Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cf/hc2cfdft_6.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 6 -dit -name hc2cfdft_6 -include hc2cf.h */
29
30 /*
31 * This function contains 58 FP additions, 44 FP multiplications,
32 * (or, 36 additions, 22 multiplications, 22 fused multiply/add),
33 * 42 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "hc2cf.h"
36
37 static void hc2cfdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
44 E TP, TT, TN, TM, TY, T13;
45 {
46 E T3, TQ, TJ, T12, Tu, TB, TX, T10, Tj, Tf, Ti, Td, Th, TU, TS;
47 {
48 E TC, TI, TF, TH, TA, Tw, TZ;
49 {
50 E T1, T2, TD, TE;
51 T1 = Ip[0];
52 T2 = Im[0];
53 TD = Rm[0];
54 TE = Rp[0];
55 TC = W[0];
56 T3 = T1 - T2;
57 TI = T1 + T2;
58 TQ = TE + TD;
59 TF = TD - TE;
60 TH = W[1];
61 }
62 {
63 E Tr, To, Ts, Tl, Tq;
64 {
65 E Tm, Tn, TG, T11;
66 Tm = Rm[WS(rs, 2)];
67 Tn = Rp[WS(rs, 2)];
68 TG = TC * TF;
69 T11 = TH * TF;
70 Tr = Ip[WS(rs, 2)];
71 TA = Tn + Tm;
72 To = Tm - Tn;
73 TJ = FNMS(TH, TI, TG);
74 T12 = FMA(TC, TI, T11);
75 Ts = Im[WS(rs, 2)];
76 }
77 Tl = W[8];
78 Tq = W[9];
79 {
80 E Tz, Ty, TW, Tx, Tt, Tp;
81 Tw = W[6];
82 Tx = Tr - Ts;
83 Tt = Tr + Ts;
84 Tp = Tl * To;
85 Tz = W[7];
86 Ty = Tw * Tx;
87 TW = Tl * Tt;
88 Tu = FNMS(Tq, Tt, Tp);
89 TZ = Tz * Tx;
90 TB = FNMS(Tz, TA, Ty);
91 TX = FMA(Tq, To, TW);
92 }
93 }
94 {
95 E T5, T6, Ta, Tb;
96 T5 = Ip[WS(rs, 1)];
97 T10 = FMA(Tw, TA, TZ);
98 T6 = Im[WS(rs, 1)];
99 Ta = Rp[WS(rs, 1)];
100 Tb = Rm[WS(rs, 1)];
101 {
102 E T4, Tg, T7, Tc, T9, T8, TR;
103 T4 = W[5];
104 Tg = T5 - T6;
105 T7 = T5 + T6;
106 Tj = Ta + Tb;
107 Tc = Ta - Tb;
108 T9 = W[4];
109 T8 = T4 * T7;
110 Tf = W[2];
111 Ti = W[3];
112 TR = T9 * T7;
113 Td = FMA(T9, Tc, T8);
114 Th = Tf * Tg;
115 TU = Ti * Tg;
116 TS = FNMS(T4, Tc, TR);
117 }
118 }
119 }
120 {
121 E Te, T1d, TK, Tv, T1a, T1b, Tk, TV;
122 TP = Td + T3;
123 Te = T3 - Td;
124 Tk = FNMS(Ti, Tj, Th);
125 TV = FMA(Tf, Tj, TU);
126 T1d = TQ + TS;
127 TT = TQ - TS;
128 TN = TJ - TB;
129 TK = TB + TJ;
130 Tv = Tk + Tu;
131 TM = Tu - Tk;
132 TY = TV - TX;
133 T1a = TV + TX;
134 T1b = T10 + T12;
135 T13 = T10 - T12;
136 {
137 E T1g, TL, T1e, T1c, T19, T1f;
138 T1g = Tv - TK;
139 TL = Tv + TK;
140 T1e = T1a + T1b;
141 T1c = T1a - T1b;
142 T19 = FNMS(KP500000000, TL, Te);
143 Ip[0] = KP500000000 * (Te + TL);
144 T1f = FNMS(KP500000000, T1e, T1d);
