Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t2_5.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 07:36:09 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 5 -name t2_5 -include t.h */
29
30 /*
31 * This function contains 44 FP additions, 40 FP multiplications,
32 * (or, 14 additions, 10 multiplications, 30 fused multiply/add),
33 * 47 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "t.h"
36
37 static void t2_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT m;
45 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
46 E Ta, T1, TO, Tp, TS, Ti, TL, TC, To, TE, Ts, TF, T2, T8, T5;
47 E TT, Tt, TG;
48 T2 = W[0];
49 Ta = W[3];
50 T8 = W[2];
51 T5 = W[1];
52 {
53 E Tq, Tr, Te, T9;
54 T1 = ri[0];
55 Te = T2 * Ta;
56 T9 = T2 * T8;
57 TO = ii[0];
58 {
59 E T3, Tf, Tm, Tj, Tb, T4, T6, Tc, Tg;
60 T3 = ri[WS(rs, 1)];
61 Tf = FMA(T5, T8, Te);
62 Tm = FNMS(T5, T8, Te);
63 Tj = FMA(T5, Ta, T9);
64 Tb = FNMS(T5, Ta, T9);
65 T4 = T2 * T3;
66 T6 = ii[WS(rs, 1)];
67 Tc = ri[WS(rs, 4)];
68 Tg = ii[WS(rs, 4)];
69 {
70 E Tk, Tl, Tn, TD;
71 {
72 E T7, Tz, Th, TB, Ty, Td, TA;
73 Tk = ri[WS(rs, 2)];
74 T7 = FMA(T5, T6, T4);
75 Ty = T2 * T6;
76 Td = Tb * Tc;
77 TA = Tb * Tg;
78 Tl = Tj * Tk;
79 Tz = FNMS(T5, T3, Ty);
80 Th = FMA(Tf, Tg, Td);
81 TB = FNMS(Tf, Tc, TA);
82 Tn = ii[WS(rs, 2)];
83 Tp = ri[WS(rs, 3)];
84 TS = T7 - Th;
85 Ti = T7 + Th;
86 TL = Tz + TB;
87 TC = Tz - TB;
88 TD = Tj * Tn;
89 Tq = T8 * Tp;
90 Tr = ii[WS(rs, 3)];
91 }
92 To = FMA(Tm, Tn, Tl);
93 TE = FNMS(Tm, Tk, TD);
94 }
95 }
96 Ts = FMA(Ta, Tr, Tq);
97 TF = T8 * Tr;
98 }
99 TT = To - Ts;
100 Tt = To + Ts;
101 TG = FNMS(Ta, Tp, TF);
102 {
103 E TU, TW, TV, TR, Tw, Tu;
104 TU = FMA(KP618033988, TT, TS);
105 TW = FNMS(KP618033988, TS, TT);
106 Tw = Ti - Tt;
107 Tu = Ti + Tt;
108 {
109 E TM, TH, Tv, TI, TK;
110 TM = TE + TG;
111 TH = TE - TG;
112 ri[0] = T1 + Tu;
113 Tv = FNMS(KP250000000, Tu, T1);
114 TI = FMA(KP618033988, TH, TC);
115 TK = FNMS(KP618033988, TC, TH);
116 {
117 E TQ, TN, TJ, Tx, TP;
118 TQ = TL - TM;
119 TN = TL + TM;
120 TJ = FNMS(KP559016994, Tw, Tv);
121 Tx = FMA(KP559016994, Tw, Tv);
122 ii[0] = TN + TO;
123 TP = FNMS(KP250000000, TN, TO);
124 ri[WS(rs, 1)] = FMA(KP951056516, TI, Tx);
125 ri[WS(rs, 4)] = FNMS(KP951056516, TI, Tx);
126 ri[WS(rs, 3)] = FMA(KP951056516, TK, TJ);
127 ri[WS(rs, 2)] = FNMS(KP951056516, TK, TJ);
128 TV = FNMS(KP559016994, TQ, TP);
129 TR = FMA(KP559016994, TQ, TP);
130 }
131 }
132 ii[WS(rs, 4)] = FMA(KP951056516, TU, TR);
133 ii[WS(rs, 1)] = FNMS(KP951056516, TU, TR);
134 ii[WS(rs, 3)] = FNMS(KP951056516, TW, TV);
135 ii[WS(rs, 2)] = FMA(KP951056516, TW, TV);
136 }
137 }
138 }
139 }
140
141 static const tw_instr twinstr[] = {
142 {TW_CEXP, 0, 1},
143 {TW_CEXP, 0, 3},
144 {TW_NEXT, 1, 0}
145 };
146
147 static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {14, 10, 30, 0}, 0, 0, 0 };
148
149 void X(codelet_t2_5) (planner *p) {
150 X(kdft_dit_register) (p, t2_5, &desc);
151 }
