Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/n1_5.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 07:35:42 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 5 -name n1_5 -include n.h */
29
30 /*
31 * This function contains 32 FP additions, 18 FP multiplications,
32 * (or, 14 additions, 0 multiplications, 18 fused multiply/add),
33 * 37 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "n.h"
36
37 static void n1_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT i;
45 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
46 E Tq, Ti, Tk, Tu, Tw, Tp, Tb, Tj, Tr, Tv;
47 {
48 E T1, Tl, Ts, Tt, T8, Ta, Te, Tm, Tn, Th, To, T9;
49 T1 = ri[0];
50 Tl = ii[0];
51 {
52 E T2, T3, T5, T6;
53 T2 = ri[WS(is, 1)];
54 T3 = ri[WS(is, 4)];
55 T5 = ri[WS(is, 2)];
56 T6 = ri[WS(is, 3)];
57 {
58 E Tc, T4, T7, Td, Tf, Tg;
59 Tc = ii[WS(is, 1)];
60 Ts = T2 - T3;
61 T4 = T2 + T3;
62 Tt = T5 - T6;
63 T7 = T5 + T6;
64 Td = ii[WS(is, 4)];
65 Tf = ii[WS(is, 2)];
66 Tg = ii[WS(is, 3)];
67 T8 = T4 + T7;
68 Ta = T4 - T7;
69 Te = Tc - Td;
70 Tm = Tc + Td;
71 Tn = Tf + Tg;
72 Th = Tf - Tg;
73 }
74 }
75 ro[0] = T1 + T8;
76 To = Tm + Tn;
77 Tq = Tm - Tn;
78 Ti = FMA(KP618033988, Th, Te);
79 Tk = FNMS(KP618033988, Te, Th);
80 io[0] = Tl + To;
81 T9 = FNMS(KP250000000, T8, T1);
82 Tu = FMA(KP618033988, Tt, Ts);
83 Tw = FNMS(KP618033988, Ts, Tt);
84 Tp = FNMS(KP250000000, To, Tl);
85 Tb = FMA(KP559016994, Ta, T9);
86 Tj = FNMS(KP559016994, Ta, T9);
87 }
88 Tr = FMA(KP559016994, Tq, Tp);
89 Tv = FNMS(KP559016994, Tq, Tp);
90 ro[WS(os, 2)] = FNMS(KP951056516, Tk, Tj);
91 ro[WS(os, 3)] = FMA(KP951056516, Tk, Tj);
92 ro[WS(os, 1)] = FMA(KP951056516, Ti, Tb);
93 ro[WS(os, 4)] = FNMS(KP951056516, Ti, Tb);
94 io[WS(os, 2)] = FMA(KP951056516, Tw, Tv);
95 io[WS(os, 3)] = FNMS(KP951056516, Tw, Tv);
96 io[WS(os, 4)] = FMA(KP951056516, Tu, Tr);
97 io[WS(os, 1)] = FNMS(KP951056516, Tu, Tr);
98 }
99 }
100 }
101
102 static const kdft_desc desc = { 5, "n1_5", {14, 0, 18, 0}, &GENUS, 0, 0, 0, 0 };
103
104 void X(codelet_n1_5) (planner *p) {
105 X(kdft_register) (p, n1_5, &desc);
106 }
107
108 #else /* HAVE_FMA */
109
110 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 5 -name n1_5 -include n.h */
111
112 /*
113 * This function contains 32 FP additions, 12 FP multiplications,
114 * (or, 26 additions, 6 multiplications, 6 fused multiply/add),
115 * 21 stack variables, 4 constants, and 20 memory accesses
116 */
117 #include "n.h"
118
119 static void n1_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
120 {
121 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
122 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
123 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
124 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
125 {
126 INT i;
127 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(20, is), MAKE_VOLATILE_STRIDE(20, os)) {
128 E T1, To, T8, Tt, T9, Ts, Te, Tp, Th, Tn;
129 T1 = ri[0];
130 To = ii[0];
131 {
132 E T2, T3, T4, T5, T6, T7;
133 T2 = ri[WS(is, 1)];
134 T3 = ri[WS(is, 4)];
135 T4 = T2 + T3;
136 T5 = ri[WS(is, 2)];
137 T6 = ri[WS(is, 3)];
138 T7 = T5 + T6;
139 T8 = T4 + T7;
140 Tt = T5 - T6;
141 T9 = KP559016994 * (T4 - T7);
142 Ts = T2 - T3;
143 }
144 {
145 E Tc, Td, Tl, Tf, Tg, Tm;
146 Tc = ii[WS(is, 1)];
147 Td = ii[WS(is, 4)];
148 Tl = Tc + Td;
149 Tf = ii[WS(is, 2)];
150 Tg = ii[WS(is, 3)];
151 Tm = Tf + Tg;
152 Te = Tc - Td;
153 Tp = Tl + Tm;
154 Th = Tf - Tg;
155 Tn = KP559016994 * (Tl - Tm);
156 }
157 ro[0] = T1 + T8;
158 io[0] = To + Tp;
159 {
160 E Ti, Tk, Tb, Tj, Ta;
161 Ti = FMA(KP951056516, Te, KP587785252 * Th);
162 Tk = FNMS(KP587785252, Te, KP951056516 * Th);
163 Ta = FNMS(KP250000000, T8, T1);
164 Tb = T9 + Ta;
165 Tj = Ta - T9;
166 ro[WS(os, 4)] = Tb - Ti;
167 ro[WS(os, 3)] = Tj + Tk;
168 ro[WS(os, 1)] = Tb + Ti;
169 ro[WS(os, 2)] = Tj - Tk;
170 }
171 {
172 E Tu, Tv, Tr, Tw, Tq;
173 Tu = FMA(KP951056516, Ts, KP587785252 * Tt);
174 Tv = FNMS(KP587785252, Ts, KP951056516 * Tt);
175 Tq = FNMS(KP250000000, Tp, To);
176 Tr = Tn + Tq;
177 Tw = Tq - Tn;
178 io[WS(os, 1)] = Tr - Tu;
179 io[WS(os, 3)] = Tw - Tv;
180 io[WS(os, 4)] = Tu + Tr;
181 io[WS(os, 2)] = Tv + Tw;
182 }
183 }
184 }
185 }
186
187 static const kdft_desc desc = { 5, "n1_5", {26, 6, 6, 0}, &GENUS, 0, 0, 0, 0 };
188
189 void X(codelet_n1_5) (planner *p) {
190 X(kdft_register) (p, n1_5, &desc);
191 }
192
193 #endif /* HAVE_FMA */