Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/t2sv_16.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 08:06:11 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_twiddle.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 16 -name t2sv_16 -include dft/simd/ts.h */
29
30 /*
31 * This function contains 196 FP additions, 134 FP multiplications,
32 * (or, 104 additions, 42 multiplications, 92 fused multiply/add),
33 * 90 stack variables, 3 constants, and 64 memory accesses
34 */
35 #include "dft/simd/ts.h"
36
37 static void t2sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT m;
44 for (m = mb, W = W + (mb * 8); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 8), MAKE_VOLATILE_STRIDE(32, rs)) {
45 V T2, Tf, TM, TO, T3, T6, T5, Th, Tz, Ti, T7, TZ, TT, Tq, TW;
46 V Tb, Tu, TP, TI, TF, TC, T1z, T1O, T1D, T1L, Tm, T1f, T1p, T1j, T1m;
47 {
48 V TN, TS, T4, Tp, Ta, Tt, Tl, Tg;
49 T2 = LDW(&(W[0]));
50 Tf = LDW(&(W[TWVL * 2]));
51 Tg = VMUL(T2, Tf);
52 TM = LDW(&(W[TWVL * 6]));
53 TN = VMUL(T2, TM);
54 TO = LDW(&(W[TWVL * 7]));
55 TS = VMUL(T2, TO);
56 T3 = LDW(&(W[TWVL * 4]));
57 T4 = VMUL(T2, T3);
58 Tp = VMUL(Tf, T3);
59 T6 = LDW(&(W[TWVL * 5]));
60 Ta = VMUL(T2, T6);
61 Tt = VMUL(Tf, T6);
62 T5 = LDW(&(W[TWVL * 1]));
63 Th = LDW(&(W[TWVL * 3]));
64 Tl = VMUL(T2, Th);
65 Tz = VFMA(T5, Th, Tg);
66 Ti = VFNMS(T5, Th, Tg);
67 T7 = VFMA(T5, T6, T4);
68 TZ = VFNMS(Th, T3, Tt);
69 TT = VFNMS(T5, TM, TS);
70 Tq = VFNMS(Th, T6, Tp);
71 TW = VFMA(Th, T6, Tp);
72 Tb = VFNMS(T5, T3, Ta);
73 Tu = VFMA(Th, T3, Tt);
74 TP = VFMA(T5, TO, TN);
75 TI = VFMA(T5, T3, Ta);
76 TF = VFNMS(T5, T6, T4);
77 {
78 V T1y, T1C, T1e, T1i;
79 T1y = VMUL(Tz, T3);
80 T1C = VMUL(Tz, T6);
81 TC = VFNMS(T5, Tf, Tl);
82 T1z = VFMA(TC, T6, T1y);
83 T1O = VFMA(TC, T3, T1C);
84 T1D = VFNMS(TC, T3, T1C);
85 T1L = VFNMS(TC, T6, T1y);
86 T1e = VMUL(Ti, T3);
87 T1i = VMUL(Ti, T6);
88 Tm = VFMA(T5, Tf, Tl);
89 T1f = VFMA(Tm, T6, T1e);
90 T1p = VFMA(Tm, T3, T1i);
91 T1j = VFNMS(Tm, T3, T1i);
92 T1m = VFNMS(Tm, T6, T1e);
93 }
94 }
95 {
96 V Te, T1U, T3A, T3L, T1G, T2D, T2A, T3h, T1R, T2B, T2I, T3i, Tx, T3M, T1Z;
97 V T3w, TL, T26, T25, T37, T1d, T2o, T2l, T3c, T1s, T2m, T2t, T3d, T12, T28;
98 V T2d, T38;
99 {
100 V T1, T3z, T8, T9, Tc, T3x, Td, T3y;
101 T1 = LD(&(ri[0]), ms, &(ri[0]));
102 T3z = LD(&(ii[0]), ms, &(ii[0]));
103 T8 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
104 T9 = VMUL(T7, T8);
105 Tc = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
106 T3x = VMUL(T7, Tc);
107 Td = VFMA(Tb, Tc, T9);
108 Te = VADD(T1, Td);
109 T1U = VSUB(T1, Td);
110 T3y = VFNMS(Tb, T8, T3x);
111 T3A = VADD(T3y, T3z);
112 T3L = VSUB(T3z, T3y);
113 }
114 {
115 V T1u, T1v, T1w, T2w, T1A, T1B, T1E, T2y;
116 T1u = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
117 T1v = VMUL(TM, T1u);
118 T1w = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
119 T2w = VMUL(TM, T1w);
120 T1A = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
121 T1B = VMUL(T1z, T1A);
122 T1E = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
123 T2y = VMUL(T1z, T1E);
124 {
125 V T1x, T1F, T2x, T2z;
126 T1x = VFMA(TO, T1w, T1v);
127 T1F = VFMA(T1D, T1E, T1B);
128 T1G = VADD(T1x, T1F);
