Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n2fv_16.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 07:37:23 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 16 -name n2fv_16 -with-ostride 2 -include n2f.h -store-multiple 2 */
29
30 /*
31 * This function contains 72 FP additions, 34 FP multiplications,
32 * (or, 38 additions, 0 multiplications, 34 fused multiply/add),
33 * 62 stack variables, 3 constants, and 40 memory accesses
34 */
35 #include "n2f.h"
36
37 static void n2fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT i;
44 const R *xi;
45 R *xo;
46 xi = ri;
47 xo = ro;
48 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
49 V T7, Tu, TF, TB, T13, TL, TO, TX, TC, Te, TP, Th, TQ, Tk, TW;
50 V T16;
51 {
52 V TH, TU, Tz, Tf, TK, TV, TA, TM, Ta, TN, Td, Tg, Ti, Tj;
53 {
54 V T1, T2, T4, T5, To, Tp, Tr, Ts;
55 T1 = LD(&(xi[0]), ivs, &(xi[0]));
56 T2 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
57 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
58 T5 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
59 To = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
60 Tp = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
61 Tr = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
62 Ts = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
63 {
64 V T8, TJ, Tq, TI, Tt, T9, Tb, Tc, T3, T6;
65 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
66 TH = VSUB(T1, T2);
67 T3 = VADD(T1, T2);
68 TU = VSUB(T4, T5);
69 T6 = VADD(T4, T5);
70 TJ = VSUB(To, Tp);
71 Tq = VADD(To, Tp);
72 TI = VSUB(Tr, Ts);
73 Tt = VADD(Tr, Ts);
74 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
75 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
76 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
77 T7 = VSUB(T3, T6);
78 Tz = VADD(T3, T6);
79 Tf = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
80 TK = VADD(TI, TJ);
81 TV = VSUB(TJ, TI);
82 TA = VADD(Tt, Tq);
83 Tu = VSUB(Tq, Tt);
84 TM = VSUB(T8, T9);
85 Ta = VADD(T8, T9);
86 TN = VSUB(Tb, Tc);
87 Td = VADD(Tb, Tc);
88 Tg = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
89 Ti = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
90 Tj = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
91 }
92 }
93 TF = VSUB(Tz, TA);
94 TB = VADD(Tz, TA);
95 T13 = VFNMS(LDK(KP707106781), TK, TH);
96 TL = VFMA(LDK(KP707106781), TK, TH);
97 TO = VFNMS(LDK(KP414213562), TN, TM);
98 TX = VFMA(LDK(KP414213562), TM, TN);
99 TC = VADD(Ta, Td);
100 Te = VSUB(Ta, Td);
101 TP = VSUB(Tf, Tg);
102 Th = VADD(Tf, Tg);
103 TQ = VSUB(Tj, Ti);
104 Tk = VADD(Ti, Tj);
105 TW = VFNMS(LDK(KP707106781), TV, TU);
106 T16 = VFMA(LDK(KP707106781), TV, TU);
107 }
108 {
109 V TY, TR, Tl, TD;
110 TY = VFMA(LDK(KP414213562), TP, TQ);
111 TR = VFNMS(LDK(KP414213562), TQ, TP);
