Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1bv_15.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 07:37:04 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 15 -name n1bv_15 -include n1b.h */
29
30 /*
31 * This function contains 78 FP additions, 49 FP multiplications,
32 * (or, 36 additions, 7 multiplications, 42 fused multiply/add),
33 * 78 stack variables, 8 constants, and 30 memory accesses
34 */
35 #include "n1b.h"
36
37 static void n1bv_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
40 DVK(KP910592997, +0.910592997310029334643087372129977886038870291);
41 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
42 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
43 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
44 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
45 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
46 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
47 {
48 INT i;
49 const R *xi;
50 R *xo;
51 xi = ii;
52 xo = io;
53 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(30, is), MAKE_VOLATILE_STRIDE(30, os)) {
54 V Tb, TH, Tw, TA, Th, T11, T5, Ti, T12, Ta, Tx, Te, Tq, T16, Tj;
55 V T1, T2, T3;
56 T1 = LD(&(xi[0]), ivs, &(xi[0]));
57 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
58 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
59 {
60 V T6, T7, T8, Tm, Tn, To;
61 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
62 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
63 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
64 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
65 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
66 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
67 {
68 V T4, Tc, T9, Td, Tp;
69 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
70 T4 = VADD(T2, T3);
71 TH = VSUB(T2, T3);
72 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
73 Tw = VSUB(T7, T8);
74 T9 = VADD(T7, T8);
75 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
76 Tp = VADD(Tn, To);
77 TA = VSUB(Tn, To);
78 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
79 T11 = VADD(T1, T4);
80 T5 = VFNMS(LDK(KP500000000), T4, T1);
81 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
82 T12 = VADD(T6, T9);
83 Ta = VFNMS(LDK(KP500000000), T9, T6);
84 Tx = VSUB(Tc, Td);
85 Te = VADD(Tc, Td);
86 Tq = VFNMS(LDK(KP500000000), Tp, Tm);
87 T16 = VADD(Tm, Tp);
88 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
89 }
90 }
91 {
92 V TI, Ty, T13, Tf, Tz, Tk;
93 TI = VADD(Tw, Tx);
94 Ty = VSUB(Tw, Tx);
95 T13 = VADD(Tb, Te);
96 Tf = VFNMS(LDK(KP500000000), Te, Tb);
97 Tz = VSUB(Ti, Tj);
98 Tk = VADD(Ti, Tj);
99 {
100 V T1d, T14, Tg, TE, TJ, TB, T15, Tl;
101 T1d = VSUB(T12, T13);
102 T14 = VADD(T12, T13);
103 Tg = VADD(Ta, Tf);
104 TE = VSUB(Ta, Tf);
105 TJ = VADD(Tz, TA);
106 TB = VSUB(Tz, TA);
107 T15 = VADD(Th, Tk);
108 Tl = VFNMS(LDK(KP500000000), Tk, Th);
