Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2sv_16.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 07:37:48 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n2sv_16 -with-ostride 1 -include 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 * 110 stack variables, 3 constants, and 72 memory accesses
34 */
35 #include "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(KP707106781, +0.707106781186547524400844362104849039284835938);
41 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
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 T2p, T2q, T2r, T2s, T2x, T2y, T2z, T2A, T1M, T1N, T1L, T1P, T2F, T2G, T2H;
46 V T2I, T1O, T1Q;
47 {
48 V T1l, T1H, T1R, T7, T1x, TN, TC, T25, T1E, T1b, T1Z, Tt, T2h, T22, T1D;
49 V T1g, T1n, TQ, T11, Ti, Te, T26, T1m, TT, T1S, TJ, TZ, T1V, TW, Tl;
50 V T12, T13;
51 {
52 V Tq, T1c, Tp, T20, T1a, Tr, T1d, T1e;
53 {
54 V T1, T2, Tw, Tx, T4, T5, Tz, TA;
55 T1 = LD(&(ri[0]), ivs, &(ri[0]));
56 T2 = LD(&(ri[WS(is, 8)]), ivs, &(ri[0]));
57 Tw = LD(&(ii[0]), ivs, &(ii[0]));
58 Tx = LD(&(ii[WS(is, 8)]), ivs, &(ii[0]));
59 T4 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
60 T5 = LD(&(ri[WS(is, 12)]), ivs, &(ri[0]));
61 Tz = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
62 TA = LD(&(ii[WS(is, 12)]), ivs, &(ii[0]));
63 {
64 V Tn, TL, T3, T1k, Ty, T1j, T6, TM, TB, To, T18, T19;
65 Tn = LD(&(ri[WS(is, 15)]), ivs, &(ri[WS(is, 1)]));
66 TL = VSUB(T1, T2);
67 T3 = VADD(T1, T2);
68 T1k = VSUB(Tw, Tx);
69 Ty = VADD(Tw, Tx);
70 T1j = VSUB(T4, T5);
71 T6 = VADD(T4, T5);
72 TM = VSUB(Tz, TA);
73 TB = VADD(Tz, TA);
74 To = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
75 T18 = LD(&(ii[WS(is, 15)]), ivs, &(ii[WS(is, 1)]));
76 T19 = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
77 Tq = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
78 T1l = VADD(T1j, T1k);
79 T1H = VSUB(T1k, T1j);
80 T1R = VSUB(T3, T6);
81 T7 = VADD(T3, T6);
82 T1x = VADD(TL, TM);
83 TN = VSUB(TL, TM);
84 TC = VADD(Ty, TB);
85 T25 = VSUB(Ty, TB);
86 T1c = VSUB(Tn, To);
87 Tp = VADD(Tn, To);
88 T20 = VADD(T18, T19);
89 T1a = VSUB(T18, T19);
90 Tr = LD(&(ri[WS(is, 11)]), ivs, &(ri[WS(is, 1)]));
91 T1d = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
92 T1e = LD(&(ii[WS(is, 11)]), ivs, &(ii[WS(is, 1)]));
93 }
94 }
95 {
96 V Tb, Ta, TF, Tc, TG, TH, TP, TO;
97 {
98 V T8, T9, TD, TE;
99 T8 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
