Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/t2sv_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 07:39:26 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 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include ts.h */
29
30 /*
31 * This function contains 74 FP additions, 50 FP multiplications,
32 * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
33 * 64 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "ts.h"
36
37 static void t2sv_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 * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
43 V T1m, T1l, T1k, T1u, T1n, T1o;
44 {
45 V T2, T3, Tl, Tn, T5, T6;
46 T2 = LDW(&(W[0]));
47 T3 = LDW(&(W[TWVL * 2]));
48 Tl = LDW(&(W[TWVL * 4]));
49 Tn = LDW(&(W[TWVL * 5]));
50 T5 = LDW(&(W[TWVL * 1]));
51 T6 = LDW(&(W[TWVL * 3]));
52 {
53 V T1, T1s, TK, T1r, Td, Tk, TG, TC, TY, Tu, TW, TL, TM, TO, TQ;
54 V Tx, Tz, TD, TH;
55 {
56 V T8, T4, Tm, Tr, Tc, Ta;
57 T1 = LD(&(ri[0]), ms, &(ri[0]));
58 T1s = LD(&(ii[0]), ms, &(ii[0]));
59 T8 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
60 T4 = VMUL(T2, T3);
61 Tm = VMUL(T2, Tl);
62 Tr = VMUL(T2, Tn);
63 Tc = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
64 Ta = VMUL(T2, T6);
65 {
66 V Tp, Tt, Tg, T7, Tf, To, Ts, Ti, Tb, Tj;
67 Tp = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
68 Tt = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
69 Tg = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
70 T7 = VFNMS(T5, T6, T4);
71 Tf = VFMA(T5, T6, T4);
72 To = VFMA(T5, Tn, Tm);
73 Ts = VFNMS(T5, Tl, Tr);
74 Ti = VFNMS(T5, T3, Ta);
75 Tb = VFMA(T5, T3, Ta);
76 Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
77 TK = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
78 {
79 V T1q, T9, Th, TF;
80 T1q = VMUL(T7, Tc);
81 T9 = VMUL(T7, T8);
82 Th = VMUL(Tf, Tg);
83 TF = VMUL(Tf, Tn);
84 {
85 V TB, TX, Tq, TV;
86 TB = VMUL(Tf, Tl);
87 TX = VMUL(To, Tt);
88 Tq = VMUL(To, Tp);
89 TV = VMUL(Tf, Tj);
90 T1r = VFNMS(Tb, T8, T1q);
91 Td = VFMA(Tb, Tc, T9);
92 Tk = VFMA(Ti, Tj, Th);
93 TG = VFNMS(Ti, Tl, TF);
94 TC = VFMA(Ti, Tn, TB);
95 TY = VFNMS(Ts, Tp, TX);
96 Tu = VFMA(Ts, Tt, Tq);
97 TW = VFNMS(Ti, Tg, TV);
98 TL = VMUL(Tl, TK);
99 }
100 }
101 TM = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
102 TO = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
103 TQ = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
104 Tx = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
105 Tz = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
106 TD = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
107 TH = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
108 }
109 }
110 {
111 V Te, T1p, T1g, T10, TS, T18, T1d, T1t, T1x, T1y, Tv, TJ, T11, T16;
112 {
113 V TN, T1a, TR, T1c, TA, T13, TI, T15;
114 {
115 V TU, T19, TP, T1b, Ty, T12, TE, T14, TZ;
116 TU = VSUB(T1, Td);
117 Te = VADD(T1, Td);
118 TN = VFMA(Tn, TM, TL);
119 T19 = VMUL(Tl, TM);
120 TP = VMUL(T3, TO);
121 T1b = VMUL(T3, TQ);
122 Ty = VMUL(T2, Tx);
123 T12 = VMUL(T2, Tz);
124 TE = VMUL(TC, TD);
125 T14 = VMUL(TC, TH);
126 T1p = VADD(TW, TY);
127 TZ = VSUB(TW, TY);
128 T1a = VFNMS(Tn, TK, T19);
129 TR = VFMA(T6, TQ, TP);
130 T1c = VFNMS(T6, TO, T1b);
131 TA = VFMA(T5, Tz, Ty);
132 T13 = VFNMS(T5, Tx, T12);
133 TI = VFMA(TG, TH, TE);
134 T15 = VFNMS(TG, TD, T14);
135 T1g = VSUB(TU, TZ);
136 T10 = VADD(TU, TZ);
137 }
138 TS = VADD(TN, TR);
139 T18 = VSUB(TN, TR);
140 T1d = VSUB(T1a, T1c);
141 T1m = VADD(T1a, T1c);
142 T1t = VADD(T1r, T1s);
143 T1x = VSUB(T1s, T1r);
144 T1y = VSUB(Tk, Tu);
145 Tv = VADD(Tk, Tu);
146 TJ = VADD(TA, TI);
147 T11 = VSUB(TA, TI);
