Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/n1_12.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
1 /*
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
3 * Copyright (c) 2003, 2007-11 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 Sun Nov 25 07:35:43 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include 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 * 63 stack variables, 2 constants, and 48 memory accesses
34 */
35 #include "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 TT, TW, TF, T1q, TY, TQ, TX, T1n;
45 {
46 E TA, TS, TR, T5, Ts, Tz, TD, TV, TU, Ta, Tx, TC, T1d, Th, TJ;
47 E TG, Tg, T1u, T1c, T1f, TM, TN, Tk, T1i;
48 {
49 E T6, Tt, Tu, Tv, T9;
50 {
51 E T1, To, Tp, Tq, T4, T2, T3, T7, T8, Tr;
52 T1 = ri[0];
53 T2 = ri[WS(is, 4)];
54 T3 = ri[WS(is, 8)];
55 To = ii[0];
56 Tp = ii[WS(is, 4)];
57 Tq = ii[WS(is, 8)];
58 T4 = T2 + T3;
59 TA = T3 - T2;
60 T6 = ri[WS(is, 6)];
61 TS = Tp - Tq;
62 Tr = Tp + Tq;
63 TR = FNMS(KP500000000, T4, T1);
64 T5 = T1 + T4;
65 T7 = ri[WS(is, 10)];
66 Ts = To + Tr;
67 Tz = FNMS(KP500000000, Tr, To);
68 T8 = ri[WS(is, 2)];
69 Tt = ii[WS(is, 6)];
70 Tu = ii[WS(is, 10)];
71 Tv = ii[WS(is, 2)];
72 T9 = T7 + T8;
73 TD = T8 - T7;
74 }
75 {
76 E Tc, T1a, TH, TI, Tf, Td, Te, Tw, Ti, Tj, T1b;
77 Tc = ri[WS(is, 3)];
78 TV = Tu - Tv;
79 Tw = Tu + Tv;
80 TU = FNMS(KP500000000, T9, T6);
81 Ta = T6 + T9;
82 Td = ri[WS(is, 7)];
83 Tx = Tt + Tw;
84 TC = FNMS(KP500000000, Tw, Tt);
85 Te = ri[WS(is, 11)];
86 T1a = ii[WS(is, 3)];
87 TH = ii[WS(is, 7)];
88 TI = ii[WS(is, 11)];
89 Tf = Td + Te;
90 T1d = Te - Td;
91 Th = ri[WS(is, 9)];
92 TJ = TH - TI;
93 T1b = TH + TI;
94 TG = FNMS(KP500000000, Tf, Tc);
95 Tg = Tc + Tf;
96 Ti = ri[WS(is, 1)];
97 T1u = T1a + T1b;
98 T1c = FNMS(KP500000000, T1b, T1a);
99 Tj = ri[WS(is, 5)];
100 T1f = ii[WS(is, 9)];
101 TM = ii[WS(is, 1)];
102 TN = ii[WS(is, 5)];
103 Tk = Ti + Tj;
104 T1i = Tj - Ti;
105 }
106 }
107 {
108 E T1t, TO, TL, T1h, T1w, Tb, T1g, Tl;
109 T1t = T5 - Ta;
110 Tb = T5 + Ta;
111 TO = TM - TN;
112 T1g = TM + TN;
113 TL = FNMS(KP500000000, Tk, Th);
114 Tl = Th + Tk;
115 {
116 E T1x, Ty, T1v, Tn, Tm, T1y;
117 T1x = Ts + Tx;
118 Ty = Ts - Tx;
119 T1v = T1f + T1g;
120 T1h = FNMS(KP500000000, T1g, T1f);
121 Tn = Tg - Tl;
122 Tm = Tg + Tl;
123 T1y = T1u + T1v;
124 T1w = T1u - T1v;
125 ro[0] = Tb + Tm;
126 ro[WS(os, 6)] = Tb - Tm;
127 io[WS(os, 3)] = Tn + Ty;
128 io[0] = T1x + T1y;
129 io[WS(os, 6)] = T1x - T1y;
130 io[WS(os, 9)] = Ty - Tn;
131 }
132 {
133 E TB, TE, T1o, T11, T1p, TK, TP, T15, T1k, T18, T14, T16, T1l, T1m;
134 {
135 E T1e, T1j, TZ, T10, T12, T13;
136 TB = FNMS(KP866025403, TA, Tz);
137 TZ = FMA(KP866025403, TA, Tz);
138 T10 = FMA(KP866025403, TD, TC);
139 TE = FNMS(KP866025403, TD, TC);
140 T1o = FNMS(KP866025403, T1d, T1c);
141 T1e = FMA(KP866025403, T1d, T1c);
142 ro[WS(os, 9)] = T1t + T1w;
143 ro[WS(os, 3)] = T1t - T1w;
144 T1l = TZ + T10;
145 T11 = TZ - T10;
