Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2bv_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 07:37:46 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include n2b.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 * 79 stack variables, 4 constants, and 50 memory accesses
34 */
35 #include "n2b.h"
36
37 static void n2bv_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 = ii;
48 xo = io;
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 T1H, T1I, TS, TA, TN, TV, T1M, T1N, T1O, T1P, T1R, T1S, TK, TU, TR;
51 V Tl;
52 {
53 V T3, TE, T1r, T13, Ta, TL, Tz, TG, Ts, TF, Th, TM, T1u, T1C, T1n;
54 V T1a, T1m, T1h, T1x, T1D, Tk, Ti;
55 {
56 V T1, T2, TC, TD;
57 T1 = LD(&(xi[0]), ivs, &(xi[0]));
58 T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
59 TC = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
60 TD = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
61 {
62 V T14, T6, T1c, Tv, Tm, T1f, Ty, T17, T9, Tn, Tp, T1b, Td, Tq, Te;
63 V Tf, T15, To;
64 {
65 V Tw, Tx, T7, T8, Tb, Tc;
66 {
67 V T4, T5, Tt, Tu, T11, T12;
68 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
69 T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
70 Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
71 Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
73 T3 = VSUB(T1, T2);
74 T11 = VADD(T1, T2);
75 TE = VSUB(TC, TD);
76 T12 = VADD(TC, TD);
77 T14 = VADD(T4, T5);
78 T6 = VSUB(T4, T5);
79 T1c = VADD(Tt, Tu);
80 Tv = VSUB(Tt, Tu);
81 Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
82 T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
83 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
84 T1r = VADD(T11, T12);
85 T13 = VSUB(T11, T12);
86 }
87 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
88 Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
89 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
90 T1f = VADD(Tw, Tx);
91 Ty = VSUB(Tw, Tx);
92 T17 = VADD(T7, T8);
93 T9 = VSUB(T7, T8);
94 Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
95 Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
96 T1b = VADD(Tb, Tc);
97 Td = VSUB(Tb, Tc);
98 Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
99 Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
100 Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
101 }
102 Ta = VADD(T6, T9);
103 TL = VSUB(T6, T9);
104 T15 = VADD(Tm, Tn);
105 To = VSUB(Tm, Tn);
106 Tz = VSUB(Tv, Ty);
107 TG = VADD(Tv, Ty);
108 {
109 V T1d, T1v, T18, Tr, T1e, Tg, T16, T1s;
110 T1d = VSUB(T1b, T1c);
111 T1v = VADD(T1b, T1c);
112 T18 = VADD(Tp, Tq);
113 Tr = VSUB(Tp, Tq);
114 T1e = VADD(Te, Tf);
115 Tg = VSUB(Te, Tf);
116 T16 = VSUB(T14, T15);
117 T1s = VADD(T14, T15);
118 {
119 V T1t, T19, T1w, T1g;
120 T1t = VADD(T17, T18);
121 T19 = VSUB(T17, T18);
122 Ts = VSUB(To, Tr);
123 TF = VADD(To, Tr);
124 T1w = VADD(T1e, T1f);
125 T1g = VSUB(T1e, T1f);
