Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hf2_8.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 08:06:35 EDT 2018 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_hc2hc.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -dit -name hf2_8 -include rdft/scalar/hf.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 "rdft/scalar/hf.h"
36
37 static void hf2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
40 {
41 INT m;
42 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
43 E T2, T3, Tl, Tn, T5, T6, Tf, T7, Ts, Tb, To, Ti, TC, TG;
44 {
45 E T4, Tm, Tr, Ta, TB, TF;
46 T2 = W[0];
47 T3 = W[2];
48 T4 = T2 * T3;
49 Tl = W[4];
50 Tm = T2 * Tl;
51 Tn = W[5];
52 Tr = T2 * Tn;
53 T5 = W[1];
54 T6 = W[3];
55 Ta = T2 * T6;
56 Tf = FMA(T5, T6, T4);
57 T7 = FNMS(T5, T6, T4);
58 Ts = FNMS(T5, Tl, Tr);
59 Tb = FMA(T5, T3, Ta);
60 To = FMA(T5, Tn, Tm);
61 TB = Tf * Tl;
62 TF = Tf * Tn;
63 Ti = FNMS(T5, T3, Ta);
64 TC = FMA(Ti, Tn, TB);
65 TG = FNMS(Ti, Tl, TF);
66 }
67 {
68 E T1, T1s, Td, T1r, Tu, TY, Tk, TW, TN, TR, T18, T1a, T1c, T1d, TA;
69 E TI, T11, T13, T15, T16;
70 T1 = cr[0];
71 T1s = ci[0];
72 {
73 E T8, T9, Tc, T1q;
74 T8 = cr[WS(rs, 4)];
75 T9 = T7 * T8;
76 Tc = ci[WS(rs, 4)];
77 T1q = T7 * Tc;
78 Td = FMA(Tb, Tc, T9);
79 T1r = FNMS(Tb, T8, T1q);
80 }
81 {
82 E Tp, Tq, Tt, TX;
83 Tp = cr[WS(rs, 6)];
84 Tq = To * Tp;
85 Tt = ci[WS(rs, 6)];
86 TX = To * Tt;
87 Tu = FMA(Ts, Tt, Tq);
88 TY = FNMS(Ts, Tp, TX);
89 }
90 {
91 E Tg, Th, Tj, TV;
92 Tg = cr[WS(rs, 2)];
93 Th = Tf * Tg;
94 Tj = ci[WS(rs, 2)];
95 TV = Tf * Tj;
96 Tk = FMA(Ti, Tj, Th);
97 TW = FNMS(Ti, Tg, TV);
98 }
99 {
100 E TK, TL, TM, T19, TO, TP, TQ, T1b;
101 TK = cr[WS(rs, 7)];
102 TL = Tl * TK;
103 TM = ci[WS(rs, 7)];
104 T19 = Tl * TM;
105 TO = cr[WS(rs, 3)];
106 TP = T3 * TO;
107 TQ = ci[WS(rs, 3)];
108 T1b = T3 * TQ;
109 TN = FMA(Tn, TM, TL);
110 TR = FMA(T6, TQ, TP);
111 T18 = TN - TR;
112 T1a = FNMS(Tn, TK, T19);
113 T1c = FNMS(T6, TO, T1b);
114 T1d = T1a - T1c;
115 }
116 {
117 E Tx, Ty, Tz, T12, TD, TE, TH, T14;
118 Tx = cr[WS(rs, 1)];
119 Ty = T2 * Tx;
120 Tz = ci[WS(rs, 1)];
121 T12 = T2 * Tz;
122 TD = cr[WS(rs, 5)];
123 TE = TC * TD;
124 TH = ci[WS(rs, 5)];
125 T14 = TC * TH;
126 TA = FMA(T5, Tz, Ty);
127 TI = FMA(TG, TH, TE);
128 T11 = TA - TI;
129 T13 = FNMS(T5, Tx, T12);
130 T15 = FNMS(TG, TD, T14);
131 T16 = T13 - T15;
132 }
133 {
134 E T10, T1g, T1z, T1B, T1f, T1A, T1j, T1C;
135 {
136 E TU, TZ, T1x, T1y;
137 TU = T1 - Td;
138 TZ = TW - TY;
139 T10 = TU + TZ;
140 T1g = TU - TZ;
141 T1x = Tk - Tu;
142 T1y = T1s - T1r;
143 T1z = T1x + T1y;
144 T1B = T1y - T1x;
145 }
146 {
147 E T17, T1e, T1h, T1i;
148 T17 = T11 + T16;
149 T1e = T18 - T1d;
150 T1f = T17 + T1e;
151 T1A = T1e - T17;