146 T1j = FMA(KP866025403, T1i, T1h);
147 T1p = FNMS(KP866025403, T1i, T1h);
148 TK = FNMS(KP866025403, TJ, TG);
149 T12 = FMA(KP866025403, TJ, TG);
150 T13 = FMA(KP866025403, TO, TL);
151 TP = FNMS(KP866025403, TO, TL);
152 TT = FNMS(KP866025403, TS, TR);
153 T15 = FMA(KP866025403, TS, TR);
154 T1m = T1e + T1j;
155 T1k = T1e - T1j;
156 T18 = T12 + T13;
157 T14 = T12 - T13;
158 T16 = FMA(KP866025403, TV, TU);
159 TW = FNMS(KP866025403, TV, TU);
160 }
161 io[WS(os, 10)] = T1l - T1m;
162 io[WS(os, 4)] = T1l + T1m;
163 io[WS(os, 7)] = T11 + T14;
164 io[WS(os, 1)] = T11 - T14;
165 {
166 E T17, T19, T1r, T1s;
167 T17 = T15 + T16;
168 T19 = T15 - T16;
169 ro[WS(os, 7)] = T19 - T1k;
170 ro[WS(os, 1)] = T19 + T1k;
171 ro[WS(os, 4)] = T17 + T18;
172 ro[WS(os, 10)] = T17 - T18;
173 T1r = TB + TE;
174 TF = TB - TE;
175 T1s = T1o + T1p;
176 T1q = T1o - T1p;
177 TY = TK + TP;
178 TQ = TK - TP;
179 io[WS(os, 2)] = T1r - T1s;
180 io[WS(os, 8)] = T1r + T1s;
181 }
182 }
183 }
184 }
185 io[WS(os, 11)] = TF + TQ;
186 io[WS(os, 5)] = TF - TQ;
187 TX = TT + TW;
188 T1n = TT - TW;
189 ro[WS(os, 11)] = T1n - T1q;
190 ro[WS(os, 5)] = T1n + T1q;
191 ro[WS(os, 8)] = TX + TY;
192 ro[WS(os, 2)] = TX - TY;
193 }
194 }
195 }
196
197 static const kdft_desc desc = { 12, "n1_12", {72, 0, 24, 0}, &GENUS, 0, 0, 0, 0 };
198
199 void X(codelet_n1_12) (planner *p) {
200 X(kdft_register) (p, n1_12, &desc);
201 }
202
203 #else /* HAVE_FMA */
204
205 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 12 -name n1_12 -include n.h */
206
207 /*
208 * This function contains 96 FP additions, 16 FP multiplications,
209 * (or, 88 additions, 8 multiplications, 8 fused multiply/add),
210 * 43 stack variables, 2 constants, and 48 memory accesses
211 */
212 #include "n.h"
213
214 static void n1_12(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
215 {
216 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
217 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
218 {
219 INT i;
220 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)) {
221 E T5, TR, TA, Ts, TS, Tz, Ta, TU, TD, Tx, TV, TC, Tg, T1a, TG;
222 E TJ, T1u, T1d, Tl, T1f, TL, TO, T1v, T1i;
223 {
224 E T1, T2, T3, T4;
225 T1 = ri[0];
226 T2 = ri[WS(is, 4)];
227 T3 = ri[WS(is, 8)];
228 T4 = T2 + T3;
229 T5 = T1 + T4;
230 TR = FNMS(KP500000000, T4, T1);
231 TA = KP866025403 * (T3 - T2);
232 }
233 {
234 E To, Tp, Tq, Tr;
235 To = ii[0];
236 Tp = ii[WS(is, 4)];
237 Tq = ii[WS(is, 8)];
238 Tr = Tp + Tq;
239 Ts = To + Tr;
240 TS = KP866025403 * (Tp - Tq);
241 Tz = FNMS(KP500000000, Tr, To);
242 }
243 {
244 E T6, T7, T8, T9;
245 T6 = ri[WS(is, 6)];
246 T7 = ri[WS(is, 10)];
247 T8 = ri[WS(is, 2)];
248 T9 = T7 + T8;
249 Ta = T6 + T9;
250 TU = FNMS(KP500000000, T9, T6);
251 TD = KP866025403 * (T8 - T7);
252 }
253 {
254 E Tt, Tu, Tv, Tw;
255 Tt = ii[WS(is, 6)];
256 Tu = ii[WS(is, 10)];
257 Tv = ii[WS(is, 2)];
258 Tw = Tu + Tv;
259 Tx = Tt + Tw;
260 TV = KP866025403 * (Tu - Tv);
