Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t3fv_20.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 07:38:55 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3fv_20 -include t3f.h */
29
30 /*
31 * This function contains 138 FP additions, 118 FP multiplications,
32 * (or, 92 additions, 72 multiplications, 46 fused multiply/add),
33 * 90 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "t3f.h"
36
37 static void t3fv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
41 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
43 {
44 INT m;
45 R *x;
46 x = ri;
47 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
48 V T1k, T1w, T1r, T1z, T1o, T1y, T1v, T1h;
49 {
50 V T2, T8, T3, Td;
51 T2 = LDW(&(W[0]));
52 T8 = LDW(&(W[TWVL * 2]));
53 T3 = LDW(&(W[TWVL * 4]));
54 Td = LDW(&(W[TWVL * 6]));
55 {
56 V T7, TM, T1F, T23, T1p, Tp, T1j, T27, T1P, T1I, T1i, T1L, T28, T1S, T1q;
57 V TE, T1n, T1d, T26, T2e;
58 {
59 V T1, TK, T5, TH;
60 T1 = LD(&(x[0]), ms, &(x[0]));
61 TK = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
62 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
63 TH = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
64 {
65 V TA, Tx, TU, T1O, T14, Th, T1G, T1R, T1b, T1J, To, Ts, TV, Tv, TO;
66 V TQ, TT, Ty, TB;
67 {
68 V Tq, Tt, T17, T1a, Tk, Tn;
69 {
70 V Tl, Ti, T15, T18, TZ, Tc, T6, Tb, Tf, T10, T12, TL;
71 {
72 V TJ, Ta, T9, T4;
73 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
74 TA = VZMULJ(T2, T8);
75 T9 = VZMUL(T2, T8);
76 Tx = VZMUL(T8, T3);
77 Tl = VZMULJ(T8, T3);
78 T4 = VZMUL(T2, T3);
79 Tq = VZMULJ(T2, T3);
80 Tt = VZMULJ(T2, Td);
81 Ti = VZMULJ(T8, Td);
82 T15 = VZMULJ(TA, Td);
83 T18 = VZMULJ(TA, T3);
84 TU = VZMUL(TA, T3);
85 TJ = VZMULJ(T9, Td);
86 TZ = VZMUL(T9, T3);
87 Tc = VZMULJ(T9, T3);
88 T6 = VZMULJ(T4, T5);
89 Tb = VZMULJ(T9, Ta);
90 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
91 T10 = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
92 T12 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
93 TL = VZMULJ(TJ, TK);
94 }
95 {
96 V T1D, T11, T13, T19, T1E, Tg, T16, TI, Te, Tj, Tm;
97 T16 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
98 TI = VZMULJ(Tc, TH);
99 Te = VZMULJ(Tc, Td);
100 T7 = VSUB(T1, T6);
101 T1D = VADD(T1, T6);
102 T11 = VZMULJ(TZ, T10);
103 T13 = VZMULJ(T8, T12);
104 T19 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
105 T17 = VZMULJ(T15, T16);
106 TM = VSUB(TI, TL);
107 T1E = VADD(TI, TL);
108 Tg = VZMULJ(Te, Tf);
109 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
110 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
111 T1O = VADD(T11, T13);
112 T14 = VSUB(T11, T13);
113 T1a = VZMULJ(T18, T19);
114 T1F = VSUB(T1D, T1E);
