Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t1_10.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 08:04:14 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 -n 10 -name t1_10 -include dft/scalar/t.h */
29
30 /*
31 * This function contains 102 FP additions, 72 FP multiplications,
32 * (or, 48 additions, 18 multiplications, 54 fused multiply/add),
33 * 47 stack variables, 4 constants, and 40 memory accesses
34 */
35 #include "dft/scalar/t.h"
36
37 static void t1_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 * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
46 E T8, T23, T12, T1U, TM, TZ, T10, T1F, T1G, T1P, T16, T17, T18, T1s, T1x;
47 E T25, Tl, Ty, Tz, T1I, T1J, T1O, T13, T14, T15, T1h, T1m, T24;
48 {
49 E T1, T1T, T3, T6, T4, T1R, T2, T7, T1S, T5;
50 T1 = ri[0];
51 T1T = ii[0];
52 T3 = ri[WS(rs, 5)];
53 T6 = ii[WS(rs, 5)];
54 T2 = W[8];
55 T4 = T2 * T3;
56 T1R = T2 * T6;
57 T5 = W[9];
58 T7 = FMA(T5, T6, T4);
59 T1S = FNMS(T5, T3, T1R);
60 T8 = T1 - T7;
61 T23 = T1T - T1S;
62 T12 = T1 + T7;
63 T1U = T1S + T1T;
64 }
65 {
66 E TF, T1p, TY, T1w, TL, T1r, TS, T1u;
67 {
68 E TB, TE, TC, T1o, TA, TD;
69 TB = ri[WS(rs, 4)];
70 TE = ii[WS(rs, 4)];
71 TA = W[6];
72 TC = TA * TB;
73 T1o = TA * TE;
74 TD = W[7];
75 TF = FMA(TD, TE, TC);
76 T1p = FNMS(TD, TB, T1o);
77 }
78 {
79 E TU, TX, TV, T1v, TT, TW;
80 TU = ri[WS(rs, 1)];
81 TX = ii[WS(rs, 1)];
82 TT = W[0];
83 TV = TT * TU;
84 T1v = TT * TX;
85 TW = W[1];
86 TY = FMA(TW, TX, TV);
87 T1w = FNMS(TW, TU, T1v);
88 }
89 {
90 E TH, TK, TI, T1q, TG, TJ;
91 TH = ri[WS(rs, 9)];
92 TK = ii[WS(rs, 9)];
93 TG = W[16];
94 TI = TG * TH;
95 T1q = TG * TK;
96 TJ = W[17];
97 TL = FMA(TJ, TK, TI);
98 T1r = FNMS(TJ, TH, T1q);
99 }
100 {
101 E TO, TR, TP, T1t, TN, TQ;
102 TO = ri[WS(rs, 6)];
103 TR = ii[WS(rs, 6)];
104 TN = W[10];
105 TP = TN * TO;
106 T1t = TN * TR;
107 TQ = W[11];
108 TS = FMA(TQ, TR, TP);
109 T1u = FNMS(TQ, TO, T1t);
110 }
111 TM = TF - TL;
112 TZ = TS - TY;
113 T10 = TM + TZ;
114 T1F = T1p + T1r;
115 T1G = T1u + T1w;
116 T1P = T1F + T1G;
117 T16 = TF + TL;
118 T17 = TS + TY;
119 T18 = T16 + T17;
120 T1s = T1p - T1r;
121 T1x = T1u - T1w;
122 T25 = T1s + T1x;
123 }
124 {
125 E Te, T1e, Tx, T1l, Tk, T1g, Tr, T1j;
126 {
127 E Ta, Td, Tb, T1d, T9, Tc;
128 Ta = ri[WS(rs, 2)];
129 Td = ii[WS(rs, 2)];
130 T9 = W[2];
131 Tb = T9 * Ta;
132 T1d = T9 * Td;
133 Tc = W[3];
134 Te = FMA(Tc, Td, Tb);
135 T1e = FNMS(Tc, Ta, T1d);
136 }
137 {
138 E Tt, Tw, Tu, T1k, Ts, Tv;
139 Tt = ri[WS(rs, 3)];
140 Tw = ii[WS(rs, 3)];
141 Ts = W[4];
142 Tu = Ts * Tt;
143 T1k = Ts * Tw;
144 Tv = W[5];
145 Tx = FMA(Tv, Tw, Tu);
146 T1l = FNMS(Tv, Tt, T1k);
147 }
148 {
149 E Tg, Tj, Th, T1f, Tf, Ti;
150 Tg = ri[WS(rs, 7)];
151 Tj = ii[WS(rs, 7)];
152 Tf = W[12];
153 Th = Tf * Tg;
154 T1f = Tf * Tj;
155 Ti = W[13];
156 Tk = FMA(Ti, Tj, Th);