126 Th = VADD(Td, Tg);
127 TM = VSUB(Td, Tg);
128 T1u = VADD(T1s, T1t);
129 T1C = VSUB(T1s, T1t);
130 T1n = VSUB(T16, T19);
131 T1a = VADD(T16, T19);
132 T1m = VSUB(T1d, T1g);
133 T1h = VADD(T1d, T1g);
134 T1x = VADD(T1v, T1w);
135 T1D = VSUB(T1v, T1w);
136 }
137 }
138 }
139 }
140 Tk = VSUB(Ta, Th);
141 Ti = VADD(Ta, Th);
142 {
143 V TJ, T1k, T1A, TZ, Tj, T1E, T1G, TI, T10, T1j, T1z, T1i, T1y, TH;
144 TJ = VSUB(TF, TG);
145 TH = VADD(TF, TG);
146 T1i = VADD(T1a, T1h);
147 T1k = VSUB(T1a, T1h);
148 T1y = VADD(T1u, T1x);
149 T1A = VSUB(T1u, T1x);
150 TZ = VADD(T3, Ti);
151 Tj = VFNMS(LDK(KP250000000), Ti, T3);
152 T1E = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1D, T1C));
153 T1G = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1C, T1D));
154 TI = VFNMS(LDK(KP250000000), TH, TE);
155 T10 = VADD(TE, TH);
156 T1j = VFNMS(LDK(KP250000000), T1i, T13);
157 T1H = VADD(T1r, T1y);
158 STM2(&(xo[0]), T1H, ovs, &(xo[0]));
159 T1z = VFNMS(LDK(KP250000000), T1y, T1r);
160 T1I = VADD(T13, T1i);
161 STM2(&(xo[20]), T1I, ovs, &(xo[0]));
162 {
163 V T1J, T1K, T1p, T1l, T1o, T1q, T1F, T1B, T1L, T1Q;
164 TS = VFNMS(LDK(KP618033988), Ts, Tz);
165 TA = VFMA(LDK(KP618033988), Tz, Ts);
166 TN = VFMA(LDK(KP618033988), TM, TL);
167 TV = VFNMS(LDK(KP618033988), TL, TM);
168 T1J = VFMAI(T10, TZ);
169 STM2(&(xo[10]), T1J, ovs, &(xo[2]));
170 T1K = VFNMSI(T10, TZ);
171 STM2(&(xo[30]), T1K, ovs, &(xo[2]));
172 T1p = VFMA(LDK(KP559016994), T1k, T1j);
173 T1l = VFNMS(LDK(KP559016994), T1k, T1j);
174 T1o = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1n, T1m));
175 T1q = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1m, T1n));
176 T1F = VFNMS(LDK(KP559016994), T1A, T1z);
177 T1B = VFMA(LDK(KP559016994), T1A, T1z);
178 T1L = VFNMSI(T1q, T1p);
179 STM2(&(xo[28]), T1L, ovs, &(xo[0]));
180 STN2(&(xo[28]), T1L, T1K, ovs);
181 T1M = VFMAI(T1q, T1p);
182 STM2(&(xo[12]), T1M, ovs, &(xo[0]));
183 T1N = VFMAI(T1o, T1l);
184 STM2(&(xo[36]), T1N, ovs, &(xo[0]));
185 T1O = VFNMSI(T1o, T1l);
186 STM2(&(xo[4]), T1O, ovs, &(xo[0]));
187 T1P = VFMAI(T1E, T1B);
188 STM2(&(xo[32]), T1P, ovs, &(xo[0]));
189 T1Q = VFNMSI(T1E, T1B);
190 STM2(&(xo[8]), T1Q, ovs, &(xo[0]));
191 STN2(&(xo[8]), T1Q, T1J, ovs);
192 T1R = VFNMSI(T1G, T1F);
193 STM2(&(xo[24]), T1R, ovs, &(xo[0]));
194 T1S = VFMAI(T1G, T1F);
195 STM2(&(xo[16]), T1S, ovs, &(xo[0]));
196 TK = VFMA(LDK(KP559016994), TJ, TI);
197 TU = VFNMS(LDK(KP559016994), TJ, TI);
198 TR = VFNMS(LDK(KP559016994), Tk, Tj);
199 Tl = VFMA(LDK(KP559016994), Tk, Tj);
200 }
201 }
202 }
203 {
204 V TY, TW, TO, TQ, TB, TP, TX, TT;
205 TY = VFMA(LDK(KP951056516), TV, TU);
206 TW = VFNMS(LDK(KP951056516), TV, TU);
207 TO = VFMA(LDK(KP951056516), TN, TK);
208 TQ = VFNMS(LDK(KP951056516), TN, TK);
209 TB = VFNMS(LDK(KP951056516), TA, Tl);
210 TP = VFMA(LDK(KP951056516), TA, Tl);
211 TX = VFNMS(LDK(KP951056516), TS, TR);
