Statistics
| Branch: | Revision:

ffmpeg / libavcodec / faandct.c @ 6636b7e8

History | View | Annotate | Download (6.62 KB)

1 65e4c8c9 Michael Niedermayer
/*
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 5509bffa Diego Biurrun
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 65e4c8c9 Michael Niedermayer
 *
19 6f5cf8c5 Michael Niedermayer
 * this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
20 65e4c8c9 Michael Niedermayer
 */
21
22
/**
23
 * @file faandct.c
24 115329f1 Diego Biurrun
 * @brief
25 65e4c8c9 Michael Niedermayer
 *     Floating point AAN DCT
26
 * @author Michael Niedermayer <michaelni@gmx.at>
27
 */
28
29
#include "dsputil.h"
30
#include "faandct.h"
31
32
#define FLOAT float
33 b4c3816c Michael Niedermayer
#ifdef FAAN_POSTSCALE
34
#    define SCALE(x) postscale[x]
35
#else
36
#    define SCALE(x) 1
37
#endif
38 65e4c8c9 Michael Niedermayer
39
//numbers generated by simple c code (not as accurate as they could be)
40
/*
41
for(i=0; i<8; i++){
42
    printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
43
}
44
*/
45
#define B0 1.00000000000000000000
46
#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
47
#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
48
#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
49
#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
50
#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
51
#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
52
#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
53
54
55
#define A1 0.70710678118654752438 // cos(pi*4/16)
56
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
57
#define A5 0.38268343236508977170 // cos(pi*6/16)
58
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
59
60
static FLOAT postscale[64]={
61
B0*B0, B0*B1, B0*B2, B0*B3, B0*B4, B0*B5, B0*B6, B0*B7,
62
B1*B0, B1*B1, B1*B2, B1*B3, B1*B4, B1*B5, B1*B6, B1*B7,
63
B2*B0, B2*B1, B2*B2, B2*B3, B2*B4, B2*B5, B2*B6, B2*B7,
64
B3*B0, B3*B1, B3*B2, B3*B3, B3*B4, B3*B5, B3*B6, B3*B7,
65
B4*B0, B4*B1, B4*B2, B4*B3, B4*B4, B4*B5, B4*B6, B4*B7,
66
B5*B0, B5*B1, B5*B2, B5*B3, B5*B4, B5*B5, B5*B6, B5*B7,
67
B6*B0, B6*B1, B6*B2, B6*B3, B6*B4, B6*B5, B6*B6, B6*B7,
68
B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
69
};
70
71 d43fb4e8 Michael Niedermayer
static always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
72 65e4c8c9 Michael Niedermayer
{
73
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
74
    FLOAT tmp10, tmp11, tmp12, tmp13;
75
    FLOAT z1, z2, z3, z4, z5, z11, z13;
76
    int i;
77
78
    for (i=0; i<8*8; i+=8) {
79
        tmp0= data[0 + i] + data[7 + i];
80
        tmp7= data[0 + i] - data[7 + i];
81
        tmp1= data[1 + i] + data[6 + i];
82
        tmp6= data[1 + i] - data[6 + i];
83
        tmp2= data[2 + i] + data[5 + i];
84
        tmp5= data[2 + i] - data[5 + i];
85
        tmp3= data[3 + i] + data[4 + i];
86
        tmp4= data[3 + i] - data[4 + i];
87 115329f1 Diego Biurrun
88 65e4c8c9 Michael Niedermayer
        tmp10= tmp0 + tmp3;
89
        tmp13= tmp0 - tmp3;
90
        tmp11= tmp1 + tmp2;
91
        tmp12= tmp1 - tmp2;
92 115329f1 Diego Biurrun
93 65e4c8c9 Michael Niedermayer
        temp[0 + i]= tmp10 + tmp11;
94
        temp[4 + i]= tmp10 - tmp11;
95 115329f1 Diego Biurrun
96 65e4c8c9 Michael Niedermayer
        z1= (tmp12 + tmp13)*A1;
97
        temp[2 + i]= tmp13 + z1;
98
        temp[6 + i]= tmp13 - z1;
99 115329f1 Diego Biurrun
100 65e4c8c9 Michael Niedermayer
        tmp10= tmp4 + tmp5;
101
        tmp11= tmp5 + tmp6;
102
        tmp12= tmp6 + tmp7;
103
104
        z5= (tmp10 - tmp12) * A5;
105
        z2= tmp10*A2 + z5;
106
        z4= tmp12*A4 + z5;
107
        z3= tmp11*A1;
108
109
        z11= tmp7 + z3;
110
        z13= tmp7 - z3;
111
112
        temp[5 + i]= z13 + z2;
113
        temp[3 + i]= z13 - z2;
114
        temp[1 + i]= z11 + z4;
115
        temp[7 + i]= z11 - z4;
116 115329f1 Diego Biurrun
    }
117 d43fb4e8 Michael Niedermayer
}
118
119
void ff_faandct(DCTELEM * data)
120
{
121
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
122
    FLOAT tmp10, tmp11, tmp12, tmp13;