157 T1g = FNMS(Ti, Tg, T1f);
158 }
159 {
160 E Tn, Tq, To, T1i, Tm, Tp;
161 Tn = ri[WS(rs, 8)];
162 Tq = ii[WS(rs, 8)];
163 Tm = W[14];
164 To = Tm * Tn;
165 T1i = Tm * Tq;
166 Tp = W[15];
167 Tr = FMA(Tp, Tq, To);
168 T1j = FNMS(Tp, Tn, T1i);
169 }
170 Tl = Te - Tk;
171 Ty = Tr - Tx;
172 Tz = Tl + Ty;
173 T1I = T1e + T1g;
174 T1J = T1j + T1l;
175 T1O = T1I + T1J;
176 T13 = Te + Tk;
177 T14 = Tr + Tx;
178 T15 = T13 + T14;
179 T1h = T1e - T1g;
180 T1m = T1j - T1l;
181 T24 = T1h + T1m;
182 }
183 {
184 E T1b, T11, T1a, T1z, T1B, T1n, T1y, T1A, T1c;
185 T1b = Tz - T10;
186 T11 = Tz + T10;
187 T1a = FNMS(KP250000000, T11, T8);
188 T1n = T1h - T1m;
189 T1y = T1s - T1x;
190 T1z = FMA(KP618033988, T1y, T1n);
191 T1B = FNMS(KP618033988, T1n, T1y);
192 ri[WS(rs, 5)] = T8 + T11;
193 T1A = FNMS(KP559016994, T1b, T1a);
194 ri[WS(rs, 7)] = FNMS(KP951056516, T1B, T1A);
195 ri[WS(rs, 3)] = FMA(KP951056516, T1B, T1A);
196 T1c = FMA(KP559016994, T1b, T1a);
197 ri[WS(rs, 9)] = FNMS(KP951056516, T1z, T1c);
198 ri[WS(rs, 1)] = FMA(KP951056516, T1z, T1c);
199 }
200 {
201 E T28, T26, T27, T2c, T2e, T2a, T2b, T2d, T29;
202 T28 = T24 - T25;
203 T26 = T24 + T25;
204 T27 = FNMS(KP250000000, T26, T23);
205 T2a = Tl - Ty;
206 T2b = TM - TZ;
207 T2c = FMA(KP618033988, T2b, T2a);
208 T2e = FNMS(KP618033988, T2a, T2b);
209 ii[WS(rs, 5)] = T26 + T23;
210 T2d = FNMS(KP559016994, T28, T27);
211 ii[WS(rs, 3)] = FNMS(KP951056516, T2e, T2d);
212 ii[WS(rs, 7)] = FMA(KP951056516, T2e, T2d);
213 T29 = FMA(KP559016994, T28, T27);
214 ii[WS(rs, 1)] = FNMS(KP951056516, T2c, T29);
215 ii[WS(rs, 9)] = FMA(KP951056516, T2c, T29);
216 }
217 {
218 E T1D, T19, T1C, T1L, T1N, T1H, T1K, T1M, T1E;
219 T1D = T15 - T18;
220 T19 = T15 + T18;
221 T1C = FNMS(KP250000000, T19, T12);
222 T1H = T1F - T1G;
223 T1K = T1I - T1J;
224 T1L = FNMS(KP618033988, T1K, T1H);
225 T1N = FMA(KP618033988, T1H, T1K);
226 ri[0] = T12 + T19;
227 T1M = FMA(KP559016994, T1D, T1C);
228 ri[WS(rs, 4)] = FNMS(KP951056516, T1N, T1M);
229 ri[WS(rs, 6)] = FMA(KP951056516, T1N, T1M);
230 T1E = FNMS(KP559016994, T1D, T1C);
231 ri[WS(rs, 2)] = FNMS(KP951056516, T1L, T1E);
232 ri[WS(rs, 8)] = FMA(KP951056516, T1L, T1E);
233 }
234 {
235 E T1W, T1Q, T1V, T20, T22, T1Y, T1Z, T21, T1X;
236 T1W = T1O - T1P;
237 T1Q = T1O + T1P;
238 T1V = FNMS(KP250000000, T1Q, T1U);
239 T1Y = T16 - T17;
240 T1Z = T13 - T14;
241 T20 = FNMS(KP618033988, T1Z, T1Y);
242 T22 = FMA(KP618033988, T1Y, T1Z);
243 ii[0] = T1Q + T1U;
244 T21 = FMA(KP559016994, T1W, T1V);
245 ii[WS(rs, 4)] = FMA(KP951056516, T22, T21);
246 ii[WS(rs, 6)] = FNMS(KP951056516, T22, T21);
247 T1X = FNMS(KP559016994, T1W, T1V);
248 ii[WS(rs, 2)] = FMA(KP951056516, T20, T1X);
249 ii[WS(rs, 8)] = FNMS(KP951056516, T20, T1X);
250 }
251 }
252 }
253 }
254
255 static const tw_instr twinstr[] = {
256 {TW_FULL, 0, 10},
257 {TW_NEXT, 1, 0}
258 };
259
