Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t3fv_16.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 07:38:49 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3fv_16 -include t3f.h */
29
30 /*
31 * This function contains 98 FP additions, 86 FP multiplications,
32 * (or, 64 additions, 52 multiplications, 34 fused multiply/add),
33 * 70 stack variables, 3 constants, and 32 memory accesses
34 */
35 #include "t3f.h"
36
37 static void t3fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
41 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
42 {
43 INT m;
44 R *x;
45 x = ri;
46 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) {
47 V T13, Tg, TY, T14, T1A, T1q, T1f, T1x, T1r, T1i, Tt, T16, TB, T1j, T1k;
48 V TH;
49 {
50 V T2, T8, Tu, T3;
51 T2 = LDW(&(W[0]));
52 T8 = LDW(&(W[TWVL * 2]));
53 Tu = LDW(&(W[TWVL * 6]));
54 T3 = LDW(&(W[TWVL * 4]));
55 {
56 V Ty, T1o, Tf, T1b, T7, Tr, TR, TX, T1g, Tl, To, Tw, TG, Tz, T1p;
57 V T1e, TC;
58 {
59 V T1, T5, Ta, Td;
60 T1 = LD(&(x[0]), ms, &(x[0]));
61 T5 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
62 Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
63 Td = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
64 {
65 V Tx, TO, TE, Tb, Tm, Tp, TN, Te, T6, TW, TP, TS;
66 {
67 V TM, T9, TL, Tc, TU, T4, TV;
68 TM = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
69 Tx = VZMULJ(T2, T8);
70 T9 = VZMUL(T2, T8);
71 TL = VZMULJ(T2, Tu);
72 TO = VZMULJ(T8, T3);
73 Tc = VZMUL(T8, T3);
74 TU = VZMUL(T2, T3);
75 T4 = VZMULJ(T2, T3);
76 TV = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
77 TE = VZMUL(Tx, T3);
78 Ty = VZMULJ(Tx, T3);
79 Tb = VZMULJ(T9, Ta);
80 Tm = VZMULJ(T9, T3);
81 Tp = VZMUL(T9, T3);
82 TN = VZMULJ(TL, TM);
83 Te = VZMULJ(Tc, Td);
84 T6 = VZMULJ(T4, T5);
85 TW = VZMULJ(TU, TV);
86 }
87 TP = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
88 TS = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
89 {
90 V TQ, TT, Ti, Tk, Tn, Th, Tq, Tj;
91 Th = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
92 Tq = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
93 Tj = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
94 T1o = VSUB(Tb, Te);
95 Tf = VADD(Tb, Te);
96 T1b = VSUB(T1, T6);
97 T7 = VADD(T1, T6);
98 TQ = VZMULJ(TO, TP);
99 TT = VZMULJ(Tx, TS);
100 Ti = VZMULJ(T2, Th);
101 Tr = VZMULJ(Tp, Tq);
102 Tk = VZMULJ(T3, Tj);
103 Tn = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
104 {
105 V T1d, T1c, Tv, TF;
106 Tv = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
107 TF = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
108 T1d = VSUB(TN, TQ);
109 TR = VADD(TN, TQ);
110 T1c = VSUB(TT, TW);
111 TX = VADD(TT, TW);
112 T1g = VSUB(Ti, Tk);
113 Tl = VADD(Ti, Tk);
114 To = VZMULJ(Tm, Tn);
115 Tw = VZMULJ(Tu, Tv);
116 TG = VZMULJ(TE, TF);
117 Tz = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
118 T1p = VSUB(T1d, T1c);
119 T1e = VADD(T1c, T1d);
120 TC = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
121 }
122 }
123 }
124 }
125 {
126 V T1h, Ts, TA, TD;
127 T13 = VADD(T7, Tf);
128 Tg = VSUB(T7, Tf);
