Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cb/hb2_5.c @ 19:26056e866c29

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW to comparison table
author Chris Cannam
date Tue, 06 Oct 2015 13:08:39 +0100
parents
children
comparison
equal deleted inserted replaced
18:8db794ca3e0b 19:26056e866c29
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 Tue Mar 4 13:50:29 EST 2014 */
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 -sign 1 -twiddle-log3 -precompute-twiddles -n 5 -dif -name hb2_5 -include hb.h */
29
30 /*
31 * This function contains 44 FP additions, 40 FP multiplications,
32 * (or, 14 additions, 10 multiplications, 30 fused multiply/add),
33 * 51 stack variables, 4 constants, and 20 memory accesses
34 */
35 #include "hb.h"
36
37 static void hb2_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) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
46 E T9, TB, Tz, Tm, T1, TG, TO, TJ, TC, Tn, Tg, To, Tf, Tw, TQ;
47 E T8, Tb, Th, Ta, Ti, Tp;
48 T9 = W[0];
49 TB = W[3];
50 Tz = W[2];
51 Tm = W[1];
52 {
53 E T4, Tu, T5, T6;
54 T1 = cr[0];
55 {
56 E TF, TA, T2, T3;
57 TF = T9 * TB;
58 TA = T9 * Tz;
59 T2 = cr[WS(rs, 1)];
60 T3 = ci[0];
61 TG = FMA(Tm, Tz, TF);
62 TO = FNMS(Tm, Tz, TF);
63 TJ = FMA(Tm, TB, TA);
64 TC = FNMS(Tm, TB, TA);
65 T4 = T2 + T3;
66 Tu = T2 - T3;
67 T5 = cr[WS(rs, 2)];
68 T6 = ci[WS(rs, 1)];
69 }
70 Tn = ci[WS(rs, 4)];
71 {
72 E Td, Te, T7, Tv;
73 Td = ci[WS(rs, 3)];
74 Te = cr[WS(rs, 4)];
75 T7 = T5 + T6;
76 Tv = T5 - T6;
77 Tg = ci[WS(rs, 2)];
78 To = Td - Te;
79 Tf = Td + Te;
80 Tw = FMA(KP618033988, Tv, Tu);
81 TQ = FNMS(KP618033988, Tu, Tv);
82 T8 = T4 + T7;
83 Tb = T4 - T7;
84 Th = cr[WS(rs, 3)];
85 }
86 }
87 cr[0] = T1 + T8;
88 Ta = FNMS(KP250000000, T8, T1);
89 Ti = Tg + Th;
90 Tp = Tg - Th;
91 {
92 E Tc, TK, Ts, Tq;
93 Tc = FMA(KP559016994, Tb, Ta);
94 TK = FNMS(KP559016994, Tb, Ta);
95 Ts = To - Tp;
96 Tq = To + Tp;
97 {
98 E Tj, TL, Tr, TM, TT;
99 Tj = FMA(KP618033988, Ti, Tf);
100 TL = FNMS(KP618033988, Tf, Ti);
101 ci[0] = Tn + Tq;
102 Tr = FNMS(KP250000000, Tq, Tn);
103 TM = FMA(KP951056516, TL, TK);
104 TT = FNMS(KP951056516, TL, TK);
105 {
106 E Tk, TD, Tt, TP;
107 Tk = FNMS(KP951056516, Tj, Tc);
108 TD = FMA(KP951056516, Tj, Tc);
109 Tt = FMA(KP559016994, Ts, Tr);
110 TP = FNMS(KP559016994, Ts, Tr);
111 {
112 E TW, TU, TS, TN;
113 TW = TB * TT;
114 TU = Tz * TT;
115 TS = TO * TM;
116 TN = TJ * TM;
117 {
118 E TI, TE, Ty, Tl;
119 TI = TG * TD;
120 TE = TC * TD;
121 Ty = Tm * Tk;
122 Tl = T9 * Tk;
123 {
124 E TR, TV, Tx, TH;
125 TR = FNMS(KP951056516, TQ, TP);
126 TV = FMA(KP951056516, TQ, TP);
127 Tx = FMA(KP951056516, Tw, Tt);
128 TH = FNMS(KP951056516, Tw, Tt);
129 ci[WS(rs, 3)] = FMA(Tz, TV, TW);
130 cr[WS(rs, 3)] = FNMS(TB, TV, TU);
131 ci[WS(rs, 2)] = FMA(TJ, TR, TS);
132 cr[WS(rs, 2)] = FNMS(TO, TR, TN);
133 ci[WS(rs, 4)] = FMA(TC, TH, TI);
134 cr[WS(rs, 4)] = FNMS(TG, TH, TE);
135 ci[WS(rs, 1)] = FMA(T9, Tx, Ty);
136 cr[WS(rs, 1)] = FNMS(Tm, Tx, Tl);
137 }
138 }
139 }
140 }
141 }
142 }
143 }
144 }
145 }
146
147 static const tw_instr twinstr[] = {
148 {TW_CEXP, 1, 1},
149 {TW_CEXP, 1, 3},
150 {TW_NEXT, 1, 0}
151 };
152
153 static const hc2hc_desc desc = { 5, "hb2_5", twinstr, &GENUS, {14, 10, 30, 0} };
154
155 void X(codelet_hb2_5) (planner *p) {
