Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cb/hb_5.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 5 -dif -name hb_5 -include rdft/scalar/hb.h */
29
30 /*
31 * This function contains 40 FP additions, 34 FP multiplications,
32 * (or, 14 additions, 8 multiplications, 26 fused multiply/add),
33 * 27 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "rdft/scalar/hb.h"
36
37 static void hb_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
40 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
41 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
42 DK(KP618033988, +0.618033988749894848204586834365638117720309180);
43 {
44 INT m;
45 for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(10, rs)) {
46 E T1, Tb, TM, Tw, T8, Ta, Tn, Tj, TH, Ts, Tq, Tr;
47 {
48 E T4, Tu, T7, Tv;
49 T1 = cr[0];
50 {
51 E T2, T3, T5, T6;
52 T2 = cr[WS(rs, 1)];
53 T3 = ci[0];
54 T4 = T2 + T3;
55 Tu = T2 - T3;
56 T5 = cr[WS(rs, 2)];
57 T6 = ci[WS(rs, 1)];
58 T7 = T5 + T6;
59 Tv = T5 - T6;
60 }
61 Tb = T4 - T7;
62 TM = FNMS(KP618033988, Tu, Tv);
63 Tw = FMA(KP618033988, Tv, Tu);
64 T8 = T4 + T7;
65 Ta = FNMS(KP250000000, T8, T1);
66 }
67 {
68 E Tf, To, Ti, Tp;
69 Tn = ci[WS(rs, 4)];
70 {
71 E Td, Te, Tg, Th;
72 Td = ci[WS(rs, 3)];
73 Te = cr[WS(rs, 4)];
74 Tf = Td + Te;
75 To = Td - Te;
76 Tg = ci[WS(rs, 2)];
77 Th = cr[WS(rs, 3)];
78 Ti = Tg + Th;
79 Tp = Tg - Th;
80 }
81 Tj = FMA(KP618033988, Ti, Tf);
82 TH = FNMS(KP618033988, Tf, Ti);
83 Ts = To - Tp;
84 Tq = To + Tp;
85 Tr = FNMS(KP250000000, Tq, Tn);
86 }
87 cr[0] = T1 + T8;
88 ci[0] = Tn + Tq;
89 {
90 E Tk, TA, Tx, TD, Tc, Tt;
91 Tc = FMA(KP559016994, Tb, Ta);
92 Tk = FNMS(KP951056516, Tj, Tc);
93 TA = FMA(KP951056516, Tj, Tc);
94 Tt = FMA(KP559016994, Ts, Tr);
95 Tx = FMA(KP951056516, Tw, Tt);
96 TD = FNMS(KP951056516, Tw, Tt);
97 {
98 E T9, Tl, Tm, Ty;
99 T9 = W[0];
100 Tl = T9 * Tk;
101 Tm = W[1];
102 Ty = Tm * Tk;
103 cr[WS(rs, 1)] = FNMS(Tm, Tx, Tl);
104 ci[WS(rs, 1)] = FMA(T9, Tx, Ty);
105 }
106 {
107 E Tz, TB, TC, TE;
108 Tz = W[6];
109 TB = Tz * TA;
110 TC = W[7];
111 TE = TC * TA;
112 cr[WS(rs, 4)] = FNMS(TC, TD, TB);
113 ci[WS(rs, 4)] = FMA(Tz, TD, TE);
114 }
115 }
116 {
117 E TI, TQ, TN, TT, TG, TL;
118 TG = FNMS(KP559016994, Tb, Ta);
119 TI = FMA(KP951056516, TH, TG);
120 TQ = FNMS(KP951056516, TH, TG);
121 TL = FNMS(KP559016994, Ts, Tr);
122 TN = FNMS(KP951056516, TM, TL);
123 TT = FMA(KP951056516, TM, TL);
124 {
125 E TF, TJ, TK, TO;
126 TF = W[2];
127 TJ = TF * TI;
128 TK = W[3];
129 TO = TK * TI;
130 cr[WS(rs, 2)] = FNMS(TK, TN, TJ);
131 ci[WS(rs, 2)] = FMA(TF, TN, TO);
132 }
133 {
134 E TP, TR, TS, TU;
135 TP = W[4];
136 TR = TP * TQ;
137 TS = W[5];
138 TU = TS * TQ;
139 cr[WS(rs, 3)] = FNMS(TS, TT, TR);
140 ci[WS(rs, 3)] = FMA(TP, TT, TU);
141 }
142 }
143 }
144 }
145 }
146
147 static const tw_instr twinstr[] = {
148 {TW_FULL, 1, 5},
149 {TW_NEXT, 1, 0}
150 };
151
152 static const hc2hc_desc desc = { 5, "hb_5", twinstr, &GENUS, {14, 8, 26, 0} };
153
