Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cb/hb_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 08:07:32 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 -sign 1 -n 10 -dif -name hb_10 -include rdft/scalar/hb.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 "rdft/scalar/hb.h"
36
37 static void hb_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT m;
45 for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
46 E TH, T1B, TB, T11, T1E, T1G, TK, TM, T1x, T1V, T3, T1g, Tl, T1I, T1J;
47 E TO, TP, T1p, Ti, Tk, T1n, T1o, TF, TG;
48 TF = ci[WS(rs, 9)];
49 TG = cr[WS(rs, 5)];
50 TH = TF - TG;
51 T1B = TF + TG;
52 {
53 E Tp, T1u, Tz, T1s, Ts, T1v, Tw, T1r;
54 {
55 E Tn, To, Tx, Ty;
56 Tn = ci[WS(rs, 5)];
57 To = cr[WS(rs, 9)];
58 Tp = Tn - To;
59 T1u = Tn + To;
60 Tx = ci[WS(rs, 6)];
61 Ty = cr[WS(rs, 8)];
62 Tz = Tx - Ty;
63 T1s = Tx + Ty;
64 }
65 {
66 E Tq, Tr, Tu, Tv;
67 Tq = ci[WS(rs, 8)];
68 Tr = cr[WS(rs, 6)];
69 Ts = Tq - Tr;
70 T1v = Tq + Tr;
71 Tu = ci[WS(rs, 7)];
72 Tv = cr[WS(rs, 7)];
73 Tw = Tu - Tv;
74 T1r = Tu + Tv;
75 }
76 {
77 E Tt, TA, T1C, T1D;
78 Tt = Tp - Ts;
79 TA = Tw - Tz;
80 TB = FNMS(KP618033988, TA, Tt);
81 T11 = FMA(KP618033988, Tt, TA);
82 T1C = T1r - T1s;
83 T1D = T1u - T1v;
84 T1E = T1C + T1D;
85 T1G = T1C - T1D;
86 }
87 {
88 E TI, TJ, T1t, T1w;
89 TI = Tw + Tz;
90 TJ = Tp + Ts;
91 TK = TI + TJ;
92 TM = TI - TJ;
93 T1t = T1r + T1s;
94 T1w = T1u + T1v;
95 T1x = FMA(KP618033988, T1w, T1t);
96 T1V = FNMS(KP618033988, T1t, T1w);
97 }
98 }
99 {
100 E Td, T1k, Tg, T1l, Th, T1m, T6, T1h, T9, T1i, Ta, T1j, T1, T2;
101 T1 = cr[0];
102 T2 = ci[WS(rs, 4)];
103 T3 = T1 + T2;
104 T1g = T1 - T2;
105 {
106 E Tb, Tc, Te, Tf;
107 Tb = cr[WS(rs, 4)];
108 Tc = ci[0];
109 Td = Tb + Tc;
110 T1k = Tb - Tc;
111 Te = ci[WS(rs, 3)];
112 Tf = cr[WS(rs, 1)];
113 Tg = Te + Tf;
114 T1l = Te - Tf;
115 }
116 Th = Td + Tg;
117 T1m = T1k + T1l;
118 {
119 E T4, T5, T7, T8;
120 T4 = cr[WS(rs, 2)];
121 T5 = ci[WS(rs, 2)];
122 T6 = T4 + T5;
123 T1h = T4 - T5;
124 T7 = ci[WS(rs, 1)];
125 T8 = cr[WS(rs, 3)];
126 T9 = T7 + T8;
127 T1i = T7 - T8;
128 }
129 Ta = T6 + T9;
130 T1j = T1h + T1i;
131 Tl = Ta - Th;
132 T1I = T1h - T1i;
133 T1J = T1k - T1l;
134 TO = Td - Tg;
135 TP = T6 - T9;
136 T1p = T1j - T1m;
137 Ti = Ta + Th;
