Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1fv_14.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 07:36:52 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 14 -name n1fv_14 -include n1f.h */
29
30 /*
31 * This function contains 74 FP additions, 48 FP multiplications,
32 * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
33 * 63 stack variables, 6 constants, and 28 memory accesses
34 */
35 #include "n1f.h"
36
37 static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
40 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
41 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
42 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
43 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
45 {
46 INT i;
47 const R *xi;
48 R *xo;
49 xi = ri;
50 xo = ro;
51 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
52 V TH, T3, TP, Tn, Ta, Ts, TW, TK, TO, Tk, TM, Tg, TL, Td, T1;
53 V T2;
54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
55 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
56 {
57 V Ti, TI, T6, TJ, T9, Tj, Te, Tf, Tb, Tc;
58 {
59 V T4, T5, T7, T8, Tl, Tm;
60 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
61 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
62 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
63 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
64 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
65 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
66 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
67 TH = VADD(T1, T2);
68 T3 = VSUB(T1, T2);
69 TI = VADD(T4, T5);
70 T6 = VSUB(T4, T5);
71 TJ = VADD(T7, T8);
72 T9 = VSUB(T7, T8);
73 TP = VADD(Tl, Tm);
74 Tn = VSUB(Tl, Tm);
75 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
76 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
77 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
78 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
79 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
80 }
81 Ta = VADD(T6, T9);
82 Ts = VSUB(T9, T6);
83 TW = VSUB(TJ, TI);
84 TK = VADD(TI, TJ);
85 TO = VADD(Ti, Tj);
86 Tk = VSUB(Ti, Tj);
87 TM = VADD(Te, Tf);
88 Tg = VSUB(Te, Tf);
89 TL = VADD(Tb, Tc);
90 Td = VSUB(Tb, Tc);
91 }
92 {
93 V T18, TB, T13, TY, TG, Tw, T11, Tr, T16, TT, Tz, TE, TU, TQ;
94 TU = VSUB(TO, TP);
95 TQ = VADD(TO, TP);
96 {
97 V Tt, To, TV, TN;
98 Tt = VSUB(Tn, Tk);
99 To = VADD(Tk, Tn);
100 TV = VSUB(TL, TM);
101 TN = VADD(TL, TM);
102 {
103 V Tu, Th, TZ, T17;
104 Tu = VSUB(Tg, Td);
105 Th = VADD(Td, Tg);
106 TZ = VFNMS(LDK(KP356895867), TK, TQ);
107 T17 = VFNMS(LDK(KP554958132), TU, TW);
108 {
109 V Tp, TA, T14, TR;
110 Tp = VFNMS(LDK(KP356895867), Ta, To);
111 TA = VFMA(LDK(KP554958132), Tt, Ts);
112 ST(&(xo[0]), VADD(TH, VADD(TK, VADD(TN, TQ))), ovs, &(xo[0]));
113 T14 = VFNMS(LDK(KP356895867), TN, TK);
114 TR = VFNMS(LDK(KP356895867), TQ, TN);
115 {
116 V T12, TX, Tx, TC;
117 T12 = VFMA(LDK(KP554958132), TV, TU);
118 TX = VFMA(LDK(KP554958132), TW, TV);
119 ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)]));
120 Tx = VFNMS(LDK(KP356895867), Th, Ta);
121 TC = VFNMS(LDK(KP356895867), To, Th);
122 {
123 V TF, Tv, T10, Tq;
124 TF = VFNMS(LDK(KP554958132), Ts, Tu);
125 Tv = VFMA(LDK(KP554958132), Tu, Tt);
126 T10 = VFNMS(LDK(KP692021471), TZ, TN);
127 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
