Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.3/dft/simd/common/n1fv_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 07:36:52 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 15 -name n1fv_15 -include n1f.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 "n1f.h"
36
37 static void n1fv_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 = ri;
52 xo = ro;
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, TX, TM, TQ, Th, TB, T5, Ti, Ta, TC, TN, Te, TG, Tq, 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 TX = VSUB(T3, T2);
72 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
73 TM = VSUB(T8, T7);
74 T9 = VADD(T7, T8);
75 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
76 Tp = VADD(Tn, To);
77 TQ = VSUB(To, Tn);
78 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
79 TB = VFNMS(LDK(KP500000000), T4, T1);
80 T5 = VADD(T1, T4);
81 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
82 Ta = VADD(T6, T9);
83 TC = VFNMS(LDK(KP500000000), T9, T6);
84 TN = VSUB(Td, Tc);
85 Te = VADD(Tc, Td);
86 TG = VFNMS(LDK(KP500000000), Tp, Tm);
87 Tq = VADD(Tm, Tp);
88 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
89 }
90 }
91 {
92 V TY, TO, Tf, TD, TP, Tk;
93 TY = VADD(TM, TN);
94 TO = VSUB(TM, TN);
95 Tf = VADD(Tb, Te);
96 TD = VFNMS(LDK(KP500000000), Te, Tb);
97 TP = VSUB(Tj, Ti);
98 Tk = VADD(Ti, Tj);
99 {
100 V Tx, Tg, TE, TU, TZ, TR, Tl, TF;
101 Tx = VSUB(Ta, Tf);
102 Tg = VADD(Ta, Tf);
103 TE = VADD(TC, TD);
104 TU = VSUB(TC, TD);
105 TZ = VADD(TP, TQ);
106 TR = VSUB(TP, TQ);
107 Tl = VADD(Th, Tk);
108 TF = VFNMS(LDK(KP500000000), Tk, Th);
109 {
110 V T12, T10, T18, TS, Tw, Tr, TH, TV, T11, T1g;
111 T12 = VSUB(TY, TZ);
112 T10 = VADD(TY, TZ);
113 T18 = VFNMS(LDK(KP618033988), TO, TR);
114 TS = VFMA(LDK(KP618033988), TR, TO);
115 Tw = VSUB(Tl, Tq);
116 Tr = VADD(Tl, Tq);
117 TH = VADD(TF, TG);
118 TV = VSUB(TF, TG);
119 T11 = VFNMS(LDK(KP250000000), T10, TX);
120 T1g = VMUL(LDK(KP866025403), VADD(TX, T10));
121 {
122 V TA, Ty, Tu, TK, TI, T1a, TW, T1b, T13, Tt, Ts, TJ, T1f;
123 TA = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tx));
124 Ty = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tx, Tw));
125 Ts = VADD(Tg, Tr);
126 Tu = VSUB(Tg, Tr);
127 TK = VSUB(TE, TH);
128 TI = VADD(TE, TH);
129 T1a = VFNMS(LDK(KP618033988), TU, TV);
130 TW = VFMA(LDK(KP618033988), TV, TU);
131 T1b = VFNMS(LDK(KP559016994), T12, T11);
132 T13 = VFMA(LDK(KP559016994), T12, T11);
133 ST(&(xo[0]), VADD(T5, Ts), ovs, &(xo[0]));
134 Tt = VFNMS(LDK(KP250000000), Ts, T5);
135 TJ = VFNMS(LDK(KP250000000), TI, TB);
136 T1f = VADD(TB, TI);
137 {
138 V T1c, T1e, T16, T14, Tv, Tz, T17, TL;
139 T1c = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1b, T1a));
140 T1e = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1b, T1a));
141 T16 = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T13, TW));
142 T14 = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T13, TW));
143 Tv = VFNMS(LDK(KP559016994), Tu, Tt);
144 Tz = VFMA(LDK(KP559016994), Tu, Tt);
145 T17 = VFNMS(LDK(KP559016994), TK, TJ);
146 TL = VFMA(LDK(KP559016994), TK, TJ);