100 T9 = LD(&(ri[WS(is, 10)]), ivs, &(ri[0]));
101 TD = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
102 TE = LD(&(ii[WS(is, 10)]), ivs, &(ii[0]));
103 Tb = LD(&(ri[WS(is, 14)]), ivs, &(ri[0]));
104 {
105 V T17, Ts, T21, T1f;
106 T17 = VSUB(Tq, Tr);
107 Ts = VADD(Tq, Tr);
108 T21 = VADD(T1d, T1e);
109 T1f = VSUB(T1d, T1e);
110 TP = VSUB(T8, T9);
111 Ta = VADD(T8, T9);
112 TO = VSUB(TD, TE);
113 TF = VADD(TD, TE);
114 T1E = VSUB(T1a, T17);
115 T1b = VADD(T17, T1a);
116 T1Z = VSUB(Tp, Ts);
117 Tt = VADD(Tp, Ts);
118 T2h = VADD(T20, T21);
119 T22 = VSUB(T20, T21);
120 T1D = VADD(T1c, T1f);
121 T1g = VSUB(T1c, T1f);
122 Tc = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
123 }
124 TG = LD(&(ii[WS(is, 14)]), ivs, &(ii[0]));
125 TH = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
126 }
127 T1n = VADD(TP, TO);
128 TQ = VSUB(TO, TP);
129 {
130 V Tg, Th, TX, TR, Td, TS, TI, TY, Tj, Tk;
131 Tg = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
132 Th = LD(&(ri[WS(is, 9)]), ivs, &(ri[WS(is, 1)]));
133 TX = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
134 TR = VSUB(Tb, Tc);
135 Td = VADD(Tb, Tc);
136 TS = VSUB(TG, TH);
137 TI = VADD(TG, TH);
138 TY = LD(&(ii[WS(is, 9)]), ivs, &(ii[WS(is, 1)]));
139 Tj = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
140 T11 = VSUB(Tg, Th);
141 Ti = VADD(Tg, Th);
142 Tk = LD(&(ri[WS(is, 13)]), ivs, &(ri[WS(is, 1)]));
143 Te = VADD(Ta, Td);
144 T26 = VSUB(Td, Ta);
145 T1m = VSUB(TR, TS);
146 TT = VADD(TR, TS);
147 T1S = VSUB(TF, TI);
148 TJ = VADD(TF, TI);
149 TZ = VSUB(TX, TY);
150 T1V = VADD(TX, TY);
151 TW = VSUB(Tj, Tk);
152 Tl = VADD(Tj, Tk);
153 T12 = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
154 T13 = LD(&(ii[WS(is, 13)]), ivs, &(ii[WS(is, 1)]));
155 }
156 }
157 }
158 {
159 V T2f, Tf, T2j, TK, Tm, T1U, T10, T1B, T14, T1W;
160 T2f = VSUB(T7, Te);
161 Tf = VADD(T7, Te);
162 T2j = VADD(TC, TJ);
163 TK = VSUB(TC, TJ);
164 Tm = VADD(Ti, Tl);
165 T1U = VSUB(Ti, Tl);
166 T10 = VADD(TW, TZ);
167 T1B = VSUB(TZ, TW);
168 T14 = VSUB(T12, T13);
169 T1W = VADD(T12, T13);
170 {
171 V T29, T1T, T27, T2d, T2b, T23, T15, T1A, T2l, T2m, T2n, T2o, T2i, T2k, T1Y;
172 V T2a;
173 {
174 V Tv, Tu, T1X, T2g;
175 T29 = VSUB(T1R, T1S);
176 T1T = VADD(T1R, T1S);
177 T27 = VSUB(T25, T26);
178 T2d = VADD(T26, T25);
179 T2b = VADD(T1Z, T22);
180 T23 = VSUB(T1Z, T22);
181 Tv = VSUB(Tt, Tm);
182 Tu = VADD(Tm, Tt);
183 T1X = VSUB(T1V, T1W);
