Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/n2bv_16.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 08:05:12 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 16 -name n2bv_16 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
29
30 /*
31 * This function contains 72 FP additions, 34 FP multiplications,
32 * (or, 38 additions, 0 multiplications, 34 fused multiply/add),
33 * 38 stack variables, 3 constants, and 40 memory accesses
34 */
35 #include "dft/simd/n2b.h"
36
37 static void n2bv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
38 {
39 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
40 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
41 DVK(KP414213562, +0.414213562373095048801688724209698078569671875);
42 {
43 INT i;
44 const R *xi;
45 R *xo;
46 xi = ii;
47 xo = io;
48 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
49 V T7, TU, Tz, TH, Tu, TV, TA, TK, Te, TX, TC, TO, Tl, TY, TD;
50 V TR;
51 {
52 V T1, T2, T3, T4, T5, T6;
53 T1 = LD(&(xi[0]), ivs, &(xi[0]));
54 T2 = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
55 T3 = VADD(T1, T2);
56 T4 = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
57 T5 = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
58 T6 = VADD(T4, T5);
59 T7 = VSUB(T3, T6);
60 TU = VSUB(T4, T5);
61 Tz = VADD(T3, T6);
62 TH = VSUB(T1, T2);
63 }
64 {
65 V Tq, TI, Tt, TJ;
66 {
67 V To, Tp, Tr, Ts;
68 To = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
69 Tp = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
70 Tq = VADD(To, Tp);
71 TI = VSUB(To, Tp);
72 Tr = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
73 Ts = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
74 Tt = VADD(Tr, Ts);
75 TJ = VSUB(Tr, Ts);
76 }
77 Tu = VSUB(Tq, Tt);
78 TV = VSUB(TI, TJ);
79 TA = VADD(Tq, Tt);
80 TK = VADD(TI, TJ);
81 }
82 {
83 V Ta, TM, Td, TN;
84 {
85 V T8, T9, Tb, Tc;
86 T8 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
87 T9 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
88 Ta = VADD(T8, T9);
89 TM = VSUB(T8, T9);
90 Tb = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
91 Tc = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
92 Td = VADD(Tb, Tc);
93 TN = VSUB(Tb, Tc);
94 }
95 Te = VSUB(Ta, Td);
96 TX = VFMA(LDK(KP414213562), TM, TN);
97 TC = VADD(Ta, Td);
98 TO = VFNMS(LDK(KP414213562), TN, TM);
99 }
100 {
101 V Th, TP, Tk, TQ;
102 {
103 V Tf, Tg, Ti, Tj;
104 Tf = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
105 Tg = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
106 Th = VADD(Tf, Tg);
107 TP = VSUB(Tf, Tg);
108 Ti = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
109 Tj = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
110 Tk = VADD(Ti, Tj);
111 TQ = VSUB(Tj, Ti);
112 }
113 Tl = VSUB(Th, Tk);
114 TY = VFMA(LDK(KP414213562), TP, TQ);
115 TD = VADD(Th, Tk);
116 TR = VFNMS(LDK(KP414213562), TQ, TP);
