Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_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 07:35:49 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 -n 8 -name t1_8 -include t.h */
29
30 /*
31 * This function contains 66 FP additions, 36 FP multiplications,
32 * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
33 * 61 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "t.h"
36
37 static void t1_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 * 14); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
43 E T1g, T1f, T1e, Tm, T1q, T1o, T1p, TN, T1h, T1i;
44 {
45 E T1, T1m, T1l, T7, TS, Tk, TQ, Te, To, Tr, T17, TM, T12, Tu, TW;
46 E Tp, Tx, Tt, Tq, Tw;
47 {
48 E T3, T6, T2, T5;
49 T1 = ri[0];
50 T1m = ii[0];
51 T3 = ri[WS(rs, 4)];
52 T6 = ii[WS(rs, 4)];
53 T2 = W[6];
54 T5 = W[7];
55 {
56 E Ta, Td, T9, Tc;
57 {
58 E Tg, Tj, Ti, TR, Th, T1k, T4, Tf;
59 Tg = ri[WS(rs, 6)];
60 Tj = ii[WS(rs, 6)];
61 T1k = T2 * T6;
62 T4 = T2 * T3;
63 Tf = W[10];
64 Ti = W[11];
65 T1l = FNMS(T5, T3, T1k);
66 T7 = FMA(T5, T6, T4);
67 TR = Tf * Tj;
68 Th = Tf * Tg;
69 Ta = ri[WS(rs, 2)];
70 Td = ii[WS(rs, 2)];
71 TS = FNMS(Ti, Tg, TR);
72 Tk = FMA(Ti, Tj, Th);
73 T9 = W[2];
74 Tc = W[3];
75 }
76 {
77 E TB, TE, TH, T13, TC, TK, TG, TD, TJ, TP, Tb, TA, Tn;
78 TB = ri[WS(rs, 7)];
79 TE = ii[WS(rs, 7)];
80 TP = T9 * Td;
81 Tb = T9 * Ta;
82 TA = W[12];
83 TH = ri[WS(rs, 3)];
84 TQ = FNMS(Tc, Ta, TP);
85 Te = FMA(Tc, Td, Tb);
86 T13 = TA * TE;
87 TC = TA * TB;
88 TK = ii[WS(rs, 3)];
89 TG = W[4];
90 TD = W[13];
91 TJ = W[5];
92 {
93 E T14, TF, T16, TL, T15, TI;
94 To = ri[WS(rs, 1)];
95 T15 = TG * TK;
96 TI = TG * TH;
97 T14 = FNMS(TD, TB, T13);
98 TF = FMA(TD, TE, TC);
99 T16 = FNMS(TJ, TH, T15);
100 TL = FMA(TJ, TK, TI);
101 Tr = ii[WS(rs, 1)];
102 Tn = W[0];
103 T17 = T14 - T16;
104 T1g = T14 + T16;
105 TM = TF + TL;
106 T12 = TF - TL;
107 }
108 Tu = ri[WS(rs, 5)];
109 TW = Tn * Tr;
110 Tp = Tn * To;
111 Tx = ii[WS(rs, 5)];
112 Tt = W[8];
113 Tq = W[1];
114 Tw = W[9];
115 }
116 }
117 }
118 {
119 E T8, T1j, T1n, Tz, T1a, TU, Tl, T1b, T1c, T1v, T1t, T1w, T19, T1u, T1d;
120 {
121 E T1r, T10, TV, T1s, T11, T18;
122 {
123 E TO, TX, Ts, TZ, Ty, TT, TY, Tv;
124 T8 = T1 + T7;
125 TO = T1 - T7;
126 TY = Tt * Tx;
127 Tv = Tt * Tu;
128 TX = FNMS(Tq, To, TW);
129 Ts = FMA(Tq, Tr, Tp);
130 TZ = FNMS(Tw, Tu, TY);
131 Ty = FMA(Tw, Tx, Tv);
132 TT = TQ - TS;
133 T1j = TQ + TS;
134 T1n = T1l + T1m;
135 T1r = T1m - T1l;
136 T10 = TX - TZ;
137 T1f = TX + TZ;
138 Tz = Ts + Ty;
139 TV = Ts - Ty;
140 T1a = TO - TT;
141 TU = TO + TT;
