Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cb/hb_6.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 08:07:31 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 -sign 1 -n 6 -dif -name hb_6 -include rdft/scalar/hb.h */
29
30 /*
31 * This function contains 46 FP additions, 32 FP multiplications,
32 * (or, 24 additions, 10 multiplications, 22 fused multiply/add),
33 * 31 stack variables, 2 constants, and 24 memory accesses
34 */
35 #include "rdft/scalar/hb.h"
36
37 static void hb_6(R *cr, R *ci, const R *W, stride rs, 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 - 1) * 10); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
44 E Td, Tn, TO, TJ, TN, Tk, Tr, T3, TC, Ts, TQ, Ta, Tm, TF, TG;
45 {
46 E Tb, Tc, Tg, TH, Tj, TI;
47 Tb = ci[WS(rs, 5)];
48 Tc = cr[WS(rs, 3)];
49 Td = Tb - Tc;
50 {
51 E Te, Tf, Th, Ti;
52 Te = ci[WS(rs, 3)];
53 Tf = cr[WS(rs, 5)];
54 Tg = Te - Tf;
55 TH = Te + Tf;
56 Th = ci[WS(rs, 4)];
57 Ti = cr[WS(rs, 4)];
58 Tj = Th - Ti;
59 TI = Th + Ti;
60 }
61 Tn = Tj - Tg;
62 TO = TH - TI;
63 TJ = TH + TI;
64 TN = Tb + Tc;
65 Tk = Tg + Tj;
66 Tr = FNMS(KP500000000, Tk, Td);
67 }
68 {
69 E T6, TD, T9, TE, T1, T2;
70 T1 = cr[0];
71 T2 = ci[WS(rs, 2)];
72 T3 = T1 + T2;
73 TC = T1 - T2;
74 {
75 E T4, T5, T7, T8;
76 T4 = cr[WS(rs, 2)];
77 T5 = ci[0];
78 T6 = T4 + T5;
79 TD = T4 - T5;
80 T7 = ci[WS(rs, 1)];
81 T8 = cr[WS(rs, 1)];
82 T9 = T7 + T8;
83 TE = T7 - T8;
84 }
85 Ts = T6 - T9;
86 TQ = TD - TE;
87 Ta = T6 + T9;
88 Tm = FNMS(KP500000000, Ta, T3);
89 TF = TD + TE;
90 TG = FNMS(KP500000000, TF, TC);
91 }
92 cr[0] = T3 + Ta;
93 ci[0] = Td + Tk;
94 {
95 E To, Tt, Tp, Tu, Tl, Tq;
96 To = FNMS(KP866025403, Tn, Tm);
97 Tt = FNMS(KP866025403, Ts, Tr);
98 Tl = W[2];
99 Tp = Tl * To;
100 Tu = Tl * Tt;
101 Tq = W[3];
102 cr[WS(rs, 2)] = FNMS(Tq, Tt, Tp);
103 ci[WS(rs, 2)] = FMA(Tq, To, Tu);
104 }
105 {
106 E T13, TZ, T11, T12, T14, T10;
107 T13 = TN + TO;
108 T10 = TC + TF;
109 TZ = W[4];
110 T11 = TZ * T10;
111 T12 = W[5];
112 T14 = T12 * T10;
113 cr[WS(rs, 3)] = FNMS(T12, T13, T11);
114 ci[WS(rs, 3)] = FMA(TZ, T13, T14);
115 }
116 {
117 E Tw, Tz, Tx, TA, Tv, Ty;
118 Tw = FMA(KP866025403, Tn, Tm);
119 Tz = FMA(KP866025403, Ts, Tr);
120 Tv = W[6];
121 Tx = Tv * Tw;
122 TA = Tv * Tz;
123 Ty = W[7];
124 cr[WS(rs, 4)] = FNMS(Ty, Tz, Tx);
125 ci[WS(rs, 4)] = FMA(Ty, Tw, TA);
126 }
127 {
128 E TR, TX, TT, TV, TW, TY, TB, TL, TM, TS, TP, TU, TK;
129 TP = FNMS(KP500000000, TO, TN);
130 TR = FMA(KP866025403, TQ, TP);
131 TX = FNMS(KP866025403, TQ, TP);
132 TU = FMA(KP866025403, TJ, TG);
133 TT = W[8];
134 TV = TT * TU;
135 TW = W[9];
136 TY = TW * TU;
137 TK = FNMS(KP866025403, TJ, TG);
138 TB = W[0];
139 TL = TB * TK;
140 TM = W[1];
141 TS = TM * TK;
142 cr[WS(rs, 1)] = FNMS(TM, TR, TL);
143 ci[WS(rs, 1)] = FMA(TB, TR, TS);
144 cr[WS(rs, 5)] = FNMS(TW, TX, TV);
145 ci[WS(rs, 5)] = FMA(TT, TX, TY);
146 }
147 }
148 }
149 }
150
151 static const tw_instr twinstr[] = {
152 {TW_FULL, 1, 6},
153 {TW_NEXT, 1, 0}
154 };
155
156 static const hc2hc_desc desc = { 6, "hb_6", twinstr, &GENUS, {24, 10, 22, 0} };
157
158 void X(codelet_hb_6) (planner *p) {
