Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/q1_3.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 08:04:30 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_twidsq.native -fma -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 3 -name q1_3 -include dft/scalar/q.h */
29
30 /*
31 * This function contains 48 FP additions, 42 FP multiplications,
32 * (or, 18 additions, 12 multiplications, 30 fused multiply/add),
33 * 35 stack variables, 2 constants, and 36 memory accesses
34 */
35 #include "dft/scalar/q.h"
36
37 static void q1_3(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
38 {
39 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
40 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
41 {
42 INT m;
43 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
44 E T1, T4, T6, Tg, Td, Te, T9, Tf, Tp, Ts, Tu, TE, TB, TC, Tx;
45 E TD, TZ, T10, TV, T11, TN, TQ, TS, T12;
46 {
47 E T2, T3, Tv, Tw;
48 T1 = rio[0];
49 T2 = rio[WS(rs, 1)];
50 T3 = rio[WS(rs, 2)];
51 T4 = T2 + T3;
52 T6 = FNMS(KP500000000, T4, T1);
53 Tg = T3 - T2;
54 {
55 E T7, T8, Tq, Tr;
56 Td = iio[0];
57 T7 = iio[WS(rs, 1)];
58 T8 = iio[WS(rs, 2)];
59 Te = T7 + T8;
60 T9 = T7 - T8;
61 Tf = FNMS(KP500000000, Te, Td);
62 Tp = rio[WS(vs, 1)];
63 Tq = rio[WS(vs, 1) + WS(rs, 1)];
64 Tr = rio[WS(vs, 1) + WS(rs, 2)];
65 Ts = Tq + Tr;
66 Tu = FNMS(KP500000000, Ts, Tp);
67 TE = Tr - Tq;
68 }
69 TB = iio[WS(vs, 1)];
70 Tv = iio[WS(vs, 1) + WS(rs, 1)];
71 Tw = iio[WS(vs, 1) + WS(rs, 2)];
72 TC = Tv + Tw;
73 Tx = Tv - Tw;
74 TD = FNMS(KP500000000, TC, TB);
75 {
76 E TT, TU, TO, TP;
77 TZ = iio[WS(vs, 2)];
78 TT = iio[WS(vs, 2) + WS(rs, 1)];
79 TU = iio[WS(vs, 2) + WS(rs, 2)];
80 T10 = TT + TU;
81 TV = TT - TU;
82 T11 = FNMS(KP500000000, T10, TZ);
83 TN = rio[WS(vs, 2)];
84 TO = rio[WS(vs, 2) + WS(rs, 1)];
85 TP = rio[WS(vs, 2) + WS(rs, 2)];
86 TQ = TO + TP;
87 TS = FNMS(KP500000000, TQ, TN);
88 T12 = TP - TO;
89 }
90 }
91 rio[0] = T1 + T4;
92 iio[0] = Td + Te;
93 rio[WS(rs, 1)] = Tp + Ts;
94 iio[WS(rs, 1)] = TB + TC;
95 iio[WS(rs, 2)] = TZ + T10;
96 rio[WS(rs, 2)] = TN + TQ;
97 {
98 E Ta, Th, Tb, Ti, T5, Tc;
99 Ta = FMA(KP866025403, T9, T6);
100 Th = FMA(KP866025403, Tg, Tf);
101 T5 = W[0];
102 Tb = T5 * Ta;
103 Ti = T5 * Th;
104 Tc = W[1];
105 rio[WS(vs, 1)] = FMA(Tc, Th, Tb);
106 iio[WS(vs, 1)] = FNMS(Tc, Ta, Ti);
107 }
108 {
109 E T16, T19, T17, T1a, T15, T18;
110 T16 = FNMS(KP866025403, TV, TS);
111 T19 = FNMS(KP866025403, T12, T11);
112 T15 = W[2];
113 T17 = T15 * T16;
114 T1a = T15 * T19;
115 T18 = W[3];
116 rio[WS(vs, 2) + WS(rs, 2)] = FMA(T18, T19, T17);
117 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(T18, T16, T1a);
