Browse code

Replacement reference DCT implementation. patch by Dylan Yudaken, dyudaken gmail com

Originally committed as revision 18273 to svn://svn.ffmpeg.org/ffmpeg/trunk

Dylan Yudaken authored on 2009/04/01 00:48:47
Showing 1 changed files
1 1
new file mode 100644
... ...
@@ -0,0 +1,121 @@
0
+/*
1
+ * reference discrete cosine transform (double precision)
2
+ * Copyright (C) 2009 Dylan Yudaken
3
+ *
4
+ * This file is part of FFmpeg.
5
+ *
6
+ * FFmpeg is free software; you can redistribute it and/or
7
+ * modify it under the terms of the GNU Lesser General Public
8
+ * License as published by the Free Software Foundation; either
9
+ * version 2.1 of the License, or (at your option) any later version.
10
+ *
11
+ * FFmpeg is distributed in the hope that it will be useful,
12
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
13
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
+ * Lesser General Public License for more details.
15
+ *
16
+ * You should have received a copy of the GNU Lesser General Public
17
+ * License along with FFmpeg; if not, write to the Free Software
18
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
+ */
20
+
21
+/**
22
+ * @file libavcodec/dctref.c
23
+ * reference discrete cosine transform (double precision)
24
+ *
25
+ * @author Dylan Yudaken (dyudaken at gmail)
26
+ *
27
+ * @note This file could be optimized a lot, but is for
28
+ * reference and so readability is better.
29
+ */
30
+
31
+#include "libavutil/mathematics.h"
32
+static double coefficients[8 * 8];
33
+
34
+/**
35
+ * Initialize the double precision discrete cosine transform
36
+ * functions fdct & idct.
37
+ */
38
+av_cold void ff_ref_dct_init(void)
39
+{
40
+    unsigned int i, j;
41
+
42
+    for (j = 0; j < 8; ++j) {
43
+        coefficients[j] = sqrt(0.125);
44
+        for (i = 8; i < 64; i += 8) {
45
+            coefficients[i + j] = 0.5 * cos(i * (j + 0.5) * M_PI / 64.0);
46
+        }
47
+    }
48
+}
49
+
50
+/**
51
+ * Transform 8x8 block of data with a double precision forward DCT <br>
52
+ * This is a reference implementation.
53
+ *
54
+ * @param block pointer to 8x8 block of data to transform
55
+ */
56
+void ff_ref_fdct(short *block)
57
+{
58
+    /* implement the equation: block = coefficients * block * coefficients' */
59
+
60
+    unsigned int i, j, k;
61
+    double out[8 * 8];
62
+
63
+    /* out = coefficients * block */
64
+    for (i = 0; i < 64; i += 8) {
65
+        for (j = 0; j < 8; ++j) {
66
+            double tmp = 0;
67
+            for (k = 0; k < 8; ++k) {
68
+                tmp += coefficients[i + k] * block[k * 8 + j];
69
+            }
70
+            out[i + j] = tmp * 8;
71
+        }
72
+    }
73
+
74
+    /* block = out * (coefficients') */
75
+    for (j = 0; j < 8; ++j) {
76
+        for (i = 0; i < 64; i += 8) {
77
+            double tmp = 0;
78
+            for (k = 0; k < 8; ++k) {
79
+                tmp += out[i + k] * coefficients[j * 8 + k];
80
+            }
81
+            block[i + j] = floor(tmp + 0.499999999999);
82
+        }
83
+    }
84
+}
85
+
86
+/**
87
+ * Transform 8x8 block of data with a double precision inverse DCT <br>
88
+ * This is a reference implementation.
89
+ *
90
+ * @param block pointer to 8x8 block of data to transform
91
+ */
92
+void ff_ref_idct(short *block)
93
+{
94
+    /* implement the equation: block = (coefficients') * block * coefficients */
95
+
96
+    unsigned int i, j, k;
97
+    double out[8 * 8];
98
+
99
+    /* out = block * coefficients */
100
+    for (i = 0; i < 64; i += 8) {
101
+        for (j = 0; j < 8; ++j) {
102
+            double tmp = 0;
103
+            for (k = 0; k < 8; ++k) {
104
+                tmp += block[i + k] * coefficients[k * 8 + j];
105
+            }
106
+            out[i + j] = tmp;
107
+        }
108
+    }
109
+
110
+    /* block = (coefficients') * out */
111
+    for (i = 0; i < 8; ++i) {
112
+        for (j = 0; j < 8; ++j) {
113
+            double tmp = 0;
114
+            for (k = 0; k < 64; k += 8) {
115
+                tmp += coefficients[k + i] * out[k + j];
116
+            }
117
+            block[i * 8 + j] = floor(tmp + 0.5);
118
+        }
119
+    }
120
+}