Clarify which BSD license.
[libav.git] / libavcodec / faandct.c
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 z1, z2, z3, 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 z1= (tmp12 + tmp13)*A1;
102 temp[2 + i]= tmp13 + z1;
103 temp[6 + i]= tmp13 - z1;
104
105 tmp10= tmp4 + tmp5;
106 tmp11= tmp5 + tmp6;
107 tmp12= tmp6 + tmp7;
108
109 z5= (tmp10 - tmp12) * A5;
110 z2= tmp10*A2 + z5;
111 z4= tmp12*A4 + z5;
112 z3= tmp11*A1;
113
114 z11= tmp7 + z3;
115 z13= tmp7 - z3;
116
117 temp[5 + i]= z13 + z2;
118 temp[3 + i]= z13 - z2;
119 temp[1 + i]= z11 + z4;
120 temp[7 + i]= z11 - z4;
121 }
122 }
123
124 void ff_faandct(DCTELEM * data)
125 {
126 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
127 FLOAT tmp10, tmp11, tmp12, tmp13;
128 FLOAT z1, z2, z3, z4, z5, z11, z13;
129 FLOAT temp[64];
130 int i;
131
132 emms_c();
133
134 row_fdct(temp, data);
135
136 for (i=0; i<8; i++) {
137 tmp0= temp[8*0 + i] + temp[8*7 + i];
138 tmp7= temp[8*0 + i] - temp[8*7 + i];
139 tmp1= temp[8*1 + i] + temp[8*6 + i];
140 tmp6= temp[8*1 + i] - temp[8*6 + i];
141 tmp2= temp[8*2 + i] + temp[8*5 + i];
142 tmp5= temp[8*2 + i] - temp[8*5 + i];
143 tmp3= temp[8*3 + i] + temp[8*4 + i];
144 tmp4= temp[8*3 + i] - temp[8*4 + i];
145
146 tmp10= tmp0 + tmp3;
147 tmp13= tmp0 - tmp3;
148 tmp11= tmp1 + tmp2;
149 tmp12= tmp1 - tmp2;
150
151 data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
152 data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
153
154 z1= (tmp12 + tmp13)* A1;
155 data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + z1));
156 data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - z1));
157
158 tmp10= tmp4 + tmp5;
159 tmp11= tmp5 + tmp6;
160 tmp12= tmp6 + tmp7;
161
162 z5= (tmp10 - tmp12) * A5;
163 z2= tmp10*A2 + z5;
164 z4= tmp12*A4 + z5;
165 z3= tmp11*A1;
166
167 z11= tmp7 + z3;
168 z13= tmp7 - z3;
169
170 data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
171 data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
172 data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
173 data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
174 }
175 }
176
177 void ff_faandct248(DCTELEM * data)
178 {
179 FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
180 FLOAT tmp10, tmp11, tmp12, tmp13;
181 FLOAT z1;
182 FLOAT temp[64];
183 int i;
184
185 emms_c();
186
187 row_fdct(temp, data);
188
189 for (i=0; i<8; i++) {
190 tmp0 = temp[8*0 + i] + temp[8*1 + i];
191 tmp1 = temp[8*2 + i] + temp[8*3 + i];
192 tmp2 = temp[8*4 + i] + temp[8*5 + i];
193 tmp3 = temp[8*6 + i] + temp[8*7 + i];
194 tmp4 = temp[8*0 + i] - temp[8*1 + i];
195 tmp5 = temp[8*2 + i] - temp[8*3 + i];
196 tmp6 = temp[8*4 + i] - temp[8*5 + i];
197 tmp7 = temp[8*6 + i] - temp[8*7 + i];
198
199 tmp10 = tmp0 + tmp3;
200 tmp11 = tmp1 + tmp2;
201 tmp12 = tmp1 - tmp2;
202 tmp13 = tmp0 - tmp3;
203
204 data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
205 data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
206
207 z1 = (tmp12 + tmp13)* A1;
208 data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
209 data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
210
211 tmp10 = tmp4 + tmp7;
212 tmp11 = tmp5 + tmp6;
213 tmp12 = tmp5 - tmp6;
214 tmp13 = tmp4 - tmp7;
215
216 data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
217 data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
218
219 z1 = (tmp12 + tmp13)* A1;
220 data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + z1));
221 data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - z1));
222 }
223 }