Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.5/rdft/scalar/r2cf/hf_5.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
children
comparison
equal deleted inserted replaced
41:481f5f8c5634 42:2cd0e3b3e1fd
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 Sat Jul 30 16:46:15 EDT 2016 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2hc.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -n 5 -dit -name hf_5 -include 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 * 43 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "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(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, TJ, TK, TA, TR, Te, TC, Tk, TE, Tq;
47 {
48 E Tg, Tj, Tm, TB, Th, Tp, Tl, Ti, To, TD, Tn;
49 T1 = cr[0];
50 TJ = ci[0];
51 {
52 E T9, Tc, Ty, Ta, Tb, Tx, T7, Tf, Tz, Td;
53 {
54 E T3, T6, T8, Tw, T4, T2, T5;
55 T3 = cr[WS(rs, 1)];
56 T6 = ci[WS(rs, 1)];
57 T2 = W[0];
58 T9 = cr[WS(rs, 4)];
59 Tc = ci[WS(rs, 4)];
60 T8 = W[6];
61 Tw = T2 * T6;
62 T4 = T2 * T3;
63 T5 = W[1];
64 Ty = T8 * Tc;
65 Ta = T8 * T9;
66 Tb = W[7];
67 Tx = FNMS(T5, T3, Tw);
68 T7 = FMA(T5, T6, T4);
69 }
70 Tg = cr[WS(rs, 2)];
71 Tz = FNMS(Tb, T9, Ty);
72 Td = FMA(Tb, Tc, Ta);
73 Tj = ci[WS(rs, 2)];
74 Tf = W[2];
75 TK = Tx + Tz;
76 TA = Tx - Tz;
77 TR = Td - T7;
78 Te = T7 + Td;
79 Tm = cr[WS(rs, 3)];
80 TB = Tf * Tj;
81 Th = Tf * Tg;
82 Tp = ci[WS(rs, 3)];
83 Tl = W[4];
84 Ti = W[3];
85 To = W[5];
86 }
87 TD = Tl * Tp;
88 Tn = Tl * Tm;
89 TC = FNMS(Ti, Tg, TB);
90 Tk = FMA(Ti, Tj, Th);
91 TE = FNMS(To, Tm, TD);
92 Tq = FMA(To, Tp, Tn);
93 }
94 {
95 E TG, TI, TO, TS, TU, Tu, TN, Tt, TL, TF;
96 TL = TC + TE;
97 TF = TC - TE;
98 {
99 E Tr, TQ, TM, Ts;
100 Tr = Tk + Tq;
101 TQ = Tk - Tq;
102 TG = FMA(KP618033988, TF, TA);
103 TI = FNMS(KP618033988, TA, TF);
104 TO = TK - TL;
105 TM = TK + TL;
106 TS = FMA(KP618033988, TR, TQ);
107 TU = FNMS(KP618033988, TQ, TR);
108 Tu = Te - Tr;
109 Ts = Te + Tr;
110 ci[WS(rs, 4)] = TM + TJ;
111 TN = FNMS(KP250000000, TM, TJ);
112 cr[0] = T1 + Ts;
113 Tt = FNMS(KP250000000, Ts, T1);
114 }
115 {
116 E TT, TP, Tv, TH;
117 TT = FMA(KP559016994, TO, TN);
118 TP = FNMS(KP559016994, TO, TN);
119 Tv = FMA(KP559016994, Tu, Tt);
120 TH = FNMS(KP559016994, Tu, Tt);
121 ci[WS(rs, 2)] = FMA(KP951056516, TS, TP);
122 cr[WS(rs, 3)] = FMS(KP951056516, TS, TP);
123 ci[WS(rs, 3)] = FMA(KP951056516, TU, TT);
124 cr[WS(rs, 4)] = FMS(KP951056516, TU, TT);
125 ci[WS(rs, 1)] = FMA(KP951056516, TI, TH);
126 cr[WS(rs, 2)] = FNMS(KP951056516, TI, TH);
127 cr[WS(rs, 1)] = FMA(KP951056516, TG, Tv);
128 ci[0] = FNMS(KP951056516, TG, Tv);
129 }
130 }
131 }
132 }
133 }
134
135 static const tw_instr twinstr[] = {
136 {TW_FULL, 1, 5},
137 {TW_NEXT, 1, 0}
138 };
139
140 static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {14, 8, 26, 0} };
