Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cb/r2cb_32.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 07:41:09 EST 2012 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_r2cb.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 32 -name r2cb_32 -include r2cb.h */
29
30 /*
31 * This function contains 156 FP additions, 84 FP multiplications,
32 * (or, 72 additions, 0 multiplications, 84 fused multiply/add),
33 * 82 stack variables, 9 constants, and 64 memory accesses
34 */
35 #include "r2cb.h"
36
37 static void r2cb_32(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
38 {
39 DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
40 DK(KP668178637, +0.668178637919298919997757686523080761552472251);
41 DK(KP1_961570560, +1.961570560806460898252364472268478073947867462);
42 DK(KP198912367, +0.198912367379658006911597622644676228597850501);
43 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
44 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
45 DK(KP414213562, +0.414213562373095048801688724209698078569671875);
46 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
47 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
48 {
49 INT i;
50 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(128, rs), MAKE_VOLATILE_STRIDE(128, csr), MAKE_VOLATILE_STRIDE(128, csi)) {
51 E T1F, T1C, T1H, T1z, T1G, T1I;
52 {
53 E T8, T1t, Tz, T1R, T5, T1S, T1u, TE, T1w, TP, T1U, Tg, T2m, T1X, T1x;
54 E TK, T1D, T1d, T20, To, T2p, T28, T1A, TW, T11, T1e, Tv, T25, T23, T2q;
55 E T16, T1f, TA, TD;
56 {
57 E T4, Ty, T1, T2, T6, T7;
58 T4 = Cr[WS(csr, 8)];
59 Ty = Ci[WS(csi, 8)];
60 T1 = Cr[0];
61 T2 = Cr[WS(csr, 16)];
62 T6 = Cr[WS(csr, 4)];
63 T7 = Cr[WS(csr, 12)];
64 {
65 E TB, Tx, T3, TC;
66 TB = Ci[WS(csi, 4)];
67 Tx = T1 - T2;
68 T3 = T1 + T2;
69 TA = T6 - T7;
70 T8 = T6 + T7;
71 TC = Ci[WS(csi, 12)];
72 T1t = FMA(KP2_000000000, Ty, Tx);
73 Tz = FNMS(KP2_000000000, Ty, Tx);
74 T1R = FNMS(KP2_000000000, T4, T3);
75 T5 = FMA(KP2_000000000, T4, T3);
76 TD = TB + TC;
77 T1S = TB - TC;
78 }
79 }
80 {
81 E Td, TG, Tc, T1V, TO, Te, TH, TI;
82 {
83 E Ta, Tb, TM, TN;
84 Ta = Cr[WS(csr, 2)];
85 T1u = TA + TD;
86 TE = TA - TD;
87 Tb = Cr[WS(csr, 14)];
88 TM = Ci[WS(csi, 2)];
89 TN = Ci[WS(csi, 14)];
90 Td = Cr[WS(csr, 10)];
91 TG = Ta - Tb;
92 Tc = Ta + Tb;
93 T1V = TM - TN;
94 TO = TM + TN;
