Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/dft/scalar/codelets/t2_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 08:04:25 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_twiddle.native -fma -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 5 -name t2_5 -include dft/scalar/t.h */
29
30 /*
31 * This function contains 44 FP additions, 40 FP multiplications,
32 * (or, 14 additions, 10 multiplications, 30 fused multiply/add),
33 * 38 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "dft/scalar/t.h"
36
37 static void t2_5(R *ri, R *ii, 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 * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
46 E T2, Ta, T8, T5, Tb, Tm, Tf, Tj, T9, Te;
47 T2 = W[0];
48 Ta = W[3];
49 T8 = W[2];
50 T9 = T2 * T8;
51 Te = T2 * Ta;
52 T5 = W[1];
53 Tb = FNMS(T5, Ta, T9);
54 Tm = FNMS(T5, T8, Te);
55 Tf = FMA(T5, T8, Te);
56 Tj = FMA(T5, Ta, T9);
57 {
58 E T1, TO, T7, Th, Ti, Tz, TB, TL, To, Ts, Tt, TE, TG, TM;
59 T1 = ri[0];
60 TO = ii[0];
61 {
62 E T3, T4, T6, Ty, Tc, Td, Tg, TA;
63 T3 = ri[WS(rs, 1)];
64 T4 = T2 * T3;
65 T6 = ii[WS(rs, 1)];
66 Ty = T2 * T6;
67 Tc = ri[WS(rs, 4)];
68 Td = Tb * Tc;
69 Tg = ii[WS(rs, 4)];
70 TA = Tb * Tg;
71 T7 = FMA(T5, T6, T4);
72 Th = FMA(Tf, Tg, Td);
73 Ti = T7 + Th;
74 Tz = FNMS(T5, T3, Ty);
75 TB = FNMS(Tf, Tc, TA);
76 TL = Tz + TB;
77 }
78 {
79 E Tk, Tl, Tn, TD, Tp, Tq, Tr, TF;
80 Tk = ri[WS(rs, 2)];
81 Tl = Tj * Tk;
82 Tn = ii[WS(rs, 2)];
83 TD = Tj * Tn;
84 Tp = ri[WS(rs, 3)];
85 Tq = T8 * Tp;
86 Tr = ii[WS(rs, 3)];
87 TF = T8 * Tr;
88 To = FMA(Tm, Tn, Tl);
89 Ts = FMA(Ta, Tr, Tq);
90 Tt = To + Ts;
91 TE = FNMS(Tm, Tk, TD);
92 TG = FNMS(Ta, Tp, TF);
93 TM = TE + TG;
94 }
95 {
96 E Tw, Tu, Tv, TI, TK, TC, TH, TJ, Tx;
97 Tw = Ti - Tt;
98 Tu = Ti + Tt;
99 Tv = FNMS(KP250000000, Tu, T1);
100 TC = Tz - TB;
101 TH = TE - TG;
102 TI = FMA(KP618033988, TH, TC);
103 TK = FNMS(KP618033988, TC, TH);
104 ri[0] = T1 + Tu;
105 TJ = FNMS(KP559016994, Tw, Tv);
106 ri[WS(rs, 2)] = FNMS(KP951056516, TK, TJ);
107 ri[WS(rs, 3)] = FMA(KP951056516, TK, TJ);
108 Tx = FMA(KP559016994, Tw, Tv);
109 ri[WS(rs, 4)] = FNMS(KP951056516, TI, Tx);
110 ri[WS(rs, 1)] = FMA(KP951056516, TI, Tx);
111 }
112 {
113 E TQ, TN, TP, TU, TW, TS, TT, TV, TR;
114 TQ = TL - TM;
115 TN = TL + TM;
116 TP = FNMS(KP250000000, TN, TO);
117 TS = T7 - Th;
118 TT = To - Ts;
119 TU = FMA(KP618033988, TT, TS);
120 TW = FNMS(KP618033988, TS, TT);
121 ii[0] = TN + TO;
122 TV = FNMS(KP559016994, TQ, TP);
123 ii[WS(rs, 2)] = FMA(KP951056516, TW, TV);
124 ii[WS(rs, 3)] = FNMS(KP951056516, TW, TV);
125 TR = FMA(KP559016994, TQ, TP);
126 ii[WS(rs, 1)] = FNMS(KP951056516, TU, TR);
127 ii[WS(rs, 4)] = FMA(KP951056516, TU, TR);
128 }
129 }
130 }
131 }
132 }
133
134 static const tw_instr twinstr[] = {
135 {TW_CEXP, 0, 1},
136 {TW_CEXP, 0, 3},
137 {TW_NEXT, 1, 0}
138 };
139
140 static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {14, 10, 30, 0}, 0, 0, 0 };
141
142 void X(codelet_t2_5) (planner *p) {
143 X(kdft_dit_register) (p, t2_5, &desc);
144 }
145 #else
146
