Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/n1_12.c @ 82:d0c2a83c1364

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
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
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 08:04:10 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_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */
29
30 /*
31 * This function contains 96 FP additions, 24 FP multiplications,
32 * (or, 72 additions, 0 multiplications, 24 fused multiply/add),
33 * 43 stack variables, 2 constants, and 48 memory accesses
34 */
35 #include "dft/scalar/n.h"
36
37 static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT i;
43 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) {
44 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1d, TG;
45 E TJ, T1u, T1c, Tl, T1i, TL, TO, T1v, T1h;
46 {
47 E T1, T2, T3, T4;
48 T1 = ri[0];
49 T2 = ri[WS(is, 4)];
50 T3 = ri[WS(is, 8)];
51 T4 = T2 + T3;
52 T5 = T1 + T4;
53 TR = FNMS(KP500000000, T4, T1);
54 TA = T3 - T2;
55 }
56 {
57 E To, Tp, Tq, Tr;
58 To = ii[0];
59 Tp = ii[WS(is, 4)];
60 Tq = ii[WS(is, 8)];
61 Tr = Tp + Tq;
62 Ts = To + Tr;
63 TS = Tp - Tq;
64 Tz = FNMS(KP500000000, Tr, To);
65 }
66 {
67 E T6, T7, T8, T9;
68 T6 = ri[WS(is, 6)];
69 T7 = ri[WS(is, 10)];
70 T8 = ri[WS(is, 2)];
71 T9 = T7 + T8;
72 Ta = T6 + T9;
73 TU = FNMS(KP500000000, T9, T6);
74 TD = T8 - T7;
75 }
76 {
77 E Tt, Tu, Tv, Tw;
78 Tt = ii[WS(is, 6)];
79 Tu = ii[WS(is, 10)];
80 Tv = ii[WS(is, 2)];
81 Tw = Tu + Tv;
82 Tx = Tt + Tw;
83 TV = Tu - Tv;
84 TC = FNMS(KP500000000, Tw, Tt);
85 }
86 {
87 E Tc, Td, Te, Tf;
88 Tc = ri[WS(is, 3)];
89 Td = ri[WS(is, 7)];
90 Te = ri[WS(is, 11)];
91 Tf = Td + Te;
92 Tg = Tc + Tf;
93 T1d = Te - Td;
94 TG = FNMS(KP500000000, Tf, Tc);
95 }
96 {
97 E T1a, TH, TI, T1b;
98 T1a = ii[WS(is, 3)];
99 TH = ii[WS(is, 7)];
100 TI = ii[WS(is, 11)];
101 T1b = TH + TI;
102 TJ = TH - TI;
103 T1u = T1a + T1b;
104 T1c = FNMS(KP500000000, T1b, T1a);
105 }
106 {
107 E Th, Ti, Tj, Tk;
108 Th = ri[WS(is, 9)];
109 Ti = ri[WS(is, 1)];
110 Tj = ri[WS(is, 5)];
111 Tk = Ti + Tj;
112 Tl = Th + Tk;
113 T1i = Tj - Ti;
114 TL = FNMS(KP500000000, Tk, Th);
115 }
116 {
117 E T1f, TM, TN, T1g;
118 T1f = ii[WS(is, 9)];
119 TM = ii[WS(is, 1)];
120 TN = ii[WS(is, 5)];
121 T1g = TM + TN;
122 TO = TM - TN;
123 T1v = T1f + T1g;
124 T1h = FNMS(KP500000000, T1g, T1f);
125 }
126 {
127 E Tb, Tm, T1t, T1w;
128 Tb = T5 + Ta;
129 Tm = Tg + Tl;
130 ro[WS(os, 6)] = Tb - Tm;
131 ro[0] = Tb + Tm;
132 {
133 E T1x, T1y, Tn, Ty;
134 T1x = Ts + Tx;
135 T1y = T1u + T1v;
136 io[WS(os, 6)] = T1x - T1y;
137 io[0] = T1x + T1y;
138 Tn = Tg - Tl;
139 Ty = Ts - Tx;
140 io[WS(os, 3)] = Tn + Ty;
141 io[WS(os, 9)] = Ty - Tn;
142 }
143 T1t = T5 - Ta;
144 T1w = T1u - T1v;
145 ro[WS(os, 3)] = T1t - T1w;
146 ro[WS(os, 9)] = T1t + T1w;
147 {
148 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
149 {
150 E TZ, T10, T1e, T1j;
151 TZ = FMA(KP866025403, TA, Tz);
152 T10 = FMA(KP866025403, TD, TC);
153 T11 = TZ - T10;
154 T1l = TZ + T10;
155 T1e = FMA(KP866025403, T1d, T1c);
156 T1j = FMA(KP866025403, T1i, T1h);
