Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2bv_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 07:37:29 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 n2bv_8 -with-ostride 2 -include n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 26 FP additions, 10 FP multiplications,
32 * (or, 16 additions, 0 multiplications, 10 fused multiply/add),
33 * 38 stack variables, 1 constants, and 20 memory accesses
34 */
35 #include "n2b.h"
36
37 static void n2bv_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 Tr, Ts, Tt, Tu, Tg, Ti, Tb, Th;
76 Tr = VFMAI(To, Tl);
77 STM2(&(xo[4]), Tr, ovs, &(xo[0]));
78 Ts = VFNMSI(To, Tl);
79 STM2(&(xo[12]), Ts, ovs, &(xo[0]));
80 Tt = VADD(Tp, Tq);
81 STM2(&(xo[0]), Tt, ovs, &(xo[0]));
82 Tu = VSUB(Tp, Tq);
83 STM2(&(xo[8]), Tu, ovs, &(xo[0]));
84 Tg = VFNMS(LDK(KP707106781), Tf, Te);
85 Ti = VFMA(LDK(KP707106781), Tf, Te);
86 Tb = VFNMS(LDK(KP707106781), Ta, T3);
87 Th = VFMA(LDK(KP707106781), Ta, T3);
88 {
89 V Tv, Tw, Tx, Ty;
90 Tv = VFNMSI(Ti, Th);
91 STM2(&(xo[14]), Tv, ovs, &(xo[2]));
92 STN2(&(xo[12]), Ts, Tv, ovs);
93 Tw = VFMAI(Ti, Th);
94 STM2(&(xo[2]), Tw, ovs, &(xo[2]));
95 STN2(&(xo[0]), Tt, Tw, ovs);
96 Tx = VFMAI(Tg, Tb);
97 STM2(&(xo[10]), Tx, ovs, &(xo[2]));
98 STN2(&(xo[8]), Tu, Tx, ovs);
99 Ty = VFNMSI(Tg, Tb);
100 STM2(&(xo[6]), Ty, ovs, &(xo[2]));
101 STN2(&(xo[4]), Tr, Ty, ovs);
102 }
103 }
104 }
105 }
106 }
107 }
108 VLEAVE();
109 }
110
111 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {16, 0, 10, 0}, &GENUS, 0, 2, 0, 0 };
112
113 void XSIMD(codelet_n2bv_8) (planner *p) {
114 X(kdft_register) (p, n2bv_8, &desc);
115 }
116
117 #else /* HAVE_FMA */
118
119 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 8 -name n2bv_8 -with-ostride 2 -include n2b.h -store-multiple 2 */
120
121 /*
122 * This function contains 26 FP additions, 2 FP multiplications,
123 * (or, 26 additions, 2 multiplications, 0 fused multiply/add),
124 * 24 stack variables, 1 constants, and 20 memory accesses
125 */
126 #include "n2b.h"
127
128 static void n2bv_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
129 {
130 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
131 {
132 INT i;
133 const R *xi;
134 R *xo;
135 xi = ii;
136 xo = io;
137 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)) {
138 V Ta, Tk, Te, Tj, T7, Tn, Tf, Tm, Tr, Tu;
139 {
140 V T8, T9, Tc, Td;
141 T8 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
142 T9 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
143 Ta = VSUB(T8, T9);
144 Tk = VADD(T8, T9);
145 Tc = LD(&(xi[0]), ivs, &(xi[0]));
146 Td = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
147 Te = VSUB(Tc, Td);
148 Tj = VADD(Tc, Td);
149 {
150 V T1, T2, T3, T4, T5, T6;
151 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
152 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
153 T3 = VSUB(T1, T2);
154 T4 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
155 T5 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
156 T6 = VSUB(T4, T5);
157 T7 = VMUL(LDK(KP707106781), VSUB(T3, T6));
158 Tn = VADD(T4, T5);
159 Tf = VMUL(LDK(KP707106781), VADD(T3, T6));
160 Tm = VADD(T1, T2);
161 }
162 }
163 {
164 V Ts, Tb, Tg, Tp, Tq, Tt;
165 Tb = VBYI(VSUB(T7, Ta));
166 Tg = VSUB(Te, Tf);
167 Tr = VADD(Tb, Tg);
168 STM2(&(xo[6]), Tr, ovs, &(xo[2]));
169 Ts = VSUB(Tg, Tb);
170 STM2(&(xo[10]), Ts, ovs, &(xo[2]));
171 Tp = VADD(Tj, Tk);
172 Tq = VADD(Tm, Tn);
173 Tt = VSUB(Tp, Tq);
174 STM2(&(xo[8]), Tt, ovs, &(xo[0]));
175 STN2(&(xo[8]), Tt, Ts, ovs);
176 Tu = VADD(Tp, Tq);
177 STM2(&(xo[0]), Tu, ovs, &(xo[0]));
178 }
179 {
180 V Tw, Th, Ti, Tv;
181 Th = VBYI(VADD(Ta, T7));
182 Ti = VADD(Te, Tf);
183 Tv = VADD(Th, Ti);
184 STM2(&(xo[2]), Tv, ovs, &(xo[2]));
185 STN2(&(xo[0]), Tu, Tv, ovs);
186 Tw = VSUB(Ti, Th);
187 STM2(&(xo[14]), Tw, ovs, &(xo[2]));
188 {
189 V Tl, To, Tx, Ty;
190 Tl = VSUB(Tj, Tk);
191 To = VBYI(VSUB(Tm, Tn));
192 Tx = VSUB(Tl, To);
193 STM2(&(xo[12]), Tx, ovs, &(xo[0]));
194 STN2(&(xo[12]), Tx, Tw, ovs);
195 Ty = VADD(Tl, To);
196 STM2(&(xo[4]), Ty, ovs, &(xo[0]));
197 STN2(&(xo[4]), Ty, Tr, ovs);
198 }
199 }
200 }
201 }
202 VLEAVE();
203 }
204
205 static const kdft_desc desc = { 8, XSIMD_STRING("n2bv_8"), {26, 2, 0, 0}, &GENUS, 0, 2, 0, 0 };
206
207 void XSIMD(codelet_n2bv_8) (planner *p) {
208 X(kdft_register) (p, n2bv_8, &desc);
209 }
210
211 #endif /* HAVE_FMA */