply/add),
426 * 52 stack variables, 3 constants, and 64 memory accesses
427 */
428 #include "dft/scalar/t.h"
429
430 static void t1_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
431 {
432 DK(KP382683432, +0.382683432365089771728459984030398866761344562);
433 DK(KP923879532, +0.923879532511286756128183189396788286822416626);
434 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
435 {
436 INT m;
437 for (m = mb, W = W + (mb * 30); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 30, MAKE_VOLATILE_STRIDE(32, rs)) {
438 E T7, T37, T1t, T2U, Ti, T38, T1w, T2R, Tu, T2s, T1C, T2c, TF, T2t, T1H;
439 E T2d, T1f, T1q, T2B, T2C, T2D, T2E, T1Z, T2j, T24, T2k, TS, T13, T2w, T2x;
440 E T2y, T2z, T1O, T2g, T1T, T2h;
441 {
442 E T1, T2T, T6, T2S;
443 T1 = ri[0];
444 T2T = ii[0];
445 {
446 E T3, T5, T2, T4;
447 T3 = ri[WS(rs, 8)];
448 T5 = ii[WS(rs, 8)];
449 T2 = W[14];
450 T4 = W[15];
451 T6 = FMA(T2, T3, T4 * T5);
452 T2S = FNMS(T4, T3, T2 * T5);
453 }
454 T7 = T1 + T6;
455 T37 = T2T - T2S;
456 T1t = T1 - T6;
457 T2U = T2S + T2T;
458 }
459 {
460 E Tc, T1u, Th, T1v;
461 {
462 E T9, Tb, T8, Ta;
463 T9 = ri[WS(rs, 4)];
464 Tb = ii[WS(rs, 4)];
465 T8 = W[6];
466 Ta = W[7];
467 Tc = FMA(T8, T9, Ta * Tb);
468 T1u = FNMS(Ta, T9, T8 * Tb);
469 }
470 {
471 E Te, Tg, Td, Tf;
472 Te = ri[WS(rs, 12)];
473 Tg = ii[WS(rs, 12)];
474 Td = W[22];
475 Tf = W[23];
476 Th = FMA(Td, Te, Tf * Tg);
477 T1v = FNMS(Tf, Te, Td * Tg);
478 }
479 Ti = Tc + Th;
480 T38 = Tc - Th;
481 T1w = T1u - T1v;
482 T2R = T1u + T1v;
483 }
484 {
485 E To, T1y, Tt, T1z, T1A, T1B;
486 {
487 E Tl, Tn, Tk, Tm;
488 Tl = ri[WS(rs, 2)];
489 Tn = ii[WS(rs, 2)];
490 Tk = W[2];
491 Tm = W[3];
492 To = FMA(Tk, Tl, Tm * Tn);
493 T1y = FNMS(Tm, Tl, Tk * Tn);
494 }
495 {
496 E Tq, Ts, Tp, Tr;
497 Tq = ri[WS(rs, 10)];
498 Ts = ii[WS(rs, 10)];
499 Tp = W[18];
500 Tr = W[19];
501 Tt = FMA(Tp, Tq, Tr * Ts);
502 T1z = FNMS(Tr, Tq, Tp * Ts);
503 }
504 Tu = To + Tt;
505 T2s = T1y + T1z;
506 T1A = T1y - T1z;
507 T1B = To - Tt;
508 T1C = T1A - T1B;
509 T2c = T1B + T1A;
510 }
511 {
512 E Tz, T1E, TE, T1F, T1D, T1G;
513 {
514 E Tw, Ty, Tv, Tx;
515 Tw = ri[WS(rs, 14)];
516 Ty = ii[WS(rs, 14)];
517 Tv = W[26];
518 Tx = W[27];
519 Tz = FMA(Tv, Tw, Tx * Ty);
520 T1E = FNMS(Tx, Tw, Tv * Ty);
521 }
522 {
523 E TB, TD, TA, TC;
524 TB = ri[WS(rs, 6)];
525 TD = ii[WS(rs, 6)];
526 TA = W[10];
527 TC = W[11];
528 TE = FMA(TA, TB, TC * TD);
529 T1F = FNMS(TC, TB, TA * TD);
530 }
531 TF = Tz + TE;
532 T2t = T1E + T1F;
533 T1D = Tz - TE;
534 T1G = T1E - T1F;
535 T1H = T1D + T1G;
536 T2d = T1D - T1G;
537 }
538 {
539 E T19, T20, T1p, T1X, T1e, T21, T1k, T1W;
540 {
541 E T16, T18, T15, T17;
542 T16 = ri[WS(rs, 15)];
543 T18 = ii[WS(rs, 15)];
544 T15 = W[28];
545 T17 = W[29];
546 T19 = FMA(T15, T16, T17 * T18);
