Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_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 07:40:44 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 -n 8 -dit -name hc2cfdft_8 -include 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 * 55 stack variables, 2 constants, and 32 memory accesses
34 */
35 #include "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 T1A, T1w, T1z, T1x, T1H, T1v, T1L, T1F;
45 {
46 E Ty, T14, TO, T1o, Tv, TG, T16, T1m, Ta, T19, T1h, TV, T10, TX, TZ;
47 E Tk, T1i, TY, T1b, TF, TB, T1l;
48 {
49 E TH, TN, TK, TM;
50 {
51 E Tw, Tx, TI, TJ;
52 Tw = Ip[0];
53 Tx = Im[0];
54 TI = Rm[0];
55 TJ = Rp[0];
56 TH = W[0];
57 Ty = Tw - Tx;
58 TN = Tw + Tx;
59 T14 = TJ + TI;
60 TK = TI - TJ;
61 TM = W[1];
62 }
63 {
64 E Ts, Tp, Tt, Tm, Tr;
65 {
66 E Tn, To, TL, T1n;
67 Tn = Ip[WS(rs, 2)];
68 To = Im[WS(rs, 2)];
69 TL = TH * TK;
70 T1n = TM * TK;
71 Ts = Rp[WS(rs, 2)];
72 TF = Tn + To;
73 Tp = Tn - To;
74 TO = FNMS(TM, TN, TL);
75 T1o = FMA(TH, TN, T1n);
76 Tt = Rm[WS(rs, 2)];
77 }
78 Tm = W[6];
79 Tr = W[7];
80 {
81 E TE, TD, T15, TC, Tu, Tq;
82 TB = W[8];
83 TC = Tt - Ts;
84 Tu = Ts + Tt;
85 Tq = Tm * Tp;
86 TE = W[9];
87 TD = TB * TC;
88 T15 = Tm * Tu;
89 Tv = FNMS(Tr, Tu, Tq);
90 T1l = TE * TC;
91 TG = FNMS(TE, TF, TD);
92 T16 = FMA(Tr, Tp, T15);
93 }
94 }
95 }
96 {
97 E TU, TR, TT, T1g, TS;
98 {
99 E T2, T3, T7, T8;
100 T2 = Ip[WS(rs, 1)];
101 T1m = FMA(TB, TF, T1l);
102 T3 = Im[WS(rs, 1)];
103 T7 = Rp[WS(rs, 1)];
104 T8 = Rm[WS(rs, 1)];
105 {
106 E T1, T4, T9, T6, T5, TQ, T18;
107 T1 = W[2];
108 TU = T2 + T3;
109 T4 = T2 - T3;
110 TR = T7 - T8;
111 T9 = T7 + T8;
112 T6 = W[3];
113 T5 = T1 * T4;
114 TQ = W[4];
115 T18 = T1 * T9;
116 TT = W[5];
117 Ta = FNMS(T6, T9, T5);
118 T1g = TQ * TU;
119 TS = TQ * TR;
120 T19 = FMA(T6, T4, T18);
121 }
122 }
123 {
124 E Tc, Td, Th, Ti;
125 Tc = Ip[WS(rs, 3)];
126 T1h = FNMS(TT, TR, T1g);
127 TV = FMA(TT, TU, TS);
128 Td = Im[WS(rs, 3)];
129 Th = Rp[WS(rs, 3)];
130 Ti = Rm[WS(rs, 3)];
131 {
132 E Tb, Te, Tj, Tg, Tf, TW, T1a;
133 Tb = W[10];
134 T10 = Tc + Td;
135 Te = Tc - Td;
136 TX = Th - Ti;
137 Tj = Th + Ti;
138 Tg = W[11];
139 Tf = Tb * Te;
140 TW = W[12];
141 T1a = Tb * Tj;
142 TZ = W[13];
143 Tk = FNMS(Tg, Tj, Tf);
144 T1i = TW * T10;
145 TY = TW * TX;
146 T1b = FMA(Tg, Te, T1a);
147 }
148 }
149 }
150 {
151 E T1E, T1t, TA, T1s, T1D, T1u, T1e, T13, T1r, T1d;
152 {
153 E TP, T1f, T1q, T12, T17, T1c;
154 {
155 E Tl, T11, Tz, T1p, T1k, T1j;
156 T1E = Ta - Tk;
157 Tl = Ta + Tk;
158 T1j = FNMS(TZ, TX, T1i);
159 T11 = FMA(TZ, T10, TY);
160 Tz = Tv + Ty;
161 T1t = Ty - Tv;
162 T1A = T1o - T1m;
163 T1p = T1m + T1o;
164 T1k = T1h + T1j;
165 T1w = T1j - T1h;
166 T1z = TO - TG;
167 TP = TG + TO;
168 T1f = Tz - Tl;
169 TA = Tl + Tz;
170 T1s = T1k + T1p;
171 T1q = T1k - T1p;
172 T12 = TV + T11;
173 T1x = TV - T11;
174 T1D = T14 - T16;
175 T17 = T14 + T16;
176 T1c = T19 + T1b;
177 T1u = T19 - T1b;
178 }
179 Im[WS(rs, 1)] = KP500000000 * (T1q - T1f);
180 T1e = T12 + TP;