129 T1h = VSUB(To, Tr);
130 Ts = VADD(To, Tr);
131 TY = VSUB(TR, TX);
132 T14 = VADD(TX, TR);
133 TA = VZMULJ(Ty, Tz);
134 T1A = VFMA(LDK(KP707106781), T1p, T1o);
135 T1q = VFNMS(LDK(KP707106781), T1p, T1o);
136 T1f = VFMA(LDK(KP707106781), T1e, T1b);
137 T1x = VFNMS(LDK(KP707106781), T1e, T1b);
138 TD = VZMULJ(T8, TC);
139 T1r = VFMA(LDK(KP414213562), T1g, T1h);
140 T1i = VFNMS(LDK(KP414213562), T1h, T1g);
141 Tt = VSUB(Tl, Ts);
142 T16 = VADD(Tl, Ts);
143 TB = VADD(Tw, TA);
144 T1j = VSUB(Tw, TA);
145 T1k = VSUB(TG, TD);
146 TH = VADD(TD, TG);
147 }
148 }
149 }
150 {
151 V T15, T19, T1l, T1s, TI, T17;
152 T15 = VADD(T13, T14);
153 T19 = VSUB(T13, T14);
154 T1l = VFNMS(LDK(KP414213562), T1k, T1j);
155 T1s = VFMA(LDK(KP414213562), T1j, T1k);
156 TI = VSUB(TB, TH);
157 T17 = VADD(TB, TH);
158 {
159 V T1y, T1t, T1B, T1m;
160 T1y = VADD(T1r, T1s);
161 T1t = VSUB(T1r, T1s);
162 T1B = VSUB(T1l, T1i);
163 T1m = VADD(T1i, T1l);
164 {
165 V T18, T1a, TJ, TZ;
166 T18 = VADD(T16, T17);
167 T1a = VSUB(T17, T16);
168 TJ = VADD(Tt, TI);
169 TZ = VSUB(TI, Tt);
170 {
171 V T1u, T1w, T1z, T1D;
172 T1u = VFNMS(LDK(KP923879532), T1t, T1q);
173 T1w = VFMA(LDK(KP923879532), T1t, T1q);
174 T1z = VFNMS(LDK(KP923879532), T1y, T1x);
175 T1D = VFMA(LDK(KP923879532), T1y, T1x);
176 {
177 V T1n, T1v, T1C, T1E;
178 T1n = VFNMS(LDK(KP923879532), T1m, T1f);
179 T1v = VFMA(LDK(KP923879532), T1m, T1f);
180 T1C = VFNMS(LDK(KP923879532), T1B, T1A);
181 T1E = VFMA(LDK(KP923879532), T1B, T1A);
182 ST(&(x[WS(rs, 12)]), VFNMSI(T1a, T19), ms, &(x[0]));
183 ST(&(x[WS(rs, 4)]), VFMAI(T1a, T19), ms, &(x[0]));
184 ST(&(x[0]), VADD(T15, T18), ms, &(x[0]));
185 ST(&(x[WS(rs, 8)]), VSUB(T15, T18), ms, &(x[0]));
186 {
187 V T10, T12, TK, T11;
188 T10 = VFNMS(LDK(KP707106781), TZ, TY);
189 T12 = VFMA(LDK(KP707106781), TZ, TY);
190 TK = VFNMS(LDK(KP707106781), TJ, Tg);
191 T11 = VFMA(LDK(KP707106781), TJ, Tg);
192 ST(&(x[WS(rs, 1)]), VFNMSI(T1w, T1v), ms, &(x[WS(rs, 1)]));
193 ST(&(x[WS(rs, 15)]), VFMAI(T1w, T1v), ms, &(x[WS(rs, 1)]));
194 ST(&(x[WS(rs, 7)]), VFMAI(T1u, T1n), ms, &(x[WS(rs, 1)]));
195 ST(&(x[WS(rs, 9)]), VFNMSI(T1u, T1n), ms, &(x[WS(rs, 1)]));
196 ST(&(x[WS(rs, 3)]), VFMAI(T1E, T1D), ms, &(x[WS(rs, 1)]));
197 ST(&(x[WS(rs, 13)]), VFNMSI(T1E, T1D), ms, &(x[WS(rs, 1)]));
198 ST(&(x[WS(rs, 11)]), VFMAI(T1C, T1z), ms, &(x[WS(rs, 1)]));
199 ST(&(x[WS(rs, 5)]), VFNMSI(T1C, T1z), ms, &(x[WS(rs, 1)]));
200 ST(&(x[WS(rs, 14)]), VFNMSI(T12, T11), ms, &(x[0]));
201 ST(&(x[WS(rs, 2)]), VFMAI(T12, T11), ms, &(x[0]));
202 ST(&(x[WS(rs, 10)]), VFMAI(T10, TK), ms, &(x[0]));
203 ST(&(x[WS(rs, 6)]), VFNMSI(T10, TK), ms, &(x[0]));
204 }
205 }
206 }
207 }
208 }
209 }
210 }
211 }
212 VLEAVE();
213 }
214
215 static const tw_instr twinstr[] = {
216 VTW(0, 1),
217 VTW(0, 3),
218 VTW(0, 9),
219 VTW(0, 15),
220 {TW_NEXT, VL, 0}
221 };
222
223 static const ct_desc desc = { 16, XSIMD_STRING("t3fv_16"), twinstr, &GENUS, {64, 52, 34, 0}, 0, 0, 0 };
224
225 void XSIMD(codelet_t3fv_16) (planner *p) {
226 X(kdft_dit_register) (p, t3fv_16, &desc);
227 }
228 #else /* HAVE_FMA */
229