117 }
118 {
119 V T1b, T1c, T1d, T1e;
120 {
121 V TB, TE, TF, TG;
122 TB = VSUB(Tz, TA);
123 TE = VSUB(TC, TD);
124 T1b = VFNMSI(TE, TB);
125 STM2(&(xo[24]), T1b, ovs, &(xo[0]));
126 T1c = VFMAI(TE, TB);
127 STM2(&(xo[8]), T1c, ovs, &(xo[0]));
128 TF = VADD(Tz, TA);
129 TG = VADD(TC, TD);
130 T1d = VSUB(TF, TG);
131 STM2(&(xo[16]), T1d, ovs, &(xo[0]));
132 T1e = VADD(TF, TG);
133 STM2(&(xo[0]), T1e, ovs, &(xo[0]));
134 }
135 {
136 V T1f, T1g, T1h, T1i;
137 {
138 V Tn, Tx, Tw, Ty, Tm, Tv;
139 Tm = VADD(Te, Tl);
140 Tn = VFNMS(LDK(KP707106781), Tm, T7);
141 Tx = VFMA(LDK(KP707106781), Tm, T7);
142 Tv = VSUB(Te, Tl);
143 Tw = VFNMS(LDK(KP707106781), Tv, Tu);
144 Ty = VFMA(LDK(KP707106781), Tv, Tu);
145 T1f = VFNMSI(Tw, Tn);
146 STM2(&(xo[12]), T1f, ovs, &(xo[0]));
147 T1g = VFNMSI(Ty, Tx);
148 STM2(&(xo[28]), T1g, ovs, &(xo[0]));
149 T1h = VFMAI(Tw, Tn);
150 STM2(&(xo[20]), T1h, ovs, &(xo[0]));
151 T1i = VFMAI(Ty, Tx);
152 STM2(&(xo[4]), T1i, ovs, &(xo[0]));
153 }
154 {
155 V TT, T11, T10, T12;
156 {
157 V TL, TS, TW, TZ;
158 TL = VFMA(LDK(KP707106781), TK, TH);
159 TS = VADD(TO, TR);
160 TT = VFNMS(LDK(KP923879532), TS, TL);
161 T11 = VFMA(LDK(KP923879532), TS, TL);
162 TW = VFMA(LDK(KP707106781), TV, TU);
163 TZ = VSUB(TX, TY);
164 T10 = VFNMS(LDK(KP923879532), TZ, TW);
165 T12 = VFMA(LDK(KP923879532), TZ, TW);
166 }
167 {
168 V T1j, T1k, T1l, T1m;
169 T1j = VFNMSI(T10, TT);
170 STM2(&(xo[14]), T1j, ovs, &(xo[2]));
171 STN2(&(xo[12]), T1f, T1j, ovs);
172 T1k = VFMAI(T12, T11);
173 STM2(&(xo[2]), T1k, ovs, &(xo[2]));
174 STN2(&(xo[0]), T1e, T1k, ovs);
175 T1l = VFMAI(T10, TT);
176 STM2(&(xo[18]), T1l, ovs, &(xo[2]));
177 STN2(&(xo[16]), T1d, T1l, ovs);
178 T1m = VFNMSI(T12, T11);
179 STM2(&(xo[30]), T1m, ovs, &(xo[2]));
180 STN2(&(xo[28]), T1g, T1m, ovs);
181 }
182 }
183 {
184 V T15, T19, T18, T1a;
185 {
186 V T13, T14, T16, T17;
187 T13 = VFNMS(LDK(KP707106781), TK, TH);
188 T14 = VADD(TX, TY);
189 T15 = VFNMS(LDK(KP923879532), T14, T13);
190 T19 = VFMA(LDK(KP923879532), T14, T13);
191 T16 = VFNMS(LDK(KP707106781), TV, TU);
192 T17 = VSUB(TO, TR);
193 T18 = VFMA(LDK(KP923879532), T17, T16);
194 T1a = VFNMS(LDK(KP923879532), T17, T16);
195 }
196 {
197 V T1n, T1o, T1p, T1q;
198 T1n = VFMAI(T18, T15);
199 STM2(&(xo[10]), T1n, ovs, &(xo[2]));
200 STN2(&(xo[8]), T1c, T1n, ovs);
201 T1o = VFMAI(T1a, T19);
202 STM2(&(xo[26]), T1o, ovs, &(xo[2]));
203 STN2(&(xo[24]), T1b, T1o, ovs);
204 T1p = VFNMSI(T18, T15);
205 STM2(&(xo[22]), T1p, ovs, &(xo[2]));
206 STN2(&(xo[20]), T1h, T1p, ovs);
207 T1q = VFNMSI(T1a, T19);
208 STM2(&(xo[6]), T1q, ovs, &(xo[2]));
209 STN2(&(xo[4]), T1i, T1q, ovs);
210 }
211 }
212 }
213 }
214 }
215 }
216 VLEAVE();
217 }
218
219 static const kdft_desc desc = { 16, XSIMD_STRING("n2bv_16"), {38, 0, 34, 0}, &GENUS, 0, 2, 0, 0 };
220
221 void XSIMD(codelet_n2bv_16) (planner *p) {