145 Rp[0] = KP500000000 * (T1d + T1e);
146 Im[WS(rs, 1)] = -(KP500000000 * (FNMS(KP866025403, T1c, T19)));
147 Ip[WS(rs, 2)] = KP500000000 * (FMA(KP866025403, T1c, T19));
148 Rm[WS(rs, 1)] = KP500000000 * (FMA(KP866025403, T1g, T1f));
149 Rp[WS(rs, 2)] = KP500000000 * (FNMS(KP866025403, T1g, T1f));
150 }
151 }
152 }
153 {
154 E TO, T16, T14, T18, T17, T15;
155 TO = TM + TN;
156 T16 = TN - TM;
157 T14 = TY + T13;
158 T18 = T13 - TY;
159 T17 = FMA(KP500000000, TO, TP);
160 Im[WS(rs, 2)] = KP500000000 * (TO - TP);
161 T15 = FNMS(KP500000000, T14, TT);
162 Rm[WS(rs, 2)] = KP500000000 * (TT + T14);
163 Im[0] = -(KP500000000 * (FNMS(KP866025403, T18, T17)));
164 Ip[WS(rs, 1)] = KP500000000 * (FMA(KP866025403, T18, T17));
165 Rm[0] = KP500000000 * (FNMS(KP866025403, T16, T15));
166 Rp[WS(rs, 1)] = KP500000000 * (FMA(KP866025403, T16, T15));
167 }
168 }
169 }
170 }
171
172 static const tw_instr twinstr[] = {
173 {TW_FULL, 1, 6},
174 {TW_NEXT, 1, 0}
175 };
176
177 static const hc2c_desc desc = { 6, "hc2cfdft_6", twinstr, &GENUS, {36, 22, 22, 0} };
178
179 void X(codelet_hc2cfdft_6) (planner *p) {
180 X(khc2c_register) (p, hc2cfdft_6, &desc, HC2C_VIA_DFT);
181 }
182 #else /* HAVE_FMA */
183
184 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 6 -dit -name hc2cfdft_6 -include hc2cf.h */
185
186 /*
187 * This function contains 58 FP additions, 36 FP multiplications,
188 * (or, 44 additions, 22 multiplications, 14 fused multiply/add),
189 * 40 stack variables, 3 constants, and 24 memory accesses
190 */
191 #include "hc2cf.h"
192
193 static void hc2cfdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
194 {
195 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
196 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
197 DK(KP433012701, +0.433012701892219323381861585376468091735701313);
198 {
199 INT m;
200 for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 10, MAKE_VOLATILE_STRIDE(24, rs)) {
201 E T3, TM, Tc, TN, Ts, T10, TI, TR, TF, T11, TH, TU;
202 {
203 E T1, T2, TD, Tz, TA, TB, T7, Tf, Tb, Th, Tq, Tw, Tm, Tu, T4;
204 E T8;
205 {
206 E T5, T6, T9, Ta;
207 T1 = Ip[0];
208 T2 = Im[0];
209 TD = T1 + T2;
210 Tz = Rm[0];
211 TA = Rp[0];
212 TB = Tz - TA;
213 T5 = Ip[WS(rs, 1)];
214 T6 = Im[WS(rs, 1)];
215 T7 = T5 + T6;
216 Tf = T5 - T6;
217 T9 = Rp[WS(rs, 1)];
218 Ta = Rm[WS(rs, 1)];
219 Tb = T9 - Ta;
220 Th = T9 + Ta;
