Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t2_8.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
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 07:35:59 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include t.h */
29
30 /*
31 * This function contains 74 FP additions, 50 FP multiplications,
32 * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
33 * 64 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "t.h"
36
37 static void t2_8(R *ri, R *ii, 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 * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
43 E TS, T1m, TJ, T1l, T1k, Tw, T1w, T1u;
44 {
45 E T2, T3, Tl, Tn, T5, T4, Tm, Tr, T6;
46 T2 = W[0];
47 T3 = W[2];
48 Tl = W[4];
49 Tn = W[5];
50 T5 = W[1];
51 T4 = T2 * T3;
52 Tm = T2 * Tl;
53 Tr = T2 * Tn;
54 T6 = W[3];
55 {
56 E T1, T1s, TG, Td, T1r, Tu, TY, Tk, TW, T18, T1d, TD, TH, TA, T13;
57 E TE, T14;
58 {
59 E To, Ts, Tf, T7, T8, Ti, Tb, T9, Tc, TC, Ta, TF, TB, Tg, Th;
60 E Tj;
61 T1 = ri[0];
62 To = FMA(T5, Tn, Tm);
63 Ts = FNMS(T5, Tl, Tr);
64 Tf = FMA(T5, T6, T4);
65 T7 = FNMS(T5, T6, T4);
66 Ta = T2 * T6;
67 T1s = ii[0];
68 T8 = ri[WS(rs, 4)];
69 TF = Tf * Tn;
70 TB = Tf * Tl;
71 Ti = FNMS(T5, T3, Ta);
72 Tb = FMA(T5, T3, Ta);
73 T9 = T7 * T8;
74 Tc = ii[WS(rs, 4)];
75 TG = FNMS(Ti, Tl, TF);
76 TC = FMA(Ti, Tn, TB);
77 {
78 E Tp, T1q, Tt, Tq, TX;
79 Tp = ri[WS(rs, 6)];
80 Td = FMA(Tb, Tc, T9);
81 T1q = T7 * Tc;
82 Tt = ii[WS(rs, 6)];
83 Tq = To * Tp;
84 Tg = ri[WS(rs, 2)];
85 T1r = FNMS(Tb, T8, T1q);
86 TX = To * Tt;
87 Tu = FMA(Ts, Tt, Tq);
88 Th = Tf * Tg;
89 Tj = ii[WS(rs, 2)];
90 TY = FNMS(Ts, Tp, TX);
91 }
92 {
93 E TO, TQ, TN, TP, T1a, T1b;
94 {
95 E TK, TM, TL, T19, TV;
96 TK = ri[WS(rs, 7)];
97 TM = ii[WS(rs, 7)];
98 Tk = FMA(Ti, Tj, Th);
99 TV = Tf * Tj;
100 TL = Tl * TK;
101 T19 = Tl * TM;
102 TO = ri[WS(rs, 3)];
103 TW = FNMS(Ti, Tg, TV);
104 TQ = ii[WS(rs, 3)];
105 TN = FMA(Tn, TM, TL);
106 TP = T3 * TO;
107 T1a = FNMS(Tn, TK, T19);
108 T1b = T3 * TQ;
109 }
110 {
111 E Tx, Tz, Ty, T12, T1c, TR;
112 Tx = ri[WS(rs, 1)];
113 TR = FMA(T6, TQ, TP);
114 Tz = ii[WS(rs, 1)];
115 T1c = FNMS(T6, TO, T1b);
116 Ty = T2 * Tx;
117 T18 = TN - TR;
118 TS = TN + TR;
119 T12 = T2 * Tz;
120 T1d = T1a - T1c;
121 T1m = T1a + T1c;
122 TD = ri[WS(rs, 5)];
123 TH = ii[WS(rs, 5)];
124 TA = FMA(T5, Tz, Ty);
125 T13 = FNMS(T5, Tx, T12);
126 TE = TC * TD;
127 T14 = TC * TH;
128 }
129 }
130 }
131 {
132 E Te, T1p, T1t, Tv;
133 {
134 E T1g, T10, T1z, T1B, T1A, T1j, T1C, T1f;
135 {
136 E T1x, T11, T16, T1y;
137 {
138 E TU, TZ, TI, T15;
139 Te = T1 + Td;
140 TU = T1 - Td;
141 TZ = TW - TY;
142 T1p = TW + TY;
143 TI = FMA(TG, TH, TE);
144 T15 = FNMS(TG, TD, T14);
145 T1t = T1r + T1s;
146 T1x = T1s - T1r;
147 T1g = TU - TZ;