115 T23 = VADD(T1D, T1E);
116 Th = VSUB(Tb, Tg);
117 T1G = VADD(Tb, Tg);
118 Tk = VZMULJ(Ti, Tj);
119 Tn = VZMULJ(Tl, Tm);
120 }
121 }
122 {
123 V Tr, Tu, TN, TP, TS;
124 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
125 T1R = VADD(T17, T1a);
126 T1b = VSUB(T17, T1a);
127 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
128 TN = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
129 TP = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
130 TS = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
131 T1J = VADD(Tk, Tn);
132 To = VSUB(Tk, Tn);
133 Ts = VZMULJ(Tq, Tr);
134 TV = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
135 Tv = VZMULJ(Tt, Tu);
136 TO = VZMULJ(T3, TN);
137 TQ = VZMULJ(Td, TP);
138 TT = VZMULJ(T2, TS);
139 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
140 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
141 }
142 }
143 {
144 V T1N, Tw, T1H, TR, Tz, TC, T1c, TX, T1K, TW;
145 T1p = VSUB(Th, To);
146 Tp = VADD(Th, To);
147 TW = VZMULJ(TU, TV);
148 T1N = VADD(Ts, Tv);
149 Tw = VSUB(Ts, Tv);
150 T1H = VADD(TO, TQ);
151 TR = VSUB(TO, TQ);
152 Tz = VZMULJ(Tx, Ty);
153 TC = VZMULJ(TA, TB);
154 T1j = VSUB(T1b, T14);
155 T1c = VADD(T14, T1b);
156 TX = VSUB(TT, TW);
157 T1K = VADD(TT, TW);
158 T27 = VADD(T1N, T1O);
159 T1P = VSUB(T1N, T1O);
160 {
161 V TD, T1Q, T24, TY, T25;
162 TD = VSUB(Tz, TC);
163 T1Q = VADD(Tz, TC);
164 T1I = VSUB(T1G, T1H);
165 T24 = VADD(T1G, T1H);
166 TY = VADD(TR, TX);
167 T1i = VSUB(TX, TR);
168 T25 = VADD(T1J, T1K);
169 T1L = VSUB(T1J, T1K);
170 T28 = VADD(T1Q, T1R);
171 T1S = VSUB(T1Q, T1R);
172 T1q = VSUB(Tw, TD);
173 TE = VADD(Tw, TD);
174 T1n = VSUB(T1c, TY);
175 T1d = VADD(TY, T1c);
176 T26 = VADD(T24, T25);
177 T2e = VSUB(T24, T25);
178 }
179 }
180 }
181 }
182 {
183 V T1M, T1Z, T1Y, T1T, T29, T2f, T1g, TF, T1m, T1e;
184 T1M = VADD(T1I, T1L);
185 T1Z = VSUB(T1I, T1L);
186 T1Y = VSUB(T1P, T1S);
187 T1T = VADD(T1P, T1S);
188 T29 = VADD(T27, T28);
189 T2f = VSUB(T27, T28);
190 T1g = VSUB(Tp, TE);
191 TF = VADD(Tp, TE);
192 T1m = VFNMS(LDK(KP250000000), T1d, TM);
193 T1e = VADD(TM, T1d);
194 {
195 V T1W, T2c, T1f, T2i, T2g, T22, T20, T1V, T2b, T1U, T2a, TG;
196 T1k = VFMA(LDK(KP618033988), T1j, T1i);
197 T1w = VFNMS(LDK(KP618033988), T1i, T1j);
198 T1W = VSUB(T1M, T1T);
199 T1U = VADD(T1M, T1T);
200 T2c = VSUB(T26, T29);
201 T2a = VADD(T26, T29);
202 T1f = VFNMS(LDK(KP250000000), TF, T7);
203 TG = VADD(T7, TF);
204 T2i = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T2e, T2f));
205 T2g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T2f, T2e));
206 T22 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1Y, T1Z));
207 T20 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1Z, T1Y));