128 Tq = VFNMS(LDK(KP692021471), Tp, Th);
129 TB = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TA, Tu));
130 {
131 V T15, TS, Ty, TD;
132 T15 = VFNMS(LDK(KP692021471), T14, TQ);
133 TS = VFNMS(LDK(KP692021471), TR, TK);
134 T13 = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), T12, TW));
135 TY = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TX, TU));
136 Ty = VFNMS(LDK(KP692021471), Tx, To);
137 TD = VFNMS(LDK(KP692021471), TC, Ta);
138 TG = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TF, Tt));
139 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
140 T11 = VFNMS(LDK(KP900968867), T10, TH);
141 Tr = VFNMS(LDK(KP900968867), Tq, T3);
142 T16 = VFNMS(LDK(KP900968867), T15, TH);
143 TT = VFNMS(LDK(KP900968867), TS, TH);
144 Tz = VFNMS(LDK(KP900968867), Ty, T3);
145 TE = VFNMS(LDK(KP900968867), TD, T3);
146 }
147 }
148 }
149 }
150 }
151 }
152 ST(&(xo[WS(os, 12)]), VFNMSI(T13, T11), ovs, &(xo[0]));
153 ST(&(xo[WS(os, 2)]), VFMAI(T13, T11), ovs, &(xo[0]));
154 ST(&(xo[WS(os, 9)]), VFMAI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
155 ST(&(xo[WS(os, 5)]), VFNMSI(Tw, Tr), ovs, &(xo[WS(os, 1)]));
156 ST(&(xo[WS(os, 8)]), VFNMSI(T18, T16), ovs, &(xo[0]));
157 ST(&(xo[WS(os, 6)]), VFMAI(T18, T16), ovs, &(xo[0]));
158 ST(&(xo[WS(os, 10)]), VFNMSI(TY, TT), ovs, &(xo[0]));
159 ST(&(xo[WS(os, 4)]), VFMAI(TY, TT), ovs, &(xo[0]));
160 ST(&(xo[WS(os, 1)]), VFMAI(TB, Tz), ovs, &(xo[WS(os, 1)]));
161 ST(&(xo[WS(os, 13)]), VFNMSI(TB, Tz), ovs, &(xo[WS(os, 1)]));
162 ST(&(xo[WS(os, 3)]), VFMAI(TG, TE), ovs, &(xo[WS(os, 1)]));
163 ST(&(xo[WS(os, 11)]), VFNMSI(TG, TE), ovs, &(xo[WS(os, 1)]));
164 }
165 }
166 }
167 VLEAVE();
168 }
169
170 static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), {32, 6, 42, 0}, &GENUS, 0, 0, 0, 0 };
171
172 void XSIMD(codelet_n1fv_14) (planner *p) {
173 X(kdft_register) (p, n1fv_14, &desc);
174 }
175
176 #else /* HAVE_FMA */
177
178 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 14 -name n1fv_14 -include n1f.h */
179
180 /*
181 * This function contains 74 FP additions, 36 FP multiplications,
182 * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
183 * 33 stack variables, 6 constants, and 28 memory accesses
184 */
185 #include "n1f.h"
186
187 static void n1fv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
188 {
189 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
190 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
191 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
192 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
193 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
194 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
195 {
196 INT i;
197 const R *xi;
198 R *xo;
199 xi = ri;
200 xo = ro;
201 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
202 V T3, Ty, To, TK, Tr, TE, Ta, TJ, Tq, TB, Th, TL, Ts, TH, T1;
203 V T2;
204 T1 = LD(&(xi[0]), ivs, &(xi[0]));
205 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
206 T3 = VSUB(T1, T2);
207 Ty = VADD(T1, T2);
208 {
209 V Tk, TC, Tn, TD;
210 {
211 V Ti, Tj, Tl, Tm;
212 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
213 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
214 Tk = VSUB(Ti, Tj);
215 TC = VADD(Ti, Tj);
216 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
217 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
218 Tn = VSUB(Tl, Tm);
