Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n1bv_15.c @ 167:bd3cc4d1df30

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam <cannam@all-day-breakfast.com>
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
166:cbd6d7e562c7 167:bd3cc4d1df30
1 /*
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
3 * Copyright (c) 2003, 2007-14 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 Thu May 24 08:04:58 EDT 2018 */
23
24 #include "dft/codelet-dft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_notw_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 15 -name n1bv_15 -include dft/simd/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 * 53 stack variables, 8 constants, and 30 memory accesses
34 */
35 #include "dft/simd/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(KP910592997, +0.910592997310029334643087372129977886038870291);
40 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
41 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
42 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
44 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
45 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
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 T5, T11, TH, Ty, TE, TF, TB, Tg, Tr, Ts, T12, T13, T14, T15, T16;
55 V T17, TK, TM, TZ, T10;
56 {
57 V T1, T2, T3, T4;
58 T1 = LD(&(xi[0]), ivs, &(xi[0]));
59 T2 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
60 T3 = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
61 T4 = VADD(T2, T3);
62 T5 = VFNMS(LDK(KP500000000), T4, T1);
63 T11 = VADD(T1, T4);
64 TH = VSUB(T2, T3);
65 }
66 {
67 V T6, T9, Ta, Tw, Tm, Tp, Tq, TA, Tb, Te, Tf, Tx, Th, Tk, Tl;
68 V Tz, TI, TJ;
69 {
70 V T7, T8, Tn, To;
71 T6 = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
72 T7 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
73 T8 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
74 T9 = VADD(T7, T8);
75 Ta = VFNMS(LDK(KP500000000), T9, T6);
76 Tw = VSUB(T7, T8);
77 Tm = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
78 Tn = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
79 To = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
80 Tp = VADD(Tn, To);
81 Tq = VFNMS(LDK(KP500000000), Tp, Tm);
82 TA = VSUB(Tn, To);
83 }
84 {
85 V Tc, Td, Ti, Tj;
86 Tb = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
87 Tc = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
88 Td = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
89 Te = VADD(Tc, Td);
90 Tf = VFNMS(LDK(KP500000000), Te, Tb);
91 Tx = VSUB(Tc, Td);
92 Th = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
93 Ti = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
94 Tj = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
95 Tk = VADD(Ti, Tj);
96 Tl = VFNMS(LDK(KP500000000), Tk, Th);
97 Tz = VSUB(Ti, Tj);
98 }
99 Ty = VSUB(Tw, Tx);
100 TE = VSUB(Ta, Tf);
101 TF = VSUB(Tl, Tq);
102 TB = VSUB(Tz, TA);
103 Tg = VADD(Ta, Tf);
