Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2sv_16.c @ 82:d0c2a83c1364

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
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
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 08:05:19 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.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n2sv_16 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
29
30 /*
31 * This function contains 144 FP additions, 40 FP multiplications,
32 * (or, 104 additions, 0 multiplications, 40 fused multiply/add),
33 * 74 stack variables, 3 constants, and 72 memory accesses
34 */
35 #include "dft/simd/n2s.h"
36
37 static void n2sv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT i;
44 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
45 V T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
46 V T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
47 V T1U, T1A;
48 {
49 V T3, TL, Ty, T1k, T6, T1j, TB, TM;
50 {
51 V T1, T2, Tw, Tx;
52 T1 = LD(&(ri[0]), ivs, &(ri[0]));
53 T2 = LD(&(ri[WS(is, 8)]), ivs, &(ri[0]));
54 T3 = VADD(T1, T2);
55 TL = VSUB(T1, T2);
56 Tw = LD(&(ii[0]), ivs, &(ii[0]));
57 Tx = LD(&(ii[WS(is, 8)]), ivs, &(ii[0]));
58 Ty = VADD(Tw, Tx);
59 T1k = VSUB(Tw, Tx);
60 }
61 {
62 V T4, T5, Tz, TA;
63 T4 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
64 T5 = LD(&(ri[WS(is, 12)]), ivs, &(ri[0]));
65 T6 = VADD(T4, T5);
66 T1j = VSUB(T4, T5);
67 Tz = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
68 TA = LD(&(ii[WS(is, 12)]), ivs, &(ii[0]));
69 TB = VADD(Tz, TA);
70 TM = VSUB(Tz, TA);
71 }
72 T7 = VADD(T3, T6);
73 T1R = VSUB(T3, T6);
74 T25 = VSUB(Ty, TB);
75 TC = VADD(Ty, TB);
76 TN = VSUB(TL, TM);
77 T1x = VADD(TL, TM);
78 T1H = VSUB(T1k, T1j);
79 T1l = VADD(T1j, T1k);
80 }
81 {
82 V Tp, T1c, T1a, T20, Ts, T17, T1f, T21;
83 {
84 V Tn, To, T18, T19;
85 Tn = LD(&(ri[WS(is, 15)]), ivs, &(ri[WS(is, 1)]));
86 To = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
87 Tp = VADD(Tn, To);
88 T1c = VSUB(Tn, To);
89 T18 = LD(&(ii[WS(is, 15)]), ivs, &(ii[WS(is, 1)]));
90 T19 = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
91 T1a = VSUB(T18, T19);
92 T20 = VADD(T18, T19);
93 }
94 {
95 V Tq, Tr, T1d, T1e;
96 Tq = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
97 Tr = LD(&(ri[WS(is, 11)]), ivs, &(ri[WS(is, 1)]));
98 Ts = VADD(Tq, Tr);
99 T17 = VSUB(Tq, Tr);
100 T1d = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
101 T1e = LD(&(ii[WS(is, 11)]), ivs, &(ii[WS(is, 1)]));