152 T1h = T11 - T16;
153 T1i = T18 + T1d;
154 T1j = T1h + T1i;
155 T1C = T1i - T1h;
156 }
157 ci[WS(rs, 2)] = FNMS(KP707106781, T1f, T10);
158 cr[WS(rs, 5)] = FMS(KP707106781, T1C, T1B);
159 ci[WS(rs, 6)] = FMA(KP707106781, T1C, T1B);
160 cr[WS(rs, 1)] = FMA(KP707106781, T1f, T10);
161 cr[WS(rs, 3)] = FNMS(KP707106781, T1j, T1g);
162 cr[WS(rs, 7)] = FMS(KP707106781, T1A, T1z);
163 ci[WS(rs, 4)] = FMA(KP707106781, T1A, T1z);
164 ci[0] = FMA(KP707106781, T1j, T1g);
165 }
166 {
167 E Tw, T1k, T1u, T1w, TT, T1v, T1n, T1o;
168 {
169 E Te, Tv, T1p, T1t;
170 Te = T1 + Td;
171 Tv = Tk + Tu;
172 Tw = Te + Tv;
173 T1k = Te - Tv;
174 T1p = TW + TY;
175 T1t = T1r + T1s;
176 T1u = T1p + T1t;
177 T1w = T1t - T1p;
178 }
179 {
180 E TJ, TS, T1l, T1m;
181 TJ = TA + TI;
182 TS = TN + TR;
183 TT = TJ + TS;
184 T1v = TS - TJ;
185 T1l = T1a + T1c;
186 T1m = T13 + T15;
187 T1n = T1l - T1m;
188 T1o = T1m + T1l;
189 }
190 ci[WS(rs, 3)] = Tw - TT;
191 cr[WS(rs, 6)] = T1v - T1w;
192 ci[WS(rs, 5)] = T1v + T1w;
193 cr[0] = Tw + TT;
194 cr[WS(rs, 2)] = T1k - T1n;
195 cr[WS(rs, 4)] = T1o - T1u;
196 ci[WS(rs, 7)] = T1o + T1u;
197 ci[WS(rs, 1)] = T1k + T1n;
198 }
199 }
200 }
201 }
202 }
203
204 static const tw_instr twinstr[] = {
205 {TW_CEXP, 1, 1},
206 {TW_CEXP, 1, 3},
207 {TW_CEXP, 1, 7},
208 {TW_NEXT, 1, 0}
209 };
210
211 static const hc2hc_desc desc = { 8, "hf2_8", twinstr, &GENUS, {44, 20, 30, 0} };
212
213 void X(codelet_hf2_8) (planner *p) {
214 X(khc2hc_register) (p, hf2_8, &desc);
215 }
216 #else
217
218 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 8 -dit -name hf2_8 -include rdft/scalar/hf.h */
219
220 /*
221 * This function contains 74 FP additions, 44 FP multiplications,
222 * (or, 56 additions, 26 multiplications, 18 fused multiply/add),
223 * 42 stack variables, 1 constants, and 32 memory accesses
224 */
225 #include "rdft/scalar/hf.h"
226
227 static void hf2_8(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
228 {
229 DK(KP707106781, +0.707106781186547524400844362104849039284835938);
230 {
231 INT m;
232 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
233 E T2, T5, T3, T6, T8, Tc, Tg, Ti, Tl, Tm, Tn, Tz, Tp, Tx;
234 {
235 E T4, Tb, T7, Ta;
236 T2 = W[0];
237 T5 = W[1];
238 T3 = W[2];
239 T6 = W[3];
240 T4 = T2 * T3;
241 Tb = T5 * T3;
242 T7 = T5 * T6;
243 Ta = T2 * T6;
244 T8 = T4 - T7;
245 Tc = Ta + Tb;
246 Tg = T4 + T7;
247 Ti = Ta - Tb;
248 Tl = W[4];
249 Tm = W[5];
250 Tn = FMA(T2, Tl, T5 * Tm);
251 Tz = FNMS(Ti, Tl, Tg * Tm);
252 Tp = FNMS(T5, Tl, T2 * Tm);
253 Tx = FMA(Tg, Tl, Ti * Tm);
254 }
255 {
256 E Tf, T1j, TL, T1d, TJ, T16, TV, TY, Ts, T1i, TO, T1a, TC, T17, TQ;
257 E TT;
258 {
259 E T1, T1c, Te, T1b, T9, Td;
260 T1 = cr[0];
261 T1c = ci[0];
