Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2bv_14.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 08:05:11 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 14 -name n2bv_14 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 74 FP additions, 48 FP multiplications,
32 * (or, 32 additions, 6 multiplications, 42 fused multiply/add),
33 * 51 stack variables, 6 constants, and 35 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
37 static void n2bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP801937735, +0.801937735804838252472204639014890102331838324);
40 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
41 DVK(KP554958132, +0.554958132087371191422194871006410481067288862);
42 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
43 DVK(KP692021471, +0.692021471630095869627814897002069140197260599);
44 DVK(KP356895867, +0.356895867892209443894399510021300583399127187);
45 {
46 INT i;
47 const R *xi;
48 R *xo;
49 xi = ii;
50 xo = io;
51 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
52 V T3, TH, Ts, TV, TW, Tt, Tu, TU, Ta, To, Th, Tp, TC, Tx, TK;
53 V TQ, TN, TR, T14, TZ, T1, T2;
54 T1 = LD(&(xi[0]), ivs, &(xi[0]));
55 T2 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
56 T3 = VSUB(T1, T2);
57 TH = VADD(T1, T2);
58 {
59 V T6, TI, T9, TJ, Tn, TP, Tk, TO, Tg, TM, Td, TL;
60 {
61 V T4, T5, Ti, Tj;
62 T4 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
63 T5 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
64 T6 = VSUB(T4, T5);
65 TI = VADD(T4, T5);
66 {
67 V T7, T8, Tl, Tm;
68 T7 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
69 T8 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
70 T9 = VSUB(T7, T8);
71 TJ = VADD(T7, T8);
72 Tl = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
73 Tm = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
74 Tn = VSUB(Tl, Tm);
75 TP = VADD(Tl, Tm);
76 }
77 Ti = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
78 Tj = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
79 Tk = VSUB(Ti, Tj);
80 TO = VADD(Ti, Tj);
81 {
82 V Te, Tf, Tb, Tc;
83 Te = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
84 Tf = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
85 Tg = VSUB(Te, Tf);
86 TM = VADD(Te, Tf);
87 Tb = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
88 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
89 Td = VSUB(Tb, Tc);
90 TL = VADD(Tb, Tc);
91 }
92 }
93 Ts = VSUB(Tk, Tn);
94 TV = VSUB(TP, TO);
95 TW = VSUB(TM, TL);
96 Tt = VSUB(Td, Tg);
97 Tu = VSUB(T6, T9);
98 TU = VSUB(TI, TJ);
99 Ta = VADD(T6, T9);
100 To = VADD(Tk, Tn);
101 Th = VADD(Td, Tg);
102 Tp = VFNMS(LDK(KP356895867), To, Th);
103 TC = VFNMS(LDK(KP356895867), Th, Ta);
104 Tx = VFNMS(LDK(KP356895867), Ta, To);
105 TK = VADD(TI, TJ);
106 TQ = VADD(TO, TP);
107 TN = VADD(TL, TM);
108 TR = VFNMS(LDK(KP356895867), TK, TQ);
109 T14 = VFNMS(LDK(KP356895867), TQ, TN);