152 #else /* HAVE_FMA */
153
154 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 5 -name t2_5 -include t.h */
155
156 /*
157 * This function contains 44 FP additions, 32 FP multiplications,
158 * (or, 30 additions, 18 multiplications, 14 fused multiply/add),
159 * 37 stack variables, 4 constants, and 20 memory accesses
160 */
161 #include "t.h"
162
163 static void t2_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
164 {
165 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
166 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
167 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
168 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
169 {
170 INT m;
171 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
172 E T2, T4, T7, T9, Tb, Tl, Tf, Tj;
173 {
174 E T8, Te, Ta, Td;
175 T2 = W[0];
176 T4 = W[1];
177 T7 = W[2];
178 T9 = W[3];
179 T8 = T2 * T7;
180 Te = T4 * T7;
181 Ta = T4 * T9;
182 Td = T2 * T9;
183 Tb = T8 - Ta;
184 Tl = Td - Te;
185 Tf = Td + Te;
186 Tj = T8 + Ta;
187 }
188 {
189 E T1, TI, Ty, TB, TN, TM, TF, TG, TH, Ti, Tr, Ts;
190 T1 = ri[0];
191 TI = ii[0];
192 {
193 E T6, Tw, Tq, TA, Th, Tx, Tn, Tz;
194 {
195 E T3, T5, To, Tp;
196 T3 = ri[WS(rs, 1)];
197 T5 = ii[WS(rs, 1)];
198 T6 = FMA(T2, T3, T4 * T5);
199 Tw = FNMS(T4, T3, T2 * T5);
200 To = ri[WS(rs, 3)];
201 Tp = ii[WS(rs, 3)];
202 Tq = FMA(T7, To, T9 * Tp);
203 TA = FNMS(T9, To, T7 * Tp);
204 }
205 {
206 E Tc, Tg, Tk, Tm;
207 Tc = ri[WS(rs, 4)];
208 Tg = ii[WS(rs, 4)];
209 Th = FMA(Tb, Tc, Tf * Tg);
210 Tx = FNMS(Tf, Tc, Tb * Tg);
211 Tk = ri[WS(rs, 2)];
212 Tm = ii[WS(rs, 2)];
213 Tn = FMA(Tj, Tk, Tl * Tm);
214 Tz = FNMS(Tl, Tk, Tj * Tm);
215 }
216 Ty = Tw - Tx;
217 TB = Tz - TA;
218 TN = Tn - Tq;
219 TM = T6 - Th;
220 TF = Tw + Tx;
221 TG = Tz + TA;
222 TH = TF + TG;
223 Ti = T6 + Th;
224 Tr = Tn + Tq;
225 Ts = Ti + Tr;
226 }
227 ri[0] = T1 + Ts;
228 ii[0] = TH + TI;
229 {
230 E TC, TE, Tv, TD, Tt, Tu;
231 TC = FMA(KP951056516, Ty, KP587785252 * TB);
232 TE = FNMS(KP587785252, Ty, KP951056516 * TB);
233 Tt = KP559016994 * (Ti - Tr);
234 Tu = FNMS(KP250000000, Ts, T1);
235 Tv = Tt + Tu;
236 TD = Tu - Tt;
237 ri[WS(rs, 4)] = Tv - TC;
238 ri[WS(rs, 3)] = TD + TE;
239 ri[WS(rs, 1)] = Tv + TC;
240 ri[WS(rs, 2)] = TD - TE;
241 }
242 {
243 E TO, TP, TL, TQ, TJ, TK;
244 TO = FMA(KP951056516, TM, KP587785252 * TN);
245 TP = FNMS(KP587785252, TM, KP951056516 * TN);
246 TJ = KP559016994 * (TF - TG);
247 TK = FNMS(KP250000000, TH, TI);
248 TL = TJ + TK;
249 TQ = TK - TJ;
250 ii[WS(rs, 1)] = TL - TO;
251 ii[WS(rs, 3)] = TQ - TP;
252 ii[WS(rs, 4)] = TO + TL;
253 ii[WS(rs, 2)] = TP + TQ;
254 }
255 }
256 }
257 }
258 }
259
260 static const tw_instr twinstr[] = {
261 {TW_CEXP, 0, 1},
262 {TW_CEXP, 0, 3},
263 {TW_NEXT, 1, 0}
264 };
265
266 static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {30, 18, 14, 0}, 0, 0, 0 };
267
268 void X(codelet_t2_5) (planner *p) {
269 X(kdft_dit_register) (p, t2_5, &desc);
270 }
271 #endif /* HAVE_FMA */