260 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {48, 18, 54, 0}, 0, 0, 0 };
261
262 void X(codelet_t1_10) (planner *p) {
263 X(kdft_dit_register) (p, t1_10, &desc);
264 }
265 #else
266
267 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 10 -name t1_10 -include dft/scalar/t.h */
268
269 /*
270 * This function contains 102 FP additions, 60 FP multiplications,
271 * (or, 72 additions, 30 multiplications, 30 fused multiply/add),
272 * 45 stack variables, 4 constants, and 40 memory accesses
273 */
274 #include "dft/scalar/t.h"
275
276 static void t1_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
277 {
278 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
279 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
280 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
281 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
282 {
283 INT m;
284 for (m = mb, W = W + (mb * 18); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
285 E T7, T1O, TT, T1C, TF, TQ, TR, T1o, T1p, T1y, TX, TY, TZ, T1d, T1g;
286 E T1M, Ti, Tt, Tu, T1r, T1s, T1x, TU, TV, TW, T16, T19, T1L;
287 {
288 E T1, T1B, T6, T1A;
289 T1 = ri[0];
290 T1B = ii[0];
291 {
292 E T3, T5, T2, T4;
293 T3 = ri[WS(rs, 5)];
294 T5 = ii[WS(rs, 5)];
295 T2 = W[8];
296 T4 = W[9];
297 T6 = FMA(T2, T3, T4 * T5);
298 T1A = FNMS(T4, T3, T2 * T5);
299 }
300 T7 = T1 - T6;
301 T1O = T1B - T1A;
302 TT = T1 + T6;
303 T1C = T1A + T1B;
304 }
305 {
306 E Tz, T1b, TP, T1f, TE, T1c, TK, T1e;
307 {
308 E Tw, Ty, Tv, Tx;
309 Tw = ri[WS(rs, 4)];
310 Ty = ii[WS(rs, 4)];
311 Tv = W[6];
312 Tx = W[7];
313 Tz = FMA(Tv, Tw, Tx * Ty);
314 T1b = FNMS(Tx, Tw, Tv * Ty);
315 }
316 {
317 E TM, TO, TL, TN;
318 TM = ri[WS(rs, 1)];
319 TO = ii[WS(rs, 1)];
320 TL = W[0];
321 TN = W[1];
322 TP = FMA(TL, TM, TN * TO);
323 T1f = FNMS(TN, TM, TL * TO);
324 }
325 {
326 E TB, TD, TA, TC;
327 TB = ri[WS(rs, 9)];
328 TD = ii[WS(rs, 9)];
329 TA = W[16];
330 TC = W[17];
331 TE = FMA(TA, TB, TC * TD);
332 T1c = FNMS(TC, TB, TA * TD);
333 }
334 {
335 E TH, TJ, TG, TI;
336 TH = ri[WS(rs, 6)];
337 TJ = ii[WS(rs, 6)];
338 TG = W[10];
339 TI = W[11];
340 TK = FMA(TG, TH, TI * TJ);
341 T1e = FNMS(TI, TH, TG * TJ);
342 }
343 TF = Tz - TE;
344 TQ = TK - TP;
345 TR = TF + TQ;
346 T1o = T1b + T1c;
347 T1p = T1e + T1f;
348 T1y = T1o + T1p;
349 TX = Tz + TE;
350 TY = TK + TP;
351 TZ = TX + TY;
352 T1d = T1b - T1c;
353 T1g = T1e - T1f;
354 T1M = T1d + T1g;
355 }
356 {
357 E Tc, T14, Ts, T18, Th, T15, Tn, T17;
358 {
359 E T9, Tb, T8, Ta;
360 T9 = ri[WS(rs, 2)];
361 Tb = ii[WS(rs, 2)];
362 T8 = W[2];
363 Ta = W[3];
364 Tc = FMA(T8, T9, Ta * Tb);
365 T14 = FNMS(Ta, T9, T8 * Tb);
366 }
367 {
368 E Tp, Tr, To, Tq;
369 Tp = ri[WS(rs, 3)];
370 Tr = ii[WS(rs, 3)];
371 To = W[4];
372 Tq = W[5];
373 Ts = FMA(To, Tp, Tq * Tr);
374 T18 = FNMS(Tq, Tp, To * Tr);
375 }
376 {
377 E Te, Tg, Td, Tf;