159 X(khc2hc_register) (p, hb_6, &desc);
160 }
161 #else
162
163 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 6 -dif -name hb_6 -include rdft/scalar/hb.h */
164
165 /*
166 * This function contains 46 FP additions, 28 FP multiplications,
167 * (or, 32 additions, 14 multiplications, 14 fused multiply/add),
168 * 27 stack variables, 2 constants, and 24 memory accesses
169 */
170 #include "rdft/scalar/hb.h"
171
172 static void hb_6(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
173 {
174 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
175 DK(KP866025403, +0.866025403784438646763723170752936183471402627);
176 {
177 INT m;
178 for (m = mb, W = W + ((mb - 1) * 10); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 10, MAKE_VOLATILE_STRIDE(12, rs)) {
179 E T3, Ty, Ta, TO, Tr, TB, Td, TE, Tk, TL, Tn, TH;
180 {
181 E T1, T2, Tb, Tc;
182 T1 = cr[0];
183 T2 = ci[WS(rs, 2)];
184 T3 = T1 + T2;
185 Ty = T1 - T2;
186 {
187 E T6, Tz, T9, TA;
188 {
189 E T4, T5, T7, T8;
190 T4 = cr[WS(rs, 2)];
191 T5 = ci[0];
192 T6 = T4 + T5;
193 Tz = T4 - T5;
194 T7 = ci[WS(rs, 1)];
195 T8 = cr[WS(rs, 1)];
196 T9 = T7 + T8;
197 TA = T7 - T8;
198 }
199 Ta = T6 + T9;
200 TO = KP866025403 * (Tz - TA);
201 Tr = KP866025403 * (T6 - T9);
202 TB = Tz + TA;
203 }
204 Tb = ci[WS(rs, 5)];
205 Tc = cr[WS(rs, 3)];
206 Td = Tb - Tc;
207 TE = Tb + Tc;
208 {
209 E Tg, TG, Tj, TF;
210 {
211 E Te, Tf, Th, Ti;
212 Te = ci[WS(rs, 3)];
213 Tf = cr[WS(rs, 5)];
214 Tg = Te - Tf;
215 TG = Te + Tf;
216 Th = ci[WS(rs, 4)];
217 Ti = cr[WS(rs, 4)];
218 Tj = Th - Ti;
219 TF = Th + Ti;
220 }
221 Tk = Tg + Tj;
222 TL = KP866025403 * (TG + TF);
223 Tn = KP866025403 * (Tj - Tg);
224 TH = TF - TG;
225 }
226 }
227 cr[0] = T3 + Ta;
228 ci[0] = Td + Tk;
229 {
230 E TC, TI, Tx, TD;
231 TC = Ty + TB;
232 TI = TE - TH;
233 Tx = W[4];
234 TD = W[5];
235 cr[WS(rs, 3)] = FNMS(TD, TI, Tx * TC);
236 ci[WS(rs, 3)] = FMA(TD, TC, Tx * TI);
237 }
238 {
239 E To, Tu, Ts, Tw, Tm, Tq;
240 Tm = FNMS(KP500000000, Ta, T3);
241 To = Tm - Tn;
242 Tu = Tm + Tn;
243 Tq = FNMS(KP500000000, Tk, Td);
244 Ts = Tq - Tr;
245 Tw = Tr + Tq;
246 {
247 E Tl, Tp, Tt, Tv;
248 Tl = W[2];
249 Tp = W[3];
250 cr[WS(rs, 2)] = FNMS(Tp, Ts, Tl * To);
251 ci[WS(rs, 2)] = FMA(Tl, Ts, Tp * To);
252 Tt = W[6];
253 Tv = W[7];
254 cr[WS(rs, 4)] = FNMS(Tv, Tw, Tt * Tu);
255 ci[WS(rs, 4)] = FMA(Tt, Tw, Tv * Tu);
256 }
257 }
258 {
259 E TM, TS, TQ, TU, TK, TP;
260 TK = FNMS(KP500000000, TB, Ty);
261 TM = TK - TL;
262 TS = TK + TL;
263 TP = FMA(KP500000000, TH, TE);
264 TQ = TO + TP;
265 TU = TP - TO;
266 {
267 E TJ, TN, TR, TT;
268 TJ = W[0];
269 TN = W[1];
270 cr[WS(rs, 1)] = FNMS(TN, TQ, TJ * TM);
271 ci[WS(rs, 1)] = FMA(TN, TM, TJ * TQ);
272 TR = W[8];
273 TT = W[9];
274 cr[WS(rs, 5)] = FNMS(TT, TU, TR * TS);
275 ci[WS(rs, 5)] = FMA(TT, TS, TR * TU);
276 }
277 }
278 }
279 }
280 }
281
282 static const tw_instr twinstr[] = {
283 {TW_FULL, 1, 6},
284 {TW_NEXT, 1, 0}
285 };
286
287 static const hc2hc_desc desc = { 6, "hb_6", twinstr, &GENUS, {32, 14, 14, 0} };
288
289 void X(codelet_hb_6) (planner *p) {
290 X(khc2hc_register) (p, hb_6, &desc);
291 }
292 #endif