Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_12.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 07:40:45 EST 2012 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hc2cfdft_12 -include hc2cf.h */
29
30 /*
31 * This function contains 142 FP additions, 92 FP multiplications,
32 * (or, 96 additions, 46 multiplications, 46 fused multiply/add),
33 * 71 stack variables, 2 constants, and 48 memory accesses
34 */
35 #include "hc2cf.h"
36
37 static void hc2cfdft_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
44 E T2z, T2M;
45 {
46 E To, T1E, T2H, T1m, T1W, Tl, T1J, T2i, T2K, T1B, T2I, T2e, T19, T2E, T2C;
47 E T27, T1M, Tz, T2B, T1f, T1O, TJ, TT, T1Q;
48 {
49 E T2b, T1s, T1A, T2d;
50 {
51 E T1u, T1z, T1v, T2c, T1i, Te, T1l, Tj, Tf, T1H, T4, T1o, T1, T1r, T9;
52 E T1n, T5;
53 {
54 E T1x, T1y, T1t, Tm, Tn;
55 Tm = Ip[0];
56 Tn = Im[0];
57 T1x = Rp[0];
58 T1y = Rm[0];
59 T1t = W[0];
60 T1u = Tm + Tn;
61 To = Tm - Tn;
62 {
63 E Th, Ti, Tb, Tc, Td;
64 Tc = Ip[WS(rs, 4)];
65 T1z = T1x - T1y;
66 T1E = T1x + T1y;
67 Td = Im[WS(rs, 4)];
68 T1v = T1t * T1u;
69 Th = Rp[WS(rs, 4)];
70 T2c = T1t * T1z;
71 T1i = Tc + Td;
72 Te = Tc - Td;
73 Ti = Rm[WS(rs, 4)];
74 Tb = W[14];
75 {
76 E T7, T8, T2, T3;
77 T2 = Ip[WS(rs, 2)];
78 T1l = Th - Ti;
79 Tj = Th + Ti;
80 Tf = Tb * Te;
81 T3 = Im[WS(rs, 2)];
82 T7 = Rp[WS(rs, 2)];
83 T1H = Tb * Tj;
84 T8 = Rm[WS(rs, 2)];
85 T4 = T2 - T3;
86 T1o = T2 + T3;
87 T1 = W[6];
88 T1r = T7 - T8;
89 T9 = T7 + T8;
90 T1n = W[8];
91 T5 = T1 * T4;
92 }
93 }
94 }
95 {
96 E T1F, T2a, T1p, T1h, T1k;
97 T1F = T1 * T9;
98 T2a = T1n * T1r;
99 T1p = T1n * T1o;
100 T1h = W[16];
101 T1k = W[17];
102 {
103 E T1G, Ta, Tk, T1I, T1q, T1w;
104 {
105 E T6, Tg, T2G, T1j;
106 T6 = W[7];
107 Tg = W[15];
108 T2G = T1h * T1l;
109 T1j = T1h * T1i;
110 T1G = FMA(T6, T4, T1F);
111 Ta = FNMS(T6, T9, T5);
112 T2H = FMA(T1k, T1i, T2G);
113 T1m = FNMS(T1k, T1l, T1j);
114 Tk = FNMS(Tg, Tj, Tf);
115 T1I = FMA(Tg, Te, T1H);
116 }
117 T1q = W[9];
118 T1w = W[1];
119 T1W = Ta - Tk;
120 Tl = Ta + Tk;
121 T1J = T1G + T1I;
122 T2i = T1I - T1G;
123 T2b = FMA(T1q, T1o, T2a);
124 T1s = FNMS(T1q, T1r, T1p);
125 T1A = FNMS(T1w, T1z, T1v);
126 T2d = FMA(T1w, T1u, T2c);
127 }
128 }
129 }
130 {
131 E T11, Tt, T10, TX, Ty, TZ, T23, T1b, TN, TS, T1e, T1P, TO, T17, TD;
132 E T16, T13, T14, TI, TA;
133 {
134 E Tw, Tx, Tr, Ts, TK;
135 Tr = Ip[WS(rs, 3)];
136 Ts = Im[WS(rs, 3)];
137 T2K = T1s - T1A;
138 T1B = T1s + T1A;
139 T2I = T2b + T2d;