138 Tk = FNMS(KP250000000, Ti, T3);
139 T1n = T1j + T1m;
140 T1o = FNMS(KP250000000, T1n, T1g);
141 }
142 cr[0] = T3 + Ti;
143 ci[0] = TH + TK;
144 {
145 E T2d, T29, T2b, T2c, T2e, T2a;
146 T2d = T1B + T1E;
147 T2a = T1g + T1n;
148 T29 = W[8];
149 T2b = T29 * T2a;
150 T2c = W[9];
151 T2e = T2c * T2a;
152 cr[WS(rs, 5)] = FNMS(T2c, T2d, T2b);
153 ci[WS(rs, 5)] = FMA(T29, T2d, T2e);
154 }
155 {
156 E TQ, T16, TC, TU, TN, T15, T12, T1a, Tm, TL, T10;
157 TQ = FNMS(KP618033988, TP, TO);
158 T16 = FMA(KP618033988, TO, TP);
159 Tm = FNMS(KP559016994, Tl, Tk);
160 TC = FMA(KP951056516, TB, Tm);
161 TU = FNMS(KP951056516, TB, Tm);
162 TL = FNMS(KP250000000, TK, TH);
163 TN = FNMS(KP559016994, TM, TL);
164 T15 = FMA(KP559016994, TM, TL);
165 T10 = FMA(KP559016994, Tl, Tk);
166 T12 = FMA(KP951056516, T11, T10);
167 T1a = FNMS(KP951056516, T11, T10);
168 {
169 E TR, TE, TS, Tj, TD;
170 TR = FNMS(KP951056516, TQ, TN);
171 TE = W[3];
172 TS = TE * TC;
173 Tj = W[2];
174 TD = Tj * TC;
175 cr[WS(rs, 2)] = FNMS(TE, TR, TD);
176 ci[WS(rs, 2)] = FMA(Tj, TR, TS);
177 }
178 {
179 E T1d, T1c, T1e, T19, T1b;
180 T1d = FMA(KP951056516, T16, T15);
181 T1c = W[11];
182 T1e = T1c * T1a;
183 T19 = W[10];
184 T1b = T19 * T1a;
185 cr[WS(rs, 6)] = FNMS(T1c, T1d, T1b);
186 ci[WS(rs, 6)] = FMA(T19, T1d, T1e);
187 }
188 {
189 E TX, TW, TY, TT, TV;
190 TX = FMA(KP951056516, TQ, TN);
191 TW = W[15];
192 TY = TW * TU;
193 TT = W[14];
194 TV = TT * TU;
195 cr[WS(rs, 8)] = FNMS(TW, TX, TV);
196 ci[WS(rs, 8)] = FMA(TT, TX, TY);
197 }
198 {
199 E T17, T14, T18, TZ, T13;
200 T17 = FNMS(KP951056516, T16, T15);
201 T14 = W[7];
202 T18 = T14 * T12;
203 TZ = W[6];
204 T13 = TZ * T12;
205 cr[WS(rs, 4)] = FNMS(T14, T17, T13);
206 ci[WS(rs, 4)] = FMA(TZ, T17, T18);
207 }
208 }
209 {
210 E T1K, T20, T1y, T1O, T1H, T1Z, T1W, T24, T1q, T1F, T1U;
211 T1K = FMA(KP618033988, T1J, T1I);
212 T20 = FNMS(KP618033988, T1I, T1J);
213 T1q = FMA(KP559016994, T1p, T1o);
214 T1y = FNMS(KP951056516, T1x, T1q);
215 T1O = FMA(KP951056516, T1x, T1q);
216 T1F = FNMS(KP250000000, T1E, T1B);
217 T1H = FMA(KP559016994, T1G, T1F);
218 T1Z = FNMS(KP559016994, T1G, T1F);
219 T1U = FNMS(KP559016994, T1p, T1o);
220 T1W = FNMS(KP951056516, T1V, T1U);
221 T24 = FMA(KP951056516, T1V, T1U);
222 {
223 E T1L, T1A, T1M, T1f, T1z;