141
142 void X(codelet_hf_5) (planner *p) {
143 X(khc2hc_register) (p, hf_5, &desc);
144 }
145 #else /* HAVE_FMA */
146
147 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -n 5 -dit -name hf_5 -include hf.h */
148
149 /*
150 * This function contains 40 FP additions, 28 FP multiplications,
151 * (or, 26 additions, 14 multiplications, 14 fused multiply/add),
152 * 29 stack variables, 4 constants, and 20 memory accesses
153 */
154 #include "hf.h"
155
156 static void hf_5(R *cr, R *ci, 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 - 1) * 8); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 8, MAKE_VOLATILE_STRIDE(10, rs)) {
165 E T1, TE, Tu, Tx, TC, TB, TF, TG, TH, Tc, Tn, To;
166 T1 = cr[0];
167 TE = ci[0];
168 {
169 E T6, Ts, Tm, Tw, Tb, Tt, Th, Tv;
170 {
171 E T3, T5, T2, T4;
172 T3 = cr[WS(rs, 1)];
173 T5 = ci[WS(rs, 1)];
174 T2 = W[0];
175 T4 = W[1];
176 T6 = FMA(T2, T3, T4 * T5);
177 Ts = FNMS(T4, T3, T2 * T5);
178 }
179 {
180 E Tj, Tl, Ti, Tk;
181 Tj = cr[WS(rs, 3)];
182 Tl = ci[WS(rs, 3)];
183 Ti = W[4];
184 Tk = W[5];
185 Tm = FMA(Ti, Tj, Tk * Tl);
186 Tw = FNMS(Tk, Tj, Ti * Tl);
187 }
188 {
189 E T8, Ta, T7, T9;
190 T8 = cr[WS(rs, 4)];
191 Ta = ci[WS(rs, 4)];
192 T7 = W[6];
193 T9 = W[7];
194 Tb = FMA(T7, T8, T9 * Ta);
195 Tt = FNMS(T9, T8, T7 * Ta);
196 }
197 {
198 E Te, Tg, Td, Tf;
199 Te = cr[WS(rs, 2)];
200 Tg = ci[WS(rs, 2)];
201 Td = W[2];
202 Tf = W[3];
203 Th = FMA(Td, Te, Tf * Tg);
204 Tv = FNMS(Tf, Te, Td * Tg);
205 }
206 Tu = Ts - Tt;
207 Tx = Tv - Tw;
208 TC = Th - Tm;
209 TB = Tb - T6;
210 TF = Ts + Tt;
211 TG = Tv + Tw;
212 TH = TF + TG;
213 Tc = T6 + Tb;
214 Tn = Th + Tm;
215 To = Tc + Tn;
216 }
217 cr[0] = T1 + To;
218 {
219 E Ty, TA, Tr, Tz, Tp, Tq;
220 Ty = FMA(KP951056516, Tu, KP587785252 * Tx);
221 TA = FNMS(KP587785252, Tu, KP951056516 * Tx);
222 Tp = KP559016994 * (Tc - Tn);
223 Tq = FNMS(KP250000000, To, T1);
224 Tr = Tp + Tq;
225 Tz = Tq - Tp;
226 ci[0] = Tr - Ty;
227 ci[WS(rs, 1)] = Tz + TA;
228 cr[WS(rs, 1)] = Tr + Ty;
229 cr[WS(rs, 2)] = Tz - TA;
230 }
231 ci[WS(rs, 4)] = TH + TE;
232 {
233 E TD, TL, TK, TM, TI, TJ;
234 TD = FMA(KP587785252, TB, KP951056516 * TC);
235 TL = FNMS(KP587785252, TC, KP951056516 * TB);
236 TI = FNMS(KP250000000, TH, TE);
237 TJ = KP559016994 * (TF - TG);
238 TK = TI - TJ;
239 TM = TJ + TI;
240 cr[WS(rs, 3)] = TD - TK;
241 ci[WS(rs, 3)] = TL + TM;
242 ci[WS(rs, 2)] = TD + TK;
243 cr[WS(rs, 4)] = TL - TM;
244 }
245 }
246 }
247 }
248
249 static const tw_instr twinstr[] = {
250 {TW_FULL, 1, 5},
251 {TW_NEXT, 1, 0}
252 };
253
254 static const hc2hc_desc desc = { 5, "hf_5", twinstr, &GENUS, {26, 14, 14, 0} };
255
256 void X(codelet_hf_5) (planner *p) {
257 X(khc2hc_register) (p, hf_5, &desc);
258 }
259 #endif /* HAVE_FMA */