142 T1s = Te - Tk;
143 Tl = Te + Tk;
144 }
145 T1b = T10 - TV;
146 T11 = TV + T10;
147 T18 = T12 - T17;
148 T1c = T12 + T17;
149 T1v = T1s + T1r;
150 T1t = T1r - T1s;
151 T1w = T18 - T11;
152 T19 = T11 + T18;
153 }
154 ii[WS(rs, 3)] = FMA(KP707106781, T1w, T1v);
155 ii[WS(rs, 7)] = FNMS(KP707106781, T1w, T1v);
156 ri[WS(rs, 1)] = FMA(KP707106781, T19, TU);
157 ri[WS(rs, 5)] = FNMS(KP707106781, T19, TU);
158 T1u = T1b + T1c;
159 T1d = T1b - T1c;
160 ii[WS(rs, 1)] = FMA(KP707106781, T1u, T1t);
161 ii[WS(rs, 5)] = FNMS(KP707106781, T1u, T1t);
162 ri[WS(rs, 3)] = FMA(KP707106781, T1d, T1a);
163 ri[WS(rs, 7)] = FNMS(KP707106781, T1d, T1a);
164 T1e = T8 - Tl;
165 Tm = T8 + Tl;
166 T1q = T1n - T1j;
167 T1o = T1j + T1n;
168 T1p = TM - Tz;
169 TN = Tz + TM;
170 }
171 }
172 ii[WS(rs, 2)] = T1p + T1q;
173 ii[WS(rs, 6)] = T1q - T1p;
174 ri[0] = Tm + TN;
175 ri[WS(rs, 4)] = Tm - TN;
176 T1h = T1f - T1g;
177 T1i = T1f + T1g;
178 ii[0] = T1i + T1o;
179 ii[WS(rs, 4)] = T1o - T1i;
180 ri[WS(rs, 2)] = T1e + T1h;
181 ri[WS(rs, 6)] = T1e - T1h;
182 }
183 }
184 }
185
186 static const tw_instr twinstr[] = {
187 {TW_FULL, 0, 8},
188 {TW_NEXT, 1, 0}
189 };
190
191 static const ct_desc desc = { 8, "t1_8", twinstr, &GENUS, {44, 14, 22, 0}, 0, 0, 0 };
192
193 void X(codelet_t1_8) (planner *p) {
194 X(kdft_dit_register) (p, t1_8, &desc);
195 }
196 #else /* HAVE_FMA */
197
198 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 8 -name t1_8 -include t.h */
199
200 /*
201 * This function contains 66 FP additions, 32 FP multiplications,
202 * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
203 * 28 stack variables, 1 constants, and 32 memory accesses
204 */
205 #include "t.h"
206
207 static void t1_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
208 {
209 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
210 {
211 INT m;
212 for (m = mb, W = W + (mb * 14); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 14, MAKE_VOLATILE_STRIDE(16, rs)) {
213 E T7, T1e, TH, T19, TF, T13, TR, TU, Ti, T1f, TK, T16, Tu, T12, TM;
214 E TP;
215 {
216 E T1, T18, T6, T17;
217 T1 = ri[0];
218 T18 = ii[0];
219 {
220 E T3, T5, T2, T4;
221 T3 = ri[WS(rs, 4)];
222 T5 = ii[WS(rs, 4)];
223 T2 = W[6];
224 T4 = W[7];
225 T6 = FMA(T2, T3, T4 * T5);
226 T17 = FNMS(T4, T3, T2 * T5);
227 }
228 T7 = T1 + T6;
229 T1e = T18 - T17;
230 TH = T1 - T6;
231 T19 = T17 + T18;
232 }
233 {
234 E Tz, TS, TE, TT;
235 {
236 E Tw, Ty, Tv, Tx;
237 Tw = ri[WS(rs, 7)];
238 Ty = ii[WS(rs, 7)];
239 Tv = W[12];
240 Tx = W[13];
241 Tz = FMA(Tv, Tw, Tx * Ty);
242 TS = FNMS(Tx, Tw, Tv * Ty);
243 }
244 {