184 T2g = VADD(T1V, T1W);
185 T15 = VSUB(T11, T14);
186 T1A = VADD(T11, T14);
187 T2l = VSUB(TK, Tv);
188 STM4(&(io[12]), T2l, ovs, &(io[0]));
189 T2m = VADD(Tv, TK);
190 STM4(&(io[4]), T2m, ovs, &(io[0]));
191 T2n = VADD(Tf, Tu);
192 STM4(&(ro[0]), T2n, ovs, &(ro[0]));
193 T2o = VSUB(Tf, Tu);
194 STM4(&(ro[8]), T2o, ovs, &(ro[0]));
195 T2i = VSUB(T2g, T2h);
196 T2k = VADD(T2g, T2h);
197 T1Y = VADD(T1U, T1X);
198 T2a = VSUB(T1X, T1U);
199 }
200 {
201 V T1I, T1y, T1t, T16, T1v, TV, T1r, T1p, T2t, T2u, T2v, T2w, T1h, T1s, TU;
202 V T1o;
203 T1I = VADD(TQ, TT);
204 TU = VSUB(TQ, TT);
205 T1o = VSUB(T1m, T1n);
206 T1y = VADD(T1n, T1m);
207 T1t = VFNMS(LDK(KP414213562), T10, T15);
208 T16 = VFMA(LDK(KP414213562), T15, T10);
209 T2p = VADD(T2f, T2i);
210 STM4(&(ro[4]), T2p, ovs, &(ro[0]));
211 T2q = VSUB(T2f, T2i);
212 STM4(&(ro[12]), T2q, ovs, &(ro[0]));
213 T2r = VADD(T2j, T2k);
214 STM4(&(io[0]), T2r, ovs, &(io[0]));
215 T2s = VSUB(T2j, T2k);
216 STM4(&(io[8]), T2s, ovs, &(io[0]));
217 {
218 V T28, T24, T2e, T2c;
219 T28 = VSUB(T23, T1Y);
220 T24 = VADD(T1Y, T23);
221 T2e = VADD(T2a, T2b);
222 T2c = VSUB(T2a, T2b);
223 T1v = VFNMS(LDK(KP707106781), TU, TN);
224 TV = VFMA(LDK(KP707106781), TU, TN);
225 T1r = VFMA(LDK(KP707106781), T1o, T1l);
226 T1p = VFNMS(LDK(KP707106781), T1o, T1l);
227 T2t = VFNMS(LDK(KP707106781), T28, T27);
228 STM4(&(io[14]), T2t, ovs, &(io[0]));
229 T2u = VFMA(LDK(KP707106781), T28, T27);
230 STM4(&(io[6]), T2u, ovs, &(io[0]));
231 T2v = VFMA(LDK(KP707106781), T24, T1T);
232 STM4(&(ro[2]), T2v, ovs, &(ro[0]));
233 T2w = VFNMS(LDK(KP707106781), T24, T1T);
234 STM4(&(ro[10]), T2w, ovs, &(ro[0]));
235 T2x = VFNMS(LDK(KP707106781), T2e, T2d);
236 STM4(&(io[10]), T2x, ovs, &(io[0]));
237 T2y = VFMA(LDK(KP707106781), T2e, T2d);
238 STM4(&(io[2]), T2y, ovs, &(io[0]));
239 T2z = VFMA(LDK(KP707106781), T2c, T29);
240 STM4(&(ro[6]), T2z, ovs, &(ro[0]));
241 T2A = VFNMS(LDK(KP707106781), T2c, T29);
242 STM4(&(ro[14]), T2A, ovs, &(ro[0]));
243 T1h = VFNMS(LDK(KP414213562), T1g, T1b);
244 T1s = VFMA(LDK(KP414213562), T1b, T1g);
245 }
246 {
247 V T1z, T1J, T1K, T1G, T2B, T2C, T2D, T2E, T1C, T1F;
248 T1M = VFNMS(LDK(KP414213562), T1A, T1B);
249 T1C = VFMA(LDK(KP414213562), T1B, T1A);
250 T1F = VFNMS(LDK(KP414213562), T1E, T1D);
251 T1N = VFMA(LDK(KP414213562), T1D, T1E);
252 {
253 V T1q, T1i, T1w, T1u;
254 T1q = VADD(T16, T1h);
255 T1i = VSUB(T16, T1h);
256 T1w = VADD(T1t, T1s);