112 Tl = VSUB(Th, Tk);
113 TD = VADD(Th, Tk);
114 {
115 V TS, T17, TZ, T14;
116 TS = VADD(TO, TR);
117 T17 = VSUB(TR, TO);
118 TZ = VSUB(TX, TY);
119 T14 = VADD(TX, TY);
120 {
121 V TE, TG, Tm, Tv;
122 TE = VADD(TC, TD);
123 TG = VSUB(TD, TC);
124 Tm = VADD(Te, Tl);
125 Tv = VSUB(Tl, Te);
126 {
127 V T18, T1a, TT, T11;
128 T18 = VFNMS(LDK(KP923879532), T17, T16);
129 T1a = VFMA(LDK(KP923879532), T17, T16);
130 TT = VFNMS(LDK(KP923879532), TS, TL);
131 T11 = VFMA(LDK(KP923879532), TS, TL);
132 {
133 V T15, T19, T10, T12;
134 T15 = VFNMS(LDK(KP923879532), T14, T13);
135 T19 = VFMA(LDK(KP923879532), T14, T13);
136 T10 = VFNMS(LDK(KP923879532), TZ, TW);
137 T12 = VFMA(LDK(KP923879532), TZ, TW);
138 {
139 V T1b, T1c, T1d, T1e;
140 T1b = VFMAI(TG, TF);
141 STM2(&(xo[8]), T1b, ovs, &(xo[0]));
142 T1c = VFNMSI(TG, TF);
143 STM2(&(xo[24]), T1c, ovs, &(xo[0]));
144 T1d = VADD(TB, TE);
145 STM2(&(xo[0]), T1d, ovs, &(xo[0]));
146 T1e = VSUB(TB, TE);
147 STM2(&(xo[16]), T1e, ovs, &(xo[0]));
148 {
149 V Tw, Ty, Tn, Tx;
150 Tw = VFNMS(LDK(KP707106781), Tv, Tu);
151 Ty = VFMA(LDK(KP707106781), Tv, Tu);
152 Tn = VFNMS(LDK(KP707106781), Tm, T7);
153 Tx = VFMA(LDK(KP707106781), Tm, T7);
154 {
155 V T1f, T1g, T1h, T1i;
156 T1f = VFMAI(T1a, T19);
157 STM2(&(xo[6]), T1f, ovs, &(xo[2]));
158 T1g = VFNMSI(T1a, T19);
159 STM2(&(xo[26]), T1g, ovs, &(xo[2]));
160 STN2(&(xo[24]), T1c, T1g, ovs);
161 T1h = VFMAI(T18, T15);
162 STM2(&(xo[22]), T1h, ovs, &(xo[2]));
163 T1i = VFNMSI(T18, T15);
164 STM2(&(xo[10]), T1i, ovs, &(xo[2]));
165 STN2(&(xo[8]), T1b, T1i, ovs);
166 {
167 V T1j, T1k, T1l, T1m;
168 T1j = VFNMSI(T12, T11);
169 STM2(&(xo[2]), T1j, ovs, &(xo[2]));
170 STN2(&(xo[0]), T1d, T1j, ovs);
171 T1k = VFMAI(T12, T11);
172 STM2(&(xo[30]), T1k, ovs, &(xo[2]));
173 T1l = VFMAI(T10, TT);
174 STM2(&(xo[14]), T1l, ovs, &(xo[2]));
175 T1m = VFNMSI(T10, TT);
176 STM2(&(xo[18]), T1m, ovs, &(xo[2]));
177 STN2(&(xo[16]), T1e, T1m, ovs);
178 {
179 V T1n, T1o, T1p, T1q;
180 T1n = VFNMSI(Ty, Tx);
181 STM2(&(xo[28]), T1n, ovs, &(xo[0]));
182 STN2(&(xo[28]), T1n, T1k, ovs);
183 T1o = VFMAI(Ty, Tx);
184 STM2(&(xo[4]), T1o, ovs, &(xo[0]));
185 STN2(&(xo[4]), T1o, T1f, ovs);
186 T1p = VFMAI(Tw, Tn);
187 STM2(&(xo[20]), T1p, ovs, &(xo[0]));
188 STN2(&(xo[20]), T1p, T1h, ovs);
189 T1q = VFNMSI(Tw, Tn);
190 STM2(&(xo[12]), T1q, ovs, &(xo[0]));
191 STN2(&(xo[12]), T1q, T1l, ovs);
192 }
193 }
194 }
195 }
196 }
197 }
198 }
199 }
200 }
201 }
202 }
203 }
204 VLEAVE();
205 }
206
207 static const kdft_desc desc = { 16, XSIMD_STRING("n2fv_16"), {38, 0, 34, 0}, &GENUS, 0, 2, 0, 0 };
208
209 void XSIMD(codelet_n2fv_16) (planner *p) {
210 X(kdft_register) (p, n2fv_16, &desc);
211 }
212
213 #else /* HAVE_FMA */
214