147 ST(&(xo[WS(os, 10)]), VFMAI(T1g, T1f), ovs, &(xo[0]));
148 ST(&(xo[WS(os, 5)]), VFNMSI(T1g, T1f), ovs, &(xo[WS(os, 1)]));
149 {
150 V T19, T1d, T15, TT;
151 ST(&(xo[WS(os, 12)]), VFMAI(Ty, Tv), ovs, &(xo[0]));
152 ST(&(xo[WS(os, 3)]), VFNMSI(Ty, Tv), ovs, &(xo[WS(os, 1)]));
153 ST(&(xo[WS(os, 9)]), VFMAI(TA, Tz), ovs, &(xo[WS(os, 1)]));
154 ST(&(xo[WS(os, 6)]), VFNMSI(TA, Tz), ovs, &(xo[0]));
155 T19 = VFMA(LDK(KP823639103), T18, T17);
156 T1d = VFNMS(LDK(KP823639103), T18, T17);
157 T15 = VFNMS(LDK(KP823639103), TS, TL);
158 TT = VFMA(LDK(KP823639103), TS, TL);
159 ST(&(xo[WS(os, 2)]), VFMAI(T1c, T19), ovs, &(xo[0]));
160 ST(&(xo[WS(os, 13)]), VFNMSI(T1c, T19), ovs, &(xo[WS(os, 1)]));
161 ST(&(xo[WS(os, 7)]), VFMAI(T1e, T1d), ovs, &(xo[WS(os, 1)]));
162 ST(&(xo[WS(os, 8)]), VFNMSI(T1e, T1d), ovs, &(xo[0]));
163 ST(&(xo[WS(os, 4)]), VFMAI(T16, T15), ovs, &(xo[0]));
164 ST(&(xo[WS(os, 11)]), VFNMSI(T16, T15), ovs, &(xo[WS(os, 1)]));
165 ST(&(xo[WS(os, 14)]), VFMAI(T14, TT), ovs, &(xo[0]));
166 ST(&(xo[WS(os, 1)]), VFNMSI(T14, TT), 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("n1fv_15"), {36, 7, 42, 0}, &GENUS, 0, 0, 0, 0 };
179
180 void XSIMD(codelet_n1fv_15) (planner *p) {
181 X(kdft_register) (p, n1fv_15, &desc);
182 }
183
184 #else /* HAVE_FMA */
185
186 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 15 -name n1fv_15 -include n1f.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 "n1f.h"
194
195 static void n1fv_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(KP587785252, +0.587785252292473129168705954639072768597652438);
201 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
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 = ri;
212 xo = ro;
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 T5, T10, TB, TO, TU, TV, TR, Ta, Tf, Tg, Tl, Tq, Tr, TE, TH;
215 V TI, TZ, T11, T1f, T1g;
216 {
217 V T1, T2, T3, T4;
218 T1 = LD(&(xi[0]), ivs, &(xi[0]));
219 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
220 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
221 T4 = VADD(T2, T3);
222 T5 = VADD(T1, T4);
223 T10 = VSUB(T3, T2);
224 TB = VFNMS(LDK(KP500000000), T4, T1);
225 }
226 {
227 V T6, T9, TC, TP, Tm, Tp, TG, TN, Tb, Te, TD, TQ, Th, Tk, TF;
228 V TM, TX, TY;
229 {
230 V T7, T8, Tn, To;
231 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
232 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
233 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
234 T9 = VADD(T7, T8);
235 TC = VFNMS(LDK(KP500000000), T9, T6);
236 TP = VSUB(T8, T7);
237 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
238 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
239 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
240 Tp = VADD(Tn, To);
241 TG = VFNMS(LDK(KP500000000), Tp, Tm);
242 TN = VSUB(To, Tn);
243 }
244 {
245 V Tc, Td, Ti, Tj;
246 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
247 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
248 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
249 Te = VADD(Tc, Td);
250 TD = VFNMS(LDK(KP500000000), Te, Tb);
251 TQ = VSUB(Td, Tc);
252 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
