Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hc2cfdft2_8.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 08:07:16 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_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -dit -name hc2cfdft2_8 -include rdft/scalar/hc2cf.h */
29
30 /*
31 * This function contains 90 FP additions, 66 FP multiplications,
32 * (or, 60 additions, 36 multiplications, 30 fused multiply/add),
33 * 45 stack variables, 2 constants, and 32 memory accesses
34 */
35 #include "rdft/scalar/hc2cf.h"
36
37 static void hc2cfdft2_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(32, rs)) {
44 E T1, T2, Th, Tj, T4, T5, T6, Tk, TB, Tq, Tw, Tc, TM, TQ;
45 {
46 E T3, Ti, Tp, Tb, TL, TP;
47 T1 = W[0];
48 T2 = W[2];
49 T3 = T1 * T2;
50 Th = W[4];
51 Ti = T1 * Th;
52 Tj = W[5];
53 Tp = T1 * Tj;
54 T4 = W[1];
55 T5 = W[3];
56 Tb = T1 * T5;
57 T6 = FMA(T4, T5, T3);
58 Tk = FMA(T4, Tj, Ti);
59 TB = FMA(T4, T2, Tb);
60 Tq = FNMS(T4, Th, Tp);
61 Tw = FNMS(T4, T5, T3);
62 TL = T6 * Th;
63 TP = T6 * Tj;
64 Tc = FNMS(T4, T2, Tb);
65 TM = FMA(Tc, Tj, TL);
66 TQ = FNMS(Tc, Th, TP);
67 }
68 {
69 E TI, T1a, TY, T1u, TF, T1s, TS, T1c, Tg, T1n, T13, T1f, Tu, T1p, T17;
70 E T1h;
71 {
72 E TG, TH, TX, TT, TU, TV, TW, T1t;
73 TG = Ip[0];
74 TH = Im[0];
75 TX = TG + TH;
76 TT = Rm[0];
77 TU = Rp[0];
78 TV = TT - TU;
79 TI = TG - TH;
80 T1a = TU + TT;
81 TW = T1 * TV;
82 TY = FNMS(T4, TX, TW);
83 T1t = T4 * TV;
84 T1u = FMA(T1, TX, T1t);
85 }
86 {
87 E Tz, TR, TE, TN;
88 {
89 E Tx, Ty, TC, TD;
90 Tx = Ip[WS(rs, 2)];
91 Ty = Im[WS(rs, 2)];
92 Tz = Tx - Ty;
93 TR = Tx + Ty;
94 TC = Rp[WS(rs, 2)];
95 TD = Rm[WS(rs, 2)];
96 TE = TC + TD;
97 TN = TD - TC;
98 }
99 {
100 E TA, T1r, TO, T1b;
101 TA = Tw * Tz;
102 TF = FNMS(TB, TE, TA);
103 T1r = TQ * TN;
104 T1s = FMA(TM, TR, T1r);
105 TO = TM * TN;
106 TS = FNMS(TQ, TR, TO);
107 T1b = Tw * TE;
108 T1c = FMA(TB, Tz, T1b);
109 }
110 }
111 {
112 E T9, T12, Tf, T10;
113 {
114 E T7, T8, Td, Te;
115 T7 = Ip[WS(rs, 1)];
116 T8 = Im[WS(rs, 1)];
117 T9 = T7 - T8;
118 T12 = T7 + T8;
119 Td = Rp[WS(rs, 1)];
120 Te = Rm[WS(rs, 1)];
121 Tf = Td + Te;
122 T10 = Td - Te;
123 }
124 {
125 E Ta, T1m, T11, T1e;
126 Ta = T6 * T9;
127 Tg = FNMS(Tc, Tf, Ta);
128 T1m = T2 * T12;
129 T1n = FNMS(T5, T10, T1m);
130 T11 = T2 * T10;
131 T13 = FMA(T5, T12, T11);
132 T1e = T6 * Tf;
133 T1f = FMA(Tc, T9, T1e);
134 }
135 }
136 {
137 E Tn, T16, Tt, T14;
138 {
139 E Tl, Tm, Tr, Ts;