378 Te = ri[WS(rs, 7)];
379 Tg = ii[WS(rs, 7)];
380 Td = W[12];
381 Tf = W[13];
382 Th = FMA(Td, Te, Tf * Tg);
383 T15 = FNMS(Tf, Te, Td * Tg);
384 }
385 {
386 E Tk, Tm, Tj, Tl;
387 Tk = ri[WS(rs, 8)];
388 Tm = ii[WS(rs, 8)];
389 Tj = W[14];
390 Tl = W[15];
391 Tn = FMA(Tj, Tk, Tl * Tm);
392 T17 = FNMS(Tl, Tk, Tj * Tm);
393 }
394 Ti = Tc - Th;
395 Tt = Tn - Ts;
396 Tu = Ti + Tt;
397 T1r = T14 + T15;
398 T1s = T17 + T18;
399 T1x = T1r + T1s;
400 TU = Tc + Th;
401 TV = Tn + Ts;
402 TW = TU + TV;
403 T16 = T14 - T15;
404 T19 = T17 - T18;
405 T1L = T16 + T19;
406 }
407 {
408 E T11, TS, T12, T1i, T1k, T1a, T1h, T1j, T13;
409 T11 = KP559016994 * (Tu - TR);
410 TS = Tu + TR;
411 T12 = FNMS(KP250000000, TS, T7);
412 T1a = T16 - T19;
413 T1h = T1d - T1g;
414 T1i = FMA(KP951056516, T1a, KP587785252 * T1h);
415 T1k = FNMS(KP587785252, T1a, KP951056516 * T1h);
416 ri[WS(rs, 5)] = T7 + TS;
417 T1j = T12 - T11;
418 ri[WS(rs, 7)] = T1j - T1k;
419 ri[WS(rs, 3)] = T1j + T1k;
420 T13 = T11 + T12;
421 ri[WS(rs, 9)] = T13 - T1i;
422 ri[WS(rs, 1)] = T13 + T1i;
423 }
424 {
425 E T1N, T1P, T1Q, T1U, T1W, T1S, T1T, T1V, T1R;
426 T1N = KP559016994 * (T1L - T1M);
427 T1P = T1L + T1M;
428 T1Q = FNMS(KP250000000, T1P, T1O);
429 T1S = Ti - Tt;
430 T1T = TF - TQ;
431 T1U = FMA(KP951056516, T1S, KP587785252 * T1T);
432 T1W = FNMS(KP587785252, T1S, KP951056516 * T1T);
433 ii[WS(rs, 5)] = T1P + T1O;
434 T1V = T1Q - T1N;
435 ii[WS(rs, 3)] = T1V - T1W;
436 ii[WS(rs, 7)] = T1W + T1V;
437 T1R = T1N + T1Q;
438 ii[WS(rs, 1)] = T1R - T1U;
439 ii[WS(rs, 9)] = T1U + T1R;
440 }
441 {
442 E T1m, T10, T1l, T1u, T1w, T1q, T1t, T1v, T1n;
443 T1m = KP559016994 * (TW - TZ);
444 T10 = TW + TZ;
445 T1l = FNMS(KP250000000, T10, TT);
446 T1q = T1o - T1p;
447 T1t = T1r - T1s;
448 T1u = FNMS(KP587785252, T1t, KP951056516 * T1q);
449 T1w = FMA(KP951056516, T1t, KP587785252 * T1q);
450 ri[0] = TT + T10;
451 T1v = T1m + T1l;
452 ri[WS(rs, 4)] = T1v - T1w;
453 ri[WS(rs, 6)] = T1v + T1w;
454 T1n = T1l - T1m;
455 ri[WS(rs, 2)] = T1n - T1u;
456 ri[WS(rs, 8)] = T1n + T1u;
457 }
458 {
459 E T1H, T1z, T1G, T1F, T1J, T1D, T1E, T1K, T1I;
460 T1H = KP559016994 * (T1x - T1y);
461 T1z = T1x + T1y;
462 T1G = FNMS(KP250000000, T1z, T1C);
463 T1D = TX - TY;
464 T1E = TU - TV;
465 T1F = FNMS(KP587785252, T1E, KP951056516 * T1D);
466 T1J = FMA(KP951056516, T1E, KP587785252 * T1D);
467 ii[0] = T1z + T1C;
468 T1K = T1H + T1G;
469 ii[WS(rs, 4)] = T1J + T1K;
470 ii[WS(rs, 6)] = T1K - T1J;
471 T1I = T1G - T1H;
472 ii[WS(rs, 2)] = T1F + T1I;
473 ii[WS(rs, 8)] = T1I - T1F;
474 }
475 }
476 }
477 }
478
479 static const tw_instr twinstr[] = {
480 {TW_FULL, 0, 10},
481 {TW_NEXT, 1, 0}
482 };
483
484 static const ct_desc desc = { 10, "t1_10", twinstr, &GENUS, {72, 30, 30, 0}, 0, 0, 0 };
485
486 void X(codelet_t1_10) (planner *p) {
487 X(kdft_dit_register) (p, t1_10, &desc);
488 }
489 #endif