157 T1k = T1e - T1j;
158 T1m = T1e + T1j;
159 }
160 {
161 E T12, T13, T15, T16;
162 T12 = FMA(KP866025403, TJ, TG);
163 T13 = FMA(KP866025403, TO, TL);
164 T14 = T12 - T13;
165 T18 = T12 + T13;
166 T15 = FMA(KP866025403, TS, TR);
167 T16 = FMA(KP866025403, TV, TU);
168 T17 = T15 + T16;
169 T19 = T15 - T16;
170 }
171 io[WS(os, 1)] = T11 - T14;
172 ro[WS(os, 1)] = T19 + T1k;
173 io[WS(os, 7)] = T11 + T14;
174 ro[WS(os, 7)] = T19 - T1k;
175 ro[WS(os, 10)] = T17 - T18;
176 io[WS(os, 10)] = T1l - T1m;
177 ro[WS(os, 4)] = T17 + T18;
178 io[WS(os, 4)] = T1l + T1m;
179 }
180 {
181 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
182 {
183 E TB, TE, T1o, T1p;
184 TB = FNMS(KP866025403, TA, Tz);
185 TE = FNMS(KP866025403, TD, TC);
186 TF = TB - TE;
187 T1r = TB + TE;
188 T1o = FNMS(KP866025403, T1d, T1c);
189 T1p = FNMS(KP866025403, T1i, T1h);
190 T1q = T1o - T1p;
191 T1s = T1o + T1p;
192 }
193 {
194 E TK, TP, TT, TW;
195 TK = FNMS(KP866025403, TJ, TG);
196 TP = FNMS(KP866025403, TO, TL);
197 TQ = TK - TP;
198 TY = TK + TP;
199 TT = FNMS(KP866025403, TS, TR);
200 TW = FNMS(KP866025403, TV, TU);
201 TX = TT + TW;
202 T1n = TT - TW;
203 }
204 io[WS(os, 5)] = TF - TQ;
205 ro[WS(os, 5)] = T1n + T1q;
206 io[WS(os, 11)] = TF + TQ;
207 ro[WS(os, 11)] = T1n - T1q;
208 ro[WS(os, 2)] = TX - TY;
209 io[WS(os, 2)] = T1r - T1s;
210 ro[WS(os, 8)] = TX + TY;
211 io[WS(os, 8)] = T1r + T1s;
212 }
213 }
214 }
215 }
216 }
217
218 static const kdft_desc desc = { 12, "n1_12", {72, 0, 24, 0}, &GENUS, 0, 0, 0, 0 };
219
220 void X(codelet_n1_12) (planner *p) {
221 X(kdft_register) (p, n1_12, &desc);
222 }
223
224 #else
225
226 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include dft/scalar/n.h */
227
228 /*
229 * This function contains 96 FP additions, 16 FP multiplications,
230 * (or, 88 additions, 8 multiplications, 8 fused multiply/add),
231 * 43 stack variables, 2 constants, and 48 memory accesses
232 */
233 #include "dft/scalar/n.h"
234
235 static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
236 {
237 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
238 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
239 {
240 INT i;
241 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(48, is), MAKE_VOLATILE_STRIDE(48, os)) {
242 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1a, TG;
243 E TJ, T1u, T1d, Tl, T1f, TL, TO, T1v, T1i;
244 {
245 E T1, T2, T3, T4;
246 T1 = ri[0];
247 T2 = ri[WS(is, 4)];
248 T3 = ri[WS(is, 8)];
249 T4 = T2 + T3;
250 T5 = T1 + T4;
251 TR = FNMS(KP500000000, T4, T1);
252 TA = KP866025403 * (T3 - T2);
253 }
254 {
255 E To, Tp, Tq, Tr;
256 To = ii[0];
257 Tp = ii[WS(is, 4)];
258 Tq = ii[WS(is, 8)];
259 Tr = Tp + Tq;
260 Ts = To + Tr;
261 TS = KP866025403 * (Tp - Tq);
262 Tz = FNMS(KP500000000, Tr, To);
263 }
264 {
265 E T6, T7, T8, T9;
266 T6 = ri[WS(is, 6)];
267 T7 = ri[WS(is, 10)];
268 T8 = ri[WS(is, 2)];
269 T9 = T7 + T8;
270 Ta = T6 + T9;
271 TU = FNMS(KP500000000, T9, T6);
272 TD = KP866025403 * (T8 - T7);
273 }
274 {
275 E Tt, Tu, Tv, Tw;
276 Tt = ii[WS(is, 6)];
277 Tu = ii[WS(is, 10)];