118 }
119 {
120 E TI, TL, TJ, TM, TH, TK;
121 TI = FNMS(KP866025403, Tx, Tu);
122 TL = FNMS(KP866025403, TE, TD);
123 TH = W[2];
124 TJ = TH * TI;
125 TM = TH * TL;
126 TK = W[3];
127 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TK, TL, TJ);
128 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TK, TI, TM);
129 }
130 {
131 E Ty, TF, Tz, TG, Tt, TA;
132 Ty = FMA(KP866025403, Tx, Tu);
133 TF = FMA(KP866025403, TE, TD);
134 Tt = W[0];
135 Tz = Tt * Ty;
136 TG = Tt * TF;
137 TA = W[1];
138 rio[WS(vs, 1) + WS(rs, 1)] = FMA(TA, TF, Tz);
139 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(TA, Ty, TG);
140 }
141 {
142 E TW, T13, TX, T14, TR, TY;
143 TW = FMA(KP866025403, TV, TS);
144 T13 = FMA(KP866025403, T12, T11);
145 TR = W[0];
146 TX = TR * TW;
147 T14 = TR * T13;
148 TY = W[1];
149 rio[WS(vs, 1) + WS(rs, 2)] = FMA(TY, T13, TX);
150 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(TY, TW, T14);
151 }
152 {
153 E Tk, Tn, Tl, To, Tj, Tm;
154 Tk = FNMS(KP866025403, T9, T6);
155 Tn = FNMS(KP866025403, Tg, Tf);
156 Tj = W[2];
157 Tl = Tj * Tk;
158 To = Tj * Tn;
159 Tm = W[3];
160 rio[WS(vs, 2)] = FMA(Tm, Tn, Tl);
161 iio[WS(vs, 2)] = FNMS(Tm, Tk, To);
162 }
163 }
164 }
165 }
166
167 static const tw_instr twinstr[] = {
168 {TW_FULL, 0, 3},
169 {TW_NEXT, 1, 0}
170 };
171
172 static const ct_desc desc = { 3, "q1_3", twinstr, &GENUS, {18, 12, 30, 0}, 0, 0, 0 };
173
174 void X(codelet_q1_3) (planner *p) {
175 X(kdft_difsq_register) (p, q1_3, &desc);
176 }
177 #else
178
179 /* Generated by: ../../../genfft/gen_twidsq.native -compact -variables 4 -pipeline-latency 4 -reload-twiddle -dif -n 3 -name q1_3 -include dft/scalar/q.h */
180
181 /*
182 * This function contains 48 FP additions, 36 FP multiplications,
183 * (or, 30 additions, 18 multiplications, 18 fused multiply/add),
184 * 35 stack variables, 2 constants, and 36 memory accesses
185 */
186 #include "dft/scalar/q.h"
187
188 static void q1_3(R *rio, R *iio, const R *W, stride rs, stride vs, INT mb, INT me, INT ms)
189 {
190 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
191 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
192 {
193 INT m;
194 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, rio = rio + ms, iio = iio + ms, W = W + 4, MAKE_VOLATILE_STRIDE(6, rs), MAKE_VOLATILE_STRIDE(0, vs)) {
195 E T1, T4, T6, Tc, Td, Te, T9, Tf, Tl, To, Tq, Tw, Tx, Ty, Tt;
196 E Tz, TR, TS, TN, TT, TF, TI, TK, TQ;
197 {
198 E T2, T3, Tr, Ts;
199 T1 = rio[0];
200 T2 = rio[WS(rs, 1)];
201 T3 = rio[WS(rs, 2)];
202 T4 = T2 + T3;
203 T6 = FNMS(KP500000000, T4, T1);
204 Tc = KP866025403 * (T3 - T2);
205 {
206 E T7, T8, Tm, Tn;