129 T2D = VSUB(T1x, T1F);
130 T2x = VFNMS(TO, T1u, T2w);
131 T2z = VFNMS(T1D, T1A, T2y);
132 T2A = VSUB(T2x, T2z);
133 T3h = VADD(T2x, T2z);
134 }
135 }
136 {
137 V T1H, T1I, T1J, T2E, T1M, T1N, T1P, T2G;
138 T1H = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
139 T1I = VMUL(Tf, T1H);
140 T1J = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
141 T2E = VMUL(Tf, T1J);
142 T1M = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
143 T1N = VMUL(T1L, T1M);
144 T1P = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
145 T2G = VMUL(T1L, T1P);
146 {
147 V T1K, T1Q, T2F, T2H;
148 T1K = VFMA(Th, T1J, T1I);
149 T1Q = VFMA(T1O, T1P, T1N);
150 T1R = VADD(T1K, T1Q);
151 T2B = VSUB(T1K, T1Q);
152 T2F = VFNMS(Th, T1H, T2E);
153 T2H = VFNMS(T1O, T1M, T2G);
154 T2I = VSUB(T2F, T2H);
155 T3i = VADD(T2F, T2H);
156 }
157 }
158 {
159 V Tj, Tk, Tn, T1V, Tr, Ts, Tv, T1X;
160 Tj = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
161 Tk = VMUL(Ti, Tj);
162 Tn = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
163 T1V = VMUL(Ti, Tn);
164 Tr = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
165 Ts = VMUL(Tq, Tr);
166 Tv = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
167 T1X = VMUL(Tq, Tv);
168 {
169 V To, Tw, T1W, T1Y;
170 To = VFMA(Tm, Tn, Tk);
171 Tw = VFMA(Tu, Tv, Ts);
172 Tx = VADD(To, Tw);
173 T3M = VSUB(To, Tw);
174 T1W = VFNMS(Tm, Tj, T1V);
175 T1Y = VFNMS(Tu, Tr, T1X);
176 T1Z = VSUB(T1W, T1Y);
177 T3w = VADD(T1W, T1Y);
178 }
179 }
180 {
181 V TA, TB, TD, T21, TG, TH, TJ, T23;
182 TA = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
183 TB = VMUL(Tz, TA);
184 TD = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
185 T21 = VMUL(Tz, TD);
186 TG = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
187 TH = VMUL(TF, TG);
188 TJ = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
189 T23 = VMUL(TF, TJ);
190 {
191 V TE, TK, T22, T24;
192 TE = VFMA(TC, TD, TB);
193 TK = VFMA(TI, TJ, TH);
194 TL = VADD(TE, TK);
195 T26 = VSUB(TE, TK);
196 T22 = VFNMS(TC, TA, T21);
197 T24 = VFNMS(TI, TG, T23);
198 T25 = VSUB(T22, T24);
199 T37 = VADD(T22, T24);
200 }
201 }
202 {
203 V T15, T16, T17, T2h, T19, T1a, T1b, T2j;
204 T15 = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
205 T16 = VMUL(T2, T15);
206 T17 = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
207 T2h = VMUL(T2, T17);
208 T19 = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
209 T1a = VMUL(T3, T19);
210 T1b = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
211 T2j = VMUL(T3, T1b);
212 {
213 V T18, T1c, T2i, T2k;
214 T18 = VFMA(T5, T17, T16);
215 T1c = VFMA(T6, T1b, T1a);
216 T1d = VADD(T18, T1c);
217 T2o = VSUB(T18, T1c);
218 T2i = VFNMS(T5, T15, T2h);
219 T2k = VFNMS(T6, T19, T2j);
220 T2l = VSUB(T2i, T2k);
221 T3c = VADD(T2i, T2k);
222 }
223 }
224 {
225 V T1g, T1h, T1k, T2p, T1n, T1o, T1q, T2r;
226 T1g = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
227 T1h = VMUL(T1f, T1g);
228 T1k = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
229 T2p = VMUL(T1f, T1k);
230 T1n = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
231 T1o = VMUL(T1m, T1n);
232 T1q = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
233 T2r = VMUL(T1m, T1q);
234 {