104 Tr = VADD(Tl, Tq);
105 Ts = VADD(Tg, Tr);
106 T12 = VADD(T6, T9);
107 T13 = VADD(Tb, Te);
108 T14 = VADD(T12, T13);
109 T15 = VADD(Th, Tk);
110 T16 = VADD(Tm, Tp);
111 T17 = VADD(T15, T16);
112 TI = VADD(Tw, Tx);
113 TJ = VADD(Tz, TA);
114 TK = VADD(TI, TJ);
115 TM = VSUB(TI, TJ);
116 }
117 TZ = VADD(T5, Ts);
118 T10 = VMUL(LDK(KP866025403), VADD(TH, TK));
119 ST(&(xo[WS(os, 5)]), VFNMSI(T10, TZ), ovs, &(xo[WS(os, 1)]));
120 ST(&(xo[WS(os, 10)]), VFMAI(T10, TZ), ovs, &(xo[0]));
121 {
122 V T1a, T18, T19, T1e, T1g, T1c, T1d, T1f, T1b;
123 T1a = VSUB(T14, T17);
124 T18 = VADD(T14, T17);
125 T19 = VFNMS(LDK(KP250000000), T18, T11);
126 T1c = VSUB(T15, T16);
127 T1d = VSUB(T12, T13);
128 T1e = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), T1d, T1c));
129 T1g = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), T1c, T1d));
130 ST(&(xo[0]), VADD(T11, T18), ovs, &(xo[0]));
131 T1f = VFMA(LDK(KP559016994), T1a, T19);
132 ST(&(xo[WS(os, 6)]), VFMAI(T1g, T1f), ovs, &(xo[0]));
133 ST(&(xo[WS(os, 9)]), VFNMSI(T1g, T1f), ovs, &(xo[WS(os, 1)]));
134 T1b = VFNMS(LDK(KP559016994), T1a, T19);
135 ST(&(xo[WS(os, 3)]), VFMAI(T1e, T1b), ovs, &(xo[WS(os, 1)]));
136 ST(&(xo[WS(os, 12)]), VFNMSI(T1e, T1b), ovs, &(xo[0]));
137 }
138 {
139 V TC, TG, TU, TS, TN, TV, Tv, TR, TL, Tt, Tu;
140 TC = VFMA(LDK(KP618033988), TB, Ty);
141 TG = VFMA(LDK(KP618033988), TF, TE);
142 TU = VFNMS(LDK(KP618033988), TE, TF);
143 TS = VFNMS(LDK(KP618033988), Ty, TB);
144 TL = VFNMS(LDK(KP250000000), TK, TH);
145 TN = VFMA(LDK(KP559016994), TM, TL);
146 TV = VFNMS(LDK(KP559016994), TM, TL);
147 Tt = VFNMS(LDK(KP250000000), Ts, T5);
148 Tu = VSUB(Tg, Tr);
149 Tv = VFMA(LDK(KP559016994), Tu, Tt);
150 TR = VFNMS(LDK(KP559016994), Tu, Tt);
151 {
152 V TD, TO, TX, TY;
153 TD = VFNMS(LDK(KP823639103), TC, Tv);
154 TO = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), TN, TG));
155 ST(&(xo[WS(os, 1)]), VFMAI(TO, TD), ovs, &(xo[WS(os, 1)]));
156 ST(&(xo[WS(os, 14)]), VFNMSI(TO, TD), ovs, &(xo[0]));
157 TX = VFMA(LDK(KP823639103), TS, TR);
158 TY = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), TV, TU));
159 ST(&(xo[WS(os, 7)]), VFNMSI(TY, TX), ovs, &(xo[WS(os, 1)]));
160 ST(&(xo[WS(os, 8)]), VFMAI(TY, TX), ovs, &(xo[0]));
161 }
162 {
163 V TP, TQ, TT, TW;
164 TP = VFMA(LDK(KP823639103), TC, Tv);
165 TQ = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), TN, TG));
166 ST(&(xo[WS(os, 4)]), VFNMSI(TQ, TP), ovs, &(xo[0]));
167 ST(&(xo[WS(os, 11)]), VFMAI(TQ, TP), ovs, &(xo[WS(os, 1)]));
168 TT = VFNMS(LDK(KP823639103), TS, TR);
169 TW = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), TV, TU));
170 ST(&(xo[WS(os, 2)]), VFNMSI(TW, TT), ovs, &(xo[0]));
171 ST(&(xo[WS(os, 13)]), VFMAI(TW, TT), ovs, &(xo[WS(os, 1)]));
172 }
173 }
174 }
175 }
176 VLEAVE();
177 }
178
179 static const kdft_desc desc = { 15, XSIMD_STRING("n1bv_15"), {36, 7, 42, 0}, &GENUS, 0, 0, 0, 0 };
180
181 void XSIMD(codelet_n1bv_15) (planner *p) {