140 Tl = Ip[WS(rs, 3)];
141 Tm = Im[WS(rs, 3)];
142 Tn = Tl - Tm;
143 T16 = Tl + Tm;
144 Tr = Rp[WS(rs, 3)];
145 Ts = Rm[WS(rs, 3)];
146 Tt = Tr + Ts;
147 T14 = Tr - Ts;
148 }
149 {
150 E To, T1o, T15, T1g;
151 To = Tk * Tn;
152 Tu = FNMS(Tq, Tt, To);
153 T1o = Th * T16;
154 T1p = FNMS(Tj, T14, T1o);
155 T15 = Th * T14;
156 T17 = FMA(Tj, T16, T15);
157 T1g = Tk * Tt;
158 T1h = FMA(Tq, Tn, T1g);
159 }
160 }
161 {
162 E TK, T1l, T1w, T1y, T19, T1k, T1j, T1x;
163 {
164 E Tv, TJ, T1q, T1v;
165 Tv = Tg + Tu;
166 TJ = TF + TI;
167 TK = Tv + TJ;
168 T1l = TJ - Tv;
169 T1q = T1n + T1p;
170 T1v = T1s + T1u;
171 T1w = T1q - T1v;
172 T1y = T1q + T1v;
173 }
174 {
175 E TZ, T18, T1d, T1i;
176 TZ = TS + TY;
177 T18 = T13 + T17;
178 T19 = TZ - T18;
179 T1k = T18 + TZ;
180 T1d = T1a + T1c;
181 T1i = T1f + T1h;
182 T1j = T1d - T1i;
183 T1x = T1d + T1i;
184 }
185 Ip[0] = KP500000000 * (TK + T19);
186 Rp[0] = KP500000000 * (T1x + T1y);
187 Im[WS(rs, 3)] = KP500000000 * (T19 - TK);
188 Rm[WS(rs, 3)] = KP500000000 * (T1x - T1y);
189 Rm[WS(rs, 1)] = KP500000000 * (T1j - T1k);
190 Im[WS(rs, 1)] = KP500000000 * (T1w - T1l);
191 Rp[WS(rs, 2)] = KP500000000 * (T1j + T1k);
192 Ip[WS(rs, 2)] = KP500000000 * (T1l + T1w);
193 }
194 {
195 E T1B, T1N, T1L, T1R, T1E, T1O, T1H, T1P;
196 {
197 E T1z, T1A, T1J, T1K;
198 T1z = TI - TF;
199 T1A = T1f - T1h;
200 T1B = T1z - T1A;
201 T1N = T1A + T1z;
202 T1J = T1a - T1c;
203 T1K = Tg - Tu;
204 T1L = T1J - T1K;
205 T1R = T1J + T1K;
206 }
207 {
208 E T1C, T1D, T1F, T1G;
209 T1C = T1p - T1n;
210 T1D = T13 - T17;
211 T1E = T1C + T1D;
212 T1O = T1C - T1D;
213 T1F = TY - TS;
214 T1G = T1u - T1s;
215 T1H = T1F - T1G;
216 T1P = T1F + T1G;
217 }
218 {
219 E T1I, T1S, T1M, T1Q;
220 T1I = T1E + T1H;
221 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1I, T1B));
222 Im[WS(rs, 2)] = -(KP500000000 * (FNMS(KP707106781, T1I, T1B)));
223 T1S = T1O + T1P;
224 Rm[WS(rs, 2)] = KP500000000 * (FNMS(KP707106781, T1S, T1R));
225 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1S, T1R));
226 T1M = T1H - T1E;
227 Rm[0] = KP500000000 * (FNMS(KP707106781, T1M, T1L));
228 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1M, T1L));
229 T1Q = T1O - T1P;
230 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1Q, T1N));
231 Im[0] = -(KP500000000 * (FNMS(KP707106781, T1Q, T1N)));
232 }
233 }
234 }
235 }
236 }
237 }
238
239 static const tw_instr twinstr[] = {
240 {TW_CEXP, 1, 1},
241 {TW_CEXP, 1, 3},