235 V T1l, T1r, T2q, T2s;
236 T1l = VFMA(T1j, T1k, T1h);
237 T1r = VFMA(T1p, T1q, T1o);
238 T1s = VADD(T1l, T1r);
239 T2m = VSUB(T1l, T1r);
240 T2q = VFNMS(T1j, T1g, T2p);
241 T2s = VFNMS(T1p, T1n, T2r);
242 T2t = VSUB(T2q, T2s);
243 T3d = VADD(T2q, T2s);
244 }
245 }
246 {
247 V TQ, TR, TU, T29, TX, TY, T10, T2b;
248 TQ = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
249 TR = VMUL(TP, TQ);
250 TU = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
251 T29 = VMUL(TP, TU);
252 TX = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
253 TY = VMUL(TW, TX);
254 T10 = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
255 T2b = VMUL(TW, T10);
256 {
257 V TV, T11, T2a, T2c;
258 TV = VFMA(TT, TU, TR);
259 T11 = VFMA(TZ, T10, TY);
260 T12 = VADD(TV, T11);
261 T28 = VSUB(TV, T11);
262 T2a = VFNMS(TT, TQ, T29);
263 T2c = VFNMS(TZ, TX, T2b);
264 T2d = VSUB(T2a, T2c);
265 T38 = VADD(T2a, T2c);
266 }
267 }
268 {
269 V T14, T3q, T3C, T3E, T1T, T3D, T3t, T3u;
270 {
271 V Ty, T13, T3v, T3B;
272 Ty = VADD(Te, Tx);
273 T13 = VADD(TL, T12);
274 T14 = VADD(Ty, T13);
275 T3q = VSUB(Ty, T13);
276 T3v = VADD(T37, T38);
277 T3B = VADD(T3w, T3A);
278 T3C = VADD(T3v, T3B);
279 T3E = VSUB(T3B, T3v);
280 }
281 {
282 V T1t, T1S, T3r, T3s;
283 T1t = VADD(T1d, T1s);
284 T1S = VADD(T1G, T1R);
285 T1T = VADD(T1t, T1S);
286 T3D = VSUB(T1S, T1t);
287 T3r = VADD(T3c, T3d);
288 T3s = VADD(T3h, T3i);
289 T3t = VSUB(T3r, T3s);
290 T3u = VADD(T3r, T3s);
291 }
292 ST(&(ri[WS(rs, 8)]), VSUB(T14, T1T), ms, &(ri[0]));
293 ST(&(ii[WS(rs, 8)]), VSUB(T3C, T3u), ms, &(ii[0]));
294 ST(&(ri[0]), VADD(T14, T1T), ms, &(ri[0]));
295 ST(&(ii[0]), VADD(T3u, T3C), ms, &(ii[0]));
296 ST(&(ri[WS(rs, 12)]), VSUB(T3q, T3t), ms, &(ri[0]));
297 ST(&(ii[WS(rs, 12)]), VSUB(T3E, T3D), ms, &(ii[0]));
298 ST(&(ri[WS(rs, 4)]), VADD(T3q, T3t), ms, &(ri[0]));
299 ST(&(ii[WS(rs, 4)]), VADD(T3D, T3E), ms, &(ii[0]));
300 }
301 {
302 V T3a, T3m, T3H, T3J, T3f, T3n, T3k, T3o;
303 {
304 V T36, T39, T3F, T3G;
305 T36 = VSUB(Te, Tx);
306 T39 = VSUB(T37, T38);
307 T3a = VADD(T36, T39);
308 T3m = VSUB(T36, T39);
309 T3F = VSUB(T12, TL);
310 T3G = VSUB(T3A, T3w);
311 T3H = VADD(T3F, T3G);
312 T3J = VSUB(T3G, T3F);
313 }
314 {
315 V T3b, T3e, T3g, T3j;
316 T3b = VSUB(T1d, T1s);
317 T3e = VSUB(T3c, T3d);
318 T3f = VADD(T3b, T3e);
319 T3n = VSUB(T3e, T3b);
320 T3g = VSUB(T1G, T1R);
321 T3j = VSUB(T3h, T3i);
322 T3k = VSUB(T3g, T3j);
323 T3o = VADD(T3g, T3j);
324 }
325 {
326 V T3l, T3I, T3p, T3K;
327 T3l = VADD(T3f, T3k);
328 ST(&(ri[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3l, T3a), ms, &(ri[0]));
329 ST(&(ri[WS(rs, 2)]), VFMA(LDK(KP707106781), T3l, T3a), ms, &(ri[0]));
330 T3I = VADD(T3n, T3o);
331 ST(&(ii[WS(rs, 2)]), VFMA(LDK(KP707106781), T3I, T3H), ms, &(ii[0]));
332 ST(&(ii[WS(rs, 10)]), VFNMS(LDK(KP707106781), T3I, T3H), ms, &(ii[0]));
333 T3p = VSUB(T3n, T3o);
334 ST(&(ri[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3p, T3m), ms, &(ri[0]));
335 ST(&(ri[WS(rs, 6)]), VFMA(LDK(KP707106781), T3p, T3m), ms, &(ri[0]));
336 T3K = VSUB(T3k, T3f);
337 ST(&(ii[WS(rs, 6)]), VFMA(LDK(KP707106781), T3K, T3J), ms, &(ii[0]));
338 ST(&(ii[WS(rs, 14)]), VFNMS(LDK(KP707106781), T3K, T3J), ms, &(ii[0]));
339 }
340 }
