Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2fv_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 08:05:11 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_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n2fv_20 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
29
30 /*
31 * This function contains 104 FP additions, 50 FP multiplications,
32 * (or, 58 additions, 4 multiplications, 46 fused multiply/add),
33 * 57 stack variables, 4 constants, and 50 memory accesses
34 */
35 #include "dft/simd/n2f.h"
36
37 static void n2fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
40 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
43 {
44 INT i;
45 const R *xi;
46 R *xo;
47 xi = ri;
48 xo = ro;
49 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
50 V T3, T1r, Tm, T13, TG, TN, TO, TH, T16, T19, T1a, T1v, T1w, T1x, T1s;
51 V T1t, T1u, T1d, T1g, T1h, Ti, TE, TB, TL;
52 {
53 V T1, T2, T11, Tk, Tl, T12;
54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
55 T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
56 T11 = VADD(T1, T2);
57 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
58 Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
59 T12 = VADD(Tk, Tl);
60 T3 = VSUB(T1, T2);
61 T1r = VADD(T11, T12);
62 Tm = VSUB(Tk, Tl);
63 T13 = VSUB(T11, T12);
64 }
65 {
66 V T6, T14, Tw, T1c, Tz, T1f, T9, T17, Td, T1b, Tp, T15, Ts, T18, Tg;
67 V T1e;
68 {
69 V T4, T5, Tu, Tv;
70 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
71 T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
72 T6 = VSUB(T4, T5);
73 T14 = VADD(T4, T5);
74 Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
75 Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
76 Tw = VSUB(Tu, Tv);
77 T1c = VADD(Tu, Tv);
78 }
79 {
80 V Tx, Ty, T7, T8;
81 Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
82 Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
83 Tz = VSUB(Tx, Ty);
84 T1f = VADD(Tx, Ty);
85 T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
86 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
87 T9 = VSUB(T7, T8);
88 T17 = VADD(T7, T8);
89 }
90 {
91 V Tb, Tc, Tn, To;
92 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
93 Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
94 Td = VSUB(Tb, Tc);
95 T1b = VADD(Tb, Tc);
96 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
97 To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
98 Tp = VSUB(Tn, To);
99 T15 = VADD(Tn, To);
100 }
101 {
102 V Tq, Tr, Te, Tf;
103 Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
104 Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
105 Ts = VSUB(Tq, Tr);
106 T18 = VADD(Tq, Tr);
107 Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
108 Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
109 Tg = VSUB(Te, Tf);
110 T1e = VADD(Te, Tf);
111 }
112 TG = VSUB(Ts, Tp);
113 TN = VSUB(T6, T9);
114 TO = VSUB(Td, Tg);
115 TH = VSUB(Tz, Tw);
116 T16 = VSUB(T14, T15);
117 T19 = VSUB(T17, T18);