140 T2e = T2b - T2d;
141 Tw = Rp[WS(rs, 3)];
142 T11 = Tr + Ts;
143 Tt = Tr - Ts;
144 Tx = Rm[WS(rs, 3)];
145 T10 = W[12];
146 TX = W[13];
147 {
148 E TL, TY, TM, TQ, TR;
149 TL = Ip[WS(rs, 1)];
150 Ty = Tw + Tx;
151 TY = Tx - Tw;
152 TM = Im[WS(rs, 1)];
153 TQ = Rp[WS(rs, 1)];
154 TR = Rm[WS(rs, 1)];
155 TZ = TX * TY;
156 T23 = T10 * TY;
157 T1b = TL + TM;
158 TN = TL - TM;
159 TS = TQ + TR;
160 T1e = TQ - TR;
161 }
162 TK = W[2];
163 {
164 E TG, TH, TB, TC;
165 TB = Ip[WS(rs, 5)];
166 TC = Im[WS(rs, 5)];
167 TG = Rp[WS(rs, 5)];
168 T1P = TK * TS;
169 TO = TK * TN;
170 T17 = TB + TC;
171 TD = TB - TC;
172 TH = Rm[WS(rs, 5)];
173 T16 = W[20];
174 T13 = W[21];
175 T14 = TH - TG;
176 TI = TG + TH;
177 TA = W[18];
178 }
179 }
180 {
181 E T12, T1N, TE, T18, T24, T26, T25, T15;
182 T12 = FMA(T10, T11, TZ);
183 T15 = T13 * T14;
184 T25 = T16 * T14;
185 T1N = TA * TI;
186 TE = TA * TD;
187 T18 = FMA(T16, T17, T15);
188 T24 = FNMS(TX, T11, T23);
189 T26 = FNMS(T13, T17, T25);
190 {
191 E Tv, T1L, Tu, Tq;
192 Tq = W[10];
193 T19 = T12 + T18;
194 T2E = T18 - T12;
195 Tv = W[11];
196 T2C = T24 + T26;
197 T27 = T24 - T26;
198 T1L = Tq * Ty;
199 Tu = Tq * Tt;
200 {
201 E T1d, T2A, T1c, T1a, TF, TP;
202 T1a = W[4];
203 T1d = W[5];
204 T1M = FMA(Tv, Tt, T1L);
205 Tz = FNMS(Tv, Ty, Tu);
206 T2A = T1a * T1e;
207 T1c = T1a * T1b;
208 TF = W[19];
209 TP = W[3];
210 T2B = FMA(T1d, T1b, T2A);
211 T1f = FNMS(T1d, T1e, T1c);
212 T1O = FMA(TF, TD, T1N);
213 TJ = FNMS(TF, TI, TE);
214 TT = FNMS(TP, TS, TO);
215 T1Q = FMA(TP, TN, T1P);
216 }
217 }
218 }
219 }
220 }
221 {
222 E T2h, T2D, T1Z, T2l, T2J, T22, T2k, T29, T30, T1U, T1V, T1Y, T2Z, T1T;
223 {
224 E T2Y, TW, T2V, T1D, T1K, T1S;
225 {
226 E Tp, T2W, TU, T1R, T2X, T1g, TV, T1C;
227 T2h = FNMS(KP500000000, Tl, To);
228 Tp = Tl + To;
229 T2W = T2C - T2B;
230 T2D = FMA(KP500000000, T2C, T2B);
231 T1Z = TJ - TT;
232 TU = TJ + TT;
233 T1R = T1O + T1Q;
234 T2l = T1Q - T1O;
235 T2J = FNMS(KP500000000, T2I, T2H);
236 T2X = T2H + T2I;
237 T1g = T19 + T1f;
238 T22 = FNMS(KP500000000, T19, T1f);
239 T2k = FNMS(KP500000000, TU, Tz);
240 TV = Tz + TU;
241 T1C = T1m + T1B;
242 T29 = FNMS(KP500000000, T1B, T1m);
243 T2Y = T2W - T2X;
244 T30 = T2W + T2X;
245 TW = Tp - TV;
246 T2V = TV + Tp;
247 T1U = T1g + T1C;
248 T1D = T1g - T1C;
249 T1V = FNMS(KP500000000, T1J, T1E);
250 T1K = T1E + T1J;
251 T1S = T1M + T1R;
252 T1Y = FNMS(KP500000000, T1R, T1M);
253 }
254 Ip[WS(rs, 3)] = KP500000000 * (TW + T1D);
