Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2sv_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 07:37:47 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 -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */
29
30 /*
31 * This function contains 16 FP additions, 0 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 0 fused multiply/add),
33 * 25 stack variables, 0 constants, and 18 memory accesses
34 */
35 #include "n2s.h"
36
37 static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 {
40 INT i;
41 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
42 V T1, T2, T7, T8, T4, T5, Tc, Td;
43 T1 = LD(&(ri[0]), ivs, &(ri[0]));
44 T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
45 T7 = LD(&(ii[0]), ivs, &(ii[0]));
46 T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
47 T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
48 T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
49 Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
50 Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
51 {
52 V T3, Tb, T9, Tf, T6, Ta, Te, Tg;
53 T3 = VADD(T1, T2);
54 Tb = VSUB(T1, T2);
55 T9 = VSUB(T7, T8);
56 Tf = VADD(T7, T8);
57 T6 = VADD(T4, T5);
58 Ta = VSUB(T4, T5);
59 Te = VSUB(Tc, Td);
60 Tg = VADD(Tc, Td);
61 {
62 V Th, Ti, Tj, Tk;
63 Th = VADD(Ta, T9);
64 STM4(&(io[3]), Th, ovs, &(io[1]));
65 Ti = VSUB(T9, Ta);
66 STM4(&(io[1]), Ti, ovs, &(io[1]));
67 Tj = VADD(T3, T6);
68 STM4(&(ro[0]), Tj, ovs, &(ro[0]));
69 Tk = VSUB(T3, T6);
70 STM4(&(ro[2]), Tk, ovs, &(ro[0]));
71 {
72 V Tl, Tm, Tn, To;
73 Tl = VADD(Tf, Tg);
74 STM4(&(io[0]), Tl, ovs, &(io[0]));
75 Tm = VSUB(Tf, Tg);
76 STM4(&(io[2]), Tm, ovs, &(io[0]));
77 STN4(&(io[0]), Tl, Ti, Tm, Th, ovs);
78 Tn = VSUB(Tb, Te);
79 STM4(&(ro[3]), Tn, ovs, &(ro[1]));
80 To = VADD(Tb, Te);
81 STM4(&(ro[1]), To, ovs, &(ro[1]));
82 STN4(&(ro[0]), Tj, To, Tk, Tn, ovs);
83 }
84 }
85 }
86 }
87 }
88 VLEAVE();
89 }
90
91 static const kdft_desc desc = { 4, XSIMD_STRING("n2sv_4"), {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 };
92
93 void XSIMD(codelet_n2sv_4) (planner *p) {
94 X(kdft_register) (p, n2sv_4, &desc);
95 }
96
97 #else /* HAVE_FMA */
98
99 /* Generated by: ../../../genfft/gen_notw.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2sv_4 -with-ostride 1 -include n2s.h -store-multiple 4 */
100
101 /*
102 * This function contains 16 FP additions, 0 FP multiplications,
103 * (or, 16 additions, 0 multiplications, 0 fused multiply/add),
104 * 17 stack variables, 0 constants, and 18 memory accesses
105 */
106 #include "n2s.h"
107
108 static void n2sv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
109 {
110 {
111 INT i;
112 for (i = v; i > 0; i = i - (2 * VL), ri = ri + ((2 * VL) * ivs), ii = ii + ((2 * VL) * ivs), ro = ro + ((2 * VL) * ovs), io = io + ((2 * VL) * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
113 V T3, Tb, T9, Tf, T6, Ta, Te, Tg;
114 {
115 V T1, T2, T7, T8;
116 T1 = LD(&(ri[0]), ivs, &(ri[0]));
117 T2 = LD(&(ri[WS(is, 2)]), ivs, &(ri[0]));
118 T3 = VADD(T1, T2);
119 Tb = VSUB(T1, T2);
120 T7 = LD(&(ii[0]), ivs, &(ii[0]));
121 T8 = LD(&(ii[WS(is, 2)]), ivs, &(ii[0]));
122 T9 = VSUB(T7, T8);
123 Tf = VADD(T7, T8);
124 }
125 {
126 V T4, T5, Tc, Td;
127 T4 = LD(&(ri[WS(is, 1)]), ivs, &(ri[WS(is, 1)]));
128 T5 = LD(&(ri[WS(is, 3)]), ivs, &(ri[WS(is, 1)]));
129 T6 = VADD(T4, T5);
130 Ta = VSUB(T4, T5);
131 Tc = LD(&(ii[WS(is, 1)]), ivs, &(ii[WS(is, 1)]));
132 Td = LD(&(ii[WS(is, 3)]), ivs, &(ii[WS(is, 1)]));
133 Te = VSUB(Tc, Td);
134 Tg = VADD(Tc, Td);
135 }
136 {
137 V Th, Ti, Tj, Tk;
138 Th = VSUB(T3, T6);
139 STM4(&(ro[2]), Th, ovs, &(ro[0]));
140 Ti = VSUB(Tf, Tg);
141 STM4(&(io[2]), Ti, ovs, &(io[0]));
142 Tj = VADD(T3, T6);
143 STM4(&(ro[0]), Tj, ovs, &(ro[0]));
144 Tk = VADD(Tf, Tg);
145 STM4(&(io[0]), Tk, ovs, &(io[0]));
146 {
147 V Tl, Tm, Tn, To;
148 Tl = VSUB(T9, Ta);
149 STM4(&(io[1]), Tl, ovs, &(io[1]));
150 Tm = VADD(Tb, Te);
151 STM4(&(ro[1]), Tm, ovs, &(ro[1]));
152 Tn = VADD(Ta, T9);
153 STM4(&(io[3]), Tn, ovs, &(io[1]));
154 STN4(&(io[0]), Tk, Tl, Ti, Tn, ovs);
155 To = VSUB(Tb, Te);
156 STM4(&(ro[3]), To, ovs, &(ro[1]));
157 STN4(&(ro[0]), Tj, Tm, Th, To, ovs);
158 }
159 }
160 }
161 }
162 VLEAVE();
163 }
164
165 static const kdft_desc desc = { 4, XSIMD_STRING("n2sv_4"), {16, 0, 0, 0}, &GENUS, 0, 1, 0, 0 };
166
167 void XSIMD(codelet_n2sv_4) (planner *p) {
168 X(kdft_register) (p, n2sv_4, &desc);
169 }
170
171 #endif /* HAVE_FMA */