212 TT = VFMA(LDK(KP951056516), TS, TR);
213 {
214 V T1T, T1U, T1V, T1W;
215 T1T = VFMAI(TQ, TP);
216 STM2(&(xo[18]), T1T, ovs, &(xo[2]));
217 STN2(&(xo[16]), T1S, T1T, ovs);
218 T1U = VFNMSI(TQ, TP);
219 STM2(&(xo[22]), T1U, ovs, &(xo[2]));
220 STN2(&(xo[20]), T1I, T1U, ovs);
221 T1V = VFMAI(TO, TB);
222 STM2(&(xo[2]), T1V, ovs, &(xo[2]));
223 STN2(&(xo[0]), T1H, T1V, ovs);
224 T1W = VFNMSI(TO, TB);
225 STM2(&(xo[38]), T1W, ovs, &(xo[2]));
226 STN2(&(xo[36]), T1N, T1W, ovs);
227 {
228 V T1X, T1Y, T1Z, T20;
229 T1X = VFMAI(TW, TT);
230 STM2(&(xo[34]), T1X, ovs, &(xo[2]));
231 STN2(&(xo[32]), T1P, T1X, ovs);
232 T1Y = VFNMSI(TW, TT);
233 STM2(&(xo[6]), T1Y, ovs, &(xo[2]));
234 STN2(&(xo[4]), T1O, T1Y, ovs);
235 T1Z = VFMAI(TY, TX);
236 STM2(&(xo[26]), T1Z, ovs, &(xo[2]));
237 STN2(&(xo[24]), T1R, T1Z, ovs);
238 T20 = VFNMSI(TY, TX);
239 STM2(&(xo[14]), T20, ovs, &(xo[2]));
240 STN2(&(xo[12]), T1M, T20, ovs);
241 }
242 }
243 }
244 }
245 }
246 VLEAVE();
247 }
248
249 static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), {58, 4, 46, 0}, &GENUS, 0, 2, 0, 0 };
250
251 void XSIMD(codelet_n2bv_20) (planner *p) {
252 X(kdft_register) (p, n2bv_20, &desc);
253 }
254
255 #else /* HAVE_FMA */
256
257 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 20 -name n2bv_20 -with-ostride 2 -include n2b.h -store-multiple 2 */
258
259 /*
260 * This function contains 104 FP additions, 24 FP multiplications,
261 * (or, 92 additions, 12 multiplications, 12 fused multiply/add),
262 * 57 stack variables, 4 constants, and 50 memory accesses
263 */
264 #include "n2b.h"
265
266 static void n2bv_20(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
267 {
268 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
269 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
270 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
271 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
272 {
273 INT i;
274 const R *xi;
275 R *xo;
276 xi = ii;
277 xo = io;
278 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)) {
279 V T3, T1y, TH, T1i, Ts, TL, TM, Tz, T13, T16, T1j, T1u, T1v, T1w, T1r;
280 V T1s, T1t, T1a, T1d, T1k, Ti, Tk, TE, TI;
281 {
282 V T1, T2, T1g, TF, TG, T1h;
283 T1 = LD(&(xi[0]), ivs, &(xi[0]));
284 T2 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
285 T1g = VADD(T1, T2);
286 TF = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
287 TG = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
288 T1h = VADD(TF, TG);
289 T3 = VSUB(T1, T2);
290 T1y = VADD(T1g, T1h);
291 TH = VSUB(TF, TG);
292 T1i = VSUB(T1g, T1h);
293 }
294 {
295 V T6, T11, Tv, T19, Ty, T1c, T9, T14, Td, T18, To, T12, Tr, T15, Tg;
296 V T1b;
297 {
298 V T4, T5, Tt, Tu;
299 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
300 T5 = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
301 T6 = VSUB(T4, T5);
302 T11 = VADD(T4, T5);