102 T1f = VSUB(T1d, T1e);
103 T21 = VADD(T1d, T1e);
104 }
105 Tt = VADD(Tp, Ts);
106 T22 = VSUB(T20, T21);
107 T2h = VADD(T20, T21);
108 T1b = VADD(T17, T1a);
109 T1g = VSUB(T1c, T1f);
110 T1E = VSUB(T1a, T17);
111 T1Z = VSUB(Tp, Ts);
112 T1D = VADD(T1c, T1f);
113 }
114 {
115 V Ta, TP, TF, TO, Td, TR, TI, TS;
116 {
117 V T8, T9, TD, TE;
118 T8 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
119 T9 = LD(&(ri[WS(is, 10)]), ivs, &(ri[0]));
120 Ta = VADD(T8, T9);
121 TP = VSUB(T8, T9);
122 TD = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
123 TE = LD(&(ii[WS(is, 10)]), ivs, &(ii[0]));
124 TF = VADD(TD, TE);
125 TO = VSUB(TD, TE);
126 }
127 {
128 V Tb, Tc, TG, TH;
129 Tb = LD(&(ri[WS(is, 14)]), ivs, &(ri[0]));
130 Tc = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
131 Td = VADD(Tb, Tc);
132 TR = VSUB(Tb, Tc);
133 TG = LD(&(ii[WS(is, 14)]), ivs, &(ii[0]));
134 TH = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
135 TI = VADD(TG, TH);
136 TS = VSUB(TG, TH);
137 }
138 Te = VADD(Ta, Td);
139 T1S = VSUB(TF, TI);
140 T26 = VSUB(Td, Ta);
141 TJ = VADD(TF, TI);
142 TQ = VSUB(TO, TP);
143 T1m = VSUB(TR, TS);
144 T1n = VADD(TP, TO);
145 TT = VADD(TR, TS);
146 }
147 {
148 V Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
149 {
150 V Tg, Th, TX, TY;
151 Tg = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
152 Th = LD(&(ri[WS(is, 9)]), ivs, &(ri[WS(is, 1)]));
153 Ti = VADD(Tg, Th);
154 T11 = VSUB(Tg, Th);
155 TX = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
156 TY = LD(&(ii[WS(is, 9)]), ivs, &(ii[WS(is, 1)]));
157 TZ = VSUB(TX, TY);
158 T1V = VADD(TX, TY);
159 }
160 {
161 V Tj, Tk, T12, T13;
162 Tj = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
163 Tk = LD(&(ri[WS(is, 13)]), ivs, &(ri[WS(is, 1)]));
164 Tl = VADD(Tj, Tk);
165 TW = VSUB(Tj, Tk);
166 T12 = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
167 T13 = LD(&(ii[WS(is, 13)]), ivs, &(ii[WS(is, 1)]));
168 T14 = VSUB(T12, T13);
169 T1W = VADD(T12, T13);
170 }
171 Tm = VADD(Ti, Tl);
172 T1X = VSUB(T1V, T1W);
173 T2g = VADD(T1V, T1W);
174 T10 = VADD(TW, TZ);
175 T15 = VSUB(T11, T14);
176 T1B = VSUB(TZ, TW);
177 T1U = VSUB(Ti, Tl);
178 T1A = VADD(T11, T14);
179 }
180 {
181 V T2l, T2m, T2n, T2o, T2p, T2q, T2r, T2s;
182 {
183 V Tf, Tu, T2j, T2k;
184 Tf = VADD(T7, Te);
185 Tu = VADD(Tm, Tt);
186 T2l = VSUB(Tf, Tu);
187 STM4(&(ro[8]), T2l, ovs, &(ro[0]));
188 T2m = VADD(Tf, Tu);
189 STM4(&(ro[0]), T2m, ovs, &(ro[0]));
