Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t2_10.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 08:04:25 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 -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include dft/scalar/t.h */
29
30 /*
31 * This function contains 114 FP additions, 94 FP multiplications,
32 * (or, 48 additions, 28 multiplications, 66 fused multiply/add),
33 * 63 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "dft/scalar/t.h"
36
37 static void t2_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
43 {
44 INT m;
45 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) {
46 E T2, T3, T8, Tc, T5, T6, Tl, T7, TB, TF, T12, TY, To, Ts, Tw;
47 E Tb, Td, Th;
48 {
49 E TA, TX, TE, T11, Ta, T4;
50 T2 = W[0];
51 T3 = W[2];
52 T4 = T2 * T3;
53 T8 = W[4];
54 TA = T2 * T8;
55 TX = T3 * T8;
56 Tc = W[5];
57 TE = T2 * Tc;
58 T11 = T3 * Tc;
59 T5 = W[1];
60 T6 = W[3];
61 Ta = T2 * T6;
62 Tl = FMA(T5, T6, T4);
63 T7 = FNMS(T5, T6, T4);
64 TB = FMA(T5, Tc, TA);
65 TF = FNMS(T5, T8, TE);
66 T12 = FNMS(T6, T8, T11);
67 TY = FMA(T6, Tc, TX);
68 {
69 E Tr, Tv, T9, Tg;
70 Tr = Tl * T8;
71 Tv = Tl * Tc;
72 To = FNMS(T5, T3, Ta);
73 Ts = FMA(To, Tc, Tr);
74 Tw = FNMS(To, T8, Tv);
75 T9 = T7 * T8;
76 Tg = T7 * Tc;
77 Tb = FMA(T5, T3, Ta);
78 Td = FMA(Tb, Tc, T9);
79 Th = FNMS(Tb, T8, Tg);
80 }
81 }
82 {
83 E Tk, T1c, T24, T2d, TW, T19, T1a, T1P, T1Q, T1Z, T1g, T1h, T1i, T1C, T1H;
84 E T2f, Tz, TM, TN, T1S, T1T, T1Y, T1d, T1e, T1f, T1r, T1w, T2e;
85 {
86 E T1, T23, Te, Tf, Ti, T21, Tj, T22;
87 T1 = ri[0];
88 T23 = ii[0];
89 Te = ri[WS(rs, 5)];
90 Tf = Td * Te;
91 Ti = ii[WS(rs, 5)];
92 T21 = Td * Ti;
93 Tj = FMA(Th, Ti, Tf);
94 Tk = T1 - Tj;
95 T1c = T1 + Tj;
96 T22 = FNMS(Th, Te, T21);
97 T24 = T22 + T23;
98 T2d = T23 - T22;
99 }
100 {
101 E TR, T1z, T18, T1G, TV, T1B, T14, T1E;
102 {
103 E TO, TP, TQ, T1y;
104 TO = ri[WS(rs, 4)];
105 TP = T7 * TO;
106 TQ = ii[WS(rs, 4)];
107 T1y = T7 * TQ;
108 TR = FMA(Tb, TQ, TP);
109 T1z = FNMS(Tb, TO, T1y);
110 }
111 {
112 E T15, T16, T17, T1F;
113 T15 = ri[WS(rs, 1)];
114 T16 = T2 * T15;
115 T17 = ii[WS(rs, 1)];
116 T1F = T2 * T17;
117 T18 = FMA(T5, T17, T16);
118 T1G = FNMS(T5, T15, T1F);
119 }
120 {
121 E TS, TT, TU, T1A;
122 TS = ri[WS(rs, 9)];
123 TT = T8 * TS;
124 TU = ii[WS(rs, 9)];
125 T1A = T8 * TU;
126 TV = FMA(Tc, TU, TT);
127 T1B = FNMS(Tc, TS, T1A);
128 }
129 {
130 E TZ, T10, T13, T1D;
131 TZ = ri[WS(rs, 6)];