230 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -no-generate-bytw -n 16 -name t3fv_16 -include t3f.h */
231
232 /*
233 * This function contains 98 FP additions, 64 FP multiplications,
234 * (or, 94 additions, 60 multiplications, 4 fused multiply/add),
235 * 51 stack variables, 3 constants, and 32 memory accesses
236 */
237 #include "t3f.h"
238
239 static void t3fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
240 {
241 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
242 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
243 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
244 {
245 INT m;
246 R *x;
247 x = ri;
248 for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(16, rs)) {
249 V T4, T5, T6, To, T1, Ty, T7, T8, TO, TV, Te, Tp, TB, TH, Ts;
250 T4 = LDW(&(W[0]));
251 T5 = LDW(&(W[TWVL * 2]));
252 T6 = VZMULJ(T4, T5);
253 To = VZMUL(T4, T5);
254 T1 = LDW(&(W[TWVL * 6]));
255 Ty = VZMULJ(T4, T1);
256 T7 = LDW(&(W[TWVL * 4]));
257 T8 = VZMULJ(T6, T7);
258 TO = VZMUL(T5, T7);
259 TV = VZMULJ(T4, T7);
260 Te = VZMUL(T6, T7);
261 Tp = VZMULJ(To, T7);
262 TB = VZMULJ(T5, T7);
263 TH = VZMUL(T4, T7);
264 Ts = VZMUL(To, T7);
265 {
266 V TY, T1f, TR, T1g, T1q, T1r, TL, TZ, T1l, T1m, T1n, Ti, T12, T1i, T1j;
267 V T1k, Tw, T11, TU, TX, TW;
268 TU = LD(&(x[0]), ms, &(x[0]));
269 TW = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
270 TX = VZMULJ(TV, TW);
271 TY = VSUB(TU, TX);
272 T1f = VADD(TU, TX);
273 {
274 V TN, TQ, TM, TP;
275 TM = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
276 TN = VZMULJ(To, TM);
277 TP = LD(&(x[WS(rs, 12)]), ms, &(x[0]));
278 TQ = VZMULJ(TO, TP);
279 TR = VSUB(TN, TQ);
280 T1g = VADD(TN, TQ);
281 }
282 {
283 V TA, TJ, TD, TG, TE, TK;
284 {
285 V Tz, TI, TC, TF;
286 Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0]));
287 TA = VZMULJ(Ty, Tz);
288 TI = LD(&(x[WS(rs, 10)]), ms, &(x[0]));
289 TJ = VZMULJ(TH, TI);
290 TC = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
291 TD = VZMULJ(TB, TC);
292 TF = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
293 TG = VZMULJ(T6, TF);
294 }
295 T1q = VADD(TA, TD);
296 T1r = VADD(TG, TJ);
297 TE = VSUB(TA, TD);
298 TK = VSUB(TG, TJ);
299 TL = VMUL(LDK(KP707106781), VSUB(TE, TK));
300 TZ = VMUL(LDK(KP707106781), VADD(TK, TE));
301 }
302 {
303 V T3, Tg, Ta, Td, Tb, Th;
304 {
305 V T2, Tf, T9, Tc;
306 T2 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)]));
307 T3 = VZMULJ(T1, T2);
308 Tf = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)]));
309 Tg = VZMULJ(Te, Tf);
310 T9 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
311 Ta = VZMULJ(T8, T9);
312 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
313 Td = VZMULJ(T5, Tc);
314 }
315 T1l = VADD(T3, Ta);
316 T1m = VADD(Td, Tg);
317 T1n = VSUB(T1l, T1m);
318 Tb = VSUB(T3, Ta);
319 Th = VSUB(Td, Tg);
320 Ti = VFNMS(LDK(KP923879532), Th, VMUL(LDK(KP382683432), Tb));
321 T12 = VFMA(LDK(KP923879532), Tb, VMUL(LDK(KP382683432), Th));
322 }
323 {
324 V Tk, Tu, Tm, Tr, Tn, Tv;
325 {
326 V Tj, Tt, Tl, Tq;
327 Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
