Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hf_9.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 08:06:29 EDT 2018 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include rdft/scalar/hf.h */
29
30 /*
31 * This function contains 96 FP additions, 88 FP multiplications,
32 * (or, 24 additions, 16 multiplications, 72 fused multiply/add),
33 * 55 stack variables, 10 constants, and 36 memory accesses
34 */
35 #include "rdft/scalar/hf.h"
36
37 static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP852868531, +0.852868531952443209628250963940074071936020296);
40 DK(KP492403876, +0.492403876506104029683371512294761506835321626);
41 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
42 DK(KP777861913, +0.777861913430206160028177977318626690410586096);
43 DK(KP839099631, +0.839099631177280011763127298123181364687434283);
44 DK(KP954188894, +0.954188894138671133499268364187245676532219158);
45 DK(KP363970234, +0.363970234266202361351047882776834043890471784);
46 DK(KP176326980, +0.176326980708464973471090386868618986121633062);
47 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
48 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
49 {
50 INT m;
51 for (m = mb, W = W + ((mb - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
52 E T1, T1P, Te, T1S, T10, T1Q, T1a, T1d, Ty, T18, Tl, T13, T19, T1c, T1l;
53 E T1r, TS, T1p, TF, T1o, T1g, T1q;
54 T1 = cr[0];
55 T1P = ci[0];
56 {
57 E T3, T6, T4, TW, T9, Tc, Ta, TY, T2, T8;
58 T3 = cr[WS(rs, 3)];
59 T6 = ci[WS(rs, 3)];
60 T2 = W[4];
61 T4 = T2 * T3;
62 TW = T2 * T6;
63 T9 = cr[WS(rs, 6)];
64 Tc = ci[WS(rs, 6)];
65 T8 = W[10];
66 Ta = T8 * T9;
67 TY = T8 * Tc;
68 {
69 E T7, TX, Td, TZ, T5, Tb;
70 T5 = W[5];
71 T7 = FMA(T5, T6, T4);
72 TX = FNMS(T5, T3, TW);
73 Tb = W[11];
74 Td = FMA(Tb, Tc, Ta);
75 TZ = FNMS(Tb, T9, TY);
76 Te = T7 + Td;
77 T1S = Td - T7;
78 T10 = TX - TZ;
79 T1Q = TX + TZ;
80 }
81 }
82 {
83 E Th, Tk, Ti, T12, Tx, T17, Tr, T15, Tg, Tj;
84 Th = cr[WS(rs, 1)];
85 Tk = ci[WS(rs, 1)];
86 Tg = W[0];
87 Ti = Tg * Th;
88 T12 = Tg * Tk;
89 {
90 E Tt, Tw, Tu, T16, Ts, Tv;
91 Tt = cr[WS(rs, 7)];
92 Tw = ci[WS(rs, 7)];
93 Ts = W[12];
94 Tu = Ts * Tt;
95 T16 = Ts * Tw;
96 Tv = W[13];
97 Tx = FMA(Tv, Tw, Tu);
98 T17 = FNMS(Tv, Tt, T16);
99 }
100 {
101 E Tn, Tq, To, T14, Tm, Tp;
102 Tn = cr[WS(rs, 4)];
103 Tq = ci[WS(rs, 4)];
104 Tm = W[6];
105 To = Tm * Tn;
106 T14 = Tm * Tq;
107 Tp = W[7];
108 Tr = FMA(Tp, Tq, To);
109 T15 = FNMS(Tp, Tn, T14);
110 }
111 T1a = Tr - Tx;