261 TC = FNMS(KP500000000, Tw, Tt);
262 }
263 {
264 E Tc, Td, Te, Tf;
265 Tc = ri[WS(is, 3)];
266 Td = ri[WS(is, 7)];
267 Te = ri[WS(is, 11)];
268 Tf = Td + Te;
269 Tg = Tc + Tf;
270 T1a = KP866025403 * (Te - Td);
271 TG = FNMS(KP500000000, Tf, Tc);
272 }
273 {
274 E T1b, TH, TI, T1c;
275 T1b = ii[WS(is, 3)];
276 TH = ii[WS(is, 7)];
277 TI = ii[WS(is, 11)];
278 T1c = TH + TI;
279 TJ = KP866025403 * (TH - TI);
280 T1u = T1b + T1c;
281 T1d = FNMS(KP500000000, T1c, T1b);
282 }
283 {
284 E Th, Ti, Tj, Tk;
285 Th = ri[WS(is, 9)];
286 Ti = ri[WS(is, 1)];
287 Tj = ri[WS(is, 5)];
288 Tk = Ti + Tj;
289 Tl = Th + Tk;
290 T1f = KP866025403 * (Tj - Ti);
291 TL = FNMS(KP500000000, Tk, Th);
292 }
293 {
294 E T1g, TM, TN, T1h;
295 T1g = ii[WS(is, 9)];
296 TM = ii[WS(is, 1)];
297 TN = ii[WS(is, 5)];
298 T1h = TM + TN;
299 TO = KP866025403 * (TM - TN);
300 T1v = T1g + T1h;
301 T1i = FNMS(KP500000000, T1h, T1g);
302 }
303 {
304 E Tb, Tm, T1t, T1w;
305 Tb = T5 + Ta;
306 Tm = Tg + Tl;
307 ro[WS(os, 6)] = Tb - Tm;
308 ro[0] = Tb + Tm;
309 {
310 E T1x, T1y, Tn, Ty;
311 T1x = Ts + Tx;
312 T1y = T1u + T1v;
313 io[WS(os, 6)] = T1x - T1y;
314 io[0] = T1x + T1y;
315 Tn = Tg - Tl;
316 Ty = Ts - Tx;
317 io[WS(os, 3)] = Tn + Ty;
318 io[WS(os, 9)] = Ty - Tn;
319 }
320 T1t = T5 - Ta;
321 T1w = T1u - T1v;
322 ro[WS(os, 3)] = T1t - T1w;
323 ro[WS(os, 9)] = T1t + T1w;
324 {
325 E T11, T1l, T1k, T1m, T14, T18, T17, T19;
326 {
327 E TZ, T10, T1e, T1j;
328 TZ = TA + Tz;
329 T10 = TD + TC;
330 T11 = TZ - T10;
331 T1l = TZ + T10;
332 T1e = T1a + T1d;
333 T1j = T1f + T1i;
334 T1k = T1e - T1j;
335 T1m = T1e + T1j;
336 }
337 {
338 E T12, T13, T15, T16;
339 T12 = TG + TJ;
340 T13 = TL + TO;
341 T14 = T12 - T13;
342 T18 = T12 + T13;
343 T15 = TR + TS;
344 T16 = TU + TV;
345 T17 = T15 + T16;
346 T19 = T15 - T16;
347 }
348 io[WS(os, 1)] = T11 - T14;
349 ro[WS(os, 1)] = T19 + T1k;
350 io[WS(os, 7)] = T11 + T14;
351 ro[WS(os, 7)] = T19 - T1k;
352 ro[WS(os, 10)] = T17 - T18;
353 io[WS(os, 10)] = T1l - T1m;
354 ro[WS(os, 4)] = T17 + T18;
355 io[WS(os, 4)] = T1l + T1m;
356 }
357 {
358 E TF, T1r, T1q, T1s, TQ, TY, TX, T1n;
359 {
360 E TB, TE, T1o, T1p;
361 TB = Tz - TA;
362 TE = TC - TD;
363 TF = TB - TE;
364 T1r = TB + TE;
365 T1o = T1d - T1a;
366 T1p = T1i - T1f;
367 T1q = T1o - T1p;
368 T1s = T1o + T1p;
369 }
370 {
371 E TK, TP, TT, TW;
372 TK = TG - TJ;
373 TP = TL - TO;
374 TQ = TK - TP;
375 TY = TK + TP;
376 TT = TR - TS;
377 TW = TU - TV;
378 TX = TT + TW;
379 T1n = TT - TW;
380 }
381 io[WS(os, 5)] = TF - TQ;
382 ro[WS(os, 5)] = T1n + T1q;
383 io[WS(os, 11)] = TF + TQ;
384 ro[WS(os, 11)] = T1n - T1q;
385 ro[WS(os, 2)] = TX - TY;
386 io[WS(os, 2)] = T1r - T1s;
387 ro[WS(os, 8)] = TX + TY;
388 io[WS(os, 8)] = T1r + T1s;
389 }
390 }
391 }
392 }
393 }
394
395 static const kdft_desc desc = { 12, "n1_12", {88, 8, 8, 0}, &GENUS, 0, 0, 0, 0 };
396
397 void X(codelet_n1_12) (planner *p) {
398 X(kdft_register) (p, n1_12, &desc);
399 }
400
401 #endif /* HAVE_FMA */