219 TD = VADD(Tl, Tm);
220 }
221 To = VADD(Tk, Tn);
222 TK = VSUB(TC, TD);
223 Tr = VSUB(Tn, Tk);
224 TE = VADD(TC, TD);
225 }
226 {
227 V T6, Tz, T9, TA;
228 {
229 V T4, T5, T7, T8;
230 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
231 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
232 T6 = VSUB(T4, T5);
233 Tz = VADD(T4, T5);
234 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
235 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
236 T9 = VSUB(T7, T8);
237 TA = VADD(T7, T8);
238 }
239 Ta = VADD(T6, T9);
240 TJ = VSUB(TA, Tz);
241 Tq = VSUB(T9, T6);
242 TB = VADD(Tz, TA);
243 }
244 {
245 V Td, TF, Tg, TG;
246 {
247 V Tb, Tc, Te, Tf;
248 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
249 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
250 Td = VSUB(Tb, Tc);
251 TF = VADD(Tb, Tc);
252 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
253 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
254 Tg = VSUB(Te, Tf);
255 TG = VADD(Te, Tf);
256 }
257 Th = VADD(Td, Tg);
258 TL = VSUB(TF, TG);
259 Ts = VSUB(Tg, Td);
260 TH = VADD(TF, TG);
261 }
262 ST(&(xo[WS(os, 7)]), VADD(T3, VADD(Ta, VADD(Th, To))), ovs, &(xo[WS(os, 1)]));
263 ST(&(xo[0]), VADD(Ty, VADD(TB, VADD(TH, TE))), ovs, &(xo[0]));
264 {
265 V Tt, Tp, TP, TQ;
266 Tt = VBYI(VFNMS(LDK(KP781831482), Tr, VFNMS(LDK(KP433883739), Ts, VMUL(LDK(KP974927912), Tq))));
267 Tp = VFMA(LDK(KP623489801), To, VFNMS(LDK(KP900968867), Th, VFNMS(LDK(KP222520933), Ta, T3)));
268 ST(&(xo[WS(os, 5)]), VSUB(Tp, Tt), ovs, &(xo[WS(os, 1)]));
269 ST(&(xo[WS(os, 9)]), VADD(Tp, Tt), ovs, &(xo[WS(os, 1)]));
270 TP = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
271 TQ = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TB, Ty)));
272 ST(&(xo[WS(os, 2)]), VADD(TP, TQ), ovs, &(xo[0]));
273 ST(&(xo[WS(os, 12)]), VSUB(TQ, TP), ovs, &(xo[0]));
274 }
275 {
276 V Tv, Tu, TM, TI;
277 Tv = VBYI(VFMA(LDK(KP781831482), Tq, VFMA(LDK(KP974927912), Ts, VMUL(LDK(KP433883739), Tr))));
278 Tu = VFMA(LDK(KP623489801), Ta, VFNMS(LDK(KP900968867), To, VFNMS(LDK(KP222520933), Th, T3)));
279 ST(&(xo[WS(os, 13)]), VSUB(Tu, Tv), ovs, &(xo[WS(os, 1)]));
280 ST(&(xo[WS(os, 1)]), VADD(Tu, Tv), ovs, &(xo[WS(os, 1)]));
281 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
282 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TH, Ty)));
283 ST(&(xo[WS(os, 6)]), VSUB(TI, TM), ovs, &(xo[0]));
284 ST(&(xo[WS(os, 8)]), VADD(TM, TI), ovs, &(xo[0]));
285 }
286 {
287 V TO, TN, Tx, Tw;
288 TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
289 TN = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP222520933), TE, VFNMS(LDK(KP900968867), TB, Ty)));
290 ST(&(xo[WS(os, 4)]), VSUB(TN, TO), ovs, &(xo[0]));
291 ST(&(xo[WS(os, 10)]), VADD(TO, TN), ovs, &(xo[0]));
292 Tx = VBYI(VFMA(LDK(KP433883739), Tq, VFNMS(LDK(KP781831482), Ts, VMUL(LDK(KP974927912), Tr))));
293 Tw = VFMA(LDK(KP623489801), Th, VFNMS(LDK(KP222520933), To, VFNMS(LDK(KP900968867), Ta, T3)));
294 ST(&(xo[WS(os, 11)]), VSUB(Tw, Tx), ovs, &(xo[WS(os, 1)]));
295 ST(&(xo[WS(os, 3)]), VADD(Tw, Tx), ovs, &(xo[WS(os, 1)]));
296 }
297 }
298 }
299 VLEAVE();
300 }
301
302 static const kdft_desc desc = { 14, XSIMD_STRING("n1fv_14"), {50, 12, 24, 0}, &GENUS, 0, 0, 0, 0 };
303
304 void XSIMD(codelet_n1fv_14) (planner *p) {
305 X(kdft_register) (p, n1fv_14, &desc);
306 }
307
308 #endif /* HAVE_FMA */