Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/t3bv_20.c @ 167:bd3cc4d1df30

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 <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
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 08:06:09 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_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 20 -name t3bv_20 -include dft/simd/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 * 73 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "dft/simd/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(KP618033988, +0.618033988749894848204586834365638117720309180);
41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
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 T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
49 V T1d, Ti, Tt, T11;
50 T2 = LDW(&(W[0]));
51 T8 = LDW(&(W[TWVL * 2]));
52 T9 = VZMUL(T2, T8);
53 TA = VZMULJ(T2, T8);
54 T3 = LDW(&(W[TWVL * 4]));
55 Tc = VZMULJ(T9, T3);
56 T4 = VZMUL(T2, T3);
57 TV = VZMUL(T9, T3);
58 T14 = VZMULJ(TA, T3);
59 Tl = VZMULJ(T8, T3);
60 Tq = VZMULJ(T2, T3);
61 Tx = VZMUL(T8, T3);
62 TQ = VZMUL(TA, T3);
63 Td = LDW(&(W[TWVL * 6]));
64 Te = VZMULJ(Tc, Td);
65 T1d = VZMULJ(T9, Td);
66 Ti = VZMULJ(T8, Td);
67 Tt = VZMULJ(T2, Td);
68 T11 = VZMULJ(TA, Td);
69 {
70 V T7, T1g, T1F, T23, TU, T1n, T1o, T18, Tp, TE, TF, T27, T28, T29, T1P;
71 V T1S, T1T, T1h, T1i, T1j, T24, T25, T26, T1I, T1L, T1M, T1B, T1C;
72 {
73 V T1, T1f, T6, T1c, T1e, T5, T1b, T1D, T1E;
74 T1 = LD(&(x[0]), ms, &(x[0]));
75 T1e = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
76 T1f = VZMUL(T1d, T1e);
77 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
78 T6 = VZMUL(T4, T5);
79 T1b = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
80 T1c = VZMUL(Tc, T1b);
81 T7 = VSUB(T1, T6);
82 T1g = VSUB(T1c, T1f);
83 T1D = VADD(T1, T6);
84 T1E = VADD(T1c, T1f);
85 T1F = VSUB(T1D, T1E);
86 T23 = VADD(T1D, T1E);
87 }
88 {
89 V Th, T1G, T10, T1O, T17, T1R, To, T1J, Tw, T1N, TN, T1H, TT, T1K, TD;
90 V T1Q;
91 {
92 V Tb, Tg, Ta, Tf;
93 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
94 Tb = VZMUL(T9, Ta);
95 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
96 Tg = VZMUL(Te, Tf);
97 Th = VSUB(Tb, Tg);
98 T1G = VADD(Tb, Tg);
99 }
100 {
101 V TX, TZ, TW, TY;
102 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
103 TX = VZMUL(TV, TW);
104 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
105 TZ = VZMUL(T8, TY);
106 T10 = VSUB(TX, TZ);
107 T1O = VADD(TX, TZ);
108 }
109 {
110 V T13, T16, T12, T15;
111 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
112 T13 = VZMUL(T11, T12);
113 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
114 T16 = VZMUL(T14, T15);