257 T1u = VSUB(T1s, T1t);
258 T1L = VFNMS(LDK(KP707106781), T1y, T1x);
259 T1z = VFMA(LDK(KP707106781), T1y, T1x);
260 T1P = VFMA(LDK(KP707106781), T1I, T1H);
261 T1J = VFNMS(LDK(KP707106781), T1I, T1H);
262 T1K = VSUB(T1F, T1C);
263 T1G = VADD(T1C, T1F);
264 T2B = VFMA(LDK(KP923879532), T1q, T1p);
265 STM4(&(io[15]), T2B, ovs, &(io[1]));
266 T2C = VFNMS(LDK(KP923879532), T1q, T1p);
267 STM4(&(io[7]), T2C, ovs, &(io[1]));
268 T2D = VFMA(LDK(KP923879532), T1i, TV);
269 STM4(&(ro[3]), T2D, ovs, &(ro[1]));
270 T2E = VFNMS(LDK(KP923879532), T1i, TV);
271 STM4(&(ro[11]), T2E, ovs, &(ro[1]));
272 T2F = VFMA(LDK(KP923879532), T1w, T1v);
273 STM4(&(ro[15]), T2F, ovs, &(ro[1]));
274 T2G = VFNMS(LDK(KP923879532), T1w, T1v);
275 STM4(&(ro[7]), T2G, ovs, &(ro[1]));
276 T2H = VFMA(LDK(KP923879532), T1u, T1r);
277 STM4(&(io[3]), T2H, ovs, &(io[1]));
278 T2I = VFNMS(LDK(KP923879532), T1u, T1r);
279 STM4(&(io[11]), T2I, ovs, &(io[1]));
280 }
281 {
282 V T2J, T2K, T2L, T2M;
283 T2J = VFNMS(LDK(KP923879532), T1G, T1z);
284 STM4(&(ro[9]), T2J, ovs, &(ro[1]));
285 STN4(&(ro[8]), T2o, T2J, T2w, T2E, ovs);
286 T2K = VFMA(LDK(KP923879532), T1G, T1z);
287 STM4(&(ro[1]), T2K, ovs, &(ro[1]));
288 STN4(&(ro[0]), T2n, T2K, T2v, T2D, ovs);
289 T2L = VFNMS(LDK(KP923879532), T1K, T1J);
290 STM4(&(io[13]), T2L, ovs, &(io[1]));
291 STN4(&(io[12]), T2l, T2L, T2t, T2B, ovs);
292 T2M = VFMA(LDK(KP923879532), T1K, T1J);
293 STM4(&(io[5]), T2M, ovs, &(io[1]));
294 STN4(&(io[4]), T2m, T2M, T2u, T2C, ovs);
295 }
296 }
297 }
298 }
299 }
300 }
301 T1O = VSUB(T1M, T1N);
302 T1Q = VADD(T1M, T1N);
303 {
304 V T2N, T2O, T2P, T2Q;
305 T2N = VFMA(LDK(KP923879532), T1Q, T1P);
306 STM4(&(io[1]), T2N, ovs, &(io[1]));
307 STN4(&(io[0]), T2r, T2N, T2y, T2H, ovs);
308 T2O = VFNMS(LDK(KP923879532), T1Q, T1P);
309 STM4(&(io[9]), T2O, ovs, &(io[1]));
310 STN4(&(io[8]), T2s, T2O, T2x, T2I, ovs);
311 T2P = VFMA(LDK(KP923879532), T1O, T1L);
312 STM4(&(ro[5]), T2P, ovs, &(ro[1]));
313 STN4(&(ro[4]), T2p, T2P, T2z, T2G, ovs);
314 T2Q = VFNMS(LDK(KP923879532), T1O, T1L);
315 STM4(&(ro[13]), T2Q, ovs, &(ro[1]));
316 STN4(&(ro[12]), T2q, T2Q, T2A, T2F, ovs);
317 }
318 }
319 }
320 VLEAVE();
321 }
322
323 static const kdft_desc desc = { 16, XSIMD_STRING("n2sv_16"), {104, 0, 40, 0}, &GENUS, 0, 1, 0, 0 };
324
325 void XSIMD(codelet_n2sv_16) (planner *p) {
326 X(kdft_register) (p, n2sv_16, &desc);
327 }
328
329 #else /* HAVE_FMA */
330