341 {
342 V T20, T3N, T3T, T2Q, T2f, T3O, T30, T34, T2T, T3U, T2v, T2N, T2X, T33, T2K;
343 V T2O;
344 {
345 V T27, T2e, T2n, T2u;
346 T20 = VSUB(T1U, T1Z);
347 T3N = VSUB(T3L, T3M);
348 T3T = VADD(T3M, T3L);
349 T2Q = VADD(T1U, T1Z);
350 T27 = VSUB(T25, T26);
351 T2e = VADD(T28, T2d);
352 T2f = VSUB(T27, T2e);
353 T3O = VADD(T27, T2e);
354 {
355 V T2Y, T2Z, T2R, T2S;
356 T2Y = VADD(T2D, T2I);
357 T2Z = VSUB(T2A, T2B);
358 T30 = VFNMS(LDK(KP414213562), T2Z, T2Y);
359 T34 = VFMA(LDK(KP414213562), T2Y, T2Z);
360 T2R = VADD(T26, T25);
361 T2S = VSUB(T28, T2d);
362 T2T = VADD(T2R, T2S);
363 T3U = VSUB(T2S, T2R);
364 }
365 T2n = VADD(T2l, T2m);
366 T2u = VSUB(T2o, T2t);
367 T2v = VFMA(LDK(KP414213562), T2u, T2n);
368 T2N = VFNMS(LDK(KP414213562), T2n, T2u);
369 {
370 V T2V, T2W, T2C, T2J;
371 T2V = VADD(T2o, T2t);
372 T2W = VSUB(T2l, T2m);
373 T2X = VFMA(LDK(KP414213562), T2W, T2V);
374 T33 = VFNMS(LDK(KP414213562), T2V, T2W);
375 T2C = VADD(T2A, T2B);
376 T2J = VSUB(T2D, T2I);
377 T2K = VFNMS(LDK(KP414213562), T2J, T2C);
378 T2O = VFMA(LDK(KP414213562), T2C, T2J);
379 }
380 }
381 {
382 V T2g, T2L, T3V, T3W;
383 T2g = VFMA(LDK(KP707106781), T2f, T20);
384 T2L = VSUB(T2v, T2K);
385 ST(&(ri[WS(rs, 11)]), VFNMS(LDK(KP923879532), T2L, T2g), ms, &(ri[WS(rs, 1)]));
386 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP923879532), T2L, T2g), ms, &(ri[WS(rs, 1)]));
387 T3V = VFMA(LDK(KP707106781), T3U, T3T);
388 T3W = VSUB(T2O, T2N);
389 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP923879532), T3W, T3V), ms, &(ii[WS(rs, 1)]));
390 ST(&(ii[WS(rs, 11)]), VFNMS(LDK(KP923879532), T3W, T3V), ms, &(ii[WS(rs, 1)]));
391 }
392 {
393 V T2M, T2P, T3X, T3Y;
394 T2M = VFNMS(LDK(KP707106781), T2f, T20);
395 T2P = VADD(T2N, T2O);
396 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP923879532), T2P, T2M), ms, &(ri[WS(rs, 1)]));
397 ST(&(ri[WS(rs, 15)]), VFMA(LDK(KP923879532), T2P, T2M), ms, &(ri[WS(rs, 1)]));
398 T3X = VFNMS(LDK(KP707106781), T3U, T3T);
399 T3Y = VADD(T2v, T2K);
400 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP923879532), T3Y, T3X), ms, &(ii[WS(rs, 1)]));
401 ST(&(ii[WS(rs, 15)]), VFMA(LDK(KP923879532), T3Y, T3X), ms, &(ii[WS(rs, 1)]));
402 }
403 {
404 V T2U, T31, T3P, T3Q;
405 T2U = VFMA(LDK(KP707106781), T2T, T2Q);
406 T31 = VADD(T2X, T30);
407 ST(&(ri[WS(rs, 9)]), VFNMS(LDK(KP923879532), T31, T2U), ms, &(ri[WS(rs, 1)]));
408 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP923879532), T31, T2U), ms, &(ri[WS(rs, 1)]));
409 T3P = VFMA(LDK(KP707106781), T3O, T3N);
410 T3Q = VADD(T33, T34);
411 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP923879532), T3Q, T3P), ms, &(ii[WS(rs, 1)]));
412 ST(&(ii[WS(rs, 9)]), VFNMS(LDK(KP923879532), T3Q, T3P), ms, &(ii[WS(rs, 1)]));
413 }
414 {
415 V T32, T35, T3R, T3S;
416 T32 = VFNMS(LDK(KP707106781), T2T, T2Q);
417 T35 = VSUB(T33, T34);
418 ST(&(ri[WS(rs, 13)]), VFNMS(LDK(KP923879532), T35, T32), ms, &(ri[WS(rs, 1)]));
419 ST(&(ri[WS(rs, 5)]), VFMA(LDK(KP923879532), T35, T32), ms, &(ri[WS(rs, 1)]));
420 T3R = VFNMS(LDK(KP707106781), T3O, T3N);
421 T3S = VSUB(T30, T2X);
422 ST(&(ii[WS(rs, 5)]), VFMA(LDK(KP923879532), T3S, T3R), ms, &(ii[WS(rs, 1)]));