115 T17 = VSUB(T13, T16);
116 T1R = VADD(T13, T16);
117 }
118 {
119 V Tk, Tn, Tj, Tm;
120 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
121 Tk = VZMUL(Ti, Tj);
122 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
123 Tn = VZMUL(Tl, Tm);
124 To = VSUB(Tk, Tn);
125 T1J = VADD(Tk, Tn);
126 }
127 {
128 V Ts, Tv, Tr, Tu;
129 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
130 Ts = VZMUL(Tq, Tr);
131 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
132 Tv = VZMUL(Tt, Tu);
133 Tw = VSUB(Ts, Tv);
134 T1N = VADD(Ts, Tv);
135 }
136 {
137 V TK, TM, TJ, TL;
138 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
139 TK = VZMUL(T3, TJ);
140 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
141 TM = VZMUL(Td, TL);
142 TN = VSUB(TK, TM);
143 T1H = VADD(TK, TM);
144 }
145 {
146 V TP, TS, TO, TR;
147 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
148 TP = VZMUL(T2, TO);
149 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
150 TS = VZMUL(TQ, TR);
151 TT = VSUB(TP, TS);
152 T1K = VADD(TP, TS);
153 }
154 {
155 V Tz, TC, Ty, TB;
156 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
157 Tz = VZMUL(Tx, Ty);
158 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
159 TC = VZMUL(TA, TB);
160 TD = VSUB(Tz, TC);
161 T1Q = VADD(Tz, TC);
162 }
163 TU = VSUB(TN, TT);
164 T1n = VSUB(Th, To);
165 T1o = VSUB(Tw, TD);
166 T18 = VSUB(T10, T17);
167 Tp = VADD(Th, To);
168 TE = VADD(Tw, TD);
169 TF = VADD(Tp, TE);
170 T27 = VADD(T1N, T1O);
171 T28 = VADD(T1Q, T1R);
172 T29 = VADD(T27, T28);
173 T1P = VSUB(T1N, T1O);
174 T1S = VSUB(T1Q, T1R);
175 T1T = VADD(T1P, T1S);
176 T1h = VADD(TN, TT);
177 T1i = VADD(T10, T17);
178 T1j = VADD(T1h, T1i);
179 T24 = VADD(T1G, T1H);
180 T25 = VADD(T1J, T1K);
181 T26 = VADD(T24, T25);
182 T1I = VSUB(T1G, T1H);
183 T1L = VSUB(T1J, T1K);
184 T1M = VADD(T1I, T1L);
185 }
186 T1B = VADD(T7, TF);
187 T1C = VADD(T1g, T1j);
188 ST(&(x[WS(rs, 15)]), VFNMSI(T1C, T1B), ms, &(x[WS(rs, 1)]));
189 ST(&(x[WS(rs, 5)]), VFMAI(T1C, T1B), ms, &(x[WS(rs, 1)]));
190 {
191 V T2c, T2a, T2b, T2g, T2i, T2e, T2f, T2h, T2d;
192 T2c = VSUB(T26, T29);
193 T2a = VADD(T26, T29);
194 T2b = VFNMS(LDK(KP250000000), T2a, T23);
195 T2e = VSUB(T24, T25);
196 T2f = VSUB(T27, T28);
197 T2g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T2f, T2e));
198 T2i = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T2e, T2f));
199 ST(&(x[0]), VADD(T23, T2a), ms, &(x[0]));
200 T2h = VFNMS(LDK(KP559016994), T2c, T2b);
201 ST(&(x[WS(rs, 8)]), VFMAI(T2i, T2h), ms, &(x[0]));
202 ST(&(x[WS(rs, 12)]), VFNMSI(T2i, T2h), ms, &(x[0]));
203 T2d = VFMA(LDK(KP559016994), T2c, T2b);
204 ST(&(x[WS(rs, 4)]), VFNMSI(T2g, T2d), ms, &(x[0]));
