PandA-2024.02
examples
CHStone
CHStone
jpeg
chenidct.c
Go to the documentation of this file.
1
/*
2
+--------------------------------------------------------------------------+
3
| CHStone : A suite of Benchmark Programs for C-based High-Level Synthesis |
4
| ======================================================================== |
5
| |
6
| * Collected and Modified : Y. Hara, H. Tomiyama, S. Honda, |
7
| H. Takada and K. Ishii |
8
| Nagoya University, Japan |
9
| |
10
| * Remarks : |
11
| 1. This source code is reformatted to follow CHStone's style. |
12
| 2. Test vectors are added for CHStone. |
13
| 3. If "main_result" is 0 at the end of the program, the program is |
14
| successfully executed. |
15
| 4. Follow the copyright of each benchmark program. |
16
+--------------------------------------------------------------------------+
17
*/
18
/*
19
* IDCT transformation of Chen algorithm
20
*
21
* @(#) $Id: chenidct.c,v 1.2 2003/07/18 10:19:21 honda Exp $
22
*/
23
/*************************************************************
24
Copyright (C) 1990, 1991, 1993 Andy C. Hung, all rights reserved.
25
PUBLIC DOMAIN LICENSE: Stanford University Portable Video Research
26
Group. If you use this software, you agree to the following: This
27
program package is purely experimental, and is licensed "as is".
28
Permission is granted to use, modify, and distribute this program
29
without charge for any purpose, provided this license/ disclaimer
30
notice appears in the copies. No warranty or maintenance is given,
31
either expressed or implied. In no event shall the author(s) be
32
liable to you or a third party for any special, incidental,
33
consequential, or other damages, arising out of the use or inability
34
to use the program for any purpose (or the loss of data), even if we
35
have been advised of such possibilities. Any public reference or
36
advertisement of this source code should refer to it as the Portable
37
Video Research Group (PVRG) code, and not by any author(s) (or
38
Stanford University) name.
39
*************************************************************/
40
/*
41
************************************************************
42
chendct.c
43
44
A simple DCT algorithm that seems to have fairly nice arithmetic
45
properties.
46
47
W. H. Chen, C. H. Smith and S. C. Fralick "A fast computational
48
algorithm for the discrete cosine transform," IEEE Trans. Commun.,
49
vol. COM-25, pp. 1004-1009, Sept 1977.
50
51
************************************************************
52
*/
53
54
#define LS(r,s) ((r) << (s))
55
#define RS(r,s) ((r) >> (s))
/* Caution with rounding... */
56
57
#define MSCALE(expr) RS((expr),9)
58
59
/* Cos constants */
60
61
#define c1d4 362L
62
63
#define c1d8 473L
64
#define c3d8 196L
65
66
#define c1d16 502L
67
#define c3d16 426L
68
#define c5d16 284L
69
#define c7d16 100L
70
71
72
/*
73
*
74
* ChenIDCT() implements the Chen inverse dct. Note that there are two
75
* input vectors that represent x=input, and y=output, and must be
76
* defined (and storage allocated) before this routine is called.
77
*/
78
void
79
ChenIDct
(
int
*
x
,
int
*y)
80
{
81
register
int
i;
82
register
int
*aptr;
83
register
int
a0
,
a1
,
a2
,
a3
;
84
register
int
b0
,
b1
,
b2
, b3;
85
register
int
c0
,
c1
,
c2
, c3;
86
87
/* Loop over columns */
88
89
for
(i = 0; i < 8; i++)
90
{
91
aptr = x + i;
92
b0 =
LS
(*aptr, 2);
93
aptr += 8;
94
a0 =
LS
(*aptr, 2);
95
aptr += 8;
96
b2 =
LS
(*aptr, 2);
97
aptr += 8;
98
a1 =
LS
(*aptr, 2);
99
aptr += 8;
100
b1 =
LS
(*aptr, 2);
101
aptr += 8;
102
a2 =
LS
(*aptr, 2);
103
aptr += 8;
104
b3 =
LS
(*aptr, 2);
105
aptr += 8;
106
a3 =
LS
(*aptr, 2);
107
108
/* Split into even mode b0 = x0 b1 = x4 b2 = x2 b3 = x6.
109
And the odd terms a0 = x1 a1 = x3 a2 = x5 a3 = x7.