148 T16 = VSUB(T13, T15);
149 T1l = VADD(T13, T15);
150 }
151 {
152 V Tw, T1w, T1v, TT;
153 {
154 V T1i, T1e, T1B, T1z, T1h, T17;
155 T1i = VADD(T18, T1d);
156 T1e = VSUB(T18, T1d);
157 T1B = VADD(T1y, T1x);
158 T1z = VSUB(T1x, T1y);
159 T1h = VSUB(T16, T11);
160 T17 = VADD(T11, T16);
161 T1k = VSUB(Te, Tv);
162 Tw = VADD(Te, Tv);
163 {
164 V T1A, T1j, T1C, T1f;
165 T1A = VADD(T1h, T1i);
166 T1j = VSUB(T1h, T1i);
167 T1C = VSUB(T1e, T17);
168 T1f = VADD(T17, T1e);
169 T1w = VSUB(T1t, T1p);
170 T1u = VADD(T1p, T1t);
171 T1v = VSUB(TS, TJ);
172 TT = VADD(TJ, TS);
173 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
174 ST(&(ii[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
175 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
176 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
177 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
178 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
179 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
180 ST(&(ri[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
181 }
182 }
183 ST(&(ri[WS(rs, 4)]), VSUB(Tw, TT), ms, &(ri[0]));
184 ST(&(ri[0]), VADD(Tw, TT), ms, &(ri[0]));
185 ST(&(ii[WS(rs, 6)]), VSUB(T1w, T1v), ms, &(ii[0]));
186 ST(&(ii[WS(rs, 2)]), VADD(T1v, T1w), ms, &(ii[0]));
187 }
188 }
189 }
190 }
191 T1n = VSUB(T1l, T1m);
192 T1o = VADD(T1l, T1m);
193 ST(&(ii[0]), VADD(T1o, T1u), ms, &(ii[0]));
194 ST(&(ii[WS(rs, 4)]), VSUB(T1u, T1o), ms, &(ii[0]));
195 ST(&(ri[WS(rs, 2)]), VADD(T1k, T1n), ms, &(ri[0]));
196 ST(&(ri[WS(rs, 6)]), VSUB(T1k, T1n), ms, &(ri[0]));
197 }
198 }
199 VLEAVE();
200 }
201
202 static const tw_instr twinstr[] = {
203 VTW(0, 1),
204 VTW(0, 3),
205 VTW(0, 7),
206 {TW_NEXT, (2 * VL), 0}
207 };
208
209 static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
210
211 void XSIMD(codelet_t2sv_8) (planner *p) {
212 X(kdft_dit_register) (p, t2sv_8, &desc);
213 }
214 #else /* HAVE_FMA */
215
216 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include ts.h */
217
218 /*
219 * This function contains 74 FP additions, 44 FP multiplications,
220 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
221 * 42 stack variables, 1 constants, and 32 memory accesses
222 */
223 #include "ts.h"
224
225 static void t2sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
226 {
227 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
228 {
229 INT m;
230 for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
231 V T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
232 {
233 V T4, Tb, T7, Ta;
234 T2 = LDW(&(W[0]));
235 T5 = LDW(&(W[TWVL * 1]));
236 T3 = LDW(&(W[TWVL * 2]));
237 T6 = LDW(&(W[TWVL * 3]));
238 T4 = VMUL(T2, T3);
239 Tb = VMUL(T5, T3);
240 T7 = VMUL(T5, T6);
241 Ta = VMUL(T2, T6);
242 T8 = VSUB(T4, T7);
243 Tc = VADD(Ta, Tb);
244 Tg = VADD(T4, T7);
245 Ti = VSUB(Ta, Tb);
246 Tl = LDW(&(W[TWVL * 4]));
247 Tm = LDW(&(W[TWVL * 5]));
248 Tn = VFMA(T2, Tl, VMUL(T5, Tm));
249 Tz = VFNMS(Ti, Tl, VMUL(Tg, Tm));
250 Tp = VFNMS(T5, Tl, VMUL(T2, Tm));
251 Tx = VFMA(Tg, Tl, VMUL(Ti, Tm));
252 }
253 {
254 V Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
255 V TT;
256 {
257 V T1, T1c, Te, T1b, T9, Td;
258 T1 = LD(&(ri[0]), ms, &(ri[0]));
259 T1c = LD(&(ii[0]), ms, &(ii[0]));
260 T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
261 Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
262 Te = VFMA(T8, T9, VMUL(Tc, Td));
263 T1b = VFNMS(Tc, T9, VMUL(T8, Td));
264 Tf = VADD(T1, Te);
265 T1i = VSUB(T1c, T1b);
266 TL = VSUB(T1, Te);
267 T1d = VADD(T1b, T1c);
268 }
269 {
270 V TF, TW, TI, TX;
271 {
272 V TD, TE, TG, TH;
273 TD = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