132 T10 = TY * TZ;
133 T13 = ii[WS(rs, 6)];
134 T1D = TY * T13;
135 T14 = FMA(T12, T13, T10);
136 T1E = FNMS(T12, TZ, T1D);
137 }
138 TW = TR - TV;
139 T19 = T14 - T18;
140 T1a = TW + T19;
141 T1P = T1z + T1B;
142 T1Q = T1E + T1G;
143 T1Z = T1P + T1Q;
144 T1g = TR + TV;
145 T1h = T14 + T18;
146 T1i = T1g + T1h;
147 T1C = T1z - T1B;
148 T1H = T1E - T1G;
149 T2f = T1C + T1H;
150 }
151 {
152 E Tq, T1o, TL, T1v, Ty, T1q, TH, T1t;
153 {
154 E Tm, Tn, Tp, T1n;
155 Tm = ri[WS(rs, 2)];
156 Tn = Tl * Tm;
157 Tp = ii[WS(rs, 2)];
158 T1n = Tl * Tp;
159 Tq = FMA(To, Tp, Tn);
160 T1o = FNMS(To, Tm, T1n);
161 }
162 {
163 E TI, TJ, TK, T1u;
164 TI = ri[WS(rs, 3)];
165 TJ = T3 * TI;
166 TK = ii[WS(rs, 3)];
167 T1u = T3 * TK;
168 TL = FMA(T6, TK, TJ);
169 T1v = FNMS(T6, TI, T1u);
170 }
171 {
172 E Tt, Tu, Tx, T1p;
173 Tt = ri[WS(rs, 7)];
174 Tu = Ts * Tt;
175 Tx = ii[WS(rs, 7)];
176 T1p = Ts * Tx;
177 Ty = FMA(Tw, Tx, Tu);
178 T1q = FNMS(Tw, Tt, T1p);
179 }
180 {
181 E TC, TD, TG, T1s;
182 TC = ri[WS(rs, 8)];
183 TD = TB * TC;
184 TG = ii[WS(rs, 8)];
185 T1s = TB * TG;
186 TH = FMA(TF, TG, TD);
187 T1t = FNMS(TF, TC, T1s);
188 }
189 Tz = Tq - Ty;
190 TM = TH - TL;
191 TN = Tz + TM;
192 T1S = T1o + T1q;
193 T1T = T1t + T1v;
194 T1Y = T1S + T1T;
195 T1d = Tq + Ty;
196 T1e = TH + TL;
197 T1f = T1d + T1e;
198 T1r = T1o - T1q;
199 T1w = T1t - T1v;
200 T2e = T1r + T1w;
201 }
202 {
203 E T1l, T1b, T1k, T1J, T1L, T1x, T1I, T1K, T1m;
204 T1l = TN - T1a;
205 T1b = TN + T1a;
206 T1k = FNMS(KP250000000, T1b, Tk);
207 T1x = T1r - T1w;
208 T1I = T1C - T1H;
209 T1J = FMA(KP618033988, T1I, T1x);
210 T1L = FNMS(KP618033988, T1x, T1I);
211 ri[WS(rs, 5)] = Tk + T1b;
212 T1K = FNMS(KP559016994, T1l, T1k);
213 ri[WS(rs, 7)] = FNMS(KP951056516, T1L, T1K);
214 ri[WS(rs, 3)] = FMA(KP951056516, T1L, T1K);
215 T1m = FMA(KP559016994, T1l, T1k);
216 ri[WS(rs, 9)] = FNMS(KP951056516, T1J, T1m);
217 ri[WS(rs, 1)] = FMA(KP951056516, T1J, T1m);
218 }
219 {
220 E T2i, T2g, T2h, T2m, T2o, T2k, T2l, T2n, T2j;
221 T2i = T2e - T2f;
222 T2g = T2e + T2f;
223 T2h = FNMS(KP250000000, T2g, T2d);
224 T2k = Tz - TM;
225 T2l = TW - T19;
226 T2m = FMA(KP618033988, T2l, T2k);
227 T2o = FNMS(KP618033988, T2k, T2l);
228 ii[WS(rs, 5)] = T2g + T2d;
229 T2n = FNMS(KP559016994, T2i, T2h);
230 ii[WS(rs, 3)] = FNMS(KP951056516, T2o, T2n);
231 ii[WS(rs, 7)] = FMA(KP951056516, T2o, T2n);
232 T2j = FMA(KP559016994, T2i, T2h);
233 ii[WS(rs, 1)] = FNMS(KP951056516, T2m, T2j);