95 Te = Cr[WS(csr, 6)];
96 TH = Ci[WS(csi, 10)];
97 TI = Ci[WS(csi, 6)];
98 }
99 {
100 E Tl, TS, Tk, T26, T1c, Tm, TT, TU;
101 {
102 E Ti, Tj, T1a, T1b;
103 Ti = Cr[WS(csr, 1)];
104 {
105 E TL, Tf, T1W, TJ;
106 TL = Td - Te;
107 Tf = Td + Te;
108 T1W = TH - TI;
109 TJ = TH + TI;
110 T1w = TO - TL;
111 TP = TL + TO;
112 T1U = Tc - Tf;
113 Tg = Tc + Tf;
114 T2m = T1W + T1V;
115 T1X = T1V - T1W;
116 T1x = TG + TJ;
117 TK = TG - TJ;
118 Tj = Cr[WS(csr, 15)];
119 }
120 T1a = Ci[WS(csi, 1)];
121 T1b = Ci[WS(csi, 15)];
122 Tl = Cr[WS(csr, 9)];
123 TS = Ti - Tj;
124 Tk = Ti + Tj;
125 T26 = T1a - T1b;
126 T1c = T1a + T1b;
127 Tm = Cr[WS(csr, 7)];
128 TT = Ci[WS(csi, 9)];
129 TU = Ci[WS(csi, 7)];
130 }
131 {
132 E Ts, TX, Tr, T22, T10, Tt, T13, T14;
133 {
134 E Tp, Tq, TY, TZ;
135 Tp = Cr[WS(csr, 5)];
136 {
137 E T19, Tn, T27, TV;
138 T19 = Tl - Tm;
139 Tn = Tl + Tm;
140 T27 = TT - TU;
141 TV = TT + TU;
142 T1D = T1c - T19;
143 T1d = T19 + T1c;
144 T20 = Tk - Tn;
145 To = Tk + Tn;
146 T2p = T27 + T26;
147 T28 = T26 - T27;
148 T1A = TS + TV;
149 TW = TS - TV;
150 Tq = Cr[WS(csr, 11)];
151 }
152 TY = Ci[WS(csi, 5)];
153 TZ = Ci[WS(csi, 11)];
154 Ts = Cr[WS(csr, 3)];
155 TX = Tp - Tq;
156 Tr = Tp + Tq;
157 T22 = TY - TZ;
158 T10 = TY + TZ;
159 Tt = Cr[WS(csr, 13)];
160 T13 = Ci[WS(csi, 3)];
161 T14 = Ci[WS(csi, 13)];
162 }
163 {
164 E T12, Tu, T21, T15;
165 T11 = TX - T10;
166 T1e = TX + T10;
167 T12 = Ts - Tt;
168 Tu = Ts + Tt;
169 T21 = T14 - T13;
170 T15 = T13 + T14;
171 Tv = Tr + Tu;
172 T25 = Tr - Tu;
173 T23 = T21 - T22;
174 T2q = T22 + T21;
175 T16 = T12 - T15;
176 T1f = T12 + T15;
177 }
178 }
179 }
180 }
181 {
182 E T1B, T1E, T1l, T1m, T1p, T1o, T1T, T1Y, T29, T2g, T2j, T2f, T2h, T24;
183 {
184 E T1g, T17, T2n, T2t, T2u, T2s;
185 {
186 E T2o, Tw, T2w, T2r, T2l, T9, Th, T2v;
187 T2o = To - Tv;
188 Tw = To + Tv;
189 T2w = T2q + T2p;
190 T2r = T2p - T2q;
191 T1g = T1e - T1f;
192 T1B = T1e + T1f;
193 T17 = T11 + T16;
194 T1E = T16 - T11;
195 T2l = FNMS(KP2_000000000, T8, T5);
196 T9 = FMA(KP2_000000000, T8, T5);
197 Th = FMA(KP2_000000000, Tg, T9);
198 T2v = FNMS(KP2_000000000, Tg, T9);
199 T2n = FNMS(KP2_000000000, T2m, T2l);
200 T2t = FMA(KP2_000000000, T2m, T2l);
201 R0[WS(rs, 4)] = FNMS(KP2_000000000, T2w, T2v);
202 R0[WS(rs, 12)] = FMA(KP2_000000000, T2w, T2v);
203 R0[0] = FMA(KP2_000000000, Tw, Th);