182 X(kdft_register) (p, n1bv_15, &desc);
183 }
184
185 #else
186
187 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 15 -name n1bv_15 -include dft/simd/n1b.h */
188
189 /*
190 * This function contains 78 FP additions, 25 FP multiplications,
191 * (or, 64 additions, 11 multiplications, 14 fused multiply/add),
192 * 55 stack variables, 10 constants, and 30 memory accesses
193 */
194 #include "dft/simd/n1b.h"
195
196 static void n1bv_15(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
197 {
198 DVK(KP216506350, +0.216506350946109661690930792688234045867850657);
199 DVK(KP509036960, +0.509036960455127183450980863393907648510733164);
200 DVK(KP823639103, +0.823639103546331925877420039278190003029660514);
201 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
202 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
203 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
204 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
205 DVK(KP866025403, +0.866025403784438646763723170752936183471402627);
206 DVK(KP484122918, +0.484122918275927110647408174972799951354115213);
207 DVK(KP500000000, +0.500000000000000000000000000000000000000000000);
208 {
209 INT i;
210 const R *xi;
211 R *xo;
212 xi = ii;
213 xo = io;
214 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)) {
215 V Ti, T11, TH, Ts, TL, TM, Tz, TC, TD, TI, T12, T13, T14, T15, T16;
216 V T17, Tf, Tj, TZ, T10;
217 {
218 V TF, Tg, Th, TG;
219 TF = LD(&(xi[0]), ivs, &(xi[0]));
220 Tg = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
221 Th = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
222 TG = VADD(Tg, Th);
223 Ti = VSUB(Tg, Th);
224 T11 = VADD(TF, TG);
225 TH = VFNMS(LDK(KP500000000), TG, TF);
226 }
227 {
228 V Tm, Tn, T3, To, Tw, Tx, Td, Ty, Tp, Tq, T6, Tr, Tt, Tu, Ta;
229 V Tv, T7, Te;
230 {
231 V T1, T2, Tb, Tc;
232 Tm = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
233 T1 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
234 T2 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
235 Tn = VADD(T1, T2);
236 T3 = VSUB(T1, T2);
237 To = VFNMS(LDK(KP500000000), Tn, Tm);
238 Tw = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
239 Tb = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
240 Tc = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
241 Tx = VADD(Tb, Tc);
242 Td = VSUB(Tb, Tc);
243 Ty = VFNMS(LDK(KP500000000), Tx, Tw);
244 }
245 {
246 V T4, T5, T8, T9;
247 Tp = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
248 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
249 T5 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
250 Tq = VADD(T4, T5);
251 T6 = VSUB(T4, T5);
252 Tr = VFNMS(LDK(KP500000000), Tq, Tp);
253 Tt = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
254 T8 = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
255 T9 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
256 Tu = VADD(T8, T9);
257 Ta = VSUB(T8, T9);
258 Tv = VFNMS(LDK(KP500000000), Tu, Tt);
