Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t3bv_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 07:39:23 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 t3bv_20 -include t3b.h -sign 1 */
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 "t3b.h"
36
37 static void t3bv_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 = ii;
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 T19, T1u, T1p, T1x, T1m, T1w, T1t, TI;
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, T1g, T1F, T23, T1n, Tp, T18, T27, T1P, T1I, TU, T1L, T28, T1S, T1o;
57 V TE, T1l, T1j, T26, T2e;
58 {
59 V T1, T1e, T5, T1b;
60 T1 = LD(&(x[0]), ms, &(x[0]));
61 T1e = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
62 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
63 T1b = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
64 {
65 V TA, Tx, TQ, T1O, T10, Th, T1G, T1R, T17, T1J, To, Ts, TR, Tv, TK;
66 V TM, TP, Ty, TB;
67 {
68 V Tq, Tt, T13, T16, Tk, Tn;
69 {
70 V Tl, Ti, T11, T14, TV, Tc, T6, Tb, Tf, TW, TY, T1f;
71 {
72 V T1d, 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 T11 = VZMULJ(TA, Td);
83 T14 = VZMULJ(TA, T3);
84 TQ = VZMUL(TA, T3);
85 T1d = VZMULJ(T9, Td);
86 TV = VZMUL(T9, T3);
87 Tc = VZMULJ(T9, T3);
88 T6 = VZMUL(T4, T5);
89 Tb = VZMUL(T9, Ta);
90 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
91 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
92 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
93 T1f = VZMUL(T1d, T1e);
94 }
95 {
96 V T1D, TX, TZ, T15, T1E, Tg, T12, T1c, Te, Tj, Tm;
97 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
98 T1c = VZMUL(Tc, T1b);
99 Te = VZMULJ(Tc, Td);
100 T7 = VSUB(T1, T6);
101 T1D = VADD(T1, T6);
102 TX = VZMUL(TV, TW);
103 TZ = VZMUL(T8, TY);
104 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
105 T13 = VZMUL(T11, T12);
106 T1g = VSUB(T1c, T1f);
107 T1E = VADD(T1c, T1f);
108 Tg = VZMUL(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(TX, TZ);
112 T10 = VSUB(TX, TZ);
113 T16 = VZMUL(T14, T15);
114 T1F = VSUB(T1D, T1E);
115 T23 = VADD(T1D, T1E);
116 Th = VSUB(Tb, Tg);
117 T1G = VADD(Tb, Tg);
118 Tk = VZMUL(Ti, Tj);
119 Tn = VZMUL(Tl, Tm);
120 }
121 }
122 {
123 V Tr, Tu, TJ, TL, TO;
124 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
125 T1R = VADD(T13, T16);
126 T17 = VSUB(T13, T16);
127 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
128 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
129 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
130 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
131 T1J = VADD(Tk, Tn);
132 To = VSUB(Tk, Tn);
133 Ts = VZMUL(Tq, Tr);
134 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
135 Tv = VZMUL(Tt, Tu);
136 TK = VZMUL(T3, TJ);
137 TM = VZMUL(Td, TL);
138 TP = VZMUL(T2, TO);
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, TN, Tz, TC, T1i, TT, T1K, TS;
145 T1n = VSUB(Th, To);
146 Tp = VADD(Th, To);
147 TS = VZMUL(TQ, TR);
148 T1N = VADD(Ts, Tv);
149 Tw = VSUB(Ts, Tv);
150 T1H = VADD(TK, TM);
151 TN = VSUB(TK, TM);
152 Tz = VZMUL(Tx, Ty);
153 TC = VZMUL(TA, TB);
154 T18 = VSUB(T10, T17);
155 T1i = VADD(T10, T17);
156 TT = VSUB(TP, TS);
157 T1K = VADD(TP, TS);
158 T27 = VADD(T1N, T1O);
159 T1P = VSUB(T1N, T1O);
160 {
161 V TD, T1Q, T24, T1h, T25;
162 TD = VSUB(Tz, TC);
163 T1Q = VADD(Tz, TC);
164 T1I = VSUB(T1G, T1H);