423 ST(&(ii[WS(rs, 13)]), VFNMS(LDK(KP923879532), T3S, T3R), ms, &(ii[WS(rs, 1)]));
424 }
425 }
426 }
427 }
428 }
429 VLEAVE();
430 }
431
432 static const tw_instr twinstr[] = {
433 VTW(0, 1),
434 VTW(0, 3),
435 VTW(0, 9),
436 VTW(0, 15),
437 {TW_NEXT, (2 * VL), 0}
438 };
439
440 static const ct_desc desc = { 16, XSIMD_STRING("t2sv_16"), twinstr, &GENUS, {104, 42, 92, 0}, 0, 0, 0 };
441
442 void XSIMD(codelet_t2sv_16) (planner *p) {
443 X(kdft_dit_register) (p, t2sv_16, &desc);
444 }
445 #else
446
447 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 16 -name t2sv_16 -include dft/simd/ts.h */
448
449 /*
450 * This function contains 196 FP additions, 108 FP multiplications,
451 * (or, 156 additions, 68 multiplications, 40 fused multiply/add),
452 * 82 stack variables, 3 constants, and 64 memory accesses
453 */
454 #include "dft/simd/ts.h"
455
456 static void t2sv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
457 {
458 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
459 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
460 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
461 {
462 INT m;
463 for (m = mb, W = W + (mb * 8); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 8), MAKE_VOLATILE_STRIDE(32, rs)) {
464 V T2, T5, Tg, Ti, Tk, To, TE, TC, T6, T3, T8, TW, TJ, Tt, TU;
465 V Tc, Tx, TH, TN, TO, TP, TR, T1f, T1k, T1b, T1i, T1y, T1H, T1u, T1F;
466 {
467 V T7, Tv, Ta, Ts, T4, Tw, Tb, Tr;
468 {
469 V Th, Tn, Tj, Tm;
470 T2 = LDW(&(W[0]));
471 T5 = LDW(&(W[TWVL * 1]));
472 Tg = LDW(&(W[TWVL * 2]));
473 Ti = LDW(&(W[TWVL * 3]));
474 Th = VMUL(T2, Tg);
475 Tn = VMUL(T5, Tg);
476 Tj = VMUL(T5, Ti);
477 Tm = VMUL(T2, Ti);
478 Tk = VSUB(Th, Tj);
479 To = VADD(Tm, Tn);
480 TE = VSUB(Tm, Tn);
481 TC = VADD(Th, Tj);
482 T6 = LDW(&(W[TWVL * 5]));
483 T7 = VMUL(T5, T6);
484 Tv = VMUL(Tg, T6);
485 Ta = VMUL(T2, T6);
486 Ts = VMUL(Ti, T6);
487 T3 = LDW(&(W[TWVL * 4]));
488 T4 = VMUL(T2, T3);
489 Tw = VMUL(Ti, T3);
490 Tb = VMUL(T5, T3);
491 Tr = VMUL(Tg, T3);
492 }
493 T8 = VADD(T4, T7);
494 TW = VSUB(Tv, Tw);
495 TJ = VADD(Ta, Tb);
496 Tt = VSUB(Tr, Ts);
497 TU = VADD(Tr, Ts);
498 Tc = VSUB(Ta, Tb);
499 Tx = VADD(Tv, Tw);
500 TH = VSUB(T4, T7);
501 TN = LDW(&(W[TWVL * 6]));
502 TO = LDW(&(W[TWVL * 7]));
503 TP = VFMA(T2, TN, VMUL(T5, TO));
504 TR = VFNMS(T5, TN, VMUL(T2, TO));
505 {
506 V T1d, T1e, T19, T1a;
507 T1d = VMUL(Tk, T6);
508 T1e = VMUL(To, T3);
509 T1f = VSUB(T1d, T1e);
510 T1k = VADD(T1d, T1e);
511 T19 = VMUL(Tk, T3);
512 T1a = VMUL(To, T6);
513 T1b = VADD(T19, T1a);
514 T1i = VSUB(T19, T1a);
515 }
516 {
517 V T1w, T1x, T1s, T1t;
518 T1w = VMUL(TC, T6);
519 T1x = VMUL(TE, T3);
520 T1y = VSUB(T1w, T1x);
521 T1H = VADD(T1w, T1x);
522 T1s = VMUL(TC, T3);
523 T1t = VMUL(TE, T6);
524 T1u = VADD(T1s, T1t);
525 T1F = VSUB(T1s, T1t);
526 }
527 }
528 {
529 V Tf, T3r, T1N, T3e, TA, T3s, T1Q, T3b, TM, T2M, T1W, T2w, TZ, T2N, T21;
530 V T2x, T1B, T1K, T2V, T2W, T2X, T2Y, T2j, T2D, T2o, T2E, T18, T1n, T2Q, T2R;
531 V T2S, T2T, T28, T2A, T2d, T2B;
532 {
533 V T1, T3d, Te, T3c, T9, Td;
534 T1 = LD(&(ri[0]), ms, &(ri[0]));
535 T3d = LD(&(ii[0]), ms, &(ii[0]));
536 T9 = LD(&(ri[WS(rs, 8)]), ms, &(ri[0]));
537 Td = LD(&(ii[WS(rs, 8)]), ms, &(ii[0]));
538 Te = VFMA(T8, T9, VMUL(Tc, Td));