109 {
110 V TM, TK, TS, TC, T1c, T17, Tr, TF, TL, T10;
111 TM = VSUB(TI, TJ);
112 TK = VADD(TI, TJ);
113 TS = VFNMS(LDK(KP618033988), Ty, TB);
114 TC = VFMA(LDK(KP618033988), TB, Ty);
115 T1c = VSUB(T15, T16);
116 T17 = VADD(T15, T16);
117 Tr = VADD(Tl, Tq);
118 TF = VSUB(Tl, Tq);
119 TL = VFNMS(LDK(KP250000000), TK, TH);
120 T10 = VMUL(LDK(KP866025403), VADD(TH, TK));
121 {
122 V T1g, T1e, T1a, Tu, Ts, TU, TG, TV, TN, T19, T18, Tt, TZ;
123 T1g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1c, T1d));
124 T1e = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1d, T1c));
125 T18 = VADD(T14, T17);
126 T1a = VSUB(T14, T17);
127 Tu = VSUB(Tg, Tr);
128 Ts = VADD(Tg, Tr);
129 TU = VFNMS(LDK(KP618033988), TE, TF);
130 TG = VFMA(LDK(KP618033988), TF, TE);
131 TV = VFNMS(LDK(KP559016994), TM, TL);
132 TN = VFMA(LDK(KP559016994), TM, TL);
133 ST(&(xo[0]), VADD(T11, T18), ovs, &(xo[0]));
134 T19 = VFNMS(LDK(KP250000000), T18, T11);
135 Tt = VFNMS(LDK(KP250000000), Ts, T5);
136 TZ = VADD(T5, Ts);
137 {
138 V TW, TY, TQ, TO, T1b, T1f, TR, Tv;
139 TW = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), TV, TU));
140 TY = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), TV, TU));
141 TQ = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), TN, TG));
142 TO = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), TN, TG));
143 T1b = VFNMS(LDK(KP559016994), T1a, T19);
144 T1f = VFMA(LDK(KP559016994), T1a, T19);
145 TR = VFNMS(LDK(KP559016994), Tu, Tt);
146 Tv = VFMA(LDK(KP559016994), Tu, Tt);
147 ST(&(xo[WS(os, 10)]), VFMAI(T10, TZ), ovs, &(xo[0]));
148 ST(&(xo[WS(os, 5)]), VFNMSI(T10, TZ), ovs, &(xo[WS(os, 1)]));
149 {
150 V TT, TX, TP, TD;
151 ST(&(xo[WS(os, 12)]), VFNMSI(T1e, T1b), ovs, &(xo[0]));
152 ST(&(xo[WS(os, 3)]), VFMAI(T1e, T1b), ovs, &(xo[WS(os, 1)]));
153 ST(&(xo[WS(os, 9)]), VFNMSI(T1g, T1f), ovs, &(xo[WS(os, 1)]));
154 ST(&(xo[WS(os, 6)]), VFMAI(T1g, T1f), ovs, &(xo[0]));
155 TT = VFNMS(LDK(KP823639103), TS, TR);
156 TX = VFMA(LDK(KP823639103), TS, TR);
157 TP = VFMA(LDK(KP823639103), TC, Tv);
158 TD = VFNMS(LDK(KP823639103), TC, Tv);
159 ST(&(xo[WS(os, 13)]), VFMAI(TW, TT), ovs, &(xo[WS(os, 1)]));
160 ST(&(xo[WS(os, 2)]), VFNMSI(TW, TT), ovs, &(xo[0]));
161 ST(&(xo[WS(os, 8)]), VFMAI(TY, TX), ovs, &(xo[0]));
162 ST(&(xo[WS(os, 7)]), VFNMSI(TY, TX), ovs, &(xo[WS(os, 1)]));
163 ST(&(xo[WS(os, 11)]), VFMAI(TQ, TP), ovs, &(xo[WS(os, 1)]));
164 ST(&(xo[WS(os, 4)]), VFNMSI(TQ, TP), ovs, &(xo[0]));
165 ST(&(xo[WS(os, 14)]), VFNMSI(TO, TD), ovs, &(xo[0]));
166 ST(&(xo[WS(os, 1)]), VFMAI(TO, TD), ovs, &(xo[WS(os, 1)]));
167 }
168 }
169 }
170 }
171 }
172 }
173 }
174 }
175 VLEAVE();
176 }
177
178 static const kdft_desc desc = { 15, XSIMD_STRING("n1bv_15"), {36, 7, 42, 0}, &GENUS, 0, 0, 0, 0 };
179
180 void XSIMD(codelet_n1bv_15) (planner *p) {
181 X(kdft_register) (p, n1bv_15, &desc);
182 }
183
184 #else /* HAVE_FMA */
185