208 ST(&(x[WS(rs, 10)]), VADD(T1F, T1U), ms, &(x[0]));
209 T1V = VFNMS(LDK(KP250000000), T1U, T1F);
210 ST(&(x[0]), VADD(T23, T2a), ms, &(x[0]));
211 T2b = VFNMS(LDK(KP250000000), T2a, T23);
212 ST(&(x[WS(rs, 15)]), VFMAI(T1e, TG), ms, &(x[WS(rs, 1)]));
213 ST(&(x[WS(rs, 5)]), VFNMSI(T1e, TG), ms, &(x[WS(rs, 1)]));
214 T1r = VFMA(LDK(KP618033988), T1q, T1p);
215 T1z = VFNMS(LDK(KP618033988), T1p, T1q);
216 {
217 V T21, T1X, T2h, T2d;
218 T21 = VFMA(LDK(KP559016994), T1W, T1V);
219 T1X = VFNMS(LDK(KP559016994), T1W, T1V);
220 T2h = VFNMS(LDK(KP559016994), T2c, T2b);
221 T2d = VFMA(LDK(KP559016994), T2c, T2b);
222 ST(&(x[WS(rs, 18)]), VFNMSI(T20, T1X), ms, &(x[0]));
223 ST(&(x[WS(rs, 2)]), VFMAI(T20, T1X), ms, &(x[0]));
224 ST(&(x[WS(rs, 14)]), VFMAI(T22, T21), ms, &(x[0]));
225 ST(&(x[WS(rs, 6)]), VFNMSI(T22, T21), ms, &(x[0]));
226 ST(&(x[WS(rs, 16)]), VFNMSI(T2g, T2d), ms, &(x[0]));
227 ST(&(x[WS(rs, 4)]), VFMAI(T2g, T2d), ms, &(x[0]));
228 ST(&(x[WS(rs, 12)]), VFMAI(T2i, T2h), ms, &(x[0]));
229 ST(&(x[WS(rs, 8)]), VFNMSI(T2i, T2h), ms, &(x[0]));
230 T1o = VFNMS(LDK(KP559016994), T1n, T1m);
231 T1y = VFMA(LDK(KP559016994), T1n, T1m);
232 T1v = VFNMS(LDK(KP559016994), T1g, T1f);
233 T1h = VFMA(LDK(KP559016994), T1g, T1f);
234 }
235 }
236 }
237 }
238 }
239 {
240 V T1C, T1A, T1s, T1u, T1l, T1t, T1B, T1x;
241 T1C = VFMA(LDK(KP951056516), T1z, T1y);
242 T1A = VFNMS(LDK(KP951056516), T1z, T1y);
243 T1s = VFMA(LDK(KP951056516), T1r, T1o);
244 T1u = VFNMS(LDK(KP951056516), T1r, T1o);
245 T1l = VFMA(LDK(KP951056516), T1k, T1h);
246 T1t = VFNMS(LDK(KP951056516), T1k, T1h);
247 T1B = VFMA(LDK(KP951056516), T1w, T1v);
248 T1x = VFNMS(LDK(KP951056516), T1w, T1v);
249 ST(&(x[WS(rs, 11)]), VFMAI(T1u, T1t), ms, &(x[WS(rs, 1)]));
250 ST(&(x[WS(rs, 9)]), VFNMSI(T1u, T1t), ms, &(x[WS(rs, 1)]));
251 ST(&(x[WS(rs, 19)]), VFMAI(T1s, T1l), ms, &(x[WS(rs, 1)]));
252 ST(&(x[WS(rs, 1)]), VFNMSI(T1s, T1l), ms, &(x[WS(rs, 1)]));
253 ST(&(x[WS(rs, 3)]), VFMAI(T1A, T1x), ms, &(x[WS(rs, 1)]));
254 ST(&(x[WS(rs, 17)]), VFNMSI(T1A, T1x), ms, &(x[WS(rs, 1)]));
255 ST(&(x[WS(rs, 7)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)]));
256 ST(&(x[WS(rs, 13)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)]));
257 }
258 }
259 }
260 VLEAVE();
261 }
262
263 static const tw_instr twinstr[] = {
264 VTW(0, 1),
265 VTW(0, 3),
266 VTW(0, 9),
267 VTW(0, 19),
268 {TW_NEXT, VL, 0}
269 };
270
271 static const ct_desc desc = { 20, XSIMD_STRING("t3fv_20"), twinstr, &GENUS, {92, 72, 46, 0}, 0, 0, 0 };
272
273 void XSIMD(codelet_t3fv_20) (planner *p) {
274 X(kdft_dit_register) (p, t3fv_20, &desc);
275 }
276 #else /* HAVE_FMA */
277
278 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3fv_20 -include t3f.h */