118 T1a = VADD(T16, T19);
119 T1v = VADD(T1b, T1c);
120 T1w = VADD(T1e, T1f);
121 T1x = VADD(T1v, T1w);
122 T1s = VADD(T14, T15);
123 T1t = VADD(T17, T18);
124 T1u = VADD(T1s, T1t);
125 T1d = VSUB(T1b, T1c);
126 T1g = VSUB(T1e, T1f);
127 T1h = VADD(T1d, T1g);
128 {
129 V Ta, Th, Tt, TA;
130 Ta = VADD(T6, T9);
131 Th = VADD(Td, Tg);
132 Ti = VADD(Ta, Th);
133 TE = VSUB(Ta, Th);
134 Tt = VADD(Tp, Ts);
135 TA = VADD(Tw, Tz);
136 TB = VADD(Tt, TA);
137 TL = VSUB(TA, Tt);
138 }
139 }
140 {
141 V T1I, T1J, T1K, T1L, T1N, T1H, Tj, TC;
142 Tj = VADD(T3, Ti);
143 TC = VADD(Tm, TB);
144 T1H = VFNMSI(TC, Tj);
145 STM2(&(xo[10]), T1H, ovs, &(xo[2]));
146 T1I = VFMAI(TC, Tj);
147 STM2(&(xo[30]), T1I, ovs, &(xo[2]));
148 {
149 V T1A, T1y, T1z, T1E, T1G, T1C, T1D, T1F, T1B, T1M;
150 T1A = VSUB(T1u, T1x);
151 T1y = VADD(T1u, T1x);
152 T1z = VFNMS(LDK(KP250000000), T1y, T1r);
153 T1C = VSUB(T1s, T1t);
154 T1D = VSUB(T1v, T1w);
155 T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C));
156 T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D));
157 T1J = VADD(T1r, T1y);
158 STM2(&(xo[0]), T1J, ovs, &(xo[0]));
159 T1F = VFNMS(LDK(KP559016994), T1A, T1z);
160 T1K = VFNMSI(T1G, T1F);
161 STM2(&(xo[16]), T1K, ovs, &(xo[0]));
162 T1L = VFMAI(T1G, T1F);
163 STM2(&(xo[24]), T1L, ovs, &(xo[0]));
164 T1B = VFMA(LDK(KP559016994), T1A, T1z);
165 T1M = VFMAI(T1E, T1B);
166 STM2(&(xo[8]), T1M, ovs, &(xo[0]));
167 STN2(&(xo[8]), T1M, T1H, ovs);
168 T1N = VFNMSI(T1E, T1B);
169 STM2(&(xo[32]), T1N, ovs, &(xo[0]));
170 }
171 {
172 V T1O, T1P, T1R, T1S;
173 {
174 V T1k, T1i, T1j, T1o, T1q, T1m, T1n, T1p, T1Q, T1l;
175 T1k = VSUB(T1a, T1h);
176 T1i = VADD(T1a, T1h);
177 T1j = VFNMS(LDK(KP250000000), T1i, T13);
178 T1m = VSUB(T1d, T1g);
179 T1n = VSUB(T16, T19);
180 T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m));
181 T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n));
182 T1O = VADD(T13, T1i);
183 STM2(&(xo[20]), T1O, ovs, &(xo[0]));
184 T1p = VFMA(LDK(KP559016994), T1k, T1j);
185 T1P = VFNMSI(T1q, T1p);
186 STM2(&(xo[12]), T1P, ovs, &(xo[0]));
187 T1Q = VFMAI(T1q, T1p);
188 STM2(&(xo[28]), T1Q, ovs, &(xo[0]));
189 STN2(&(xo[28]), T1Q, T1I, ovs);
190 T1l = VFNMS(LDK(KP559016994), T1k, T1j);
191 T1R = VFMAI(T1o, T1l);
192 STM2(&(xo[4]), T1R, ovs, &(xo[0]));
193 T1S = VFNMSI(T1o, T1l);
194 STM2(&(xo[36]), T1S, ovs, &(xo[0]));
195 }
196 {
197 V TI, TP, TX, TU, TM, TW, TF, TT, TK, TD;
198 TI = VFMA(LDK(KP618033988), TH, TG);
199 TP = VFMA(LDK(KP618033988), TO, TN);
200 TX = VFNMS(LDK(KP618033988), TN, TO);
201 TU = VFNMS(LDK(KP618033988), TG, TH);
202 TK = VFNMS(LDK(KP250000000), TB, Tm);
203 TM = VFNMS(LDK(KP559016994), TL, TK);
204 TW = VFMA(LDK(KP559016994), TL, TK);
205 TD = VFNMS(LDK(KP250000000), Ti, T3);
206 TF = VFMA(LDK(KP559016994), TE, TD);
207 TT = VFNMS(LDK(KP559016994), TE, TD);
208 {
209 V TJ, TQ, T1T, T1U;
210 TJ = VFMA(LDK(KP951056516), TI, TF);
211 TQ = VFMA(LDK(KP951056516), TP, TM);
212 T1T = VFNMSI(TQ, TJ);