110
*/
111
112
c0 =
MSCALE
((
c7d16
* a0) - (
c1d16
* a3));
113
c1 =
MSCALE
((
c3d16
* a2) - (
c5d16
* a1));
114
c2 =
MSCALE
((
c3d16
* a1) + (
c5d16
* a2));
115
c3 =
MSCALE
((
c1d16
* a0) + (
c7d16
* a3));
116
117
/* First Butterfly on even terms. */
118
119
a0 =
MSCALE
(
c1d4
* (b0 + b1));
120
a1 =
MSCALE
(
c1d4
* (b0 - b1));
121
122
a2 =
MSCALE
((
c3d8
* b2) - (
c1d8
* b3));
123
a3 =
MSCALE
((
c1d8
* b2) + (
c3d8
* b3));
124
125
b0 = a0 +
a3
;
126
b1 = a1 +
a2
;
127
b2 = a1 -
a2
;
128
b3 = a0 -
a3
;
129
130
/* Second Butterfly */
131
132
a0 = c0 +
c1
;
133
a1 = c0 -
c1
;
134
a2 = c3 -
c2
;
135
a3 = c3 +
c2
;
136
137
c0 =
a0
;
138
c1 =
MSCALE
(
c1d4
* (a2 - a1));
139
c2 =
MSCALE
(
c1d4
* (a2 + a1));
140
c3 =
a3
;
141
142
aptr = y + i;
143
*aptr = b0 + c3;
144
aptr += 8;
145
*aptr = b1 +
c2
;
146
aptr += 8;
147
*aptr = b2 +
c1
;
148
aptr += 8;
149
*aptr = b3 +
c0
;
150
aptr += 8;
151
*aptr = b3 -
c0
;
152
aptr += 8;
153
*aptr = b2 -
c1
;
154
aptr += 8;
155
*aptr = b1 -
c2
;
156
aptr += 8;
157
*aptr = b0 - c3;
158
}
159
160
/* Loop over rows */
161
162
for
(i = 0; i < 8; i++)
163
{
164
aptr = y +
LS
(i, 3);
165
b0 = *(aptr++);
166
a0 = *(aptr++);
167
b2 = *(aptr++);
168
a1 = *(aptr++);
169
b1 = *(aptr++);
170
a2 = *(aptr++);
171
b3 = *(aptr++);
172
a3 = *(aptr);
173
174
/*
175
Split into even mode b0 = x0 b1 = x4 b2 = x2 b3 = x6.
176
And the odd terms a0 = x1 a1 = x3 a2 = x5 a3 = x7.
177
*/
178
179
c0 =
MSCALE
((
c7d16
* a0) - (
c1d16
* a3));
180
c1 =
MSCALE
((
c3d16
* a2) - (
c5d16
* a1));
181
c2 =
MSCALE
((
c3d16
* a1) + (
c5d16
* a2));
182
c3 =
MSCALE
((
c1d16
* a0) + (
c7d16
* a3));
183
184
/* First Butterfly on even terms. */
185
186
a0 =
MSCALE
(
c1d4
* (b0 + b1));
187
a1 =
MSCALE
(
c1d4
* (b0 - b1));
188
189
a2 =
MSCALE
((
c3d8
* b2) - (
c1d8
* b3));
190
a3 =
MSCALE
((
c1d8
* b2) + (
c3d8
* b3));
191
192
/* Calculate last set of b's */
193
194
b0 = a0 +
a3
;
195
b1 = a1 +
a2
;
196
b2 = a1 -
a2
;
197
b3 = a0 -
a3
;
198
199
/* Second Butterfly */
200
201
a0 = c0 +
c1
;
202
a1 = c0 -
c1
;
203
a2 = c3 -
c2
;
204
a3 = c3 +
c2
;
205
206
c0 =
a0
;
207
c1 =
MSCALE
(
c1d4
* (a2 - a1));
208
c2 =
MSCALE
(
c1d4
* (a2 + a1));
209
c3 =
a3
;
210
211
aptr = y +
LS
(i, 3);
212
*(aptr++) = b0 + c3;
213
*(aptr++) = b1 + c2;
214
*(aptr++) = b2 + c1;
215
*(aptr++) = b3 + c0;
216
*(aptr++) = b3 - c0;
217
*(aptr++) = b2 - c1;
218
*(aptr++) = b1 - c2;
219
*(aptr) = b0 - c3;
220
}
221
222
/*
223
Retrieve correct accuracy. We have additional factor
224
of 16 that must be removed.
225
*/
226
227
for
(i = 0, aptr = y; i < 64; i++, aptr++)
228
*aptr = (((*aptr < 0) ? (*aptr - 8) : (*aptr + 8)) / 16);
229
}
230
231
/*END*/
c2
TVMArray c2[1]
Definition:
11_conv2d_b.parallel.c:17
c1
TVMArray c1[1]
Definition:
09_conv2d_a.parallel.c:17
ChenIDct
void ChenIDct(int *x, int *y)
Definition:
chenidct.c:79
c1d16
#define c1d16
Definition:
chenidct.c:66
b0
TVMArray b0[1]
Definition:
09_conv2d_a.parallel.c:13
c1d4
#define c1d4
Definition:
chenidct.c:61
LS
#define LS(r, s)
Definition:
chenidct.c:54
c3d8
#define c3d8
Definition:
chenidct.c:64
MSCALE
#define MSCALE(expr)
Definition:
chenidct.c:57
a3
TVMArray a3[1]
Definition:
04_dense_a.parallel.c:12
c3d16
#define c3d16
Definition:
chenidct.c:67
c5d16
#define c5d16
Definition:
chenidct.c:68
a0
TVMArray a0[1]
Definition:
01_vecmul_a.wrapper.c:4
b2
TVMArray b2[1]
Definition:
09_conv2d_a.parallel.c:15
c7d16
#define c7d16
Definition:
chenidct.c:69
a2
TVMArray a2[1]
Definition:
01_vecmul_a.wrapper.c:6
a1
TVMArray a1[1]
Definition:
01_vecmul_a.wrapper.c:5
b1
TVMArray b1[1]
Definition:
09_conv2d_a.parallel.c:14
c1d8
#define c1d8
Definition:
chenidct.c:63
x
x
Return the smallest n such that 2^n >= _x.
Definition:
math_function.hpp:170
c0
TVMArray c0[1]
Definition:
09_conv2d_a.parallel.c:16
Generated on Mon Feb 12 2024 13:02:49 for PandA-2024.02 by
1.8.13