204 R0[WS(rs, 8)] = FNMS(KP2_000000000, Tw, Th);
205 T2u = T2o + T2r;
206 T2s = T2o - T2r;
207 }
208 {
209 E T1j, TR, T18, T1h, TF, TQ;
210 T1l = FNMS(KP1_414213562, TE, Tz);
211 TF = FMA(KP1_414213562, TE, Tz);
212 TQ = FNMS(KP414213562, TP, TK);
213 T1m = FMA(KP414213562, TK, TP);
214 R0[WS(rs, 2)] = FMA(KP1_414213562, T2s, T2n);
215 R0[WS(rs, 10)] = FNMS(KP1_414213562, T2s, T2n);
216 R0[WS(rs, 6)] = FNMS(KP1_414213562, T2u, T2t);
217 R0[WS(rs, 14)] = FMA(KP1_414213562, T2u, T2t);
218 T1j = FNMS(KP1_847759065, TQ, TF);
219 TR = FMA(KP1_847759065, TQ, TF);
220 T1p = FNMS(KP707106781, T17, TW);
221 T18 = FMA(KP707106781, T17, TW);
222 T1h = FMA(KP707106781, T1g, T1d);
223 T1o = FNMS(KP707106781, T1g, T1d);
224 {
225 E T2d, T2e, T1k, T1i;
226 T1T = FNMS(KP2_000000000, T1S, T1R);
227 T2d = FMA(KP2_000000000, T1S, T1R);
228 T2e = T1U + T1X;
229 T1Y = T1U - T1X;
230 T29 = T25 + T28;
231 T2g = T28 - T25;
232 T1k = FMA(KP198912367, T18, T1h);
233 T1i = FNMS(KP198912367, T1h, T18);
234 T2j = FMA(KP1_414213562, T2e, T2d);
235 T2f = FNMS(KP1_414213562, T2e, T2d);
236 R1[WS(rs, 4)] = FNMS(KP1_961570560, T1k, T1j);
237 R1[WS(rs, 12)] = FMA(KP1_961570560, T1k, T1j);
238 R1[0] = FMA(KP1_961570560, T1i, TR);
239 R1[WS(rs, 8)] = FNMS(KP1_961570560, T1i, TR);
240 T2h = T20 - T23;
241 T24 = T20 + T23;
242 }
243 }
244 }
245 {
246 E T1v, T1y, T1M, T1P, T1L, T1N;
247 {
248 E T1r, T1n, T2k, T2i;
249 T2k = FMA(KP414213562, T2g, T2h);
250 T2i = FNMS(KP414213562, T2h, T2g);
251 T1r = FMA(KP1_847759065, T1m, T1l);
252 T1n = FNMS(KP1_847759065, T1m, T1l);
253 R0[WS(rs, 7)] = FNMS(KP1_847759065, T2k, T2j);
254 R0[WS(rs, 15)] = FMA(KP1_847759065, T2k, T2j);
255 R0[WS(rs, 11)] = FMA(KP1_847759065, T2i, T2f);
256 R0[WS(rs, 3)] = FNMS(KP1_847759065, T2i, T2f);
257 {
258 E T1J, T1K, T1s, T1q;
259 T1v = FNMS(KP1_414213562, T1u, T1t);
260 T1J = FMA(KP1_414213562, T1u, T1t);
261 T1K = FMA(KP414213562, T1w, T1x);
262 T1y = FNMS(KP414213562, T1x, T1w);
263 T1F = FNMS(KP707106781, T1E, T1D);
264 T1M = FMA(KP707106781, T1E, T1D);
265 T1s = FMA(KP668178637, T1o, T1p);
266 T1q = FNMS(KP668178637, T1p, T1o);
267 T1P = FMA(KP1_847759065, T1K, T1J);
268 T1L = FNMS(KP1_847759065, T1K, T1J);
269 R1[WS(rs, 6)] = FNMS(KP1_662939224, T1s, T1r);
270 R1[WS(rs, 14)] = FMA(KP1_662939224, T1s, T1r);
271 R1[WS(rs, 10)] = FMA(KP1_662939224, T1q, T1n);
