Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/simd/common/t2sv_8.c @ 82:d0c2a83c1364

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
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
comparison
equal deleted inserted replaced
81:7029a4916348 82:d0c2a83c1364
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 08:06:11 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_twiddle.native -fma -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include dft/simd/ts.h */
29
30 /*
31 * This function contains 74 FP additions, 50 FP multiplications,
32 * (or, 44 additions, 20 multiplications, 30 fused multiply/add),
33 * 48 stack variables, 1 constants, and 32 memory accesses
34 */
35 #include "dft/simd/ts.h"
36
37 static void t2sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
43 V T2, T3, Tl, Tn, T5, T6, Tf, T7, Ts, Tb, To, Ti, TC, TG;
44 {
45 V T4, Tm, Tr, Ta, TB, TF;
46 T2 = LDW(&(W[0]));
47 T3 = LDW(&(W[TWVL * 2]));
48 T4 = VMUL(T2, T3);
49 Tl = LDW(&(W[TWVL * 4]));
50 Tm = VMUL(T2, Tl);
51 Tn = LDW(&(W[TWVL * 5]));
52 Tr = VMUL(T2, Tn);
53 T5 = LDW(&(W[TWVL * 1]));
54 T6 = LDW(&(W[TWVL * 3]));
55 Ta = VMUL(T2, T6);
56 Tf = VFMA(T5, T6, T4);
57 T7 = VFNMS(T5, T6, T4);
58 Ts = VFNMS(T5, Tl, Tr);
59 Tb = VFMA(T5, T3, Ta);
60 To = VFMA(T5, Tn, Tm);
61 TB = VMUL(Tf, Tl);
62 TF = VMUL(Tf, Tn);
63 Ti = VFNMS(T5, T3, Ta);
64 TC = VFMA(Ti, Tn, TB);
65 TG = VFNMS(Ti, Tl, TF);
66 }
67 {
68 V T1, T1s, Td, T1r, Tu, TY, Tk, TW, TN, TR, T18, T1a, T1c, T1d, TA;
69 V TI, T11, T13, T15, T16;
70 T1 = LD(&(ri[0]), ms, &(ri[0]));
71 T1s = LD(&(ii[0]), ms, &(ii[0]));
72 {
73 V T8, T9, Tc, T1q;
74 T8 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
75 T9 = VMUL(T7, T8);
76 Tc = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
77 T1q = VMUL(T7, Tc);
78 Td = VFMA(Tb, Tc, T9);
79 T1r = VFNMS(Tb, T8, T1q);
80 }
81 {
82 V Tp, Tq, Tt, TX;
83 Tp = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
84 Tq = VMUL(To, Tp);
85 Tt = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
86 TX = VMUL(To, Tt);
87 Tu = VFMA(Ts, Tt, Tq);
88 TY = VFNMS(Ts, Tp, TX);
89 }
90 {
91 V Tg, Th, Tj, TV;
92 Tg = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
93 Th = VMUL(Tf, Tg);
94 Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
95 TV = VMUL(Tf, Tj);
96 Tk = VFMA(Ti, Tj, Th);
97 TW = VFNMS(Ti, Tg, TV);
98 }
99 {
100 V TK, TL, TM, T19, TO, TP, TQ, T1b;
101 TK = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
102 TL = VMUL(Tl, TK);
103 TM = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
104 T19 = VMUL(Tl, TM);
105 TO = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
106 TP = VMUL(T3, TO);
107 TQ = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
108 T1b = VMUL(T3, TQ);
109 TN = VFMA(Tn, TM, TL);
110 TR = VFMA(T6, TQ, TP);
111 T18 = VSUB(TN, TR);
112 T1a = VFNMS(Tn, TK, T19);