156 X(khc2hc_register) (p, hb2_5, &desc);
157 }
158 #else /* HAVE_FMA */
159
160 /* Generated by: ../../../genfft/gen_hc2hc.native -compact -variables 4 -pipeline-latency 4 -sign 1 -twiddle-log3 -precompute-twiddles -n 5 -dif -name hb2_5 -include hb.h */
161
162 /*
163 * This function contains 44 FP additions, 32 FP multiplications,
164 * (or, 30 additions, 18 multiplications, 14 fused multiply/add),
165 * 33 stack variables, 4 constants, and 20 memory accesses
166 */
167 #include "hb.h"
168
169 static void hb2_5(R *cr, R *ci, const R *W, stride rs, INT mb, INT me, INT ms)
170 {
171 DK(KP250000000, +0.250000000000000000000000000000000000000000000);
172 DK(KP587785252, +0.587785252292473129168705954639072768597652438);
173 DK(KP951056516, +0.951056516295153572116439333379382143405698634);
174 DK(KP559016994, +0.559016994374947424102293417182819058860154590);
175 {
176 INT m;
177 for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, cr = cr + ms, ci = ci - ms, W = W + 4, MAKE_VOLATILE_STRIDE(10, rs)) {
178 E Th, Tk, Ti, Tl, Tn, TP, Tx, TN;
179 {
180 E Tj, Tw, Tm, Tv;
181 Th = W[0];
182 Tk = W[1];
183 Ti = W[2];
184 Tl = W[3];
185 Tj = Th * Ti;
186 Tw = Tk * Ti;
187 Tm = Tk * Tl;
188 Tv = Th * Tl;
189 Tn = Tj + Tm;
190 TP = Tv + Tw;
191 Tx = Tv - Tw;
192 TN = Tj - Tm;
193 }
194 {
195 E T1, Tp, TK, TA, T8, To, T9, Tt, TI, TC, Tg, TB;
196 {
197 E T4, Ty, T7, Tz;
198 T1 = cr[0];
199 {
200 E T2, T3, T5, T6;
201 T2 = cr[WS(rs, 1)];
202 T3 = ci[0];
203 T4 = T2 + T3;
204 Ty = T2 - T3;
205 T5 = cr[WS(rs, 2)];
206 T6 = ci[WS(rs, 1)];
207 T7 = T5 + T6;
208 Tz = T5 - T6;
209 }
210 Tp = KP559016994 * (T4 - T7);
211 TK = FMA(KP951056516, Ty, KP587785252 * Tz);
212 TA = FNMS(KP951056516, Tz, KP587785252 * Ty);
213 T8 = T4 + T7;
214 To = FNMS(KP250000000, T8, T1);
215 }
216 {
217 E Tc, Tr, Tf, Ts;
218 T9 = ci[WS(rs, 4)];
219 {
220 E Ta, Tb, Td, Te;
221 Ta = ci[WS(rs, 3)];
222 Tb = cr[WS(rs, 4)];
223 Tc = Ta - Tb;
224 Tr = Ta + Tb;
225 Td = ci[WS(rs, 2)];
226 Te = cr[WS(rs, 3)];
227 Tf = Td - Te;
228 Ts = Td + Te;
229 }
230 Tt = FNMS(KP951056516, Ts, KP587785252 * Tr);
231 TI = FMA(KP951056516, Tr, KP587785252 * Ts);
232 TC = KP559016994 * (Tc - Tf);
233 Tg = Tc + Tf;
234 TB = FNMS(KP250000000, Tg, T9);
235 }
236 cr[0] = T1 + T8;
237 ci[0] = T9 + Tg;
238 {
239 E Tu, TF, TE, TG, Tq, TD;
240 Tq = To - Tp;
241 Tu = Tq - Tt;
242 TF = Tq + Tt;
243 TD = TB - TC;
244 TE = TA + TD;
245 TG = TD - TA;
246 cr[WS(rs, 2)] = FNMS(Tx, TE, Tn * Tu);
247 ci[WS(rs, 2)] = FMA(Tn, TE, Tx * Tu);
248 cr[WS(rs, 3)] = FNMS(Tl, TG, Ti * TF);
249 ci[WS(rs, 3)] = FMA(Ti, TG, Tl * TF);
250 }
251 {
252 E TJ, TO, TM, TQ, TH, TL;
253 TH = Tp + To;
254 TJ = TH - TI;
255 TO = TH + TI;
256 TL = TC + TB;
257 TM = TK + TL;
258 TQ = TL - TK;
259 cr[WS(rs, 1)] = FNMS(Tk, TM, Th * TJ);
260 ci[WS(rs, 1)] = FMA(Th, TM, Tk * TJ);
261 cr[WS(rs, 4)] = FNMS(TP, TQ, TN * TO);
262 ci[WS(rs, 4)] = FMA(TN, TQ, TP * TO);
263 }
264 }
265 }
266 }
267 }
268
269 static const tw_instr twinstr[] = {
270 {TW_CEXP, 1, 1},
271 {TW_CEXP, 1, 3},
272 {TW_NEXT, 1, 0}
273 };
274
275 static const hc2hc_desc desc = { 5, "hb2_5", twinstr, &GENUS, {30, 18, 14, 0} };
276
277 void X(codelet_hb2_5) (planner *p) {
278 X(khc2hc_register) (p, hb2_5, &desc);
279 }
280 #endif /* HAVE_FMA */