Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hc2cfdft_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 08:07:10 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 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
29
30 /*
31 * This function contains 82 FP additions, 52 FP multiplications,
32 * (or, 60 additions, 30 multiplications, 22 fused multiply/add),
33 * 31 stack variables, 2 constants, and 32 memory accesses
34 */
35 #include "rdft/scalar/hc2cf.h"
36
37 static void hc2cfdft_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) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
44 E Ty, T14, TO, T1o, Tv, T16, TG, T1m, Ta, T19, TV, T1h, Tk, T1b, T11;
45 E T1j;
46 {
47 E Tw, Tx, TN, TI, TJ, TK;
48 Tw = Ip[0];
49 Tx = Im[0];
50 TN = Tw + Tx;
51 TI = Rm[0];
52 TJ = Rp[0];
53 TK = TI - TJ;
54 Ty = Tw - Tx;
55 T14 = TJ + TI;
56 {
57 E TH, TL, TM, T1n;
58 TH = W[0];
59 TL = TH * TK;
60 TM = W[1];
61 T1n = TM * TK;
62 TO = FNMS(TM, TN, TL);
63 T1o = FMA(TH, TN, T1n);
64 }
65 }
66 {
67 E Tp, TF, Tu, TC;
68 {
69 E Tn, To, Ts, Tt;
70 Tn = Ip[WS(rs, 2)];
71 To = Im[WS(rs, 2)];
72 Tp = Tn - To;
73 TF = Tn + To;
74 Ts = Rp[WS(rs, 2)];
75 Tt = Rm[WS(rs, 2)];
76 Tu = Ts + Tt;
77 TC = Tt - Ts;
78 }
79 {
80 E Tq, T15, Tm, Tr;
81 Tm = W[6];
82 Tq = Tm * Tp;
83 T15 = Tm * Tu;
84 Tr = W[7];
85 Tv = FNMS(Tr, Tu, Tq);
86 T16 = FMA(Tr, Tp, T15);
87 }
88 {
89 E TB, TD, TE, T1l;
90 TB = W[8];
91 TD = TB * TC;
92 TE = W[9];
93 T1l = TE * TC;
94 TG = FNMS(TE, TF, TD);
95 T1m = FMA(TB, TF, T1l);
96 }
97 }
98 {
99 E T4, TU, T9, TR;
100 {
101 E T2, T3, T7, T8;
102 T2 = Ip[WS(rs, 1)];
103 T3 = Im[WS(rs, 1)];
104 T4 = T2 - T3;
105 TU = T2 + T3;
106 T7 = Rp[WS(rs, 1)];
107 T8 = Rm[WS(rs, 1)];
108 T9 = T7 + T8;
109 TR = T7 - T8;
110 }
111 {
112 E T5, T18, T1, T6;
113 T1 = W[2];
114 T5 = T1 * T4;
115 T18 = T1 * T9;
116 T6 = W[3];
117 Ta = FNMS(T6, T9, T5);
118 T19 = FMA(T6, T4, T18);
119 }
120 {
121 E TS, T1g, TQ, TT;
122 TQ = W[4];
123 TS = TQ * TR;
124 T1g = TQ * TU;
125 TT = W[5];
126 TV = FMA(TT, TU, TS);
127 T1h = FNMS(TT, TR, T1g);
128 }
129 }
130 {
131 E Te, T10, Tj, TX;
132 {
133 E Tc, Td, Th, Ti;
134 Tc = Ip[WS(rs, 3)];
135 Td = Im[WS(rs, 3)];
136 Te = Tc - Td;
137 T10 = Tc + Td;
138 Th = Rp[WS(rs, 3)];
139 Ti = Rm[WS(rs, 3)];
140 Tj = Th + Ti;
141 TX = Th - Ti;
142 }
143 {
144 E Tf, T1a, Tb, Tg;
145 Tb = W[10];
146 Tf = Tb * Te;
147 T1a = Tb * Tj;
148 Tg = W[11];
149 Tk = FNMS(Tg, Tj, Tf);
150 T1b = FMA(Tg, Te, T1a);
151 }
152 {
153 E TY, T1i, TW, TZ;
154 TW = W[12];
155 TY = TW * TX;
156 T1i = TW * T10;
157 TZ = W[13];
158 T11 = FMA(TZ, T10, TY);
159 T1j = FNMS(TZ, TX, T1i);
160 }
161 }
162 {
163 E TA, T1f, T1q, T1s, T13, T1e, T1d, T1r;
164 {
165 E Tl, Tz, T1k, T1p;
166 Tl = Ta + Tk;
167 Tz = Tv + Ty;
168 TA = Tl + Tz;
169 T1f = Tz - Tl;
170 T1k = T1h + T1j;
171 T1p = T1m + T1o;
172 T1q = T1k - T1p;
173 T1s = T1k + T1p;
174 }
175 {
176 E TP, T12, T17, T1c;
177 TP = TG + TO;
178 T12 = TV + T11;
179 T13 = TP - T12;
180 T1e = T12 + TP;
181 T17 = T14 + T16;
182 T1c = T19 + T1b;
183 T1d = T17 - T1c;
184 T1r = T17 + T1c;
185 }