255 Im[WS(rs, 2)] = KP500000000 * (T1D - TW);
256 Im[WS(rs, 5)] = KP500000000 * (T2Y - T2V);
257 T2Z = T1K - T1S;
258 T1T = T1K + T1S;
259 Ip[0] = KP500000000 * (T2V + T2Y);
260 }
261 {
262 E T2v, T1X, T2Q, T2F, T2R, T2L, T2w, T20, T2t, T28, T2p, T2j;
263 Rm[WS(rs, 2)] = KP500000000 * (T2Z + T30);
264 Rp[WS(rs, 3)] = KP500000000 * (T2Z - T30);
265 Rp[0] = KP500000000 * (T1T + T1U);
266 Rm[WS(rs, 5)] = KP500000000 * (T1T - T1U);
267 T2v = FMA(KP866025403, T1W, T1V);
268 T1X = FNMS(KP866025403, T1W, T1V);
269 T2Q = FMA(KP866025403, T2E, T2D);
270 T2F = FNMS(KP866025403, T2E, T2D);
271 T2R = FMA(KP866025403, T2K, T2J);
272 T2L = FNMS(KP866025403, T2K, T2J);
273 T2w = FMA(KP866025403, T1Z, T1Y);
274 T20 = FNMS(KP866025403, T1Z, T1Y);
275 T2t = FMA(KP866025403, T27, T22);
276 T28 = FNMS(KP866025403, T27, T22);
277 T2p = FMA(KP866025403, T2i, T2h);
278 T2j = FNMS(KP866025403, T2i, T2h);
279 {
280 E T2T, T2q, T2s, T2U;
281 {
282 E T21, T2f, T2S, T2n, T2P, T2m, T2o, T2g;
283 T2T = T1X - T20;
284 T21 = T1X + T20;
285 T2q = FMA(KP866025403, T2l, T2k);
286 T2m = FNMS(KP866025403, T2l, T2k);
287 T2s = FMA(KP866025403, T2e, T29);
288 T2f = FNMS(KP866025403, T2e, T29);
289 T2S = T2Q + T2R;
290 T2U = T2R - T2Q;
291 T2n = T2j - T2m;
292 T2P = T2m + T2j;
293 T2o = T2f - T28;
294 T2g = T28 + T2f;
295 Im[WS(rs, 3)] = KP500000000 * (T2S - T2P);
296 Ip[WS(rs, 2)] = KP500000000 * (T2P + T2S);
297 Rm[WS(rs, 3)] = KP500000000 * (T21 + T2g);
298 Rp[WS(rs, 2)] = KP500000000 * (T21 - T2g);
299 Ip[WS(rs, 5)] = KP500000000 * (T2n + T2o);
300 Im[0] = KP500000000 * (T2o - T2n);
301 }
302 {
303 E T2y, T2x, T2N, T2O, T2r, T2u;
304 T2z = T2q + T2p;
305 T2r = T2p - T2q;
306 T2u = T2s - T2t;
307 T2y = T2t + T2s;
308 T2x = T2v + T2w;
309 T2N = T2v - T2w;
310 Rp[WS(rs, 5)] = KP500000000 * (T2T + T2U);
311 Rm[0] = KP500000000 * (T2T - T2U);
312 Im[WS(rs, 4)] = KP500000000 * (T2u - T2r);
313 Ip[WS(rs, 1)] = KP500000000 * (T2r + T2u);
314 T2O = T2L - T2F;
315 T2M = T2F + T2L;
316 Rp[WS(rs, 1)] = KP500000000 * (T2N + T2O);
317 Rm[WS(rs, 4)] = KP500000000 * (T2N - T2O);
318 Rp[WS(rs, 4)] = KP500000000 * (T2x + T2y);
319 Rm[WS(rs, 1)] = KP500000000 * (T2x - T2y);
320 }
321 }
322 }
323 }
324 }
325 Im[WS(rs, 1)] = -(KP500000000 * (T2z + T2M));
326 Ip[WS(rs, 4)] = KP500000000 * (T2z - T2M);
327 }
328 }
329 }
330
331 static const tw_instr twinstr[] = {
332 {TW_FULL, 1, 12},
333 {TW_NEXT, 1, 0}
334 };
335
