Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/rdft/scalar/r2cb/hc2cbdft_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 07:42:04 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 6 -dif -name hc2cbdft_6 -include hc2cb.h */
29
30 /*
31 * This function contains 58 FP additions, 32 FP multiplications,
32 * (or, 36 additions, 10 multiplications, 22 fused multiply/add),
33 * 52 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "hc2cb.h"
36
37 static void hc2cbdft_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 T18, T1b, T16, T1e, T1a, T1f, T19, T1g, T1c;
45 {
46 E Tw, T4, TV, Tj, TP, TH, Tr, TY, T5, T6, Ta, Ty;
47 {
48 E Tg, TF, Tf, TD, Tp, Th;
49 {
50 E Td, Te, Tn, To;
51 Td = Ip[WS(rs, 1)];
52 Te = Im[WS(rs, 1)];
53 Tn = Ip[0];
54 To = Im[WS(rs, 2)];
55 Tg = Ip[WS(rs, 2)];
56 TF = Te + Td;
57 Tf = Td - Te;
58 TD = Tn + To;
59 Tp = Tn - To;
60 Th = Im[0];
61 }
62 {
63 E T2, T3, T8, T9;
64 T2 = Rp[0];
65 T3 = Rm[WS(rs, 2)];
66 {
67 E Tq, TE, Ti, TG;
68 T8 = Rm[WS(rs, 1)];
69 TE = Tg + Th;
70 Ti = Tg - Th;
71 Tw = T2 - T3;
72 T4 = T2 + T3;
73 TG = TE - TF;
74 TV = TF + TE;
75 Tq = Tf + Ti;
76 Tj = Tf - Ti;
77 TP = FNMS(KP500000000, TG, TD);
78 TH = TD + TG;
79 T9 = Rp[WS(rs, 1)];
80 Tr = FNMS(KP500000000, Tq, Tp);
81 TY = Tp + Tq;
82 }
83 T5 = Rp[WS(rs, 2)];
84 T6 = Rm[0];
85 Ta = T8 + T9;
86 Ty = T8 - T9;
87 }
88 }
89 {
90 E TO, TT, Ts, TA, TR, Tc, TN, TW, TS, Tx, T7;
91 Tx = T5 - T6;
92 T7 = T5 + T6;
93 TO = W[0];
94 TT = W[1];
95 {
96 E Tz, TQ, Tb, TU;
97 Tz = Tx + Ty;
98 TQ = Tx - Ty;
99 Tb = T7 + Ta;
100 Ts = T7 - Ta;
101 TU = FNMS(KP500000000, Tz, Tw);
102 TA = Tw + Tz;
103 TR = FMA(KP866025403, TQ, TP);
104 T18 = FNMS(KP866025403, TQ, TP);
105 Tc = FNMS(KP500000000, Tb, T4);
106 TN = T4 + Tb;
107 T1b = FMA(KP866025403, TV, TU);
108 TW = FNMS(KP866025403, TV, TU);
109 TS = TO * TR;
110 }
111 {
112 E T15, Tt, T12, T1, Tm, TI, TM, Tl, TJ;
113 {
114 E Tv, TC, TB, TL, Tk, TZ, TX, T10;
115 T15 = FMA(KP866025403, Ts, Tr);
116 Tt = FNMS(KP866025403, Ts, Tr);
117 TZ = TO * TW;
118 TX = FMA(TT, TW, TS);
119 Tv = W[4];
120 TC = W[5];
121 T10 = FNMS(TT, TR, TZ);
122 Rm[0] = TN + TX;
123 Rp[0] = TN - TX;
124 TB = Tv * TA;
125 Im[0] = T10 - TY;
126 Ip[0] = TY + T10;
127 TL = TC * TA;
128 Tk = FNMS(KP866025403, Tj, Tc);
129 T12 = FMA(KP866025403, Tj, Tc);
130 T1 = W[3];
131 Tm = W[2];
132 TI = FNMS(TC, TH, TB);
133 TM = FMA(Tv, TH, TL);
134 Tl = T1 * Tk;
135 TJ = Tm * Tk;
136 }
137 {
138 E T11, T14, T13, T1d, T17, Tu, TK;
139 Tu = FMA(Tm, Tt, Tl);
140 TK = FNMS(T1, Tt, TJ);
141 T11 = W[6];
142 T14 = W[7];
143 Im[WS(rs, 1)] = TI - Tu;
144 Ip[WS(rs, 1)] = Tu + TI;
145 Rm[WS(rs, 1)] = TK + TM;
146 Rp[WS(rs, 1)] = TK - TM;
147 T13 = T11 * T12;
148 T1d = T14 * T12;
149 T17 = W[8];
150 T16 = FNMS(T14, T15, T13);
151 T1e = FMA(T11, T15, T1d);
152 T1a = W[9];
153 T1f = T17 * T1b;
154 T19 = T17 * T18;
155 }
156 }
157 }
158 }
159 T1g = FNMS(T1a, T18, T1f);
160 T1c = FMA(T1a, T1b, T19);
161 Im[WS(rs, 2)] = T1g - T1e;
162 Ip[WS(rs, 2)] = T1e + T1g;
163 Rm[WS(rs, 2)] = T16 + T1c;
164 Rp[WS(rs, 2)] = T16 - T1c;
165 }
166 }
167 }
168
169 static const tw_instr twinstr[] = {
170 {TW_FULL, 1, 6},
171 {TW_NEXT, 1, 0}
172 };
173
174 static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {36, 10, 22, 0} };
175
176 void X(codelet_hc2cbdft_6) (planner *p) {
177 X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT);