181 T13 = TP - T12;
182 T1r = T17 + T1c;
183 T1d = T17 - T1c;
184 Ip[WS(rs, 2)] = KP500000000 * (T1f + T1q);
185 }
186 Im[WS(rs, 3)] = KP500000000 * (T13 - TA);
187 Ip[0] = KP500000000 * (TA + T13);
188 Rm[WS(rs, 3)] = KP500000000 * (T1r - T1s);
189 Rp[0] = KP500000000 * (T1r + T1s);
190 Rp[WS(rs, 2)] = KP500000000 * (T1d + T1e);
191 Rm[WS(rs, 1)] = KP500000000 * (T1d - T1e);
192 T1H = T1u + T1t;
193 T1v = T1t - T1u;
194 T1L = T1D + T1E;
195 T1F = T1D - T1E;
196 }
197 }
198 {
199 E T1y, T1I, T1B, T1J;
200 T1y = T1w + T1x;
201 T1I = T1w - T1x;
202 T1B = T1z - T1A;
203 T1J = T1z + T1A;
204 {
205 E T1M, T1K, T1C, T1G;
206 T1M = T1I + T1J;
207 T1K = T1I - T1J;
208 T1C = T1y + T1B;
209 T1G = T1B - T1y;
210 Im[0] = -(KP500000000 * (FNMS(KP707106781, T1K, T1H)));
211 Ip[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1K, T1H));
212 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1M, T1L));
213 Rm[WS(rs, 2)] = KP500000000 * (FNMS(KP707106781, T1M, T1L));
214 Rp[WS(rs, 3)] = KP500000000 * (FMA(KP707106781, T1G, T1F));
215 Rm[0] = KP500000000 * (FNMS(KP707106781, T1G, T1F));
216 Im[WS(rs, 2)] = -(KP500000000 * (FNMS(KP707106781, T1C, T1v)));
217 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP707106781, T1C, T1v));
218 }
219 }
220 }
221 }
222 }
223
224 static const tw_instr twinstr[] = {
225 {TW_FULL, 1, 8},
226 {TW_NEXT, 1, 0}
227 };
228
229 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {60, 30, 22, 0} };
230
231 void X(codelet_hc2cfdft_8) (planner *p) {
232 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
233 }
234 #else /* HAVE_FMA */
235
236 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 8 -dit -name hc2cfdft_8 -include hc2cf.h */
237
238 /*
239 * This function contains 82 FP additions, 44 FP multiplications,
240 * (or, 68 additions, 30 multiplications, 14 fused multiply/add),
241 * 39 stack variables, 2 constants, and 32 memory accesses
242 */
243 #include "hc2cf.h"
244
245 static void hc2cfdft_8(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
246 {
247 DK(KP353553390, +0.353553390593273762200422181052424519642417969);
248 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
249 {
250 INT m;
251 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)) {
252 E Tv, TX, Ts, TY, TE, T1a, TJ, T19, T1l, T1m, T9, T10, Ti, T11, TP;
253 E T16, TU, T17, T1i, T1j;
254 {
255 E Tt, Tu, TD, Tz, TA, TB, Tn, TI, Tr, TG, Tk, To;
256 Tt = Ip[0];
257 Tu = Im[0];
258 TD = Tt + Tu;
259 Tz = Rm[0];
260 TA = Rp[0];
261 TB = Tz - TA;
262 {
263 E Tl, Tm, Tp, Tq;
264 Tl = Ip[WS(rs, 2)];
265 Tm = Im[WS(rs, 2)];
266 Tn = Tl - Tm;
267 TI = Tl + Tm;
268 Tp = Rp[WS(rs, 2)];
269 Tq = Rm[WS(rs, 2)];
270 Tr = Tp + Tq;
271 TG = Tp - Tq;
272 }
273 Tv = Tt - Tu;
274 TX = TA + Tz;
275 Tk = W[6];
276 To = W[7];
277 Ts = FNMS(To, Tr, Tk * Tn);
278 TY = FMA(Tk, Tr, To * Tn);
279 {
280 E Ty, TC, TF, TH;
281 Ty = W[0];
282 TC = W[1];
283 TE = FNMS(TC, TD, Ty * TB);
284 T1a = FMA(TC, TB, Ty * TD);
285 TF = W[8];
286 TH = W[9];
287 TJ = FMA(TF, TG, TH * TI);
