Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cb/hc2cbdft_8.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 07:42:04 EST 2012 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cbdft_8 -include hc2cb.h */
29
30 /*
31 * This function contains 82 FP additions, 36 FP multiplications,
32 * (or, 60 additions, 14 multiplications, 22 fused multiply/add),
33 * 55 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "hc2cb.h"
36
37 static void hc2cbdft_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 {
41 INT m;
42 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)) {
43 E T1m, T1r, T1i, T1u, T1o, T1v, T1n, T1w, T1s;
44 {
45 E T1k, Tl, T1p, TE, TP, T1g, TM, T1b, T1f, T1a, TU, Tf, T1l, TH, Tw;
46 E T1q;
47 {
48 E TA, T3, TN, Tk, Th, T6, TO, TD, Tb, Tm, Ta, TK, Tp, Tc, Ts;
49 E Tt;
50 {
51 E T4, T5, TB, TC;
52 {
53 E T1, T2, Ti, Tj;
54 T1 = Rp[0];
55 T2 = Rm[WS(rs, 3)];
56 Ti = Ip[0];
57 Tj = Im[WS(rs, 3)];
58 T4 = Rp[WS(rs, 2)];
59 TA = T1 - T2;
60 T3 = T1 + T2;
61 TN = Ti - Tj;
62 Tk = Ti + Tj;
63 T5 = Rm[WS(rs, 1)];
64 TB = Ip[WS(rs, 2)];
65 TC = Im[WS(rs, 1)];
66 }
67 {
68 E T8, T9, Tn, To;
69 T8 = Rp[WS(rs, 1)];
70 Th = T4 - T5;
71 T6 = T4 + T5;
72 TO = TB - TC;
73 TD = TB + TC;
74 T9 = Rm[WS(rs, 2)];
75 Tn = Ip[WS(rs, 1)];
76 To = Im[WS(rs, 2)];
77 Tb = Rm[0];
78 Tm = T8 - T9;
79 Ta = T8 + T9;
80 TK = Tn - To;
81 Tp = Tn + To;
82 Tc = Rp[WS(rs, 3)];
83 Ts = Im[0];
84 Tt = Ip[WS(rs, 3)];
85 }
86 }
87 {
88 E Tr, Td, Tu, TL, Te, T7;
89 T1k = Tk - Th;
90 Tl = Th + Tk;
91 Tr = Tb - Tc;
92 Td = Tb + Tc;
93 TL = Tt - Ts;
94 Tu = Ts + Tt;
95 T1p = TA + TD;
96 TE = TA - TD;
97 TP = TN + TO;
98 T1g = TN - TO;
99 TM = TK + TL;
100 T1b = TL - TK;
101 T1f = Ta - Td;
102 Te = Ta + Td;
103 T1a = T3 - T6;
104 T7 = T3 + T6;
105 {
106 E Tq, TF, TG, Tv;
107 Tq = Tm + Tp;
108 TF = Tm - Tp;
109 TG = Tr - Tu;
110 Tv = Tr + Tu;
111 TU = T7 - Te;
112 Tf = T7 + Te;
113 T1l = TF - TG;
114 TH = TF + TG;
115 Tw = Tq - Tv;
116 T1q = Tq + Tv;
117 }
118 }
119 }
120 {
121 E TX, T10, T1c, T13, T1h, T1E, T1H, T1C, T1K, T1G, T1L, T1F;
122 {
123 E TQ, Tx, T1y, TI, Tg, Tz;
124 TX = TP - TM;
125 TQ = TM + TP;
126 Tx = FMA(KP707106781, Tw, Tl);
127 T10 = FNMS(KP707106781, Tw, Tl);
128 T1c = T1a + T1b;
129 T1y = T1a - T1b;
130 T13 = FNMS(KP707106781, TH, TE);
131 TI = FMA(KP707106781, TH, TE);
132 Tg = W[0];
133 Tz = W[1];
134 {
135 E T1B, T1A, T1x, T1J, T1z, T1D;
136 {
137 E TR, Ty, TS, TJ;
138 T1B = T1g - T1f;
139 T1h = T1f + T1g;
140 T1A = W[11];
141 TR = Tg * TI;
142 Ty = Tg * Tx;
143 T1x = W[10];
144 T1J = T1A * T1y;
145 TS = FNMS(Tz, Tx, TR);
146 TJ = FMA(Tz, TI, Ty);
147 T1z = T1x * T1y;
148 T1m = FMA(KP707106781, T1l, T1k);
149 T1E = FNMS(KP707106781, T1l, T1k);
150 Im[0] = TS - TQ;
151 Ip[0] = TQ + TS;
152 Rm[0] = Tf + TJ;
153 Rp[0] = Tf - TJ;
154 T1H = FMA(KP707106781, T1q, T1p);
155 T1r = FNMS(KP707106781, T1q, T1p);
156 T1D = W[12];
157 }