278 Tv = ii[WS(is, 2)];
279 Tw = Tu + Tv;
280 Tx = Tt + Tw;
281 TV = KP866025403 * (Tu - Tv);
282 TC = FNMS(KP500000000, Tw, Tt);
283 }
284 {
285 E Tc, Td, Te, Tf;
286 Tc = ri[WS(is, 3)];
287 Td = ri[WS(is, 7)];
288 Te = ri[WS(is, 11)];
289 Tf = Td + Te;
290 Tg = Tc + Tf;
291 T1a = KP866025403 * (Te - Td);
292 TG = FNMS(KP500000000, Tf, Tc);
293 }
294 {
295 E T1b, TH, TI, T1c;
296 T1b = ii[WS(is, 3)];
297 TH = ii[WS(is, 7)];
298 TI = ii[WS(is, 11)];
299 T1c = TH + TI;
300 TJ = KP866025403 * (TH - TI);
301 T1u = T1b + T1c;
302 T1d = FNMS(KP500000000, T1c, T1b);
303 }
304 {
305 E Th, Ti, Tj, Tk;
306 Th = ri[WS(is, 9)];
307 Ti = ri[WS(is, 1)];
308 Tj = ri[WS(is, 5)];
309 Tk = Ti + Tj;
310 Tl = Th + Tk;
311 T1f = KP866025403 * (Tj - Ti);
312 TL = FNMS(KP500000000, Tk, Th);
313 }
314 {
315 E T1g, TM, TN, T1h;
316 T1g = ii[WS(is, 9)];
317 TM = ii[WS(is, 1)];
318 TN = ii[WS(is, 5)];
319 T1h = TM + TN;
320 TO = KP866025403 * (TM - TN);
321 T1v = T1g + T1h;
322 T1i = FNMS(KP500000000, T1h, T1g);
323 }
324 {
325 E Tb, Tm, T1t, T1w;
326 Tb = T5 + Ta;
327 Tm = Tg + Tl;
328 ro[WS(os, 6)] = Tb - Tm;
329 ro[0] = Tb + Tm;
330 {
331 E T1x, T1y, Tn, Ty;
332 T1x = Ts + Tx;
333 T1y = T1u + T1v;
334 io[WS(os, 6)] = T1x - T1y;
335 io[0] = T1x + T1y;
336 Tn = Tg - Tl;
337 Ty = Ts - Tx;
338 io[WS(os, 3)] = Tn + Ty;
339 io[WS(os, 9)] = Ty - Tn;
340 }
341 T1t = T5 - Ta;
342 T1w = T1u - T1v;
343 ro[WS(os, 3)] = T1t - T1w;
344 ro[WS(os, 9)] = T1t + T1w;
345 {
346 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
347 {
348 E TZ, T10, T1e, T1j;
349 TZ = TA + Tz;
350 T10 = TD + TC;
351 T11 = TZ - T10;
352 T1l = TZ + T10;
353 T1e = T1a + T1d;
354 T1j = T1f + T1i;
355 T1k = T1e - T1j;
356 T1m = T1e + T1j;
357 }
358 {
359 E T12, T13, T15, T16;
360 T12 = TG + TJ;
361 T13 = TL + TO;
362 T14 = T12 - T13;
363 T18 = T12 + T13;
364 T15 = TR + TS;
365 T16 = TU + TV;
366 T17 = T15 + T16;
367 T19 = T15 - T16;
368 }
369 io[WS(os, 1)] = T11 - T14;
370 ro[WS(os, 1)] = T19 + T1k;
371 io[WS(os, 7)] = T11 + T14;
372 ro[WS(os, 7)] = T19 - T1k;
373 ro[WS(os, 10)] = T17 - T18;
374 io[WS(os, 10)] = T1l - T1m;
375 ro[WS(os, 4)] = T17 + T18;
376 io[WS(os, 4)] = T1l + T1m;
377 }
378 {
379 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
380 {
381 E TB, TE, T1o, T1p;
382 TB = Tz - TA;
383 TE = TC - TD;
384 TF = TB - TE;
385 T1r = TB + TE;
386 T1o = T1d - T1a;
387 T1p = T1i - T1f;
388 T1q = T1o - T1p;
389 T1s = T1o + T1p;
390 }
391 {
392 E TK, TP, TT, TW;
393 TK = TG - TJ;
394 TP = TL - TO;
395 TQ = TK - TP;
396 TY = TK + TP;
397 TT = TR - TS;
398 TW = TU - TV;
399 TX = TT + TW;
400 T1n = TT - TW;
401 }
402 io[WS(os, 5)] = TF - TQ;
403 ro[WS(os, 5)] = T1n + T1q;
404 io[WS(os, 11)] = TF + TQ;
405 ro[WS(os, 11)] = T1n - T1q;
406 ro[WS(os, 2)] = TX - TY;
407 io[WS(os, 2)] = T1r - T1s;
408 ro[WS(os, 8)] = TX + TY;
409 io[WS(os, 8)] = T1r + T1s;
410 }
411 }
412 }
413 }
414 }
415
416 static const kdft_desc desc = { 12, "n1_12", {88, 8, 8, 0}, &GENUS, 0, 0, 0, 0 };
417
418 void X(codelet_n1_12) (planner *p) {
419 X(kdft_register) (p, n1_12, &desc);
420 }
421
422 #endif