165 T24 = VADD(T1G, T1H);
166 T1h = VADD(TN, TT);
167 TU = VSUB(TN, TT);
168 T25 = VADD(T1J, T1K);
169 T1L = VSUB(T1J, T1K);
170 T28 = VADD(T1Q, T1R);
171 T1S = VSUB(T1Q, T1R);
172 T1o = VSUB(Tw, TD);
173 TE = VADD(Tw, TD);
174 T1l = VSUB(T1h, T1i);
175 T1j = VADD(T1h, T1i);
176 T26 = VADD(T24, T25);
177 T2e = VSUB(T24, T25);
178 }
179 }
180 }
181 }
182 {
183 V T1M, T1Z, T1Y, T1T, T29, T2f, TH, TF, T1k, T1C;
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 TH = VSUB(Tp, TE);
191 TF = VADD(Tp, TE);
192 T1k = VFNMS(LDK(KP250000000), T1j, T1g);
193 T1C = VADD(T1g, T1j);
194 {
195 V T1W, T2c, TG, T2i, T2g, T22, T20, T1V, T2b, T1U, T2a, T1B;
196 T19 = VFMA(LDK(KP618033988), T18, TU);
197 T1u = VFNMS(LDK(KP618033988), TU, T18);
198 T1W = VSUB(T1M, T1T);
199 T1U = VADD(T1M, T1T);
200 T2c = VSUB(T26, T29);
201 T2a = VADD(T26, T29);
202 TG = VFNMS(LDK(KP250000000), TF, T7);
203 T1B = 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, 5)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)]));
213 ST(&(x[WS(rs, 15)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)]));
214 T1p = VFMA(LDK(KP618033988), T1o, T1n);
215 T1x = VFNMS(LDK(KP618033988), T1n, T1o);
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)]), VFMAI(T20, T1X), ms, &(x[0]));
223 ST(&(x[WS(rs, 2)]), VFNMSI(T20, T1X), ms, &(x[0]));
224 ST(&(x[WS(rs, 14)]), VFNMSI(T22, T21), ms, &(x[0]));
225 ST(&(x[WS(rs, 6)]), VFMAI(T22, T21), ms, &(x[0]));
226 ST(&(x[WS(rs, 16)]), VFMAI(T2g, T2d), ms, &(x[0]));
227 ST(&(x[WS(rs, 4)]), VFNMSI(T2g, T2d), ms, &(x[0]));
228 ST(&(x[WS(rs, 12)]), VFNMSI(T2i, T2h), ms, &(x[0]));
229 ST(&(x[WS(rs, 8)]), VFMAI(T2i, T2h), ms, &(x[0]));
230 T1m = VFMA(LDK(KP559016994), T1l, T1k);
231 T1w = VFNMS(LDK(KP559016994), T1l, T1k);
232 T1t = VFNMS(LDK(KP559016994), TH, TG);
233 TI = VFMA(LDK(KP559016994), TH, TG);
234 }
235 }
236 }
237 }
238 }
239 {
240 V T1A, T1y, T1q, T1s, T1a, T1r, T1z, T1v;
241 T1A = VFMA(LDK(KP951056516), T1x, T1w);
242 T1y = VFNMS(LDK(KP951056516), T1x, T1w);
243 T1q = VFMA(LDK(KP951056516), T1p, T1m);
244 T1s = VFNMS(LDK(KP951056516), T1p, T1m);
245 T1a = VFNMS(LDK(KP951056516), T19, TI);
246 T1r = VFMA(LDK(KP951056516), T19, TI);
247 T1z = VFNMS(LDK(KP951056516), T1u, T1t);
248 T1v = VFMA(LDK(KP951056516), T1u, T1t);
249 ST(&(x[WS(rs, 9)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)]));
250 ST(&(x[WS(rs, 11)]), VFNMSI(T1s, T1r), ms, &(x[WS(rs, 1)]));
251 ST(&(x[WS(rs, 1)]), VFMAI(T1q, T1a), ms, &(x[WS(rs, 1)]));
252 ST(&(x[WS(rs, 19)]), VFNMSI(T1q, T1a), ms, &(x[WS(rs, 1)]));
253 ST(&(x[WS(rs, 17)]), VFMAI(T1y, T1v), ms, &(x[WS(rs, 1)]));
254 ST(&(x[WS(rs, 3)]), VFNMSI(T1y, T1v), ms, &(x[WS(rs, 1)]));
255 ST(&(x[WS(rs, 13)]), VFMAI(T1A, T1z), ms, &(x[WS(rs, 1)]));
256 ST(&(x[WS(rs, 7)]), VFNMSI(T1A, T1z), 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("t3bv_20"), twinstr, &GENUS, {92, 72, 46, 0}, 0, 0, 0 };
272
273 void XSIMD(codelet_t3bv_20) (planner *p) {
274 X(kdft_dit_register) (p, t3bv_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 t3bv_20 -include t3b.h -sign 1 */
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 "t3b.h"
286
287 static void t3bv_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 = ii;
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, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