539 T3c = VFNMS(Tc, T9, VMUL(T8, Td));
540 Tf = VADD(T1, Te);
541 T3r = VSUB(T3d, T3c);
542 T1N = VSUB(T1, Te);
543 T3e = VADD(T3c, T3d);
544 }
545 {
546 V Tq, T1O, Tz, T1P;
547 {
548 V Tl, Tp, Tu, Ty;
549 Tl = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
550 Tp = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
551 Tq = VFMA(Tk, Tl, VMUL(To, Tp));
552 T1O = VFNMS(To, Tl, VMUL(Tk, Tp));
553 Tu = LD(&(ri[WS(rs, 12)]), ms, &(ri[0]));
554 Ty = LD(&(ii[WS(rs, 12)]), ms, &(ii[0]));
555 Tz = VFMA(Tt, Tu, VMUL(Tx, Ty));
556 T1P = VFNMS(Tx, Tu, VMUL(Tt, Ty));
557 }
558 TA = VADD(Tq, Tz);
559 T3s = VSUB(Tq, Tz);
560 T1Q = VSUB(T1O, T1P);
561 T3b = VADD(T1O, T1P);
562 }
563 {
564 V TG, T1S, TL, T1T, T1U, T1V;
565 {
566 V TD, TF, TI, TK;
567 TD = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
568 TF = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
569 TG = VFMA(TC, TD, VMUL(TE, TF));
570 T1S = VFNMS(TE, TD, VMUL(TC, TF));
571 TI = LD(&(ri[WS(rs, 10)]), ms, &(ri[0]));
572 TK = LD(&(ii[WS(rs, 10)]), ms, &(ii[0]));
573 TL = VFMA(TH, TI, VMUL(TJ, TK));
574 T1T = VFNMS(TJ, TI, VMUL(TH, TK));
575 }
576 TM = VADD(TG, TL);
577 T2M = VADD(T1S, T1T);
578 T1U = VSUB(T1S, T1T);
579 T1V = VSUB(TG, TL);
580 T1W = VSUB(T1U, T1V);
581 T2w = VADD(T1V, T1U);
582 }
583 {
584 V TT, T1Y, TY, T1Z, T1X, T20;
585 {
586 V TQ, TS, TV, TX;
587 TQ = LD(&(ri[WS(rs, 14)]), ms, &(ri[0]));
588 TS = LD(&(ii[WS(rs, 14)]), ms, &(ii[0]));
589 TT = VFMA(TP, TQ, VMUL(TR, TS));
590 T1Y = VFNMS(TR, TQ, VMUL(TP, TS));
591 TV = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
592 TX = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
593 TY = VFMA(TU, TV, VMUL(TW, TX));
594 T1Z = VFNMS(TW, TV, VMUL(TU, TX));
595 }
596 TZ = VADD(TT, TY);
597 T2N = VADD(T1Y, T1Z);
598 T1X = VSUB(TT, TY);
599 T20 = VSUB(T1Y, T1Z);
600 T21 = VADD(T1X, T20);
601 T2x = VSUB(T1X, T20);
602 }
603 {
604 V T1r, T2k, T1J, T2h, T1A, T2l, T1E, T2g;
605 {
606 V T1p, T1q, T1G, T1I;
607 T1p = LD(&(ri[WS(rs, 15)]), ms, &(ri[WS(rs, 1)]));
608 T1q = LD(&(ii[WS(rs, 15)]), ms, &(ii[WS(rs, 1)]));
609 T1r = VFMA(TN, T1p, VMUL(TO, T1q));
610 T2k = VFNMS(TO, T1p, VMUL(TN, T1q));
611 T1G = LD(&(ri[WS(rs, 11)]), ms, &(ri[WS(rs, 1)]));
612 T1I = LD(&(ii[WS(rs, 11)]), ms, &(ii[WS(rs, 1)]));
613 T1J = VFMA(T1F, T1G, VMUL(T1H, T1I));
614 T2h = VFNMS(T1H, T1G, VMUL(T1F, T1I));
615 }
616 {
617 V T1v, T1z, T1C, T1D;
618 T1v = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
619 T1z = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
620 T1A = VFMA(T1u, T1v, VMUL(T1y, T1z));
621 T2l = VFNMS(T1y, T1v, VMUL(T1u, T1z));
622 T1C = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
623 T1D = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
624 T1E = VFMA(Tg, T1C, VMUL(Ti, T1D));
625 T2g = VFNMS(Ti, T1C, VMUL(Tg, T1D));
626 }
627 T1B = VADD(T1r, T1A);
628 T1K = VADD(T1E, T1J);
629 T2V = VSUB(T1B, T1K);
630 T2W = VADD(T2k, T2l);
631 T2X = VADD(T2g, T2h);
632 T2Y = VSUB(T2W, T2X);
633 {
634 V T2f, T2i, T2m, T2n;
635 T2f = VSUB(T1r, T1A);
636 T2i = VSUB(T2g, T2h);
637 T2j = VSUB(T2f, T2i);
638 T2D = VADD(T2f, T2i);
639 T2m = VSUB(T2k, T2l);
640 T2n = VSUB(T1E, T1J);
641 T2o = VADD(T2m, T2n);
