Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1bv_7.c @ 10:37bf6b4a2645

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
9:c0fb53affa76 10:37bf6b4a2645
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 07:36:58 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include n1b.h */
29
30 /*
31 * This function contains 30 FP additions, 24 FP multiplications,
32 * (or, 9 additions, 3 multiplications, 21 fused multiply/add),
33 * 37 stack variables, 6 constants, and 14 memory accesses
34 */
35 #include "n1b.h"
36
37 static void n1bv_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
40 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
41 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
42 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
43 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
44 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
45 {
46 INT i;
47 const R *xi;
48 R *xo;
49 xi = ii;
50 xo = io;
51 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(14, is), MAKE_VOLATILE_STRIDE(14, os)) {
52 V T1, T2, T3, T8, T9, T5, T6;
53 T1 = LD(&(xi[0]), ivs, &(xi[0]));
54 T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
55 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
56 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
57 T9 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
58 T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
59 T6 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
60 {
61 V Tg, T4, Te, Ta, Tf, T7;
62 Tg = VSUB(T2, T3);
63 T4 = VADD(T2, T3);
64 Te = VSUB(T8, T9);
65 Ta = VADD(T8, T9);
66 Tf = VSUB(T5, T6);
67 T7 = VADD(T5, T6);
68 {
69 V Tr, Tj, Tm, Th, To, Tb;
70 Tr = VFMA(LDK(KP554958132), Te, Tg);
71 Tj = VFNMS(LDK(KP356895867), T4, Ta);
72 Tm = VFMA(LDK(KP554958132), Tf, Te);
73 Th = VFNMS(LDK(KP554958132), Tg, Tf);
74 ST(&(xo[0]), VADD(T1, VADD(T4, VADD(T7, Ta))), ovs, &(xo[0]));
75 To = VFNMS(LDK(KP356895867), T7, T4);
76 Tb = VFNMS(LDK(KP356895867), Ta, T7);
77 {
78 V Ts, Tk, Tn, Ti;
79 Ts = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Tr, Tf));
80 Tk = VFNMS(LDK(KP692021471), Tj, T7);
81 Tn = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tm, Tg));
82 Ti = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Th, Te));
83 {
84 V Tp, Tc, Tl, Tq, Td;
85 Tp = VFNMS(LDK(KP692021471), To, Ta);
86 Tc = VFNMS(LDK(KP692021471), Tb, T4);
87 Tl = VFNMS(LDK(KP900968867), Tk, T1);
88 Tq = VFNMS(LDK(KP900968867), Tp, T1);
89 Td = VFNMS(LDK(KP900968867), Tc, T1);
90 ST(&(xo[WS(os, 5)]), VFNMSI(Tn, Tl), ovs, &(xo[WS(os, 1)]));
91 ST(&(xo[WS(os, 2)]), VFMAI(Tn, Tl), ovs, &(xo[0]));
92 ST(&(xo[WS(os, 6)]), VFNMSI(Ts, Tq), ovs, &(xo[0]));
93 ST(&(xo[WS(os, 1)]), VFMAI(Ts, Tq), ovs, &(xo[WS(os, 1)]));
94 ST(&(xo[WS(os, 4)]), VFNMSI(Ti, Td), ovs, &(xo[0]));
95 ST(&(xo[WS(os, 3)]), VFMAI(Ti, Td), ovs, &(xo[WS(os, 1)]));
96 }
97 }
98 }
99 }
100 }
101 }
102 VLEAVE();
103 }
104
105 static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {9, 3, 21, 0}, &GENUS, 0, 0, 0, 0 };
106
107 void XSIMD(codelet_n1bv_7) (planner *p) {
108 X(kdft_register) (p, n1bv_7, &desc);
109 }
110
111 #else /* HAVE_FMA */
112
113 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include n1b.h */
114
115 /*
116 * This function contains 30 FP additions, 18 FP multiplications,
117 * (or, 18 additions, 6 multiplications, 12 fused multiply/add),
118 * 24 stack variables, 6 constants, and 14 memory accesses
119 */
120 #include "n1b.h"
121
122 static void n1bv_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
123 {
124 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
125 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
126 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
127 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
128 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
129 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
130 {
131 INT i;
132 const R *xi;
133 R *xo;
134 xi = ii;
135 xo = io;
136 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(14, is), MAKE_VOLATILE_STRIDE(14, os)) {
137 V Tb, T9, Tc, T3, Te, T6, Td, T7, T8, Ti, Tj;
138 Tb = LD(&(xi[0]), ivs, &(xi[0]));
139 T7 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
140 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
141 T9 = VSUB(T7, T8);
142 Tc = VADD(T7, T8);
143 {
144 V T1, T2, T4, T5;
145 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
146 T2 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
147 T3 = VSUB(T1, T2);
148 Te = VADD(T1, T2);
149 T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
150 T5 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
151 T6 = VSUB(T4, T5);
152 Td = VADD(T4, T5);
153 }
154 ST(&(xo[0]), VADD(Tb, VADD(Te, VADD(Tc, Td))), ovs, &(xo[0]));
155 Ti = VBYI(VFNMS(LDK(KP781831482), T6, VFNMS(LDK(KP433883739), T9, VMUL(LDK(KP974927912), T3))));
156 Tj = VFMA(LDK(KP623489801), Td, VFNMS(LDK(KP900968867), Tc, VFNMS(LDK(KP222520933), Te, Tb)));
157 ST(&(xo[WS(os, 2)]), VADD(Ti, Tj), ovs, &(xo[0]));
158 ST(&(xo[WS(os, 5)]), VSUB(Tj, Ti), ovs, &(xo[WS(os, 1)]));
159 {
160 V Ta, Tf, Tg, Th;
161 Ta = VBYI(VFMA(LDK(KP433883739), T3, VFNMS(LDK(KP781831482), T9, VMUL(LDK(KP974927912), T6))));
162 Tf = VFMA(LDK(KP623489801), Tc, VFNMS(LDK(KP222520933), Td, VFNMS(LDK(KP900968867), Te, Tb)));
163 ST(&(xo[WS(os, 3)]), VADD(Ta, Tf), ovs, &(xo[WS(os, 1)]));
164 ST(&(xo[WS(os, 4)]), VSUB(Tf, Ta), ovs, &(xo[0]));
165 Tg = VBYI(VFMA(LDK(KP781831482), T3, VFMA(LDK(KP974927912), T9, VMUL(LDK(KP433883739), T6))));
166 Th = VFMA(LDK(KP623489801), Te, VFNMS(LDK(KP900968867), Td, VFNMS(LDK(KP222520933), Tc, Tb)));
167 ST(&(xo[WS(os, 1)]), VADD(Tg, Th), ovs, &(xo[WS(os, 1)]));
168 ST(&(xo[WS(os, 6)]), VSUB(Th, Tg), ovs, &(xo[0]));
169 }
170 }
171 }
172 VLEAVE();
173 }
174
175 static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {18, 6, 12, 0}, &GENUS, 0, 0, 0, 0 };
176
177 void XSIMD(codelet_n1bv_7) (planner *p) {
178 X(kdft_register) (p, n1bv_7, &desc);
179 }
180
181 #endif /* HAVE_FMA */