190 T2j = VADD(TC, TJ);
191 T2k = VADD(T2g, T2h);
192 T2n = VSUB(T2j, T2k);
193 STM4(&(io[8]), T2n, ovs, &(io[0]));
194 T2o = VADD(T2j, T2k);
195 STM4(&(io[0]), T2o, ovs, &(io[0]));
196 }
197 {
198 V Tv, TK, T2f, T2i;
199 Tv = VSUB(Tt, Tm);
200 TK = VSUB(TC, TJ);
201 T2p = VADD(Tv, TK);
202 STM4(&(io[4]), T2p, ovs, &(io[0]));
203 T2q = VSUB(TK, Tv);
204 STM4(&(io[12]), T2q, ovs, &(io[0]));
205 T2f = VSUB(T7, Te);
206 T2i = VSUB(T2g, T2h);
207 T2r = VSUB(T2f, T2i);
208 STM4(&(ro[12]), T2r, ovs, &(ro[0]));
209 T2s = VADD(T2f, T2i);
210 STM4(&(ro[4]), T2s, ovs, &(ro[0]));
211 }
212 {
213 V T2t, T2u, T2v, T2w, T2x, T2y, T2z, T2A;
214 {
215 V T1T, T27, T24, T28, T1Y, T23;
216 T1T = VADD(T1R, T1S);
217 T27 = VSUB(T25, T26);
218 T1Y = VADD(T1U, T1X);
219 T23 = VSUB(T1Z, T22);
220 T24 = VADD(T1Y, T23);
221 T28 = VSUB(T23, T1Y);
222 T2t = VFNMS(LDK(KP707106781), T24, T1T);
223 STM4(&(ro[10]), T2t, ovs, &(ro[0]));
224 T2u = VFMA(LDK(KP707106781), T28, T27);
225 STM4(&(io[6]), T2u, ovs, &(io[0]));
226 T2v = VFMA(LDK(KP707106781), T24, T1T);
227 STM4(&(ro[2]), T2v, ovs, &(ro[0]));
228 T2w = VFNMS(LDK(KP707106781), T28, T27);
229 STM4(&(io[14]), T2w, ovs, &(io[0]));
230 }
231 {
232 V T29, T2d, T2c, T2e, T2a, T2b;
233 T29 = VSUB(T1R, T1S);
234 T2d = VADD(T26, T25);
235 T2a = VSUB(T1X, T1U);
236 T2b = VADD(T1Z, T22);
237 T2c = VSUB(T2a, T2b);
238 T2e = VADD(T2a, T2b);
239 T2x = VFNMS(LDK(KP707106781), T2c, T29);
240 STM4(&(ro[14]), T2x, ovs, &(ro[0]));
241 T2y = VFMA(LDK(KP707106781), T2e, T2d);
242 STM4(&(io[2]), T2y, ovs, &(io[0]));
243 T2z = VFMA(LDK(KP707106781), T2c, T29);
244 STM4(&(ro[6]), T2z, ovs, &(ro[0]));
245 T2A = VFNMS(LDK(KP707106781), T2e, T2d);
246 STM4(&(io[10]), T2A, ovs, &(io[0]));
247 }
248 {
249 V T2B, T2C, T2D, T2E, T2F, T2G, T2H, T2I;
250 {
251 V TV, T1v, T1p, T1r, T1i, T1q, T1u, T1w, TU, T1o;
252 TU = VSUB(TQ, TT);
253 TV = VFMA(LDK(KP707106781), TU, TN);
254 T1v = VFNMS(LDK(KP707106781), TU, TN);
255 T1o = VSUB(T1m, T1n);
256 T1p = VFNMS(LDK(KP707106781), T1o, T1l);
257 T1r = VFMA(LDK(KP707106781), T1o, T1l);
258 {
259 V T16, T1h, T1s, T1t;
260 T16 = VFMA(LDK(KP414213562), T15, T10);
261 T1h = VFNMS(LDK(KP414213562), T1g, T1b);
262 T1i = VSUB(T16, T1h);
263 T1q = VADD(T16, T1h);
264 T1s = VFMA(LDK(KP414213562), T1b, T1g);
265 T1t = VFNMS(LDK(KP414213562), T10, T15);
266 T1u = VSUB(T1s, T1t);
267 T1w = VADD(T1t, T1s);
268 }
269 T2B = VFNMS(LDK(KP923879532), T1i, TV);
270 STM4(&(ro[11]), T2B, ovs, &(ro[1]));
271 T2C = VFNMS(LDK(KP923879532), T1u, T1r);
