Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/scalar/codelets/n1_8.c @ 95:89f5e221ed7b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW3
author Chris Cannam <cannam@all-day-breakfast.com>
date Wed, 20 Mar 2013 15:35:50 +0000
parents
children
comparison
equal deleted inserted replaced
94:d278df1123f9 95:89f5e221ed7b
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 07:35:42 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_notw.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 8 -name n1_8 -include n.h */
29
30 /*
31 * This function contains 52 FP additions, 8 FP multiplications,
32 * (or, 44 additions, 0 multiplications, 8 fused multiply/add),
33 * 36 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "n.h"
36
37 static void n1_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT i;
42 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
43 E TF, TE, TD, TI;
44 {
45 E Tn, T3, TC, Ti, TB, T6, To, Tl, Td, TN, Tz, TH, Ta, Tq, Tt;
46 E TM;
47 {
48 E T4, T5, Tj, Tk;
49 {
50 E T1, T2, Tg, Th;
51 T1 = ri[0];
52 T2 = ri[WS(is, 4)];
53 Tg = ii[0];
54 Th = ii[WS(is, 4)];
55 T4 = ri[WS(is, 2)];
56 Tn = T1 - T2;
57 T3 = T1 + T2;
58 TC = Tg - Th;
59 Ti = Tg + Th;
60 T5 = ri[WS(is, 6)];
61 }
62 Tj = ii[WS(is, 2)];
63 Tk = ii[WS(is, 6)];
64 {
65 E Tb, Tc, Tw, Tx;
66 Tb = ri[WS(is, 7)];
67 TB = T4 - T5;
68 T6 = T4 + T5;
69 To = Tj - Tk;
70 Tl = Tj + Tk;
71 Tc = ri[WS(is, 3)];
72 Tw = ii[WS(is, 7)];
73 Tx = ii[WS(is, 3)];
74 {
75 E T8, Tv, Ty, T9, Tr, Ts;
76 T8 = ri[WS(is, 1)];
77 Td = Tb + Tc;
78 Tv = Tb - Tc;
79 TN = Tw + Tx;
80 Ty = Tw - Tx;
81 T9 = ri[WS(is, 5)];
82 Tr = ii[WS(is, 1)];
83 Ts = ii[WS(is, 5)];
84 Tz = Tv - Ty;
85 TH = Tv + Ty;
86 Ta = T8 + T9;
87 Tq = T8 - T9;
88 Tt = Tr - Ts;
89 TM = Tr + Ts;
90 }
91 }
92 }
93 {
94 E TL, TG, Tu, Tf, Tm, TO;
95 {
96 E T7, Te, TP, TQ;
97 TL = T3 - T6;
98 T7 = T3 + T6;
99 TG = Tt - Tq;
100 Tu = Tq + Tt;
101 Te = Ta + Td;
102 Tf = Td - Ta;
103 Tm = Ti - Tl;
104 TP = Ti + Tl;
105 TQ = TM + TN;
106 TO = TM - TN;
107 ro[0] = T7 + Te;
108 ro[WS(os, 4)] = T7 - Te;
109 io[0] = TP + TQ;
110 io[WS(os, 4)] = TP - TQ;
111 }
112 {
113 E Tp, TA, TJ, TK;
114 TF = Tn - To;
115 Tp = Tn + To;
116 io[WS(os, 6)] = Tm - Tf;
117 io[WS(os, 2)] = Tf + Tm;
118 ro[WS(os, 2)] = TL + TO;
119 ro[WS(os, 6)] = TL - TO;
120 TA = Tu + Tz;
121 TE = Tz - Tu;
122 TD = TB + TC;
123 TJ = TC - TB;
124 TK = TG + TH;
125 TI = TG - TH;
126 ro[WS(os, 1)] = FMA(KP707106781, TA, Tp);
127 ro[WS(os, 5)] = FNMS(KP707106781, TA, Tp);
128 io[WS(os, 1)] = FMA(KP707106781, TK, TJ);
129 io[WS(os, 5)] = FNMS(KP707106781, TK, TJ);
130 }
131 }
132 }
133 io[WS(os, 3)] = FMA(KP707106781, TE, TD);
134 io[WS(os, 7)] = FNMS(KP707106781, TE, TD);
135 ro[WS(os, 3)] = FMA(KP707106781, TI, TF);
136 ro[WS(os, 7)] = FNMS(KP707106781, TI, TF);
137 }
138 }
139 }
140
141 static const kdft_desc desc = { 8, "n1_8", {44, 0, 8, 0}, &GENUS, 0, 0, 0, 0 };
142