299 V T1g, Ti, Tt, T11;
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 TV = VZMUL(T9, T3);
308 T14 = VZMULJ(TA, T3);
309 Tl = VZMULJ(T8, T3);
310 Tq = VZMULJ(T2, T3);
311 Tx = VZMUL(T8, T3);
312 TQ = VZMUL(TA, T3);
313 Td = LDW(&(W[TWVL * 6]));
314 Te = VZMULJ(Tc, Td);
315 T1g = VZMULJ(T9, Td);
316 Ti = VZMULJ(T8, Td);
317 Tt = VZMULJ(T2, Td);
318 T11 = VZMULJ(TA, Td);
319 {
320 V T7, T1j, T1U, T2a, TU, T1n, T1o, T18, Tp, TE, TF, T26, T27, T28, T1M;
321 V T1P, T1W, T1b, T1c, T1k, T23, T24, T25, T1F, T1I, T1V, T1B, T1C;
322 {
323 V T1, T1i, T6, T1f, T1h, T5, T1e, T1S, T1T;
324 T1 = LD(&(x[0]), ms, &(x[0]));
325 T1h = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
326 T1i = VZMUL(T1g, T1h);
327 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
328 T6 = VZMUL(T4, T5);
329 T1e = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
330 T1f = VZMUL(Tc, T1e);
331 T7 = VSUB(T1, T6);
332 T1j = VSUB(T1f, T1i);
333 T1S = VADD(T1, T6);
334 T1T = VADD(T1f, T1i);
335 T1U = VSUB(T1S, T1T);
336 T2a = VADD(T1S, T1T);
337 }
338 {
339 V Th, T1D, T10, T1L, T17, T1O, To, T1G, Tw, T1K, TN, T1E, TT, T1H, TD;
340 V T1N;
341 {
342 V Tb, Tg, Ta, Tf;
343 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
344 Tb = VZMUL(T9, Ta);
345 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
346 Tg = VZMUL(Te, Tf);
347 Th = VSUB(Tb, Tg);
348 T1D = VADD(Tb, Tg);
349 }
350 {
351 V TX, TZ, TW, TY;
352 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
353 TX = VZMUL(TV, TW);
354 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
355 TZ = VZMUL(T8, TY);
356 T10 = VSUB(TX, TZ);
357 T1L = VADD(TX, TZ);
358 }
359 {
360 V T13, T16, T12, T15;
361 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
362 T13 = VZMUL(T11, T12);
363 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
364 T16 = VZMUL(T14, T15);
365 T17 = VSUB(T13, T16);
366 T1O = VADD(T13, T16);
367 }
368 {
369 V Tk, Tn, Tj, Tm;
370 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
371 Tk = VZMUL(Ti, Tj);
372 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
373 Tn = VZMUL(Tl, Tm);
374 To = VSUB(Tk, Tn);
375 T1G = VADD(Tk, Tn);
376 }
377 {
378 V Ts, Tv, Tr, Tu;
379 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
380 Ts = VZMUL(Tq, Tr);
381 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
382 Tv = VZMUL(Tt, Tu);
383 Tw = VSUB(Ts, Tv);
384 T1K = VADD(Ts, Tv);
385 }
386 {
387 V TK, TM, TJ, TL;
388 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
389 TK = VZMUL(T3, TJ);
390 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
391 TM = VZMUL(Td, TL);
392 TN = VSUB(TK, TM);
393 T1E = VADD(TK, TM);
394 }
395 {
396 V TP, TS, TO, TR;
397 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
398 TP = VZMUL(T2, TO);
399 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
400 TS = VZMUL(TQ, TR);
401 TT = VSUB(TP, TS);
402 T1H = VADD(TP, TS);
403 }
404 {
405 V Tz, TC, Ty, TB;
406 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
407 Tz = VZMUL(Tx, Ty);
408 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
409 TC = VZMUL(TA, TB);
410 TD = VSUB(Tz, TC);
411 T1N = VADD(Tz, TC);
412 }
413 TU = VSUB(TN, TT);
414 T1n = VSUB(Th, To);
415 T1o = VSUB(Tw, TD);
416 T18 = VSUB(T10, T17);
417 Tp = VADD(Th, To);
418 TE = VADD(Tw, TD);
419 TF = VADD(Tp, TE);
420 T26 = VADD(T1K, T1L);
421 T27 = VADD(T1N, T1O);
