Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2bv_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 07:37:30 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 14 -name n2bv_14 -with-ostride 2 -include n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 74 FP additions, 48 FP multiplications,
32 * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
33 * 65 stack variables, 6 constants, and 35 memory accesses
34 */
35 #include "n2b.h"
36
37 static void n2bv_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 = ii;
50 xo = io;
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, Tu, TU, 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 Tu = VSUB(T6, T9);
83 TU = VSUB(TI, TJ);
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 T19, T1a, T13, TG, TY, T18, TB, Tw, TT, Tz, T11, T16, TE, Tr, TV;
94 V TQ;
95 TV = VSUB(TP, TO);
96 TQ = VADD(TO, TP);
97 {
98 V Ts, To, TW, TN;
99 Ts = VSUB(Tk, Tn);
100 To = VADD(Tk, Tn);
101 TW = VSUB(TM, TL);
102 TN = VADD(TL, TM);
103 {
104 V Tt, Th, TR, T12;
105 Tt = VSUB(Td, Tg);
106 Th = VADD(Td, Tg);
107 TR = VFNMS(LDK(KP356895867), TK, TQ);
108 T12 = VFNMS(LDK(KP554958132), TV, TU);
109 {
110 V Tx, TF, TZ, T14;
111 Tx = VFNMS(LDK(KP356895867), Ta, To);
112 TF = VFMA(LDK(KP554958132), Ts, Tu);
113 T19 = VADD(TH, VADD(TK, VADD(TN, TQ)));
114 STM2(&(xo[0]), T19, ovs, &(xo[0]));
115 TZ = VFNMS(LDK(KP356895867), TN, TK);
116 T14 = VFNMS(LDK(KP356895867), TQ, TN);
117 {
118 V TX, T17, TC, Tp;
119 TX = VFMA(LDK(KP554958132), TW, TV);
120 T17 = VFMA(LDK(KP554958132), TU, TW);
121 T1a = VADD(T3, VADD(Ta, VADD(Th, To)));
122 STM2(&(xo[14]), T1a, ovs, &(xo[2]));
123 TC = VFNMS(LDK(KP356895867), Th, Ta);
124 Tp = VFNMS(LDK(KP356895867), To, Th);
125 {
126 V TA, Tv, TS, Ty;
127 TA = VFMA(LDK(KP554958132), Tt, Ts);
128 Tv = VFNMS(LDK(KP554958132), Tu, Tt);
129 TS = VFNMS(LDK(KP692021471), TR, TN);
130 T13 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T12, TW));
131 Ty = VFNMS(LDK(KP692021471), Tx, Th);
132 TG = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TF, Tt));
133 {
134 V T10, T15, TD, Tq;
135 T10 = VFNMS(LDK(KP692021471), TZ, TQ);
136 T15 = VFNMS(LDK(KP692021471), T14, TK);
137 TY = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TX, TU));
138 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
139 TD = VFNMS(LDK(KP692021471), TC, To);
140 Tq = VFNMS(LDK(KP692021471), Tp, Ta);
141 TB = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TA, Tu));
142 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
143 TT = VFNMS(LDK(KP900968867), TS, TH);
144 Tz = VFNMS(LDK(KP900968867), Ty, T3);
145 T11 = VFNMS(LDK(KP900968867), T10, TH);
146 T16 = VFNMS(LDK(KP900968867), T15, TH);
147 TE = VFNMS(LDK(KP900968867), TD, T3);
148 Tr = VFNMS(LDK(KP900968867), Tq, T3);
149 }
150 }
151 }
152 }
153 }
154 }
155 {
156 V T1b, T1c, T1d, T1e;
157 T1b = VFMAI(TY, TT);
158 STM2(&(xo[4]), T1b, ovs, &(xo[0]));
159 T1c = VFNMSI(TY, TT);
160 STM2(&(xo[24]), T1c, ovs, &(xo[0]));
161 T1d = VFMAI(TB, Tz);
162 STM2(&(xo[18]), T1d, ovs, &(xo[2]));
163 T1e = VFNMSI(TB, Tz);
164 STM2(&(xo[10]), T1e, ovs, &(xo[2]));
165 {
166 V T1f, T1g, T1h, T1i;
167 T1f = VFMAI(T13, T11);