279
280 /*
281 * This function contains 138 FP additions, 92 FP multiplications,
282 * (or, 126 additions, 80 multiplications, 12 fused multiply/add),
283 * 73 stack variables, 4 constants, and 40 memory accesses
284 */
285 #include "t3f.h"
286
287 static void t3fv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
288 {
289 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
290 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
291 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
292 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
293 {
294 INT m;
295 R *x;
296 x = ri;
297 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(20, rs)) {
298 V T2, T8, T9, TA, T3, Tc, T4, TZ, T18, Tl, Tq, Tx, TU, Td, Te;
299 V T15, Ti, Tt, TJ;
300 T2 = LDW(&(W[0]));
301 T8 = LDW(&(W[TWVL * 2]));
302 T9 = VZMUL(T2, T8);
303 TA = VZMULJ(T2, T8);
304 T3 = LDW(&(W[TWVL * 4]));
305 Tc = VZMULJ(T9, T3);
306 T4 = VZMUL(T2, T3);
307 TZ = VZMUL(T9, T3);
308 T18 = VZMULJ(TA, T3);
309 Tl = VZMULJ(T8, T3);
310 Tq = VZMULJ(T2, T3);
311 Tx = VZMUL(T8, T3);
312 TU = VZMUL(TA, T3);
313 Td = LDW(&(W[TWVL * 6]));
314 Te = VZMULJ(Tc, Td);
315 T15 = VZMULJ(TA, Td);
316 Ti = VZMULJ(T8, Td);
317 Tt = VZMULJ(T2, Td);
318 TJ = VZMULJ(T9, Td);
319 {
320 V T7, TM, T1U, T2d, T1i, T1p, T1q, T1j, Tp, TE, TF, T26, T27, T2b, T1M;
321 V T1P, T1V, TY, T1c, T1d, T23, T24, T2a, T1F, T1I, T1W, TG, T1e;
322 {
323 V T1, TL, T6, TI, TK, T5, TH, T1S, T1T;
324 T1 = LD(&(x[0]), ms, &(x[0]));
325 TK = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
326 TL = VZMULJ(TJ, TK);
327 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
328 T6 = VZMULJ(T4, T5);
329 TH = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
330 TI = VZMULJ(Tc, TH);
331 T7 = VSUB(T1, T6);
332 TM = VSUB(TI, TL);
333 T1S = VADD(T1, T6);
334 T1T = VADD(TI, TL);
335 T1U = VSUB(T1S, T1T);
336 T2d = VADD(T1S, T1T);
337 }
338 {
339 V Th, T1K, T14, T1E, T1b, T1H, To, T1N, Tw, T1D, TR, T1L, TX, T1O, TD;
340 V T1G;
341 {
342 V Tb, Tg, Ta, Tf;
343 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
344 Tb = VZMULJ(T9, Ta);
345 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
346 Tg = VZMULJ(Te, Tf);
347 Th = VSUB(Tb, Tg);
348 T1K = VADD(Tb, Tg);
349 }
350 {
351 V T11, T13, T10, T12;
352 T10 = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
353 T11 = VZMULJ(TZ, T10);
354 T12 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
355 T13 = VZMULJ(T8, T12);
356 T14 = VSUB(T11, T13);
357 T1E = VADD(T11, T13);
358 }
359 {
360 V T17, T1a, T16, T19;
361 T16 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
362 T17 = VZMULJ(T15, T16);
363 T19 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