245 E TB, TD, TA, TC;
246 TB = ri[WS(rs, 3)];
247 TD = ii[WS(rs, 3)];
248 TA = W[4];
249 TC = W[5];
250 TE = FMA(TA, TB, TC * TD);
251 TT = FNMS(TC, TB, TA * TD);
252 }
253 TF = Tz + TE;
254 T13 = TS + TT;
255 TR = Tz - TE;
256 TU = TS - TT;
257 }
258 {
259 E Tc, TI, Th, TJ;
260 {
261 E T9, Tb, T8, Ta;
262 T9 = ri[WS(rs, 2)];
263 Tb = ii[WS(rs, 2)];
264 T8 = W[2];
265 Ta = W[3];
266 Tc = FMA(T8, T9, Ta * Tb);
267 TI = FNMS(Ta, T9, T8 * Tb);
268 }
269 {
270 E Te, Tg, Td, Tf;
271 Te = ri[WS(rs, 6)];
272 Tg = ii[WS(rs, 6)];
273 Td = W[10];
274 Tf = W[11];
275 Th = FMA(Td, Te, Tf * Tg);
276 TJ = FNMS(Tf, Te, Td * Tg);
277 }
278 Ti = Tc + Th;
279 T1f = Tc - Th;
280 TK = TI - TJ;
281 T16 = TI + TJ;
282 }
283 {
284 E To, TN, Tt, TO;
285 {
286 E Tl, Tn, Tk, Tm;
287 Tl = ri[WS(rs, 1)];
288 Tn = ii[WS(rs, 1)];
289 Tk = W[0];
290 Tm = W[1];
291 To = FMA(Tk, Tl, Tm * Tn);
292 TN = FNMS(Tm, Tl, Tk * Tn);
293 }
294 {
295 E Tq, Ts, Tp, Tr;
296 Tq = ri[WS(rs, 5)];
297 Ts = ii[WS(rs, 5)];
298 Tp = W[8];
299 Tr = W[9];
300 Tt = FMA(Tp, Tq, Tr * Ts);
301 TO = FNMS(Tr, Tq, Tp * Ts);
302 }
303 Tu = To + Tt;
304 T12 = TN + TO;
305 TM = To - Tt;
306 TP = TN - TO;
307 }
308 {
309 E Tj, TG, T1b, T1c;
310 Tj = T7 + Ti;
311 TG = Tu + TF;
312 ri[WS(rs, 4)] = Tj - TG;
313 ri[0] = Tj + TG;
314 {
315 E T15, T1a, T11, T14;
316 T15 = T12 + T13;
317 T1a = T16 + T19;
318 ii[0] = T15 + T1a;
319 ii[WS(rs, 4)] = T1a - T15;
320 T11 = T7 - Ti;
321 T14 = T12 - T13;
322 ri[WS(rs, 6)] = T11 - T14;
323 ri[WS(rs, 2)] = T11 + T14;
324 }
325 T1b = TF - Tu;
326 T1c = T19 - T16;
327 ii[WS(rs, 2)] = T1b + T1c;
328 ii[WS(rs, 6)] = T1c - T1b;
329 {
330 E TX, T1g, T10, T1d, TY, TZ;
331 TX = TH - TK;
332 T1g = T1e - T1f;
333 TY = TP - TM;
334 TZ = TR + TU;
335 T10 = KP707106781 * (TY - TZ);
336 T1d = KP707106781 * (TY + TZ);
337 ri[WS(rs, 7)] = TX - T10;
338 ii[WS(rs, 5)] = T1g - T1d;
339 ri[WS(rs, 3)] = TX + T10;
340 ii[WS(rs, 1)] = T1d + T1g;
341 }
342 {
343 E TL, T1i, TW, T1h, TQ, TV;
344 TL = TH + TK;
345 T1i = T1f + T1e;
346 TQ = TM + TP;
347 TV = TR - TU;
348 TW = KP707106781 * (TQ + TV);
349 T1h = KP707106781 * (TV - TQ);
350 ri[WS(rs, 5)] = TL - TW;
351 ii[WS(rs, 7)] = T1i - T1h;
352 ri[WS(rs, 1)] = TL + TW;
353 ii[WS(rs, 3)] = T1h + T1i;
354 }
355 }
356 }
357 }
358 }
359
360 static const tw_instr twinstr[] = {
361 {TW_FULL, 0, 8},
362 {TW_NEXT, 1, 0}
363 };
364
365 static const ct_desc desc = { 8, "t1_8", twinstr, &GENUS, {52, 18, 14, 0}, 0, 0, 0 };
366
367 void X(codelet_t1_8) (planner *p) {
368 X(kdft_dit_register) (p, t1_8, &desc);
369 }
370 #endif /* HAVE_FMA */