642 T2E = VSUB(T2m, T2n);
643 }
644 }
645 {
646 V T14, T24, T1m, T2b, T17, T25, T1h, T2a;
647 {
648 V T12, T13, T1j, T1l;
649 T12 = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
650 T13 = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
651 T14 = VFMA(T2, T12, VMUL(T5, T13));
652 T24 = VFNMS(T5, T12, VMUL(T2, T13));
653 T1j = LD(&(ri[WS(rs, 13)]), ms, &(ri[WS(rs, 1)]));
654 T1l = LD(&(ii[WS(rs, 13)]), ms, &(ii[WS(rs, 1)]));
655 T1m = VFMA(T1i, T1j, VMUL(T1k, T1l));
656 T2b = VFNMS(T1k, T1j, VMUL(T1i, T1l));
657 }
658 {
659 V T15, T16, T1c, T1g;
660 T15 = LD(&(ri[WS(rs, 9)]), ms, &(ri[WS(rs, 1)]));
661 T16 = LD(&(ii[WS(rs, 9)]), ms, &(ii[WS(rs, 1)]));
662 T17 = VFMA(T3, T15, VMUL(T6, T16));
663 T25 = VFNMS(T6, T15, VMUL(T3, T16));
664 T1c = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
665 T1g = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
666 T1h = VFMA(T1b, T1c, VMUL(T1f, T1g));
667 T2a = VFNMS(T1f, T1c, VMUL(T1b, T1g));
668 }
669 T18 = VADD(T14, T17);
670 T1n = VADD(T1h, T1m);
671 T2Q = VSUB(T18, T1n);
672 T2R = VADD(T24, T25);
673 T2S = VADD(T2a, T2b);
674 T2T = VSUB(T2R, T2S);
675 {
676 V T26, T27, T29, T2c;
677 T26 = VSUB(T24, T25);
678 T27 = VSUB(T1h, T1m);
679 T28 = VADD(T26, T27);
680 T2A = VSUB(T26, T27);
681 T29 = VSUB(T14, T17);
682 T2c = VSUB(T2a, T2b);
683 T2d = VSUB(T29, T2c);
684 T2B = VADD(T29, T2c);
685 }
686 }
687 {
688 V T23, T2r, T3A, T3C, T2q, T3B, T2u, T3x;
689 {
690 V T1R, T22, T3y, T3z;
691 T1R = VSUB(T1N, T1Q);
692 T22 = VMUL(LDK(KP707106781), VSUB(T1W, T21));
693 T23 = VADD(T1R, T22);
694 T2r = VSUB(T1R, T22);
695 T3y = VMUL(LDK(KP707106781), VSUB(T2x, T2w));
696 T3z = VADD(T3s, T3r);
697 T3A = VADD(T3y, T3z);
698 T3C = VSUB(T3z, T3y);
699 }
700 {
701 V T2e, T2p, T2s, T2t;
702 T2e = VFMA(LDK(KP923879532), T28, VMUL(LDK(KP382683432), T2d));
703 T2p = VFNMS(LDK(KP923879532), T2o, VMUL(LDK(KP382683432), T2j));
704 T2q = VADD(T2e, T2p);
705 T3B = VSUB(T2p, T2e);
706 T2s = VFNMS(LDK(KP923879532), T2d, VMUL(LDK(KP382683432), T28));
707 T2t = VFMA(LDK(KP382683432), T2o, VMUL(LDK(KP923879532), T2j));
708 T2u = VSUB(T2s, T2t);
709 T3x = VADD(T2s, T2t);
710 }
711 ST(&(ri[WS(rs, 11)]), VSUB(T23, T2q), ms, &(ri[WS(rs, 1)]));
712 ST(&(ii[WS(rs, 11)]), VSUB(T3A, T3x), ms, &(ii[WS(rs, 1)]));
713 ST(&(ri[WS(rs, 3)]), VADD(T23, T2q), ms, &(ri[WS(rs, 1)]));
714 ST(&(ii[WS(rs, 3)]), VADD(T3x, T3A), ms, &(ii[WS(rs, 1)]));
715 ST(&(ri[WS(rs, 15)]), VSUB(T2r, T2u), ms, &(ri[WS(rs, 1)]));
716 ST(&(ii[WS(rs, 15)]), VSUB(T3C, T3B), ms, &(ii[WS(rs, 1)]));
717 ST(&(ri[WS(rs, 7)]), VADD(T2r, T2u), ms, &(ri[WS(rs, 1)]));
718 ST(&(ii[WS(rs, 7)]), VADD(T3B, T3C), ms, &(ii[WS(rs, 1)]));
719 }
720 {
721 V T2P, T31, T3m, T3o, T30, T3n, T34, T3j;
722 {
723 V T2L, T2O, T3k, T3l;
724 T2L = VSUB(Tf, TA);
725 T2O = VSUB(T2M, T2N);
726 T2P = VADD(T2L, T2O);
727 T31 = VSUB(T2L, T2O);
728 T3k = VSUB(TZ, TM);
729 T3l = VSUB(T3e, T3b);
730 T3m = VADD(T3k, T3l);
731 T3o = VSUB(T3l, T3k);
732 }
733 {
734 V T2U, T2Z, T32, T33;
735 T2U = VADD(T2Q, T2T);
736 T2Z = VSUB(T2V, T2Y);
737 T30 = VMUL(LDK(KP707106781), VADD(T2U, T2Z));
738 T3n = VMUL(LDK(KP707106781), VSUB(T2Z, T2U));
739 T32 = VSUB(T2T, T2Q);
740 T33 = VADD(T2V, T2Y);