234 ii[WS(rs, 9)] = FMA(KP951056516, T2m, T2j);
235 }
236 {
237 E T1N, T1j, T1M, T1V, T1X, T1R, T1U, T1W, T1O;
238 T1N = T1f - T1i;
239 T1j = T1f + T1i;
240 T1M = FNMS(KP250000000, T1j, T1c);
241 T1R = T1P - T1Q;
242 T1U = T1S - T1T;
243 T1V = FNMS(KP618033988, T1U, T1R);
244 T1X = FMA(KP618033988, T1R, T1U);
245 ri[0] = T1c + T1j;
246 T1W = FMA(KP559016994, T1N, T1M);
247 ri[WS(rs, 4)] = FNMS(KP951056516, T1X, T1W);
248 ri[WS(rs, 6)] = FMA(KP951056516, T1X, T1W);
249 T1O = FNMS(KP559016994, T1N, T1M);
250 ri[WS(rs, 2)] = FNMS(KP951056516, T1V, T1O);
251 ri[WS(rs, 8)] = FMA(KP951056516, T1V, T1O);
252 }
253 {
254 E T26, T20, T25, T2a, T2c, T28, T29, T2b, T27;
255 T26 = T1Y - T1Z;
256 T20 = T1Y + T1Z;
257 T25 = FNMS(KP250000000, T20, T24);
258 T28 = T1g - T1h;
259 T29 = T1d - T1e;
260 T2a = FNMS(KP618033988, T29, T28);
261 T2c = FMA(KP618033988, T28, T29);
262 ii[0] = T20 + T24;
263 T2b = FMA(KP559016994, T26, T25);
264 ii[WS(rs, 4)] = FMA(KP951056516, T2c, T2b);
265 ii[WS(rs, 6)] = FNMS(KP951056516, T2c, T2b);
266 T27 = FNMS(KP559016994, T26, T25);
267 ii[WS(rs, 2)] = FMA(KP951056516, T2a, T27);
268 ii[WS(rs, 8)] = FNMS(KP951056516, T2a, T27);
269 }
270 }
271 }
272 }
273 }
274
275 static const tw_instr twinstr[] = {
276 {TW_CEXP, 0, 1},
277 {TW_CEXP, 0, 3},
278 {TW_CEXP, 0, 9},
279 {TW_NEXT, 1, 0}
280 };
281
282 static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {48, 28, 66, 0}, 0, 0, 0 };
283
284 void X(codelet_t2_10) (planner *p) {
285 X(kdft_dit_register) (p, t2_10, &desc);
286 }
287 #else
288
289 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 10 -name t2_10 -include dft/scalar/t.h */
290
291 /*
292 * This function contains 114 FP additions, 80 FP multiplications,
293 * (or, 76 additions, 42 multiplications, 38 fused multiply/add),
294 * 63 stack variables, 4 constants, and 40 memory accesses
295 */
296 #include "dft/scalar/t.h"
297
298 static void t2_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
299 {
300 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
301 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
302 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
303 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
304 {
305 INT m;
306 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(20, rs)) {
307 E T2, T5, T3, T6, T8, Tm, Tc, Tk, T9, Td, Te, TM, TO, Tg, Tp;
308 E Tv, Tx, Tr;
309 {
310 E T4, Tb, T7, Ta;
311 T2 = W[0];