222 X(kdft_register) (p, n2bv_16, &desc);
223 }
224
225 #else
226
227 /* Generated by: ../../../genfft/gen_notw_c.native -simd -compact -variables 4 -pipeline-latency 8 -sign 1 -n 16 -name n2bv_16 -with-ostride 2 -include dft/simd/n2b.h -store-multiple 2 */
228
229 /*
230 * This function contains 72 FP additions, 12 FP multiplications,
231 * (or, 68 additions, 8 multiplications, 4 fused multiply/add),
232 * 38 stack variables, 3 constants, and 40 memory accesses
233 */
234 #include "dft/simd/n2b.h"
235
236 static void n2bv_16(const R *ri, const R *ii, R *ro, R *io, stride is, stride os, INT v, INT ivs, INT ovs)
237 {
238 DVK(KP382683432, +0.382683432365089771728459984030398866761344562);
239 DVK(KP923879532, +0.923879532511286756128183189396788286822416626);
240 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
241 {
242 INT i;
243 const R *xi;
244 R *xo;
245 xi = ii;
246 xo = io;
247 for (i = v; i > 0; i = i - VL, xi = xi + (VL * ivs), xo = xo + (VL * ovs), MAKE_VOLATILE_STRIDE(32, is), MAKE_VOLATILE_STRIDE(32, os)) {
248 V Tp, T13, Tu, TY, Tm, T14, Tv, TU, T7, T16, Tx, TN, Te, T17, Ty;
249 V TQ;
250 {
251 V Tn, To, TX, Ts, Tt, TW;
252 Tn = LD(&(xi[WS(is, 4)]), ivs, &(xi[0]));
253 To = LD(&(xi[WS(is, 12)]), ivs, &(xi[0]));
254 TX = VADD(Tn, To);
255 Ts = LD(&(xi[0]), ivs, &(xi[0]));
256 Tt = LD(&(xi[WS(is, 8)]), ivs, &(xi[0]));
257 TW = VADD(Ts, Tt);
258 Tp = VSUB(Tn, To);
259 T13 = VADD(TW, TX);
260 Tu = VSUB(Ts, Tt);
261 TY = VSUB(TW, TX);
262 }
263 {
264 V Ti, TS, Tl, TT;
265 {
266 V Tg, Th, Tj, Tk;
267 Tg = LD(&(xi[WS(is, 2)]), ivs, &(xi[0]));
268 Th = LD(&(xi[WS(is, 10)]), ivs, &(xi[0]));
269 Ti = VSUB(Tg, Th);
270 TS = VADD(Tg, Th);
271 Tj = LD(&(xi[WS(is, 14)]), ivs, &(xi[0]));
272 Tk = LD(&(xi[WS(is, 6)]), ivs, &(xi[0]));
273 Tl = VSUB(Tj, Tk);
274 TT = VADD(Tj, Tk);
275 }
276 Tm = VMUL(LDK(KP707106781), VSUB(Ti, Tl));
277 T14 = VADD(TS, TT);
278 Tv = VMUL(LDK(KP707106781), VADD(Ti, Tl));
279 TU = VSUB(TS, TT);
280 }
281 {
282 V T3, TL, T6, TM;
283 {
284 V T1, T2, T4, T5;
285 T1 = LD(&(xi[WS(is, 1)]), ivs, &(xi[WS(is, 1)]));
286 T2 = LD(&(xi[WS(is, 9)]), ivs, &(xi[WS(is, 1)]));
287 T3 = VSUB(T1, T2);
288 TL = VADD(T1, T2);
289 T4 = LD(&(xi[WS(is, 5)]), ivs, &(xi[WS(is, 1)]));
290 T5 = LD(&(xi[WS(is, 13)]), ivs, &(xi[WS(is, 1)]));
291 T6 = VSUB(T4, T5);
292 TM = VADD(T4, T5);
293 }
294 T7 = VFNMS(LDK(KP382683432), T6, VMUL(LDK(KP923879532), T3));
295 T16 = VADD(TL, TM);
296 Tx = VFMA(LDK(KP382683432), T3, VMUL(LDK(KP923879532), T6));
297 TN = VSUB(TL, TM);
298 }
299 {
300 V Ta, TO, Td, TP;
301 {
302 V T8, T9, Tb, Tc;
303 T8 = LD(&(xi[WS(is, 15)]), ivs, &(xi[WS(is, 1)]));
304 T9 = LD(&(xi[WS(is, 7)]), ivs, &(xi[WS(is, 1)]));
305 Ta = VSUB(T8, T9);
306 TO = VADD(T8, T9);
307 Tb = LD(&(xi[WS(is, 3)]), ivs, &(xi[WS(is, 1)]));
308 Tc = LD(&(xi[WS(is, 11)]), ivs, &(xi[WS(is, 1)]));
309 Td = VSUB(Tb, Tc);