221 {
222 E To, Tp, Tk, Tl;
223 To = Rp[WS(rs, 2)];
224 Tp = Rm[WS(rs, 2)];
225 Tq = To - Tp;
226 Tw = To + Tp;
227 Tk = Ip[WS(rs, 2)];
228 Tl = Im[WS(rs, 2)];
229 Tm = Tk + Tl;
230 Tu = Tk - Tl;
231 }
232 }
233 T3 = T1 - T2;
234 TM = TA + Tz;
235 T4 = W[5];
236 T8 = W[4];
237 Tc = FMA(T4, T7, T8 * Tb);
238 TN = FNMS(T4, Tb, T8 * T7);
239 {
240 E Ti, TP, Tr, TQ;
241 {
242 E Te, Tg, Tj, Tn;
243 Te = W[2];
244 Tg = W[3];
245 Ti = FNMS(Tg, Th, Te * Tf);
246 TP = FMA(Tg, Tf, Te * Th);
247 Tj = W[9];
248 Tn = W[8];
249 Tr = FMA(Tj, Tm, Tn * Tq);
250 TQ = FNMS(Tj, Tq, Tn * Tm);
251 }
252 Ts = Ti - Tr;
253 T10 = TP + TQ;
254 TI = Ti + Tr;
255 TR = TP - TQ;
256 }
257 {
258 E Tx, TS, TE, TT;
259 {
260 E Tt, Tv, Ty, TC;
261 Tt = W[6];
262 Tv = W[7];
263 Tx = FNMS(Tv, Tw, Tt * Tu);
264 TS = FMA(Tv, Tu, Tt * Tw);
265 Ty = W[0];
266 TC = W[1];
267 TE = FNMS(TC, TD, Ty * TB);
268 TT = FMA(TC, TB, Ty * TD);
269 }
270 TF = Tx + TE;
271 T11 = TS + TT;
272 TH = TE - Tx;
273 TU = TS - TT;
274 }
275 }
276 {
277 E T12, Td, TG, TZ;
278 T12 = KP433012701 * (T10 - T11);
279 Td = T3 - Tc;
280 TG = Ts + TF;
281 TZ = FNMS(KP250000000, TG, KP500000000 * Td);
282 Ip[0] = KP500000000 * (Td + TG);
283 Im[WS(rs, 1)] = T12 - TZ;
284 Ip[WS(rs, 2)] = TZ + T12;
285 }
286 {
287 E T16, T13, T14, T15;
288 T16 = KP433012701 * (Ts - TF);
289 T13 = TM + TN;
290 T14 = T10 + T11;
291 T15 = FNMS(KP250000000, T14, KP500000000 * T13);
292 Rp[WS(rs, 2)] = T15 - T16;
293 Rp[0] = KP500000000 * (T13 + T14);
294 Rm[WS(rs, 1)] = T16 + T15;
295 }
296 {
297 E TY, TJ, TK, TX;
298 TY = KP433012701 * (TU - TR);
299 TJ = TH - TI;
300 TK = Tc + T3;
301 TX = FMA(KP500000000, TK, KP250000000 * TJ);
302 Im[WS(rs, 2)] = KP500000000 * (TJ - TK);
303 Im[0] = TY - TX;
304 Ip[WS(rs, 1)] = TX + TY;
305 }
306 {
307 E TL, TO, TV, TW;
308 TL = KP433012701 * (TI + TH);
309 TO = TM - TN;
310 TV = TR + TU;
311 TW = FNMS(KP250000000, TV, KP500000000 * TO);
312 Rp[WS(rs, 1)] = TL + TW;
313 Rm[WS(rs, 2)] = KP500000000 * (TO + TV);
314 Rm[0] = TW - TL;
315 }
316 }
317 }
318 }
319
320 static const tw_instr twinstr[] = {
321 {TW_FULL, 1, 6},
322 {TW_NEXT, 1, 0}
323 };
324
325 static const hc2c_desc desc = { 6, "hc2cfdft_6", twinstr, &GENUS, {44, 22, 14, 0} };
326
327 void X(codelet_hc2cfdft_6) (planner *p) {
328 X(khc2c_register) (p, hc2cfdft_6, &desc, HC2C_VIA_DFT);
329 }
330 #endif /* HAVE_FMA */