158 T1C = FNMS(T1A, T1B, T1z);
159 T1K = FMA(T1x, T1B, T1J);
160 T1G = W[13];
161 T1L = T1D * T1H;
162 T1F = T1D * T1E;
163 }
164 }
165 {
166 E TY, T16, T12, T17, T11;
167 {
168 E TW, TT, T15, TV, TZ, T1M, T1I;
169 TW = W[7];
170 T1M = FNMS(T1G, T1E, T1L);
171 T1I = FMA(T1G, T1H, T1F);
172 TT = W[6];
173 T15 = TW * TU;
174 Im[WS(rs, 3)] = T1M - T1K;
175 Ip[WS(rs, 3)] = T1K + T1M;
176 Rm[WS(rs, 3)] = T1C + T1I;
177 Rp[WS(rs, 3)] = T1C - T1I;
178 TV = TT * TU;
179 TZ = W[8];
180 TY = FNMS(TW, TX, TV);
181 T16 = FMA(TT, TX, T15);
182 T12 = W[9];
183 T17 = TZ * T13;
184 T11 = TZ * T10;
185 }
186 {
187 E T1e, T19, T1t, T1d, T1j, T18, T14;
188 T1e = W[3];
189 T18 = FNMS(T12, T10, T17);
190 T14 = FMA(T12, T13, T11);
191 T19 = W[2];
192 T1t = T1e * T1c;
193 Im[WS(rs, 2)] = T18 - T16;
194 Ip[WS(rs, 2)] = T16 + T18;
195 Rm[WS(rs, 2)] = TY + T14;
196 Rp[WS(rs, 2)] = TY - T14;
197 T1d = T19 * T1c;
198 T1j = W[4];
199 T1i = FNMS(T1e, T1h, T1d);
200 T1u = FMA(T19, T1h, T1t);
201 T1o = W[5];
202 T1v = T1j * T1r;
203 T1n = T1j * T1m;
204 }
205 }
206 }
207 }
208 T1w = FNMS(T1o, T1m, T1v);
209 T1s = FMA(T1o, T1r, T1n);
210 Im[WS(rs, 1)] = T1w - T1u;
211 Ip[WS(rs, 1)] = T1u + T1w;
212 Rm[WS(rs, 1)] = T1i + T1s;
213 Rp[WS(rs, 1)] = T1i - T1s;
214 }
215 }
216 }
217
218 static const tw_instr twinstr[] = {
219 {TW_FULL, 1, 8},
220 {TW_NEXT, 1, 0}
221 };
222
223 static const hc2c_desc desc = { 8, "hc2cbdft_8", twinstr, &GENUS, {60, 14, 22, 0} };
224
225 void X(codelet_hc2cbdft_8) (planner *p) {
226 X(khc2c_register) (p, hc2cbdft_8, &desc, HC2C_VIA_DFT);
227 }
228 #else /* HAVE_FMA */
229
230 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 8 -dif -name hc2cbdft_8 -include hc2cb.h */
231
232 /*
233 * This function contains 82 FP additions, 32 FP multiplications,
234 * (or, 68 additions, 18 multiplications, 14 fused multiply/add),
235 * 30 stack variables, 1 constants, and 32 memory accesses
236 */
237 #include "hc2cb.h"
238
239 static void hc2cbdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
240 {
241 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
242 {
243 INT m;
244 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)) {
245 E T7, T1d, T1h, Tl, TG, T14, T19, TO, Te, TL, T18, T15, TB, T1e, Tw;
246 E T1i;
247 {
248 E T3, TC, Tk, TM, T6, Th, TF, TN;
249 {
250 E T1, T2, Ti, Tj;
251 T1 = Rp[0];
252 T2 = Rm[WS(rs, 3)];
253 T3 = T1 + T2;
254 TC = T1 - T2;
255 Ti = Ip[0];
256 Tj = Im[WS(rs, 3)];
257 Tk = Ti + Tj;
258 TM = Ti - Tj;
259 }
260 {
261 E T4, T5, TD, TE;
262 T4 = Rp[WS(rs, 2)];
263 T5 = Rm[WS(rs, 1)];
264 T6 = T4 + T5;
265 Th = T4 - T5;
266 TD = Ip[WS(rs, 2)];
267 TE = Im[WS(rs, 1)];
268 TF = TD + TE;
269 TN = TD - TE;
270 }
271 T7 = T3 + T6;
272 T1d = Tk - Th;
273 T1h = TC + TF;
274 Tl = Th + Tk;
275 TG = TC - TF;
276 T14 = T3 - T6;
277 T19 = TM - TN;
278 TO = TM + TN;
279 }
280 {
281 E Ta, Tm, Tp, TJ, Td, Tr, Tu, TK;
282 {
283 E T8, T9, Tn, To;