213 STM2(&(xo[2]), T1T, ovs, &(xo[2]));
214 STN2(&(xo[0]), T1J, T1T, ovs);
215 T1U = VFMAI(TQ, TJ);
216 STM2(&(xo[38]), T1U, ovs, &(xo[2]));
217 STN2(&(xo[36]), T1S, T1U, ovs);
218 }
219 {
220 V TZ, T10, T1V, T1W;
221 TZ = VFMA(LDK(KP951056516), TU, TT);
222 T10 = VFMA(LDK(KP951056516), TX, TW);
223 T1V = VFNMSI(T10, TZ);
224 STM2(&(xo[26]), T1V, ovs, &(xo[2]));
225 STN2(&(xo[24]), T1L, T1V, ovs);
226 T1W = VFMAI(T10, TZ);
227 STM2(&(xo[14]), T1W, ovs, &(xo[2]));
228 STN2(&(xo[12]), T1P, T1W, ovs);
229 }
230 {
231 V TR, TS, T1X, T1Y;
232 TR = VFNMS(LDK(KP951056516), TI, TF);
233 TS = VFNMS(LDK(KP951056516), TP, TM);
234 T1X = VFNMSI(TS, TR);
235 STM2(&(xo[18]), T1X, ovs, &(xo[2]));
236 STN2(&(xo[16]), T1K, T1X, ovs);
237 T1Y = VFMAI(TS, TR);
238 STM2(&(xo[22]), T1Y, ovs, &(xo[2]));
239 STN2(&(xo[20]), T1O, T1Y, ovs);
240 }
241 {
242 V TV, TY, T1Z, T20;
243 TV = VFNMS(LDK(KP951056516), TU, TT);
244 TY = VFNMS(LDK(KP951056516), TX, TW);
245 T1Z = VFNMSI(TY, TV);
246 STM2(&(xo[34]), T1Z, ovs, &(xo[2]));
247 STN2(&(xo[32]), T1N, T1Z, ovs);
248 T20 = VFMAI(TY, TV);
249 STM2(&(xo[6]), T20, ovs, &(xo[2]));
250 STN2(&(xo[4]), T1R, T20, ovs);
251 }
252 }
253 }
254 }
255 }
256 }
257 VLEAVE();
258 }
259
260 static const kdft_desc desc = { 20, XSIMD_STRING("n2fv_20"), {58, 4, 46, 0}, &GENUS, 0, 2, 0, 0 };
261
262 void XSIMD(codelet_n2fv_20) (planner *p) {
263 X(kdft_register) (p, n2fv_20, &desc);
264 }
265
266 #else
267
268 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 20 -name n2fv_20 -with-ostride 2 -include dft/simd/n2f.h -store-multiple 2 */
269
270 /*
271 * This function contains 104 FP additions, 24 FP multiplications,
272 * (or, 92 additions, 12 multiplications, 12 fused multiply/add),
273 * 57 stack variables, 4 constants, and 50 memory accesses
274 */
275 #include "dft/simd/n2f.h"
276
277 static void n2fv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
278 {
279 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
280 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
281 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
282 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
283 {
284 INT i;
285 const R *xi;
286 R *xo;
287 xi = ri;
288 xo = ro;
289 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(40, is), MAKE_VOLATILE_STRIDE(40, os)) {
290 V T3, T1B, Tm, T1i, TG, TN, TO, TH, T13, T16, T1k, T1u, T1v, T1z, T1r;
291 V T1s, T1y, T1a, T1d, T1j, Ti, TD, TB, TL;
292 {
293 V T1, T2, T1g, Tk, Tl, T1h;
294 T1 = LD(&(xi[0]), ivs, &(xi[0]));
295 T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
296 T1g = VADD(T1, T2);
297 Tk = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
298 Tl = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
299 T1h = VADD(Tk, Tl);
300 T3 = VSUB(T1, T2);
301 T1B = VADD(T1g, T1h);
302 Tm = VSUB(Tk, Tl);
303 T1i = VSUB(T1g, T1h);
304 }
305 {
306 V T6, T18, Tw, T12, Tz, T15, T9, T1b, Td, T11, Tp, T19, Ts, T1c, Tg;
307 V T14;
308 {
309 V T4, T5, Tu, Tv;
