Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t1sv_8.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 07:39:24 EST 2012 */
23
24 #include "codelet-dft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_twiddle.native -fma -reorder-insns -schedule-for-pipeline -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1sv_8 -include ts.h */
29
30 /*
31 * This function contains 66 FP additions, 36 FP multiplications,
32 * (or, 44 additions, 14 multiplications, 22 fused multiply/add),
33 * 59 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "ts.h"
36
37 static void t1sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 for (m = mb, W = W + (mb * 14); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 14), MAKE_VOLATILE_STRIDE(16, rs)) {
43 V T1, T1m, T1l, T7, TS, Tk, TQ, Te, To, Tr, Tu, T14, TF, Tx, T16;
44 V TL, Tt, TW, Tp, Tq, Tw;
45 {
46 V T3, T6, T2, T5;
47 T1 = LD(&(ri[0]), ms, &(ri[0]));
48 T1m = LD(&(ii[0]), ms, &(ii[0]));
49 T3 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
50 T6 = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
51 T2 = LDW(&(W[TWVL * 6]));
52 T5 = LDW(&(W[TWVL * 7]));
53 {
54 V Tg, Tj, Ti, Ta, Td, T1k, T4, T9, Tc, TR, Th, Tf;
55 Tg = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
56 Tj = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
57 Tf = LDW(&(W[TWVL * 10]));
58 Ti = LDW(&(W[TWVL * 11]));
59 Ta = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
60 Td = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
61 T1k = VMUL(T2, T6);
62 T4 = VMUL(T2, T3);
63 T9 = LDW(&(W[TWVL * 2]));
64 Tc = LDW(&(W[TWVL * 3]));
65 TR = VMUL(Tf, Tj);
66 Th = VMUL(Tf, Tg);
67 {
68 V TB, TE, TH, TK, TG, TD, TJ, T13, TC, TA, TP, Tb, T15, TI, Tn;
69 TB = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
70 TE = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
71 T1l = VFNMS(T5, T3, T1k);
72 T7 = VFMA(T5, T6, T4);
73 TP = VMUL(T9, Td);
74 Tb = VMUL(T9, Ta);
75 TS = VFNMS(Ti, Tg, TR);
76 Tk = VFMA(Ti, Tj, Th);
77 TA = LDW(&(W[TWVL * 12]));
78 TH = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
79 TK = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
80 TG = LDW(&(W[TWVL * 4]));
81 TQ = VFNMS(Tc, Ta, TP);
82 Te = VFMA(Tc, Td, Tb);
83 TD = LDW(&(W[TWVL * 13]));
84 TJ = LDW(&(W[TWVL * 5]));
85 T13 = VMUL(TA, TE);
86 TC = VMUL(TA, TB);
87 To = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
88 T15 = VMUL(TG, TK);
89 TI = VMUL(TG, TH);
90 Tr = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
91 Tn = LDW(&(W[0]));
92 Tu = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
93 T14 = VFNMS(TD, TB, T13);
94 TF = VFMA(TD, TE, TC);
95 Tx = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
96 T16 = VFNMS(TJ, TH, T15);
97 TL = VFMA(TJ, TK, TI);
98 Tt = LDW(&(W[TWVL * 8]));
99 TW = VMUL(Tn, Tr);
100 Tp = VMUL(Tn, To);