148 T10 = TU + TZ;
149 T11 = TA - TI;
150 TJ = TA + TI;
151 T1l = T13 + T15;
152 T16 = T13 - T15;
153 T1y = Tk - Tu;
154 Tv = Tk + Tu;
155 }
156 {
157 E T1i, T1e, T17, T1h;
158 T1i = T18 + T1d;
159 T1e = T18 - T1d;
160 T17 = T11 + T16;
161 T1h = T16 - T11;
162 T1z = T1x - T1y;
163 T1B = T1y + T1x;
164 T1A = T1h + T1i;
165 T1j = T1h - T1i;
166 T1C = T1e - T17;
167 T1f = T17 + T1e;
168 }
169 }
170 ri[WS(rs, 7)] = FNMS(KP707106781, T1j, T1g);
171 ii[WS(rs, 7)] = FNMS(KP707106781, T1C, T1B);
172 ri[WS(rs, 1)] = FMA(KP707106781, T1f, T10);
173 ri[WS(rs, 5)] = FNMS(KP707106781, T1f, T10);
174 ii[WS(rs, 1)] = FMA(KP707106781, T1A, T1z);
175 ii[WS(rs, 5)] = FNMS(KP707106781, T1A, T1z);
176 ri[WS(rs, 3)] = FMA(KP707106781, T1j, T1g);
177 ii[WS(rs, 3)] = FMA(KP707106781, T1C, T1B);
178 }
179 T1k = Te - Tv;
180 Tw = Te + Tv;
181 T1w = T1t - T1p;
182 T1u = T1p + T1t;
183 }
184 }
185 }
186 {
187 E TT, T1v, T1n, T1o;
188 TT = TJ + TS;
189 T1v = TS - TJ;
190 T1n = T1l - T1m;
191 T1o = T1l + T1m;
192 ii[WS(rs, 2)] = T1v + T1w;
193 ii[WS(rs, 6)] = T1w - T1v;
194 ri[0] = Tw + TT;
195 ri[WS(rs, 4)] = Tw - TT;
196 ii[0] = T1o + T1u;
197 ii[WS(rs, 4)] = T1u - T1o;
198 ri[WS(rs, 2)] = T1k + T1n;
199 ri[WS(rs, 6)] = T1k - T1n;
200 }
201 }
202 }
203 }
204
205 static const tw_instr twinstr[] = {
206 {TW_CEXP, 0, 1},
207 {TW_CEXP, 0, 3},
208 {TW_CEXP, 0, 7},
209 {TW_NEXT, 1, 0}
210 };
211
212 static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
213
214 void X(codelet_t2_8) (planner *p) {
215 X(kdft_dit_register) (p, t2_8, &desc);
216 }
217 #else /* HAVE_FMA */
218
219 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -name t2_8 -include t.h */
220
221 /*
222 * This function contains 74 FP additions, 44 FP multiplications,
223 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
224 * 42 stack variables, 1 constants, and 32 memory accesses
225 */
226 #include "t.h"
227
228 static void t2_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
229 {
230 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
231 {
232 INT m;
233 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
234 E T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
235 {
236 E T4, Tb, T7, Ta;
237 T2 = W[0];
238 T5 = W[1];
239 T3 = W[2];
240 T6 = W[3];
241 T4 = T2 * T3;
242 Tb = T5 * T3;
243 T7 = T5 * T6;
244 Ta = T2 * T6;
245 T8 = T4 - T7;
246 Tc = Ta + Tb;
247 Tg = T4 + T7;
248 Ti = Ta - Tb;
249 Tl = W[4];
250 Tm = W[5];
251 Tn = FMA(T2, Tl, T5 * Tm);
252 Tz = FNMS(Ti, Tl, Tg * Tm);
253 Tp = FNMS(T5, Tl, T2 * Tm);
254 Tx = FMA(Tg, Tl, Ti * Tm);
255 }
256 {
257 E Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
258 E TT;
259 {
260 E T1, T1c, Te, T1b, T9, Td;
261 T1 = ri[0];
262 T1c = ii[0];
263 T9 = ri[WS(rs, 4)];