331 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n2sv_16 -with-ostride 1 -include n2s.h -store-multiple 4 */
332
333 /*
334 * This function contains 144 FP additions, 24 FP multiplications,
335 * (or, 136 additions, 16 multiplications, 8 fused multiply/add),
336 * 74 stack variables, 3 constants, and 72 memory accesses
337 */
338 #include "n2s.h"
339
340 static void n2sv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
341 {
342 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
343 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
344 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
345 {
346 INT i;
347 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)) {
348 V T7, T1R, T25, TC, TN, T1x, T1H, T1l, Tt, T22, T2h, T1b, T1g, T1E, T1Z;
349 V T1D, Te, T1S, T26, TJ, TQ, T1m, T1n, TT, Tm, T1X, T2g, T10, T15, T1B;
350 V T1U, T1A;
351 {
352 V T3, TL, Ty, T1k, T6, T1j, TB, TM;
353 {
354 V T1, T2, Tw, Tx;
355 T1 = LD(&(ri[0]), ivs, &(ri[0]));
356 T2 = LD(&(ri[WS(is, 8)]), ivs, &(ri[0]));
357 T3 = VADD(T1, T2);
358 TL = VSUB(T1, T2);
359 Tw = LD(&(ii[0]), ivs, &(ii[0]));
360 Tx = LD(&(ii[WS(is, 8)]), ivs, &(ii[0]));
361 Ty = VADD(Tw, Tx);
362 T1k = VSUB(Tw, Tx);
363 }
364 {
365 V T4, T5, Tz, TA;
366 T4 = LD(&(ri[WS(is, 4)]), ivs, &(ri[0]));
367 T5 = LD(&(ri[WS(is, 12)]), ivs, &(ri[0]));
368 T6 = VADD(T4, T5);
369 T1j = VSUB(T4, T5);
370 Tz = LD(&(ii[WS(is, 4)]), ivs, &(ii[0]));
371 TA = LD(&(ii[WS(is, 12)]), ivs, &(ii[0]));
372 TB = VADD(Tz, TA);
373 TM = VSUB(Tz, TA);
374 }
375 T7 = VADD(T3, T6);
376 T1R = VSUB(T3, T6);
377 T25 = VSUB(Ty, TB);
378 TC = VADD(Ty, TB);
379 TN = VSUB(TL, TM);
380 T1x = VADD(TL, TM);
381 T1H = VSUB(T1k, T1j);
382 T1l = VADD(T1j, T1k);
383 }
384 {
385 V Tp, T17, T1f, T20, Ts, T1c, T1a, T21;
386 {
387 V Tn, To, T1d, T1e;
388 Tn = LD(&(ri[WS(is, 15)]), ivs, &(ri[WS(is, 1)]));
389 To = LD(&(ri[WS(is, 7)]), ivs, &(ri[WS(is, 1)]));
390 Tp = VADD(Tn, To);
391 T17 = VSUB(Tn, To);
392 T1d = LD(&(ii[WS(is, 15)]), ivs, &(ii[WS(is, 1)]));
393 T1e = LD(&(ii[WS(is, 7)]), ivs, &(ii[WS(is, 1)]));
394 T1f = VSUB(T1d, T1e);
395 T20 = VADD(T1d, T1e);
396 }
397 {
398 V Tq, Tr, T18, T19;
399 Tq = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
400 Tr = LD(&(ri[WS(is, 11)]), ivs, &(ri[WS(is, 1)]));
401 Ts = VADD(Tq, Tr);
402 T1c = VSUB(Tq, Tr);
403 T18 = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
404 T19 = LD(&(ii[WS(is, 11)]), ivs, &(ii[WS(is, 1)]));