205 ST(&(x[WS(rs, 16)]), VFMAI(T2g, T2d), ms, &(x[0]));
206 }
207 {
208 V T1W, T1U, T1V, T20, T22, T1Y, T1Z, T21, T1X;
209 T1W = VSUB(T1M, T1T);
210 T1U = VADD(T1M, T1T);
211 T1V = VFNMS(LDK(KP250000000), T1U, T1F);
212 T1Y = VSUB(T1P, T1S);
213 T1Z = VSUB(T1I, T1L);
214 T20 = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1Z, T1Y));
215 T22 = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1Y, T1Z));
216 ST(&(x[WS(rs, 10)]), VADD(T1F, T1U), ms, &(x[0]));
217 T21 = VFMA(LDK(KP559016994), T1W, T1V);
218 ST(&(x[WS(rs, 6)]), VFMAI(T22, T21), ms, &(x[0]));
219 ST(&(x[WS(rs, 14)]), VFNMSI(T22, T21), ms, &(x[0]));
220 T1X = VFNMS(LDK(KP559016994), T1W, T1V);
221 ST(&(x[WS(rs, 2)]), VFNMSI(T20, T1X), ms, &(x[0]));
222 ST(&(x[WS(rs, 18)]), VFMAI(T20, T1X), ms, &(x[0]));
223 }
224 {
225 V T19, T1p, T1x, T1u, T1m, T1w, TI, T1t;
226 T19 = VFMA(LDK(KP618033988), T18, TU);
227 T1p = VFMA(LDK(KP618033988), T1o, T1n);
228 T1x = VFNMS(LDK(KP618033988), T1n, T1o);
229 T1u = VFNMS(LDK(KP618033988), TU, T18);
230 {
231 V T1k, T1l, TG, TH;
232 T1k = VFNMS(LDK(KP250000000), T1j, T1g);
233 T1l = VSUB(T1h, T1i);
234 T1m = VFMA(LDK(KP559016994), T1l, T1k);
235 T1w = VFNMS(LDK(KP559016994), T1l, T1k);
236 TG = VFNMS(LDK(KP250000000), TF, T7);
237 TH = VSUB(Tp, TE);
238 TI = VFMA(LDK(KP559016994), TH, TG);
239 T1t = VFNMS(LDK(KP559016994), TH, TG);
240 }
241 {
242 V T1a, T1q, T1z, T1A;
243 T1a = VFNMS(LDK(KP951056516), T19, TI);
244 T1q = VFMA(LDK(KP951056516), T1p, T1m);
245 ST(&(x[WS(rs, 19)]), VFNMSI(T1q, T1a), ms, &(x[WS(rs, 1)]));
246 ST(&(x[WS(rs, 1)]), VFMAI(T1q, T1a), ms, &(x[WS(rs, 1)]));
247 T1z = VFNMS(LDK(KP951056516), T1u, T1t);
248 T1A = VFMA(LDK(KP951056516), T1x, T1w);
249 ST(&(x[WS(rs, 7)]), VFNMSI(T1A, T1z), ms, &(x[WS(rs, 1)]));
250 ST(&(x[WS(rs, 13)]), VFMAI(T1A, T1z), ms, &(x[WS(rs, 1)]));
251 }
252 {
253 V T1r, T1s, T1v, T1y;
254 T1r = VFMA(LDK(KP951056516), T19, TI);
255 T1s = VFNMS(LDK(KP951056516), T1p, T1m);
256 ST(&(x[WS(rs, 11)]), VFNMSI(T1s, T1r), ms, &(x[WS(rs, 1)]));
257 ST(&(x[WS(rs, 9)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)]));
258 T1v = VFMA(LDK(KP951056516), T1u, T1t);
259 T1y = VFNMS(LDK(KP951056516), T1x, T1w);
260 ST(&(x[WS(rs, 3)]), VFNMSI(T1y, T1v), ms, &(x[WS(rs, 1)]));
261 ST(&(x[WS(rs, 17)]), VFMAI(T1y, T1v), ms, &(x[WS(rs, 1)]));
262 }
263 }
264 }
265 }
266 }
267 VLEAVE();
268 }
269
270 static const tw_instr twinstr[] = {
271 VTW(0, 1),
272 VTW(0, 3),
273 VTW(0, 9),
274 VTW(0, 19),
275 {TW_NEXT, VL, 0}
276 };
277
278 static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {92, 72, 46, 0}, 0, 0, 0 };
279
280 void XSIMD(codelet_t3bv_20) (planner *p) {
281 X(kdft_dit_register) (p, t3bv_20, &desc);
282 }
283 #else
284