253 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
254 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
255 Tk = VADD(Ti, Tj);
256 TF = VFNMS(LDK(KP500000000), Tk, Th);
257 TM = VSUB(Tj, Ti);
258 }
259 TO = VSUB(TM, TN);
260 TU = VSUB(TF, TG);
261 TV = VSUB(TC, TD);
262 TR = VSUB(TP, TQ);
263 Ta = VADD(T6, T9);
264 Tf = VADD(Tb, Te);
265 Tg = VADD(Ta, Tf);
266 Tl = VADD(Th, Tk);
267 Tq = VADD(Tm, Tp);
268 Tr = VADD(Tl, Tq);
269 TE = VADD(TC, TD);
270 TH = VADD(TF, TG);
271 TI = VADD(TE, TH);
272 TX = VADD(TP, TQ);
273 TY = VADD(TM, TN);
274 TZ = VMUL(LDK(KP484122918), VSUB(TX, TY));
275 T11 = VADD(TX, TY);
276 }
277 T1f = VADD(TB, TI);
278 T1g = VBYI(VMUL(LDK(KP866025403), VADD(T10, T11)));
279 ST(&(xo[WS(os, 5)]), VSUB(T1f, T1g), ovs, &(xo[WS(os, 1)]));
280 ST(&(xo[WS(os, 10)]), VADD(T1f, T1g), ovs, &(xo[0]));
281 {
282 V Tu, Ts, Tt, Ty, TA, Tw, Tx, Tz, Tv;
283 Tu = VMUL(LDK(KP559016994), VSUB(Tg, Tr));
284 Ts = VADD(Tg, Tr);
285 Tt = VFNMS(LDK(KP250000000), Ts, T5);
286 Tw = VSUB(Tl, Tq);
287 Tx = VSUB(Ta, Tf);
288 Ty = VBYI(VFNMS(LDK(KP587785252), Tx, VMUL(LDK(KP951056516), Tw)));
289 TA = VBYI(VFMA(LDK(KP951056516), Tx, VMUL(LDK(KP587785252), Tw)));
290 ST(&(xo[0]), VADD(T5, Ts), ovs, &(xo[0]));
291 Tz = VADD(Tu, Tt);
292 ST(&(xo[WS(os, 6)]), VSUB(Tz, TA), ovs, &(xo[0]));
293 ST(&(xo[WS(os, 9)]), VADD(TA, Tz), ovs, &(xo[WS(os, 1)]));
294 Tv = VSUB(Tt, Tu);
295 ST(&(xo[WS(os, 3)]), VSUB(Tv, Ty), ovs, &(xo[WS(os, 1)]));
296 ST(&(xo[WS(os, 12)]), VADD(Ty, Tv), ovs, &(xo[0]));
297 }
298 {
299 V TS, TW, T1b, T18, T13, T1a, TL, T17, T12, TJ, TK;
300 TS = VFNMS(LDK(KP509036960), TR, VMUL(LDK(KP823639103), TO));
301 TW = VFNMS(LDK(KP587785252), TV, VMUL(LDK(KP951056516), TU));
302 T1b = VFMA(LDK(KP951056516), TV, VMUL(LDK(KP587785252), TU));
303 T18 = VFMA(LDK(KP823639103), TR, VMUL(LDK(KP509036960), TO));
304 T12 = VFNMS(LDK(KP216506350), T11, VMUL(LDK(KP866025403), T10));
305 T13 = VSUB(TZ, T12);
306 T1a = VADD(TZ, T12);
307 TJ = VFNMS(LDK(KP250000000), TI, TB);
308 TK = VMUL(LDK(KP559016994), VSUB(TE, TH));
309 TL = VSUB(TJ, TK);
310 T17 = VADD(TK, TJ);
311 {
312 V TT, T14, T1d, T1e;
313 TT = VSUB(TL, TS);
314 T14 = VBYI(VSUB(TW, T13));
315 ST(&(xo[WS(os, 8)]), VSUB(TT, T14), ovs, &(xo[0]));
316 ST(&(xo[WS(os, 7)]), VADD(TT, T14), ovs, &(xo[WS(os, 1)]));
317 T1d = VSUB(T17, T18);
318 T1e = VBYI(VADD(T1b, T1a));
319 ST(&(xo[WS(os, 11)]), VSUB(T1d, T1e), ovs, &(xo[WS(os, 1)]));
320 ST(&(xo[WS(os, 4)]), VADD(T1d, T1e), ovs, &(xo[0]));
321 }
322 {
323 V T15, T16, T19, T1c;
324 T15 = VADD(TL, TS);
325 T16 = VBYI(VADD(TW, T13));
326 ST(&(xo[WS(os, 13)]), VSUB(T15, T16), ovs, &(xo[WS(os, 1)]));
327 ST(&(xo[WS(os, 2)]), VADD(T15, T16), ovs, &(xo[0]));
328 T19 = VADD(T17, T18);
329 T1c = VBYI(VSUB(T1a, T1b));
330 ST(&(xo[WS(os, 14)]), VSUB(T19, T1c), ovs, &(xo[0]));
331 ST(&(xo[WS(os, 1)]), VADD(T19, T1c), ovs, &(xo[WS(os, 1)]));
332 }
333 }
334 }
335 }
336 VLEAVE();
337 }
338
339 static const kdft_desc desc = { 15, XSIMD_STRING("n1fv_15"), {64, 11, 14, 0}, &GENUS, 0, 0, 0, 0 };
340
341 void XSIMD(codelet_n1fv_15) (planner *p) {
342 X(kdft_register) (p, n1fv_15, &desc);
343 }
344
345 #endif /* HAVE_FMA */