272 STM4(&(io[11]), T2C, ovs, &(io[1]));
273 T2D = VFMA(LDK(KP923879532), T1i, TV);
274 STM4(&(ro[3]), T2D, ovs, &(ro[1]));
275 T2E = VFMA(LDK(KP923879532), T1u, T1r);
276 STM4(&(io[3]), T2E, ovs, &(io[1]));
277 T2F = VFNMS(LDK(KP923879532), T1q, T1p);
278 STM4(&(io[7]), T2F, ovs, &(io[1]));
279 T2G = VFNMS(LDK(KP923879532), T1w, T1v);
280 STM4(&(ro[7]), T2G, ovs, &(ro[1]));
281 T2H = VFMA(LDK(KP923879532), T1q, T1p);
282 STM4(&(io[15]), T2H, ovs, &(io[1]));
283 T2I = VFMA(LDK(KP923879532), T1w, T1v);
284 STM4(&(ro[15]), T2I, ovs, &(ro[1]));
285 }
286 {
287 V T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
288 T1y = VADD(T1n, T1m);
289 T1z = VFMA(LDK(KP707106781), T1y, T1x);
290 T1L = VFNMS(LDK(KP707106781), T1y, T1x);
291 T1I = VADD(TQ, TT);
292 T1J = VFNMS(LDK(KP707106781), T1I, T1H);
293 T1P = VFMA(LDK(KP707106781), T1I, T1H);
294 {
295 V T1C, T1F, T1M, T1N;
296 T1C = VFMA(LDK(KP414213562), T1B, T1A);
297 T1F = VFNMS(LDK(KP414213562), T1E, T1D);
298 T1G = VADD(T1C, T1F);
299 T1K = VSUB(T1F, T1C);
300 T1M = VFNMS(LDK(KP414213562), T1A, T1B);
301 T1N = VFMA(LDK(KP414213562), T1D, T1E);
302 T1O = VSUB(T1M, T1N);
303 T1Q = VADD(T1M, T1N);
304 }
305 {
306 V T2J, T2K, T2L, T2M;
307 T2J = VFNMS(LDK(KP923879532), T1G, T1z);
308 STM4(&(ro[9]), T2J, ovs, &(ro[1]));
309 STN4(&(ro[8]), T2l, T2J, T2t, T2B, ovs);
310 T2K = VFNMS(LDK(KP923879532), T1Q, T1P);
311 STM4(&(io[9]), T2K, ovs, &(io[1]));
312 STN4(&(io[8]), T2n, T2K, T2A, T2C, ovs);
313 T2L = VFMA(LDK(KP923879532), T1G, T1z);
314 STM4(&(ro[1]), T2L, ovs, &(ro[1]));
315 STN4(&(ro[0]), T2m, T2L, T2v, T2D, ovs);
316 T2M = VFMA(LDK(KP923879532), T1Q, T1P);
317 STM4(&(io[1]), T2M, ovs, &(io[1]));
318 STN4(&(io[0]), T2o, T2M, T2y, T2E, ovs);
319 }
320 {
321 V T2N, T2O, T2P, T2Q;
322 T2N = VFNMS(LDK(KP923879532), T1K, T1J);
323 STM4(&(io[13]), T2N, ovs, &(io[1]));
324 STN4(&(io[12]), T2q, T2N, T2w, T2H, ovs);
325 T2O = VFNMS(LDK(KP923879532), T1O, T1L);
326 STM4(&(ro[13]), T2O, ovs, &(ro[1]));
327 STN4(&(ro[12]), T2r, T2O, T2x, T2I, ovs);
328 T2P = VFMA(LDK(KP923879532), T1K, T1J);
329 STM4(&(io[5]), T2P, ovs, &(io[1]));
330 STN4(&(io[4]), T2p, T2P, T2u, T2F, ovs);
331 T2Q = VFMA(LDK(KP923879532), T1O, T1L);
332 STM4(&(ro[5]), T2Q, ovs, &(ro[1]));
333 STN4(&(ro[4]), T2s, T2Q, T2z, T2G, ovs);
334 }
335 }
336 }
337 }
338 }
339 }
340 }
341 VLEAVE();
342 }
343
344 static const kdft_desc desc = { 16, XSIMD_STRING("n2sv_16"), {104, 0, 40, 0}, &GENUS, 0, 1, 0, 0 };
345