264 Td = ii[WS(rs, 4)];
265 Te = FMA(T8, T9, Tc * Td);
266 T1b = FNMS(Tc, T9, T8 * Td);
267 Tf = T1 + Te;
268 T1i = T1c - T1b;
269 TL = T1 - Te;
270 T1d = T1b + T1c;
271 }
272 {
273 E TF, TW, TI, TX;
274 {
275 E TD, TE, TG, TH;
276 TD = ri[WS(rs, 7)];
277 TE = ii[WS(rs, 7)];
278 TF = FMA(Tl, TD, Tm * TE);
279 TW = FNMS(Tm, TD, Tl * TE);
280 TG = ri[WS(rs, 3)];
281 TH = ii[WS(rs, 3)];
282 TI = FMA(T3, TG, T6 * TH);
283 TX = FNMS(T6, TG, T3 * TH);
284 }
285 TJ = TF + TI;
286 T17 = TW + TX;
287 TV = TF - TI;
288 TY = TW - TX;
289 }
290 {
291 E Tk, TM, Tr, TN;
292 {
293 E Th, Tj, To, Tq;
294 Th = ri[WS(rs, 2)];
295 Tj = ii[WS(rs, 2)];
296 Tk = FMA(Tg, Th, Ti * Tj);
297 TM = FNMS(Ti, Th, Tg * Tj);
298 To = ri[WS(rs, 6)];
299 Tq = ii[WS(rs, 6)];
300 Tr = FMA(Tn, To, Tp * Tq);
301 TN = FNMS(Tp, To, Tn * Tq);
302 }
303 Ts = Tk + Tr;
304 T1j = Tk - Tr;
305 TO = TM - TN;
306 T1a = TM + TN;
307 }
308 {
309 E Tw, TR, TB, TS;
310 {
311 E Tu, Tv, Ty, TA;
312 Tu = ri[WS(rs, 1)];
313 Tv = ii[WS(rs, 1)];
314 Tw = FMA(T2, Tu, T5 * Tv);
315 TR = FNMS(T5, Tu, T2 * Tv);
316 Ty = ri[WS(rs, 5)];
317 TA = ii[WS(rs, 5)];
318 TB = FMA(Tx, Ty, Tz * TA);
319 TS = FNMS(Tz, Ty, Tx * TA);
320 }
321 TC = Tw + TB;
322 T16 = TR + TS;
323 TQ = Tw - TB;
324 TT = TR - TS;
325 }
326 {
327 E Tt, TK, T1f, T1g;
328 Tt = Tf + Ts;
329 TK = TC + TJ;
330 ri[WS(rs, 4)] = Tt - TK;
331 ri[0] = Tt + TK;
332 {
333 E T19, T1e, T15, T18;
334 T19 = T16 + T17;
335 T1e = T1a + T1d;
336 ii[0] = T19 + T1e;
337 ii[WS(rs, 4)] = T1e - T19;
338 T15 = Tf - Ts;
339 T18 = T16 - T17;
340 ri[WS(rs, 6)] = T15 - T18;
341 ri[WS(rs, 2)] = T15 + T18;
342 }
343 T1f = TJ - TC;
344 T1g = T1d - T1a;
345 ii[WS(rs, 2)] = T1f + T1g;
346 ii[WS(rs, 6)] = T1g - T1f;
347 {
348 E T11, T1k, T14, T1h, T12, T13;
349 T11 = TL - TO;
350 T1k = T1i - T1j;
351 T12 = TT - TQ;
352 T13 = TV + TY;
353 T14 = KP707106781 * (T12 - T13);
354 T1h = KP707106781 * (T12 + T13);
355 ri[WS(rs, 7)] = T11 - T14;
356 ii[WS(rs, 5)] = T1k - T1h;
357 ri[WS(rs, 3)] = T11 + T14;
358 ii[WS(rs, 1)] = T1h + T1k;
359 }
360 {
361 E TP, T1m, T10, T1l, TU, TZ;
362 TP = TL + TO;
363 T1m = T1j + T1i;
364 TU = TQ + TT;
365 TZ = TV - TY;
366 T10 = KP707106781 * (TU + TZ);
367 T1l = KP707106781 * (TZ - TU);
368 ri[WS(rs, 5)] = TP - T10;
369 ii[WS(rs, 7)] = T1m - T1l;
370 ri[WS(rs, 1)] = TP + T10;
371 ii[WS(rs, 3)] = T1l + T1m;
372 }
373 }
374 }
375 }
376 }
377 }
378
379 static const tw_instr twinstr[] = {
380 {TW_CEXP, 0, 1},
381 {TW_CEXP, 0, 3},
382 {TW_CEXP, 0, 7},
383 {TW_NEXT, 1, 0}
384 };
385
386 static const ct_desc desc = { 8, "t2_8", twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
387
388 void X(codelet_t2_8) (planner *p) {
389 X(kdft_dit_register) (p, t2_8, &desc);
390 }
391 #endif /* HAVE_FMA */