422 T28 = VADD(T26, T27);
423 T1M = VSUB(T1K, T1L);
424 T1P = VSUB(T1N, T1O);
425 T1W = VADD(T1M, T1P);
426 T1b = VADD(TN, TT);
427 T1c = VADD(T10, T17);
428 T1k = VADD(T1b, T1c);
429 T23 = VADD(T1D, T1E);
430 T24 = VADD(T1G, T1H);
431 T25 = VADD(T23, T24);
432 T1F = VSUB(T1D, T1E);
433 T1I = VSUB(T1G, T1H);
434 T1V = VADD(T1F, T1I);
435 }
436 T1B = VADD(T7, TF);
437 T1C = VBYI(VADD(T1j, T1k));
438 ST(&(x[WS(rs, 15)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)]));
439 ST(&(x[WS(rs, 5)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)]));
440 {
441 V T29, T2b, T2c, T2g, T2i, T2e, T2f, T2h, T2d;
442 T29 = VMUL(LDK(KP559016994), VSUB(T25, T28));
443 T2b = VADD(T25, T28);
444 T2c = VFNMS(LDK(KP250000000), T2b, T2a);
445 T2e = VSUB(T23, T24);
446 T2f = VSUB(T26, T27);
447 T2g = VBYI(VFMA(LDK(KP951056516), T2e, VMUL(LDK(KP587785252), T2f)));
448 T2i = VBYI(VFNMS(LDK(KP951056516), T2f, VMUL(LDK(KP587785252), T2e)));
449 ST(&(x[0]), VADD(T2a, T2b), ms, &(x[0]));
450 T2h = VSUB(T2c, T29);
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 T2d = VADD(T29, T2c);
454 ST(&(x[WS(rs, 4)]), VSUB(T2d, T2g), ms, &(x[0]));
455 ST(&(x[WS(rs, 16)]), VADD(T2g, T2d), ms, &(x[0]));
456 }
457 {
458 V T1Z, T1X, T1Y, T1R, T21, T1J, T1Q, T22, 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(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J)));
465 T21 = VBYI(VFMA(LDK(KP951056516), T1J, VMUL(LDK(KP587785252), T1Q)));
466 ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0]));
467 T22 = VADD(T1Z, T1Y);
468 ST(&(x[WS(rs, 6)]), VADD(T21, T22), ms, &(x[0]));
469 ST(&(x[WS(rs, 14)]), VSUB(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 T19, T1p, T1w, T1u, T1m, T1x, TI, T1t;
476 T19 = VFNMS(LDK(KP951056516), T18, VMUL(LDK(KP587785252), TU));
477 T1p = VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1n));
478 T1w = VFMA(LDK(KP951056516), T1n, VMUL(LDK(KP587785252), T1o));
479 T1u = VFMA(LDK(KP951056516), TU, VMUL(LDK(KP587785252), T18));
480 {
481 V T1d, T1l, TG, TH;
482 T1d = VMUL(LDK(KP559016994), VSUB(T1b, T1c));
483 T1l = VFNMS(LDK(KP250000000), T1k, T1j);
484 T1m = VSUB(T1d, T1l);
485 T1x = VADD(T1d, T1l);
486 TG = VFNMS(LDK(KP250000000), TF, T7);
487 TH = VMUL(LDK(KP559016994), VSUB(Tp, TE));
488 TI = VSUB(TG, TH);
489 T1t = VADD(TH, TG);
490 }
491 {
492 V T1a, T1q, T1z, T1A;
493 T1a = VSUB(TI, T19);
494 T1q = VBYI(VSUB(T1m, T1p));
495 ST(&(x[WS(rs, 17)]), VSUB(T1a, T1q), ms, &(x[WS(rs, 1)]));
496 ST(&(x[WS(rs, 3)]), VADD(T1a, T1q), ms, &(x[WS(rs, 1)]));
497 T1z = VADD(T1t, T1u);
498 T1A = VBYI(VSUB(T1x, T1w));
499 ST(&(x[WS(rs, 11)]), VSUB(T1z, T1A), ms, &(x[WS(rs, 1)]));
500 ST(&(x[WS(rs, 9)]), VADD(T1z, T1A), ms, &(x[WS(rs, 1)]));
501 }
502 {
503 V T1r, T1s, T1v, T1y;
504 T1r = VADD(TI, T19);
505 T1s = VBYI(VADD(T1p, T1m));
506 ST(&(x[WS(rs, 13)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)]));
507 ST(&(x[WS(rs, 7)]), VADD(T1r, T1s), ms, &(x[WS(rs, 1)]));
508 T1v = VSUB(T1t, T1u);
509 T1y = VBYI(VADD(T1w, T1x));
510 ST(&(x[WS(rs, 19)]), VSUB(T1v, T1y), ms, &(x[WS(rs, 1)]));
511 ST(&(x[WS(rs, 1)]), VADD(T1v, T1y), 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("t3bv_20"), twinstr, &GENUS, {126, 80, 12, 0}, 0, 0, 0 };
529
530 void XSIMD(codelet_t3bv_20) (planner *p) {
531 X(kdft_dit_register) (p, t3bv_20, &desc);
532 }
533 #endif /* HAVE_FMA */