346 void XSIMD(codelet_n2sv_16) (planner *p) {
347 X(kdft_register) (p, n2sv_16, &desc);
348 }
349
350 #else
351
352 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n2sv_16 -with-ostride 1 -include dft/simd/n2s.h -store-multiple 4 */
353
354 /*
355 * This function contains 144 FP additions, 24 FP multiplications,
356 * (or, 136 additions, 16 multiplications, 8 fused multiply/add),
357 * 74 stack variables, 3 constants, and 72 memory accesses
358 */
359 #include "dft/simd/n2s.h"
360
361 static void n2sv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
362 {
363 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
364 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
365 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
366 {
367 INT i;
368 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(64, is), MAKE_VOLATILE_STRIDE(64, os)) {
369 V T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
370 V T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
371 V T1U, T1A;
372 {
373 V T3, TL, Ty, T1k, T6, T1j, TB, TM;
374 {
375 V T1, T2, Tw, Tx;
376 T1 = LD(&(ri[0]), ivs, &(ri[0]));
377 T2 = LD(&(ri[WS(is, 8)]), ivs, &(ri[0]));
378 T3 = VADD(T1, T2);
379 TL = VSUB(T1, T2);
380 Tw = LD(&(ii[0]), ivs, &(ii[0]));
381 Tx = LD(&(ii[WS(is, 8)]), ivs, &(ii[0]));
382 Ty = VADD(Tw, Tx);
383 T1k = VSUB(Tw, Tx);
384 }
385 {
386 V T4, T5, Tz, TA;
387 T4 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
388 T5 = LD(&(ri[WS(is, 12)]), ivs, &(ri[0]));
389 T6 = VADD(T4, T5);
390 T1j = VSUB(T4, T5);
391 Tz = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
392 TA = LD(&(ii[WS(is, 12)]), ivs, &(ii[0]));
393 TB = VADD(Tz, TA);
394 TM = VSUB(Tz, TA);
395 }
396 T7 = VADD(T3, T6);
397 T1R = VSUB(T3, T6);
398 T25 = VSUB(Ty, TB);
399 TC = VADD(Ty, TB);
400 TN = VSUB(TL, TM);
401 T1x = VADD(TL, TM);
402 T1H = VSUB(T1k, T1j);
403 T1l = VADD(T1j, T1k);
404 }
405 {
406 V Tp, T17, T1f, T20, Ts, T1c, T1a, T21;
407 {
408 V Tn, To, T1d, T1e;
409 Tn = LD(&(ri[WS(is, 15)]), ivs, &(ri[WS(is, 1)]));
410 To = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
411 Tp = VADD(Tn, To);
412 T17 = VSUB(Tn, To);
413 T1d = LD(&(ii[WS(is, 15)]), ivs, &(ii[WS(is, 1)]));
414 T1e = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
415 T1f = VSUB(T1d, T1e);
416 T20 = VADD(T1d, T1e);
417 }
418 {
419 V Tq, Tr, T18, T19;
420 Tq = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
421 Tr = LD(&(ri[WS(is, 11)]), ivs, &(ri[WS(is, 1)]));
422 Ts = VADD(Tq, Tr);
423 T1c = VSUB(Tq, Tr);