186 Ip[0] = KP500000000 * (TA + T13);
187 Rp[0] = KP500000000 * (T1r + T1s);
188 Im[WS(rs, 3)] = KP500000000 * (T13 - TA);
189 Rm[WS(rs, 3)] = KP500000000 * (T1r - T1s);
190 Rm[WS(rs, 1)] = KP500000000 * (T1d - T1e);
191 Im[WS(rs, 1)] = KP500000000 * (T1q - T1f);
192 Rp[WS(rs, 2)] = KP500000000 * (T1d + T1e);
193 Ip[WS(rs, 2)] = KP500000000 * (T1f + T1q);
194 }
195 {
196 E T1v, T1H, T1F, T1L, T1y, T1I, T1B, T1J;
197 {
198 E T1t, T1u, T1D, T1E;
199 T1t = Ty - Tv;
200 T1u = T19 - T1b;
201 T1v = T1t - T1u;
202 T1H = T1u + T1t;
203 T1D = T14 - T16;
204 T1E = Ta - Tk;
205 T1F = T1D - T1E;
206 T1L = T1D + T1E;
207 }
208 {
209 E T1w, T1x, T1z, T1A;
210 T1w = T1j - T1h;
211 T1x = TV - T11;
212 T1y = T1w + T1x;
213 T1I = T1w - T1x;
214 T1z = TO - TG;
215 T1A = T1o - T1m;
216 T1B = T1z - T1A;
217 T1J = T1z + T1A;
218 }
219 {
220 E T1C, T1M, T1G, T1K;
221 T1C = T1y + T1B;
222 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1C, T1v));
223 Im[WS(rs, 2)] = -(KP500000000 * (FNMS(KP707106781, T1C, T1v)));
224 T1M = T1I + T1J;
225 Rm[WS(rs, 2)] = KP500000000 * (FNMS(KP707106781, T1M, T1L));
226 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1M, T1L));
227 T1G = T1B - T1y;
228 Rm[0] = KP500000000 * (FNMS(KP707106781, T1G, T1F));
229 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1G, T1F));
230 T1K = T1I - T1J;
231 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1K, T1H));
232 Im[0] = -(KP500000000 * (FNMS(KP707106781, T1K, T1H)));
233 }
234 }
235 }
236 }
237 }
238
239 static const tw_instr twinstr[] = {
240 {TW_FULL, 1, 8},
241 {TW_NEXT, 1, 0}
242 };
243
244 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {60, 30, 22, 0} };
245
246 void X(codelet_hc2cfdft_8) (planner *p) {
247 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
248 }
249 #else
250
251 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include rdft/scalar/hc2cf.h */
252
253 /*
254 * This function contains 82 FP additions, 44 FP multiplications,
255 * (or, 68 additions, 30 multiplications, 14 fused multiply/add),
256 * 39 stack variables, 2 constants, and 32 memory accesses
257 */
258 #include "rdft/scalar/hc2cf.h"
259
260 static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
261 {
262 DK(KP353553390, +0.353553390593273762200422181052424519642417969);
263 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
264 {
265 INT m;
266 for (m = mb, W = W + ((mb - 1) * 14); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 14, MAKE_VOLATILE_STRIDE(32, rs)) {
267 E Tv, TX, Ts, TY, TE, T1a, TJ, T19, T1l, T1m, T9, T10, Ti, T11, TP;
268 E T16, TU, T17, T1i, T1j;
269 {
270 E Tt, Tu, TD, Tz, TA, TB, Tn, TI, Tr, TG, Tk, To;
271 Tt = Ip[0];
272 Tu = Im[0];
273 TD = Tt + Tu;
274 Tz = Rm[0];
275 TA = Rp[0];
276 TB = Tz - TA;
277 {
278 E Tl, Tm, Tp, Tq;
279 Tl = Ip[WS(rs, 2)];
280 Tm = Im[WS(rs, 2)];
281 Tn = Tl - Tm;
282 TI = Tl + Tm;
283 Tp = Rp[WS(rs, 2)];
284 Tq = Rm[WS(rs, 2)];
285 Tr = Tp + Tq;
286 TG = Tp - Tq;
287 }
288 Tv = Tt - Tu;
289 TX = TA + Tz;
290 Tk = W[6];
291 To = W[7];
292 Ts = FNMS(To, Tr, Tk * Tn);
293 TY = FMA(Tk, Tr, To * Tn);
294 {
295 E Ty, TC, TF, TH;
296 Ty = W[0];
