Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/t1_4.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 07:35:47 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 -n 4 -name t1_4 -include t.h */
29
30 /*
31 * This function contains 22 FP additions, 12 FP multiplications,
32 * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
33 * 31 stack variables, 0 constants, and 16 memory accesses
34 */
35 #include "t.h"
36
37 static void t1_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 {
40 INT m;
41 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs)) {
42 E To, Te, Tm, T8, Tw, Tx, Tq, Tk;
43 {
44 E T1, Tv, Tu, T7, Tg, Tj, Tf, Ti, Tp, Th;
45 T1 = ri[0];
46 Tv = ii[0];
47 {
48 E T3, T6, T2, T5;
49 T3 = ri[WS(rs, 2)];
50 T6 = ii[WS(rs, 2)];
51 T2 = W[2];
52 T5 = W[3];
53 {
54 E Ta, Td, Tc, Tn, Tb, Tt, T4, T9;
55 Ta = ri[WS(rs, 1)];
56 Td = ii[WS(rs, 1)];
57 Tt = T2 * T6;
58 T4 = T2 * T3;
59 T9 = W[0];
60 Tc = W[1];
61 Tu = FNMS(T5, T3, Tt);
62 T7 = FMA(T5, T6, T4);
63 Tn = T9 * Td;
64 Tb = T9 * Ta;
65 Tg = ri[WS(rs, 3)];
66 Tj = ii[WS(rs, 3)];
67 To = FNMS(Tc, Ta, Tn);
68 Te = FMA(Tc, Td, Tb);
69 Tf = W[4];
70 Ti = W[5];
71 }
72 }
73 Tm = T1 - T7;
74 T8 = T1 + T7;
75 Tw = Tu + Tv;
76 Tx = Tv - Tu;
77 Tp = Tf * Tj;
78 Th = Tf * Tg;
79 Tq = FNMS(Ti, Tg, Tp);
80 Tk = FMA(Ti, Tj, Th);
81 }
82 {
83 E Ts, Tr, Tl, Ty;
84 Ts = To + Tq;
85 Tr = To - Tq;
86 Tl = Te + Tk;
87 Ty = Te - Tk;
88 ri[WS(rs, 1)] = Tm + Tr;
89 ri[WS(rs, 3)] = Tm - Tr;
90 ii[WS(rs, 2)] = Tw - Ts;
91 ii[0] = Ts + Tw;
92 ii[WS(rs, 3)] = Ty + Tx;
93 ii[WS(rs, 1)] = Tx - Ty;
94 ri[0] = T8 + Tl;
95 ri[WS(rs, 2)] = T8 - Tl;
96 }
97 }
98 }
99 }
100
101 static const tw_instr twinstr[] = {
102 {TW_FULL, 0, 4},
103 {TW_NEXT, 1, 0}
104 };
105
106 static const ct_desc desc = { 4, "t1_4", twinstr, &GENUS, {16, 6, 6, 0}, 0, 0, 0 };
107
108 void X(codelet_t1_4) (planner *p) {
109 X(kdft_dit_register) (p, t1_4, &desc);
110 }
111 #else /* HAVE_FMA */
112
113 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -n 4 -name t1_4 -include t.h */
114
115 /*
116 * This function contains 22 FP additions, 12 FP multiplications,
117 * (or, 16 additions, 6 multiplications, 6 fused multiply/add),
118 * 13 stack variables, 0 constants, and 16 memory accesses
119 */
120 #include "t.h"
121
122 static void t1_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
123 {
124 {
125 INT m;
126 for (m = mb, W = W + (mb * 6); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 6, MAKE_VOLATILE_STRIDE(8, rs)) {
127 E T1, Tp, T6, To, Tc, Tk, Th, Tl;
128 T1 = ri[0];
129 Tp = ii[0];
130 {
131 E T3, T5, T2, T4;
132 T3 = ri[WS(rs, 2)];
133 T5 = ii[WS(rs, 2)];
134 T2 = W[2];
135 T4 = W[3];
136 T6 = FMA(T2, T3, T4 * T5);
137 To = FNMS(T4, T3, T2 * T5);
138 }
139 {
140 E T9, Tb, T8, Ta;
141 T9 = ri[WS(rs, 1)];
142 Tb = ii[WS(rs, 1)];
143 T8 = W[0];
144 Ta = W[1];
145 Tc = FMA(T8, T9, Ta * Tb);
146 Tk = FNMS(Ta, T9, T8 * Tb);
147 }
148 {
149 E Te, Tg, Td, Tf;
150 Te = ri[WS(rs, 3)];
151 Tg = ii[WS(rs, 3)];
152 Td = W[4];
153 Tf = W[5];
154 Th = FMA(Td, Te, Tf * Tg);
155 Tl = FNMS(Tf, Te, Td * Tg);
156 }
157 {
158 E T7, Ti, Tn, Tq;
159 T7 = T1 + T6;
160 Ti = Tc + Th;
161 ri[WS(rs, 2)] = T7 - Ti;
162 ri[0] = T7 + Ti;
163 Tn = Tk + Tl;
164 Tq = To + Tp;
165 ii[0] = Tn + Tq;
166 ii[WS(rs, 2)] = Tq - Tn;
167 }
168 {
169 E Tj, Tm, Tr, Ts;
170 Tj = T1 - T6;
171 Tm = Tk - Tl;
172 ri[WS(rs, 3)] = Tj - Tm;
173 ri[WS(rs, 1)] = Tj + Tm;
174 Tr = Tp - To;
175 Ts = Tc - Th;
176 ii[WS(rs, 1)] = Tr - Ts;
177 ii[WS(rs, 3)] = Ts + Tr;
178 }
179 }
180 }
181 }
182
183 static const tw_instr twinstr[] = {
184 {TW_FULL, 0, 4},
185 {TW_NEXT, 1, 0}
186 };
187
188 static const ct_desc desc = { 4, "t1_4", twinstr, &GENUS, {16, 6, 6, 0}, 0, 0, 0 };
189
190 void X(codelet_t1_4) (planner *p) {
191 X(kdft_dit_register) (p, t1_4, &desc);
192 }
193 #endif /* HAVE_FMA */