112 T1d = T15 - T17;
113 Ty = Tr + Tx;
114 T18 = T15 + T17;
115 Tj = W[1];
116 Tl = FMA(Tj, Tk, Ti);
117 T13 = FNMS(Tj, Th, T12);
118 T19 = FNMS(KP500000000, T18, T13);
119 T1c = FNMS(KP500000000, Ty, Tl);
120 }
121 {
122 E TB, TE, TC, T1n, TR, T1k, TL, T1i, TA, TD;
123 TB = cr[WS(rs, 2)];
124 TE = ci[WS(rs, 2)];
125 TA = W[2];
126 TC = TA * TB;
127 T1n = TA * TE;
128 {
129 E TN, TQ, TO, T1j, TM, TP;
130 TN = cr[WS(rs, 8)];
131 TQ = ci[WS(rs, 8)];
132 TM = W[14];
133 TO = TM * TN;
134 T1j = TM * TQ;
135 TP = W[15];
136 TR = FMA(TP, TQ, TO);
137 T1k = FNMS(TP, TN, T1j);
138 }
139 {
140 E TH, TK, TI, T1h, TG, TJ;
141 TH = cr[WS(rs, 5)];
142 TK = ci[WS(rs, 5)];
143 TG = W[8];
144 TI = TG * TH;
145 T1h = TG * TK;
146 TJ = W[9];
147 TL = FMA(TJ, TK, TI);
148 T1i = FNMS(TJ, TH, T1h);
149 }
150 T1l = T1i - T1k;
151 T1r = TR - TL;
152 TS = TL + TR;
153 T1p = T1i + T1k;
154 TD = W[3];
155 TF = FMA(TD, TE, TC);
156 T1o = FNMS(TD, TB, T1n);
157 T1g = FNMS(KP500000000, TS, TF);
158 T1q = FNMS(KP500000000, T1p, T1o);
159 }
160 {
161 E Tf, T21, TU, T24, T1O, T22, T1L, T23;
162 Tf = T1 + Te;
163 T21 = T1Q + T1P;
164 {
165 E Tz, TT, T1M, T1N;
166 Tz = Tl + Ty;
167 TT = TF + TS;
168 TU = Tz + TT;
169 T24 = TT - Tz;
170 T1M = T13 + T18;
171 T1N = T1o + T1p;
172 T1O = T1M - T1N;
173 T22 = T1M + T1N;
174 }
175 cr[0] = Tf + TU;
176 ci[WS(rs, 8)] = T22 + T21;
177 T1L = FNMS(KP500000000, TU, Tf);
178 ci[WS(rs, 2)] = FNMS(KP866025403, T1O, T1L);
179 cr[WS(rs, 3)] = FMA(KP866025403, T1O, T1L);
180 T23 = FNMS(KP500000000, T22, T21);
181 cr[WS(rs, 6)] = FMS(KP866025403, T24, T23);
182 ci[WS(rs, 5)] = FMA(KP866025403, T24, T23);
183 }
184 {
185 E T11, T1z, T1T, T1X, T1f, T1w, T1t, T1x, T1u, T1Y, T1C, T1I, T1F, T1J, T1G;
186 E T1U, TV, T1R;
187 TV = FNMS(KP500000000, Te, T1);
188 T11 = FNMS(KP866025403, T10, TV);
189 T1z = FMA(KP866025403, T10, TV);
190 T1R = FNMS(KP500000000, T1Q, T1P);
191 T1T = FMA(KP866025403, T1S, T1R);
192 T1X = FNMS(KP866025403, T1S, T1R);
193 {
194 E T1b, T1e, T1m, T1s;
195 T1b = FMA(KP866025403, T1a, T19);
196 T1e = FNMS(KP866025403, T1d, T1c);
197 T1f = FMA(KP176326980, T1e, T1b);
198 T1w = FNMS(KP176326980, T1b, T1e);
199 T1m = FNMS(KP866025403, T1l, T1g);
200 T1s = FNMS(KP866025403, T1r, T1q);
201 T1t = FNMS(KP363970234, T1s, T1m);
202 T1x = FMA(KP363970234, T1m, T1s);
203 }
204 T1u = FNMS(KP954188894, T1t, T1f);
205 T1Y = FMA(KP954188894, T1x, T1w);
206 {
207 E T1A, T1B, T1D, T1E;
208 T1A = FMA(KP866025403, T1r, T1q);
209 T1B = FMA(KP866025403, T1l, T1g);