285 /* 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 dft/simd/t3b.h -sign 1 */
286
287 /*
288 * This function contains 138 FP additions, 92 FP multiplications,
289 * (or, 126 additions, 80 multiplications, 12 fused multiply/add),
290 * 73 stack variables, 4 constants, and 40 memory accesses
291 */
292 #include "dft/simd/t3b.h"
293
294 static void t3bv_20(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
295 {
296 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
297 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
298 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
299 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
300 {
301 INT m;
302 R *x;
303 x = ii;
304 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)) {
305 V T2, T8, T9, TA, T3, Tc, T4, TV, T14, Tl, Tq, Tx, TQ, Td, Te;
306 V T1g, Ti, Tt, T11;
307 T2 = LDW(&(W[0]));
308 T8 = LDW(&(W[TWVL * 2]));
309 T9 = VZMUL(T2, T8);
310 TA = VZMULJ(T2, T8);
311 T3 = LDW(&(W[TWVL * 4]));
312 Tc = VZMULJ(T9, T3);
313 T4 = VZMUL(T2, T3);
314 TV = VZMUL(T9, T3);
315 T14 = VZMULJ(TA, T3);
316 Tl = VZMULJ(T8, T3);
317 Tq = VZMULJ(T2, T3);
318 Tx = VZMUL(T8, T3);
319 TQ = VZMUL(TA, T3);
320 Td = LDW(&(W[TWVL * 6]));
321 Te = VZMULJ(Tc, Td);
322 T1g = VZMULJ(T9, Td);
323 Ti = VZMULJ(T8, Td);
324 Tt = VZMULJ(T2, Td);
325 T11 = VZMULJ(TA, Td);
326 {
327 V T7, T1j, T1U, T2a, TU, T1n, T1o, T18, Tp, TE, TF, T26, T27, T28, T1M;
328 V T1P, T1W, T1b, T1c, T1k, T23, T24, T25, T1F, T1I, T1V, T1B, T1C;
329 {
330 V T1, T1i, T6, T1f, T1h, T5, T1e, T1S, T1T;
331 T1 = LD(&(x[0]), ms, &(x[0]));
332 T1h = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
333 T1i = VZMUL(T1g, T1h);
334 T5 = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
335 T6 = VZMUL(T4, T5);
336 T1e = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
337 T1f = VZMUL(Tc, T1e);
338 T7 = VSUB(T1, T6);
339 T1j = VSUB(T1f, T1i);
340 T1S = VADD(T1, T6);
341 T1T = VADD(T1f, T1i);
342 T1U = VSUB(T1S, T1T);
343 T2a = VADD(T1S, T1T);
344 }
345 {
346 V Th, T1D, T10, T1L, T17, T1O, To, T1G, Tw, T1K, TN, T1E, TT, T1H, TD;
347 V T1N;
348 {
349 V Tb, Tg, Ta, Tf;
350 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
351 Tb = VZMUL(T9, Ta);
352 Tf = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
353 Tg = VZMUL(Te, Tf);
354 Th = VSUB(Tb, Tg);
355 T1D = VADD(Tb, Tg);
356 }
357 {
358 V TX, TZ, TW, TY;
359 TW = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
360 TX = VZMUL(TV, TW);
361 TY = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
362 TZ = VZMUL(T8, TY);
363 T10 = VSUB(TX, TZ);
364 T1L = VADD(TX, TZ);
365 }
366 {
367 V T13, T16, T12, T15;
368 T12 = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)]));
369 T13 = VZMUL(T11, T12);