272 R1[WS(rs, 2)] = FNMS(KP1_662939224, T1q, T1n);
273 T1N = FMA(KP707106781, T1B, T1A);
274 T1C = FNMS(KP707106781, T1B, T1A);
275 }
276 }
277 {
278 E T2b, T1Z, T1Q, T1O, T2c, T2a;
279 T1Q = FMA(KP198912367, T1M, T1N);
280 T1O = FNMS(KP198912367, T1N, T1M);
281 T2b = FNMS(KP1_414213562, T1Y, T1T);
282 T1Z = FMA(KP1_414213562, T1Y, T1T);
283 R1[WS(rs, 7)] = FNMS(KP1_961570560, T1Q, T1P);
284 R1[WS(rs, 15)] = FMA(KP1_961570560, T1Q, T1P);
285 R1[WS(rs, 11)] = FMA(KP1_961570560, T1O, T1L);
286 R1[WS(rs, 3)] = FNMS(KP1_961570560, T1O, T1L);
287 T2c = FMA(KP414213562, T24, T29);
288 T2a = FNMS(KP414213562, T29, T24);
289 T1H = FMA(KP1_847759065, T1y, T1v);
290 T1z = FNMS(KP1_847759065, T1y, T1v);
291 R0[WS(rs, 5)] = FNMS(KP1_847759065, T2c, T2b);
292 R0[WS(rs, 13)] = FMA(KP1_847759065, T2c, T2b);
293 R0[WS(rs, 1)] = FMA(KP1_847759065, T2a, T1Z);
294 R0[WS(rs, 9)] = FNMS(KP1_847759065, T2a, T1Z);
295 }
296 }
297 }
298 }
299 T1G = FNMS(KP668178637, T1F, T1C);
300 T1I = FMA(KP668178637, T1C, T1F);
301 R1[WS(rs, 5)] = FNMS(KP1_662939224, T1I, T1H);
302 R1[WS(rs, 13)] = FMA(KP1_662939224, T1I, T1H);
303 R1[WS(rs, 1)] = FMA(KP1_662939224, T1G, T1z);
304 R1[WS(rs, 9)] = FNMS(KP1_662939224, T1G, T1z);
305 }
306 }
307 }
308
309 static const kr2c_desc desc = { 32, "r2cb_32", {72, 0, 84, 0}, &GENUS };
310
311 void X(codelet_r2cb_32) (planner *p) {
312 X(kr2c_register) (p, r2cb_32, &desc);
313 }
314
315 #else /* HAVE_FMA */
316
317 /* Generated by: ../../../genfft/gen_r2cb.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 32 -name r2cb_32 -include r2cb.h */
318
319 /*
320 * This function contains 156 FP additions, 50 FP multiplications,
321 * (or, 140 additions, 34 multiplications, 16 fused multiply/add),
322 * 54 stack variables, 9 constants, and 64 memory accesses
323 */
324 #include "r2cb.h"
325
326 static void r2cb_32(R *R0, R *R1, R *Cr, R *Ci, stride rs, stride csr, stride csi, INT v, INT ivs, INT ovs)
327 {
328 DK(KP1_662939224, +1.662939224605090474157576755235811513477121624);
329 DK(KP1_111140466, +1.111140466039204449485661627897065748749874382);
330 DK(KP1_961570560, +1.961570560806460898252364472268478073947867462);
331 DK(KP390180644, +0.390180644032256535696569736954044481855383236);
332 DK(KP765366864, +0.765366864730179543456919968060797733522689125);
333 DK(KP1_847759065, +1.847759065022573512256366378793576573644833252);