310 TP = VADD(Tb, Tc);
311 }
312 Te = VFMA(LDK(KP923879532), Ta, VMUL(LDK(KP382683432), Td));
313 T17 = VADD(TO, TP);
314 Ty = VFNMS(LDK(KP382683432), Ta, VMUL(LDK(KP923879532), Td));
315 TQ = VSUB(TO, TP);
316 }
317 {
318 V T1b, T1c, T1d, T1e;
319 {
320 V T15, T18, T19, T1a;
321 T15 = VSUB(T13, T14);
322 T18 = VBYI(VSUB(T16, T17));
323 T1b = VSUB(T15, T18);
324 STM2(&(xo[24]), T1b, ovs, &(xo[0]));
325 T1c = VADD(T15, T18);
326 STM2(&(xo[8]), T1c, ovs, &(xo[0]));
327 T19 = VADD(T13, T14);
328 T1a = VADD(T16, T17);
329 T1d = VSUB(T19, T1a);
330 STM2(&(xo[16]), T1d, ovs, &(xo[0]));
331 T1e = VADD(T19, T1a);
332 STM2(&(xo[0]), T1e, ovs, &(xo[0]));
333 }
334 {
335 V T1f, T1g, T1h, T1i;
336 {
337 V TV, T11, T10, T12, TR, TZ;
338 TR = VMUL(LDK(KP707106781), VSUB(TN, TQ));
339 TV = VBYI(VSUB(TR, TU));
340 T11 = VBYI(VADD(TU, TR));
341 TZ = VMUL(LDK(KP707106781), VADD(TN, TQ));
342 T10 = VSUB(TY, TZ);
343 T12 = VADD(TY, TZ);
344 T1f = VADD(TV, T10);
345 STM2(&(xo[12]), T1f, ovs, &(xo[0]));
346 T1g = VSUB(T12, T11);
347 STM2(&(xo[28]), T1g, ovs, &(xo[0]));
348 T1h = VSUB(T10, TV);
349 STM2(&(xo[20]), T1h, ovs, &(xo[0]));
350 T1i = VADD(T11, T12);
351 STM2(&(xo[4]), T1i, ovs, &(xo[0]));
352 }
353 {
354 V Tr, TB, TA, TC;
355 {
356 V Tf, Tq, Tw, Tz;
357 Tf = VSUB(T7, Te);
358 Tq = VSUB(Tm, Tp);
359 Tr = VBYI(VSUB(Tf, Tq));
360 TB = VBYI(VADD(Tq, Tf));
361 Tw = VSUB(Tu, Tv);
362 Tz = VSUB(Tx, Ty);
363 TA = VSUB(Tw, Tz);
364 TC = VADD(Tw, Tz);
365 }
366 {
367 V T1j, T1k, T1l, T1m;
368 T1j = VADD(Tr, TA);
369 STM2(&(xo[10]), T1j, ovs, &(xo[2]));
370 STN2(&(xo[8]), T1c, T1j, ovs);
371 T1k = VSUB(TC, TB);
372 STM2(&(xo[26]), T1k, ovs, &(xo[2]));
373 STN2(&(xo[24]), T1b, T1k, ovs);
374 T1l = VSUB(TA, Tr);
375 STM2(&(xo[22]), T1l, ovs, &(xo[2]));
376 STN2(&(xo[20]), T1h, T1l, ovs);
377 T1m = VADD(TB, TC);
378 STM2(&(xo[6]), T1m, ovs, &(xo[2]));
379 STN2(&(xo[4]), T1i, T1m, ovs);
380 }
381 }
382 {
383 V TF, TJ, TI, TK;
384 {
385 V TD, TE, TG, TH;
386 TD = VADD(Tu, Tv);
387 TE = VADD(T7, Te);
388 TF = VADD(TD, TE);
389 TJ = VSUB(TD, TE);
390 TG = VADD(Tp, Tm);
391 TH = VADD(Tx, Ty);
392 TI = VBYI(VADD(TG, TH));
393 TK = VBYI(VSUB(TH, TG));
394 }
395 {
396 V T1n, T1o, T1p, T1q;
397 T1n = VSUB(TF, TI);
398 STM2(&(xo[30]), T1n, ovs, &(xo[2]));
399 STN2(&(xo[28]), T1g, T1n, ovs);
400 T1o = VADD(TJ, TK);
401 STM2(&(xo[14]), T1o, ovs, &(xo[2]));
402 STN2(&(xo[12]), T1f, T1o, ovs);
403 T1p = VADD(TF, TI);
404 STM2(&(xo[2]), T1p, ovs, &(xo[2]));
405 STN2(&(xo[0]), T1e, T1p, ovs);
406 T1q = VSUB(TJ, TK);
407 STM2(&(xo[18]), T1q, ovs, &(xo[2]));
408 STN2(&(xo[16]), T1d, T1q, ovs);
409 }
410 }
411 }
412 }
413 }
414 }
415 VLEAVE();
416 }
417
418 static const kdft_desc desc = { 16, XSIMD_STRING("n2bv_16"), {68, 8, 4, 0}, &GENUS, 0, 2, 0, 0 };
419
420 void XSIMD(codelet_n2bv_16) (planner *p) {
421 X(kdft_register) (p, n2bv_16, &desc);
422 }
423
424 #endif