303 Tt = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
304 Tu = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
305 Tv = VSUB(Tt, Tu);
306 T19 = VADD(Tt, Tu);
307 }
308 {
309 V Tw, Tx, T7, T8;
310 Tw = LD(&(xi[WS(is, 17)]), ivs, &(xi[WS(is, 1)]));
311 Tx = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
312 Ty = VSUB(Tw, Tx);
313 T1c = VADD(Tw, Tx);
314 T7 = LD(&(xi[WS(is, 16)]), ivs, &(xi[0]));
315 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
316 T9 = VSUB(T7, T8);
317 T14 = VADD(T7, T8);
318 }
319 {
320 V Tb, Tc, Tm, Tn;
321 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
322 Tc = LD(&(xi[WS(is, 18)]), ivs, &(xi[0]));
323 Td = VSUB(Tb, Tc);
324 T18 = VADD(Tb, Tc);
325 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
326 Tn = LD(&(xi[WS(is, 19)]), ivs, &(xi[WS(is, 1)]));
327 To = VSUB(Tm, Tn);
328 T12 = VADD(Tm, Tn);
329 }
330 {
331 V Tp, Tq, Te, Tf;
332 Tp = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
333 Tq = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
334 Tr = VSUB(Tp, Tq);
335 T15 = VADD(Tp, Tq);
336 Te = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
337 Tf = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
338 Tg = VSUB(Te, Tf);
339 T1b = VADD(Te, Tf);
340 }
341 Ts = VSUB(To, Tr);
342 TL = VSUB(T6, T9);
343 TM = VSUB(Td, Tg);
344 Tz = VSUB(Tv, Ty);
345 T13 = VSUB(T11, T12);
346 T16 = VSUB(T14, T15);
347 T1j = VADD(T13, T16);
348 T1u = VADD(T18, T19);
349 T1v = VADD(T1b, T1c);
350 T1w = VADD(T1u, T1v);
351 T1r = VADD(T11, T12);
352 T1s = VADD(T14, T15);
353 T1t = VADD(T1r, T1s);
354 T1a = VSUB(T18, T19);
355 T1d = VSUB(T1b, T1c);
356 T1k = VADD(T1a, T1d);
357 {
358 V Ta, Th, TC, TD;
359 Ta = VADD(T6, T9);
360 Th = VADD(Td, Tg);
361 Ti = VADD(Ta, Th);
362 Tk = VMUL(LDK(KP559016994), VSUB(Ta, Th));
363 TC = VADD(To, Tr);
364 TD = VADD(Tv, Ty);
365 TE = VMUL(LDK(KP559016994), VSUB(TC, TD));
366 TI = VADD(TC, TD);
367 }
368 }
369 {
370 V T1H, T1J, T1K, T1L, T1N, T1I, TZ, T10;
371 TZ = VADD(T3, Ti);
372 T10 = VBYI(VADD(TH, TI));
373 T1H = VSUB(TZ, T10);
374 STM2(&(xo[30]), T1H, ovs, &(xo[2]));
375 T1I = VADD(TZ, T10);
376 STM2(&(xo[10]), T1I, ovs, &(xo[2]));
377 {
378 V T1x, T1z, T1A, T1E, T1G, T1C, T1D, T1F, T1B, T1M;
379 T1x = VMUL(LDK(KP559016994), VSUB(T1t, T1w));
380 T1z = VADD(T1t, T1w);
381 T1A = VFNMS(LDK(KP250000000), T1z, T1y);
382 T1C = VSUB(T1r, T1s);
383 T1D = VSUB(T1u, T1v);
384 T1E = VBYI(VFMA(LDK(KP951056516), T1C, VMUL(LDK(KP587785252), T1D)));
385 T1G = VBYI(VFNMS(LDK(KP951056516), T1D, VMUL(LDK(KP587785252), T1C)));
386 T1J = VADD(T1y, T1z);
387 STM2(&(xo[0]), T1J, ovs, &(xo[0]));
388 T1F = VSUB(T1A, T1x);
389 T1K = VSUB(T1F, T1G);
390 STM2(&(xo[16]), T1K, ovs, &(xo[0]));
391 T1L = VADD(T1G, T1F);
392 STM2(&(xo[24]), T1L, ovs, &(xo[0]));
393 T1B = VADD(T1x, T1A);
394 T1M = VSUB(T1B, T1E);
395 STM2(&(xo[8]), T1M, ovs, &(xo[0]));
396 STN2(&(xo[8]), T1M, T1I, ovs);