364 T1a = VZMULJ(T18, T19);
365 T1b = VSUB(T17, T1a);
366 T1H = VADD(T17, T1a);
367 }
368 {
369 V Tk, Tn, Tj, Tm;
370 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
371 Tk = VZMULJ(Ti, Tj);
372 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
373 Tn = VZMULJ(Tl, Tm);
374 To = VSUB(Tk, Tn);
375 T1N = VADD(Tk, Tn);
376 }
377 {
378 V Ts, Tv, Tr, Tu;
379 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
380 Ts = VZMULJ(Tq, Tr);
381 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
382 Tv = VZMULJ(Tt, Tu);
383 Tw = VSUB(Ts, Tv);
384 T1D = VADD(Ts, Tv);
385 }
386 {
387 V TO, TQ, TN, TP;
388 TN = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
389 TO = VZMULJ(T3, TN);
390 TP = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
391 TQ = VZMULJ(Td, TP);
392 TR = VSUB(TO, TQ);
393 T1L = VADD(TO, TQ);
394 }
395 {
396 V TT, TW, TS, TV;
397 TS = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
398 TT = VZMULJ(T2, TS);
399 TV = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
400 TW = VZMULJ(TU, TV);
401 TX = VSUB(TT, TW);
402 T1O = VADD(TT, TW);
403 }
404 {
405 V Tz, TC, Ty, TB;
406 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
407 Tz = VZMULJ(Tx, Ty);
408 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
409 TC = VZMULJ(TA, TB);
410 TD = VSUB(Tz, TC);
411 T1G = VADD(Tz, TC);
412 }
413 T1i = VSUB(TX, TR);
414 T1p = VSUB(Th, To);
415 T1q = VSUB(Tw, TD);
416 T1j = VSUB(T1b, T14);
417 Tp = VADD(Th, To);
418 TE = VADD(Tw, TD);
419 TF = VADD(Tp, TE);
420 T26 = VADD(T1D, T1E);
421 T27 = VADD(T1G, T1H);
422 T2b = VADD(T26, T27);
423 T1M = VSUB(T1K, T1L);
424 T1P = VSUB(T1N, T1O);
425 T1V = VADD(T1M, T1P);
426 TY = VADD(TR, TX);
427 T1c = VADD(T14, T1b);
428 T1d = VADD(TY, T1c);
429 T23 = VADD(T1K, T1L);
430 T24 = VADD(T1N, T1O);
431 T2a = VADD(T23, T24);
432 T1F = VSUB(T1D, T1E);
433 T1I = VSUB(T1G, T1H);
434 T1W = VADD(T1F, T1I);
435 }
436 TG = VADD(T7, TF);
437 T1e = VBYI(VADD(TM, T1d));
438 ST(&(x[WS(rs, 5)]), VSUB(TG, T1e), ms, &(x[WS(rs, 1)]));
439 ST(&(x[WS(rs, 15)]), VADD(TG, T1e), ms, &(x[WS(rs, 1)]));
440 {
441 V T2c, T2e, T2f, T29, T2i, T25, T28, T2h, T2g;
442 T2c = VMUL(LDK(KP559016994), VSUB(T2a, T2b));
443 T2e = VADD(T2a, T2b);
444 T2f = VFNMS(LDK(KP250000000), T2e, T2d);
445 T25 = VSUB(T23, T24);
446 T28 = VSUB(T26, T27);
447 T29 = VBYI(VFMA(LDK(KP951056516), T25, VMUL(LDK(KP587785252), T28)));
448 T2i = VBYI(VFNMS(LDK(KP587785252), T25, VMUL(LDK(KP951056516), T28)));
449 ST(&(x[0]), VADD(T2d, T2e), ms, &(x[0]));
450 T2h = VSUB(T2f, T2c);
451 ST(&(x[WS(rs, 8)]), VSUB(T2h, T2i), ms, &(x[0]));
452 ST(&(x[WS(rs, 12)]), VADD(T2i, T2h), ms, &(x[0]));
453 T2g = VADD(T2c, T2f);
454 ST(&(x[WS(rs, 4)]), VADD(T29, T2g), ms, &(x[0]));