215 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 16 -name n2fv_16 -with-ostride 2 -include n2f.h -store-multiple 2 */
216
217 /*
218 * This function contains 72 FP additions, 12 FP multiplications,
219 * (or, 68 additions, 8 multiplications, 4 fused multiply/add),
220 * 38 stack variables, 3 constants, and 40 memory accesses
221 */
222 #include "n2f.h"
223
224 static void n2fv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
225 {
226 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
227 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
228 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
229 {
230 INT i;
231 const R *xi;
232 R *xo;
233 xi = ri;
234 xo = ro;
235 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
236 V Tp, T13, Tu, TN, Tm, T14, Tv, TY, T7, T17, Ty, TT, Te, T16, Tx;
237 V TQ;
238 {
239 V Tn, To, TM, Ts, Tt, TL;
240 Tn = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
241 To = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
242 TM = VADD(Tn, To);
243 Ts = LD(&(xi[0]), ivs, &(xi[0]));
244 Tt = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
245 TL = VADD(Ts, Tt);
246 Tp = VSUB(Tn, To);
247 T13 = VADD(TL, TM);
248 Tu = VSUB(Ts, Tt);
249 TN = VSUB(TL, TM);
250 }
251 {
252 V Ti, TW, Tl, TX;
253 {
254 V Tg, Th, Tj, Tk;
255 Tg = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
256 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
257 Ti = VSUB(Tg, Th);
258 TW = VADD(Tg, Th);
259 Tj = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
260 Tk = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
261 Tl = VSUB(Tj, Tk);
262 TX = VADD(Tj, Tk);
263 }
264 Tm = VMUL(LDK(KP707106781), VSUB(Ti, Tl));
265 T14 = VADD(TX, TW);
266 Tv = VMUL(LDK(KP707106781), VADD(Tl, Ti));
267 TY = VSUB(TW, TX);
268 }
269 {
270 V T3, TR, T6, TS;
271 {
272 V T1, T2, T4, T5;
273 T1 = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
274 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
275 T3 = VSUB(T1, T2);
276 TR = VADD(T1, T2);
277 T4 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
278 T5 = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
279 T6 = VSUB(T4, T5);
280 TS = VADD(T4, T5);
281 }
282 T7 = VFNMS(LDK(KP923879532), T6, VMUL(LDK(KP382683432), T3));
283 T17 = VADD(TR, TS);
284 Ty = VFMA(LDK(KP923879532), T3, VMUL(LDK(KP382683432), T6));
285 TT = VSUB(TR, TS);
286 }
287 {
288 V Ta, TO, Td, TP;
289 {
290 V T8, T9, Tb, Tc;
291 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
292 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
293 Ta = VSUB(T8, T9);
294 TO = VADD(T8, T9);
295 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
296 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
297 Td = VSUB(Tb, Tc);
298 TP = VADD(Tb, Tc);
299 }
300 Te = VFMA(LDK(KP382683432), Ta, VMUL(LDK(KP923879532), Td));
301 T16 = VADD(TO, TP);