168 STM2(&(xo[12]), T1f, ovs, &(xo[0]));
169 STN2(&(xo[12]), T1f, T1a, ovs);
170 T1g = VFNMSI(T13, T11);
171 STM2(&(xo[16]), T1g, ovs, &(xo[0]));
172 STN2(&(xo[16]), T1g, T1d, ovs);
173 T1h = VFMAI(T18, T16);
174 STM2(&(xo[8]), T1h, ovs, &(xo[0]));
175 STN2(&(xo[8]), T1h, T1e, ovs);
176 T1i = VFNMSI(T18, T16);
177 STM2(&(xo[20]), T1i, ovs, &(xo[0]));
178 {
179 V T1j, T1k, T1l, T1m;
180 T1j = VFNMSI(TG, TE);
181 STM2(&(xo[26]), T1j, ovs, &(xo[2]));
182 STN2(&(xo[24]), T1c, T1j, ovs);
183 T1k = VFMAI(TG, TE);
184 STM2(&(xo[2]), T1k, ovs, &(xo[2]));
185 STN2(&(xo[0]), T19, T1k, ovs);
186 T1l = VFNMSI(Tw, Tr);
187 STM2(&(xo[22]), T1l, ovs, &(xo[2]));
188 STN2(&(xo[20]), T1i, T1l, ovs);
189 T1m = VFMAI(Tw, Tr);
190 STM2(&(xo[6]), T1m, ovs, &(xo[2]));
191 STN2(&(xo[4]), T1b, T1m, ovs);
192 }
193 }
194 }
195 }
196 }
197 }
198 VLEAVE();
199 }
200
201 static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {32, 6, 42, 0}, &GENUS, 0, 2, 0, 0 };
202
203 void XSIMD(codelet_n2bv_14) (planner *p) {
204 X(kdft_register) (p, n2bv_14, &desc);
205 }
206
207 #else /* HAVE_FMA */
208
209 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include n2b.h -store-multiple 2 */
210
211 /*
212 * This function contains 74 FP additions, 36 FP multiplications,
213 * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
214 * 41 stack variables, 6 constants, and 35 memory accesses
215 */
216 #include "n2b.h"
217
218 static void n2bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
219 {
220 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
221 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
222 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
223 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
224 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
225 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
226 {
227 INT i;
228 const R *xi;
229 R *xo;
230 xi = ii;
231 xo = io;
232 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)) {
233 V Tp, Ty, Tl, TL, Tq, TE, T7, TJ, Ts, TB, Te, TK, Tr, TH, Tn;
234 V To;
235 Tn = LD(&(xi[0]), ivs, &(xi[0]));
236 To = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
237 Tp = VSUB(Tn, To);
238 Ty = VADD(Tn, To);
239 {
240 V Th, TC, Tk, TD;
241 {
242 V Tf, Tg, Ti, Tj;
243 Tf = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
244 Tg = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
245 Th = VSUB(Tf, Tg);
246 TC = VADD(Tf, Tg);
247 Ti = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
248 Tj = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
249 Tk = VSUB(Ti, Tj);
250 TD = VADD(Ti, Tj);
251 }
252 Tl = VSUB(Th, Tk);
253 TL = VSUB(TD, TC);
254 Tq = VADD(Th, Tk);
255 TE = VADD(TC, TD);
256 }
257 {
258 V T3, Tz, T6, TA;
259 {
260 V T1, T2, T4, T5;
261 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
262 T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
263 T3 = VSUB(T1, T2);
264 Tz = VADD(T1, T2);
265 T4 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
266 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
267 T6 = VSUB(T4, T5);
268 TA = VADD(T4, T5);
269 }
270 T7 = VSUB(T3, T6);
271 TJ = VSUB(Tz, TA);
272 Ts = VADD(T3, T6);
273 TB = VADD(Tz, TA);