113 T1c = VFNMS(T6, TO, T1b);
114 T1d = VSUB(T1a, T1c);
115 }
116 {
117 V Tx, Ty, Tz, T12, TD, TE, TH, T14;
118 Tx = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
119 Ty = VMUL(T2, Tx);
120 Tz = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
121 T12 = VMUL(T2, Tz);
122 TD = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
123 TE = VMUL(TC, TD);
124 TH = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
125 T14 = VMUL(TC, TH);
126 TA = VFMA(T5, Tz, Ty);
127 TI = VFMA(TG, TH, TE);
128 T11 = VSUB(TA, TI);
129 T13 = VFNMS(T5, Tx, T12);
130 T15 = VFNMS(TG, TD, T14);
131 T16 = VSUB(T13, T15);
132 }
133 {
134 V T10, T1g, T1z, T1B, T1f, T1C, T1j, T1A;
135 {
136 V TU, TZ, T1x, T1y;
137 TU = VSUB(T1, Td);
138 TZ = VSUB(TW, TY);
139 T10 = VADD(TU, TZ);
140 T1g = VSUB(TU, TZ);
141 T1x = VSUB(T1s, T1r);
142 T1y = VSUB(Tk, Tu);
143 T1z = VSUB(T1x, T1y);
144 T1B = VADD(T1y, T1x);
145 }
146 {
147 V T17, T1e, T1h, T1i;
148 T17 = VADD(T11, T16);
149 T1e = VSUB(T18, T1d);
150 T1f = VADD(T17, T1e);
151 T1C = VSUB(T1e, T17);
152 T1h = VSUB(T16, T11);
153 T1i = VADD(T18, T1d);
154 T1j = VSUB(T1h, T1i);
155 T1A = VADD(T1h, T1i);
156 }
157 ST(&(ri[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
158 ST(&(ii[WS(rs, 5)]), VFNMS(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
159 ST(&(ri[WS(rs, 1)]), VFMA(LDK(KP707106781), T1f, T10), ms, &(ri[WS(rs, 1)]));
160 ST(&(ii[WS(rs, 1)]), VFMA(LDK(KP707106781), T1A, T1z), ms, &(ii[WS(rs, 1)]));
161 ST(&(ri[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
162 ST(&(ii[WS(rs, 7)]), VFNMS(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
163 ST(&(ri[WS(rs, 3)]), VFMA(LDK(KP707106781), T1j, T1g), ms, &(ri[WS(rs, 1)]));
164 ST(&(ii[WS(rs, 3)]), VFMA(LDK(KP707106781), T1C, T1B), ms, &(ii[WS(rs, 1)]));
165 }
166 {
167 V Tw, T1k, T1u, T1w, TT, T1v, T1n, T1o;
168 {
169 V Te, Tv, T1p, T1t;
170 Te = VADD(T1, Td);
171 Tv = VADD(Tk, Tu);
172 Tw = VADD(Te, Tv);
173 T1k = VSUB(Te, Tv);
174 T1p = VADD(TW, TY);
175 T1t = VADD(T1r, T1s);
176 T1u = VADD(T1p, T1t);
177 T1w = VSUB(T1t, T1p);
178 }
179 {
180 V TJ, TS, T1l, T1m;
181 TJ = VADD(TA, TI);
182 TS = VADD(TN, TR);
183 TT = VADD(TJ, TS);
184 T1v = VSUB(TS, TJ);
185 T1l = VADD(T13, T15);
186 T1m = VADD(T1a, T1c);
187 T1n = VSUB(T1l, T1m);
188 T1o = VADD(T1l, T1m);
189 }
190 ST(&(ri[WS(rs, 4)]), VSUB(Tw, TT), ms, &(ri[0]));
191 ST(&(ii[WS(rs, 4)]), VSUB(T1u, T1o), ms, &(ii[0]));
192 ST(&(ri[0]), VADD(Tw, TT), ms, &(ri[0]));
193 ST(&(ii[0]), VADD(T1o, T1u), ms, &(ii[0]));
194 ST(&(ri[WS(rs, 6)]), VSUB(T1k, T1n), ms, &(ri[0]));
195 ST(&(ii[WS(rs, 6)]), VSUB(T1w, T1v), ms, &(ii[0]));
196 ST(&(ri[WS(rs, 2)]), VADD(T1k, T1n), ms, &(ri[0]));
197 ST(&(ii[WS(rs, 2)]), VADD(T1v, T1w), ms, &(ii[0]));
198 }
199 }
200 }
201 }
202 VLEAVE();
203 }
204
205 static const tw_instr twinstr[] = {
206 VTW(0, 1),
207 VTW(0, 3),
208 VTW(0, 7),
