Rev 3 | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
2 | pj | 1 | /* |
2 | * Copyright (c) 1997-1999 Massachusetts Institute of Technology |
||
3 | * |
||
4 | * This program is free software; you can redistribute it and/or modify |
||
5 | * it under the terms of the GNU General Public License as published by |
||
6 | * the Free Software Foundation; either version 2 of the License, or |
||
7 | * (at your option) any later version. |
||
8 | * |
||
9 | * This program is distributed in the hope that it will be useful, |
||
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
||
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
||
12 | * GNU General Public License for more details. |
||
13 | * |
||
14 | * You should have received a copy of the GNU General Public License |
||
15 | * along with this program; if not, write to the Free Software |
||
16 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
||
17 | * |
||
18 | */ |
||
19 | |||
20 | /* This file was automatically generated --- DO NOT EDIT */ |
||
21 | /* Generated on Tue May 18 13:55:58 EDT 1999 */ |
||
22 | |||
107 | pj | 23 | #include <fftw-int.h> |
24 | #include <fftw.h> |
||
2 | pj | 25 | |
26 | /* Generated by: ./genfft -magic-alignment-check -magic-twiddle-load-all -magic-variables 4 -magic-loopi -hc2hc-backward 4 */ |
||
27 | |||
28 | /* |
||
29 | * This function contains 34 FP additions, 18 FP multiplications, |
||
30 | * (or, 28 additions, 12 multiplications, 6 fused multiply/add), |
||
31 | * 15 stack variables, and 32 memory accesses |
||
32 | */ |
||
33 | static const fftw_real K1_414213562 = FFTW_KONST(+1.414213562373095048801688724209698078569671875); |
||
34 | static const fftw_real K2_000000000 = FFTW_KONST(+2.000000000000000000000000000000000000000000000); |
||
35 | |||
36 | /* |
||
37 | * Generator Id's : |
||
107 | pj | 38 | * $Id: fhb_4.c,v 1.2 2003-03-24 11:14:57 pj Exp $ |
39 | * $Id: fhb_4.c,v 1.2 2003-03-24 11:14:57 pj Exp $ |
||
40 | * $Id: fhb_4.c,v 1.2 2003-03-24 11:14:57 pj Exp $ |
||
2 | pj | 41 | */ |
42 | |||
43 | void fftw_hc2hc_backward_4(fftw_real *A, const fftw_complex *W, int iostride, int m, int dist) |
||
44 | { |
||
45 | int i; |
||
46 | fftw_real *X; |
||
47 | fftw_real *Y; |
||
48 | X = A; |
||
49 | Y = A + (4 * iostride); |
||
50 | { |
||
51 | fftw_real tmp39; |
||
52 | fftw_real tmp42; |
||
53 | fftw_real tmp37; |
||
54 | fftw_real tmp40; |
||
55 | ASSERT_ALIGNED_DOUBLE(); |
||
56 | { |
||
57 | fftw_real tmp38; |
||
58 | fftw_real tmp41; |
||
59 | fftw_real tmp35; |
||
60 | fftw_real tmp36; |
||
61 | ASSERT_ALIGNED_DOUBLE(); |
||
62 | tmp38 = X[iostride]; |
||
63 | tmp39 = K2_000000000 * tmp38; |
||
64 | tmp41 = Y[-iostride]; |
||
65 | tmp42 = K2_000000000 * tmp41; |
||
66 | tmp35 = X[0]; |
||
67 | tmp36 = X[2 * iostride]; |
||
68 | tmp37 = tmp35 + tmp36; |
||
69 | tmp40 = tmp35 - tmp36; |
||
70 | } |
||
71 | X[2 * iostride] = tmp37 - tmp39; |
||
72 | X[0] = tmp37 + tmp39; |
||
73 | X[3 * iostride] = tmp40 + tmp42; |
||
74 | X[iostride] = tmp40 - tmp42; |
||
75 | } |
||
76 | X = X + dist; |
||
77 | Y = Y - dist; |
||
78 | for (i = 2; i < m; i = i + 2, X = X + dist, Y = Y - dist, W = W + 3) { |
||
79 | fftw_real tmp9; |
||
80 | fftw_real tmp28; |
||
81 | fftw_real tmp18; |
||
82 | fftw_real tmp25; |
||
83 | fftw_real tmp12; |
||
84 | fftw_real tmp24; |
||
