Mercurial > hg > js-dsp-test

comparison fft/fftw/fftw-3.3.4/rdft/scalar/r2cb/hc2cbdft2_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 13:50:45 EST 2014 */
23
24 #include "codelet-rdft.h"
25
26 #ifdef HAVE_FMA
27
28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -reorder-insns -schedule-for-pipeline -compact -variables 4 -pipeline-latency 4 -sign 1 -n 4 -dif -name hc2cbdft2_4 -include hc2cb.h */
29
30 /*
31 * This function contains 30 FP additions, 12 FP multiplications,
32 * (or, 24 additions, 6 multiplications, 6 fused multiply/add),
33 * 35 stack variables, 0 constants, and 16 memory accesses
34 */
35 #include "hc2cb.h"
36
37 static void hc2cbdft2_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) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
42 E Ty, TB, Tw, TE, TA, TF, Tz, TG, TC;
43 {
44 E T4, Tg, T3, Tm, Tc, T5, Th, Ti;
45 {
46 E T1, T2, Ta, Tb;
47 T1 = Rp[0];
48 T2 = Rm[WS(rs, 1)];
49 Ta = Ip[0];
50 Tb = Im[WS(rs, 1)];
51 T4 = Rp[WS(rs, 1)];
52 Tg = T1 - T2;
53 T3 = T1 + T2;
54 Tm = Ta - Tb;
55 Tc = Ta + Tb;
56 T5 = Rm[0];
57 Th = Ip[WS(rs, 1)];
58 Ti = Im[0];
59 }
60 {
61 E T8, Td, T7, Ts, To, Tv, Tk, Te, Tf;
62 T8 = W[0];
63 {
64 E T9, T6, Tn, Tj;
65 T9 = T4 - T5;
66 T6 = T4 + T5;
67 Tn = Th - Ti;
68 Tj = Th + Ti;
69 Ty = Tc - T9;
70 Td = T9 + Tc;
71 T7 = T3 + T6;
72 Ts = T3 - T6;
73 To = Tm + Tn;
74 Tv = Tm - Tn;
75 TB = Tg + Tj;
76 Tk = Tg - Tj;
77 Te = T8 * Td;
78 }
79 Tf = W[1];
80 {
81 E Tr, Tu, Tt, TD, Tx, Tp, Tl, Tq;
82 Tr = W[2];
83 Tp = T8 * Tk;
84 Tu = W[3];
85 Tl = FMA(Tf, Tk, Te);
86 Tt = Tr * Ts;
87 Tq = FNMS(Tf, Td, Tp);
88 TD = Tu * Ts;
89 Rm[0] = T7 + Tl;
90 Rp[0] = T7 - Tl;
91 Im[0] = Tq - To;
92 Ip[0] = To + Tq;
93 Tx = W[4];
94 Tw = FNMS(Tu, Tv, Tt);
95 TE = FMA(Tr, Tv, TD);
96 TA = W[5];
97 TF = Tx * TB;
98 Tz = Tx * Ty;
99 }
100 }
101 }
102 TG = FNMS(TA, Ty, TF);
103 TC = FMA(TA, TB, Tz);
104 Im[WS(rs, 1)] = TG - TE;
105 Ip[WS(rs, 1)] = TE + TG;
106 Rm[WS(rs, 1)] = Tw + TC;
107 Rp[WS(rs, 1)] = Tw - TC;
108 }
109 }
110 }
111
112 static const tw_instr twinstr[] = {
113 {TW_FULL, 1, 4},
114 {TW_NEXT, 1, 0}
115 };
116
117 static const hc2c_desc desc = { 4, "hc2cbdft2_4", twinstr, &GENUS, {24, 6, 6, 0} };
118
119 void X(codelet_hc2cbdft2_4) (planner *p) {
120 X(khc2c_register) (p, hc2cbdft2_4, &desc, HC2C_VIA_DFT);
121 }
122 #else /* HAVE_FMA */
123
124 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -sign 1 -n 4 -dif -name hc2cbdft2_4 -include hc2cb.h */
125
126 /*
127 * This function contains 30 FP additions, 12 FP multiplications,
128 * (or, 24 additions, 6 multiplications, 6 fused multiply/add),
129 * 19 stack variables, 0 constants, and 16 memory accesses
130 */
131 #include "hc2cb.h"
132
133 static void hc2cbdft2_4(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
134 {
135 {
136 INT m;
137 for (m = mb, W = W + ((mb - 1) * 6); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 6, MAKE_VOLATILE_STRIDE(16, rs)) {
138 E T3, Tl, T6, Tm, Td, Tj, Tx, Tv, Ts, Tq;
139 {
140 E Tf, Tc, T9, Ti;
141 {
142 E T1, T2, Ta, Tb;
143 T1 = Rp[0];
144 T2 = Rm[WS(rs, 1)];
145 T3 = T1 + T2;
146 Tf = T1 - T2;
147 Ta = Ip[0];
148 Tb = Im[WS(rs, 1)];
149 Tc = Ta + Tb;
150 Tl = Ta - Tb;
151 }
152 {
153 E T4, T5, Tg, Th;
154 T4 = Rp[WS(rs, 1)];
155 T5 = Rm[0];
156 T6 = T4 + T5;
157 T9 = T4 - T5;
158 Tg = Ip[WS(rs, 1)];
159 Th = Im[0];
160 Ti = Tg + Th;
161 Tm = Tg - Th;
162 }
163 Td = T9 + Tc;
164 Tj = Tf - Ti;
165 Tx = Tf + Ti;
166 Tv = Tc - T9;
167 Ts = Tl - Tm;
168 Tq = T3 - T6;
169 }
170 {
171 E T7, Tn, Tk, To, T8, Te;
172 T7 = T3 + T6;
173 Tn = Tl + Tm;
174 T8 = W[0];
175 Te = W[1];
176 Tk = FMA(T8, Td, Te * Tj);
177 To = FNMS(Te, Td, T8 * Tj);
178 Rp[0] = T7 - Tk;
179 Ip[0] = Tn + To;
180 Rm[0] = T7 + Tk;
181 Im[0] = To - Tn;
182 }
183 {
184 E Tt, Tz, Ty, TA;
185 {
186 E Tp, Tr, Tu, Tw;
187 Tp = W[2];
188 Tr = W[3];
189 Tt = FNMS(Tr, Ts, Tp * Tq);
190 Tz = FMA(Tr, Tq, Tp * Ts);
191 Tu = W[4];
192 Tw = W[5];
193 Ty = FMA(Tu, Tv, Tw * Tx);
194 TA = FNMS(Tw, Tv, Tu * Tx);
195 }
196 Rp[WS(rs, 1)] = Tt - Ty;
197 Ip[WS(rs, 1)] = Tz + TA;
198 Rm[WS(rs, 1)] = Tt + Ty;
199 Im[WS(rs, 1)] = TA - Tz;
200 }
201 }
202 }
203 }
204
205 static const tw_instr twinstr[] = {
206 {TW_FULL, 1, 4},
207 {TW_NEXT, 1, 0}
208 };
209
210 static const hc2c_desc desc = { 4, "hc2cbdft2_4", twinstr, &GENUS, {24, 6, 6, 0} };
211
212 void X(codelet_hc2cbdft2_4) (planner *p) {
213 X(khc2c_register) (p, hc2cbdft2_4, &desc, HC2C_VIA_DFT);
214 }
215 #endif /* HAVE_FMA */