Mercurial > hg > sv-dependency-builds

comparison src/fftw-3.3.8/rdft/scalar/r2cf/hc2cfdft_2.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 08:07:10 EDT 2018 */
23
24 #include "rdft/codelet-rdft.h"
25
26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
27
28 /* Generated by: ../../../genfft/gen_hc2cdft.native -fma -compact -variables 4 -pipeline-latency 4 -n 2 -dit -name hc2cfdft_2 -include rdft/scalar/hc2cf.h */
29
30 /*
31 * This function contains 10 FP additions, 8 FP multiplications,
32 * (or, 8 additions, 6 multiplications, 2 fused multiply/add),
33 * 16 stack variables, 1 constants, and 8 memory accesses
34 */
35 #include "rdft/scalar/hc2cf.h"
36
37 static void hc2cfdft_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
38 {
39 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
40 {
41 INT m;
42 for (m = mb, W = W + ((mb - 1) * 2); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 2, MAKE_VOLATILE_STRIDE(8, rs)) {
43 E T3, Ta, Tc, T9, Td, T4, T8, Tb, Te;
44 {
45 E T1, T2, T5, T6, T7;
46 T1 = Ip[0];
47 T2 = Im[0];
48 T3 = T1 - T2;
49 Ta = T1 + T2;
50 T5 = Rm[0];
51 T6 = Rp[0];
52 T7 = T5 - T6;
53 Tc = T6 + T5;
54 T9 = W[1];
55 Td = T9 * T7;
56 T4 = W[0];
57 T8 = T4 * T7;
58 }
59 Tb = FNMS(T9, Ta, T8);
60 Ip[0] = KP500000000 * (T3 + Tb);
61 Im[0] = KP500000000 * (Tb - T3);
62 Te = FMA(T4, Ta, Td);
63 Rm[0] = KP500000000 * (Tc - Te);
64 Rp[0] = KP500000000 * (Tc + Te);
65 }
66 }
67 }
68
69 static const tw_instr twinstr[] = {
70 {TW_FULL, 1, 2},
71 {TW_NEXT, 1, 0}
72 };
73
74 static const hc2c_desc desc = { 2, "hc2cfdft_2", twinstr, &GENUS, {8, 6, 2, 0} };
75
76 void X(codelet_hc2cfdft_2) (planner *p) {
77 X(khc2c_register) (p, hc2cfdft_2, &desc, HC2C_VIA_DFT);
78 }
79 #else
80
81 /* Generated by: ../../../genfft/gen_hc2cdft.native -compact -variables 4 -pipeline-latency 4 -n 2 -dit -name hc2cfdft_2 -include rdft/scalar/hc2cf.h */
82
83 /*
84 * This function contains 10 FP additions, 8 FP multiplications,
85 * (or, 8 additions, 6 multiplications, 2 fused multiply/add),
86 * 10 stack variables, 1 constants, and 8 memory accesses
87 */
88 #include "rdft/scalar/hc2cf.h"
89
90 static void hc2cfdft_2(R *Rp, R *Ip, R *Rm, R *Im, const R *W, stride rs, INT mb, INT me, INT ms)
91 {
92 DK(KP500000000, +0.500000000000000000000000000000000000000000000);
93 {
94 INT m;
95 for (m = mb, W = W + ((mb - 1) * 2); m < me; m = m + 1, Rp = Rp + ms, Ip = Ip + ms, Rm = Rm - ms, Im = Im - ms, W = W + 2, MAKE_VOLATILE_STRIDE(8, rs)) {
96 E T3, T9, T7, Tb;
97 {
98 E T1, T2, T5, T6;
99 T1 = Ip[0];
100 T2 = Im[0];
101 T3 = T1 - T2;
102 T9 = T1 + T2;
103 T5 = Rm[0];
104 T6 = Rp[0];
105 T7 = T5 - T6;
106 Tb = T6 + T5;
107 }
108 {
109 E Ta, Tc, T4, T8;
110 T4 = W[0];
111 T8 = W[1];
112 Ta = FNMS(T8, T9, T4 * T7);
113 Tc = FMA(T8, T7, T4 * T9);
114 Ip[0] = KP500000000 * (T3 + Ta);
115 Rp[0] = KP500000000 * (Tb + Tc);
116 Im[0] = KP500000000 * (Ta - T3);
117 Rm[0] = KP500000000 * (Tb - Tc);
118 }
119 }
120 }
121 }
122
123 static const tw_instr twinstr[] = {
124 {TW_FULL, 1, 2},
125 {TW_NEXT, 1, 0}
126 };
127
128 static const hc2c_desc desc = { 2, "hc2cfdft_2", twinstr, &GENUS, {8, 6, 2, 0} };
129
130 void X(codelet_hc2cfdft_2) (planner *p) {
131 X(khc2c_register) (p, hc2cfdft_2, &desc, HC2C_VIA_DFT);
132 }
133 #endif