85 | fftw_real tmp21; |
||
86 | fftw_real tmp29; |
||
87 | ASSERT_ALIGNED_DOUBLE(); |
||
88 | { |
||
89 | fftw_real tmp7; |
||
90 | fftw_real tmp8; |
||
91 | fftw_real tmp16; |
||
92 | fftw_real tmp17; |
||
93 | ASSERT_ALIGNED_DOUBLE(); |
||
94 | tmp7 = X[0]; |
||
95 | tmp8 = Y[-2 * iostride]; |
||
96 | tmp9 = tmp7 + tmp8; |
||
97 | tmp28 = tmp7 - tmp8; |
||
98 | tmp16 = Y[0]; |
||
99 | tmp17 = X[2 * iostride]; |
||
100 | tmp18 = tmp16 - tmp17; |
||
101 | tmp25 = tmp16 + tmp17; |
||
102 | } |
||
103 | { |
||
104 | fftw_real tmp10; |
||
105 | fftw_real tmp11; |
||
106 | fftw_real tmp19; |
||
107 | fftw_real tmp20; |
||
108 | ASSERT_ALIGNED_DOUBLE(); |
||
109 | tmp10 = X[iostride]; |
||
110 | tmp11 = Y[-3 * iostride]; |
||
111 | tmp12 = tmp10 + tmp11; |
||
112 | tmp24 = tmp10 - tmp11; |
||
113 | tmp19 = Y[-iostride]; |
||
114 | tmp20 = X[3 * iostride]; |
||
115 | tmp21 = tmp19 - tmp20; |
||
116 | tmp29 = tmp19 + tmp20; |
||
117 | } |
||
118 | X[0] = tmp9 + tmp12; |
||
119 | { |
||
120 | fftw_real tmp14; |
||
121 | fftw_real tmp22; |
||
122 | fftw_real tmp13; |
||
123 | fftw_real tmp15; |
||
124 | ASSERT_ALIGNED_DOUBLE(); |
||
125 | tmp14 = tmp9 - tmp12; |
||
126 | tmp22 = tmp18 - tmp21; |
||
127 | tmp13 = c_re(W[1]); |
||
128 | tmp15 = c_im(W[1]); |
||
129 | X[2 * iostride] = (tmp13 * tmp14) + (tmp15 * tmp22); |
||
130 | Y[-iostride] = (tmp13 * tmp22) - (tmp15 * tmp14); |
||
131 | } |
||
132 | Y[-3 * iostride] = tmp18 + tmp21; |
||
133 | { |
||
134 | fftw_real tmp26; |
||
135 | fftw_real tmp30; |
||
136 | fftw_real tmp23; |
||
137 | fftw_real tmp27; |
||
138 | ASSERT_ALIGNED_DOUBLE(); |
||
139 | tmp26 = tmp24 + tmp25; |
||
140 | tmp30 = tmp28 - tmp29; |
||
141 | tmp23 = c_re(W[0]); |
||
142 | tmp27 = c_im(W[0]); |
||
143 | Y[-2 * iostride] = (tmp23 * tmp26) - (tmp27 * tmp30); |
||
144 | X[iostride] = (tmp27 * tmp26) + (tmp23 * tmp30); |
||
145 | } |
||
146 | { |
||
147 | fftw_real tmp32; |
||
148 | fftw_real tmp34; |
||
149 | fftw_real tmp31; |
||
150 | fftw_real tmp33; |
||
151 | ASSERT_ALIGNED_DOUBLE(); |
||
152 | tmp32 = tmp25 - tmp24; |
||
153 | tmp34 = tmp28 + tmp29; |
||
154 | tmp31 = c_re(W[2]); |
||
155 | tmp33 = c_im(W[2]); |
||
156 | Y[0] = (tmp31 * tmp32) - (tmp33 * tmp34); |
||
157 | X[3 * iostride] = (tmp33 * tmp32) + (tmp31 * tmp34); |
||
158 | } |
||
159 | } |
||
160 | if (i == m) { |
||
161 | fftw_real tmp1; |
||
162 | fftw_real tmp2; |
||
163 | fftw_real tmp3; |
||
164 | fftw_real tmp4; |
||
165 | fftw_real tmp5; |
||
166 | fftw_real tmp6; |
||
167 | ASSERT_ALIGNED_DOUBLE(); |
||
168 | tmp1 = X[0]; |
||
169 | tmp2 = X[iostride]; |
||
170 | tmp3 = tmp1 - tmp2; |
||
171 | tmp4 = Y[0]; |
||
172 | tmp5 = Y[-iostride]; |
||
173 | tmp6 = tmp4 + tmp5; |
||
174 | X[0] = K2_000000000 * (tmp1 + tmp2); |
||
175 | X[2 * iostride] = -(K2_000000000 * (tmp4 - tmp5)); |
||
176 | X[iostride] = K1_414213562 * (tmp3 - tmp6); |
||
177 | X[3 * iostride] = -(K1_414213562 * (tmp3 + tmp6)); |
||
178 | } |
||
179 | } |
||
180 | |||
181 | static const int twiddle_order[] = |
||
182 | {1, 2, 3}; |
||
183 | fftw_codelet_desc fftw_hc2hc_backward_4_desc = |
||
184 | { |
||
185 | "fftw_hc2hc_backward_4", |
||
186 | (void (*)()) fftw_hc2hc_backward_4, |
||
187 | 4, |
||
188 | FFTW_BACKWARD, |
||
189 | FFTW_HC2HC, |
||
190 | 102, |
||
191 | 3, |
||
192 | twiddle_order, |
||
193 | }; |