Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n1bv_7.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 08:04:55 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_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include dft/simd/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 * 33 stack variables, 6 constants, and 14 memory accesses
34 */
35 #include "dft/simd/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(KP554958132, +0.554958132087371191422194871006410481067288862);
44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
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, T4, Tg, Ta, Te, T7, Tf, Tb, Th, Tr, To, Tm, Tj, T2, T3;
53 V Ts, Tq, Tp;
54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
55 T2 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
56 T3 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
57 T4 = VADD(T2, T3);
58 Tg = VSUB(T2, T3);
59 {
60 V T8, T9, T5, T6;
61 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
62 T9 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
63 Ta = VADD(T8, T9);
64 Te = VSUB(T8, T9);
65 T5 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
66 T6 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
67 T7 = VADD(T5, T6);
68 Tf = VSUB(T5, T6);
69 }
70 Tb = VFNMS(LDK(KP356895867), Ta, T7);
71 Th = VFNMS(LDK(KP554958132), Tg, Tf);
72 Tr = VFMA(LDK(KP554958132), Te, Tg);
73 To = VFNMS(LDK(KP356895867), T7, T4);
74 Tm = VFMA(LDK(KP554958132), Tf, Te);
75 Tj = VFNMS(LDK(KP356895867), T4, Ta);
76 ST(&(xo[0]), VADD(T1, VADD(T4, VADD(T7, Ta))), ovs, &(xo[0]));
77 Ts = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Tr, Tf));
78 Tp = VFNMS(LDK(KP692021471), To, Ta);
79 Tq = VFNMS(LDK(KP900968867), Tp, T1);
80 ST(&(xo[WS(os, 1)]), VFMAI(Ts, Tq), ovs, &(xo[WS(os, 1)]));
81 ST(&(xo[WS(os, 6)]), VFNMSI(Ts, Tq), ovs, &(xo[0]));
82 {
83 V Ti, Td, Tc, Tn, Tl, Tk;
84 Ti = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Th, Te));
85 Tc = VFNMS(LDK(KP692021471), Tb, T4);
86 Td = VFNMS(LDK(KP900968867), Tc, T1);
87 ST(&(xo[WS(os, 3)]), VFMAI(Ti, Td), ovs, &(xo[WS(os, 1)]));
88 ST(&(xo[WS(os, 4)]), VFNMSI(Ti, Td), ovs, &(xo[0]));
89 Tn = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tm, Tg));
90 Tk = VFNMS(LDK(KP692021471), Tj, T7);
91 Tl = VFNMS(LDK(KP900968867), Tk, T1);
92 ST(&(xo[WS(os, 2)]), VFMAI(Tn, Tl), ovs, &(xo[0]));
93 ST(&(xo[WS(os, 5)]), VFNMSI(Tn, Tl), ovs, &(xo[WS(os, 1)]));
94 }
95 }
96 }
97 VLEAVE();
98 }
99
100 static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {9, 3, 21, 0}, &GENUS, 0, 0, 0, 0 };
101
102 void XSIMD(codelet_n1bv_7) (planner *p) {
103 X(kdft_register) (p, n1bv_7, &desc);
104 }
105
106 #else
107
108 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 7 -name n1bv_7 -include dft/simd/n1b.h */
109
110 /*
111 * This function contains 30 FP additions, 18 FP multiplications,
112 * (or, 18 additions, 6 multiplications, 12 fused multiply/add),
113 * 24 stack variables, 6 constants, and 14 memory accesses
114 */
115 #include "dft/simd/n1b.h"
116
117 static void n1bv_7(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
118 {
119 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
120 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
121 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
122 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
123 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
124 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
125 {
126 INT i;
127 const R *xi;
128 R *xo;
129 xi = ii;
130 xo = io;
131 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)) {
132 V Tb, T9, Tc, T3, Te, T6, Td, T7, T8, Ti, Tj;
133 Tb = LD(&(xi[0]), ivs, &(xi[0]));
134 T7 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
135 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
136 T9 = VSUB(T7, T8);
137 Tc = VADD(T7, T8);
138 {
139 V T1, T2, T4, T5;
140 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
141 T2 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
142 T3 = VSUB(T1, T2);
143 Te = VADD(T1, T2);
144 T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
145 T5 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
146 T6 = VSUB(T4, T5);
147 Td = VADD(T4, T5);
148 }
149 ST(&(xo[0]), VADD(Tb, VADD(Te, VADD(Tc, Td))), ovs, &(xo[0]));
150 Ti = VBYI(VFNMS(LDK(KP781831482), T6, VFNMS(LDK(KP433883739), T9, VMUL(LDK(KP974927912), T3))));
151 Tj = VFMA(LDK(KP623489801), Td, VFNMS(LDK(KP900968867), Tc, VFNMS(LDK(KP222520933), Te, Tb)));
152 ST(&(xo[WS(os, 2)]), VADD(Ti, Tj), ovs, &(xo[0]));
153 ST(&(xo[WS(os, 5)]), VSUB(Tj, Ti), ovs, &(xo[WS(os, 1)]));
154 {
155 V Ta, Tf, Tg, Th;
156 Ta = VBYI(VFMA(LDK(KP433883739), T3, VFNMS(LDK(KP781831482), T9, VMUL(LDK(KP974927912), T6))));
157 Tf = VFMA(LDK(KP623489801), Tc, VFNMS(LDK(KP222520933), Td, VFNMS(LDK(KP900968867), Te, Tb)));
158 ST(&(xo[WS(os, 3)]), VADD(Ta, Tf), ovs, &(xo[WS(os, 1)]));
159 ST(&(xo[WS(os, 4)]), VSUB(Tf, Ta), ovs, &(xo[0]));
160 Tg = VBYI(VFMA(LDK(KP781831482), T3, VFMA(LDK(KP974927912), T9, VMUL(LDK(KP433883739), T6))));
161 Th = VFMA(LDK(KP623489801), Te, VFNMS(LDK(KP900968867), Td, VFNMS(LDK(KP222520933), Tc, Tb)));
162 ST(&(xo[WS(os, 1)]), VADD(Tg, Th), ovs, &(xo[WS(os, 1)]));
163 ST(&(xo[WS(os, 6)]), VSUB(Th, Tg), ovs, &(xo[0]));
164 }
165 }
166 }
167 VLEAVE();
168 }
169
170 static const kdft_desc desc = { 7, XSIMD_STRING("n1bv_7"), {18, 6, 12, 0}, &GENUS, 0, 0, 0, 0 };
171
172 void XSIMD(codelet_n1bv_7) (planner *p) {
173 X(kdft_register) (p, n1bv_7, &desc);
174 }
175
176 #endif