186 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 15 -name n1bv_15 -include n1b.h */
187
188 /*
189 * This function contains 78 FP additions, 25 FP multiplications,
190 * (or, 64 additions, 11 multiplications, 14 fused multiply/add),
191 * 55 stack variables, 10 constants, and 30 memory accesses
192 */
193 #include "n1b.h"
194
195 static void n1bv_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
196 {
197 DVK(KP216506350, +0.216506350946109661690930792688234045867850657);
198 DVK(KP509036960, +0.509036960455127183450980863393907648510733164);
199 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
200 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
201 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
202 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
203 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
204 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
205 DVK(KP484122918, +0.484122918275927110647408174972799951354115213);
206 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
207 {
208 INT i;
209 const R *xi;
210 R *xo;
211 xi = ii;
212 xo = io;
213 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(30, is), MAKE_VOLATILE_STRIDE(30, os)) {
214 V Ti, T11, TH, Ts, TL, TM, Tz, TC, TD, TI, T12, T13, T14, T15, T16;
215 V T17, Tf, Tj, TZ, T10;
216 {
217 V TF, Tg, Th, TG;
218 TF = LD(&(xi[0]), ivs, &(xi[0]));
219 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
220 Th = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
221 TG = VADD(Tg, Th);
222 Ti = VSUB(Tg, Th);
223 T11 = VADD(TF, TG);
224 TH = VFNMS(LDK(KP500000000), TG, TF);
225 }
226 {
227 V Tm, Tn, T3, To, Tw, Tx, Td, Ty, Tp, Tq, T6, Tr, Tt, Tu, Ta;
228 V Tv, T7, Te;
229 {
230 V T1, T2, Tb, Tc;
231 Tm = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
232 T1 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
233 T2 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
234 Tn = VADD(T1, T2);
235 T3 = VSUB(T1, T2);
236 To = VFNMS(LDK(KP500000000), Tn, Tm);
237 Tw = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
238 Tb = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
239 Tc = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
240 Tx = VADD(Tb, Tc);
241 Td = VSUB(Tb, Tc);
242 Ty = VFNMS(LDK(KP500000000), Tx, Tw);
243 }
244 {
245 V T4, T5, T8, T9;
246 Tp = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
247 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
248 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
249 Tq = VADD(T4, T5);
250 T6 = VSUB(T4, T5);
251 Tr = VFNMS(LDK(KP500000000), Tq, Tp);
252 Tt = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
253 T8 = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
254 T9 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
255 Tu = VADD(T8, T9);
256 Ta = VSUB(T8, T9);
257 Tv = VFNMS(LDK(KP500000000), Tu, Tt);
258 }
259 Ts = VSUB(To, Tr);
260 TL = VSUB(T3, T6);