224 T1L = FMA(KP951056516, T1K, T1H);
225 T1A = W[1];
226 T1M = T1A * T1y;
227 T1f = W[0];
228 T1z = T1f * T1y;
229 cr[WS(rs, 1)] = FNMS(T1A, T1L, T1z);
230 ci[WS(rs, 1)] = FMA(T1f, T1L, T1M);
231 }
232 {
233 E T27, T26, T28, T23, T25;
234 T27 = FNMS(KP951056516, T20, T1Z);
235 T26 = W[13];
236 T28 = T26 * T24;
237 T23 = W[12];
238 T25 = T23 * T24;
239 cr[WS(rs, 7)] = FNMS(T26, T27, T25);
240 ci[WS(rs, 7)] = FMA(T23, T27, T28);
241 }
242 {
243 E T1R, T1Q, T1S, T1N, T1P;
244 T1R = FNMS(KP951056516, T1K, T1H);
245 T1Q = W[17];
246 T1S = T1Q * T1O;
247 T1N = W[16];
248 T1P = T1N * T1O;
249 cr[WS(rs, 9)] = FNMS(T1Q, T1R, T1P);
250 ci[WS(rs, 9)] = FMA(T1N, T1R, T1S);
251 }
252 {
253 E T21, T1Y, T22, T1T, T1X;
254 T21 = FMA(KP951056516, T20, T1Z);
255 T1Y = W[5];
256 T22 = T1Y * T1W;
257 T1T = W[4];
258 T1X = T1T * T1W;
259 cr[WS(rs, 3)] = FNMS(T1Y, T21, T1X);
260 ci[WS(rs, 3)] = FMA(T1T, T21, T22);
261 }
262 }
263 }
264 }
265 }
266
267 static const tw_instr twinstr[] = {
268 {TW_FULL, 1, 10},
269 {TW_NEXT, 1, 0}
270 };
271
272 static const hc2hc_desc desc = { 10, "hb_10", twinstr, &GENUS, {48, 18, 54, 0} };
273
274 void X(codelet_hb_10) (planner *p) {
275 X(khc2hc_register) (p, hb_10, &desc);
276 }
277 #else
278
279 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 10 -dif -name hb_10 -include rdft/scalar/hb.h */
280
281 /*
282 * This function contains 102 FP additions, 60 FP multiplications,
283 * (or, 72 additions, 30 multiplications, 30 fused multiply/add),
284 * 41 stack variables, 4 constants, and 40 memory accesses
285 */
286 #include "rdft/scalar/hb.h"
287
288 static void hb_10(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
289 {
290 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
291 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
292 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
293 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
294 {
295 INT m;
296 for (m = mb, W = W + ((mb - 1) * 18); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 18, MAKE_VOLATILE_STRIDE(20, rs)) {
297 E T3, T18, TE, TF, T1B, T1A, T1f, T1t, Ti, Tl, TJ, T1i, Tt, TA, T1w;
298 E T1v, T1p, T1E, TM, TO;
299 {
300 E T1, T2, TH, TI;
301 T1 = cr[0];
302 T2 = ci[WS(rs, 4)];
303 T3 = T1 + T2;
304 T18 = T1 - T2;
305 {
306 E T6, T19, Tg, T1d, T9, T1a, Td, T1c;