312 T5 = W[1];
313 T3 = W[2];
314 T6 = W[3];
315 T4 = T2 * T3;
316 Tb = T5 * T3;
317 T7 = T5 * T6;
318 Ta = T2 * T6;
319 T8 = T4 - T7;
320 Tm = Ta - Tb;
321 Tc = Ta + Tb;
322 Tk = T4 + T7;
323 T9 = W[4];
324 Td = W[5];
325 Te = FMA(T8, T9, Tc * Td);
326 TM = FMA(T3, T9, T6 * Td);
327 TO = FNMS(T6, T9, T3 * Td);
328 Tg = FNMS(Tc, T9, T8 * Td);
329 Tp = FMA(Tk, T9, Tm * Td);
330 Tv = FMA(T2, T9, T5 * Td);
331 Tx = FNMS(T5, T9, T2 * Td);
332 Tr = FNMS(Tm, T9, Tk * Td);
333 }
334 {
335 E Tj, T1S, TX, T1G, TL, TU, TV, T1s, T1t, T1C, T11, T12, T13, T1h, T1k;
336 E T1Q, Tu, TD, TE, T1v, T1w, T1B, TY, TZ, T10, T1a, T1d, T1P;
337 {
338 E T1, T1F, Ti, T1E, Tf, Th;
339 T1 = ri[0];
340 T1F = ii[0];
341 Tf = ri[WS(rs, 5)];
342 Th = ii[WS(rs, 5)];
343 Ti = FMA(Te, Tf, Tg * Th);
344 T1E = FNMS(Tg, Tf, Te * Th);
345 Tj = T1 - Ti;
346 T1S = T1F - T1E;
347 TX = T1 + Ti;
348 T1G = T1E + T1F;
349 }
350 {
351 E TH, T1f, TT, T1j, TK, T1g, TQ, T1i;
352 {
353 E TF, TG, TR, TS;
354 TF = ri[WS(rs, 4)];
355 TG = ii[WS(rs, 4)];
356 TH = FMA(T8, TF, Tc * TG);
357 T1f = FNMS(Tc, TF, T8 * TG);
358 TR = ri[WS(rs, 1)];
359 TS = ii[WS(rs, 1)];
360 TT = FMA(T2, TR, T5 * TS);
361 T1j = FNMS(T5, TR, T2 * TS);
362 }
363 {
364 E TI, TJ, TN, TP;
365 TI = ri[WS(rs, 9)];
366 TJ = ii[WS(rs, 9)];
367 TK = FMA(T9, TI, Td * TJ);
368 T1g = FNMS(Td, TI, T9 * TJ);
369 TN = ri[WS(rs, 6)];
370 TP = ii[WS(rs, 6)];
371 TQ = FMA(TM, TN, TO * TP);
372 T1i = FNMS(TO, TN, TM * TP);
373 }
374 TL = TH - TK;
375 TU = TQ - TT;
376 TV = TL + TU;
377 T1s = T1f + T1g;
378 T1t = T1i + T1j;
379 T1C = T1s + T1t;
380 T11 = TH + TK;
381 T12 = TQ + TT;
382 T13 = T11 + T12;
383 T1h = T1f - T1g;
384 T1k = T1i - T1j;
385 T1Q = T1h + T1k;
386 }
387 {
388 E To, T18, TC, T1c, Tt, T19, Tz, T1b;
389 {
390 E Tl, Tn, TA, TB;
391 Tl = ri[WS(rs, 2)];
392 Tn = ii[WS(rs, 2)];
393 To = FMA(Tk, Tl, Tm * Tn);
394 T18 = FNMS(Tm, Tl, Tk * Tn);
395 TA = ri[WS(rs, 3)];
396 TB = ii[WS(rs, 3)];
397 TC = FMA(T3, TA, T6 * TB);
398 T1c = FNMS(T6, TA, T3 * TB);
399 }
400 {
401 E Tq, Ts, Tw, Ty;
402 Tq = ri[WS(rs, 7)];
403 Ts = ii[WS(rs, 7)];
404 Tt = FMA(Tp, Tq, Tr * Ts);
405 T19 = FNMS(Tr, Tq, Tp * Ts);
406 Tw = ri[WS(rs, 8)];
407 Ty = ii[WS(rs, 8)];
408 Tz = FMA(Tv, Tw, Tx * Ty);
409 T1b = FNMS(Tx, Tw, Tv * Ty);
410 }
411 Tu = To - Tt;
412 TD = Tz - TC;
413 TE = Tu + TD;
414 T1v = T18 + T19;
415 T1w = T1b + T1c;
416 T1B = T1v + T1w;
417 TY = To + Tt;
418 TZ = Tz + TC;