210 T1C = FMA(KP176326980, T1B, T1A);
211 T1I = FNMS(KP176326980, T1A, T1B);
212 T1D = FMA(KP866025403, T1d, T1c);
213 T1E = FNMS(KP866025403, T1a, T19);
214 T1F = FMA(KP839099631, T1E, T1D);
215 T1J = FNMS(KP839099631, T1D, T1E);
216 }
217 T1G = FMA(KP777861913, T1F, T1C);
218 T1U = FNMS(KP777861913, T1J, T1I);
219 cr[WS(rs, 2)] = FMA(KP984807753, T1u, T11);
220 ci[WS(rs, 7)] = FNMS(KP984807753, T1U, T1T);
221 ci[WS(rs, 6)] = FNMS(KP984807753, T1Y, T1X);
222 cr[WS(rs, 1)] = FMA(KP984807753, T1G, T1z);
223 {
224 E T1V, T1W, T1H, T1K;
225 T1V = FMA(KP492403876, T1U, T1T);
226 T1W = FNMS(KP777861913, T1F, T1C);
227 cr[WS(rs, 7)] = FMS(KP852868531, T1W, T1V);
228 ci[WS(rs, 4)] = FMA(KP852868531, T1W, T1V);
229 T1H = FNMS(KP492403876, T1G, T1z);
230 T1K = FMA(KP777861913, T1J, T1I);
231 ci[WS(rs, 1)] = FNMS(KP852868531, T1K, T1H);
232 cr[WS(rs, 4)] = FMA(KP852868531, T1K, T1H);
233 }
234 {
235 E T1v, T1y, T1Z, T20;
236 T1v = FNMS(KP492403876, T1u, T11);
237 T1y = FNMS(KP954188894, T1x, T1w);
238 ci[WS(rs, 3)] = FNMS(KP852868531, T1y, T1v);
239 ci[0] = FMA(KP852868531, T1y, T1v);
240 T1Z = FMA(KP492403876, T1Y, T1X);
241 T20 = FMA(KP954188894, T1t, T1f);
242 cr[WS(rs, 5)] = FMS(KP852868531, T20, T1Z);
243 cr[WS(rs, 8)] = -(FMA(KP852868531, T20, T1Z));
244 }
245 }
246 }
247 }
248 }
249
250 static const tw_instr twinstr[] = {
251 {TW_FULL, 1, 9},
252 {TW_NEXT, 1, 0}
253 };
254
255 static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {24, 16, 72, 0} };
256
257 void X(codelet_hf_9) (planner *p) {
258 X(khc2hc_register) (p, hf_9, &desc);
259 }
260 #else
261
262 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 9 -dit -name hf_9 -include rdft/scalar/hf.h */
263
264 /*
265 * This function contains 96 FP additions, 72 FP multiplications,
266 * (or, 60 additions, 36 multiplications, 36 fused multiply/add),
267 * 41 stack variables, 8 constants, and 36 memory accesses
268 */
269 #include "rdft/scalar/hf.h"
270
271 static void hf_9(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
272 {
273 DK(KP642787609, +0.642787609686539326322643409907263432907559884);
274 DK(KP766044443, +0.766044443118978035202392650555416673935832457);
275 DK(KP939692620, +0.939692620785908384054109277324731469936208134);
276 DK(KP342020143, +0.342020143325668733044099614682259580763083368);
277 DK(KP984807753, +0.984807753012208059366743024589523013670643252);
278 DK(KP173648177, +0.173648177666930348851716626769314796000375677);
279 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