424 T18 = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
425 T19 = LD(&(ii[WS(is, 11)]), ivs, &(ii[WS(is, 1)]));
426 T1a = VSUB(T18, T19);
427 T21 = VADD(T18, T19);
428 }
429 Tt = VADD(Tp, Ts);
430 T22 = VSUB(T20, T21);
431 T2h = VADD(T20, T21);
432 T1b = VSUB(T17, T1a);
433 T1g = VADD(T1c, T1f);
434 T1E = VSUB(T1f, T1c);
435 T1Z = VSUB(Tp, Ts);
436 T1D = VADD(T17, T1a);
437 }
438 {
439 V Ta, TP, TF, TO, Td, TR, TI, TS;
440 {
441 V T8, T9, TD, TE;
442 T8 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
443 T9 = LD(&(ri[WS(is, 10)]), ivs, &(ri[0]));
444 Ta = VADD(T8, T9);
445 TP = VSUB(T8, T9);
446 TD = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
447 TE = LD(&(ii[WS(is, 10)]), ivs, &(ii[0]));
448 TF = VADD(TD, TE);
449 TO = VSUB(TD, TE);
450 }
451 {
452 V Tb, Tc, TG, TH;
453 Tb = LD(&(ri[WS(is, 14)]), ivs, &(ri[0]));
454 Tc = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
455 Td = VADD(Tb, Tc);
456 TR = VSUB(Tb, Tc);
457 TG = LD(&(ii[WS(is, 14)]), ivs, &(ii[0]));
458 TH = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
459 TI = VADD(TG, TH);
460 TS = VSUB(TG, TH);
461 }
462 Te = VADD(Ta, Td);
463 T1S = VSUB(TF, TI);
464 T26 = VSUB(Td, Ta);
465 TJ = VADD(TF, TI);
466 TQ = VSUB(TO, TP);
467 T1m = VSUB(TR, TS);
468 T1n = VADD(TP, TO);
469 TT = VADD(TR, TS);
470 }
471 {
472 V Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
473 {
474 V Tg, Th, TX, TY;
475 Tg = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
476 Th = LD(&(ri[WS(is, 9)]), ivs, &(ri[WS(is, 1)]));
477 Ti = VADD(Tg, Th);
478 T11 = VSUB(Tg, Th);
479 TX = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
480 TY = LD(&(ii[WS(is, 9)]), ivs, &(ii[WS(is, 1)]));
481 TZ = VSUB(TX, TY);
482 T1V = VADD(TX, TY);
483 }
484 {
485 V Tj, Tk, T12, T13;
486 Tj = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
487 Tk = LD(&(ri[WS(is, 13)]), ivs, &(ri[WS(is, 1)]));
488 Tl = VADD(Tj, Tk);
489 TW = VSUB(Tj, Tk);
490 T12 = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
491 T13 = LD(&(ii[WS(is, 13)]), ivs, &(ii[WS(is, 1)]));
492 T14 = VSUB(T12, T13);
493 T1W = VADD(T12, T13);
494 }
495 Tm = VADD(Ti, Tl);
496 T1X = VSUB(T1V, T1W);
497 T2g = VADD(T1V, T1W);
498 T10 = VADD(TW, TZ);
499 T15 = VSUB(T11, T14);
500 T1B = VADD(T11, T14);
501 T1U = VSUB(Ti, Tl);
502 T1A = VSUB(TZ, TW);
503 }
504 {
505 V T2l, T2m, T2n, T2o, T2p, T2q, T2r, T2s;
506 {
507 V Tf, Tu, T2j, T2k;
508 Tf = VADD(T7, Te);
509 Tu = VADD(Tm, Tt);
510 T2l = VSUB(Tf, Tu);