110 TZ = VFNMS(LDK(KP356895867), TN, TK);
111 }
112 {
113 V T19, T1a, T1b, T1e, T1c, T1g, T1h;
114 T19 = VADD(T3, VADD(Ta, VADD(Th, To)));
115 STM2(&(xo[14]), T19, ovs, &(xo[2]));
116 T1a = VADD(TH, VADD(TK, VADD(TN, TQ)));
117 STM2(&(xo[0]), T1a, ovs, &(xo[0]));
118 {
119 V Tr, Tw, Tq, Tv;
120 Tq = VFNMS(LDK(KP692021471), Tp, Ta);
121 Tr = VFNMS(LDK(KP900968867), Tq, T3);
122 Tv = VFNMS(LDK(KP554958132), Tu, Tt);
123 Tw = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tv, Ts));
124 T1b = VFMAI(Tw, Tr);
125 STM2(&(xo[6]), T1b, ovs, &(xo[2]));
126 T1c = VFNMSI(Tw, Tr);
127 STM2(&(xo[22]), T1c, ovs, &(xo[2]));
128 }
129 {
130 V T16, T18, T15, T17, T1d;
131 T15 = VFNMS(LDK(KP692021471), T14, TK);
132 T16 = VFNMS(LDK(KP900968867), T15, TH);
133 T17 = VFMA(LDK(KP554958132), TU, TW);
134 T18 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T17, TV));
135 T1d = VFNMSI(T18, T16);
136 STM2(&(xo[20]), T1d, ovs, &(xo[0]));
137 STN2(&(xo[20]), T1d, T1c, ovs);
138 T1e = VFMAI(T18, T16);
139 STM2(&(xo[8]), T1e, ovs, &(xo[0]));
140 }
141 {
142 V Tz, TB, Ty, TA, T1f;
143 Ty = VFNMS(LDK(KP692021471), Tx, Th);
144 Tz = VFNMS(LDK(KP900968867), Ty, T3);
145 TA = VFMA(LDK(KP554958132), Tt, Ts);
146 TB = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), TA, Tu));
147 T1f = VFNMSI(TB, Tz);
148 STM2(&(xo[10]), T1f, ovs, &(xo[2]));
149 STN2(&(xo[8]), T1e, T1f, ovs);
150 T1g = VFMAI(TB, Tz);
151 STM2(&(xo[18]), T1g, ovs, &(xo[2]));
152 }
153 {
154 V TT, TY, TS, TX, T1i;
155 TS = VFNMS(LDK(KP692021471), TR, TN);
156 TT = VFNMS(LDK(KP900968867), TS, TH);
157 TX = VFMA(LDK(KP554958132), TW, TV);
158 TY = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TX, TU));
159 T1h = VFNMSI(TY, TT);
160 STM2(&(xo[24]), T1h, ovs, &(xo[0]));
161 T1i = VFMAI(TY, TT);
162 STM2(&(xo[4]), T1i, ovs, &(xo[0]));
163 STN2(&(xo[4]), T1i, T1b, ovs);
164 }
165 {
166 V T11, T13, T10, T12, T1j, T1k;
167 T10 = VFNMS(LDK(KP692021471), TZ, TQ);
168 T11 = VFNMS(LDK(KP900968867), T10, TH);
169 T12 = VFNMS(LDK(KP554958132), TV, TU);
170 T13 = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), T12, TW));
171 T1j = VFNMSI(T13, T11);
172 STM2(&(xo[16]), T1j, ovs, &(xo[0]));
173 STN2(&(xo[16]), T1j, T1g, ovs);
174 T1k = VFMAI(T13, T11);
175 STM2(&(xo[12]), T1k, ovs, &(xo[0]));
176 STN2(&(xo[12]), T1k, T19, ovs);
177 }
178 {
179 V TE, TG, TD, TF, T1l, T1m;
180 TD = VFNMS(LDK(KP692021471), TC, To);
181 TE = VFNMS(LDK(KP900968867), TD, T3);
182 TF = VFMA(LDK(KP554958132), Ts, Tu);
183 TG = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), TF, Tt));
184 T1l = VFMAI(TG, TE);
185 STM2(&(xo[2]), T1l, ovs, &(xo[2]));
186 STN2(&(xo[0]), T1a, T1l, ovs);
187 T1m = VFNMSI(TG, TE);
188 STM2(&(xo[26]), T1m, ovs, &(xo[2]));
189 STN2(&(xo[24]), T1h, T1m, ovs);
190 }
191 }
192 }
193 }
194 VLEAVE();
195 }
196
197 static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {32, 6, 42, 0}, &GENUS, 0, 2, 0, 0 };
