Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/n1_5.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 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 Thu May 24 08:04:10 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -compact -variables 4 -pipeline-latency 4 -n 5 -name n1_5 -include dft/scalar/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 * 21 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "dft/scalar/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 T1, Tl, T8, Tt, Ta, Ts, Te, Tq, Th, To;
47 T1 = ri[0];
48 Tl = ii[0];
49 {
50 E T2, T3, T4, T5, T6, T7;
51 T2 = ri[WS(is, 1)];
52 T3 = ri[WS(is, 4)];
53 T4 = T2 + T3;
54 T5 = ri[WS(is, 2)];
55 T6 = ri[WS(is, 3)];
56 T7 = T5 + T6;
57 T8 = T4 + T7;
58 Tt = T5 - T6;
59 Ta = T4 - T7;
60 Ts = T2 - T3;
61 }
62 {
63 E Tc, Td, Tm, Tf, Tg, Tn;
64 Tc = ii[WS(is, 1)];
65 Td = ii[WS(is, 4)];
66 Tm = Tc + Td;
67 Tf = ii[WS(is, 2)];
68 Tg = ii[WS(is, 3)];
69 Tn = Tf + Tg;
70 Te = Tc - Td;
71 Tq = Tm - Tn;
72 Th = Tf - Tg;
73 To = Tm + Tn;
74 }
75 ro[0] = T1 + T8;
76 io[0] = Tl + To;
77 {
78 E Ti, Tk, Tb, Tj, T9;
79 Ti = FMA(KP618033988, Th, Te);
80 Tk = FNMS(KP618033988, Te, Th);
81 T9 = FNMS(KP250000000, T8, T1);
82 Tb = FMA(KP559016994, Ta, T9);
83 Tj = FNMS(KP559016994, Ta, T9);
84 ro[WS(os, 4)] = FNMS(KP951056516, Ti, Tb);
85 ro[WS(os, 3)] = FMA(KP951056516, Tk, Tj);
86 ro[WS(os, 1)] = FMA(KP951056516, Ti, Tb);
87 ro[WS(os, 2)] = FNMS(KP951056516, Tk, Tj);
88 }
89 {
90 E Tu, Tw, Tr, Tv, Tp;
91 Tu = FMA(KP618033988, Tt, Ts);
92 Tw = FNMS(KP618033988, Ts, Tt);
93 Tp = FNMS(KP250000000, To, Tl);
94 Tr = FMA(KP559016994, Tq, Tp);
95 Tv = FNMS(KP559016994, Tq, Tp);
96 io[WS(os, 1)] = FNMS(KP951056516, Tu, Tr);
97 io[WS(os, 3)] = FNMS(KP951056516, Tw, Tv);
98 io[WS(os, 4)] = FMA(KP951056516, Tu, Tr);
99 io[WS(os, 2)] = FMA(KP951056516, Tw, Tv);
100 }
101 }
102 }
103 }
104
105 static const kdft_desc desc = { 5, "n1_5", {14, 0, 18, 0}, &GENUS, 0, 0, 0, 0 };
106
107 void X(codelet_n1_5) (planner *p) {
108 X(kdft_register) (p, n1_5, &desc);
109 }
110
111 #else
112
113 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 5 -name n1_5 -include dft/scalar/n.h */
114
115 /*
116 * This function contains 32 FP additions, 12 FP multiplications,
117 * (or, 26 additions, 6 multiplications, 6 fused multiply/add),
118 * 21 stack variables, 4 constants, and 20 memory accesses
119 */
120 #include "dft/scalar/n.h"
121
122 static void n1_5(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
123 {
124 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
125 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
126 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
127 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
128 {
129 INT i;
130 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)) {
131 E T1, To, T8, Tt, T9, Ts, Te, Tp, Th, Tn;
132 T1 = ri[0];
133 To = ii[0];
134 {
135 E T2, T3, T4, T5, T6, T7;
136 T2 = ri[WS(is, 1)];
137 T3 = ri[WS(is, 4)];
138 T4 = T2 + T3;
139 T5 = ri[WS(is, 2)];
140 T6 = ri[WS(is, 3)];
141 T7 = T5 + T6;
142 T8 = T4 + T7;
143 Tt = T5 - T6;
144 T9 = KP559016994 * (T4 - T7);
145 Ts = T2 - T3;
146 }
147 {
148 E Tc, Td, Tl, Tf, Tg, Tm;
149 Tc = ii[WS(is, 1)];
150 Td = ii[WS(is, 4)];
151 Tl = Tc + Td;
152 Tf = ii[WS(is, 2)];
153 Tg = ii[WS(is, 3)];
154 Tm = Tf + Tg;
155 Te = Tc - Td;
156 Tp = Tl + Tm;
157 Th = Tf - Tg;
158 Tn = KP559016994 * (Tl - Tm);
159 }
160 ro[0] = T1 + T8;
161 io[0] = To + Tp;
162 {
163 E Ti, Tk, Tb, Tj, Ta;
164 Ti = FMA(KP951056516, Te, KP587785252 * Th);
165 Tk = FNMS(KP587785252, Te, KP951056516 * Th);
166 Ta = FNMS(KP250000000, T8, T1);
167 Tb = T9 + Ta;
168 Tj = Ta - T9;
169 ro[WS(os, 4)] = Tb - Ti;
170 ro[WS(os, 3)] = Tj + Tk;
171 ro[WS(os, 1)] = Tb + Ti;
172 ro[WS(os, 2)] = Tj - Tk;
173 }
174 {
175 E Tu, Tv, Tr, Tw, Tq;
176 Tu = FMA(KP951056516, Ts, KP587785252 * Tt);
177 Tv = FNMS(KP587785252, Ts, KP951056516 * Tt);
178 Tq = FNMS(KP250000000, Tp, To);
179 Tr = Tn + Tq;
180 Tw = Tq - Tn;
181 io[WS(os, 1)] = Tr - Tu;
182 io[WS(os, 3)] = Tw - Tv;
183 io[WS(os, 4)] = Tu + Tr;
184 io[WS(os, 2)] = Tv + Tw;
185 }
186 }
187 }
188 }
189
190 static const kdft_desc desc = { 5, "n1_5", {26, 6, 6, 0}, &GENUS, 0, 0, 0, 0 };
191
192 void X(codelet_n1_5) (planner *p) {
193 X(kdft_register) (p, n1_5, &desc);
194 }
195
196 #endif