178 }
179 #else /* HAVE_FMA */
180
181 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hc2cbdft_6 -include hc2cb.h */
182
183 /*
184 * This function contains 58 FP additions, 28 FP multiplications,
185 * (or, 44 additions, 14 multiplications, 14 fused multiply/add),
186 * 29 stack variables, 2 constants, and 24 memory accesses
187 */
188 #include "hc2cb.h"
189
190 static void hc2cbdft_6(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
191 {
192 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
193 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
194 {
195 INT m;
196 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)) {
197 E T4, Tv, Tr, TL, Tb, Tc, Ty, TP, To, TB, Tj, TQ, Tp, Tq, TE;
198 E TM;
199 {
200 E Ta, Tx, T7, Tw, T2, T3;
201 T2 = Rp[0];
202 T3 = Rm[WS(rs, 2)];
203 T4 = T2 + T3;
204 Tv = T2 - T3;
205 {
206 E T8, T9, T5, T6;
207 T8 = Rm[WS(rs, 1)];
208 T9 = Rp[WS(rs, 1)];
209 Ta = T8 + T9;
210 Tx = T8 - T9;
211 T5 = Rp[WS(rs, 2)];
212 T6 = Rm[0];
213 T7 = T5 + T6;
214 Tw = T5 - T6;
215 }
216 Tr = KP866025403 * (T7 - Ta);
217 TL = KP866025403 * (Tw - Tx);
218 Tb = T7 + Ta;
219 Tc = FNMS(KP500000000, Tb, T4);
220 Ty = Tw + Tx;
221 TP = FNMS(KP500000000, Ty, Tv);
222 }
223 {
224 E Tf, TC, Ti, TD, Td, Te;
225 Td = Ip[WS(rs, 1)];
226 Te = Im[WS(rs, 1)];
227 Tf = Td - Te;
228 TC = Te + Td;
229 {
230 E Tm, Tn, Tg, Th;
231 Tm = Ip[0];
232 Tn = Im[WS(rs, 2)];
233 To = Tm - Tn;
234 TB = Tm + Tn;
235 Tg = Ip[WS(rs, 2)];
236 Th = Im[0];
237 Ti = Tg - Th;
238 TD = Tg + Th;
239 }
240 Tj = KP866025403 * (Tf - Ti);
241 TQ = KP866025403 * (TC + TD);
242 Tp = Tf + Ti;
243 Tq = FNMS(KP500000000, Tp, To);
244 TE = TC - TD;
245 TM = FMA(KP500000000, TE, TB);
246 }
247 {
248 E TJ, TT, TS, TU;
249 TJ = T4 + Tb;
250 TT = To + Tp;
251 {
252 E TN, TR, TK, TO;
253 TN = TL + TM;
254 TR = TP - TQ;
255 TK = W[0];
256 TO = W[1];
257 TS = FMA(TK, TN, TO * TR);
258 TU = FNMS(TO, TN, TK * TR);
259 }
260 Rp[0] = TJ - TS;
261 Ip[0] = TT + TU;
262 Rm[0] = TJ + TS;
263 Im[0] = TU - TT;
264 }
265 {
266 E TZ, T15, T14, T16;
267 {
268 E TW, TY, TV, TX;
269 TW = Tc + Tj;
270 TY = Tr + Tq;
271 TV = W[6];
272 TX = W[7];
273 TZ = FNMS(TX, TY, TV * TW);
274 T15 = FMA(TX, TW, TV * TY);
275 }
276 {
277 E T11, T13, T10, T12;
278 T11 = TM - TL;
279 T13 = TP + TQ;
280 T10 = W[8];
281 T12 = W[9];
282 T14 = FMA(T10, T11, T12 * T13);
283 T16 = FNMS(T12, T11, T10 * T13);
284 }
285 Rp[WS(rs, 2)] = TZ - T14;
286 Ip[WS(rs, 2)] = T15 + T16;
287 Rm[WS(rs, 2)] = TZ + T14;
288 Im[WS(rs, 2)] = T16 - T15;
289 }
290 {
291 E Tt, TH, TG, TI;
292 {
293 E Tk, Ts, T1, Tl;
294 Tk = Tc - Tj;
295 Ts = Tq - Tr;
296 T1 = W[3];
297 Tl = W[2];
298 Tt = FMA(T1, Tk, Tl * Ts);
299 TH = FNMS(T1, Ts, Tl * Tk);
300 }
301 {
302 E Tz, TF, Tu, TA;
303 Tz = Tv + Ty;
304 TF = TB - TE;
305 Tu = W[4];
306 TA = W[5];
307 TG = FNMS(TA, TF, Tu * Tz);
308 TI = FMA(TA, Tz, Tu * TF);
309 }
310 Ip[WS(rs, 1)] = Tt + TG;
311 Rp[WS(rs, 1)] = TH - TI;
312 Im[WS(rs, 1)] = TG - Tt;
313 Rm[WS(rs, 1)] = TH + TI;
314 }
315 }
316 }
317 }
318
319 static const tw_instr twinstr[] = {
320 {TW_FULL, 1, 6},
321 {TW_NEXT, 1, 0}
322 };
323
324 static const hc2c_desc desc = { 6, "hc2cbdft_6", twinstr, &GENUS, {44, 14, 14, 0} };
325
326 void X(codelet_hc2cbdft_6) (planner *p) {
327 X(khc2c_register) (p, hc2cbdft_6, &desc, HC2C_VIA_DFT);
328 }
329 #endif /* HAVE_FMA */