147 /* Generated by: ../../../genfft/gen_twiddle.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 5 -name t2_5 -include dft/scalar/t.h */
148
149 /*
150 * This function contains 44 FP additions, 32 FP multiplications,
151 * (or, 30 additions, 18 multiplications, 14 fused multiply/add),
152 * 37 stack variables, 4 constants, and 20 memory accesses
153 */
154 #include "dft/scalar/t.h"
155
156 static void t2_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
157 {
158 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
159 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
160 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
161 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
162 {
163 INT m;
164 for (m = mb, W = W + (mb * 4); m < me; m = m + 1, ri = ri + ms, ii = ii + ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
165 E T2, T4, T7, T9, Tb, Tl, Tf, Tj;
166 {
167 E T8, Te, Ta, Td;
168 T2 = W[0];
169 T4 = W[1];
170 T7 = W[2];
171 T9 = W[3];
172 T8 = T2 * T7;
173 Te = T4 * T7;
174 Ta = T4 * T9;
175 Td = T2 * T9;
176 Tb = T8 - Ta;
177 Tl = Td - Te;
178 Tf = Td + Te;
179 Tj = T8 + Ta;
180 }
181 {
182 E T1, TI, Ty, TB, TN, TM, TF, TG, TH, Ti, Tr, Ts;
183 T1 = ri[0];
184 TI = ii[0];
185 {
186 E T6, Tw, Tq, TA, Th, Tx, Tn, Tz;
187 {
188 E T3, T5, To, Tp;
189 T3 = ri[WS(rs, 1)];
190 T5 = ii[WS(rs, 1)];
191 T6 = FMA(T2, T3, T4 * T5);
192 Tw = FNMS(T4, T3, T2 * T5);
193 To = ri[WS(rs, 3)];
194 Tp = ii[WS(rs, 3)];
195 Tq = FMA(T7, To, T9 * Tp);
196 TA = FNMS(T9, To, T7 * Tp);
197 }
198 {
199 E Tc, Tg, Tk, Tm;
200 Tc = ri[WS(rs, 4)];
201 Tg = ii[WS(rs, 4)];
202 Th = FMA(Tb, Tc, Tf * Tg);
203 Tx = FNMS(Tf, Tc, Tb * Tg);
204 Tk = ri[WS(rs, 2)];
205 Tm = ii[WS(rs, 2)];
206 Tn = FMA(Tj, Tk, Tl * Tm);
207 Tz = FNMS(Tl, Tk, Tj * Tm);
208 }
209 Ty = Tw - Tx;
210 TB = Tz - TA;
211 TN = Tn - Tq;
212 TM = T6 - Th;
213 TF = Tw + Tx;
214 TG = Tz + TA;
215 TH = TF + TG;
216 Ti = T6 + Th;
217 Tr = Tn + Tq;
218 Ts = Ti + Tr;
219 }
220 ri[0] = T1 + Ts;
221 ii[0] = TH + TI;
222 {
223 E TC, TE, Tv, TD, Tt, Tu;
224 TC = FMA(KP951056516, Ty, KP587785252 * TB);
225 TE = FNMS(KP587785252, Ty, KP951056516 * TB);
226 Tt = KP559016994 * (Ti - Tr);
227 Tu = FNMS(KP250000000, Ts, T1);
228 Tv = Tt + Tu;
229 TD = Tu - Tt;
230 ri[WS(rs, 4)] = Tv - TC;
231 ri[WS(rs, 3)] = TD + TE;
232 ri[WS(rs, 1)] = Tv + TC;
233 ri[WS(rs, 2)] = TD - TE;
234 }
235 {
236 E TO, TP, TL, TQ, TJ, TK;
237 TO = FMA(KP951056516, TM, KP587785252 * TN);
238 TP = FNMS(KP587785252, TM, KP951056516 * TN);
239 TJ = KP559016994 * (TF - TG);
240 TK = FNMS(KP250000000, TH, TI);
241 TL = TJ + TK;
242 TQ = TK - TJ;
243 ii[WS(rs, 1)] = TL - TO;
244 ii[WS(rs, 3)] = TQ - TP;
245 ii[WS(rs, 4)] = TO + TL;
246 ii[WS(rs, 2)] = TP + TQ;
247 }
248 }
249 }
250 }
251 }
252
253 static const tw_instr twinstr[] = {
254 {TW_CEXP, 0, 1},
255 {TW_CEXP, 0, 3},
256 {TW_NEXT, 1, 0}
257 };
258
259 static const ct_desc desc = { 5, "t2_5", twinstr, &GENUS, {30, 18, 14, 0}, 0, 0, 0 };
260
261 void X(codelet_t2_5) (planner *p) {
262 X(kdft_dit_register) (p, t2_5, &desc);
263 }
264 #endif