207 Td = iio[0];
208 T7 = iio[WS(rs, 1)];
209 T8 = iio[WS(rs, 2)];
210 Te = T7 + T8;
211 T9 = KP866025403 * (T7 - T8);
212 Tf = FNMS(KP500000000, Te, Td);
213 Tl = rio[WS(vs, 1)];
214 Tm = rio[WS(vs, 1) + WS(rs, 1)];
215 Tn = rio[WS(vs, 1) + WS(rs, 2)];
216 To = Tm + Tn;
217 Tq = FNMS(KP500000000, To, Tl);
218 Tw = KP866025403 * (Tn - Tm);
219 }
220 Tx = iio[WS(vs, 1)];
221 Tr = iio[WS(vs, 1) + WS(rs, 1)];
222 Ts = iio[WS(vs, 1) + WS(rs, 2)];
223 Ty = Tr + Ts;
224 Tt = KP866025403 * (Tr - Ts);
225 Tz = FNMS(KP500000000, Ty, Tx);
226 {
227 E TL, TM, TG, TH;
228 TR = iio[WS(vs, 2)];
229 TL = iio[WS(vs, 2) + WS(rs, 1)];
230 TM = iio[WS(vs, 2) + WS(rs, 2)];
231 TS = TL + TM;
232 TN = KP866025403 * (TL - TM);
233 TT = FNMS(KP500000000, TS, TR);
234 TF = rio[WS(vs, 2)];
235 TG = rio[WS(vs, 2) + WS(rs, 1)];
236 TH = rio[WS(vs, 2) + WS(rs, 2)];
237 TI = TG + TH;
238 TK = FNMS(KP500000000, TI, TF);
239 TQ = KP866025403 * (TH - TG);
240 }
241 }
242 rio[0] = T1 + T4;
243 iio[0] = Td + Te;
244 rio[WS(rs, 1)] = Tl + To;
245 iio[WS(rs, 1)] = Tx + Ty;
246 iio[WS(rs, 2)] = TR + TS;
247 rio[WS(rs, 2)] = TF + TI;
248 {
249 E Ta, Tg, T5, Tb;
250 Ta = T6 + T9;
251 Tg = Tc + Tf;
252 T5 = W[0];
253 Tb = W[1];
254 rio[WS(vs, 1)] = FMA(T5, Ta, Tb * Tg);
255 iio[WS(vs, 1)] = FNMS(Tb, Ta, T5 * Tg);
256 }
257 {
258 E TW, TY, TV, TX;
259 TW = TK - TN;
260 TY = TT - TQ;
261 TV = W[2];
262 TX = W[3];
263 rio[WS(vs, 2) + WS(rs, 2)] = FMA(TV, TW, TX * TY);
264 iio[WS(vs, 2) + WS(rs, 2)] = FNMS(TX, TW, TV * TY);
265 }
266 {
267 E TC, TE, TB, TD;
268 TC = Tq - Tt;
269 TE = Tz - Tw;
270 TB = W[2];
271 TD = W[3];
272 rio[WS(vs, 2) + WS(rs, 1)] = FMA(TB, TC, TD * TE);
273 iio[WS(vs, 2) + WS(rs, 1)] = FNMS(TD, TC, TB * TE);
274 }
275 {
276 E Tu, TA, Tp, Tv;
277 Tu = Tq + Tt;
278 TA = Tw + Tz;
279 Tp = W[0];
280 Tv = W[1];
281 rio[WS(vs, 1) + WS(rs, 1)] = FMA(Tp, Tu, Tv * TA);
282 iio[WS(vs, 1) + WS(rs, 1)] = FNMS(Tv, Tu, Tp * TA);
283 }
284 {
285 E TO, TU, TJ, TP;
286 TO = TK + TN;
287 TU = TQ + TT;
288 TJ = W[0];
289 TP = W[1];
290 rio[WS(vs, 1) + WS(rs, 2)] = FMA(TJ, TO, TP * TU);
291 iio[WS(vs, 1) + WS(rs, 2)] = FNMS(TP, TO, TJ * TU);
292 }
293 {
294 E Ti, Tk, Th, Tj;
295 Ti = T6 - T9;
296 Tk = Tf - Tc;
297 Th = W[2];
298 Tj = W[3];
299 rio[WS(vs, 2)] = FMA(Th, Ti, Tj * Tk);
300 iio[WS(vs, 2)] = FNMS(Tj, Ti, Th * Tk);
301 }
302 }
303 }
304 }
305
306 static const tw_instr twinstr[] = {
307 {TW_FULL, 0, 3},
308 {TW_NEXT, 1, 0}
309 };
310
311 static const ct_desc desc = { 3, "q1_3", twinstr, &GENUS, {30, 18, 18, 0}, 0, 0, 0 };
312
313 void X(codelet_q1_3) (planner *p) {
314 X(kdft_difsq_register) (p, q1_3, &desc);
315 }
316 #endif