101 Tq = LDW(&(W[TWVL * 1]));
102 Tw = LDW(&(W[TWVL * 9]));
103 }
104 }
105 }
106 {
107 V T8, T1g, TM, T1j, TX, Ts, T1n, T1r, T1s, Tl, T1c, T18, TZ, Ty, T1a;
108 V TU;
109 {
110 V TO, T17, T12, TY, Tv, TT;
111 T8 = VADD(T1, T7);
112 TO = VSUB(T1, T7);
113 T17 = VSUB(T14, T16);
114 T1g = VADD(T14, T16);
115 TM = VADD(TF, TL);
116 T12 = VSUB(TF, TL);
117 TY = VMUL(Tt, Tx);
118 Tv = VMUL(Tt, Tu);
119 TT = VSUB(TQ, TS);
120 T1j = VADD(TQ, TS);
121 TX = VFNMS(Tq, To, TW);
122 Ts = VFMA(Tq, Tr, Tp);
123 T1n = VADD(T1l, T1m);
124 T1r = VSUB(T1m, T1l);
125 T1s = VSUB(Te, Tk);
126 Tl = VADD(Te, Tk);
127 T1c = VADD(T12, T17);
128 T18 = VSUB(T12, T17);
129 TZ = VFNMS(Tw, Tu, TY);
130 Ty = VFMA(Tw, Tx, Tv);
131 T1a = VSUB(TO, TT);
132 TU = VADD(TO, TT);
133 }
134 {
135 V T1v, T1t, Tm, T1e, T1o, T1q, TN, T1p, T1d, T1u, T19, T1w, T1i, T1h;
136 {
137 V T10, T1f, Tz, TV, T11, T1b;
138 T1v = VADD(T1s, T1r);
139 T1t = VSUB(T1r, T1s);
140 T10 = VSUB(TX, TZ);
141 T1f = VADD(TX, TZ);
142 Tz = VADD(Ts, Ty);
143 TV = VSUB(Ts, Ty);
144 T11 = VADD(TV, T10);
145 T1b = VSUB(T10, TV);
146 Tm = VADD(T8, Tl);
147 T1e = VSUB(T8, Tl);
148 T1o = VADD(T1j, T1n);
149 T1q = VSUB(T1n, T1j);
150 TN = VADD(Tz, TM);
151 T1p = VSUB(TM, Tz);
152 T1d = VSUB(T1b, T1c);
153 T1u = VADD(T1b, T1c);
154 T19 = VADD(T11, T18);
155 T1w = VSUB(T18, T11);
156 T1i = VADD(T1f, T1g);
157 T1h = VSUB(T1f, T1g);
158 }
159 ST(&(ii[WS(rs, 6)]), VSUB(T1q, T1p), ms, &(ii[0]));
160 ST(&(ri[0]), VADD(Tm, TN), ms, &(ri[0]));
161 ST(&(ri[WS(rs, 4)]), VSUB(Tm, TN), ms, &(ri[0]));
162 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP707106781), T1u, T1t), ms, &(ii[WS(rs, 1)]));
163 ST(&(ii[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1u, T1t), ms, &(ii[WS(rs, 1)]));
164 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP707106781), T1d, T1a), ms, &(ri[WS(rs, 1)]));
165 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1d, T1a), ms, &(ri[WS(rs, 1)]));
166 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP707106781), T1w, T1v), ms, &(ii[WS(rs, 1)]));
167 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1w, T1v), ms, &(ii[WS(rs, 1)]));
168 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP707106781), T19, TU), ms, &(ri[WS(rs, 1)]));
169 ST(&(ri[WS(rs, 5)]), VFNMS(LDK(KP707106781), T19, TU), ms, &(ri[WS(rs, 1)]));
170 ST(&(ri[WS(rs, 6)]), VSUB(T1e, T1h), ms, &(ri[0]));
171 ST(&(ii[0]), VADD(T1i, T1o), ms, &(ii[0]));
172 ST(&(ii[WS(rs, 4)]), VSUB(T1o, T1i), ms, &(ii[0]));
173 ST(&(ri[WS(rs, 2)]), VADD(T1e, T1h), ms, &(ri[0]));
174 ST(&(ii[WS(rs, 2)]), VADD(T1p, T1q), ms, &(ii[0]));
175 }
176 }
177 }
178 }
179 VLEAVE();
180 }
181
182 static const tw_instr twinstr[] = {
183 VTW(0, 1),
184 VTW(0, 2),
185 VTW(0, 3),
186 VTW(0, 4),
187 VTW(0, 5),
188 VTW(0, 6),
189 VTW(0, 7),
190 {TW_NEXT, (2 * VL), 0}
191 };
192