334 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
335 DK(KP1_414213562, +1.414213562373095048801688724209698078569671875);
336 DK(KP2_000000000, +2.000000000000000000000000000000000000000000000);
337 {
338 INT i;
339 for (i = v; i > 0; i = i - 1, R0 = R0 + ovs, R1 = R1 + ovs, Cr = Cr + ivs, Ci = Ci + ivs, MAKE_VOLATILE_STRIDE(128, rs), MAKE_VOLATILE_STRIDE(128, csr), MAKE_VOLATILE_STRIDE(128, csi)) {
340 E T9, T2c, TB, T1y, T6, T2b, Ty, T1v, Th, T2e, T2f, TD, TK, T1C, T1F;
341 E T1h, Tp, T2i, T2m, TN, T13, T1K, T1Y, T1k, Tw, TU, T1l, TW, T1V, T2j;
342 E T1R, T2l;
343 {
344 E T7, T8, T1w, Tz, TA, T1x;
345 T7 = Cr[WS(csr, 4)];
346 T8 = Cr[WS(csr, 12)];
347 T1w = T7 - T8;
348 Tz = Ci[WS(csi, 4)];
349 TA = Ci[WS(csi, 12)];
350 T1x = Tz + TA;
351 T9 = KP2_000000000 * (T7 + T8);
352 T2c = KP1_414213562 * (T1w + T1x);
353 TB = KP2_000000000 * (Tz - TA);
354 T1y = KP1_414213562 * (T1w - T1x);
355 }
356 {
357 E T5, T1u, T3, T1s;
358 {
359 E T4, T1t, T1, T2;
360 T4 = Cr[WS(csr, 8)];
361 T5 = KP2_000000000 * T4;
362 T1t = Ci[WS(csi, 8)];
363 T1u = KP2_000000000 * T1t;
364 T1 = Cr[0];
365 T2 = Cr[WS(csr, 16)];
366 T3 = T1 + T2;
367 T1s = T1 - T2;
368 }
369 T6 = T3 + T5;
370 T2b = T1s + T1u;
371 Ty = T3 - T5;
372 T1v = T1s - T1u;
373 }
374 {
375 E Td, T1A, TG, T1E, Tg, T1D, TJ, T1B;
376 {
377 E Tb, Tc, TE, TF;
378 Tb = Cr[WS(csr, 2)];
379 Tc = Cr[WS(csr, 14)];
380 Td = Tb + Tc;
381 T1A = Tb - Tc;
382 TE = Ci[WS(csi, 2)];
383 TF = Ci[WS(csi, 14)];
384 TG = TE - TF;
385 T1E = TE + TF;
386 }
387 {
388 E Te, Tf, TH, TI;
389 Te = Cr[WS(csr, 10)];
390 Tf = Cr[WS(csr, 6)];
391 Tg = Te + Tf;
392 T1D = Te - Tf;
393 TH = Ci[WS(csi, 10)];
394 TI = Ci[WS(csi, 6)];
395 TJ = TH - TI;
396 T1B = TH + TI;
397 }
398 Th = KP2_000000000 * (Td + Tg);
399 T2e = T1A + T1B;
400 T2f = T1E - T1D;
401 TD = Td - Tg;
402 TK = TG - TJ;
403 T1C = T1A - T1B;
404 T1F = T1D + T1E;
405 T1h = KP2_000000000 * (TJ + TG);
406 }
407 {
408 E Tl, T1I, TZ, T1X, To, T1W, T12, T1J;
409 {
410 E Tj, Tk, TX, TY;
411 Tj = Cr[WS(csr, 1)];
412 Tk = Cr[WS(csr, 15)];
413 Tl = Tj + Tk;
414 T1I = Tj - Tk;
415 TX = Ci[WS(csi, 1)];
416 TY = Ci[WS(csi, 15)];
417 TZ = TX - TY;
418 T1X = TX + TY;
419 }
420 {
421 E Tm, Tn, T10, T11;
422 Tm = Cr[WS(csr, 9)];
423 Tn = Cr[WS(csr, 7)];
424 To = Tm + Tn;
425 T1W = Tm - Tn;
426 T10 = Ci[WS(csi, 9)];