405 T1a = VSUB(T18, T19);
406 T21 = VADD(T18, T19);
407 }
408 Tt = VADD(Tp, Ts);
409 T22 = VSUB(T20, T21);
410 T2h = VADD(T20, T21);
411 T1b = VSUB(T17, T1a);
412 T1g = VADD(T1c, T1f);
413 T1E = VSUB(T1f, T1c);
414 T1Z = VSUB(Tp, Ts);
415 T1D = VADD(T17, T1a);
416 }
417 {
418 V Ta, TP, TF, TO, Td, TR, TI, TS;
419 {
420 V T8, T9, TD, TE;
421 T8 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
422 T9 = LD(&(ri[WS(is, 10)]), ivs, &(ri[0]));
423 Ta = VADD(T8, T9);
424 TP = VSUB(T8, T9);
425 TD = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
426 TE = LD(&(ii[WS(is, 10)]), ivs, &(ii[0]));
427 TF = VADD(TD, TE);
428 TO = VSUB(TD, TE);
429 }
430 {
431 V Tb, Tc, TG, TH;
432 Tb = LD(&(ri[WS(is, 14)]), ivs, &(ri[0]));
433 Tc = LD(&(ri[WS(is, 6)]), ivs, &(ri[0]));
434 Td = VADD(Tb, Tc);
435 TR = VSUB(Tb, Tc);
436 TG = LD(&(ii[WS(is, 14)]), ivs, &(ii[0]));
437 TH = LD(&(ii[WS(is, 6)]), ivs, &(ii[0]));
438 TI = VADD(TG, TH);
439 TS = VSUB(TG, TH);
440 }
441 Te = VADD(Ta, Td);
442 T1S = VSUB(TF, TI);
443 T26 = VSUB(Td, Ta);
444 TJ = VADD(TF, TI);
445 TQ = VSUB(TO, TP);
446 T1m = VSUB(TR, TS);
447 T1n = VADD(TP, TO);
448 TT = VADD(TR, TS);
449 }
450 {
451 V Ti, T11, TZ, T1V, Tl, TW, T14, T1W;
452 {
453 V Tg, Th, TX, TY;
454 Tg = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
455 Th = LD(&(ri[WS(is, 9)]), ivs, &(ri[WS(is, 1)]));
456 Ti = VADD(Tg, Th);
457 T11 = VSUB(Tg, Th);
458 TX = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
459 TY = LD(&(ii[WS(is, 9)]), ivs, &(ii[WS(is, 1)]));
460 TZ = VSUB(TX, TY);
461 T1V = VADD(TX, TY);
462 }
463 {
464 V Tj, Tk, T12, T13;
465 Tj = LD(&(ri[WS(is, 5)]), ivs, &(ri[WS(is, 1)]));
466 Tk = LD(&(ri[WS(is, 13)]), ivs, &(ri[WS(is, 1)]));
467 Tl = VADD(Tj, Tk);
468 TW = VSUB(Tj, Tk);
469 T12 = LD(&(ii[WS(is, 5)]), ivs, &(ii[WS(is, 1)]));
470 T13 = LD(&(ii[WS(is, 13)]), ivs, &(ii[WS(is, 1)]));
471 T14 = VSUB(T12, T13);
472 T1W = VADD(T12, T13);
473 }
474 Tm = VADD(Ti, Tl);
475 T1X = VSUB(T1V, T1W);
476 T2g = VADD(T1V, T1W);
477 T10 = VADD(TW, TZ);
478 T15 = VSUB(T11, T14);
479 T1B = VADD(T11, T14);
480 T1U = VSUB(Ti, Tl);
481 T1A = VSUB(TZ, TW);
482 }
483 {
484 V T2l, T2m, T2n, T2o, T2p, T2q, T2r, T2s;
485 {
486 V Tf, Tu, T2j, T2k;
487 Tf = VADD(T7, Te);
488 Tu = VADD(Tm, Tt);
489 T2l = VSUB(Tf, Tu);
490 STM4(&(ro[8]), T2l, ovs, &(ro[0]));
491 T2m = VADD(Tf, Tu);
492 STM4(&(ro[0]), T2m, ovs, &(ro[0]));
493 T2j = VADD(TC, TJ);