154 void X(codelet_hb_5) (planner *p) {
155 X(khc2hc_register) (p, hb_5, &desc);
156 }
157 #else
158
159 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 5 -dif -name hb_5 -include rdft/scalar/hb.h */
160
161 /*
162 * This function contains 40 FP additions, 28 FP multiplications,
163 * (or, 26 additions, 14 multiplications, 14 fused multiply/add),
164 * 27 stack variables, 4 constants, and 20 memory accesses
165 */
166 #include "rdft/scalar/hb.h"
167
168 static void hb_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
169 {
170 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
171 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
172 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
173 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
174 {
175 INT m;
176 for (m = mb, W = W + ((mb - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(10, rs)) {
177 E T1, Tj, TG, Ts, T8, Ti, T9, Tn, TD, Tu, Tg, Tt;
178 {
179 E T4, Tq, T7, Tr;
180 T1 = cr[0];
181 {
182 E T2, T3, T5, T6;
183 T2 = cr[WS(rs, 1)];
184 T3 = ci[0];
185 T4 = T2 + T3;
186 Tq = T2 - T3;
187 T5 = cr[WS(rs, 2)];
188 T6 = ci[WS(rs, 1)];
189 T7 = T5 + T6;
190 Tr = T5 - T6;
191 }
192 Tj = KP559016994 * (T4 - T7);
193 TG = FMA(KP951056516, Tq, KP587785252 * Tr);
194 Ts = FNMS(KP951056516, Tr, KP587785252 * Tq);
195 T8 = T4 + T7;
196 Ti = FNMS(KP250000000, T8, T1);
197 }
198 {
199 E Tc, Tl, Tf, Tm;
200 T9 = ci[WS(rs, 4)];
201 {
202 E Ta, Tb, Td, Te;
203 Ta = ci[WS(rs, 3)];
204 Tb = cr[WS(rs, 4)];
205 Tc = Ta - Tb;
206 Tl = Ta + Tb;
207 Td = ci[WS(rs, 2)];
208 Te = cr[WS(rs, 3)];
209 Tf = Td - Te;
210 Tm = Td + Te;
211 }
212 Tn = FNMS(KP951056516, Tm, KP587785252 * Tl);
213 TD = FMA(KP951056516, Tl, KP587785252 * Tm);
214 Tu = KP559016994 * (Tc - Tf);
215 Tg = Tc + Tf;
216 Tt = FNMS(KP250000000, Tg, T9);
217 }
218 cr[0] = T1 + T8;
219 ci[0] = T9 + Tg;
220 {
221 E To, Ty, Tw, TA, Tk, Tv;
222 Tk = Ti - Tj;
223 To = Tk - Tn;
224 Ty = Tk + Tn;
225 Tv = Tt - Tu;
226 Tw = Ts + Tv;
227 TA = Tv - Ts;
228 {
229 E Th, Tp, Tx, Tz;
230 Th = W[2];
231 Tp = W[3];
232 cr[WS(rs, 2)] = FNMS(Tp, Tw, Th * To);
233 ci[WS(rs, 2)] = FMA(Th, Tw, Tp * To);
234 Tx = W[4];
235 Tz = W[5];
236 cr[WS(rs, 3)] = FNMS(Tz, TA, Tx * Ty);
237 ci[WS(rs, 3)] = FMA(Tx, TA, Tz * Ty);
238 }
239 }
240 {
241 E TE, TK, TI, TM, TC, TH;
242 TC = Tj + Ti;
243 TE = TC - TD;
244 TK = TC + TD;
245 TH = Tu + Tt;
246 TI = TG + TH;
247 TM = TH - TG;
248 {
249 E TB, TF, TJ, TL;
250 TB = W[0];
251 TF = W[1];
252 cr[WS(rs, 1)] = FNMS(TF, TI, TB * TE);
253 ci[WS(rs, 1)] = FMA(TB, TI, TF * TE);
254 TJ = W[6];
255 TL = W[7];
256 cr[WS(rs, 4)] = FNMS(TL, TM, TJ * TK);
257 ci[WS(rs, 4)] = FMA(TJ, TM, TL * TK);
258 }
259 }
260 }
261 }
262 }
263
264 static const tw_instr twinstr[] = {
265 {TW_FULL, 1, 5},
266 {TW_NEXT, 1, 0}
267 };
268
269 static const hc2hc_desc desc = { 5, "hb_5", twinstr, &GENUS, {26, 14, 14, 0} };
270
271 void X(codelet_hb_5) (planner *p) {
272 X(khc2hc_register) (p, hb_5, &desc);
273 }
274 #endif