455 ST(&(x[WS(rs, 16)]), VSUB(T2g, T29), ms, &(x[0]));
456 }
457 {
458 V T1Z, T1X, T1Y, T1R, T22, T1J, T1Q, T21, T20;
459 T1Z = VMUL(LDK(KP559016994), VSUB(T1V, T1W));
460 T1X = VADD(T1V, T1W);
461 T1Y = VFNMS(LDK(KP250000000), T1X, T1U);
462 T1J = VSUB(T1F, T1I);
463 T1Q = VSUB(T1M, T1P);
464 T1R = VBYI(VFNMS(LDK(KP587785252), T1Q, VMUL(LDK(KP951056516), T1J)));
465 T22 = VBYI(VFMA(LDK(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J)));
466 ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0]));
467 T21 = VADD(T1Z, T1Y);
468 ST(&(x[WS(rs, 6)]), VSUB(T21, T22), ms, &(x[0]));
469 ST(&(x[WS(rs, 14)]), VADD(T22, T21), ms, &(x[0]));
470 T20 = VSUB(T1Y, T1Z);
471 ST(&(x[WS(rs, 2)]), VADD(T1R, T20), ms, &(x[0]));
472 ST(&(x[WS(rs, 18)]), VSUB(T20, T1R), ms, &(x[0]));
473 }
474 {
475 V T1k, T1r, T1z, T1w, T1o, T1y, T1h, T1v;
476 T1k = VFMA(LDK(KP951056516), T1i, VMUL(LDK(KP587785252), T1j));
477 T1r = VFMA(LDK(KP951056516), T1p, VMUL(LDK(KP587785252), T1q));
478 T1z = VFNMS(LDK(KP587785252), T1p, VMUL(LDK(KP951056516), T1q));
479 T1w = VFNMS(LDK(KP587785252), T1i, VMUL(LDK(KP951056516), T1j));
480 {
481 V T1m, T1n, T1f, T1g;
482 T1m = VFMS(LDK(KP250000000), T1d, TM);
483 T1n = VMUL(LDK(KP559016994), VSUB(T1c, TY));
484 T1o = VADD(T1m, T1n);
485 T1y = VSUB(T1n, T1m);
486 T1f = VMUL(LDK(KP559016994), VSUB(Tp, TE));
487 T1g = VFNMS(LDK(KP250000000), TF, T7);
488 T1h = VADD(T1f, T1g);
489 T1v = VSUB(T1g, T1f);
490 }
491 {
492 V T1l, T1s, T1B, T1C;
493 T1l = VADD(T1h, T1k);
494 T1s = VBYI(VSUB(T1o, T1r));
495 ST(&(x[WS(rs, 19)]), VSUB(T1l, T1s), ms, &(x[WS(rs, 1)]));
496 ST(&(x[WS(rs, 1)]), VADD(T1l, T1s), ms, &(x[WS(rs, 1)]));
497 T1B = VADD(T1v, T1w);
498 T1C = VBYI(VADD(T1z, T1y));
499 ST(&(x[WS(rs, 13)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)]));
500 ST(&(x[WS(rs, 7)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)]));
501 }
502 {
503 V T1t, T1u, T1x, T1A;
504 T1t = VSUB(T1h, T1k);
505 T1u = VBYI(VADD(T1r, T1o));
506 ST(&(x[WS(rs, 11)]), VSUB(T1t, T1u), ms, &(x[WS(rs, 1)]));
507 ST(&(x[WS(rs, 9)]), VADD(T1t, T1u), ms, &(x[WS(rs, 1)]));
508 T1x = VSUB(T1v, T1w);
509 T1A = VBYI(VSUB(T1y, T1z));
510 ST(&(x[WS(rs, 17)]), VSUB(T1x, T1A), ms, &(x[WS(rs, 1)]));
511 ST(&(x[WS(rs, 3)]), VADD(T1x, T1A), ms, &(x[WS(rs, 1)]));
512 }
513 }
514 }
515 }
516 }
517 VLEAVE();
518 }
519
520 static const tw_instr twinstr[] = {
521 VTW(0, 1),
522 VTW(0, 3),
523 VTW(0, 9),
524 VTW(0, 19),
525 {TW_NEXT, VL, 0}
526 };
527
528 static const ct_desc desc = { 20, XSIMD_STRING("t3fv_20"), twinstr, &GENUS, {126, 80, 12, 0}, 0, 0, 0 };
529
530 void XSIMD(codelet_t3fv_20) (planner *p) {
531 X(kdft_dit_register) (p, t3fv_20, &desc);
532 }
533 #endif /* HAVE_FMA */