198
199 void XSIMD(codelet_n2bv_14) (planner *p) {
200 X(kdft_register) (p, n2bv_14, &desc);
201 }
202
203 #else
204
205 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 14 -name n2bv_14 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
206
207 /*
208 * This function contains 74 FP additions, 36 FP multiplications,
209 * (or, 50 additions, 12 multiplications, 24 fused multiply/add),
210 * 41 stack variables, 6 constants, and 35 memory accesses
211 */
212 #include "dft/simd/n2b.h"
213
214 static void n2bv_14(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
215 {
216 DVK(KP900968867, +0.900968867902419126236102319507445051165919162);
217 DVK(KP222520933, +0.222520933956314404288902564496794759466355569);
218 DVK(KP623489801, +0.623489801858733530525004884004239810632274731);
219 DVK(KP781831482, +0.781831482468029808708444526674057750232334519);
220 DVK(KP974927912, +0.974927912181823607018131682993931217232785801);
221 DVK(KP433883739, +0.433883739117558120475768332848358754609990728);
222 {
223 INT i;
224 const R *xi;
225 R *xo;
226 xi = ii;
227 xo = io;
228 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(28, is), MAKE_VOLATILE_STRIDE(28, os)) {
229 V Tp, Ty, Tl, TL, Tq, TE, T7, TJ, Ts, TB, Te, TK, Tr, TH, Tn;
230 V To;
231 Tn = LD(&(xi[0]), ivs, &(xi[0]));
232 To = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
233 Tp = VSUB(Tn, To);
234 Ty = VADD(Tn, To);
235 {
236 V Th, TC, Tk, TD;
237 {
238 V Tf, Tg, Ti, Tj;
239 Tf = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
240 Tg = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
241 Th = VSUB(Tf, Tg);
242 TC = VADD(Tf, Tg);
243 Ti = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
244 Tj = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
245 Tk = VSUB(Ti, Tj);
246 TD = VADD(Ti, Tj);
247 }
248 Tl = VSUB(Th, Tk);
249 TL = VSUB(TD, TC);
250 Tq = VADD(Th, Tk);
251 TE = VADD(TC, TD);
252 }
253 {
254 V T3, Tz, T6, TA;
255 {
256 V T1, T2, T4, T5;
257 T1 = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
258 T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
259 T3 = VSUB(T1, T2);
260 Tz = VADD(T1, T2);
261 T4 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
262 T5 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
263 T6 = VSUB(T4, T5);
264 TA = VADD(T4, T5);
265 }
266 T7 = VSUB(T3, T6);
267 TJ = VSUB(Tz, TA);
268 Ts = VADD(T3, T6);
269 TB = VADD(Tz, TA);
270 }
271 {
272 V Ta, TF, Td, TG;
273 {
274 V T8, T9, Tb, Tc;
275 T8 = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
276 T9 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
277 Ta = VSUB(T8, T9);
278 TF = VADD(T8, T9);
279 Tb = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
280 Tc = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
281 Td = VSUB(Tb, Tc);
282 TG = VADD(Tb, Tc);
283 }
284 Te = VSUB(Ta, Td);
285 TK = VSUB(TG, TF);