494 T2k = VADD(T2g, T2h);
495 T2n = VSUB(T2j, T2k);
496 STM4(&(io[8]), T2n, ovs, &(io[0]));
497 T2o = VADD(T2j, T2k);
498 STM4(&(io[0]), T2o, ovs, &(io[0]));
499 }
500 {
501 V Tv, TK, T2f, T2i;
502 Tv = VSUB(Tt, Tm);
503 TK = VSUB(TC, TJ);
504 T2p = VADD(Tv, TK);
505 STM4(&(io[4]), T2p, ovs, &(io[0]));
506 T2q = VSUB(TK, Tv);
507 STM4(&(io[12]), T2q, ovs, &(io[0]));
508 T2f = VSUB(T7, Te);
509 T2i = VSUB(T2g, T2h);
510 T2r = VSUB(T2f, T2i);
511 STM4(&(ro[12]), T2r, ovs, &(ro[0]));
512 T2s = VADD(T2f, T2i);
513 STM4(&(ro[4]), T2s, ovs, &(ro[0]));
514 }
515 {
516 V T2t, T2u, T2v, T2w, T2x, T2y, T2z, T2A;
517 {
518 V T1T, T27, T24, T28, T1Y, T23;
519 T1T = VADD(T1R, T1S);
520 T27 = VSUB(T25, T26);
521 T1Y = VADD(T1U, T1X);
522 T23 = VSUB(T1Z, T22);
523 T24 = VMUL(LDK(KP707106781), VADD(T1Y, T23));
524 T28 = VMUL(LDK(KP707106781), VSUB(T23, T1Y));
525 T2t = VSUB(T1T, T24);
526 STM4(&(ro[10]), T2t, ovs, &(ro[0]));
527 T2u = VADD(T27, T28);
528 STM4(&(io[6]), T2u, ovs, &(io[0]));
529 T2v = VADD(T1T, T24);
530 STM4(&(ro[2]), T2v, ovs, &(ro[0]));
531 T2w = VSUB(T27, T28);
532 STM4(&(io[14]), T2w, ovs, &(io[0]));
533 }
534 {
535 V T29, T2d, T2c, T2e, T2a, T2b;
536 T29 = VSUB(T1R, T1S);
537 T2d = VADD(T26, T25);
538 T2a = VSUB(T1X, T1U);
539 T2b = VADD(T1Z, T22);
540 T2c = VMUL(LDK(KP707106781), VSUB(T2a, T2b));
541 T2e = VMUL(LDK(KP707106781), VADD(T2a, T2b));
542 T2x = VSUB(T29, T2c);
543 STM4(&(ro[14]), T2x, ovs, &(ro[0]));
544 T2y = VADD(T2d, T2e);
545 STM4(&(io[2]), T2y, ovs, &(io[0]));
546 T2z = VADD(T29, T2c);
547 STM4(&(ro[6]), T2z, ovs, &(ro[0]));
548 T2A = VSUB(T2d, T2e);
549 STM4(&(io[10]), T2A, ovs, &(io[0]));
550 }
551 {
552 V T2B, T2C, T2D, T2E, T2F, T2G, T2H, T2I;
553 {
554 V TV, T1r, T1p, T1v, T1i, T1q, T1u, T1w, TU, T1o;
555 TU = VMUL(LDK(KP707106781), VSUB(TQ, TT));
556 TV = VADD(TN, TU);
557 T1r = VSUB(TN, TU);
558 T1o = VMUL(LDK(KP707106781), VSUB(T1m, T1n));
559 T1p = VSUB(T1l, T1o);
560 T1v = VADD(T1l, T1o);
561 {
562 V T16, T1h, T1s, T1t;
563 T16 = VFMA(LDK(KP923879532), T10, VMUL(LDK(KP382683432), T15));
564 T1h = VFNMS(LDK(KP923879532), T1g, VMUL(LDK(KP382683432), T1b));
565 T1i = VADD(T16, T1h);
566 T1q = VSUB(T1h, T16);
567 T1s = VFNMS(LDK(KP923879532), T15, VMUL(LDK(KP382683432), T10));
568 T1t = VFMA(LDK(KP382683432), T1g, VMUL(LDK(KP923879532), T1b));
569 T1u = VSUB(T1s, T1t);
570 T1w = VADD(T1s, T1t);
571 }
572 T2B = VSUB(TV, T1i);
573 STM4(&(ro[11]), T2B, ovs, &(ro[1]));