280 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
281 {
282 INT m;
283 for (m = mb, W = W + ((mb - 1) * 16); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 16, MAKE_VOLATILE_STRIDE(18, rs)) {
284 E T1, T1B, TQ, T1A, Tc, TN, T1C, T1D, TL, T1x, T19, T1o, T1c, T1n, Tu;
285 E T1w, TW, T1k, T11, T1l;
286 {
287 E T6, TO, Tb, TP;
288 T1 = cr[0];
289 T1B = ci[0];
290 {
291 E T3, T5, T2, T4;
292 T3 = cr[WS(rs, 3)];
293 T5 = ci[WS(rs, 3)];
294 T2 = W[4];
295 T4 = W[5];
296 T6 = FMA(T2, T3, T4 * T5);
297 TO = FNMS(T4, T3, T2 * T5);
298 }
299 {
300 E T8, Ta, T7, T9;
301 T8 = cr[WS(rs, 6)];
302 Ta = ci[WS(rs, 6)];
303 T7 = W[10];
304 T9 = W[11];
305 Tb = FMA(T7, T8, T9 * Ta);
306 TP = FNMS(T9, T8, T7 * Ta);
307 }
308 TQ = KP866025403 * (TO - TP);
309 T1A = KP866025403 * (Tb - T6);
310 Tc = T6 + Tb;
311 TN = FNMS(KP500000000, Tc, T1);
312 T1C = TO + TP;
313 T1D = FNMS(KP500000000, T1C, T1B);
314 }
315 {
316 E Tz, T13, TE, T14, TJ, T15, TK, T16;
317 {
318 E Tw, Ty, Tv, Tx;
319 Tw = cr[WS(rs, 2)];
320 Ty = ci[WS(rs, 2)];
321 Tv = W[2];
322 Tx = W[3];
323 Tz = FMA(Tv, Tw, Tx * Ty);
324 T13 = FNMS(Tx, Tw, Tv * Ty);
325 }
326 {
327 E TB, TD, TA, TC;
328 TB = cr[WS(rs, 5)];
329 TD = ci[WS(rs, 5)];
330 TA = W[8];
331 TC = W[9];
332 TE = FMA(TA, TB, TC * TD);
333 T14 = FNMS(TC, TB, TA * TD);
334 }
335 {
336 E TG, TI, TF, TH;
337 TG = cr[WS(rs, 8)];
338 TI = ci[WS(rs, 8)];
339 TF = W[14];
340 TH = W[15];
341 TJ = FMA(TF, TG, TH * TI);
342 T15 = FNMS(TH, TG, TF * TI);
343 }
344 TK = TE + TJ;
345 T16 = T14 + T15;
346 TL = Tz + TK;
347 T1x = T13 + T16;
348 {
349 E T17, T18, T1a, T1b;
350 T17 = FNMS(KP500000000, T16, T13);
351 T18 = KP866025403 * (TJ - TE);
352 T19 = T17 - T18;
353 T1o = T18 + T17;
354 T1a = FNMS(KP500000000, TK, Tz);
355 T1b = KP866025403 * (T14 - T15);
356 T1c = T1a - T1b;
357 T1n = T1a + T1b;
358 }
359 }
360 {
361 E Ti, TX, Tn, TT, Ts, TU, Tt, TY;
362 {
363 E Tf, Th, Te, Tg;
364 Tf = cr[WS(rs, 1)];
365 Th = ci[WS(rs, 1)];
366 Te = W[0];
367 Tg = W[1];
368 Ti = FMA(Te, Tf, Tg * Th);
369 TX = FNMS(Tg, Tf, Te * Th);
370 }
371 {
372 E Tk, Tm, Tj, Tl;
373 Tk = cr[WS(rs, 4)];
374 Tm = ci[WS(rs, 4)];
375 Tj = W[6];
376 Tl = W[7];
377 Tn = FMA(Tj, Tk, Tl * Tm);
378 TT = FNMS(Tl, Tk, Tj * Tm);
379 }
380 {
381 E Tp, Tr, To, Tq;
382 Tp = cr[WS(rs, 7)];
383 Tr = ci[WS(rs, 7)];
384 To = W[12];
385 Tq = W[13];
386 Ts = FMA(To, Tp, Tq * Tr);
387 TU = FNMS(Tq, Tp, To * Tr);
388 }
389 Tt = Tn + Ts;
390 TY = TT + TU;
391 Tu = Ti + Tt;