242 {TW_CEXP, 1, 7},
243 {TW_NEXT, 1, 0}
244 };
245
246 static const hc2c_desc desc = { 8, "hc2cfdft2_8", twinstr, &GENUS, {60, 36, 30, 0} };
247
248 void X(codelet_hc2cfdft2_8) (planner *p) {
249 X(khc2c_register) (p, hc2cfdft2_8, &desc, HC2C_VIA_DFT);
250 }
251 #else
252
253 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -dit -name hc2cfdft2_8 -include rdft/scalar/hc2cf.h */
254
255 /*
256 * This function contains 90 FP additions, 56 FP multiplications,
257 * (or, 72 additions, 38 multiplications, 18 fused multiply/add),
258 * 51 stack variables, 2 constants, and 32 memory accesses
259 */
260 #include "rdft/scalar/hc2cf.h"
261
262 static void hc2cfdft2_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
263 {
264 DK(KP353553390, +0.353553390593273762200422181052424519642417969);
265 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
266 {
267 INT m;
268 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(32, rs)) {
269 E T1, T4, T2, T5, Tu, Ty, T7, Td, Ti, Tj, Tk, TP, To, TN;
270 {
271 E T3, Tc, T6, Tb;
272 T1 = W[0];
273 T4 = W[1];
274 T2 = W[2];
275 T5 = W[3];
276 T3 = T1 * T2;
277 Tc = T4 * T2;
278 T6 = T4 * T5;
279 Tb = T1 * T5;
280 Tu = T3 - T6;
281 Ty = Tb + Tc;
282 T7 = T3 + T6;
283 Td = Tb - Tc;
284 Ti = W[4];
285 Tj = W[5];
286 Tk = FMA(T1, Ti, T4 * Tj);
287 TP = FNMS(Td, Ti, T7 * Tj);
288 To = FNMS(T4, Ti, T1 * Tj);
289 TN = FMA(T7, Ti, Td * Tj);
290 }
291 {
292 E TF, T11, TC, T12, T1d, T1e, T1q, TM, TR, T1p, Th, Ts, T15, T14, T1a;
293 E T1b, T1m, TV, TY, T1n;
294 {
295 E TD, TE, TL, TI, TJ, TK, Tx, TQ, TB, TO;
296 TD = Ip[0];
297 TE = Im[0];
298 TL = TD + TE;
299 TI = Rm[0];
300 TJ = Rp[0];
301 TK = TI - TJ;
302 {
303 E Tv, Tw, Tz, TA;
304 Tv = Ip[WS(rs, 2)];
305 Tw = Im[WS(rs, 2)];
306 Tx = Tv - Tw;
307 TQ = Tv + Tw;
308 Tz = Rp[WS(rs, 2)];
309 TA = Rm[WS(rs, 2)];
310 TB = Tz + TA;
311 TO = Tz - TA;
312 }
313 TF = TD - TE;
314 T11 = TJ + TI;
315 TC = FNMS(Ty, TB, Tu * Tx);
316 T12 = FMA(Tu, TB, Ty * Tx);
317 T1d = FNMS(TP, TO, TN * TQ);
318 T1e = FMA(T4, TK, T1 * TL);
319 T1q = T1e - T1d;
320 TM = FNMS(T4, TL, T1 * TK);
321 TR = FMA(TN, TO, TP * TQ);
322 T1p = TR + TM;
323 }
324 {
325 E Ta, TU, Tg, TT, Tn, TX, Tr, TW;
326 {
327 E T8, T9, Te, Tf;
328 T8 = Ip[WS(rs, 1)];
329 T9 = Im[WS(rs, 1)];
330 Ta = T8 - T9;
331 TU = T8 + T9;
332 Te = Rp[WS(rs, 1)];
333 Tf = Rm[WS(rs, 1)];
334 Tg = Te + Tf;
335 TT = Te - Tf;
336 }
337 {
338 E Tl, Tm, Tp, Tq;
339 Tl = Ip[WS(rs, 3)];
340 Tm = Im[WS(rs, 3)];