427 T11 = Ci[WS(csi, 7)];
428 T12 = T10 - T11;
429 T1J = T10 + T11;
430 }
431 Tp = Tl + To;
432 T2i = T1I + T1J;
433 T2m = T1X - T1W;
434 TN = Tl - To;
435 T13 = TZ - T12;
436 T1K = T1I - T1J;
437 T1Y = T1W + T1X;
438 T1k = T12 + TZ;
439 }
440 {
441 E Ts, T1L, TT, T1M, Tv, T1O, TQ, T1P;
442 {
443 E Tq, Tr, TR, TS;
444 Tq = Cr[WS(csr, 5)];
445 Tr = Cr[WS(csr, 11)];
446 Ts = Tq + Tr;
447 T1L = Tq - Tr;
448 TR = Ci[WS(csi, 5)];
449 TS = Ci[WS(csi, 11)];
450 TT = TR - TS;
451 T1M = TR + TS;
452 }
453 {
454 E Tt, Tu, TO, TP;
455 Tt = Cr[WS(csr, 3)];
456 Tu = Cr[WS(csr, 13)];
457 Tv = Tt + Tu;
458 T1O = Tt - Tu;
459 TO = Ci[WS(csi, 13)];
460 TP = Ci[WS(csi, 3)];
461 TQ = TO - TP;
462 T1P = TP + TO;
463 }
464 Tw = Ts + Tv;
465 TU = TQ - TT;
466 T1l = TT + TQ;
467 TW = Ts - Tv;
468 {
469 E T1T, T1U, T1N, T1Q;
470 T1T = T1L + T1M;
471 T1U = T1O + T1P;
472 T1V = KP707106781 * (T1T - T1U);
473 T2j = KP707106781 * (T1T + T1U);
474 T1N = T1L - T1M;
475 T1Q = T1O - T1P;
476 T1R = KP707106781 * (T1N + T1Q);
477 T2l = KP707106781 * (T1N - T1Q);
478 }
479 }
480 {
481 E Tx, T1r, Ti, T1q, Ta;
482 Tx = KP2_000000000 * (Tp + Tw);
483 T1r = KP2_000000000 * (T1l + T1k);
484 Ta = T6 + T9;
485 Ti = Ta + Th;
486 T1q = Ta - Th;
487 R0[WS(rs, 8)] = Ti - Tx;
488 R0[WS(rs, 12)] = T1q + T1r;
489 R0[0] = Ti + Tx;
490 R0[WS(rs, 4)] = T1q - T1r;
491 }
492 {
493 E T1i, T1o, T1n, T1p, T1g, T1j, T1m;
494 T1g = T6 - T9;
495 T1i = T1g - T1h;
496 T1o = T1g + T1h;
497 T1j = Tp - Tw;
498 T1m = T1k - T1l;
499 T1n = KP1_414213562 * (T1j - T1m);
500 T1p = KP1_414213562 * (T1j + T1m);
501 R0[WS(rs, 10)] = T1i - T1n;
502 R0[WS(rs, 14)] = T1o + T1p;
503 R0[WS(rs, 2)] = T1i + T1n;
504 R0[WS(rs, 6)] = T1o - T1p;
505 }
506 {
507 E TM, T16, T15, T17;
508 {
509 E TC, TL, TV, T14;
510 TC = Ty - TB;
511 TL = KP1_414213562 * (TD - TK);
512 TM = TC + TL;
513 T16 = TC - TL;
514 TV = TN + TU;
515 T14 = TW + T13;
516 T15 = FNMS(KP765366864, T14, KP1_847759065 * TV);
517 T17 = FMA(KP765366864, TV, KP1_847759065 * T14);
518 }
519 R0[WS(rs, 9)] = TM - T15;
520 R0[WS(rs, 13)] = T16 + T17;
521 R0[WS(rs, 1)] = TM + T15;
522 R0[WS(rs, 5)] = T16 - T17;
523 }
524 {
525 E T2t, T2x, T2w, T2y;
526 {
527 E T2r, T2s, T2u, T2v;
528 T2r = T2b + T2c;
529 T2s = FMA(KP1_847759065, T2e, KP765366864 * T2f);
530 T2t = T2r - T2s;
531 T2x = T2r + T2s;
532 T2u = T2i + T2j;
533 T2v = T2m - T2l;
