Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hf_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 08:06:28 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 -n 5 -dit -name hf_5 -include rdft/scalar/hf.h */
29
30 /*
31 * This function contains 40 FP additions, 34 FP multiplications,
32 * (or, 14 additions, 8 multiplications, 26 fused multiply/add),
33 * 31 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "rdft/scalar/hf.h"
36
37 static void hf_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(KP618033988, +0.618033988749894848204586834365638117720309180);
42 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
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, TJ, T7, Tx, Td, Tz, Te, TK, Tk, TC, Tq, TE, Tr, TL;
47 T1 = cr[0];
48 TJ = ci[0];
49 {
50 E T3, T6, T4, Tw, T9, Tc, Ta, Ty, T2, T8, T5, Tb;
51 T3 = cr[WS(rs, 1)];
52 T6 = ci[WS(rs, 1)];
53 T2 = W[0];
54 T4 = T2 * T3;
55 Tw = T2 * T6;
56 T9 = cr[WS(rs, 4)];
57 Tc = ci[WS(rs, 4)];
58 T8 = W[6];
59 Ta = T8 * T9;
60 Ty = T8 * Tc;
61 T5 = W[1];
62 T7 = FMA(T5, T6, T4);
63 Tx = FNMS(T5, T3, Tw);
64 Tb = W[7];
65 Td = FMA(Tb, Tc, Ta);
66 Tz = FNMS(Tb, T9, Ty);
67 Te = T7 + Td;
68 TK = Tx + Tz;
69 }
70 {
71 E Tg, Tj, Th, TB, Tm, Tp, Tn, TD, Tf, Tl, Ti, To;
72 Tg = cr[WS(rs, 2)];
73 Tj = ci[WS(rs, 2)];
74 Tf = W[2];
75 Th = Tf * Tg;
76 TB = Tf * Tj;
77 Tm = cr[WS(rs, 3)];
78 Tp = ci[WS(rs, 3)];
79 Tl = W[4];
80 Tn = Tl * Tm;
81 TD = Tl * Tp;
82 Ti = W[3];
83 Tk = FMA(Ti, Tj, Th);
84 TC = FNMS(Ti, Tg, TB);
85 To = W[5];
86 Tq = FMA(To, Tp, Tn);
87 TE = FNMS(To, Tm, TD);
88 Tr = Tk + Tq;
89 TL = TC + TE;
90 }
91 {
92 E Tu, Ts, Tt, TG, TI, TA, TF, Tv, TH;
93 Tu = Te - Tr;
94 Ts = Te + Tr;
95 Tt = FNMS(KP250000000, Ts, T1);
96 TA = Tx - Tz;
97 TF = TC - TE;
98 TG = FMA(KP618033988, TF, TA);
99 TI = FNMS(KP618033988, TA, TF);
100 cr[0] = T1 + Ts;
101 Tv = FMA(KP559016994, Tu, Tt);
102 ci[0] = FNMS(KP951056516, TG, Tv);
103 cr[WS(rs, 1)] = FMA(KP951056516, TG, Tv);
104 TH = FNMS(KP559016994, Tu, Tt);
105 cr[WS(rs, 2)] = FNMS(KP951056516, TI, TH);
106 ci[WS(rs, 1)] = FMA(KP951056516, TI, TH);
107 }
108 {
109 E TO, TM, TN, TS, TU, TQ, TR, TT, TP;
110 TO = TK - TL;
111 TM = TK + TL;
112 TN = FNMS(KP250000000, TM, TJ);
113 TQ = Tk - Tq;
114 TR = Td - T7;
115 TS = FMA(KP618033988, TR, TQ);
116 TU = FNMS(KP618033988, TQ, TR);
117 ci[WS(rs, 4)] = TM + TJ;
118 TT = FMA(KP559016994, TO, TN);
119 cr[WS(rs, 4)] = FMS(KP951056516, TU, TT);
120 ci[WS(rs, 3)] = FMA(KP951056516, TU, TT);
121 TP = FNMS(KP559016994, TO, TN);
122 cr[WS(rs, 3)] = FMS(KP951056516, TS, TP);
123 ci[WS(rs, 2)] = FMA(KP951056516, TS, TP);
124 }
125 }
126 }
127 }
128
129 static const tw_instr twinstr[] = {
130 {TW_FULL, 1, 5},
131 {TW_NEXT, 1, 0}
132 };
133
134 static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {14, 8, 26, 0} };
135
136 void X(codelet_hf_5) (planner *p) {