193 static const ct_desc desc = { 8, XSIMD_STRING("t1sv_8"), twinstr, &GENUS, {44, 14, 22, 0}, 0, 0, 0 };
194
195 void XSIMD(codelet_t1sv_8) (planner *p) {
196 X(kdft_dit_register) (p, t1sv_8, &desc);
197 }
198 #else /* HAVE_FMA */
199
200 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -n 8 -name t1sv_8 -include ts.h */
201
202 /*
203 * This function contains 66 FP additions, 32 FP multiplications,
204 * (or, 52 additions, 18 multiplications, 14 fused multiply/add),
205 * 28 stack variables, 1 constants, and 32 memory accesses
206 */
207 #include "ts.h"
208
209 static void t1sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
210 {
211 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
212 {
213 INT m;
214 for (m = mb, W = W + (mb * 14); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 14), MAKE_VOLATILE_STRIDE(16, rs)) {
215 V T7, T1e, TH, T19, TF, T13, TR, TU, Ti, T1f, TK, T16, Tu, T12, TM;
216 V TP;
217 {
218 V T1, T18, T6, T17;
219 T1 = LD(&(ri[0]), ms, &(ri[0]));
220 T18 = LD(&(ii[0]), ms, &(ii[0]));
221 {
222 V T3, T5, T2, T4;
223 T3 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
224 T5 = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
225 T2 = LDW(&(W[TWVL * 6]));
226 T4 = LDW(&(W[TWVL * 7]));
227 T6 = VFMA(T2, T3, VMUL(T4, T5));
228 T17 = VFNMS(T4, T3, VMUL(T2, T5));
229 }
230 T7 = VADD(T1, T6);
231 T1e = VSUB(T18, T17);
232 TH = VSUB(T1, T6);
233 T19 = VADD(T17, T18);
234 }
235 {
236 V Tz, TS, TE, TT;
237 {
238 V Tw, Ty, Tv, Tx;
239 Tw = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
240 Ty = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
241 Tv = LDW(&(W[TWVL * 12]));
242 Tx = LDW(&(W[TWVL * 13]));
243 Tz = VFMA(Tv, Tw, VMUL(Tx, Ty));
244 TS = VFNMS(Tx, Tw, VMUL(Tv, Ty));
245 }
246 {
247 V TB, TD, TA, TC;
248 TB = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
249 TD = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
250 TA = LDW(&(W[TWVL * 4]));
251 TC = LDW(&(W[TWVL * 5]));
252 TE = VFMA(TA, TB, VMUL(TC, TD));
253 TT = VFNMS(TC, TB, VMUL(TA, TD));
254 }
255 TF = VADD(Tz, TE);
256 T13 = VADD(TS, TT);
257 TR = VSUB(Tz, TE);
258 TU = VSUB(TS, TT);
259 }
260 {
261 V Tc, TI, Th, TJ;
262 {
263 V T9, Tb, T8, Ta;
264 T9 = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
265 Tb = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
266 T8 = LDW(&(W[TWVL * 2]));
267 Ta = LDW(&(W[TWVL * 3]));
268 Tc = VFMA(T8, T9, VMUL(Ta, Tb));
269 TI = VFNMS(Ta, T9, VMUL(T8, Tb));
270 }
271 {
272 V Te, Tg, Td, Tf;
273 Te = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
274 Tg = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
275 Td = LDW(&(W[TWVL * 10]));
276 Tf = LDW(&(W[TWVL * 11]));
277 Th = VFMA(Td, Te, VMUL(Tf, Tg));
278 TJ = VFNMS(Tf, Te, VMUL(Td, Tg));
279 }
280 Ti = VADD(Tc, Th);
281 T1f = VSUB(Tc, Th);
282 TK = VSUB(TI, TJ);
283 T16 = VADD(TI, TJ);
284 }