288 T19 = FNMS(TH, TG, TF * TI);
289 }
290 T1l = TJ + TE;
291 T1m = T1a - T19;
292 }
293 {
294 E T4, TO, T8, TM, Td, TT, Th, TR;
295 {
296 E T2, T3, T6, T7;
297 T2 = Ip[WS(rs, 1)];
298 T3 = Im[WS(rs, 1)];
299 T4 = T2 - T3;
300 TO = T2 + T3;
301 T6 = Rp[WS(rs, 1)];
302 T7 = Rm[WS(rs, 1)];
303 T8 = T6 + T7;
304 TM = T6 - T7;
305 }
306 {
307 E Tb, Tc, Tf, Tg;
308 Tb = Ip[WS(rs, 3)];
309 Tc = Im[WS(rs, 3)];
310 Td = Tb - Tc;
311 TT = Tb + Tc;
312 Tf = Rp[WS(rs, 3)];
313 Tg = Rm[WS(rs, 3)];
314 Th = Tf + Tg;
315 TR = Tf - Tg;
316 }
317 {
318 E T1, T5, Ta, Te;
319 T1 = W[2];
320 T5 = W[3];
321 T9 = FNMS(T5, T8, T1 * T4);
322 T10 = FMA(T1, T8, T5 * T4);
323 Ta = W[10];
324 Te = W[11];
325 Ti = FNMS(Te, Th, Ta * Td);
326 T11 = FMA(Ta, Th, Te * Td);
327 {
328 E TL, TN, TQ, TS;
329 TL = W[4];
330 TN = W[5];
331 TP = FMA(TL, TM, TN * TO);
332 T16 = FNMS(TN, TM, TL * TO);
333 TQ = W[12];
334 TS = W[13];
335 TU = FMA(TQ, TR, TS * TT);
336 T17 = FNMS(TS, TR, TQ * TT);
337 }
338 T1i = T17 - T16;
339 T1j = TP - TU;
340 }
341 }
342 {
343 E T1h, T1t, T1w, T1y, T1o, T1s, T1r, T1x;
344 {
345 E T1f, T1g, T1u, T1v;
346 T1f = Tv - Ts;
347 T1g = T10 - T11;
348 T1h = KP500000000 * (T1f - T1g);
349 T1t = KP500000000 * (T1g + T1f);
350 T1u = T1i - T1j;
351 T1v = T1l + T1m;
352 T1w = KP353553390 * (T1u - T1v);
353 T1y = KP353553390 * (T1u + T1v);
354 }
355 {
356 E T1k, T1n, T1p, T1q;
357 T1k = T1i + T1j;
358 T1n = T1l - T1m;
359 T1o = KP353553390 * (T1k + T1n);
360 T1s = KP353553390 * (T1n - T1k);
361 T1p = TX - TY;
362 T1q = T9 - Ti;
363 T1r = KP500000000 * (T1p - T1q);
364 T1x = KP500000000 * (T1p + T1q);
365 }
366 Ip[WS(rs, 1)] = T1h + T1o;
367 Rp[WS(rs, 1)] = T1x + T1y;
368 Im[WS(rs, 2)] = T1o - T1h;
369 Rm[WS(rs, 2)] = T1x - T1y;
370 Rm[0] = T1r - T1s;
371 Im[0] = T1w - T1t;
372 Rp[WS(rs, 3)] = T1r + T1s;
373 Ip[WS(rs, 3)] = T1t + T1w;
374 }
375 {
376 E Tx, T15, T1c, T1e, TW, T14, T13, T1d;
377 {
378 E Tj, Tw, T18, T1b;
379 Tj = T9 + Ti;
380 Tw = Ts + Tv;
381 Tx = Tj + Tw;
382 T15 = Tw - Tj;
383 T18 = T16 + T17;
384 T1b = T19 + T1a;
385 T1c = T18 - T1b;
386 T1e = T18 + T1b;
387 }
388 {
389 E TK, TV, TZ, T12;
390 TK = TE - TJ;
391 TV = TP + TU;
392 TW = TK - TV;
393 T14 = TV + TK;
394 TZ = TX + TY;
395 T12 = T10 + T11;
396 T13 = TZ - T12;
397 T1d = TZ + T12;
398 }
399 Ip[0] = KP500000000 * (Tx + TW);
400 Rp[0] = KP500000000 * (T1d + T1e);
401 Im[WS(rs, 3)] = KP500000000 * (TW - Tx);
402 Rm[WS(rs, 3)] = KP500000000 * (T1d - T1e);
403 Rm[WS(rs, 1)] = KP500000000 * (T13 - T14);
404 Im[WS(rs, 1)] = KP500000000 * (T1c - T15);
405 Rp[WS(rs, 2)] = KP500000000 * (T13 + T14);
406 Ip[WS(rs, 2)] = KP500000000 * (T15 + T1c);
407 }
408 }
409 }
410 }
411
412 static const tw_instr twinstr[] = {
413 {TW_FULL, 1, 8},
414 {TW_NEXT, 1, 0}
415 };
416
417 static const hc2c_desc desc = { 8, "hc2cfdft_8", twinstr, &GENUS, {68, 30, 14, 0} };
418
419 void X(codelet_hc2cfdft_8) (planner *p) {
420 X(khc2c_register) (p, hc2cfdft_8, &desc, HC2C_VIA_DFT);
421 }
422 #endif /* HAVE_FMA */