511 STM4(&(ro[8]), T2l, ovs, &(ro[0]));
512 T2m = VADD(Tf, Tu);
513 STM4(&(ro[0]), T2m, ovs, &(ro[0]));
514 T2j = VADD(TC, TJ);
515 T2k = VADD(T2g, T2h);
516 T2n = VSUB(T2j, T2k);
517 STM4(&(io[8]), T2n, ovs, &(io[0]));
518 T2o = VADD(T2j, T2k);
519 STM4(&(io[0]), T2o, ovs, &(io[0]));
520 }
521 {
522 V Tv, TK, T2f, T2i;
523 Tv = VSUB(Tt, Tm);
524 TK = VSUB(TC, TJ);
525 T2p = VADD(Tv, TK);
526 STM4(&(io[4]), T2p, ovs, &(io[0]));
527 T2q = VSUB(TK, Tv);
528 STM4(&(io[12]), T2q, ovs, &(io[0]));
529 T2f = VSUB(T7, Te);
530 T2i = VSUB(T2g, T2h);
531 T2r = VSUB(T2f, T2i);
532 STM4(&(ro[12]), T2r, ovs, &(ro[0]));
533 T2s = VADD(T2f, T2i);
534 STM4(&(ro[4]), T2s, ovs, &(ro[0]));
535 }
536 {
537 V T2t, T2u, T2v, T2w, T2x, T2y, T2z, T2A;
538 {
539 V T1T, T27, T24, T28, T1Y, T23;
540 T1T = VADD(T1R, T1S);
541 T27 = VSUB(T25, T26);
542 T1Y = VADD(T1U, T1X);
543 T23 = VSUB(T1Z, T22);
544 T24 = VMUL(LDK(KP707106781), VADD(T1Y, T23));
545 T28 = VMUL(LDK(KP707106781), VSUB(T23, T1Y));
546 T2t = VSUB(T1T, T24);
547 STM4(&(ro[10]), T2t, ovs, &(ro[0]));
548 T2u = VADD(T27, T28);
549 STM4(&(io[6]), T2u, ovs, &(io[0]));
550 T2v = VADD(T1T, T24);
551 STM4(&(ro[2]), T2v, ovs, &(ro[0]));
552 T2w = VSUB(T27, T28);
553 STM4(&(io[14]), T2w, ovs, &(io[0]));
554 }
555 {
556 V T29, T2d, T2c, T2e, T2a, T2b;
557 T29 = VSUB(T1R, T1S);
558 T2d = VADD(T26, T25);
559 T2a = VSUB(T1X, T1U);
560 T2b = VADD(T1Z, T22);
561 T2c = VMUL(LDK(KP707106781), VSUB(T2a, T2b));
562 T2e = VMUL(LDK(KP707106781), VADD(T2a, T2b));
563 T2x = VSUB(T29, T2c);
564 STM4(&(ro[14]), T2x, ovs, &(ro[0]));
565 T2y = VADD(T2d, T2e);
566 STM4(&(io[2]), T2y, ovs, &(io[0]));
567 T2z = VADD(T29, T2c);
568 STM4(&(ro[6]), T2z, ovs, &(ro[0]));
569 T2A = VSUB(T2d, T2e);
570 STM4(&(io[10]), T2A, ovs, &(io[0]));
571 }
572 {
573 V T2B, T2C, T2D, T2E, T2F, T2G, T2H, T2I;
574 {
575 V TV, T1r, T1p, T1v, T1i, T1q, T1u, T1w, TU, T1o;
576 TU = VMUL(LDK(KP707106781), VSUB(TQ, TT));
577 TV = VADD(TN, TU);
578 T1r = VSUB(TN, TU);
579 T1o = VMUL(LDK(KP707106781), VSUB(T1m, T1n));
580 T1p = VSUB(T1l, T1o);
581 T1v = VADD(T1l, T1o);
582 {
583 V T16, T1h, T1s, T1t;
584 T16 = VFMA(LDK(KP923879532), T10, VMUL(LDK(KP382683432), T15));
585 T1h = VFNMS(LDK(KP923879532), T1g, VMUL(LDK(KP382683432), T1b));
586 T1i = VADD(T16, T1h);
587 T1q = VSUB(T1h, T16);
588 T1s = VFNMS(LDK(KP923879532), T15, VMUL(LDK(KP382683432), T10));
589 T1t = VFMA(LDK(KP382683432), T1g, VMUL(LDK(KP923879532), T1b));
590 T1u = VSUB(T1s, T1t);
591 T1w = VADD(T1s, T1t);