336 static const hc2c_desc desc = { 12, "hc2cfdft_12", twinstr, &GENUS, {96, 46, 46, 0} };
337
338 void X(codelet_hc2cfdft_12) (planner *p) {
339 X(khc2c_register) (p, hc2cfdft_12, &desc, HC2C_VIA_DFT);
340 }
341 #else /* HAVE_FMA */
342
343 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 12 -dit -name hc2cfdft_12 -include hc2cf.h */
344
345 /*
346 * This function contains 142 FP additions, 76 FP multiplications,
347 * (or, 112 additions, 46 multiplications, 30 fused multiply/add),
348 * 52 stack variables, 3 constants, and 48 memory accesses
349 */
350 #include "hc2cf.h"
351
352 static void hc2cfdft_12(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
353 {
354 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
355 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
356 DK(KP433012701, +0.433012701892219323381861585376468091735701313);
357 {
358 INT m;
359 for (m = mb, W = W + ((mb - 1) * 22); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 22, MAKE_VOLATILE_STRIDE(48, rs)) {
360 E Tm, T1t, T1d, T2j, Tj, T1Y, T1w, T1G, T1q, T2q, T1U, T2k, Tw, T1y, T17;
361 E T2g, TP, T21, T1B, T1J, T12, T2u, T1P, T2h;
362 {
363 E Tk, Tl, T1k, T1m, T1n, T1o, T4, T1f, T8, T1h, Th, T1c, Td, T1a, T19;
364 E T1b;
365 {
366 E T2, T3, T6, T7;
367 Tk = Ip[0];
368 Tl = Im[0];
369 T1k = Tk + Tl;
370 T1m = Rp[0];
371 T1n = Rm[0];
372 T1o = T1m - T1n;
373 T2 = Ip[WS(rs, 2)];
374 T3 = Im[WS(rs, 2)];
375 T4 = T2 - T3;
376 T1f = T2 + T3;
377 T6 = Rp[WS(rs, 2)];
378 T7 = Rm[WS(rs, 2)];
379 T8 = T6 + T7;
380 T1h = T6 - T7;
381 {
382 E Tf, Tg, Tb, Tc;
383 Tf = Rp[WS(rs, 4)];
384 Tg = Rm[WS(rs, 4)];
385 Th = Tf + Tg;
386 T1c = Tf - Tg;
387 Tb = Ip[WS(rs, 4)];
388 Tc = Im[WS(rs, 4)];
389 Td = Tb - Tc;
390 T1a = Tb + Tc;
391 }
392 }
393 Tm = Tk - Tl;
394 T1t = T1m + T1n;
395 T19 = W[16];
396 T1b = W[17];
397 T1d = FNMS(T1b, T1c, T19 * T1a);
398 T2j = FMA(T19, T1c, T1b * T1a);
399 {
400 E T9, T1u, Ti, T1v;
401 {
402 E T1, T5, Ta, Te;
403 T1 = W[6];
404 T5 = W[7];
405 T9 = FNMS(T5, T8, T1 * T4);
406 T1u = FMA(T1, T8, T5 * T4);
407 Ta = W[14];
408 Te = W[15];
409 Ti = FNMS(Te, Th, Ta * Td);
410 T1v = FMA(Ta, Th, Te * Td);
411 }
412 Tj = T9 + Ti;
413 T1Y = KP433012701 * (T1v - T1u);
414 T1w = T1u + T1v;
415 T1G = KP433012701 * (T9 - Ti);
416 }
417 {
418 E T1i, T1S, T1p, T1T;
419 {
420 E T1e, T1g, T1j, T1l;
421 T1e = W[8];
422 T1g = W[9];
423 T1i = FNMS(T1g, T1h, T1e * T1f);
424 T1S = FMA(T1e, T1h, T1g * T1f);
425 T1j = W[0];
426 T1l = W[1];
427 T1p = FNMS(T1l, T1o, T1j * T1k);