285 {
286 V To, TN, Tt, TO;
287 {
288 V Tl, Tn, Tk, Tm;
289 Tl = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
290 Tn = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
291 Tk = LDW(&(W[0]));
292 Tm = LDW(&(W[TWVL * 1]));
293 To = VFMA(Tk, Tl, VMUL(Tm, Tn));
294 TN = VFNMS(Tm, Tl, VMUL(Tk, Tn));
295 }
296 {
297 V Tq, Ts, Tp, Tr;
298 Tq = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
299 Ts = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
300 Tp = LDW(&(W[TWVL * 8]));
301 Tr = LDW(&(W[TWVL * 9]));
302 Tt = VFMA(Tp, Tq, VMUL(Tr, Ts));
303 TO = VFNMS(Tr, Tq, VMUL(Tp, Ts));
304 }
305 Tu = VADD(To, Tt);
306 T12 = VADD(TN, TO);
307 TM = VSUB(To, Tt);
308 TP = VSUB(TN, TO);
309 }
310 {
311 V Tj, TG, T1b, T1c;
312 Tj = VADD(T7, Ti);
313 TG = VADD(Tu, TF);
314 ST(&(ri[WS(rs, 4)]), VSUB(Tj, TG), ms, &(ri[0]));
315 ST(&(ri[0]), VADD(Tj, TG), ms, &(ri[0]));
316 {
317 V T15, T1a, T11, T14;
318 T15 = VADD(T12, T13);
319 T1a = VADD(T16, T19);
320 ST(&(ii[0]), VADD(T15, T1a), ms, &(ii[0]));
321 ST(&(ii[WS(rs, 4)]), VSUB(T1a, T15), ms, &(ii[0]));
322 T11 = VSUB(T7, Ti);
323 T14 = VSUB(T12, T13);
324 ST(&(ri[WS(rs, 6)]), VSUB(T11, T14), ms, &(ri[0]));
325 ST(&(ri[WS(rs, 2)]), VADD(T11, T14), ms, &(ri[0]));
326 }
327 T1b = VSUB(TF, Tu);
328 T1c = VSUB(T19, T16);
329 ST(&(ii[WS(rs, 2)]), VADD(T1b, T1c), ms, &(ii[0]));
330 ST(&(ii[WS(rs, 6)]), VSUB(T1c, T1b), ms, &(ii[0]));
331 {
332 V TX, T1g, T10, T1d, TY, TZ;
333 TX = VSUB(TH, TK);
334 T1g = VSUB(T1e, T1f);
335 TY = VSUB(TP, TM);
336 TZ = VADD(TR, TU);
337 T10 = VMUL(LDK(KP707106781), VSUB(TY, TZ));
338 T1d = VMUL(LDK(KP707106781), VADD(TY, TZ));
339 ST(&(ri[WS(rs, 7)]), VSUB(TX, T10), ms, &(ri[WS(rs, 1)]));
340 ST(&(ii[WS(rs, 5)]), VSUB(T1g, T1d), ms, &(ii[WS(rs, 1)]));
341 ST(&(ri[WS(rs, 3)]), VADD(TX, T10), ms, &(ri[WS(rs, 1)]));
342 ST(&(ii[WS(rs, 1)]), VADD(T1d, T1g), ms, &(ii[WS(rs, 1)]));
343 }
344 {
345 V TL, T1i, TW, T1h, TQ, TV;
346 TL = VADD(TH, TK);
347 T1i = VADD(T1f, T1e);
348 TQ = VADD(TM, TP);
349 TV = VSUB(TR, TU);
350 TW = VMUL(LDK(KP707106781), VADD(TQ, TV));
351 T1h = VMUL(LDK(KP707106781), VSUB(TV, TQ));
352 ST(&(ri[WS(rs, 5)]), VSUB(TL, TW), ms, &(ri[WS(rs, 1)]));
353 ST(&(ii[WS(rs, 7)]), VSUB(T1i, T1h), ms, &(ii[WS(rs, 1)]));
354 ST(&(ri[WS(rs, 1)]), VADD(TL, TW), ms, &(ri[WS(rs, 1)]));
355 ST(&(ii[WS(rs, 3)]), VADD(T1h, T1i), ms, &(ii[WS(rs, 1)]));
356 }
357 }
358 }
359 }
360 VLEAVE();
361 }
362
363 static const tw_instr twinstr[] = {
364 VTW(0, 1),
365 VTW(0, 2),
366 VTW(0, 3),
367 VTW(0, 4),
368 VTW(0, 5),
369 VTW(0, 6),
370 VTW(0, 7),
371 {TW_NEXT, (2 * VL), 0}
372 };
373
374 static const ct_desc desc = { 8, XSIMD_STRING("t1sv_8"), twinstr, &GENUS, {52, 18, 14, 0}, 0, 0, 0 };
375
376 void XSIMD(codelet_t1sv_8) (planner *p) {
377 X(kdft_dit_register) (p, t1sv_8, &desc);
378 }
379 #endif /* HAVE_FMA */