Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1bv_8.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 07:36:59 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 8 -name n1bv_8 -include n1b.h */
29
30 /*
31 * This function contains 26 FP additions, 10 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
33 * 30 stack variables, 1 constants, and 16 memory accesses
34 */
35 #include "n1b.h"
36
37 static void n1bv_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 = ii;
45 xo = io;
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 T1, T2, Tc, Td, T4, T5, T7, T8;
48 T1 = LD(&(xi[0]), ivs, &(xi[0]));
49 T2 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
50 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
51 Td = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
52 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
53 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
54 T7 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
55 T8 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
56 {
57 V T3, Tj, Te, Tk, T6, Tm, T9, Tn, Tp, Tl;
58 T3 = VSUB(T1, T2);
59 Tj = VADD(T1, T2);
60 Te = VSUB(Tc, Td);
61 Tk = VADD(Tc, Td);
62 T6 = VSUB(T4, T5);
63 Tm = VADD(T4, T5);
64 T9 = VSUB(T7, T8);
65 Tn = VADD(T7, T8);
66 Tp = VADD(Tj, Tk);
67 Tl = VSUB(Tj, Tk);
68 {
69 V Tq, To, Ta, Tf;
70 Tq = VADD(Tm, Tn);
71 To = VSUB(Tm, Tn);
72 Ta = VADD(T6, T9);
73 Tf = VSUB(T6, T9);
74 {
75 V Tg, Ti, Tb, Th;
76 ST(&(xo[WS(os, 2)]), VFMAI(To, Tl), ovs, &(xo[0]));
77 ST(&(xo[WS(os, 6)]), VFNMSI(To, Tl), ovs, &(xo[0]));
78 ST(&(xo[0]), VADD(Tp, Tq), ovs, &(xo[0]));
79 ST(&(xo[WS(os, 4)]), VSUB(Tp, Tq), ovs, &(xo[0]));
80 Tg = VFNMS(LDK(KP707106781), Tf, Te);
81 Ti = VFMA(LDK(KP707106781), Tf, Te);
82 Tb = VFNMS(LDK(KP707106781), Ta, T3);
83 Th = VFMA(LDK(KP707106781), Ta, T3);
84 ST(&(xo[WS(os, 7)]), VFNMSI(Ti, Th), ovs, &(xo[WS(os, 1)]));
85 ST(&(xo[WS(os, 1)]), VFMAI(Ti, Th), ovs, &(xo[WS(os, 1)]));
86 ST(&(xo[WS(os, 5)]), VFMAI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
87 ST(&(xo[WS(os, 3)]), VFNMSI(Tg, Tb), ovs, &(xo[WS(os, 1)]));
88 }
89 }
90 }
91 }
92 }
93 VLEAVE();
94 }
95
96 static const kdft_desc desc = { 8, XSIMD_STRING("n1bv_8"), {16, 0, 10, 0}, &GENUS, 0, 0, 0, 0 };
97
98 void XSIMD(codelet_n1bv_8) (planner *p) {
99 X(kdft_register) (p, n1bv_8, &desc);
100 }
101
102 #else /* HAVE_FMA */
103
104 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n1bv_8 -include n1b.h */
105
106 /*
107 * This function contains 26 FP additions, 2 FP multiplications,
108 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
109 * 22 stack variables, 1 constants, and 16 memory accesses
110 */
111 #include "n1b.h"
112
113 static void n1bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
114 {
115 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
116 {
117 INT i;
118 const R *xi;
119 R *xo;
120 xi = ii;
121 xo = io;
122 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)) {
123 V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm;
124 {
125 V T8, T9, Tc, Td;
126 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
127 T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
128 Ta = VSUB(T8, T9);
129 Tk = VADD(T8, T9);
130 Tc = LD(&(xi[0]), ivs, &(xi[0]));
131 Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
132 Te = VSUB(Tc, Td);
133 Tj = VADD(Tc, Td);
134 {
135 V T1, T2, T3, T4, T5, T6;
136 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
137 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
138 T3 = VSUB(T1, T2);
139 T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
140 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
141 T6 = VSUB(T4, T5);
142 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
143 Tn = VADD(T4, T5);
144 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
145 Tm = VADD(T1, T2);
146 }
147 }
148 {
149 V Tb, Tg, Tp, Tq;
150 Tb = VBYI(VSUB(T7, Ta));
151 Tg = VSUB(Te, Tf);
152 ST(&(xo[WS(os, 3)]), VADD(Tb, Tg), ovs, &(xo[WS(os, 1)]));
153 ST(&(xo[WS(os, 5)]), VSUB(Tg, Tb), ovs, &(xo[WS(os, 1)]));
154 Tp = VADD(Tj, Tk);
155 Tq = VADD(Tm, Tn);
156 ST(&(xo[WS(os, 4)]), VSUB(Tp, Tq), ovs, &(xo[0]));
157 ST(&(xo[0]), VADD(Tp, Tq), ovs, &(xo[0]));
158 }
159 {
160 V Th, Ti, Tl, To;
161 Th = VBYI(VADD(Ta, T7));
162 Ti = VADD(Te, Tf);
163 ST(&(xo[WS(os, 1)]), VADD(Th, Ti), ovs, &(xo[WS(os, 1)]));
164 ST(&(xo[WS(os, 7)]), VSUB(Ti, Th), ovs, &(xo[WS(os, 1)]));
165 Tl = VSUB(Tj, Tk);
166 To = VBYI(VSUB(Tm, Tn));
167 ST(&(xo[WS(os, 6)]), VSUB(Tl, To), ovs, &(xo[0]));
168 ST(&(xo[WS(os, 2)]), VADD(Tl, To), ovs, &(xo[0]));
169 }
170 }
171 }
172 VLEAVE();
173 }
174
175 static const kdft_desc desc = { 8, XSIMD_STRING("n1bv_8"), {26, 2, 0, 0}, &GENUS, 0, 0, 0, 0 };
176
177 void XSIMD(codelet_n1bv_8) (planner *p) {
178 X(kdft_register) (p, n1bv_8, &desc);
179 }
180
181 #endif /* HAVE_FMA */