592 }
593 T2B = VSUB(TV, T1i);
594 STM4(&(ro[11]), T2B, ovs, &(ro[1]));
595 T2C = VSUB(T1v, T1w);
596 STM4(&(io[11]), T2C, ovs, &(io[1]));
597 T2D = VADD(TV, T1i);
598 STM4(&(ro[3]), T2D, ovs, &(ro[1]));
599 T2E = VADD(T1v, T1w);
600 STM4(&(io[3]), T2E, ovs, &(io[1]));
601 T2F = VSUB(T1p, T1q);
602 STM4(&(io[15]), T2F, ovs, &(io[1]));
603 T2G = VSUB(T1r, T1u);
604 STM4(&(ro[15]), T2G, ovs, &(ro[1]));
605 T2H = VADD(T1p, T1q);
606 STM4(&(io[7]), T2H, ovs, &(io[1]));
607 T2I = VADD(T1r, T1u);
608 STM4(&(ro[7]), T2I, ovs, &(ro[1]));
609 }
610 {
611 V T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
612 T1y = VMUL(LDK(KP707106781), VADD(T1n, T1m));
613 T1z = VADD(T1x, T1y);
614 T1L = VSUB(T1x, T1y);
615 T1I = VMUL(LDK(KP707106781), VADD(TQ, TT));
616 T1J = VSUB(T1H, T1I);
617 T1P = VADD(T1H, T1I);
618 {
619 V T1C, T1F, T1M, T1N;
620 T1C = VFMA(LDK(KP382683432), T1A, VMUL(LDK(KP923879532), T1B));
621 T1F = VFNMS(LDK(KP382683432), T1E, VMUL(LDK(KP923879532), T1D));
622 T1G = VADD(T1C, T1F);
623 T1K = VSUB(T1F, T1C);
624 T1M = VFNMS(LDK(KP382683432), T1B, VMUL(LDK(KP923879532), T1A));
625 T1N = VFMA(LDK(KP923879532), T1E, VMUL(LDK(KP382683432), T1D));
626 T1O = VSUB(T1M, T1N);
627 T1Q = VADD(T1M, T1N);
628 }
629 {
630 V T2J, T2K, T2L, T2M;
631 T2J = VSUB(T1z, T1G);
632 STM4(&(ro[9]), T2J, ovs, &(ro[1]));
633 STN4(&(ro[8]), T2l, T2J, T2t, T2B, ovs);
634 T2K = VSUB(T1P, T1Q);
635 STM4(&(io[9]), T2K, ovs, &(io[1]));
636 STN4(&(io[8]), T2n, T2K, T2A, T2C, ovs);
637 T2L = VADD(T1z, T1G);
638 STM4(&(ro[1]), T2L, ovs, &(ro[1]));
639 STN4(&(ro[0]), T2m, T2L, T2v, T2D, ovs);
640 T2M = VADD(T1P, T1Q);
641 STM4(&(io[1]), T2M, ovs, &(io[1]));
642 STN4(&(io[0]), T2o, T2M, T2y, T2E, ovs);
643 }
644 {
645 V T2N, T2O, T2P, T2Q;
646 T2N = VSUB(T1J, T1K);
647 STM4(&(io[13]), T2N, ovs, &(io[1]));
648 STN4(&(io[12]), T2q, T2N, T2w, T2F, ovs);
649 T2O = VSUB(T1L, T1O);
650 STM4(&(ro[13]), T2O, ovs, &(ro[1]));
651 STN4(&(ro[12]), T2r, T2O, T2x, T2G, ovs);
652 T2P = VADD(T1J, T1K);
653 STM4(&(io[5]), T2P, ovs, &(io[1]));
654 STN4(&(io[4]), T2p, T2P, T2u, T2H, ovs);
655 T2Q = VADD(T1L, T1O);
656 STM4(&(ro[5]), T2Q, ovs, &(ro[1]));
657 STN4(&(ro[4]), T2s, T2Q, T2z, T2I, ovs);
658 }
659 }
660 }
661 }
662 }
663 }
664 }
665 VLEAVE();
666 }
667
668 static const kdft_desc desc = { 16, XSIMD_STRING("n2sv_16"), {136, 16, 8, 0}, &GENUS, 0, 1, 0, 0 };
669
670 void XSIMD(codelet_n2sv_16) (planner *p) {
671 X(kdft_register) (p, n2sv_16, &desc);
672 }
673
674 #endif