2 * Floating point AAN DCT
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 * this implementation is based upon the IJG integer AAN DCT (see jfdctfst.c)
27 * Floating point AAN DCT
28 * @author Michael Niedermayer <michaelni@gmx.at>
36 # define SCALE(x) postscale[x]
41 //numbers generated by simple c code (not as accurate as they could be)
44 printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
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
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)
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,
73 static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
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;
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];
95 temp[0 + i]= tmp10 + tmp11;
96 temp[4 + i]= tmp10 - tmp11;
98 z1= (tmp12 + tmp13)*A1;
99 temp[2 + i]= tmp13 + z1;
100 temp[6 + i]= tmp13 - z1;
106 z5= (tmp10 - tmp12) * A5;
114 temp[5 + i]= z13 + z2;
115 temp[3 + i]= z13 - z2;
116 temp[1 + i]= z11 + z4;
117 temp[7 + i]= z11 - z4;
121 void ff_faandct(DCTELEM * data)
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;
131 row_fdct(temp, data);
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];
148 data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
149 data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
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));
159 z5= (tmp10 - tmp12) * A5;
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));
174 void ff_faandct248(DCTELEM * data)
176 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
177 FLOAT tmp10, tmp11, tmp12, tmp13;
184 row_fdct(temp, data);
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];
201 data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
202 data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
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));
213 data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
214 data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
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));