534 T2w = FNMS(KP1_961570560, T2v, KP390180644 * T2u);
535 T2y = FMA(KP1_961570560, T2u, KP390180644 * T2v);
536 }
537 R1[WS(rs, 11)] = T2t - T2w;
538 R1[WS(rs, 15)] = T2x + T2y;
539 R1[WS(rs, 3)] = T2t + T2w;
540 R1[WS(rs, 7)] = T2x - T2y;
541 }
542 {
543 E T1a, T1e, T1d, T1f;
544 {
545 E T18, T19, T1b, T1c;
546 T18 = Ty + TB;
547 T19 = KP1_414213562 * (TD + TK);
548 T1a = T18 - T19;
549 T1e = T18 + T19;
550 T1b = TN - TU;
551 T1c = T13 - TW;
552 T1d = FNMS(KP1_847759065, T1c, KP765366864 * T1b);
553 T1f = FMA(KP1_847759065, T1b, KP765366864 * T1c);
554 }
555 R0[WS(rs, 11)] = T1a - T1d;
556 R0[WS(rs, 15)] = T1e + T1f;
557 R0[WS(rs, 3)] = T1a + T1d;
558 R0[WS(rs, 7)] = T1e - T1f;
559 }
560 {
561 E T25, T29, T28, T2a;
562 {
563 E T23, T24, T26, T27;
564 T23 = T1v - T1y;
565 T24 = FMA(KP765366864, T1C, KP1_847759065 * T1F);
566 T25 = T23 - T24;
567 T29 = T23 + T24;
568 T26 = T1K - T1R;
569 T27 = T1Y - T1V;
570 T28 = FNMS(KP1_662939224, T27, KP1_111140466 * T26);
571 T2a = FMA(KP1_662939224, T26, KP1_111140466 * T27);
572 }
573 R1[WS(rs, 10)] = T25 - T28;
574 R1[WS(rs, 14)] = T29 + T2a;
575 R1[WS(rs, 2)] = T25 + T28;
576 R1[WS(rs, 6)] = T29 - T2a;
577 }
578 {
579 E T2h, T2p, T2o, T2q;
580 {
581 E T2d, T2g, T2k, T2n;
582 T2d = T2b - T2c;
583 T2g = FNMS(KP1_847759065, T2f, KP765366864 * T2e);
584 T2h = T2d + T2g;
585 T2p = T2d - T2g;
586 T2k = T2i - T2j;
587 T2n = T2l + T2m;
588 T2o = FNMS(KP1_111140466, T2n, KP1_662939224 * T2k);
589 T2q = FMA(KP1_111140466, T2k, KP1_662939224 * T2n);
590 }
591 R1[WS(rs, 9)] = T2h - T2o;
592 R1[WS(rs, 13)] = T2p + T2q;
593 R1[WS(rs, 1)] = T2h + T2o;
594 R1[WS(rs, 5)] = T2p - T2q;
595 }
596 {
597 E T1H, T21, T20, T22;
598 {
599 E T1z, T1G, T1S, T1Z;
600 T1z = T1v + T1y;
601 T1G = FNMS(KP765366864, T1F, KP1_847759065 * T1C);
602 T1H = T1z + T1G;
603 T21 = T1z - T1G;
604 T1S = T1K + T1R;
605 T1Z = T1V + T1Y;
606 T20 = FNMS(KP390180644, T1Z, KP1_961570560 * T1S);
607 T22 = FMA(KP390180644, T1S, KP1_961570560 * T1Z);
608 }
609 R1[WS(rs, 8)] = T1H - T20;
610 R1[WS(rs, 12)] = T21 + T22;
611 R1[0] = T1H + T20;
612 R1[WS(rs, 4)] = T21 - T22;
613 }
614 }
615 }
616 }
617
618 static const kr2c_desc desc = { 32, "r2cb_32", {140, 34, 16, 0}, &GENUS };
619
620 void X(codelet_r2cb_32) (planner *p) {
621 X(kr2c_register) (p, r2cb_32, &desc);
622 }
623
624 #endif /* HAVE_FMA */