392 T1w = TX + TY;
393 {
394 E TS, TV, TZ, T10;
395 TS = FNMS(KP500000000, Tt, Ti);
396 TV = KP866025403 * (TT - TU);
397 TW = TS - TV;
398 T1k = TS + TV;
399 TZ = FNMS(KP500000000, TY, TX);
400 T10 = KP866025403 * (Ts - Tn);
401 T11 = TZ - T10;
402 T1l = T10 + TZ;
403 }
404 }
405 {
406 E T1y, Td, TM, T1v;
407 T1y = KP866025403 * (T1w - T1x);
408 Td = T1 + Tc;
409 TM = Tu + TL;
410 T1v = FNMS(KP500000000, TM, Td);
411 cr[0] = Td + TM;
412 cr[WS(rs, 3)] = T1v + T1y;
413 ci[WS(rs, 2)] = T1v - T1y;
414 }
415 {
416 E TR, T1I, T1e, T1K, T1i, T1H, T1f, T1J;
417 TR = TN - TQ;
418 T1I = T1D - T1A;
419 {
420 E T12, T1d, T1g, T1h;
421 T12 = FMA(KP173648177, TW, KP984807753 * T11);
422 T1d = FNMS(KP939692620, T1c, KP342020143 * T19);
423 T1e = T12 + T1d;
424 T1K = KP866025403 * (T1d - T12);
425 T1g = FNMS(KP984807753, TW, KP173648177 * T11);
426 T1h = FMA(KP342020143, T1c, KP939692620 * T19);
427 T1i = KP866025403 * (T1g + T1h);
428 T1H = T1g - T1h;
429 }
430 cr[WS(rs, 2)] = TR + T1e;
431 ci[WS(rs, 6)] = T1H + T1I;
432 T1f = FNMS(KP500000000, T1e, TR);
433 ci[0] = T1f - T1i;
434 ci[WS(rs, 3)] = T1f + T1i;
435 T1J = FMS(KP500000000, T1H, T1I);
436 cr[WS(rs, 5)] = T1J - T1K;
437 cr[WS(rs, 8)] = T1K + T1J;
438 }
439 {
440 E T1L, T1M, T1N, T1O;
441 T1L = KP866025403 * (TL - Tu);
442 T1M = T1C + T1B;
443 T1N = T1w + T1x;
444 T1O = FNMS(KP500000000, T1N, T1M);
445 cr[WS(rs, 6)] = T1L - T1O;
446 ci[WS(rs, 8)] = T1N + T1M;
447 ci[WS(rs, 5)] = T1L + T1O;
448 }
449 {
450 E T1j, T1E, T1q, T1z, T1u, T1F, T1r, T1G;
451 T1j = TN + TQ;
452 T1E = T1A + T1D;
453 {
454 E T1m, T1p, T1s, T1t;
455 T1m = FMA(KP766044443, T1k, KP642787609 * T1l);
456 T1p = FMA(KP173648177, T1n, KP984807753 * T1o);
457 T1q = T1m + T1p;
458 T1z = KP866025403 * (T1p - T1m);
459 T1s = FNMS(KP642787609, T1k, KP766044443 * T1l);
460 T1t = FNMS(KP984807753, T1n, KP173648177 * T1o);
461 T1u = KP866025403 * (T1s - T1t);
462 T1F = T1s + T1t;
463 }
464 cr[WS(rs, 1)] = T1j + T1q;
465 T1r = FNMS(KP500000000, T1q, T1j);
466 ci[WS(rs, 1)] = T1r - T1u;
467 cr[WS(rs, 4)] = T1r + T1u;
468 ci[WS(rs, 7)] = T1F + T1E;
469 T1G = FNMS(KP500000000, T1F, T1E);
470 cr[WS(rs, 7)] = T1z - T1G;
471 ci[WS(rs, 4)] = T1z + T1G;
472 }
473 }
474 }
475 }
476
477 static const tw_instr twinstr[] = {
478 {TW_FULL, 1, 9},
479 {TW_NEXT, 1, 0}
480 };
481
482 static const hc2hc_desc desc = { 9, "hf_9", twinstr, &GENUS, {60, 36, 36, 0} };
483
484 void X(codelet_hf_9) (planner *p) {
485 X(khc2hc_register) (p, hf_9, &desc);
486 }
487 #endif