741 T34 = VMUL(LDK(KP707106781), VSUB(T32, T33));
742 T3j = VMUL(LDK(KP707106781), VADD(T32, T33));
743 }
744 ST(&(ri[WS(rs, 10)]), VSUB(T2P, T30), ms, &(ri[0]));
745 ST(&(ii[WS(rs, 10)]), VSUB(T3m, T3j), ms, &(ii[0]));
746 ST(&(ri[WS(rs, 2)]), VADD(T2P, T30), ms, &(ri[0]));
747 ST(&(ii[WS(rs, 2)]), VADD(T3j, T3m), ms, &(ii[0]));
748 ST(&(ri[WS(rs, 14)]), VSUB(T31, T34), ms, &(ri[0]));
749 ST(&(ii[WS(rs, 14)]), VSUB(T3o, T3n), ms, &(ii[0]));
750 ST(&(ri[WS(rs, 6)]), VADD(T31, T34), ms, &(ri[0]));
751 ST(&(ii[WS(rs, 6)]), VADD(T3n, T3o), ms, &(ii[0]));
752 }
753 {
754 V T2z, T2H, T3u, T3w, T2G, T3v, T2K, T3p;
755 {
756 V T2v, T2y, T3q, T3t;
757 T2v = VADD(T1N, T1Q);
758 T2y = VMUL(LDK(KP707106781), VADD(T2w, T2x));
759 T2z = VADD(T2v, T2y);
760 T2H = VSUB(T2v, T2y);
761 T3q = VMUL(LDK(KP707106781), VADD(T1W, T21));
762 T3t = VSUB(T3r, T3s);
763 T3u = VADD(T3q, T3t);
764 T3w = VSUB(T3t, T3q);
765 }
766 {
767 V T2C, T2F, T2I, T2J;
768 T2C = VFMA(LDK(KP382683432), T2A, VMUL(LDK(KP923879532), T2B));
769 T2F = VFNMS(LDK(KP382683432), T2E, VMUL(LDK(KP923879532), T2D));
770 T2G = VADD(T2C, T2F);
771 T3v = VSUB(T2F, T2C);
772 T2I = VFNMS(LDK(KP382683432), T2B, VMUL(LDK(KP923879532), T2A));
773 T2J = VFMA(LDK(KP923879532), T2E, VMUL(LDK(KP382683432), T2D));
774 T2K = VSUB(T2I, T2J);
775 T3p = VADD(T2I, T2J);
776 }
777 ST(&(ri[WS(rs, 9)]), VSUB(T2z, T2G), ms, &(ri[WS(rs, 1)]));
778 ST(&(ii[WS(rs, 9)]), VSUB(T3u, T3p), ms, &(ii[WS(rs, 1)]));
779 ST(&(ri[WS(rs, 1)]), VADD(T2z, T2G), ms, &(ri[WS(rs, 1)]));
780 ST(&(ii[WS(rs, 1)]), VADD(T3p, T3u), ms, &(ii[WS(rs, 1)]));
781 ST(&(ri[WS(rs, 13)]), VSUB(T2H, T2K), ms, &(ri[WS(rs, 1)]));
782 ST(&(ii[WS(rs, 13)]), VSUB(T3w, T3v), ms, &(ii[WS(rs, 1)]));
783 ST(&(ri[WS(rs, 5)]), VADD(T2H, T2K), ms, &(ri[WS(rs, 1)]));
784 ST(&(ii[WS(rs, 5)]), VADD(T3v, T3w), ms, &(ii[WS(rs, 1)]));
785 }
786 {
787 V T11, T35, T3g, T3i, T1M, T3h, T38, T39;
788 {
789 V TB, T10, T3a, T3f;
790 TB = VADD(Tf, TA);
791 T10 = VADD(TM, TZ);
792 T11 = VADD(TB, T10);
793 T35 = VSUB(TB, T10);
794 T3a = VADD(T2M, T2N);
795 T3f = VADD(T3b, T3e);
796 T3g = VADD(T3a, T3f);
797 T3i = VSUB(T3f, T3a);
798 }
799 {
800 V T1o, T1L, T36, T37;
801 T1o = VADD(T18, T1n);
802 T1L = VADD(T1B, T1K);
803 T1M = VADD(T1o, T1L);
804 T3h = VSUB(T1L, T1o);
805 T36 = VADD(T2R, T2S);
806 T37 = VADD(T2W, T2X);
807 T38 = VSUB(T36, T37);
808 T39 = VADD(T36, T37);
809 }
810 ST(&(ri[WS(rs, 8)]), VSUB(T11, T1M), ms, &(ri[0]));
811 ST(&(ii[WS(rs, 8)]), VSUB(T3g, T39), ms, &(ii[0]));
812 ST(&(ri[0]), VADD(T11, T1M), ms, &(ri[0]));
813 ST(&(ii[0]), VADD(T39, T3g), ms, &(ii[0]));
814 ST(&(ri[WS(rs, 12)]), VSUB(T35, T38), ms, &(ri[0]));
815 ST(&(ii[WS(rs, 12)]), VSUB(T3i, T3h), ms, &(ii[0]));
816 ST(&(ri[WS(rs, 4)]), VADD(T35, T38), ms, &(ri[0]));
817 ST(&(ii[WS(rs, 4)]), VADD(T3h, T3i), ms, &(ii[0]));
818 }
819 }
820 }
821 }
822 VLEAVE();
823 }
824
825 static const tw_instr twinstr[] = {
826 VTW(0, 1),
827 VTW(0, 3),
828 VTW(0, 9),
829 VTW(0, 15),
830 {TW_NEXT, (2 * VL), 0}
831 };
832
833 static const ct_desc desc = { 16, XSIMD_STRING("t2sv_16"), twinstr, &GENUS, {156, 68, 40, 0}, 0, 0, 0 };
834
835 void XSIMD(codelet_t2sv_16) (planner *p) {
836 X(kdft_dit_register) (p, t2sv_16, &desc);
837 }
838 #endif