Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n1fv_8.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 08:04:51 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 -n 8 -name n1fv_8 -include dft/simd/n1f.h */
29
30 /*
31 * This function contains 26 FP additions, 10 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
33 * 22 stack variables, 1 constants, and 16 memory accesses
34 */
35 #include "dft/simd/n1f.h"
36
37 static void n1fv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT i;
42 const R *xi;
43 R *xo;
44 xi = ri;
45 xo = ro;
46 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
47 V T3, Tj, Te, Tk, Ta, Tn, Tf, Tm;
48 {
49 V T1, T2, Tc, Td;
50 T1 = LD(&(xi[0]), ivs, &(xi[0]));
51 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
52 T3 = VSUB(T1, T2);
53 Tj = VADD(T1, T2);
54 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
55 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
56 Te = VSUB(Tc, Td);
57 Tk = VADD(Tc, Td);
58 {
59 V T4, T5, T6, T7, T8, T9;
60 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
61 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
62 T6 = VSUB(T4, T5);
63 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
64 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
65 T9 = VSUB(T7, T8);
66 Ta = VADD(T6, T9);
67 Tn = VADD(T7, T8);
68 Tf = VSUB(T9, T6);
69 Tm = VADD(T4, T5);
70 }
71 }
72 {
73 V Tb, Tg, Tp, Tq;
74 Tb = VFMA(LDK(KP707106781), Ta, T3);
75 Tg = VFNMS(LDK(KP707106781), Tf, Te);
76 ST(&(xo[WS(os, 1)]), VFNMSI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
77 ST(&(xo[WS(os, 7)]), VFMAI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
78 Tp = VSUB(Tj, Tk);
79 Tq = VSUB(Tn, Tm);
80 ST(&(xo[WS(os, 6)]), VFNMSI(Tq, Tp), ovs, &(xo[0]));
81 ST(&(xo[WS(os, 2)]), VFMAI(Tq, Tp), ovs, &(xo[0]));
82 }
83 {
84 V Th, Ti, Tl, To;
85 Th = VFNMS(LDK(KP707106781), Ta, T3);
86 Ti = VFMA(LDK(KP707106781), Tf, Te);
87 ST(&(xo[WS(os, 5)]), VFNMSI(Ti, Th), ovs, &(xo[WS(os, 1)]));
88 ST(&(xo[WS(os, 3)]), VFMAI(Ti, Th), ovs, &(xo[WS(os, 1)]));
89 Tl = VADD(Tj, Tk);
90 To = VADD(Tm, Tn);
91 ST(&(xo[WS(os, 4)]), VSUB(Tl, To), ovs, &(xo[0]));
92 ST(&(xo[0]), VADD(Tl, To), ovs, &(xo[0]));
93 }
94 }
95 }
96 VLEAVE();
97 }
98
99 static const kdft_desc desc = { 8, XSIMD_STRING("n1fv_8"), {16, 0, 10, 0}, &GENUS, 0, 0, 0, 0 };
100
101 void XSIMD(codelet_n1fv_8) (planner *p) {
102 X(kdft_register) (p, n1fv_8, &desc);
103 }
104
105 #else
106
107 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name n1fv_8 -include dft/simd/n1f.h */
108
109 /*
110 * This function contains 26 FP additions, 2 FP multiplications,
111 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
112 * 22 stack variables, 1 constants, and 16 memory accesses
113 */
114 #include "dft/simd/n1f.h"
115
116 static void n1fv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
117 {
118 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
119 {
120 INT i;
121 const R *xi;
122 R *xo;
123 xi = ri;
124 xo = ro;
125 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(16, is), MAKE_VOLATILE_STRIDE(16, os)) {
126 V T3, Tj, Tf, Tk, Ta, Tn, Tc, Tm;
127 {
128 V T1, T2, Td, Te;
129 T1 = LD(&(xi[0]), ivs, &(xi[0]));
130 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
131 T3 = VSUB(T1, T2);
132 Tj = VADD(T1, T2);
133 Td = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
134 Te = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
135 Tf = VSUB(Td, Te);
136 Tk = VADD(Td, Te);
137 {
138 V T4, T5, T6, T7, T8, T9;
139 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
140 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
141 T6 = VSUB(T4, T5);
142 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
143 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
144 T9 = VSUB(T7, T8);
145 Ta = VMUL(LDK(KP707106781), VADD(T6, T9));
146 Tn = VADD(T7, T8);
147 Tc = VMUL(LDK(KP707106781), VSUB(T9, T6));
148 Tm = VADD(T4, T5);
149 }
150 }
151 {
152 V Tb, Tg, Tp, Tq;
153 Tb = VADD(T3, Ta);
154 Tg = VBYI(VSUB(Tc, Tf));
155 ST(&(xo[WS(os, 7)]), VSUB(Tb, Tg), ovs, &(xo[WS(os, 1)]));
156 ST(&(xo[WS(os, 1)]), VADD(Tb, Tg), ovs, &(xo[WS(os, 1)]));
157 Tp = VSUB(Tj, Tk);
158 Tq = VBYI(VSUB(Tn, Tm));
159 ST(&(xo[WS(os, 6)]), VSUB(Tp, Tq), ovs, &(xo[0]));
160 ST(&(xo[WS(os, 2)]), VADD(Tp, Tq), ovs, &(xo[0]));
161 }
162 {
163 V Th, Ti, Tl, To;
164 Th = VSUB(T3, Ta);
165 Ti = VBYI(VADD(Tf, Tc));
166 ST(&(xo[WS(os, 5)]), VSUB(Th, Ti), ovs, &(xo[WS(os, 1)]));
167 ST(&(xo[WS(os, 3)]), VADD(Th, Ti), ovs, &(xo[WS(os, 1)]));
168 Tl = VADD(Tj, Tk);
169 To = VADD(Tm, Tn);
170 ST(&(xo[WS(os, 4)]), VSUB(Tl, To), ovs, &(xo[0]));
171 ST(&(xo[0]), VADD(Tl, To), ovs, &(xo[0]));
172 }
173 }
174 }
175 VLEAVE();
176 }
177
178 static const kdft_desc desc = { 8, XSIMD_STRING("n1fv_8"), {26, 2, 0, 0}, &GENUS, 0, 0, 0, 0 };
179
180 void XSIMD(codelet_n1fv_8) (planner *p) {
181 X(kdft_register) (p, n1fv_8, &desc);
182 }
183
184 #endif