Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2fv_4.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 07:37:21 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 -n 4 -name n2fv_4 -with-ostride 2 -include n2f.h -store-multiple 2 */
29
30 /*
31 * This function contains 8 FP additions, 2 FP multiplications,
32 * (or, 6 additions, 0 multiplications, 2 fused multiply/add),
33 * 15 stack variables, 0 constants, and 10 memory accesses
34 */
35 #include "n2f.h"
36
37 static void n2fv_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 const R *xi;
42 R *xo;
43 xi = ri;
44 xo = ro;
45 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
46 V T1, T2, T4, T5;
47 T1 = LD(&(xi[0]), ivs, &(xi[0]));
48 T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
49 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
50 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
51 {
52 V T3, T7, T6, T8;
53 T3 = VSUB(T1, T2);
54 T7 = VADD(T1, T2);
55 T6 = VSUB(T4, T5);
56 T8 = VADD(T4, T5);
57 {
58 V T9, Ta, Tb, Tc;
59 T9 = VSUB(T7, T8);
60 STM2(&(xo[4]), T9, ovs, &(xo[0]));
61 Ta = VADD(T7, T8);
62 STM2(&(xo[0]), Ta, ovs, &(xo[0]));
63 Tb = VFMAI(T6, T3);
64 STM2(&(xo[6]), Tb, ovs, &(xo[2]));
65 STN2(&(xo[4]), T9, Tb, ovs);
66 Tc = VFNMSI(T6, T3);
67 STM2(&(xo[2]), Tc, ovs, &(xo[2]));
68 STN2(&(xo[0]), Ta, Tc, ovs);
69 }
70 }
71 }
72 }
73 VLEAVE();
74 }
75
76 static const kdft_desc desc = { 4, XSIMD_STRING("n2fv_4"), {6, 0, 2, 0}, &GENUS, 0, 2, 0, 0 };
77
78 void XSIMD(codelet_n2fv_4) (planner *p) {
79 X(kdft_register) (p, n2fv_4, &desc);
80 }
81
82 #else /* HAVE_FMA */
83
84 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 4 -name n2fv_4 -with-ostride 2 -include n2f.h -store-multiple 2 */
85
86 /*
87 * This function contains 8 FP additions, 0 FP multiplications,
88 * (or, 8 additions, 0 multiplications, 0 fused multiply/add),
89 * 11 stack variables, 0 constants, and 10 memory accesses
90 */
91 #include "n2f.h"
92
93 static void n2fv_4(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
94 {
95 {
96 INT i;
97 const R *xi;
98 R *xo;
99 xi = ri;
100 xo = ro;
101 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(8, is), MAKE_VOLATILE_STRIDE(8, os)) {
102 V T3, T7, T6, T8;
103 {
104 V T1, T2, T4, T5;
105 T1 = LD(&(xi[0]), ivs, &(xi[0]));
106 T2 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
107 T3 = VSUB(T1, T2);
108 T7 = VADD(T1, T2);
109 T4 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
110 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
111 T6 = VBYI(VSUB(T4, T5));
112 T8 = VADD(T4, T5);
113 }
114 {
115 V T9, Ta, Tb, Tc;
116 T9 = VSUB(T3, T6);
117 STM2(&(xo[2]), T9, ovs, &(xo[2]));
118 Ta = VADD(T7, T8);
119 STM2(&(xo[0]), Ta, ovs, &(xo[0]));
120 STN2(&(xo[0]), Ta, T9, ovs);
121 Tb = VADD(T3, T6);
122 STM2(&(xo[6]), Tb, ovs, &(xo[2]));
123 Tc = VSUB(T7, T8);
124 STM2(&(xo[4]), Tc, ovs, &(xo[0]));
125 STN2(&(xo[4]), Tc, Tb, ovs);
126 }
127 }
128 }
129 VLEAVE();
130 }
131
132 static const kdft_desc desc = { 4, XSIMD_STRING("n2fv_4"), {8, 0, 0, 0}, &GENUS, 0, 2, 0, 0 };
133
134 void XSIMD(codelet_n2fv_4) (planner *p) {
135 X(kdft_register) (p, n2fv_4, &desc);
136 }
137
138 #endif /* HAVE_FMA */