274 TE = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
275 TF = VFMA(Tl, TD, VMUL(Tm, TE));
276 TW = VFNMS(Tm, TD, VMUL(Tl, TE));
277 TG = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
278 TH = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
279 TI = VFMA(T3, TG, VMUL(T6, TH));
280 TX = VFNMS(T6, TG, VMUL(T3, TH));
281 }
282 TJ = VADD(TF, TI);
283 T17 = VADD(TW, TX);
284 TV = VSUB(TF, TI);
285 TY = VSUB(TW, TX);
286 }
287 {
288 V Tk, TM, Tr, TN;
289 {
290 V Th, Tj, To, Tq;
291 Th = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
292 Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
293 Tk = VFMA(Tg, Th, VMUL(Ti, Tj));
294 TM = VFNMS(Ti, Th, VMUL(Tg, Tj));
295 To = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
296 Tq = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
297 Tr = VFMA(Tn, To, VMUL(Tp, Tq));
298 TN = VFNMS(Tp, To, VMUL(Tn, Tq));
299 }
300 Ts = VADD(Tk, Tr);
301 T1j = VSUB(Tk, Tr);
302 TO = VSUB(TM, TN);
303 T1a = VADD(TM, TN);
304 }
305 {
306 V Tw, TR, TB, TS;
307 {
308 V Tu, Tv, Ty, TA;
309 Tu = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
310 Tv = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
311 Tw = VFMA(T2, Tu, VMUL(T5, Tv));
312 TR = VFNMS(T5, Tu, VMUL(T2, Tv));
313 Ty = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
314 TA = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
315 TB = VFMA(Tx, Ty, VMUL(Tz, TA));
316 TS = VFNMS(Tz, Ty, VMUL(Tx, TA));
317 }
318 TC = VADD(Tw, TB);
319 T16 = VADD(TR, TS);
320 TQ = VSUB(Tw, TB);
321 TT = VSUB(TR, TS);
322 }
323 {
324 V Tt, TK, T1f, T1g;
325 Tt = VADD(Tf, Ts);
326 TK = VADD(TC, TJ);
327 ST(&(ri[WS(rs, 4)]), VSUB(Tt, TK), ms, &(ri[0]));
328 ST(&(ri[0]), VADD(Tt, TK), ms, &(ri[0]));
329 {
330 V T19, T1e, T15, T18;
331 T19 = VADD(T16, T17);
332 T1e = VADD(T1a, T1d);
333 ST(&(ii[0]), VADD(T19, T1e), ms, &(ii[0]));
334 ST(&(ii[WS(rs, 4)]), VSUB(T1e, T19), ms, &(ii[0]));
335 T15 = VSUB(Tf, Ts);
336 T18 = VSUB(T16, T17);
337 ST(&(ri[WS(rs, 6)]), VSUB(T15, T18), ms, &(ri[0]));
338 ST(&(ri[WS(rs, 2)]), VADD(T15, T18), ms, &(ri[0]));
339 }
340 T1f = VSUB(TJ, TC);
341 T1g = VSUB(T1d, T1a);
342 ST(&(ii[WS(rs, 2)]), VADD(T1f, T1g), ms, &(ii[0]));
343 ST(&(ii[WS(rs, 6)]), VSUB(T1g, T1f), ms, &(ii[0]));
344 {
345 V T11, T1k, T14, T1h, T12, T13;
346 T11 = VSUB(TL, TO);
347 T1k = VSUB(T1i, T1j);
348 T12 = VSUB(TT, TQ);
349 T13 = VADD(TV, TY);
350 T14 = VMUL(LDK(KP707106781), VSUB(T12, T13));
351 T1h = VMUL(LDK(KP707106781), VADD(T12, T13));
352 ST(&(ri[WS(rs, 7)]), VSUB(T11, T14), ms, &(ri[WS(rs, 1)]));
353 ST(&(ii[WS(rs, 5)]), VSUB(T1k, T1h), ms, &(ii[WS(rs, 1)]));
354 ST(&(ri[WS(rs, 3)]), VADD(T11, T14), ms, &(ri[WS(rs, 1)]));
355 ST(&(ii[WS(rs, 1)]), VADD(T1h, T1k), ms, &(ii[WS(rs, 1)]));
356 }
357 {
358 V TP, T1m, T10, T1l, TU, TZ;
359 TP = VADD(TL, TO);
360 T1m = VADD(T1j, T1i);
361 TU = VADD(TQ, TT);
362 TZ = VSUB(TV, TY);
363 T10 = VMUL(LDK(KP707106781), VADD(TU, TZ));
364 T1l = VMUL(LDK(KP707106781), VSUB(TZ, TU));
365 ST(&(ri[WS(rs, 5)]), VSUB(TP, T10), ms, &(ri[WS(rs, 1)]));
366 ST(&(ii[WS(rs, 7)]), VSUB(T1m, T1l), ms, &(ii[WS(rs, 1)]));
367 ST(&(ri[WS(rs, 1)]), VADD(TP, T10), ms, &(ri[WS(rs, 1)]));
368 ST(&(ii[WS(rs, 3)]), VADD(T1l, T1m), ms, &(ii[WS(rs, 1)]));
369 }
370 }
371 }
372 }
373 }
374 VLEAVE();
375 }
376
377 static const tw_instr twinstr[] = {
378 VTW(0, 1),
379 VTW(0, 3),
380 VTW(0, 7),
381 {TW_NEXT, (2 * VL), 0}
382 };
383
384 static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
385
386 void XSIMD(codelet_t2sv_8) (planner *p) {
387 X(kdft_dit_register) (p, t2sv_8, &desc);
388 }
389 #endif /* HAVE_FMA */