209 {TW_NEXT, (2 * VL), 0}
210 };
211
212 static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {44, 20, 30, 0}, 0, 0, 0 };
213
214 void XSIMD(codelet_t2sv_8) (planner *p) {
215 X(kdft_dit_register) (p, t2sv_8, &desc);
216 }
217 #else
218
219 /* Generated by: ../../../genfft/gen_twiddle.native -simd -compact -variables 4 -pipeline-latency 8 -twiddle-log3 -precompute-twiddles -n 8 -name t2sv_8 -include dft/simd/ts.h */
220
221 /*
222 * This function contains 74 FP additions, 44 FP multiplications,
223 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
224 * 42 stack variables, 1 constants, and 32 memory accesses
225 */
226 #include "dft/simd/ts.h"
227
228 static void t2sv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
229 {
230 DVK(KP707106781, +0.707106781186547524400844362104849039284835938);
231 {
232 INT m;
233 for (m = mb, W = W + (mb * 6); m < me; m = m + (2 * VL), ri = ri + ((2 * VL) * ms), ii = ii + ((2 * VL) * ms), W = W + ((2 * VL) * 6), MAKE_VOLATILE_STRIDE(16, rs)) {
234 V T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
235 {
236 V T4, Tb, T7, Ta;
237 T2 = LDW(&(W[0]));
238 T5 = LDW(&(W[TWVL * 1]));
239 T3 = LDW(&(W[TWVL * 2]));
240 T6 = LDW(&(W[TWVL * 3]));
241 T4 = VMUL(T2, T3);
242 Tb = VMUL(T5, T3);
243 T7 = VMUL(T5, T6);
244 Ta = VMUL(T2, T6);
245 T8 = VSUB(T4, T7);
246 Tc = VADD(Ta, Tb);
247 Tg = VADD(T4, T7);
248 Ti = VSUB(Ta, Tb);
249 Tl = LDW(&(W[TWVL * 4]));
250 Tm = LDW(&(W[TWVL * 5]));
251 Tn = VFMA(T2, Tl, VMUL(T5, Tm));
252 Tz = VFNMS(Ti, Tl, VMUL(Tg, Tm));
253 Tp = VFNMS(T5, Tl, VMUL(T2, Tm));
254 Tx = VFMA(Tg, Tl, VMUL(Ti, Tm));
255 }
256 {
257 V Tf, T1i, TL, T1d, TJ, T17, TV, TY, Ts, T1j, TO, T1a, TC, T16, TQ;
258 V TT;
259 {
260 V T1, T1c, Te, T1b, T9, Td;
261 T1 = LD(&(ri[0]), ms, &(ri[0]));
262 T1c = LD(&(ii[0]), ms, &(ii[0]));
263 T9 = LD(&(ri[WS(rs, 4)]), ms, &(ri[0]));
264 Td = LD(&(ii[WS(rs, 4)]), ms, &(ii[0]));
265 Te = VFMA(T8, T9, VMUL(Tc, Td));
266 T1b = VFNMS(Tc, T9, VMUL(T8, Td));
267 Tf = VADD(T1, Te);
268 T1i = VSUB(T1c, T1b);
269 TL = VSUB(T1, Te);
270 T1d = VADD(T1b, T1c);
271 }
272 {
273 V TF, TW, TI, TX;
274 {
275 V TD, TE, TG, TH;
276 TD = LD(&(ri[WS(rs, 7)]), ms, &(ri[WS(rs, 1)]));
277 TE = LD(&(ii[WS(rs, 7)]), ms, &(ii[WS(rs, 1)]));
278 TF = VFMA(Tl, TD, VMUL(Tm, TE));
279 TW = VFNMS(Tm, TD, VMUL(Tl, TE));
280 TG = LD(&(ri[WS(rs, 3)]), ms, &(ri[WS(rs, 1)]));
281 TH = LD(&(ii[WS(rs, 3)]), ms, &(ii[WS(rs, 1)]));
282 TI = VFMA(T3, TG, VMUL(T6, TH));
283 TX = VFNMS(T6, TG, VMUL(T3, TH));
284 }
285 TJ = VADD(TF, TI);
286 T17 = VADD(TW, TX);
287 TV = VSUB(TF, TI);
288 TY = VSUB(TW, TX);
289 }
290 {
291 V Tk, TM, Tr, TN;
292 {
293 V Th, Tj, To, Tq;
294 Th = LD(&(ri[WS(rs, 2)]), ms, &(ri[0]));
295 Tj = LD(&(ii[WS(rs, 2)]), ms, &(ii[0]));
296 Tk = VFMA(Tg, Th, VMUL(Ti, Tj));
297 TM = VFNMS(Ti, Th, VMUL(Tg, Tj));