547 T20 = FNMS(T17, T16, T15 * T18);
548 }
549 {
550 E T1m, T1o, T1l, T1n;
551 T1m = ri[WS(rs, 11)];
552 T1o = ii[WS(rs, 11)];
553 T1l = W[20];
554 T1n = W[21];
555 T1p = FMA(T1l, T1m, T1n * T1o);
556 T1X = FNMS(T1n, T1m, T1l * T1o);
557 }
558 {
559 E T1b, T1d, T1a, T1c;
560 T1b = ri[WS(rs, 7)];
561 T1d = ii[WS(rs, 7)];
562 T1a = W[12];
563 T1c = W[13];
564 T1e = FMA(T1a, T1b, T1c * T1d);
565 T21 = FNMS(T1c, T1b, T1a * T1d);
566 }
567 {
568 E T1h, T1j, T1g, T1i;
569 T1h = ri[WS(rs, 3)];
570 T1j = ii[WS(rs, 3)];
571 T1g = W[4];
572 T1i = W[5];
573 T1k = FMA(T1g, T1h, T1i * T1j);
574 T1W = FNMS(T1i, T1h, T1g * T1j);
575 }
576 T1f = T19 + T1e;
577 T1q = T1k + T1p;
578 T2B = T1f - T1q;
579 T2C = T20 + T21;
580 T2D = T1W + T1X;
581 T2E = T2C - T2D;
582 {
583 E T1V, T1Y, T22, T23;
584 T1V = T19 - T1e;
585 T1Y = T1W - T1X;
586 T1Z = T1V - T1Y;
587 T2j = T1V + T1Y;
588 T22 = T20 - T21;
589 T23 = T1k - T1p;
590 T24 = T22 + T23;
591 T2k = T22 - T23;
592 }
593 }
594 {
595 E TM, T1K, T12, T1R, TR, T1L, TX, T1Q;
596 {
597 E TJ, TL, TI, TK;
598 TJ = ri[WS(rs, 1)];
599 TL = ii[WS(rs, 1)];
600 TI = W[0];
601 TK = W[1];
602 TM = FMA(TI, TJ, TK * TL);
603 T1K = FNMS(TK, TJ, TI * TL);
604 }
605 {
606 E TZ, T11, TY, T10;
607 TZ = ri[WS(rs, 13)];
608 T11 = ii[WS(rs, 13)];
609 TY = W[24];
610 T10 = W[25];
611 T12 = FMA(TY, TZ, T10 * T11);
612 T1R = FNMS(T10, TZ, TY * T11);
613 }
614 {
615 E TO, TQ, TN, TP;
616 TO = ri[WS(rs, 9)];
617 TQ = ii[WS(rs, 9)];
618 TN = W[16];
619 TP = W[17];
620 TR = FMA(TN, TO, TP * TQ);
621 T1L = FNMS(TP, TO, TN * TQ);
622 }
623 {
624 E TU, TW, TT, TV;
625 TU = ri[WS(rs, 5)];
626 TW = ii[WS(rs, 5)];
627 TT = W[8];
628 TV = W[9];
629 TX = FMA(TT, TU, TV * TW);
630 T1Q = FNMS(TV, TU, TT * TW);
631 }
632 TS = TM + TR;
633 T13 = TX + T12;
634 T2w = TS - T13;
635 T2x = T1K + T1L;
636 T2y = T1Q + T1R;
637 T2z = T2x - T2y;
638 {
639 E T1M, T1N, T1P, T1S;
640 T1M = T1K - T1L;
641 T1N = TX - T12;
642 T1O = T1M + T1N;
643 T2g = T1M - T1N;
644 T1P = TM - TR;
645 T1S = T1Q - T1R;
646 T1T = T1P - T1S;
647 T2h = T1P + T1S;
648 }
649 }
650 {
651 E T1J, T27, T3g, T3i, T26, T3h, T2a, T3d;
652 {
653 E T1x, T1I, T3e, T3f;
654 T1x = T1t - T1w;
655 T1I = KP707106781 * (T1C - T1H);
656 T1J = T1x + T1I;
657 T27 = T1x - T1I;
658 T3e = KP707106781 * (T2d - T2c);
659 T3f = T38 + T37;
660 T3g = T3e + T3f;
661 T3i = T3f - T3e;
662 }
663 {
664 E T1U, T25, T28, T29;
665 T1U = FMA(KP923879532, T1O, KP382683432 * T1T);
666 T25 = FNMS(KP923879532, T24, KP382683432 * T1Z);
667 T26 = T1U + T25;
668 T3h = T25 - T1U;
669 T28 = FNMS(KP923879532, T1T, KP382683432 * T1O);
670 T29 = FMA(KP382683432, T24, KP923879532 * T1Z);
671 T2a = T28 - T29;
672 T3d = T28 + T29;
673 }
674 ri[WS(rs, 11)] = T1J - T26;
675 ii[WS(rs, 11)] = T3g - T3d;
676 ri[WS(rs, 3)] = T1J + T26;
677 ii[WS(rs, 3)] = T3d + T3g;