284 T8 = Rp[WS(rs, 1)];
285 T9 = Rm[WS(rs, 2)];
286 Ta = T8 + T9;
287 Tm = T8 - T9;
288 Tn = Ip[WS(rs, 1)];
289 To = Im[WS(rs, 2)];
290 Tp = Tn + To;
291 TJ = Tn - To;
292 }
293 {
294 E Tb, Tc, Ts, Tt;
295 Tb = Rm[0];
296 Tc = Rp[WS(rs, 3)];
297 Td = Tb + Tc;
298 Tr = Tb - Tc;
299 Ts = Im[0];
300 Tt = Ip[WS(rs, 3)];
301 Tu = Ts + Tt;
302 TK = Tt - Ts;
303 }
304 Te = Ta + Td;
305 TL = TJ + TK;
306 T18 = Ta - Td;
307 T15 = TK - TJ;
308 {
309 E Tz, TA, Tq, Tv;
310 Tz = Tm - Tp;
311 TA = Tr - Tu;
312 TB = KP707106781 * (Tz + TA);
313 T1e = KP707106781 * (Tz - TA);
314 Tq = Tm + Tp;
315 Tv = Tr + Tu;
316 Tw = KP707106781 * (Tq - Tv);
317 T1i = KP707106781 * (Tq + Tv);
318 }
319 }
320 {
321 E Tf, TP, TI, TQ;
322 Tf = T7 + Te;
323 TP = TL + TO;
324 {
325 E Tx, TH, Tg, Ty;
326 Tx = Tl + Tw;
327 TH = TB + TG;
328 Tg = W[0];
329 Ty = W[1];
330 TI = FMA(Tg, Tx, Ty * TH);
331 TQ = FNMS(Ty, Tx, Tg * TH);
332 }
333 Rp[0] = Tf - TI;
334 Ip[0] = TP + TQ;
335 Rm[0] = Tf + TI;
336 Im[0] = TQ - TP;
337 }
338 {
339 E T1r, T1x, T1w, T1y;
340 {
341 E T1o, T1q, T1n, T1p;
342 T1o = T14 - T15;
343 T1q = T19 - T18;
344 T1n = W[10];
345 T1p = W[11];
346 T1r = FNMS(T1p, T1q, T1n * T1o);
347 T1x = FMA(T1p, T1o, T1n * T1q);
348 }
349 {
350 E T1t, T1v, T1s, T1u;
351 T1t = T1d - T1e;
352 T1v = T1i + T1h;
353 T1s = W[12];
354 T1u = W[13];
355 T1w = FMA(T1s, T1t, T1u * T1v);
356 T1y = FNMS(T1u, T1t, T1s * T1v);
357 }
358 Rp[WS(rs, 3)] = T1r - T1w;
359 Ip[WS(rs, 3)] = T1x + T1y;
360 Rm[WS(rs, 3)] = T1r + T1w;
361 Im[WS(rs, 3)] = T1y - T1x;
362 }
363 {
364 E TV, T11, T10, T12;
365 {
366 E TS, TU, TR, TT;
367 TS = T7 - Te;
368 TU = TO - TL;
369 TR = W[6];
370 TT = W[7];
371 TV = FNMS(TT, TU, TR * TS);
372 T11 = FMA(TT, TS, TR * TU);
373 }
374 {
375 E TX, TZ, TW, TY;
376 TX = Tl - Tw;
377 TZ = TG - TB;
378 TW = W[8];
379 TY = W[9];
380 T10 = FMA(TW, TX, TY * TZ);
381 T12 = FNMS(TY, TX, TW * TZ);
382 }
383 Rp[WS(rs, 2)] = TV - T10;
384 Ip[WS(rs, 2)] = T11 + T12;
385 Rm[WS(rs, 2)] = TV + T10;
386 Im[WS(rs, 2)] = T12 - T11;
387 }
388 {
389 E T1b, T1l, T1k, T1m;
390 {
391 E T16, T1a, T13, T17;
392 T16 = T14 + T15;
393 T1a = T18 + T19;
394 T13 = W[2];
395 T17 = W[3];
396 T1b = FNMS(T17, T1a, T13 * T16);
397 T1l = FMA(T17, T16, T13 * T1a);
398 }
399 {
400 E T1f, T1j, T1c, T1g;
401 T1f = T1d + T1e;
402 T1j = T1h - T1i;
403 T1c = W[4];
404 T1g = W[5];
405 T1k = FMA(T1c, T1f, T1g * T1j);
406 T1m = FNMS(T1g, T1f, T1c * T1j);
407 }
408 Rp[WS(rs, 1)] = T1b - T1k;
409 Ip[WS(rs, 1)] = T1l + T1m;
410 Rm[WS(rs, 1)] = T1b + T1k;
411 Im[WS(rs, 1)] = T1m - T1l;
412 }
413 }
414 }
415 }
416
417 static const tw_instr twinstr[] = {
418 {TW_FULL, 1, 8},
419 {TW_NEXT, 1, 0}
420 };
421
422 static const hc2c_desc desc = { 8, "hc2cbdft_8", twinstr, &GENUS, {68, 18, 14, 0} };
423
424 void X(codelet_hc2cbdft_8) (planner *p) {
425 X(khc2c_register) (p, hc2cbdft_8, &desc, HC2C_VIA_DFT);
426 }
427 #endif /* HAVE_FMA */