341 Tn = Tl - Tm;
342 TX = Tl + Tm;
343 Tp = Rp[WS(rs, 3)];
344 Tq = Rm[WS(rs, 3)];
345 Tr = Tp + Tq;
346 TW = Tp - Tq;
347 }
348 Th = FNMS(Td, Tg, T7 * Ta);
349 Ts = FNMS(To, Tr, Tk * Tn);
350 T15 = FMA(Tk, Tr, To * Tn);
351 T14 = FMA(T7, Tg, Td * Ta);
352 T1a = FNMS(T5, TT, T2 * TU);
353 T1b = FNMS(Tj, TW, Ti * TX);
354 T1m = T1b - T1a;
355 TV = FMA(T2, TT, T5 * TU);
356 TY = FMA(Ti, TW, Tj * TX);
357 T1n = TV - TY;
358 }
359 {
360 E T1l, T1x, T1A, T1C, T1s, T1w, T1v, T1B;
361 {
362 E T1j, T1k, T1y, T1z;
363 T1j = TF - TC;
364 T1k = T14 - T15;
365 T1l = KP500000000 * (T1j - T1k);
366 T1x = KP500000000 * (T1k + T1j);
367 T1y = T1m - T1n;
368 T1z = T1p + T1q;
369 T1A = KP353553390 * (T1y - T1z);
370 T1C = KP353553390 * (T1y + T1z);
371 }
372 {
373 E T1o, T1r, T1t, T1u;
374 T1o = T1m + T1n;
375 T1r = T1p - T1q;
376 T1s = KP353553390 * (T1o + T1r);
377 T1w = KP353553390 * (T1r - T1o);
378 T1t = T11 - T12;
379 T1u = Th - Ts;
380 T1v = KP500000000 * (T1t - T1u);
381 T1B = KP500000000 * (T1t + T1u);
382 }
383 Ip[WS(rs, 1)] = T1l + T1s;
384 Rp[WS(rs, 1)] = T1B + T1C;
385 Im[WS(rs, 2)] = T1s - T1l;
386 Rm[WS(rs, 2)] = T1B - T1C;
387 Rm[0] = T1v - T1w;
388 Im[0] = T1A - T1x;
389 Rp[WS(rs, 3)] = T1v + T1w;
390 Ip[WS(rs, 3)] = T1x + T1A;
391 }
392 {
393 E TH, T19, T1g, T1i, T10, T18, T17, T1h;
394 {
395 E Tt, TG, T1c, T1f;
396 Tt = Th + Ts;
397 TG = TC + TF;
398 TH = Tt + TG;
399 T19 = TG - Tt;
400 T1c = T1a + T1b;
401 T1f = T1d + T1e;
402 T1g = T1c - T1f;
403 T1i = T1c + T1f;
404 }
405 {
406 E TS, TZ, T13, T16;
407 TS = TM - TR;
408 TZ = TV + TY;
409 T10 = TS - TZ;
410 T18 = TZ + TS;
411 T13 = T11 + T12;
412 T16 = T14 + T15;
413 T17 = T13 - T16;
414 T1h = T13 + T16;
415 }
416 Ip[0] = KP500000000 * (TH + T10);
417 Rp[0] = KP500000000 * (T1h + T1i);
418 Im[WS(rs, 3)] = KP500000000 * (T10 - TH);
419 Rm[WS(rs, 3)] = KP500000000 * (T1h - T1i);
420 Rm[WS(rs, 1)] = KP500000000 * (T17 - T18);
421 Im[WS(rs, 1)] = KP500000000 * (T1g - T19);
422 Rp[WS(rs, 2)] = KP500000000 * (T17 + T18);
423 Ip[WS(rs, 2)] = KP500000000 * (T19 + T1g);
424 }
425 }
426 }
427 }
428 }
429
430 static const tw_instr twinstr[] = {
431 {TW_CEXP, 1, 1},
432 {TW_CEXP, 1, 3},
433 {TW_CEXP, 1, 7},
434 {TW_NEXT, 1, 0}
435 };
436
437 static const hc2c_desc desc = { 8, "hc2cfdft2_8", twinstr, &GENUS, {72, 38, 18, 0} };
438
439 void X(codelet_hc2cfdft2_8) (planner *p) {
440 X(khc2c_register) (p, hc2cfdft2_8, &desc, HC2C_VIA_DFT);
441 }
442 #endif