Statistics
| Branch: | Revision:

ffmpeg / libavcodec / faandct.c @ 58c2182d

History | View | Annotate | Download (6.67 KB)

1
/*
2
 * Floating point AAN DCT
3
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4
 *
5
 * This library is free software; you can redistribute it and/or
6
 * modify it under the terms of the GNU Lesser General Public
7
 * License as published by the Free Software Foundation; either
8
 * version 2 of the License, or (at your option) any later version.
9
 *
10
 * This library 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 GNU
13
 * Lesser General Public License for more details.
14
 *
15
 * You should have received a copy of the GNU Lesser General Public
16
 * License along with this library; if not, write to the Free Software
17
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
18
 *
19
 * this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
20
 */
21

    
22
/**
23
 * @file faandct.c
24
 * @brief 
25
 *     Floating point AAN DCT
26
 * @author Michael Niedermayer <michaelni@gmx.at>
27
 */
28

    
29
#include <math.h>
30

    
31
#include "dsputil.h"
32
#include "faandct.h"
33

    
34
#define FLOAT float
35
#ifdef FAAN_POSTSCALE
36
#    define SCALE(x) postscale[x]
37
#else
38
#    define SCALE(x) 1
39
#endif
40

    
41
//numbers generated by simple c code (not as accurate as they could be)
42
/*
43
for(i=0; i<8; i++){
44
    printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
45
}
46
*/
47
#define B0 1.00000000000000000000
48
#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
49
#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
50
#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
51
#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
52
#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
53
#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
54
#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
55

    
56

    
57
#define A1 0.70710678118654752438 // cos(pi*4/16)
58
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
59
#define A5 0.38268343236508977170 // cos(pi*6/16)
60
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
61

    
62
static FLOAT postscale[64]={
63
B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7,
64
B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7,
65
B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7,
66
B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7,
67
B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7,
68
B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7,
69
B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
70
B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
71
};
72

    
73
static always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
74
{
75
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
76
    FLOAT tmp10, tmp11, tmp12, tmp13;
77
    FLOAT z1, z2, z3, z4, z5, z11, z13;
78
    int i;
79

    
80
    for (i=0; i<8*8; i+=8) {
81
        tmp0= data[0 + i] + data[7 + i];
82
        tmp7= data[0 + i] - data[7 + i];
83
        tmp1= data[1 + i] + data[6 + i];
84
        tmp6= data[1 + i] - data[6 + i];
85
        tmp2= data[2 + i] + data[5 + i];
86
        tmp5= data[2 + i] - data[5 + i];
87
        tmp3= data[3 + i] + data[4 + i];
88
        tmp4= data[3 + i] - data[4 + i];
89
        
90
        tmp10= tmp0 + tmp3;
91
        tmp13= tmp0 - tmp3;
92
        tmp11= tmp1 + tmp2;
93
        tmp12= tmp1 - tmp2;
94
        
95
        temp[0 + i]= tmp10 + tmp11;
96
        temp[4 + i]= tmp10 - tmp11;
97
        
98
        z1= (tmp12 + tmp13)*A1;
99
        temp[2 + i]= tmp13 + z1;
100
        temp[6 + i]= tmp13 - z1;
101
        
102
        tmp10= tmp4 + tmp5;
103
        tmp11= tmp5 + tmp6;
104
        tmp12= tmp6 + tmp7;
105

    
106
        z5= (tmp10 - tmp12) * A5;
107
        z2= tmp10*A2 + z5;
108
        z4= tmp12*A4 + z5;
109
        z3= tmp11*A1;
110

    
111
        z11= tmp7 + z3;
112
        z13= tmp7 - z3;
113

    
114
        temp[5 + i]= z13 + z2;
115
        temp[3 + i]= z13 - z2;
116
        temp[1 + i]= z11 + z4;
117
        temp[7 + i]= z11 - z4;
118
    }    
119
}
120

    
121
void ff_faandct(DCTELEM * data)
122
{
123
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
124
    FLOAT tmp10, tmp11, tmp12, tmp13;
125
    FLOAT z1, z2, z3, z4, z5, z11, z13;
126
    FLOAT temp[64];
127
    int i;
128

    
129
    emms_c();
130

    
131
    row_fdct(temp, data);
132

    
133
    for (i=0; i<8; i++) {
134
        tmp0= temp[8*0 + i] + temp[8*7 + i];
135
        tmp7= temp[8*0 + i] - temp[8*7 + i];
136
        tmp1= temp[8*1 + i] + temp[8*6 + i];
137
        tmp6= temp[8*1 + i] - temp[8*6 + i];
138
        tmp2= temp[8*2 + i] + temp[8*5 + i];
139
        tmp5= temp[8*2 + i] - temp[8*5 + i];
140
        tmp3= temp[8*3 + i] + temp[8*4 + i];
141
        tmp4= temp[8*3 + i] - temp[8*4 + i];
142
        
143
        tmp10= tmp0 + tmp3;
144
        tmp13= tmp0 - tmp3;
145
        tmp11= tmp1 + tmp2;
146
        tmp12= tmp1 - tmp2;
147
        
148
        data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
149
        data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
150
        
151
        z1= (tmp12 + tmp13)* A1;
152
        data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + z1));
153
        data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - z1));
154
        
155
        tmp10= tmp4 + tmp5;
156
        tmp11= tmp5 + tmp6;
157
        tmp12= tmp6 + tmp7;
158

    
159
        z5= (tmp10 - tmp12) * A5;
160
        z2= tmp10*A2 + z5;
161
        z4= tmp12*A4 + z5;
162
        z3= tmp11*A1;
163

    
164
        z11= tmp7 + z3;
165
        z13= tmp7 - z3;
166

    
167
        data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
168
        data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
169
        data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
170
        data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
171
    }
172
}
173

    
174
void ff_faandct248(DCTELEM * data)
175
{
176
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
177
    FLOAT tmp10, tmp11, tmp12, tmp13;
178
    FLOAT z1;
179
    FLOAT temp[64];
180
    int i;
181

    
182
    emms_c();
183

    
184
    row_fdct(temp, data);
185

    
186
    for (i=0; i<8; i++) {
187
        tmp0 = temp[8*0 + i] + temp[8*1 + i];
188
        tmp1 = temp[8*2 + i] + temp[8*3 + i];
189
        tmp2 = temp[8*4 + i] + temp[8*5 + i];
190
        tmp3 = temp[8*6 + i] + temp[8*7 + i];
191
        tmp4 = temp[8*0 + i] - temp[8*1 + i];
192
        tmp5 = temp[8*2 + i] - temp[8*3 + i];
193
        tmp6 = temp[8*4 + i] - temp[8*5 + i];
194
        tmp7 = temp[8*6 + i] - temp[8*7 + i];
195
        
196
        tmp10 = tmp0 + tmp3;
197
        tmp11 = tmp1 + tmp2;
198
        tmp12 = tmp1 - tmp2;
199
        tmp13 = tmp0 - tmp3;
200
        
201
        data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
202
        data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
203
        
204
        z1 = (tmp12 + tmp13)* A1;
205
        data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
206
        data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
207
        
208
        tmp10 = tmp4 + tmp7;
209
        tmp11 = tmp5 + tmp6;
210
        tmp12 = tmp5 - tmp6;
211
        tmp13 = tmp4 - tmp7;
212

    
213
        data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
214
        data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
215

    
216
        z1 = (tmp12 + tmp13)* A1;
217
        data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
218
        data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
219
    }
220
}