678 ri[WS(rs, 15)] = T27 - T2a;
679 ii[WS(rs, 15)] = T3i - T3h;
680 ri[WS(rs, 7)] = T27 + T2a;
681 ii[WS(rs, 7)] = T3h + T3i;
682 }
683 {
684 E T2v, T2H, T32, T34, T2G, T33, T2K, T2Z;
685 {
686 E T2r, T2u, T30, T31;
687 T2r = T7 - Ti;
688 T2u = T2s - T2t;
689 T2v = T2r + T2u;
690 T2H = T2r - T2u;
691 T30 = TF - Tu;
692 T31 = T2U - T2R;
693 T32 = T30 + T31;
694 T34 = T31 - T30;
695 }
696 {
697 E T2A, T2F, T2I, T2J;
698 T2A = T2w + T2z;
699 T2F = T2B - T2E;
700 T2G = KP707106781 * (T2A + T2F);
701 T33 = KP707106781 * (T2F - T2A);
702 T2I = T2z - T2w;
703 T2J = T2B + T2E;
704 T2K = KP707106781 * (T2I - T2J);
705 T2Z = KP707106781 * (T2I + T2J);
706 }
707 ri[WS(rs, 10)] = T2v - T2G;
708 ii[WS(rs, 10)] = T32 - T2Z;
709 ri[WS(rs, 2)] = T2v + T2G;
710 ii[WS(rs, 2)] = T2Z + T32;
711 ri[WS(rs, 14)] = T2H - T2K;
712 ii[WS(rs, 14)] = T34 - T33;
713 ri[WS(rs, 6)] = T2H + T2K;
714 ii[WS(rs, 6)] = T33 + T34;
715 }
716 {
717 E T2f, T2n, T3a, T3c, T2m, T3b, T2q, T35;
718 {
719 E T2b, T2e, T36, T39;
720 T2b = T1t + T1w;
721 T2e = KP707106781 * (T2c + T2d);
722 T2f = T2b + T2e;
723 T2n = T2b - T2e;
724 T36 = KP707106781 * (T1C + T1H);
725 T39 = T37 - T38;
726 T3a = T36 + T39;
727 T3c = T39 - T36;
728 }
729 {
730 E T2i, T2l, T2o, T2p;
731 T2i = FMA(KP382683432, T2g, KP923879532 * T2h);
732 T2l = FNMS(KP382683432, T2k, KP923879532 * T2j);
733 T2m = T2i + T2l;
734 T3b = T2l - T2i;
735 T2o = FNMS(KP382683432, T2h, KP923879532 * T2g);
736 T2p = FMA(KP923879532, T2k, KP382683432 * T2j);
737 T2q = T2o - T2p;
738 T35 = T2o + T2p;
739 }
740 ri[WS(rs, 9)] = T2f - T2m;
741 ii[WS(rs, 9)] = T3a - T35;
742 ri[WS(rs, 1)] = T2f + T2m;
743 ii[WS(rs, 1)] = T35 + T3a;
744 ri[WS(rs, 13)] = T2n - T2q;
745 ii[WS(rs, 13)] = T3c - T3b;
746 ri[WS(rs, 5)] = T2n + T2q;
747 ii[WS(rs, 5)] = T3b + T3c;
748 }
749 {
750 E TH, T2L, T2W, T2Y, T1s, T2X, T2O, T2P;
751 {
752 E Tj, TG, T2Q, T2V;
753 Tj = T7 + Ti;
754 TG = Tu + TF;
755 TH = Tj + TG;
756 T2L = Tj - TG;
757 T2Q = T2s + T2t;
758 T2V = T2R + T2U;
759 T2W = T2Q + T2V;
760 T2Y = T2V - T2Q;
761 }
762 {
763 E T14, T1r, T2M, T2N;
764 T14 = TS + T13;
765 T1r = T1f + T1q;
766 T1s = T14 + T1r;
767 T2X = T1r - T14;
768 T2M = T2x + T2y;
769 T2N = T2C + T2D;
770 T2O = T2M - T2N;
771 T2P = T2M + T2N;
772 }
773 ri[WS(rs, 8)] = TH - T1s;
774 ii[WS(rs, 8)] = T2W - T2P;
775 ri[0] = TH + T1s;
776 ii[0] = T2P + T2W;
777 ri[WS(rs, 12)] = T2L - T2O;
778 ii[WS(rs, 12)] = T2Y - T2X;
779 ri[WS(rs, 4)] = T2L + T2O;
780 ii[WS(rs, 4)] = T2X + T2Y;
781 }
782 }
783 }
784 }
785
786 static const tw_instr twinstr[] = {
787 {TW_FULL, 0, 16},
788 {TW_NEXT, 1, 0}
789 };
790
791 static const ct_desc desc = { 16, "t1_16", twinstr, &GENUS, {136, 46, 38, 0}, 0, 0, 0 };
792
793 void X(codelet_t1_16) (planner *p) {
794 X(kdft_dit_register) (p, t1_16, &desc);
795 }
796 #endif