Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cf/hc2cf2_4.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 13:49:24 EST 2014 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2c.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cf2_4 -include hc2cf.h */
29
30 /*
31 * This function contains 24 FP additions, 16 FP multiplications,
32 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
33 * 33 stack variables, 0 constants, and 16 memory accesses
34 */
35 #include "hc2cf.h"
36
37 static void hc2cf2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 {
40 INT m;
41 for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(16, rs)) {
42 E Ti, Tq, To, Te, Ty, TA, Tm, Ts;
43 {
44 E T2, T6, T3, T5;
45 T2 = W[0];
46 T6 = W[3];
47 T3 = W[2];
48 T5 = W[1];
49 {
50 E T1, Tx, Td, Tw, Tj, Tl, Ta, T4, Tk, Tr;
51 T1 = Rp[0];
52 Ta = T2 * T6;
53 T4 = T2 * T3;
54 Tx = Rm[0];
55 {
56 E T8, Tb, T7, Tc;
57 T8 = Rp[WS(rs, 1)];
58 Tb = FNMS(T5, T3, Ta);
59 T7 = FMA(T5, T6, T4);
60 Tc = Rm[WS(rs, 1)];
61 {
62 E Tf, Th, T9, Tv, Tg, Tp;
63 Tf = Ip[0];
64 Th = Im[0];
65 T9 = T7 * T8;
66 Tv = T7 * Tc;
67 Tg = T2 * Tf;
68 Tp = T2 * Th;
69 Td = FMA(Tb, Tc, T9);
70 Tw = FNMS(Tb, T8, Tv);
71 Ti = FMA(T5, Th, Tg);
72 Tq = FNMS(T5, Tf, Tp);
73 }
74 Tj = Ip[WS(rs, 1)];
75 Tl = Im[WS(rs, 1)];
76 }
77 To = T1 - Td;
78 Te = T1 + Td;
79 Ty = Tw + Tx;
80 TA = Tx - Tw;
81 Tk = T3 * Tj;
82 Tr = T3 * Tl;
83 Tm = FMA(T6, Tl, Tk);
84 Ts = FNMS(T6, Tj, Tr);
85 }
86 }
87 {
88 E Tn, Tz, Tu, Tt;
89 Tn = Ti + Tm;
90 Tz = Tm - Ti;
91 Tu = Tq + Ts;
92 Tt = Tq - Ts;
93 Ip[WS(rs, 1)] = Tz + TA;
94 Im[0] = Tz - TA;
95 Rp[0] = Te + Tn;
96 Rm[WS(rs, 1)] = Te - Tn;
97 Rp[WS(rs, 1)] = To + Tt;
98 Rm[0] = To - Tt;
99 Ip[0] = Tu + Ty;
100 Im[WS(rs, 1)] = Tu - Ty;
101 }
102 }
103 }
104 }
105
106 static const tw_instr twinstr[] = {
107 {TW_CEXP, 1, 1},
108 {TW_CEXP, 1, 3},
109 {TW_NEXT, 1, 0}
110 };
111
112 static const hc2c_desc desc = { 4, "hc2cf2_4", twinstr, &GENUS, {16, 8, 8, 0} };
113
114 void X(codelet_hc2cf2_4) (planner *p) {
115 X(khc2c_register) (p, hc2cf2_4, &desc, HC2C_VIA_RDFT);
116 }
117 #else /* HAVE_FMA */
118
119 /* Generated by: ../../../genfft/gen_hc2c.native -compact -variables 4 -pipeline-latency 4 -twiddle-log3 -precompute-twiddles -n 4 -dit -name hc2cf2_4 -include hc2cf.h */
120
121 /*
122 * This function contains 24 FP additions, 16 FP multiplications,
123 * (or, 16 additions, 8 multiplications, 8 fused multiply/add),
124 * 21 stack variables, 0 constants, and 16 memory accesses
125 */
126 #include "hc2cf.h"
127
128 static void hc2cf2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
129 {
130 {
131 INT m;
132 for (m = mb, W = W + ((mb - 1) * 4); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 4, MAKE_VOLATILE_STRIDE(16, rs)) {
133 E T2, T4, T3, T5, T6, T8;
134 T2 = W[0];
135 T4 = W[1];
136 T3 = W[2];
137 T5 = W[3];
138 T6 = FMA(T2, T3, T4 * T5);
139 T8 = FNMS(T4, T3, T2 * T5);
140 {
141 E T1, Tp, Ta, To, Te, Tk, Th, Tl, T7, T9;
142 T1 = Rp[0];
143 Tp = Rm[0];
144 T7 = Rp[WS(rs, 1)];
145 T9 = Rm[WS(rs, 1)];
146 Ta = FMA(T6, T7, T8 * T9);
147 To = FNMS(T8, T7, T6 * T9);
148 {
149 E Tc, Td, Tf, Tg;
150 Tc = Ip[0];
151 Td = Im[0];
152 Te = FMA(T2, Tc, T4 * Td);
153 Tk = FNMS(T4, Tc, T2 * Td);
154 Tf = Ip[WS(rs, 1)];
155 Tg = Im[WS(rs, 1)];
156 Th = FMA(T3, Tf, T5 * Tg);
157 Tl = FNMS(T5, Tf, T3 * Tg);
158 }
159 {
160 E Tb, Ti, Tn, Tq;
161 Tb = T1 + Ta;
162 Ti = Te + Th;
163 Rm[WS(rs, 1)] = Tb - Ti;
164 Rp[0] = Tb + Ti;
165 Tn = Tk + Tl;
166 Tq = To + Tp;
167 Im[WS(rs, 1)] = Tn - Tq;
168 Ip[0] = Tn + Tq;
169 }
170 {
171 E Tj, Tm, Tr, Ts;
172 Tj = T1 - Ta;
173 Tm = Tk - Tl;
174 Rm[0] = Tj - Tm;
175 Rp[WS(rs, 1)] = Tj + Tm;
176 Tr = Th - Te;
177 Ts = Tp - To;
178 Im[0] = Tr - Ts;
179 Ip[WS(rs, 1)] = Tr + Ts;
180 }
181 }
182 }
183 }
184 }
185
186 static const tw_instr twinstr[] = {
187 {TW_CEXP, 1, 1},
188 {TW_CEXP, 1, 3},
189 {TW_NEXT, 1, 0}
190 };
191
192 static const hc2c_desc desc = { 4, "hc2cf2_4", twinstr, &GENUS, {16, 8, 8, 0} };
193
194 void X(codelet_hc2cf2_4) (planner *p) {
195 X(khc2c_register) (p, hc2cf2_4, &desc, HC2C_VIA_RDFT);
196 }
197 #endif /* HAVE_FMA */