298 To = LD(&(ri[WS(rs, 6)]), ms, &(ri[0]));
299 Tq = LD(&(ii[WS(rs, 6)]), ms, &(ii[0]));
300 Tr = VFMA(Tn, To, VMUL(Tp, Tq));
301 TN = VFNMS(Tp, To, VMUL(Tn, Tq));
302 }
303 Ts = VADD(Tk, Tr);
304 T1j = VSUB(Tk, Tr);
305 TO = VSUB(TM, TN);
306 T1a = VADD(TM, TN);
307 }
308 {
309 V Tw, TR, TB, TS;
310 {
311 V Tu, Tv, Ty, TA;
312 Tu = LD(&(ri[WS(rs, 1)]), ms, &(ri[WS(rs, 1)]));
313 Tv = LD(&(ii[WS(rs, 1)]), ms, &(ii[WS(rs, 1)]));
314 Tw = VFMA(T2, Tu, VMUL(T5, Tv));
315 TR = VFNMS(T5, Tu, VMUL(T2, Tv));
316 Ty = LD(&(ri[WS(rs, 5)]), ms, &(ri[WS(rs, 1)]));
317 TA = LD(&(ii[WS(rs, 5)]), ms, &(ii[WS(rs, 1)]));
318 TB = VFMA(Tx, Ty, VMUL(Tz, TA));
319 TS = VFNMS(Tz, Ty, VMUL(Tx, TA));
320 }
321 TC = VADD(Tw, TB);
322 T16 = VADD(TR, TS);
323 TQ = VSUB(Tw, TB);
324 TT = VSUB(TR, TS);
325 }
326 {
327 V Tt, TK, T1f, T1g;
328 Tt = VADD(Tf, Ts);
329 TK = VADD(TC, TJ);
330 ST(&(ri[WS(rs, 4)]), VSUB(Tt, TK), ms, &(ri[0]));
331 ST(&(ri[0]), VADD(Tt, TK), ms, &(ri[0]));
332 {
333 V T19, T1e, T15, T18;
334 T19 = VADD(T16, T17);
335 T1e = VADD(T1a, T1d);
336 ST(&(ii[0]), VADD(T19, T1e), ms, &(ii[0]));
337 ST(&(ii[WS(rs, 4)]), VSUB(T1e, T19), ms, &(ii[0]));
338 T15 = VSUB(Tf, Ts);
339 T18 = VSUB(T16, T17);
340 ST(&(ri[WS(rs, 6)]), VSUB(T15, T18), ms, &(ri[0]));
341 ST(&(ri[WS(rs, 2)]), VADD(T15, T18), ms, &(ri[0]));
342 }
343 T1f = VSUB(TJ, TC);
344 T1g = VSUB(T1d, T1a);
345 ST(&(ii[WS(rs, 2)]), VADD(T1f, T1g), ms, &(ii[0]));
346 ST(&(ii[WS(rs, 6)]), VSUB(T1g, T1f), ms, &(ii[0]));
347 {
348 V T11, T1k, T14, T1h, T12, T13;
349 T11 = VSUB(TL, TO);
350 T1k = VSUB(T1i, T1j);
351 T12 = VSUB(TT, TQ);
352 T13 = VADD(TV, TY);
353 T14 = VMUL(LDK(KP707106781), VSUB(T12, T13));
354 T1h = VMUL(LDK(KP707106781), VADD(T12, T13));
355 ST(&(ri[WS(rs, 7)]), VSUB(T11, T14), ms, &(ri[WS(rs, 1)]));
356 ST(&(ii[WS(rs, 5)]), VSUB(T1k, T1h), ms, &(ii[WS(rs, 1)]));
357 ST(&(ri[WS(rs, 3)]), VADD(T11, T14), ms, &(ri[WS(rs, 1)]));
358 ST(&(ii[WS(rs, 1)]), VADD(T1h, T1k), ms, &(ii[WS(rs, 1)]));
359 }
360 {
361 V TP, T1m, T10, T1l, TU, TZ;
362 TP = VADD(TL, TO);
363 T1m = VADD(T1j, T1i);
364 TU = VADD(TQ, TT);
365 TZ = VSUB(TV, TY);
366 T10 = VMUL(LDK(KP707106781), VADD(TU, TZ));
367 T1l = VMUL(LDK(KP707106781), VSUB(TZ, TU));
368 ST(&(ri[WS(rs, 5)]), VSUB(TP, T10), ms, &(ri[WS(rs, 1)]));
369 ST(&(ii[WS(rs, 7)]), VSUB(T1m, T1l), ms, &(ii[WS(rs, 1)]));
370 ST(&(ri[WS(rs, 1)]), VADD(TP, T10), ms, &(ri[WS(rs, 1)]));
371 ST(&(ii[WS(rs, 3)]), VADD(T1l, T1m), ms, &(ii[WS(rs, 1)]));
372 }
373 }
374 }
375 }
376 }
377 VLEAVE();
378 }
379
380 static const tw_instr twinstr[] = {
381 VTW(0, 1),
382 VTW(0, 3),
383 VTW(0, 7),
384 {TW_NEXT, (2 * VL), 0}
385 };
386
387 static const ct_desc desc = { 8, XSIMD_STRING("t2sv_8"), twinstr, &GENUS, {56, 26, 18, 0}, 0, 0, 0 };
388
389 void XSIMD(codelet_t2sv_8) (planner *p) {
390 X(kdft_dit_register) (p, t2sv_8, &desc);
391 }
392 #endif