Statistics
| Branch: | Revision:

ffmpeg / libavcodec / faandct.c @ c3a3d344

History | View | Annotate | Download (6.92 KB)

1
/*
2
 * Floating point AAN DCT
3
 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4
 *
5
 * this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
6
 *
7
 * This file is part of FFmpeg.
8
 *
9
 * FFmpeg is free software; you can redistribute it and/or
10
 * modify it under the terms of the GNU Lesser General Public
11
 * License as published by the Free Software Foundation; either
12
 * version 2.1 of the License, or (at your option) any later version.
13
 *
14
 * FFmpeg is distributed in the hope that it will be useful,
15
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17
 * Lesser General Public License for more details.
18
 *
19
 * You should have received a copy of the GNU Lesser General Public
20
 * License along with FFmpeg; if not, write to the Free Software
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22
 *
23
 * The AAN DCT in this file except ff_faandct248() can also be used under the
24
 * new (3 clause) BSD license.
25
 */
26

    
27
/**
28
 * @file faandct.c
29
 * @brief
30
 *     Floating point AAN DCT
31
 * @author Michael Niedermayer <michaelni@gmx.at>
32
 */
33

    
34
#include "dsputil.h"
35
#include "faandct.h"
36

    
37
#define FLOAT float
38
#ifdef FAAN_POSTSCALE
39
#    define SCALE(x) postscale[x]
40
#else
41
#    define SCALE(x) 1
42
#endif
43

    
44
//numbers generated by simple c code (not as accurate as they could be)
45
/*
46
for(i=0; i<8; i++){
47
    printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
48
}
49
*/
50
#define B0 1.00000000000000000000
51
#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
52
#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
53
#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
54
#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
55
#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
56
#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
57
#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
58

    
59

    
60
#define A1 0.70710678118654752438 // cos(pi*4/16)
61
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
62
#define A5 0.38268343236508977170 // cos(pi*6/16)
63
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
64

    
65
static FLOAT postscale[64]={
66
B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7,
67
B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7,
68
B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7,
69
B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7,
70
B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7,
71
B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7,
72
B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
73
B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
74
};
75

    
76
static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
77
{
78
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
79
    FLOAT tmp10, tmp11, tmp12, tmp13;
80
    FLOAT z2, z4, z5, z11, z13;
81
    int i;
82

    
83
    for (i=0; i<8*8; i+=8) {
84
        tmp0= data[0 + i] + data[7 + i];
85
        tmp7= data[0 + i] - data[7 + i];
86
        tmp1= data[1 + i] + data[6 + i];
87
        tmp6= data[1 + i] - data[6 + i];
88
        tmp2= data[2 + i] + data[5 + i];
89
        tmp5= data[2 + i] - data[5 + i];
90
        tmp3= data[3 + i] + data[4 + i];
91
        tmp4= data[3 + i] - data[4 + i];
92

    
93
        tmp10= tmp0 + tmp3;
94
        tmp13= tmp0 - tmp3;
95
        tmp11= tmp1 + tmp2;
96
        tmp12= tmp1 - tmp2;
97

    
98
        temp[0 + i]= tmp10 + tmp11;
99
        temp[4 + i]= tmp10 - tmp11;
100

    
101
        tmp12 += tmp13;
102
        tmp12 *= A1;
103
        temp[2 + i]= tmp13 + tmp12;
104
        temp[6 + i]= tmp13 - tmp12;
105

    
106
        tmp4 += tmp5;
107
        tmp5 += tmp6;
108
        tmp6 += tmp7;
109

    
110
#if 0
111
        z5= (tmp4 - tmp6) * A5;
112
        z2= tmp4*A2 + z5;
113
        z4= tmp6*A4 + z5;
114
#else
115
        z2= tmp4*(A2+A5) - tmp6*A5;
116
        z4= tmp6*(A4-A5) + tmp4*A5;
117
#endif
118
        tmp5*=A1;
119