574 T2C = VSUB(T1v, T1w);
575 STM4(&(io[11]), T2C, ovs, &(io[1]));
576 T2D = VADD(TV, T1i);
577 STM4(&(ro[3]), T2D, ovs, &(ro[1]));
578 T2E = VADD(T1v, T1w);
579 STM4(&(io[3]), T2E, ovs, &(io[1]));
580 T2F = VSUB(T1p, T1q);
581 STM4(&(io[15]), T2F, ovs, &(io[1]));
582 T2G = VSUB(T1r, T1u);
583 STM4(&(ro[15]), T2G, ovs, &(ro[1]));
584 T2H = VADD(T1p, T1q);
585 STM4(&(io[7]), T2H, ovs, &(io[1]));
586 T2I = VADD(T1r, T1u);
587 STM4(&(ro[7]), T2I, ovs, &(ro[1]));
588 }
589 {
590 V T1z, T1L, T1J, T1P, T1G, T1K, T1O, T1Q, T1y, T1I;
591 T1y = VMUL(LDK(KP707106781), VADD(T1n, T1m));
592 T1z = VADD(T1x, T1y);
593 T1L = VSUB(T1x, T1y);
594 T1I = VMUL(LDK(KP707106781), VADD(TQ, TT));
595 T1J = VSUB(T1H, T1I);
596 T1P = VADD(T1H, T1I);
597 {
598 V T1C, T1F, T1M, T1N;
599 T1C = VFMA(LDK(KP382683432), T1A, VMUL(LDK(KP923879532), T1B));
600 T1F = VFNMS(LDK(KP382683432), T1E, VMUL(LDK(KP923879532), T1D));
601 T1G = VADD(T1C, T1F);
602 T1K = VSUB(T1F, T1C);
603 T1M = VFNMS(LDK(KP382683432), T1B, VMUL(LDK(KP923879532), T1A));
604 T1N = VFMA(LDK(KP923879532), T1E, VMUL(LDK(KP382683432), T1D));
605 T1O = VSUB(T1M, T1N);
606 T1Q = VADD(T1M, T1N);
607 }
608 {
609 V T2J, T2K, T2L, T2M;
610 T2J = VSUB(T1z, T1G);
611 STM4(&(ro[9]), T2J, ovs, &(ro[1]));
612 STN4(&(ro[8]), T2l, T2J, T2t, T2B, ovs);
613 T2K = VSUB(T1P, T1Q);
614 STM4(&(io[9]), T2K, ovs, &(io[1]));
615 STN4(&(io[8]), T2n, T2K, T2A, T2C, ovs);
616 T2L = VADD(T1z, T1G);
617 STM4(&(ro[1]), T2L, ovs, &(ro[1]));
618 STN4(&(ro[0]), T2m, T2L, T2v, T2D, ovs);
619 T2M = VADD(T1P, T1Q);
620 STM4(&(io[1]), T2M, ovs, &(io[1]));
621 STN4(&(io[0]), T2o, T2M, T2y, T2E, ovs);
622 }
623 {
624 V T2N, T2O, T2P, T2Q;
625 T2N = VSUB(T1J, T1K);
626 STM4(&(io[13]), T2N, ovs, &(io[1]));
627 STN4(&(io[12]), T2q, T2N, T2w, T2F, ovs);
628 T2O = VSUB(T1L, T1O);
629 STM4(&(ro[13]), T2O, ovs, &(ro[1]));
630 STN4(&(ro[12]), T2r, T2O, T2x, T2G, ovs);
631 T2P = VADD(T1J, T1K);
632 STM4(&(io[5]), T2P, ovs, &(io[1]));
633 STN4(&(io[4]), T2p, T2P, T2u, T2H, ovs);
634 T2Q = VADD(T1L, T1O);
635 STM4(&(ro[5]), T2Q, ovs, &(ro[1]));
636 STN4(&(ro[4]), T2s, T2Q, T2z, T2I, ovs);
637 }
638 }
639 }
640 }
641 }
642 }
643 }
644 VLEAVE();
645 }
646
647 static const kdft_desc desc = { 16, XSIMD_STRING("n2sv_16"), {136, 16, 8, 0}, &GENUS, 0, 1, 0, 0 };
648
649 void XSIMD(codelet_n2sv_16) (planner *p) {
650 X(kdft_register) (p, n2sv_16, &desc);
651 }
652
653 #endif /* HAVE_FMA */