137 X(khc2hc_register) (p, hf_5, &desc);
138 }
139 #else
140
141 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 5 -dit -name hf_5 -include rdft/scalar/hf.h */
142
143 /*
144 * This function contains 40 FP additions, 28 FP multiplications,
145 * (or, 26 additions, 14 multiplications, 14 fused multiply/add),
146 * 29 stack variables, 4 constants, and 20 memory accesses
147 */
148 #include "rdft/scalar/hf.h"
149
150 static void hf_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
151 {
152 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
153 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
154 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
155 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
156 {
157 INT m;
158 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)) {
159 E T1, TE, Tu, Tx, TC, TB, TF, TG, TH, Tc, Tn, To;
160 T1 = cr[0];
161 TE = ci[0];
162 {
163 E T6, Ts, Tm, Tw, Tb, Tt, Th, Tv;
164 {
165 E T3, T5, T2, T4;
166 T3 = cr[WS(rs, 1)];
167 T5 = ci[WS(rs, 1)];
168 T2 = W[0];
169 T4 = W[1];
170 T6 = FMA(T2, T3, T4 * T5);
171 Ts = FNMS(T4, T3, T2 * T5);
172 }
173 {
174 E Tj, Tl, Ti, Tk;
175 Tj = cr[WS(rs, 3)];
176 Tl = ci[WS(rs, 3)];
177 Ti = W[4];
178 Tk = W[5];
179 Tm = FMA(Ti, Tj, Tk * Tl);
180 Tw = FNMS(Tk, Tj, Ti * Tl);
181 }
182 {
183 E T8, Ta, T7, T9;
184 T8 = cr[WS(rs, 4)];
185 Ta = ci[WS(rs, 4)];
186 T7 = W[6];
187 T9 = W[7];
188 Tb = FMA(T7, T8, T9 * Ta);
189 Tt = FNMS(T9, T8, T7 * Ta);
190 }
191 {
192 E Te, Tg, Td, Tf;
193 Te = cr[WS(rs, 2)];
194 Tg = ci[WS(rs, 2)];
195 Td = W[2];
196 Tf = W[3];
197 Th = FMA(Td, Te, Tf * Tg);
198 Tv = FNMS(Tf, Te, Td * Tg);
199 }
200 Tu = Ts - Tt;
201 Tx = Tv - Tw;
202 TC = Th - Tm;
203 TB = Tb - T6;
204 TF = Ts + Tt;
205 TG = Tv + Tw;
206 TH = TF + TG;
207 Tc = T6 + Tb;
208 Tn = Th + Tm;
209 To = Tc + Tn;
210 }
211 cr[0] = T1 + To;
212 {
213 E Ty, TA, Tr, Tz, Tp, Tq;
214 Ty = FMA(KP951056516, Tu, KP587785252 * Tx);
215 TA = FNMS(KP587785252, Tu, KP951056516 * Tx);
216 Tp = KP559016994 * (Tc - Tn);
217 Tq = FNMS(KP250000000, To, T1);
218 Tr = Tp + Tq;
219 Tz = Tq - Tp;
220 ci[0] = Tr - Ty;
221 ci[WS(rs, 1)] = Tz + TA;
222 cr[WS(rs, 1)] = Tr + Ty;
223 cr[WS(rs, 2)] = Tz - TA;
224 }
225 ci[WS(rs, 4)] = TH + TE;
226 {
227 E TD, TL, TK, TM, TI, TJ;
228 TD = FMA(KP587785252, TB, KP951056516 * TC);
229 TL = FNMS(KP587785252, TC, KP951056516 * TB);
230 TI = FNMS(KP250000000, TH, TE);
231 TJ = KP559016994 * (TF - TG);
232 TK = TI - TJ;
233 TM = TJ + TI;
234 cr[WS(rs, 3)] = TD - TK;
235 ci[WS(rs, 3)] = TL + TM;
236 ci[WS(rs, 2)] = TD + TK;
237 cr[WS(rs, 4)] = TL - TM;
238 }
239 }
240 }
241 }
242
243 static const tw_instr twinstr[] = {
244 {TW_FULL, 1, 5},
245 {TW_NEXT, 1, 0}
246 };
247
248 static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {26, 14, 14, 0} };
249
250 void X(codelet_hf_5) (planner *p) {
251 X(khc2hc_register) (p, hf_5, &desc);
252 }
253 #endif