328 Tk = VZMULJ(T4, Tj);
329 Tt = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)]));
330 Tu = VZMULJ(Ts, Tt);
331 Tl = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
332 Tm = VZMULJ(T7, Tl);
333 Tq = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
334 Tr = VZMULJ(Tp, Tq);
335 }
336 T1i = VADD(Tk, Tm);
337 T1j = VADD(Tr, Tu);
338 T1k = VSUB(T1i, T1j);
339 Tn = VSUB(Tk, Tm);
340 Tv = VSUB(Tr, Tu);
341 Tw = VFMA(LDK(KP382683432), Tn, VMUL(LDK(KP923879532), Tv));
342 T11 = VFNMS(LDK(KP382683432), Tv, VMUL(LDK(KP923879532), Tn));
343 }
344 {
345 V T1p, T1v, T1u, T1w;
346 {
347 V T1h, T1o, T1s, T1t;
348 T1h = VSUB(T1f, T1g);
349 T1o = VMUL(LDK(KP707106781), VADD(T1k, T1n));
350 T1p = VADD(T1h, T1o);
351 T1v = VSUB(T1h, T1o);
352 T1s = VSUB(T1q, T1r);
353 T1t = VMUL(LDK(KP707106781), VSUB(T1n, T1k));
354 T1u = VBYI(VADD(T1s, T1t));
355 T1w = VBYI(VSUB(T1t, T1s));
356 }
357 ST(&(x[WS(rs, 14)]), VSUB(T1p, T1u), ms, &(x[0]));
358 ST(&(x[WS(rs, 6)]), VADD(T1v, T1w), ms, &(x[0]));
359 ST(&(x[WS(rs, 2)]), VADD(T1p, T1u), ms, &(x[0]));
360 ST(&(x[WS(rs, 10)]), VSUB(T1v, T1w), ms, &(x[0]));
361 }
362 {
363 V T1z, T1D, T1C, T1E;
364 {
365 V T1x, T1y, T1A, T1B;
366 T1x = VADD(T1f, T1g);
367 T1y = VADD(T1r, T1q);
368 T1z = VADD(T1x, T1y);
369 T1D = VSUB(T1x, T1y);
370 T1A = VADD(T1i, T1j);
371 T1B = VADD(T1l, T1m);
372 T1C = VADD(T1A, T1B);
373 T1E = VBYI(VSUB(T1B, T1A));
374 }
375 ST(&(x[WS(rs, 8)]), VSUB(T1z, T1C), ms, &(x[0]));
376 ST(&(x[WS(rs, 4)]), VADD(T1D, T1E), ms, &(x[0]));
377 ST(&(x[0]), VADD(T1z, T1C), ms, &(x[0]));
378 ST(&(x[WS(rs, 12)]), VSUB(T1D, T1E), ms, &(x[0]));
379 }
380 {
381 V TT, T15, T14, T16;
382 {
383 V Tx, TS, T10, T13;
384 Tx = VSUB(Ti, Tw);
385 TS = VSUB(TL, TR);
386 TT = VBYI(VSUB(Tx, TS));
387 T15 = VBYI(VADD(TS, Tx));
388 T10 = VADD(TY, TZ);
389 T13 = VADD(T11, T12);
390 T14 = VSUB(T10, T13);
391 T16 = VADD(T10, T13);
392 }
393 ST(&(x[WS(rs, 7)]), VADD(TT, T14), ms, &(x[WS(rs, 1)]));
394 ST(&(x[WS(rs, 15)]), VSUB(T16, T15), ms, &(x[WS(rs, 1)]));
395 ST(&(x[WS(rs, 9)]), VSUB(T14, TT), ms, &(x[WS(rs, 1)]));
396 ST(&(x[WS(rs, 1)]), VADD(T15, T16), ms, &(x[WS(rs, 1)]));
397 }
398 {
399 V T19, T1d, T1c, T1e;
400 {
401 V T17, T18, T1a, T1b;
402 T17 = VSUB(TY, TZ);
403 T18 = VADD(Tw, Ti);
404 T19 = VADD(T17, T18);
405 T1d = VSUB(T17, T18);
406 T1a = VADD(TR, TL);
407 T1b = VSUB(T12, T11);
408 T1c = VBYI(VADD(T1a, T1b));
409 T1e = VBYI(VSUB(T1b, T1a));
410 }
411 ST(&(x[WS(rs, 13)]), VSUB(T19, T1c), ms, &(x[WS(rs, 1)]));
412 ST(&(x[WS(rs, 5)]), VADD(T1d, T1e), ms, &(x[WS(rs, 1)]));
413 ST(&(x[WS(rs, 3)]), VADD(T19, T1c), ms, &(x[WS(rs, 1)]));
414 ST(&(x[WS(rs, 11)]), VSUB(T1d, T1e), ms, &(x[WS(rs, 1)]));
415 }
416 }
417 }
418 }
419 VLEAVE();
420 }
421
422 static const tw_instr twinstr[] = {
423 VTW(0, 1),
424 VTW(0, 3),
425 VTW(0, 9),
426 VTW(0, 15),
427 {TW_NEXT, VL, 0}
428 };
429
430 static const ct_desc desc = { 16, XSIMD_STRING("t3fv_16"), twinstr, &GENUS, {94, 60, 4, 0}, 0, 0, 0 };
431
432 void XSIMD(codelet_t3fv_16) (planner *p) {
433 X(kdft_dit_register) (p, t3fv_16, &desc);
434 }
435 #endif /* HAVE_FMA */