428 T1T = FMA(T1j, T1o, T1l * T1k);
429 }
430 T1q = T1i + T1p;
431 T2q = KP433012701 * (T1i - T1p);
432 T1U = KP433012701 * (T1S - T1T);
433 T2k = T1S + T1T;
434 }
435 }
436 {
437 E Tr, TT, Tv, TV, TA, TY, TE, T10, TN, T14, TJ, T16;
438 {
439 E Tp, Tq, TC, TD;
440 Tp = Ip[WS(rs, 3)];
441 Tq = Im[WS(rs, 3)];
442 Tr = Tp - Tq;
443 TT = Tp + Tq;
444 {
445 E Tt, Tu, Ty, Tz;
446 Tt = Rp[WS(rs, 3)];
447 Tu = Rm[WS(rs, 3)];
448 Tv = Tt + Tu;
449 TV = Tt - Tu;
450 Ty = Ip[WS(rs, 5)];
451 Tz = Im[WS(rs, 5)];
452 TA = Ty - Tz;
453 TY = Ty + Tz;
454 }
455 TC = Rp[WS(rs, 5)];
456 TD = Rm[WS(rs, 5)];
457 TE = TC + TD;
458 T10 = TC - TD;
459 {
460 E TL, TM, TH, TI;
461 TL = Rp[WS(rs, 1)];
462 TM = Rm[WS(rs, 1)];
463 TN = TL + TM;
464 T14 = TM - TL;
465 TH = Ip[WS(rs, 1)];
466 TI = Im[WS(rs, 1)];
467 TJ = TH - TI;
468 T16 = TH + TI;
469 }
470 }
471 {
472 E To, Ts, T13, T15;
473 To = W[10];
474 Ts = W[11];
475 Tw = FNMS(Ts, Tv, To * Tr);
476 T1y = FMA(To, Tv, Ts * Tr);
477 T13 = W[5];
478 T15 = W[4];
479 T17 = FMA(T13, T14, T15 * T16);
480 T2g = FNMS(T13, T16, T15 * T14);
481 }
482 {
483 E TF, T1z, TO, T1A;
484 {
485 E Tx, TB, TG, TK;
486 Tx = W[18];
487 TB = W[19];
488 TF = FNMS(TB, TE, Tx * TA);
489 T1z = FMA(Tx, TE, TB * TA);
490 TG = W[2];
491 TK = W[3];
492 TO = FNMS(TK, TN, TG * TJ);
493 T1A = FMA(TG, TN, TK * TJ);
494 }
495 TP = TF + TO;
496 T21 = KP433012701 * (T1A - T1z);
497 T1B = T1z + T1A;
498 T1J = KP433012701 * (TF - TO);
499 }
500 {
501 E TW, T1O, T11, T1N;
502 {
503 E TS, TU, TX, TZ;
504 TS = W[12];
505 TU = W[13];
506 TW = FNMS(TU, TV, TS * TT);
507 T1O = FMA(TS, TV, TU * TT);
508 TX = W[20];
509 TZ = W[21];
510 T11 = FNMS(TZ, T10, TX * TY);
511 T1N = FMA(TX, T10, TZ * TY);
512 }
513 T12 = TW + T11;
514 T2u = KP433012701 * (T11 - TW);
515 T1P = KP433012701 * (T1N - T1O);
516 T2h = T1O + T1N;
517 }
518 }
519 {
520 E TR, T2f, T2m, T2o, T1s, T1E, T1D, T2n;
521 {
522 E Tn, TQ, T2i, T2l;
523 Tn = Tj + Tm;
524 TQ = Tw + TP;
525 TR = Tn - TQ;
526 T2f = TQ + Tn;
527 T2i = T2g - T2h;
528 T2l = T2j + T2k;
529 T2m = T2i - T2l;
530 T2o = T2i + T2l;
531 }
532 {
533 E T18, T1r, T1x, T1C;
534 T18 = T12 + T17;
535 T1r = T1d + T1q;
536 T1s = T18 - T1r;
537 T1E = T18 + T1r;
538 T1x = T1t + T1w;
539 T1C = T1y + T1B;
540 T1D = T1x + T1C;
541 T2n = T1x - T1C;
542 }
543 Ip[WS(rs, 3)] = KP500000000 * (TR + T1s);
544 Rp[WS(rs, 3)] = KP500000000 * (T2n - T2o);
545 Im[WS(rs, 2)] = KP500000000 * (T1s - TR);
546 Rm[WS(rs, 2)] = KP500000000 * (T2n + T2o);