370 T15 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
371 T16 = VZMUL(T14, T15);
372 T17 = VSUB(T13, T16);
373 T1O = VADD(T13, T16);
374 }
375 {
376 V Tk, Tn, Tj, Tm;
377 Tj = LD(&(x[WS(rs, 16)]), ms, &(x[0]));
378 Tk = VZMUL(Ti, Tj);
379 Tm = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
380 Tn = VZMUL(Tl, Tm);
381 To = VSUB(Tk, Tn);
382 T1G = VADD(Tk, Tn);
383 }
384 {
385 V Ts, Tv, Tr, Tu;
386 Tr = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
387 Ts = VZMUL(Tq, Tr);
388 Tu = LD(&(x[WS(rs, 18)]), ms, &(x[0]));
389 Tv = VZMUL(Tt, Tu);
390 Tw = VSUB(Ts, Tv);
391 T1K = VADD(Ts, Tv);
392 }
393 {
394 V TK, TM, TJ, TL;
395 TJ = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
396 TK = VZMUL(T3, TJ);
397 TL = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)]));
398 TM = VZMUL(Td, TL);
399 TN = VSUB(TK, TM);
400 T1E = VADD(TK, TM);
401 }
402 {
403 V TP, TS, TO, TR;
404 TO = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
405 TP = VZMUL(T2, TO);
406 TR = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
407 TS = VZMUL(TQ, TR);
408 TT = VSUB(TP, TS);
409 T1H = VADD(TP, TS);
410 }
411 {
412 V Tz, TC, Ty, TB;
413 Ty = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
414 Tz = VZMUL(Tx, Ty);
415 TB = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
416 TC = VZMUL(TA, TB);
417 TD = VSUB(Tz, TC);
418 T1N = VADD(Tz, TC);
419 }
420 TU = VSUB(TN, TT);
421 T1n = VSUB(Th, To);
422 T1o = VSUB(Tw, TD);
423 T18 = VSUB(T10, T17);
424 Tp = VADD(Th, To);
425 TE = VADD(Tw, TD);
426 TF = VADD(Tp, TE);
427 T26 = VADD(T1K, T1L);
428 T27 = VADD(T1N, T1O);
429 T28 = VADD(T26, T27);
430 T1M = VSUB(T1K, T1L);
431 T1P = VSUB(T1N, T1O);
432 T1W = VADD(T1M, T1P);
433 T1b = VADD(TN, TT);
434 T1c = VADD(T10, T17);
435 T1k = VADD(T1b, T1c);
436 T23 = VADD(T1D, T1E);
437 T24 = VADD(T1G, T1H);
438 T25 = VADD(T23, T24);
439 T1F = VSUB(T1D, T1E);
440 T1I = VSUB(T1G, T1H);
441 T1V = VADD(T1F, T1I);
442 }
443 T1B = VADD(T7, TF);
444 T1C = VBYI(VADD(T1j, T1k));
445 ST(&(x[WS(rs, 15)]), VSUB(T1B, T1C), ms, &(x[WS(rs, 1)]));
446 ST(&(x[WS(rs, 5)]), VADD(T1B, T1C), ms, &(x[WS(rs, 1)]));
447 {
448 V T29, T2b, T2c, T2g, T2i, T2e, T2f, T2h, T2d;
449 T29 = VMUL(LDK(KP559016994), VSUB(T25, T28));
450 T2b = VADD(T25, T28);
451 T2c = VFNMS(LDK(KP250000000), T2b, T2a);
452 T2e = VSUB(T23, T24);
453 T2f = VSUB(T26, T27);
454 T2g = VBYI(VFMA(LDK(KP951056516), T2e, VMUL(LDK(KP587785252), T2f)));
455 T2i = VBYI(VFNMS(LDK(KP951056516), T2f, VMUL(LDK(KP587785252), T2e)));
456 ST(&(x[0]), VADD(T2a, T2b), ms, &(x[0]));
457 T2h = VSUB(T2c, T29);
458 ST(&(x[WS(rs, 8)]), VSUB(T2h, T2i), ms, &(x[0]));
459 ST(&(x[WS(rs, 12)]), VADD(T2i, T2h), ms, &(x[0]));
460 T2d = VADD(T29, T2c);