307 {
308 E T4, T5, Te, Tf;
309 T4 = cr[WS(rs, 2)];
310 T5 = ci[WS(rs, 2)];
311 T6 = T4 + T5;
312 T19 = T4 - T5;
313 Te = ci[WS(rs, 3)];
314 Tf = cr[WS(rs, 1)];
315 Tg = Te + Tf;
316 T1d = Te - Tf;
317 }
318 {
319 E T7, T8, Tb, Tc;
320 T7 = ci[WS(rs, 1)];
321 T8 = cr[WS(rs, 3)];
322 T9 = T7 + T8;
323 T1a = T7 - T8;
324 Tb = cr[WS(rs, 4)];
325 Tc = ci[0];
326 Td = Tb + Tc;
327 T1c = Tb - Tc;
328 }
329 TE = T6 - T9;
330 TF = Td - Tg;
331 T1B = T1c - T1d;
332 T1A = T19 - T1a;
333 {
334 E T1b, T1e, Ta, Th;
335 T1b = T19 + T1a;
336 T1e = T1c + T1d;
337 T1f = T1b + T1e;
338 T1t = KP559016994 * (T1b - T1e);
339 Ta = T6 + T9;
340 Th = Td + Tg;
341 Ti = Ta + Th;
342 Tl = KP559016994 * (Ta - Th);
343 }
344 }
345 TH = ci[WS(rs, 9)];
346 TI = cr[WS(rs, 5)];
347 TJ = TH - TI;
348 T1i = TH + TI;
349 {
350 E Tp, T1j, Tz, T1n, Ts, T1k, Tw, T1m;
351 {
352 E Tn, To, Tx, Ty;
353 Tn = ci[WS(rs, 7)];
354 To = cr[WS(rs, 7)];
355 Tp = Tn - To;
356 T1j = Tn + To;
357 Tx = ci[WS(rs, 8)];
358 Ty = cr[WS(rs, 6)];
359 Tz = Tx - Ty;
360 T1n = Tx + Ty;
361 }
362 {
363 E Tq, Tr, Tu, Tv;
364 Tq = ci[WS(rs, 6)];
365 Tr = cr[WS(rs, 8)];
366 Ts = Tq - Tr;
367 T1k = Tq + Tr;
368 Tu = ci[WS(rs, 5)];
369 Tv = cr[WS(rs, 9)];
370 Tw = Tu - Tv;
371 T1m = Tu + Tv;
372 }
373 Tt = Tp - Ts;
374 TA = Tw - Tz;
375 T1w = T1m + T1n;
376 T1v = T1j + T1k;
377 {
378 E T1l, T1o, TK, TL;
379 T1l = T1j - T1k;
380 T1o = T1m - T1n;
381 T1p = T1l + T1o;
382 T1E = KP559016994 * (T1l - T1o);
383 TK = Tp + Ts;
384 TL = Tw + Tz;
385 TM = TK + TL;
386 TO = KP559016994 * (TK - TL);
387 }
388 }
389 }
390 cr[0] = T3 + Ti;
391 ci[0] = TJ + TM;
392 {
393 E T1g, T1q, T17, T1h;
394 T1g = T18 + T1f;
395 T1q = T1i + T1p;
396 T17 = W[8];
397 T1h = W[9];
398 cr[WS(rs, 5)] = FNMS(T1h, T1q, T17 * T1g);
399 ci[WS(rs, 5)] = FMA(T1h, T1g, T17 * T1q);
400 }
401 {
402 E TB, TG, T11, TX, TP, T10, Tm, TW, TN, Tk;
403 TB = FNMS(KP951056516, TA, KP587785252 * Tt);
404 TG = FNMS(KP951056516, TF, KP587785252 * TE);
405 T11 = FMA(KP951056516, TE, KP587785252 * TF);
406 TX = FMA(KP951056516, Tt, KP587785252 * TA);
407 TN = FNMS(KP250000000, TM, TJ);
408 TP = TN - TO;
409 T10 = TO + TN;
410 Tk = FNMS(KP250000000, Ti, T3);
411 Tm = Tk - Tl;
412 TW = Tl + Tk;
413 {
414 E TC, TQ, Tj, TD;
415 TC = Tm - TB;
416 TQ = TG + TP;