261 TM = VSUB(Ta, Td);
262 Tz = VSUB(Tv, Ty);
263 TC = VADD(To, Tr);
264 TD = VADD(Tv, Ty);
265 TI = VADD(TC, TD);
266 T12 = VADD(Tm, Tn);
267 T13 = VADD(Tp, Tq);
268 T14 = VADD(T12, T13);
269 T15 = VADD(Tt, Tu);
270 T16 = VADD(Tw, Tx);
271 T17 = VADD(T15, T16);
272 T7 = VADD(T3, T6);
273 Te = VADD(Ta, Td);
274 Tf = VMUL(LDK(KP484122918), VSUB(T7, Te));
275 Tj = VADD(T7, Te);
276 }
277 TZ = VADD(TH, TI);
278 T10 = VBYI(VMUL(LDK(KP866025403), VADD(Ti, Tj)));
279 ST(&(xo[WS(os, 5)]), VSUB(TZ, T10), ovs, &(xo[WS(os, 1)]));
280 ST(&(xo[WS(os, 10)]), VADD(T10, TZ), ovs, &(xo[0]));
281 {
282 V T1a, T18, T19, T1e, T1f, T1c, T1d, T1g, T1b;
283 T1a = VMUL(LDK(KP559016994), VSUB(T14, T17));
284 T18 = VADD(T14, T17);
285 T19 = VFNMS(LDK(KP250000000), T18, T11);
286 T1c = VSUB(T12, T13);
287 T1d = VSUB(T15, T16);
288 T1e = VBYI(VFNMS(LDK(KP951056516), T1d, VMUL(LDK(KP587785252), T1c)));
289 T1f = VBYI(VFMA(LDK(KP951056516), T1c, VMUL(LDK(KP587785252), T1d)));
290 ST(&(xo[0]), VADD(T11, T18), ovs, &(xo[0]));
291 T1g = VADD(T1a, T19);
292 ST(&(xo[WS(os, 6)]), VADD(T1f, T1g), ovs, &(xo[0]));
293 ST(&(xo[WS(os, 9)]), VSUB(T1g, T1f), ovs, &(xo[WS(os, 1)]));
294 T1b = VSUB(T19, T1a);
295 ST(&(xo[WS(os, 3)]), VSUB(T1b, T1e), ovs, &(xo[WS(os, 1)]));
296 ST(&(xo[WS(os, 12)]), VADD(T1e, T1b), ovs, &(xo[0]));
297 }
298 {
299 V TA, TN, TU, TS, Tl, TR, TK, TV, Tk, TE, TJ;
300 TA = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz));
301 TN = VFMA(LDK(KP823639103), TL, VMUL(LDK(KP509036960), TM));
302 TU = VFNMS(LDK(KP823639103), TM, VMUL(LDK(KP509036960), TL));
303 TS = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts));
304 Tk = VFNMS(LDK(KP216506350), Tj, VMUL(LDK(KP866025403), Ti));
305 Tl = VADD(Tf, Tk);
306 TR = VSUB(Tf, Tk);
307 TE = VMUL(LDK(KP559016994), VSUB(TC, TD));
308 TJ = VFNMS(LDK(KP250000000), TI, TH);
309 TK = VADD(TE, TJ);
310 TV = VSUB(TJ, TE);
311 {
312 V TB, TO, TX, TY;
313 TB = VBYI(VADD(Tl, TA));
314 TO = VSUB(TK, TN);
315 ST(&(xo[WS(os, 1)]), VADD(TB, TO), ovs, &(xo[WS(os, 1)]));
316 ST(&(xo[WS(os, 14)]), VSUB(TO, TB), ovs, &(xo[0]));
317 TX = VBYI(VSUB(TS, TR));
318 TY = VSUB(TV, TU);
319 ST(&(xo[WS(os, 7)]), VADD(TX, TY), ovs, &(xo[WS(os, 1)]));
320 ST(&(xo[WS(os, 8)]), VSUB(TY, TX), ovs, &(xo[0]));
321 }
322 {
323 V TP, TQ, TT, TW;
324 TP = VBYI(VSUB(Tl, TA));
325 TQ = VADD(TN, TK);
326 ST(&(xo[WS(os, 4)]), VADD(TP, TQ), ovs, &(xo[0]));
327 ST(&(xo[WS(os, 11)]), VSUB(TQ, TP), ovs, &(xo[WS(os, 1)]));
328 TT = VBYI(VADD(TR, TS));
329 TW = VADD(TU, TV);
330 ST(&(xo[WS(os, 2)]), VADD(TT, TW), ovs, &(xo[0]));
331 ST(&(xo[WS(os, 13)]), VSUB(TW, TT), ovs, &(xo[WS(os, 1)]));
332 }
333 }
334 }
335 }
336 VLEAVE();
337 }
338
339 static const kdft_desc desc = { 15, XSIMD_STRING("n1bv_15"), {64, 11, 14, 0}, &GENUS, 0, 0, 0, 0 };
340
341 void XSIMD(codelet_n1bv_15) (planner *p) {
342 X(kdft_register) (p, n1bv_15, &desc);
343 }
344
345 #endif /* HAVE_FMA */