297 TC = W[1];
298 TE = FNMS(TC, TD, Ty * TB);
299 T1a = FMA(TC, TB, Ty * TD);
300 TF = W[8];
301 TH = W[9];
302 TJ = FMA(TF, TG, TH * TI);
303 T19 = FNMS(TH, TG, TF * TI);
304 }
305 T1l = TJ + TE;
306 T1m = T1a - T19;
307 }
308 {
309 E T4, TO, T8, TM, Td, TT, Th, TR;
310 {
311 E T2, T3, T6, T7;
312 T2 = Ip[WS(rs, 1)];
313 T3 = Im[WS(rs, 1)];
314 T4 = T2 - T3;
315 TO = T2 + T3;
316 T6 = Rp[WS(rs, 1)];
317 T7 = Rm[WS(rs, 1)];
318 T8 = T6 + T7;
319 TM = T6 - T7;
320 }
321 {
322 E Tb, Tc, Tf, Tg;
323 Tb = Ip[WS(rs, 3)];
324 Tc = Im[WS(rs, 3)];
325 Td = Tb - Tc;
326 TT = Tb + Tc;
327 Tf = Rp[WS(rs, 3)];
328 Tg = Rm[WS(rs, 3)];
329 Th = Tf + Tg;
330 TR = Tf - Tg;
331 }
332 {
333 E T1, T5, Ta, Te;
334 T1 = W[2];
335 T5 = W[3];
336 T9 = FNMS(T5, T8, T1 * T4);
337 T10 = FMA(T1, T8, T5 * T4);
338 Ta = W[10];
339 Te = W[11];
340 Ti = FNMS(Te, Th, Ta * Td);
341 T11 = FMA(Ta, Th, Te * Td);
342 {
343 E TL, TN, TQ, TS;
344 TL = W[4];
345 TN = W[5];
346 TP = FMA(TL, TM, TN * TO);
347 T16 = FNMS(TN, TM, TL * TO);
348 TQ = W[12];
349 TS = W[13];
350 TU = FMA(TQ, TR, TS * TT);
351 T17 = FNMS(TS, TR, TQ * TT);
352 }
353 T1i = T17 - T16;
354 T1j = TP - TU;
355 }
356 }
357 {
358 E T1h, T1t, T1w, T1y, T1o, T1s, T1r, T1x;
359 {
360 E T1f, T1g, T1u, T1v;
361 T1f = Tv - Ts;
362 T1g = T10 - T11;
363 T1h = KP500000000 * (T1f - T1g);
364 T1t = KP500000000 * (T1g + T1f);
365 T1u = T1i - T1j;
366 T1v = T1l + T1m;
367 T1w = KP353553390 * (T1u - T1v);
368 T1y = KP353553390 * (T1u + T1v);
369 }
370 {
371 E T1k, T1n, T1p, T1q;
372 T1k = T1i + T1j;
373 T1n = T1l - T1m;
374 T1o = KP353553390 * (T1k + T1n);
375 T1s = KP353553390 * (T1n - T1k);
376 T1p = TX - TY;
377 T1q = T9 - Ti;
378 T1r = KP500000000 * (T1p - T1q);
379 T1x = KP500000000 * (T1p + T1q);
380 }
381 Ip[WS(rs, 1)] = T1h + T1o;
382 Rp[WS(rs, 1)] = T1x + T1y;
383 Im[WS(rs, 2)] = T1o - T1h;
384 Rm[WS(rs, 2)] = T1x - T1y;
385 Rm[0] = T1r - T1s;
386 Im[0] = T1w - T1t;
387 Rp[WS(rs, 3)] = T1r + T1s;
388 Ip[WS(rs, 3)] = T1t + T1w;
389 }
390 {
391 E Tx, T15, T1c, T1e, TW, T14, T13, T1d;
392 {
393 E Tj, Tw, T18, T1b;
394 Tj = T9 + Ti;
395 Tw = Ts + Tv;
396 Tx = Tj + Tw;
397 T15 = Tw - Tj;
398 T18 = T16 + T17;
399 T1b = T19 + T1a;
400 T1c = T18 - T1b;
401 T1e = T18 + T1b;
402 }
403 {
404 E TK, TV, TZ, T12;
405 TK = TE - TJ;
406 TV = TP + TU;
407 TW = TK - TV;
408 T14 = TV + TK;
409 TZ = TX + TY;
410 T12 = T10 + T11;
411 T13 = TZ - T12;
412 T1d = TZ + T12;
413 }
414 Ip[0] = KP500000000 * (Tx + TW);
415 Rp[0] = KP500000000 * (T1d + T1e);
416 Im[WS(rs, 3)] = KP500000000 * (TW - Tx);
417 Rm[WS(rs, 3)] = KP500000000 * (T1d - T1e);
418 Rm[WS(rs, 1)] = KP500000000 * (T13 - T14);
419 Im[WS(rs, 1)] = KP500000000 * (T1c - T15);
420 Rp[WS(rs, 2)] = KP500000000 * (T13 + T14);
421 Ip[WS(rs, 2)] = KP500000000 * (T15 + T1c);
422 }
423 }
424 }
425 }
426
427 static const tw_instr twinstr[] = {
428 {TW_FULL, 1, 8},
429 {TW_NEXT, 1, 0}
430 };
431
432 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {68, 30, 14, 0} };
433
434 void X(codelet_hc2cfdft_8) (planner *p) {
435 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
436 }
437 #endif