397 T1N = VADD(T1E, T1B);
398 STM2(&(xo[32]), T1N, ovs, &(xo[0]));
399 }
400 {
401 V T1O, T1P, T1R, T1S;
402 {
403 V T1n, T1l, T1m, T1f, T1p, T17, T1e, T1q, T1Q, T1o;
404 T1n = VMUL(LDK(KP559016994), VSUB(T1j, T1k));
405 T1l = VADD(T1j, T1k);
406 T1m = VFNMS(LDK(KP250000000), T1l, T1i);
407 T17 = VSUB(T13, T16);
408 T1e = VSUB(T1a, T1d);
409 T1f = VBYI(VFNMS(LDK(KP951056516), T1e, VMUL(LDK(KP587785252), T17)));
410 T1p = VBYI(VFMA(LDK(KP951056516), T17, VMUL(LDK(KP587785252), T1e)));
411 T1O = VADD(T1i, T1l);
412 STM2(&(xo[20]), T1O, ovs, &(xo[0]));
413 T1q = VADD(T1n, T1m);
414 T1P = VADD(T1p, T1q);
415 STM2(&(xo[12]), T1P, ovs, &(xo[0]));
416 T1Q = VSUB(T1q, T1p);
417 STM2(&(xo[28]), T1Q, ovs, &(xo[0]));
418 STN2(&(xo[28]), T1Q, T1H, ovs);
419 T1o = VSUB(T1m, T1n);
420 T1R = VADD(T1f, T1o);
421 STM2(&(xo[4]), T1R, ovs, &(xo[0]));
422 T1S = VSUB(T1o, T1f);
423 STM2(&(xo[36]), T1S, ovs, &(xo[0]));
424 }
425 {
426 V TA, TN, TU, TS, TK, TV, Tl, TR, TJ, Tj;
427 TA = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts));
428 TN = VFNMS(LDK(KP951056516), TM, VMUL(LDK(KP587785252), TL));
429 TU = VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TM));
430 TS = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz));
431 TJ = VFNMS(LDK(KP250000000), TI, TH);
432 TK = VSUB(TE, TJ);
433 TV = VADD(TE, TJ);
434 Tj = VFNMS(LDK(KP250000000), Ti, T3);
435 Tl = VSUB(Tj, Tk);
436 TR = VADD(Tk, Tj);
437 {
438 V TB, TO, T1T, T1U;
439 TB = VSUB(Tl, TA);
440 TO = VBYI(VSUB(TK, TN));
441 T1T = VSUB(TB, TO);
442 STM2(&(xo[34]), T1T, ovs, &(xo[2]));
443 STN2(&(xo[32]), T1N, T1T, ovs);
444 T1U = VADD(TB, TO);
445 STM2(&(xo[6]), T1U, ovs, &(xo[2]));
446 STN2(&(xo[4]), T1R, T1U, ovs);
447 }
448 {
449 V TX, TY, T1V, T1W;
450 TX = VADD(TR, TS);
451 TY = VBYI(VSUB(TV, TU));
452 T1V = VSUB(TX, TY);
453 STM2(&(xo[22]), T1V, ovs, &(xo[2]));
454 STN2(&(xo[20]), T1O, T1V, ovs);
455 T1W = VADD(TX, TY);
456 STM2(&(xo[18]), T1W, ovs, &(xo[2]));
457 STN2(&(xo[16]), T1K, T1W, ovs);
458 }
459 {
460 V TP, TQ, T1X, T1Y;
461 TP = VADD(Tl, TA);
462 TQ = VBYI(VADD(TN, TK));
463 T1X = VSUB(TP, TQ);
464 STM2(&(xo[26]), T1X, ovs, &(xo[2]));
465 STN2(&(xo[24]), T1L, T1X, ovs);
466 T1Y = VADD(TP, TQ);
467 STM2(&(xo[14]), T1Y, ovs, &(xo[2]));
468 STN2(&(xo[12]), T1P, T1Y, ovs);
469 }
470 {
471 V TT, TW, T1Z, T20;
472 TT = VSUB(TR, TS);
473 TW = VBYI(VADD(TU, TV));
474 T1Z = VSUB(TT, TW);
475 STM2(&(xo[38]), T1Z, ovs, &(xo[2]));
476 STN2(&(xo[36]), T1S, T1Z, ovs);
477 T20 = VADD(TT, TW);
478 STM2(&(xo[2]), T20, ovs, &(xo[2]));
479 STN2(&(xo[0]), T1J, T20, ovs);
480 }
481 }
482 }
483 }
484 }
485 }
486 VLEAVE();
487 }
488
489 static const kdft_desc desc = { 20, XSIMD_STRING("n2bv_20"), {92, 12, 12, 0}, &GENUS, 0, 2, 0, 0 };
490
491 void XSIMD(codelet_n2bv_20) (planner *p) {
492 X(kdft_register) (p, n2bv_20, &desc);
493 }
494
495 #endif /* HAVE_FMA */