259 }
260 Ts = VSUB(To, Tr);
261 TL = VSUB(T3, T6);
262 TM = VSUB(Ta, Td);
263 Tz = VSUB(Tv, Ty);
264 TC = VADD(To, Tr);
265 TD = VADD(Tv, Ty);
266 TI = VADD(TC, TD);
267 T12 = VADD(Tm, Tn);
268 T13 = VADD(Tp, Tq);
269 T14 = VADD(T12, T13);
270 T15 = VADD(Tt, Tu);
271 T16 = VADD(Tw, Tx);
272 T17 = VADD(T15, T16);
273 T7 = VADD(T3, T6);
274 Te = VADD(Ta, Td);
275 Tf = VMUL(LDK(KP484122918), VSUB(T7, Te));
276 Tj = VADD(T7, Te);
277 }
278 TZ = VADD(TH, TI);
279 T10 = VBYI(VMUL(LDK(KP866025403), VADD(Ti, Tj)));
280 ST(&(xo[WS(os, 5)]), VSUB(TZ, T10), ovs, &(xo[WS(os, 1)]));
281 ST(&(xo[WS(os, 10)]), VADD(T10, TZ), ovs, &(xo[0]));
282 {
283 V T1a, T18, T19, T1e, T1f, T1c, T1d, T1g, T1b;
284 T1a = VMUL(LDK(KP559016994), VSUB(T14, T17));
285 T18 = VADD(T14, T17);
286 T19 = VFNMS(LDK(KP250000000), T18, T11);
287 T1c = VSUB(T12, T13);
288 T1d = VSUB(T15, T16);
289 T1e = VBYI(VFNMS(LDK(KP951056516), T1d, VMUL(LDK(KP587785252), T1c)));
290 T1f = VBYI(VFMA(LDK(KP951056516), T1c, VMUL(LDK(KP587785252), T1d)));
291 ST(&(xo[0]), VADD(T11, T18), ovs, &(xo[0]));
292 T1g = VADD(T1a, T19);
293 ST(&(xo[WS(os, 6)]), VADD(T1f, T1g), ovs, &(xo[0]));
294 ST(&(xo[WS(os, 9)]), VSUB(T1g, T1f), ovs, &(xo[WS(os, 1)]));
295 T1b = VSUB(T19, T1a);
296 ST(&(xo[WS(os, 3)]), VSUB(T1b, T1e), ovs, &(xo[WS(os, 1)]));
297 ST(&(xo[WS(os, 12)]), VADD(T1e, T1b), ovs, &(xo[0]));
298 }
299 {
300 V TA, TN, TU, TS, Tl, TR, TK, TV, Tk, TE, TJ;
301 TA = VFMA(LDK(KP951056516), Ts, VMUL(LDK(KP587785252), Tz));
302 TN = VFMA(LDK(KP823639103), TL, VMUL(LDK(KP509036960), TM));
303 TU = VFNMS(LDK(KP823639103), TM, VMUL(LDK(KP509036960), TL));
304 TS = VFNMS(LDK(KP951056516), Tz, VMUL(LDK(KP587785252), Ts));
305 Tk = VFNMS(LDK(KP216506350), Tj, VMUL(LDK(KP866025403), Ti));
306 Tl = VADD(Tf, Tk);
307 TR = VSUB(Tf, Tk);
308 TE = VMUL(LDK(KP559016994), VSUB(TC, TD));
309 TJ = VFNMS(LDK(KP250000000), TI, TH);
310 TK = VADD(TE, TJ);
311 TV = VSUB(TJ, TE);
312 {
313 V TB, TO, TX, TY;
314 TB = VBYI(VADD(Tl, TA));
315 TO = VSUB(TK, TN);
316 ST(&(xo[WS(os, 1)]), VADD(TB, TO), ovs, &(xo[WS(os, 1)]));
317 ST(&(xo[WS(os, 14)]), VSUB(TO, TB), ovs, &(xo[0]));
318 TX = VBYI(VSUB(TS, TR));
319 TY = VSUB(TV, TU);
320 ST(&(xo[WS(os, 7)]), VADD(TX, TY), ovs, &(xo[WS(os, 1)]));
321 ST(&(xo[WS(os, 8)]), VSUB(TY, TX), ovs, &(xo[0]));
322 }
323 {
324 V TP, TQ, TT, TW;
325 TP = VBYI(VSUB(Tl, TA));
326 TQ = VADD(TN, TK);
327 ST(&(xo[WS(os, 4)]), VADD(TP, TQ), ovs, &(xo[0]));
328 ST(&(xo[WS(os, 11)]), VSUB(TQ, TP), ovs, &(xo[WS(os, 1)]));
329 TT = VBYI(VADD(TR, TS));
330 TW = VADD(TU, TV);
331 ST(&(xo[WS(os, 2)]), VADD(TT, TW), ovs, &(xo[0]));
332 ST(&(xo[WS(os, 13)]), VSUB(TW, TT), ovs, &(xo[WS(os, 1)]));
333 }
334 }
335 }
336 }
337 VLEAVE();
338 }
339
340 static const kdft_desc desc = { 15, XSIMD_STRING("n1bv_15"), {64, 11, 14, 0}, &GENUS, 0, 0, 0, 0 };
341
342 void XSIMD(codelet_n1bv_15) (planner *p) {
343 X(kdft_register) (p, n1bv_15, &desc);
344 }
345
346 #endif