274 }
275 {
276 V Ta, TF, Td, TG;
277 {
278 V T8, T9, Tb, Tc;
279 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
280 T9 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
281 Ta = VSUB(T8, T9);
282 TF = VADD(T8, T9);
283 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
284 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
285 Td = VSUB(Tb, Tc);
286 TG = VADD(Tb, Tc);
287 }
288 Te = VSUB(Ta, Td);
289 TK = VSUB(TG, TF);
290 Tr = VADD(Ta, Td);
291 TH = VADD(TF, TG);
292 }
293 {
294 V TR, TS, TU, TV;
295 TR = VADD(Tp, VADD(Ts, VADD(Tq, Tr)));
296 STM2(&(xo[14]), TR, ovs, &(xo[2]));
297 TS = VADD(Ty, VADD(TB, VADD(TE, TH)));
298 STM2(&(xo[0]), TS, ovs, &(xo[0]));
299 {
300 V TT, Tm, Tt, TQ, TP, TW;
301 Tm = VBYI(VFMA(LDK(KP433883739), T7, VFNMS(LDK(KP781831482), Tl, VMUL(LDK(KP974927912), Te))));
302 Tt = VFMA(LDK(KP623489801), Tq, VFNMS(LDK(KP222520933), Tr, VFNMS(LDK(KP900968867), Ts, Tp)));
303 TT = VADD(Tm, Tt);
304 STM2(&(xo[6]), TT, ovs, &(xo[2]));
305 TU = VSUB(Tt, Tm);
306 STM2(&(xo[22]), TU, ovs, &(xo[2]));
307 TQ = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
308 TP = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TB, Ty)));
309 TV = VSUB(TP, TQ);
310 STM2(&(xo[24]), TV, ovs, &(xo[0]));
311 TW = VADD(TP, TQ);
312 STM2(&(xo[4]), TW, ovs, &(xo[0]));
313 STN2(&(xo[4]), TW, TT, ovs);
314 }
315 {
316 V T10, TM, TI, TZ;
317 {
318 V Tu, Tv, TX, TY;
319 Tu = VBYI(VFMA(LDK(KP781831482), T7, VFMA(LDK(KP974927912), Tl, VMUL(LDK(KP433883739), Te))));
320 Tv = VFMA(LDK(KP623489801), Ts, VFNMS(LDK(KP900968867), Tr, VFNMS(LDK(KP222520933), Tq, Tp)));
321 TX = VADD(Tu, Tv);
322 STM2(&(xo[2]), TX, ovs, &(xo[2]));
323 STN2(&(xo[0]), TS, TX, ovs);
324 TY = VSUB(Tv, Tu);
325 STM2(&(xo[26]), TY, ovs, &(xo[2]));
326 STN2(&(xo[24]), TV, TY, ovs);
327 }
328 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
329 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TE, Ty)));
330 TZ = VSUB(TI, TM);
331 STM2(&(xo[12]), TZ, ovs, &(xo[0]));
332 STN2(&(xo[12]), TZ, TR, ovs);
333 T10 = VADD(TI, TM);
334 STM2(&(xo[16]), T10, ovs, &(xo[0]));
335 {
336 V T11, TO, TN, T12;
337 TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
338 TN = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP222520933), TH, VFNMS(LDK(KP900968867), TB, Ty)));
339 T11 = VSUB(TN, TO);
340 STM2(&(xo[8]), T11, ovs, &(xo[0]));
341 T12 = VADD(TN, TO);
342 STM2(&(xo[20]), T12, ovs, &(xo[0]));
343 STN2(&(xo[20]), T12, TU, ovs);
344 {
345 V Tx, Tw, T13, T14;
346 Tx = VBYI(VFNMS(LDK(KP781831482), Te, VFNMS(LDK(KP433883739), Tl, VMUL(LDK(KP974927912), T7))));
347 Tw = VFMA(LDK(KP623489801), Tr, VFNMS(LDK(KP900968867), Tq, VFNMS(LDK(KP222520933), Ts, Tp)));
348 T13 = VSUB(Tw, Tx);
349 STM2(&(xo[10]), T13, ovs, &(xo[2]));
350 STN2(&(xo[8]), T11, T13, ovs);
351 T14 = VADD(Tx, Tw);
352 STM2(&(xo[18]), T14, ovs, &(xo[2]));
353 STN2(&(xo[16]), T10, T14, ovs);
354 }
355 }
356 }
357 }
358 }
359 }
360 VLEAVE();
361 }
362
363 static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {50, 12, 24, 0}, &GENUS, 0, 2, 0, 0 };
364
365 void XSIMD(codelet_n2bv_14) (planner *p) {
366 X(kdft_register) (p, n2bv_14, &desc);
367 }
368
369 #endif /* HAVE_FMA */