    
120
        z11= tmp7 + tmp5;
121
        z13= tmp7 - tmp5;
122

    
123
        temp[5 + i]= z13 + z2;
124
        temp[3 + i]= z13 - z2;
125
        temp[1 + i]= z11 + z4;
126
        temp[7 + i]= z11 - z4;
127
    }
128
}
129

    
130
void ff_faandct(DCTELEM * data)
131
{
132
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
133
    FLOAT tmp10, tmp11, tmp12, tmp13;
134
    FLOAT z2, z4, z5, z11, z13;
135
    FLOAT temp[64];
136
    int i;
137

    
138
    emms_c();
139

    
140
    row_fdct(temp, data);
141

    
142
    for (i=0; i<8; i++) {
143
        tmp0= temp[8*0 + i] + temp[8*7 + i];
144
        tmp7= temp[8*0 + i] - temp[8*7 + i];
145
        tmp1= temp[8*1 + i] + temp[8*6 + i];
146
        tmp6= temp[8*1 + i] - temp[8*6 + i];
147
        tmp2= temp[8*2 + i] + temp[8*5 + i];
148
        tmp5= temp[8*2 + i] - temp[8*5 + i];
149
        tmp3= temp[8*3 + i] + temp[8*4 + i];
150
        tmp4= temp[8*3 + i] - temp[8*4 + i];
151

    
152
        tmp10= tmp0 + tmp3;
153
        tmp13= tmp0 - tmp3;
154
        tmp11= tmp1 + tmp2;
155
        tmp12= tmp1 - tmp2;
156

    
157
        data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
158
        data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
159

    
160
        tmp12 += tmp13;
161
        tmp12 *= A1;
162
        data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
163
        data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
164

    
165
        tmp4 += tmp5;
166
        tmp5 += tmp6;
167
        tmp6 += tmp7;
168

    
169
#if 0
170
        z5= (tmp4 - tmp6) * A5;
171
        z2= tmp4*A2 + z5;
172
        z4= tmp6*A4 + z5;
173
#else
174
        z2= tmp4*(A2+A5) - tmp6*A5;
175
        z4= tmp6*(A4-A5) + tmp4*A5;
176
#endif
177
        tmp5*=A1;
178

    
179
        z11= tmp7 + tmp5;
180
        z13= tmp7 - tmp5;
181

    
182
        data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
183
        data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
184
        data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
185
        data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
186
    }
187
}
188

    
189
void ff_faandct248(DCTELEM * data)
190
{
191
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
192
    FLOAT tmp10, tmp11, tmp12, tmp13;
193
    FLOAT temp[64];
194
    int i;
195

    
196
    emms_c();
197

    
198
    row_fdct(temp, data);
199

    
200
    for (i=0; i<8; i++) {
201
        tmp0 = temp[8*0 + i] + temp[8*1 + i];
202
        tmp1 = temp[8*2 + i] + temp[8*3 + i];
203
        tmp2 = temp[8*4 + i] + temp[8*5 + i];
204
        tmp3 = temp[8*6 + i] + temp[8*7 + i];
205
        tmp4 = temp[8*0 + i] - temp[8*1 + i];
206
        tmp5 = temp[8*2 + i] - temp[8*3 + i];
207
        tmp6 = temp[8*4 + i] - temp[8*5 + i];
208
        tmp7 = temp[8*6 + i] - temp[8*7 + i];
209

    
210
        tmp10 = tmp0 + tmp3;
211
        tmp11 = tmp1 + tmp2;
212
        tmp12 = tmp1 - tmp2;
213
        tmp13 = tmp0 - tmp3;
214

    
215
        data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
216
        data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
217

    
218
        tmp12 += tmp13;
219
        tmp12 *= A1;
220
        data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
221
        data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
222

    
223
        tmp10 = tmp4 + tmp7;
224
        tmp11 = tmp5 + tmp6;
225
        tmp12 = tmp5 - tmp6;
226
        tmp13 = tmp4 - tmp7;
227

    
228
        data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
229
        data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
230

    
231
        tmp12 += tmp13;
232
        tmp12 *= A1;
233
        data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
234
        data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
235
    }
236
}