419 T10 = TY + TZ;
420 T1a = T18 - T19;
421 T1d = T1b - T1c;
422 T1P = T1a + T1d;
423 }
424 {
425 E T15, TW, T16, T1m, T1o, T1e, T1l, T1n, T17;
426 T15 = KP559016994 * (TE - TV);
427 TW = TE + TV;
428 T16 = FNMS(KP250000000, TW, Tj);
429 T1e = T1a - T1d;
430 T1l = T1h - T1k;
431 T1m = FMA(KP951056516, T1e, KP587785252 * T1l);
432 T1o = FNMS(KP587785252, T1e, KP951056516 * T1l);
433 ri[WS(rs, 5)] = Tj + TW;
434 T1n = T16 - T15;
435 ri[WS(rs, 7)] = T1n - T1o;
436 ri[WS(rs, 3)] = T1n + T1o;
437 T17 = T15 + T16;
438 ri[WS(rs, 9)] = T17 - T1m;
439 ri[WS(rs, 1)] = T17 + T1m;
440 }
441 {
442 E T1R, T1T, T1U, T1Y, T20, T1W, T1X, T1Z, T1V;
443 T1R = KP559016994 * (T1P - T1Q);
444 T1T = T1P + T1Q;
445 T1U = FNMS(KP250000000, T1T, T1S);
446 T1W = Tu - TD;
447 T1X = TL - TU;
448 T1Y = FMA(KP951056516, T1W, KP587785252 * T1X);
449 T20 = FNMS(KP587785252, T1W, KP951056516 * T1X);
450 ii[WS(rs, 5)] = T1T + T1S;
451 T1Z = T1U - T1R;
452 ii[WS(rs, 3)] = T1Z - T20;
453 ii[WS(rs, 7)] = T20 + T1Z;
454 T1V = T1R + T1U;
455 ii[WS(rs, 1)] = T1V - T1Y;
456 ii[WS(rs, 9)] = T1Y + T1V;
457 }
458 {
459 E T1q, T14, T1p, T1y, T1A, T1u, T1x, T1z, T1r;
460 T1q = KP559016994 * (T10 - T13);
461 T14 = T10 + T13;
462 T1p = FNMS(KP250000000, T14, TX);
463 T1u = T1s - T1t;
464 T1x = T1v - T1w;
465 T1y = FNMS(KP587785252, T1x, KP951056516 * T1u);
466 T1A = FMA(KP951056516, T1x, KP587785252 * T1u);
467 ri[0] = TX + T14;
468 T1z = T1q + T1p;
469 ri[WS(rs, 4)] = T1z - T1A;
470 ri[WS(rs, 6)] = T1z + T1A;
471 T1r = T1p - T1q;
472 ri[WS(rs, 2)] = T1r - T1y;
473 ri[WS(rs, 8)] = T1r + T1y;
474 }
475 {
476 E T1L, T1D, T1K, T1J, T1N, T1H, T1I, T1O, T1M;
477 T1L = KP559016994 * (T1B - T1C);
478 T1D = T1B + T1C;
479 T1K = FNMS(KP250000000, T1D, T1G);
480 T1H = T11 - T12;
481 T1I = TY - TZ;
482 T1J = FNMS(KP587785252, T1I, KP951056516 * T1H);
483 T1N = FMA(KP951056516, T1I, KP587785252 * T1H);
484 ii[0] = T1D + T1G;
485 T1O = T1L + T1K;
486 ii[WS(rs, 4)] = T1N + T1O;
487 ii[WS(rs, 6)] = T1O - T1N;
488 T1M = T1K - T1L;
489 ii[WS(rs, 2)] = T1J + T1M;
490 ii[WS(rs, 8)] = T1M - T1J;
491 }
492 }
493 }
494 }
495 }
496
497 static const tw_instr twinstr[] = {
498 {TW_CEXP, 0, 1},
499 {TW_CEXP, 0, 3},
500 {TW_CEXP, 0, 9},
501 {TW_NEXT, 1, 0}
502 };
503
504 static const ct_desc desc = { 10, "t2_10", twinstr, &GENUS, {76, 42, 38, 0}, 0, 0, 0 };
505
506 void X(codelet_t2_10) (planner *p) {
507 X(kdft_dit_register) (p, t2_10, &desc);
508 }
509 #endif