461 ST(&(x[WS(rs, 4)]), VSUB(T2d, T2g), ms, &(x[0]));
462 ST(&(x[WS(rs, 16)]), VADD(T2g, T2d), ms, &(x[0]));
463 }
464 {
465 V T1Z, T1X, T1Y, T1R, T21, T1J, T1Q, T22, T20;
466 T1Z = VMUL(LDK(KP559016994), VSUB(T1V, T1W));
467 T1X = VADD(T1V, T1W);
468 T1Y = VFNMS(LDK(KP250000000), T1X, T1U);
469 T1J = VSUB(T1F, T1I);
470 T1Q = VSUB(T1M, T1P);
471 T1R = VBYI(VFNMS(LDK(KP951056516), T1Q, VMUL(LDK(KP587785252), T1J)));
472 T21 = VBYI(VFMA(LDK(KP951056516), T1J, VMUL(LDK(KP587785252), T1Q)));
473 ST(&(x[WS(rs, 10)]), VADD(T1U, T1X), ms, &(x[0]));
474 T22 = VADD(T1Z, T1Y);
475 ST(&(x[WS(rs, 6)]), VADD(T21, T22), ms, &(x[0]));
476 ST(&(x[WS(rs, 14)]), VSUB(T22, T21), ms, &(x[0]));
477 T20 = VSUB(T1Y, T1Z);
478 ST(&(x[WS(rs, 2)]), VADD(T1R, T20), ms, &(x[0]));
479 ST(&(x[WS(rs, 18)]), VSUB(T20, T1R), ms, &(x[0]));
480 }
481 {
482 V T19, T1p, T1w, T1u, T1m, T1x, TI, T1t;
483 T19 = VFNMS(LDK(KP951056516), T18, VMUL(LDK(KP587785252), TU));
484 T1p = VFNMS(LDK(KP951056516), T1o, VMUL(LDK(KP587785252), T1n));
485 T1w = VFMA(LDK(KP951056516), T1n, VMUL(LDK(KP587785252), T1o));
486 T1u = VFMA(LDK(KP951056516), TU, VMUL(LDK(KP587785252), T18));
487 {
488 V T1d, T1l, TG, TH;
489 T1d = VMUL(LDK(KP559016994), VSUB(T1b, T1c));
490 T1l = VFNMS(LDK(KP250000000), T1k, T1j);
491 T1m = VSUB(T1d, T1l);
492 T1x = VADD(T1d, T1l);
493 TG = VFNMS(LDK(KP250000000), TF, T7);
494 TH = VMUL(LDK(KP559016994), VSUB(Tp, TE));
495 TI = VSUB(TG, TH);
496 T1t = VADD(TH, TG);
497 }
498 {
499 V T1a, T1q, T1z, T1A;
500 T1a = VSUB(TI, T19);
501 T1q = VBYI(VSUB(T1m, T1p));
502 ST(&(x[WS(rs, 17)]), VSUB(T1a, T1q), ms, &(x[WS(rs, 1)]));
503 ST(&(x[WS(rs, 3)]), VADD(T1a, T1q), ms, &(x[WS(rs, 1)]));
504 T1z = VADD(T1t, T1u);
505 T1A = VBYI(VSUB(T1x, T1w));
506 ST(&(x[WS(rs, 11)]), VSUB(T1z, T1A), ms, &(x[WS(rs, 1)]));
507 ST(&(x[WS(rs, 9)]), VADD(T1z, T1A), ms, &(x[WS(rs, 1)]));
508 }
509 {
510 V T1r, T1s, T1v, T1y;
511 T1r = VADD(TI, T19);
512 T1s = VBYI(VADD(T1p, T1m));
513 ST(&(x[WS(rs, 13)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)]));
514 ST(&(x[WS(rs, 7)]), VADD(T1r, T1s), ms, &(x[WS(rs, 1)]));
515 T1v = VSUB(T1t, T1u);
516 T1y = VBYI(VADD(T1w, T1x));
517 ST(&(x[WS(rs, 19)]), VSUB(T1v, T1y), ms, &(x[WS(rs, 1)]));
518 ST(&(x[WS(rs, 1)]), VADD(T1v, T1y), ms, &(x[WS(rs, 1)]));
519 }
520 }
521 }
522 }
523 }
524 VLEAVE();
525 }
526
527 static const tw_instr twinstr[] = {
528 VTW(0, 1),
529 VTW(0, 3),
530 VTW(0, 9),
531 VTW(0, 19),
532 {TW_NEXT, VL, 0}
533 };
534
535 static const ct_desc desc = { 20, XSIMD_STRING("t3bv_20"), twinstr, &GENUS, {126, 80, 12, 0}, 0, 0, 0 };
536
537 void XSIMD(codelet_t3bv_20) (planner *p) {
538 X(kdft_dit_register) (p, t3bv_20, &desc);
539 }
540 #endif