310 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
311 T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
312 T6 = VSUB(T4, T5);
313 T18 = VADD(T4, T5);
314 Tu = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
315 Tv = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
316 Tw = VSUB(Tu, Tv);
317 T12 = VADD(Tu, Tv);
318 }
319 {
320 V Tx, Ty, T7, T8;
321 Tx = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
322 Ty = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
323 Tz = VSUB(Tx, Ty);
324 T15 = VADD(Tx, Ty);
325 T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
326 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
327 T9 = VSUB(T7, T8);
328 T1b = VADD(T7, T8);
329 }
330 {
331 V Tb, Tc, Tn, To;
332 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
333 Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
334 Td = VSUB(Tb, Tc);
335 T11 = VADD(Tb, Tc);
336 Tn = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
337 To = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
338 Tp = VSUB(Tn, To);
339 T19 = VADD(Tn, To);
340 }
341 {
342 V Tq, Tr, Te, Tf;
343 Tq = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
344 Tr = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
345 Ts = VSUB(Tq, Tr);
346 T1c = VADD(Tq, Tr);
347 Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
348 Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
349 Tg = VSUB(Te, Tf);
350 T14 = VADD(Te, Tf);
351 }
352 TG = VSUB(Ts, Tp);
353 TN = VSUB(T6, T9);
354 TO = VSUB(Td, Tg);
355 TH = VSUB(Tz, Tw);
356 T13 = VSUB(T11, T12);
357 T16 = VSUB(T14, T15);
358 T1k = VADD(T13, T16);
359 T1u = VADD(T11, T12);
360 T1v = VADD(T14, T15);
361 T1z = VADD(T1u, T1v);
362 T1r = VADD(T18, T19);
363 T1s = VADD(T1b, T1c);
364 T1y = VADD(T1r, T1s);
365 T1a = VSUB(T18, T19);
366 T1d = VSUB(T1b, T1c);
367 T1j = VADD(T1a, T1d);
368 {
369 V Ta, Th, Tt, TA;
370 Ta = VADD(T6, T9);
371 Th = VADD(Td, Tg);
372 Ti = VADD(Ta, Th);
373 TD = VMUL(LDK(KP559016994), VSUB(Ta, Th));
374 Tt = VADD(Tp, Ts);
375 TA = VADD(Tw, Tz);
376 TB = VADD(Tt, TA);
377 TL = VMUL(LDK(KP559016994), VSUB(TA, Tt));
378 }
379 }
380 {
381 V T1I, T1J, T1K, T1L, T1N, T1H, Tj, TC;
382 Tj = VADD(T3, Ti);
383 TC = VBYI(VADD(Tm, TB));
384 T1H = VSUB(Tj, TC);
385 STM2(&(xo[10]), T1H, ovs, &(xo[2]));
386 T1I = VADD(Tj, TC);
387 STM2(&(xo[30]), T1I, ovs, &(xo[2]));
388 {
389 V T1A, T1C, T1D, T1x, T1G, T1t, T1w, T1F, T1E, T1M;
390 T1A = VMUL(LDK(KP559016994), VSUB(T1y, T1z));
391 T1C = VADD(T1y, T1z);
392 T1D = VFNMS(LDK(KP250000000), T1C, T1B);
393 T1t = VSUB(T1r, T1s);
394 T1w = VSUB(T1u, T1v);
395 T1x = VBYI(VFMA(LDK(KP951056516), T1t, VMUL(LDK(KP587785252), T1w)));
396 T1G = VBYI(VFNMS(LDK(KP587785252), T1t, VMUL(LDK(KP951056516), T1w)));
397 T1J = VADD(T1B, T1C);
398 STM2(&(xo[0]), T1J, ovs, &(xo[0]));
399 T1F = VSUB(T1D, T1A);
400 T1K = VSUB(T1F, T1G);
401 STM2(&(xo[16]), T1K, ovs, &(xo[0]));
402 T1L = VADD(T1G, T1F);
403 STM2(&(xo[24]), T1L, ovs, &(xo[0]));
404 T1E = VADD(T1A, T1D);