302 Tx = VFNMS(LDK(KP382683432), Td, VMUL(LDK(KP923879532), Ta));
303 TQ = VSUB(TO, TP);
304 }
305 {
306 V T1b, T1c, T1d, T1e;
307 {
308 V T15, T18, T19, T1a;
309 T15 = VADD(T13, T14);
310 T18 = VADD(T16, T17);
311 T1b = VSUB(T15, T18);
312 STM2(&(xo[16]), T1b, ovs, &(xo[0]));
313 T1c = VADD(T15, T18);
314 STM2(&(xo[0]), T1c, ovs, &(xo[0]));
315 T19 = VSUB(T13, T14);
316 T1a = VBYI(VSUB(T17, T16));
317 T1d = VSUB(T19, T1a);
318 STM2(&(xo[24]), T1d, ovs, &(xo[0]));
319 T1e = VADD(T19, T1a);
320 STM2(&(xo[8]), T1e, ovs, &(xo[0]));
321 }
322 {
323 V T1f, T1g, T1h, T1i;
324 {
325 V TV, T11, T10, T12, TU, TZ;
326 TU = VMUL(LDK(KP707106781), VADD(TQ, TT));
327 TV = VADD(TN, TU);
328 T11 = VSUB(TN, TU);
329 TZ = VMUL(LDK(KP707106781), VSUB(TT, TQ));
330 T10 = VBYI(VADD(TY, TZ));
331 T12 = VBYI(VSUB(TZ, TY));
332 T1f = VSUB(TV, T10);
333 STM2(&(xo[28]), T1f, ovs, &(xo[0]));
334 T1g = VADD(T11, T12);
335 STM2(&(xo[12]), T1g, ovs, &(xo[0]));
336 T1h = VADD(TV, T10);
337 STM2(&(xo[4]), T1h, ovs, &(xo[0]));
338 T1i = VSUB(T11, T12);
339 STM2(&(xo[20]), T1i, ovs, &(xo[0]));
340 }
341 {
342 V Tr, TB, TA, TC;
343 {
344 V Tf, Tq, Tw, Tz;
345 Tf = VSUB(T7, Te);
346 Tq = VSUB(Tm, Tp);
347 Tr = VBYI(VSUB(Tf, Tq));
348 TB = VBYI(VADD(Tq, Tf));
349 Tw = VADD(Tu, Tv);
350 Tz = VADD(Tx, Ty);
351 TA = VSUB(Tw, Tz);
352 TC = VADD(Tw, Tz);
353 }
354 {
355 V T1j, T1k, T1l, T1m;
356 T1j = VADD(Tr, TA);
357 STM2(&(xo[14]), T1j, ovs, &(xo[2]));
358 STN2(&(xo[12]), T1g, T1j, ovs);
359 T1k = VSUB(TC, TB);
360 STM2(&(xo[30]), T1k, ovs, &(xo[2]));
361 STN2(&(xo[28]), T1f, T1k, ovs);
362 T1l = VSUB(TA, Tr);
363 STM2(&(xo[18]), T1l, ovs, &(xo[2]));
364 STN2(&(xo[16]), T1b, T1l, ovs);
365 T1m = VADD(TB, TC);
366 STM2(&(xo[2]), T1m, ovs, &(xo[2]));
367 STN2(&(xo[0]), T1c, T1m, ovs);
368 }
369 }
370 {
371 V TF, TJ, TI, TK;
372 {
373 V TD, TE, TG, TH;
374 TD = VSUB(Tu, Tv);
375 TE = VADD(Te, T7);
376 TF = VADD(TD, TE);
377 TJ = VSUB(TD, TE);
378 TG = VADD(Tp, Tm);
379 TH = VSUB(Ty, Tx);
380 TI = VBYI(VADD(TG, TH));
381 TK = VBYI(VSUB(TH, TG));
382 }
383 {
384 V T1n, T1o, T1p, T1q;
385 T1n = VSUB(TF, TI);
386 STM2(&(xo[26]), T1n, ovs, &(xo[2]));
387 STN2(&(xo[24]), T1d, T1n, ovs);
388 T1o = VADD(TJ, TK);
389 STM2(&(xo[10]), T1o, ovs, &(xo[2]));
390 STN2(&(xo[8]), T1e, T1o, ovs);
391 T1p = VADD(TF, TI);
392 STM2(&(xo[6]), T1p, ovs, &(xo[2]));
393 STN2(&(xo[4]), T1h, T1p, ovs);
394 T1q = VSUB(TJ, TK);
395 STM2(&(xo[22]), T1q, ovs, &(xo[2]));
396 STN2(&(xo[20]), T1i, T1q, ovs);
397 }
398 }
399 }
400 }
401 }
402 }
403 VLEAVE();
404 }
405
406 static const kdft_desc desc = { 16, XSIMD_STRING("n2fv_16"), {68, 8, 4, 0}, &GENUS, 0, 2, 0, 0 };
407
408 void XSIMD(codelet_n2fv_16) (planner *p) {
409 X(kdft_register) (p, n2fv_16, &desc);
410 }
411
412 #endif /* HAVE_FMA */