123
    FLOAT z1, z2, z3, z4, z5, z11, z13;
124
    FLOAT temp[64];
125
    int i;
126
127
    emms_c();
128
129
    row_fdct(temp, data);
130 65e4c8c9 Michael Niedermayer
131
    for (i=0; i<8; i++) {
132
        tmp0= temp[8*0 + i] + temp[8*7 + i];
133
        tmp7= temp[8*0 + i] - temp[8*7 + i];
134
        tmp1= temp[8*1 + i] + temp[8*6 + i];
135
        tmp6= temp[8*1 + i] - temp[8*6 + i];
136
        tmp2= temp[8*2 + i] + temp[8*5 + i];
137
        tmp5= temp[8*2 + i] - temp[8*5 + i];
138
        tmp3= temp[8*3 + i] + temp[8*4 + i];
139
        tmp4= temp[8*3 + i] - temp[8*4 + i];
140 115329f1 Diego Biurrun
141 65e4c8c9 Michael Niedermayer
        tmp10= tmp0 + tmp3;
142
        tmp13= tmp0 - tmp3;
143
        tmp11= tmp1 + tmp2;
144
        tmp12= tmp1 - tmp2;
145 115329f1 Diego Biurrun
146 8d00c562 Michael Niedermayer
        data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
147
        data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
148 115329f1 Diego Biurrun
149 65e4c8c9 Michael Niedermayer
        z1= (tmp12 + tmp13)* A1;
150 8d00c562 Michael Niedermayer
        data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + z1));
151
        data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - z1));
152 115329f1 Diego Biurrun
153 65e4c8c9 Michael Niedermayer
        tmp10= tmp4 + tmp5;
154
        tmp11= tmp5 + tmp6;
155
        tmp12= tmp6 + tmp7;
156
157
        z5= (tmp10 - tmp12) * A5;
158
        z2= tmp10*A2 + z5;
159
        z4= tmp12*A4 + z5;
160
        z3= tmp11*A1;
161
162
        z11= tmp7 + z3;
163
        z13= tmp7 - z3;
164
165 8d00c562 Michael Niedermayer
        data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
166
        data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
167
        data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
168
        data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
169 65e4c8c9 Michael Niedermayer
    }
170
}
171 48b1f800 Roman Shaposhnik
172
void ff_faandct248(DCTELEM * data)
173
{
174
    FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
175
    FLOAT tmp10, tmp11, tmp12, tmp13;
176 d43fb4e8 Michael Niedermayer
    FLOAT z1;
177 48b1f800 Roman Shaposhnik
    FLOAT temp[64];
178
    int i;
179
180
    emms_c();
181
182 d43fb4e8 Michael Niedermayer
    row_fdct(temp, data);
183 48b1f800 Roman Shaposhnik
184
    for (i=0; i<8; i++) {
185
        tmp0 = temp[8*0 + i] + temp[8*1 + i];
186
        tmp1 = temp[8*2 + i] + temp[8*3 + i];
187
        tmp2 = temp[8*4 + i] + temp[8*5 + i];
188
        tmp3 = temp[8*6 + i] + temp[8*7 + i];
189
        tmp4 = temp[8*0 + i] - temp[8*1 + i];
190
        tmp5 = temp[8*2 + i] - temp[8*3 + i];
191
        tmp6 = temp[8*4 + i] - temp[8*5 + i];
192
        tmp7 = temp[8*6 + i] - temp[8*7 + i];
193 115329f1 Diego Biurrun
194 48b1f800 Roman Shaposhnik
        tmp10 = tmp0 + tmp3;
195
        tmp11 = tmp1 + tmp2;
196
        tmp12 = tmp1 - tmp2;
197
        tmp13 = tmp0 - tmp3;
198 115329f1 Diego Biurrun
199 48b1f800 Roman Shaposhnik
        data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
200
        data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
201 115329f1 Diego Biurrun
202 48b1f800 Roman Shaposhnik
        z1 = (tmp12 + tmp13)* A1;
203
        data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
204
        data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
205 115329f1 Diego Biurrun
206 48b1f800 Roman Shaposhnik
        tmp10 = tmp4 + tmp7;
207 bb270c08 Diego Biurrun
        tmp11 = tmp5 + tmp6;
208
        tmp12 = tmp5 - tmp6;
209
        tmp13 = tmp4 - tmp7;
210 48b1f800 Roman Shaposhnik
211 bb270c08 Diego Biurrun
        data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
212
        data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
213 48b1f800 Roman Shaposhnik
214 bb270c08 Diego Biurrun
        z1 = (tmp12 + tmp13)* A1;
215
        data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
216
        data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
217 48b1f800 Roman Shaposhnik
    }
218
}