417 Tj = W[2];
418 TD = W[3];
419 cr[WS(rs, 2)] = FNMS(TD, TQ, Tj * TC);
420 ci[WS(rs, 2)] = FMA(TD, TC, Tj * TQ);
421 }
422 {
423 E T14, T16, T13, T15;
424 T14 = TW - TX;
425 T16 = T11 + T10;
426 T13 = W[10];
427 T15 = W[11];
428 cr[WS(rs, 6)] = FNMS(T15, T16, T13 * T14);
429 ci[WS(rs, 6)] = FMA(T15, T14, T13 * T16);
430 }
431 {
432 E TS, TU, TR, TT;
433 TS = Tm + TB;
434 TU = TP - TG;
435 TR = W[14];
436 TT = W[15];
437 cr[WS(rs, 8)] = FNMS(TT, TU, TR * TS);
438 ci[WS(rs, 8)] = FMA(TT, TS, TR * TU);
439 }
440 {
441 E TY, T12, TV, TZ;
442 TY = TW + TX;
443 T12 = T10 - T11;
444 TV = W[6];
445 TZ = W[7];
446 cr[WS(rs, 4)] = FNMS(TZ, T12, TV * TY);
447 ci[WS(rs, 4)] = FMA(TZ, TY, TV * T12);
448 }
449 }
450 {
451 E T1x, T1C, T1Q, T1N, T1F, T1R, T1u, T1M, T1D, T1s;
452 T1x = FNMS(KP951056516, T1w, KP587785252 * T1v);
453 T1C = FNMS(KP951056516, T1B, KP587785252 * T1A);
454 T1Q = FMA(KP951056516, T1A, KP587785252 * T1B);
455 T1N = FMA(KP951056516, T1v, KP587785252 * T1w);
456 T1D = FNMS(KP250000000, T1p, T1i);
457 T1F = T1D - T1E;
458 T1R = T1E + T1D;
459 T1s = FNMS(KP250000000, T1f, T18);
460 T1u = T1s - T1t;
461 T1M = T1t + T1s;
462 {
463 E T1y, T1G, T1r, T1z;
464 T1y = T1u - T1x;
465 T1G = T1C + T1F;
466 T1r = W[12];
467 T1z = W[13];
468 cr[WS(rs, 7)] = FNMS(T1z, T1G, T1r * T1y);
469 ci[WS(rs, 7)] = FMA(T1r, T1G, T1z * T1y);
470 }
471 {
472 E T1U, T1W, T1T, T1V;
473 T1U = T1M + T1N;
474 T1W = T1R - T1Q;
475 T1T = W[16];
476 T1V = W[17];
477 cr[WS(rs, 9)] = FNMS(T1V, T1W, T1T * T1U);
478 ci[WS(rs, 9)] = FMA(T1T, T1W, T1V * T1U);
479 }
480 {
481 E T1I, T1K, T1H, T1J;
482 T1I = T1u + T1x;
483 T1K = T1F - T1C;
484 T1H = W[4];
485 T1J = W[5];
486 cr[WS(rs, 3)] = FNMS(T1J, T1K, T1H * T1I);
487 ci[WS(rs, 3)] = FMA(T1H, T1K, T1J * T1I);
488 }
489 {
490 E T1O, T1S, T1L, T1P;
491 T1O = T1M - T1N;
492 T1S = T1Q + T1R;
493 T1L = W[0];
494 T1P = W[1];
495 cr[WS(rs, 1)] = FNMS(T1P, T1S, T1L * T1O);
496 ci[WS(rs, 1)] = FMA(T1L, T1S, T1P * T1O);
497 }
498 }
499 }
500 }
501 }
502
503 static const tw_instr twinstr[] = {
504 {TW_FULL, 1, 10},
505 {TW_NEXT, 1, 0}
506 };
507
508 static const hc2hc_desc desc = { 10, "hb_10", twinstr, &GENUS, {72, 30, 30, 0} };
509
510 void X(codelet_hb_10) (planner *p) {
511 X(khc2hc_register) (p, hb_10, &desc);
512 }
513 #endif