286 Tr = VADD(Ta, Td);
287 TH = VADD(TF, TG);
288 }
289 {
290 V TR, TS, TU, TV;
291 TR = VADD(Tp, VADD(Ts, VADD(Tq, Tr)));
292 STM2(&(xo[14]), TR, ovs, &(xo[2]));
293 TS = VADD(Ty, VADD(TB, VADD(TE, TH)));
294 STM2(&(xo[0]), TS, ovs, &(xo[0]));
295 {
296 V TT, Tm, Tt, TQ, TP, TW;
297 Tm = VBYI(VFMA(LDK(KP433883739), T7, VFNMS(LDK(KP781831482), Tl, VMUL(LDK(KP974927912), Te))));
298 Tt = VFMA(LDK(KP623489801), Tq, VFNMS(LDK(KP222520933), Tr, VFNMS(LDK(KP900968867), Ts, Tp)));
299 TT = VADD(Tm, Tt);
300 STM2(&(xo[6]), TT, ovs, &(xo[2]));
301 TU = VSUB(Tt, Tm);
302 STM2(&(xo[22]), TU, ovs, &(xo[2]));
303 TQ = VBYI(VFMA(LDK(KP974927912), TJ, VFMA(LDK(KP433883739), TL, VMUL(LDK(KP781831482), TK))));
304 TP = VFMA(LDK(KP623489801), TH, VFNMS(LDK(KP900968867), TE, VFNMS(LDK(KP222520933), TB, Ty)));
305 TV = VSUB(TP, TQ);
306 STM2(&(xo[24]), TV, ovs, &(xo[0]));
307 TW = VADD(TP, TQ);
308 STM2(&(xo[4]), TW, ovs, &(xo[0]));
309 STN2(&(xo[4]), TW, TT, ovs);
310 }
311 {
312 V T10, TM, TI, TZ;
313 {
314 V Tu, Tv, TX, TY;
315 Tu = VBYI(VFMA(LDK(KP781831482), T7, VFMA(LDK(KP974927912), Tl, VMUL(LDK(KP433883739), Te))));
316 Tv = VFMA(LDK(KP623489801), Ts, VFNMS(LDK(KP900968867), Tr, VFNMS(LDK(KP222520933), Tq, Tp)));
317 TX = VADD(Tu, Tv);
318 STM2(&(xo[2]), TX, ovs, &(xo[2]));
319 STN2(&(xo[0]), TS, TX, ovs);
320 TY = VSUB(Tv, Tu);
321 STM2(&(xo[26]), TY, ovs, &(xo[2]));
322 STN2(&(xo[24]), TV, TY, ovs);
323 }
324 TM = VBYI(VFNMS(LDK(KP433883739), TK, VFNMS(LDK(KP974927912), TL, VMUL(LDK(KP781831482), TJ))));
325 TI = VFMA(LDK(KP623489801), TB, VFNMS(LDK(KP900968867), TH, VFNMS(LDK(KP222520933), TE, Ty)));
326 TZ = VSUB(TI, TM);
327 STM2(&(xo[12]), TZ, ovs, &(xo[0]));
328 STN2(&(xo[12]), TZ, TR, ovs);
329 T10 = VADD(TI, TM);
330 STM2(&(xo[16]), T10, ovs, &(xo[0]));
331 {
332 V T11, TO, TN, T12;
333 TO = VBYI(VFMA(LDK(KP433883739), TJ, VFNMS(LDK(KP974927912), TK, VMUL(LDK(KP781831482), TL))));
334 TN = VFMA(LDK(KP623489801), TE, VFNMS(LDK(KP222520933), TH, VFNMS(LDK(KP900968867), TB, Ty)));
335 T11 = VSUB(TN, TO);
336 STM2(&(xo[8]), T11, ovs, &(xo[0]));
337 T12 = VADD(TN, TO);
338 STM2(&(xo[20]), T12, ovs, &(xo[0]));
339 STN2(&(xo[20]), T12, TU, ovs);
340 {
341 V Tx, Tw, T13, T14;
342 Tx = VBYI(VFNMS(LDK(KP781831482), Te, VFNMS(LDK(KP433883739), Tl, VMUL(LDK(KP974927912), T7))));
343 Tw = VFMA(LDK(KP623489801), Tr, VFNMS(LDK(KP900968867), Tq, VFNMS(LDK(KP222520933), Ts, Tp)));
344 T13 = VSUB(Tw, Tx);
345 STM2(&(xo[10]), T13, ovs, &(xo[2]));
346 STN2(&(xo[8]), T11, T13, ovs);
347 T14 = VADD(Tx, Tw);
348 STM2(&(xo[18]), T14, ovs, &(xo[2]));
349 STN2(&(xo[16]), T10, T14, ovs);
350 }
351 }
352 }
353 }
354 }
355 }
356 VLEAVE();
357 }
358
359 static const kdft_desc desc = { 14, XSIMD_STRING("n2bv_14"), {50, 12, 24, 0}, &GENUS, 0, 2, 0, 0 };
360
361 void XSIMD(codelet_n2bv_14) (planner *p) {
362 X(kdft_register) (p, n2bv_14, &desc);
363 }
364
365 #endif