547 Rm[WS(rs, 5)] = KP500000000 * (T1D - T1E);
548 Im[WS(rs, 5)] = KP500000000 * (T2m - T2f);
549 Rp[0] = KP500000000 * (T1D + T1E);
550 Ip[0] = KP500000000 * (T2f + T2m);
551 }
552 {
553 E T1H, T2b, T2s, T2B, T2v, T2A, T1K, T2c, T1Q, T29, T1Z, T25, T22, T26, T1V;
554 E T28;
555 {
556 E T1F, T2r, T2t, T1I;
557 T1F = FNMS(KP250000000, T1w, KP500000000 * T1t);
558 T1H = T1F - T1G;
559 T2b = T1F + T1G;
560 T2r = FNMS(KP500000000, T2j, KP250000000 * T2k);
561 T2s = T2q - T2r;
562 T2B = T2q + T2r;
563 T2t = FMA(KP250000000, T2h, KP500000000 * T2g);
564 T2v = T2t - T2u;
565 T2A = T2u + T2t;
566 T1I = FNMS(KP250000000, T1B, KP500000000 * T1y);
567 T1K = T1I - T1J;
568 T2c = T1I + T1J;
569 }
570 {
571 E T1M, T1X, T20, T1R;
572 T1M = FNMS(KP250000000, T12, KP500000000 * T17);
573 T1Q = T1M - T1P;
574 T29 = T1P + T1M;
575 T1X = FNMS(KP250000000, Tj, KP500000000 * Tm);
576 T1Z = T1X - T1Y;
577 T25 = T1Y + T1X;
578 T20 = FNMS(KP250000000, TP, KP500000000 * Tw);
579 T22 = T20 - T21;
580 T26 = T21 + T20;
581 T1R = FNMS(KP250000000, T1q, KP500000000 * T1d);
582 T1V = T1R - T1U;
583 T28 = T1R + T1U;
584 }
585 {
586 E T1L, T1W, T2p, T2w;
587 T1L = T1H + T1K;
588 T1W = T1Q + T1V;
589 Rp[WS(rs, 2)] = T1L - T1W;
590 Rm[WS(rs, 3)] = T1L + T1W;
591 T2p = T22 + T1Z;
592 T2w = T2s - T2v;
593 Ip[WS(rs, 2)] = T2p + T2w;
594 Im[WS(rs, 3)] = T2w - T2p;
595 }
596 {
597 E T23, T24, T2x, T2y;
598 T23 = T1Z - T22;
599 T24 = T1V - T1Q;
600 Ip[WS(rs, 5)] = T23 + T24;
601 Im[0] = T24 - T23;
602 T2x = T1H - T1K;
603 T2y = T2v + T2s;
604 Rm[0] = T2x - T2y;
605 Rp[WS(rs, 5)] = T2x + T2y;
606 }
607 {
608 E T27, T2a, T2z, T2C;
609 T27 = T25 - T26;
610 T2a = T28 - T29;
611 Ip[WS(rs, 1)] = T27 + T2a;
612 Im[WS(rs, 4)] = T2a - T27;
613 T2z = T2b - T2c;
614 T2C = T2A - T2B;
615 Rm[WS(rs, 4)] = T2z - T2C;
616 Rp[WS(rs, 1)] = T2z + T2C;
617 }
618 {
619 E T2d, T2e, T2D, T2E;
620 T2d = T2b + T2c;
621 T2e = T29 + T28;
622 Rm[WS(rs, 1)] = T2d - T2e;
623 Rp[WS(rs, 4)] = T2d + T2e;
624 T2D = T26 + T25;
625 T2E = T2A + T2B;
626 Ip[WS(rs, 4)] = T2D + T2E;
627 Im[WS(rs, 1)] = T2E - T2D;
628 }
629 }
630 }
631 }
632 }
633
634 static const tw_instr twinstr[] = {
635 {TW_FULL, 1, 12},
636 {TW_NEXT, 1, 0}
637 };
638
639 static const hc2c_desc desc = { 12, "hc2cfdft_12", twinstr, &GENUS, {112, 46, 30, 0} };
640
641 void X(codelet_hc2cfdft_12) (planner *p) {
642 X(khc2c_register) (p, hc2cfdft_12, &desc, HC2C_VIA_DFT);
643 }
644 #endif /* HAVE_FMA */