143 void X(codelet_n1_8) (planner *p) {
144 X(kdft_register) (p, n1_8, &desc);
145 }
146
147 #else /* HAVE_FMA */
148
149 /* Generated by: ../../../genfft/gen_notw.native -compact -variables 4 -pipeline-latency 4 -n 8 -name n1_8 -include n.h */
150
151 /*
152 * This function contains 52 FP additions, 4 FP multiplications,
153 * (or, 52 additions, 4 multiplications, 0 fused multiply/add),
154 * 28 stack variables, 1 constants, and 32 memory accesses
155 */
156 #include "n.h"
157
158 static void n1_8(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
159 {
160 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
161 {
162 INT i;
163 for (i = v; i > 0; i = i - 1, ri = ri + ivs, ii = ii + ivs, ro = ro + ovs, io = io + ovs, MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
164 E T3, Tn, Ti, TC, T6, TB, Tl, To, Td, TN, Tz, TH, Ta, TM, Tu;
165 E TG;
166 {
167 E T1, T2, Tj, Tk;
168 T1 = ri[0];
169 T2 = ri[WS(is, 4)];
170 T3 = T1 + T2;
171 Tn = T1 - T2;
172 {
173 E Tg, Th, T4, T5;
174 Tg = ii[0];
175 Th = ii[WS(is, 4)];
176 Ti = Tg + Th;
177 TC = Tg - Th;
178 T4 = ri[WS(is, 2)];
179 T5 = ri[WS(is, 6)];
180 T6 = T4 + T5;
181 TB = T4 - T5;
182 }
183 Tj = ii[WS(is, 2)];
184 Tk = ii[WS(is, 6)];
185 Tl = Tj + Tk;
186 To = Tj - Tk;
187 {
188 E Tb, Tc, Tv, Tw, Tx, Ty;
189 Tb = ri[WS(is, 7)];
190 Tc = ri[WS(is, 3)];
191 Tv = Tb - Tc;
192 Tw = ii[WS(is, 7)];
193 Tx = ii[WS(is, 3)];
194 Ty = Tw - Tx;
195 Td = Tb + Tc;
196 TN = Tw + Tx;
197 Tz = Tv - Ty;
198 TH = Tv + Ty;
199 }
200 {
201 E T8, T9, Tq, Tr, Ts, Tt;
202 T8 = ri[WS(is, 1)];
203 T9 = ri[WS(is, 5)];
204 Tq = T8 - T9;
205 Tr = ii[WS(is, 1)];
206 Ts = ii[WS(is, 5)];
207 Tt = Tr - Ts;
208 Ta = T8 + T9;
209 TM = Tr + Ts;
210 Tu = Tq + Tt;
211 TG = Tt - Tq;
212 }
213 }
214 {
215 E T7, Te, TP, TQ;
216 T7 = T3 + T6;
217 Te = Ta + Td;
218 ro[WS(os, 4)] = T7 - Te;
219 ro[0] = T7 + Te;
220 TP = Ti + Tl;
221 TQ = TM + TN;
222 io[WS(os, 4)] = TP - TQ;
223 io[0] = TP + TQ;
224 }
225 {
226 E Tf, Tm, TL, TO;
227 Tf = Td - Ta;
228 Tm = Ti - Tl;
229 io[WS(os, 2)] = Tf + Tm;
230 io[WS(os, 6)] = Tm - Tf;
231 TL = T3 - T6;
232 TO = TM - TN;
233 ro[WS(os, 6)] = TL - TO;
234 ro[WS(os, 2)] = TL + TO;
235 }
236 {
237 E Tp, TA, TJ, TK;
238 Tp = Tn + To;
239 TA = KP707106781 * (Tu + Tz);
240 ro[WS(os, 5)] = Tp - TA;
241 ro[WS(os, 1)] = Tp + TA;
242 TJ = TC - TB;
243 TK = KP707106781 * (TG + TH);
244 io[WS(os, 5)] = TJ - TK;
245 io[WS(os, 1)] = TJ + TK;
246 }
247 {
248 E TD, TE, TF, TI;
249 TD = TB + TC;
250 TE = KP707106781 * (Tz - Tu);
251 io[WS(os, 7)] = TD - TE;
252 io[WS(os, 3)] = TD + TE;
253 TF = Tn - To;
254 TI = KP707106781 * (TG - TH);
255 ro[WS(os, 7)] = TF - TI;
256 ro[WS(os, 3)] = TF + TI;
257 }
258 }
259 }
260 }
261
262 static const kdft_desc desc = { 8, "n1_8", {52, 4, 0, 0}, &GENUS, 0, 0, 0, 0 };
263
264 void X(codelet_n1_8) (planner *p) {
265 X(kdft_register) (p, n1_8, &desc);
266 }
267
268 #endif /* HAVE_FMA */