405 T1M = VADD(T1x, T1E);
406 STM2(&(xo[8]), T1M, ovs, &(xo[0]));
407 STN2(&(xo[8]), T1M, T1H, ovs);
408 T1N = VSUB(T1E, T1x);
409 STM2(&(xo[32]), T1N, ovs, &(xo[0]));
410 }
411 {
412 V T1O, T1P, T1R, T1S;
413 {
414 V T1n, T1l, T1m, T1f, T1q, T17, T1e, T1p, T1Q, T1o;
415 T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k));
416 T1l = VADD(T1j, T1k);
417 T1m = VFNMS(LDK(KP250000000), T1l, T1i);
418 T17 = VSUB(T13, T16);
419 T1e = VSUB(T1a, T1d);
420 T1f = VBYI(VFNMS(LDK(KP587785252), T1e, VMUL(LDK(KP951056516), T17)));
421 T1q = VBYI(VFMA(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17)));
422 T1O = VADD(T1i, T1l);
423 STM2(&(xo[20]), T1O, ovs, &(xo[0]));
424 T1p = VADD(T1n, T1m);
425 T1P = VSUB(T1p, T1q);
426 STM2(&(xo[12]), T1P, ovs, &(xo[0]));
427 T1Q = VADD(T1q, T1p);
428 STM2(&(xo[28]), T1Q, ovs, &(xo[0]));
429 STN2(&(xo[28]), T1Q, T1I, ovs);
430 T1o = VSUB(T1m, T1n);
431 T1R = VADD(T1f, T1o);
432 STM2(&(xo[4]), T1R, ovs, &(xo[0]));
433 T1S = VSUB(T1o, T1f);
434 STM2(&(xo[36]), T1S, ovs, &(xo[0]));
435 }
436 {
437 V TI, TP, TX, TU, TM, TW, TF, TT, TK, TE;
438 TI = VFMA(LDK(KP951056516), TG, VMUL(LDK(KP587785252), TH));
439 TP = VFMA(LDK(KP951056516), TN, VMUL(LDK(KP587785252), TO));
440 TX = VFNMS(LDK(KP587785252), TN, VMUL(LDK(KP951056516), TO));
441 TU = VFNMS(LDK(KP587785252), TG, VMUL(LDK(KP951056516), TH));
442 TK = VFMS(LDK(KP250000000), TB, Tm);
443 TM = VADD(TK, TL);
444 TW = VSUB(TL, TK);
445 TE = VFNMS(LDK(KP250000000), Ti, T3);
446 TF = VADD(TD, TE);
447 TT = VSUB(TE, TD);
448 {
449 V TJ, TQ, T1T, T1U;
450 TJ = VADD(TF, TI);
451 TQ = VBYI(VSUB(TM, TP));
452 T1T = VSUB(TJ, TQ);
453 STM2(&(xo[38]), T1T, ovs, &(xo[2]));
454 STN2(&(xo[36]), T1S, T1T, ovs);
455 T1U = VADD(TJ, TQ);
456 STM2(&(xo[2]), T1U, ovs, &(xo[2]));
457 STN2(&(xo[0]), T1J, T1U, ovs);
458 }
459 {
460 V TZ, T10, T1V, T1W;
461 TZ = VADD(TT, TU);
462 T10 = VBYI(VADD(TX, TW));
463 T1V = VSUB(TZ, T10);
464 STM2(&(xo[26]), T1V, ovs, &(xo[2]));
465 STN2(&(xo[24]), T1L, T1V, ovs);
466 T1W = VADD(TZ, T10);
467 STM2(&(xo[14]), T1W, ovs, &(xo[2]));
468 STN2(&(xo[12]), T1P, T1W, ovs);
469 }
470 {
471 V TR, TS, T1X, T1Y;
472 TR = VSUB(TF, TI);
473 TS = VBYI(VADD(TP, TM));
474 T1X = VSUB(TR, TS);
475 STM2(&(xo[22]), T1X, ovs, &(xo[2]));
476 STN2(&(xo[20]), T1O, T1X, ovs);
477 T1Y = VADD(TR, TS);
478 STM2(&(xo[18]), T1Y, ovs, &(xo[2]));
479 STN2(&(xo[16]), T1K, T1Y, ovs);
480 }
481 {
482 V TV, TY, T1Z, T20;
483 TV = VSUB(TT, TU);
484 TY = VBYI(VSUB(TW, TX));
485 T1Z = VSUB(TV, TY);
486 STM2(&(xo[34]), T1Z, ovs, &(xo[2]));
487 STN2(&(xo[32]), T1N, T1Z, ovs);
488 T20 = VADD(TV, TY);
489 STM2(&(xo[6]), T20, ovs, &(xo[2]));
490 STN2(&(xo[4]), T1R, T20, ovs);
491 }
492 }
493 }
494 }
495 }
496 }
497 VLEAVE();
498 }
499
500 static const kdft_desc desc = { 20, XSIMD_STRING("n2fv_20"), {92, 12, 12, 0}, &GENUS, 0, 2, 0, 0 };
501
502 void XSIMD(codelet_n2fv_20) (planner *p) {
503 X(kdft_register) (p, n2fv_20, &desc);
504 }
505
506 #endif