262 T9 = cr[WS(rs, 4)];
263 Td = ci[WS(rs, 4)];
264 Te = FMA(T8, T9, Tc * Td);
265 T1b = FNMS(Tc, T9, T8 * Td);
266 Tf = T1 + Te;
267 T1j = T1c - T1b;
268 TL = T1 - Te;
269 T1d = T1b + T1c;
270 }
271 {
272 E TF, TW, TI, TX;
273 {
274 E TD, TE, TG, TH;
275 TD = cr[WS(rs, 7)];
276 TE = ci[WS(rs, 7)];
277 TF = FMA(Tl, TD, Tm * TE);
278 TW = FNMS(Tm, TD, Tl * TE);
279 TG = cr[WS(rs, 3)];
280 TH = ci[WS(rs, 3)];
281 TI = FMA(T3, TG, T6 * TH);
282 TX = FNMS(T6, TG, T3 * TH);
283 }
284 TJ = TF + TI;
285 T16 = TW + TX;
286 TV = TF - TI;
287 TY = TW - TX;
288 }
289 {
290 E Tk, TM, Tr, TN;
291 {
292 E Th, Tj, To, Tq;
293 Th = cr[WS(rs, 2)];
294 Tj = ci[WS(rs, 2)];
295 Tk = FMA(Tg, Th, Ti * Tj);
296 TM = FNMS(Ti, Th, Tg * Tj);
297 To = cr[WS(rs, 6)];
298 Tq = ci[WS(rs, 6)];
299 Tr = FMA(Tn, To, Tp * Tq);
300 TN = FNMS(Tp, To, Tn * Tq);
301 }
302 Ts = Tk + Tr;
303 T1i = Tk - Tr;
304 TO = TM - TN;
305 T1a = TM + TN;
306 }
307 {
308 E Tw, TR, TB, TS;
309 {
310 E Tu, Tv, Ty, TA;
311 Tu = cr[WS(rs, 1)];
312 Tv = ci[WS(rs, 1)];
313 Tw = FMA(T2, Tu, T5 * Tv);
314 TR = FNMS(T5, Tu, T2 * Tv);
315 Ty = cr[WS(rs, 5)];
316 TA = ci[WS(rs, 5)];
317 TB = FMA(Tx, Ty, Tz * TA);
318 TS = FNMS(Tz, Ty, Tx * TA);
319 }
320 TC = Tw + TB;
321 T17 = TR + TS;
322 TQ = Tw - TB;
323 TT = TR - TS;
324 }
325 {
326 E Tt, TK, T1f, T1g;
327 Tt = Tf + Ts;
328 TK = TC + TJ;
329 ci[WS(rs, 3)] = Tt - TK;
330 cr[0] = Tt + TK;
331 T1f = TJ - TC;
332 T1g = T1d - T1a;
333 cr[WS(rs, 6)] = T1f - T1g;
334 ci[WS(rs, 5)] = T1f + T1g;
335 {
336 E T11, T1m, T14, T1l, T12, T13;
337 T11 = TL - TO;
338 T1m = T1j - T1i;
339 T12 = TQ - TT;
340 T13 = TV + TY;
341 T14 = KP707106781 * (T12 + T13);
342 T1l = KP707106781 * (T13 - T12);
343 cr[WS(rs, 3)] = T11 - T14;
344 ci[WS(rs, 6)] = T1l + T1m;
345 ci[0] = T11 + T14;
346 cr[WS(rs, 5)] = T1l - T1m;
347 }
348 }
349 {
350 E T19, T1e, T15, T18;
351 T19 = T17 + T16;
352 T1e = T1a + T1d;
353 cr[WS(rs, 4)] = T19 - T1e;
354 ci[WS(rs, 7)] = T19 + T1e;
355 T15 = Tf - Ts;
356 T18 = T16 - T17;
357 cr[WS(rs, 2)] = T15 - T18;
358 ci[WS(rs, 1)] = T15 + T18;
359 {
360 E TP, T1k, T10, T1h, TU, TZ;
361 TP = TL + TO;
362 T1k = T1i + T1j;
363 TU = TQ + TT;
364 TZ = TV - TY;
365 T10 = KP707106781 * (TU + TZ);
366 T1h = KP707106781 * (TZ - TU);
367 ci[WS(rs, 2)] = TP - T10;
368 ci[WS(rs, 4)] = T1h + T1k;
369 cr[WS(rs, 1)] = TP + T10;
370 cr[WS(rs, 7)] = T1h - T1k;
371 }
372 }
373 }
374 }
375 }
376 }
377
378 static const tw_instr twinstr[] = {
379 {TW_CEXP, 1, 1},
380 {TW_CEXP, 1, 3},
381 {TW_CEXP, 1, 7},
382 {TW_NEXT, 1, 0}
383 };
384
385 static const hc2hc_desc desc = { 8, "hf2_8", twinstr, &GENUS, {56, 26, 18, 0} };
386
387 void X(codelet_hf2_8) (planner *p) {
388 X(khc2hc_register) (p, hf2_8, &desc);
389 }
390 #endif