Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t2_4.c @ 167:bd3cc4d1df30

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 <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
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 08:04:19 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_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -name t2_4 -include dft/scalar/t.h */
29
30 /*
31 * This function contains 24 FP additions, 16 FP multiplications,
32 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
33 * 21 stack variables, 0 constants, and 16 memory accesses
34 */
35 #include "dft/scalar/t.h"
36
37 static void t2_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 * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
42 E T2, T6, T3, T5, T7, Tb, T4, Ta;
43 T2 = W[0];
44 T6 = W[3];
45 T3 = W[2];
46 T4 = T2 * T3;
47 Ta = T2 * T6;
48 T5 = W[1];
49 T7 = FMA(T5, T6, T4);
50 Tb = FNMS(T5, T3, Ta);
51 {
52 E T1, Tx, Td, Tw, Ti, Tq, Tm, Ts;
53 T1 = ri[0];
54 Tx = ii[0];
55 {
56 E T8, T9, Tc, Tv;
57 T8 = ri[WS(rs, 2)];
58 T9 = T7 * T8;
59 Tc = ii[WS(rs, 2)];
60 Tv = T7 * Tc;
61 Td = FMA(Tb, Tc, T9);
62 Tw = FNMS(Tb, T8, Tv);
63 }
64 {
65 E Tf, Tg, Th, Tp;
66 Tf = ri[WS(rs, 1)];
67 Tg = T2 * Tf;
68 Th = ii[WS(rs, 1)];
69 Tp = T2 * Th;
70 Ti = FMA(T5, Th, Tg);
71 Tq = FNMS(T5, Tf, Tp);
72 }
73 {
74 E Tj, Tk, Tl, Tr;
75 Tj = ri[WS(rs, 3)];
76 Tk = T3 * Tj;
77 Tl = ii[WS(rs, 3)];
78 Tr = T3 * Tl;
79 Tm = FMA(T6, Tl, Tk);
80 Ts = FNMS(T6, Tj, Tr);
81 }
82 {
83 E Te, Tn, Tu, Ty;
84 Te = T1 + Td;
85 Tn = Ti + Tm;
86 ri[WS(rs, 2)] = Te - Tn;
87 ri[0] = Te + Tn;
88 Tu = Tq + Ts;
89 Ty = Tw + Tx;
90 ii[0] = Tu + Ty;
91 ii[WS(rs, 2)] = Ty - Tu;
92 }
93 {
94 E To, Tt, Tz, TA;
95 To = T1 - Td;
96 Tt = Tq - Ts;
97 ri[WS(rs, 3)] = To - Tt;
98 ri[WS(rs, 1)] = To + Tt;
99 Tz = Tx - Tw;
100 TA = Ti - Tm;
101 ii[WS(rs, 1)] = Tz - TA;
102 ii[WS(rs, 3)] = TA + Tz;
103 }
104 }
105 }
106 }
107 }
108
109 static const tw_instr twinstr[] = {
110 {TW_CEXP, 0, 1},
111 {TW_CEXP, 0, 3},
112 {TW_NEXT, 1, 0}
113 };
114
115 static const ct_desc desc = { 4, "t2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
116
117 void X(codelet_t2_4) (planner *p) {
118 X(kdft_dit_register) (p, t2_4, &desc);
119 }
120 #else
121
122 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -name t2_4 -include dft/scalar/t.h */
123
124 /*
125 * This function contains 24 FP additions, 16 FP multiplications,
126 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
127 * 21 stack variables, 0 constants, and 16 memory accesses
128 */
129 #include "dft/scalar/t.h"
130
131 static void t2_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
132 {
133 {
134 INT m;
135 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(8, rs)) {
136 E T2, T4, T3, T5, T6, T8;
137 T2 = W[0];
138 T4 = W[1];
139 T3 = W[2];
140 T5 = W[3];
141 T6 = FMA(T2, T3, T4 * T5);
142 T8 = FNMS(T4, T3, T2 * T5);
143 {
144 E T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
145 T1 = ri[0];
146 Tp = ii[0];
147 T7 = ri[WS(rs, 2)];
148 T9 = ii[WS(rs, 2)];
149 Ta = FMA(T6, T7, T8 * T9);
150 To = FNMS(T8, T7, T6 * T9);
151 {
152 E Tc, Td, Tf, Tg;
153 Tc = ri[WS(rs, 1)];
154 Td = ii[WS(rs, 1)];
155 Te = FMA(T2, Tc, T4 * Td);
156 Tk = FNMS(T4, Tc, T2 * Td);
157 Tf = ri[WS(rs, 3)];
158 Tg = ii[WS(rs, 3)];
159 Th = FMA(T3, Tf, T5 * Tg);
160 Tl = FNMS(T5, Tf, T3 * Tg);
161 }
162 {
163 E Tb, Ti, Tn, Tq;
164 Tb = T1 + Ta;
165 Ti = Te + Th;
166 ri[WS(rs, 2)] = Tb - Ti;
167 ri[0] = Tb + Ti;
168 Tn = Tk + Tl;
169 Tq = To + Tp;
170 ii[0] = Tn + Tq;
171 ii[WS(rs, 2)] = Tq - Tn;
172 }
173 {
174 E Tj, Tm, Tr, Ts;
175 Tj = T1 - Ta;
176 Tm = Tk - Tl;
177 ri[WS(rs, 3)] = Tj - Tm;
178 ri[WS(rs, 1)] = Tj + Tm;
179 Tr = Tp - To;
180 Ts = Te - Th;
181 ii[WS(rs, 1)] = Tr - Ts;
182 ii[WS(rs, 3)] = Ts + Tr;
183 }
184 }
185 }
186 }
187 }
188
189 static const tw_instr twinstr[] = {
190 {TW_CEXP, 0, 1},
191 {TW_CEXP, 0, 3},
192 {TW_NEXT, 1, 0}
193 };
194
195 static const ct_desc desc = { 4, "t2_4", twinstr, &GENUS, {16, 8, 8, 0}, 0, 0, 0 };
196
197 void X(codelet_t2_4) (planner *p) {
198 X(kdft_dit_register) (p, t2_4, &desc);
199 }
200 #endif