FFmpeg: libavcodec/faanidct.c Source File

00001 /*
00002  * Floating point AAN IDCT
00003  * Copyright (c) 2008 Michael Niedermayer <michaelni@gmx.at>
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 #include "faanidct.h"
00022 
00023 /* To allow switching to double. */
00024 #define FLOAT float
00025 
00026 #define B0 1.0000000000000000000000
00027 #define B1 1.3870398453221474618216 // cos(pi*1/16)sqrt(2)
00028 #define B2 1.3065629648763765278566 // cos(pi*2/16)sqrt(2)
00029 #define B3 1.1758756024193587169745 // cos(pi*3/16)sqrt(2)
00030 #define B4 1.0000000000000000000000 // cos(pi*4/16)sqrt(2)
00031 #define B5 0.7856949583871021812779 // cos(pi*5/16)sqrt(2)
00032 #define B6 0.5411961001461969843997 // cos(pi*6/16)sqrt(2)
00033 #define B7 0.2758993792829430123360 // cos(pi*7/16)sqrt(2)
00034 
00035 #define A4 0.70710678118654752438 // cos(pi*4/16)
00036 #define A2 0.92387953251128675613 // cos(pi*2/16)
00037 
00038 static const FLOAT prescale[64]={
00039 B0*B0/8, B0*B1/8, B0*B2/8, B0*B3/8, B0*B4/8, B0*B5/8, B0*B6/8, B0*B7/8,
00040 B1*B0/8, B1*B1/8, B1*B2/8, B1*B3/8, B1*B4/8, B1*B5/8, B1*B6/8, B1*B7/8,
00041 B2*B0/8, B2*B1/8, B2*B2/8, B2*B3/8, B2*B4/8, B2*B5/8, B2*B6/8, B2*B7/8,
00042 B3*B0/8, B3*B1/8, B3*B2/8, B3*B3/8, B3*B4/8, B3*B5/8, B3*B6/8, B3*B7/8,
00043 B4*B0/8, B4*B1/8, B4*B2/8, B4*B3/8, B4*B4/8, B4*B5/8, B4*B6/8, B4*B7/8,
00044 B5*B0/8, B5*B1/8, B5*B2/8, B5*B3/8, B5*B4/8, B5*B5/8, B5*B6/8, B5*B7/8,
00045 B6*B0/8, B6*B1/8, B6*B2/8, B6*B3/8, B6*B4/8, B6*B5/8, B6*B6/8, B6*B7/8,
00046 B7*B0/8, B7*B1/8, B7*B2/8, B7*B3/8, B7*B4/8, B7*B5/8, B7*B6/8, B7*B7/8,
00047 };
00048 
00049 static inline void p8idct(DCTELEM data[64], FLOAT temp[64], uint8_t *dest, int stride, int x, int y, int type){
00050     int i;
00051     FLOAT av_unused tmp0;
00052     FLOAT s04, d04, s17, d17, s26, d26, s53, d53;
00053     FLOAT os07, os16, os25, os34;
00054     FLOAT od07, od16, od25, od34;
00055 
00056     for(i=0; i<y*8; i+=y){
00057         s17= temp[1*x + i] + temp[7*x + i];
00058         d17= temp[1*x + i] - temp[7*x + i];
00059         s53= temp[5*x + i] + temp[3*x + i];
00060         d53= temp[5*x + i] - temp[3*x + i];
00061 
00062         od07=  s17 + s53;
00063         od25= (s17 - s53)*(2*A4);
00064 
00065 #if 0 //these 2 are equivalent
00066         tmp0= (d17 + d53)*(2*A2);
00067         od34=  d17*( 2*B6) - tmp0;
00068         od16=  d53*(-2*B2) + tmp0;
00069 #else
00070         od34=  d17*(2*(B6-A2)) - d53*(2*A2);
00071         od16=  d53*(2*(A2-B2)) + d17*(2*A2);
00072 #endif
00073 
00074         od16 -= od07;
00075         od25 -= od16;
00076         od34 += od25;
00077 
00078         s26 = temp[2*x + i] + temp[6*x + i];
00079         d26 = temp[2*x + i] - temp[6*x + i];
00080         d26*= 2*A4;
00081         d26-= s26;
00082 
00083         s04= temp[0*x + i] + temp[4*x + i];
00084         d04= temp[0*x + i] - temp[4*x + i];
00085 
00086         os07= s04 + s26;
00087         os34= s04 - s26;
00088         os16= d04 + d26;
00089         os25= d04 - d26;
00090 
00091         if(type==0){
00092             temp[0*x + i]= os07 + od07;
00093             temp[7*x + i]= os07 - od07;
00094             temp[1*x + i]= os16 + od16;
00095             temp[6*x + i]= os16 - od16;
00096             temp[2*x + i]= os25 + od25;
00097             temp[5*x + i]= os25 - od25;
00098             temp[3*x + i]= os34 - od34;
00099             temp[4*x + i]= os34 + od34;
00100         }else if(type==1){
00101             data[0*x + i]= lrintf(os07 + od07);
00102             data[7*x + i]= lrintf(os07 - od07);
00103             data[1*x + i]= lrintf(os16 + od16);
00104             data[6*x + i]= lrintf(os16 - od16);
00105             data[2*x + i]= lrintf(os25 + od25);
00106             data[5*x + i]= lrintf(os25 - od25);
00107             data[3*x + i]= lrintf(os34 - od34);
00108             data[4*x + i]= lrintf(os34 + od34);
00109         }else if(type==2){
00110             dest[0*stride + i]= av_clip_uint8(((int)dest[0*stride + i]) + lrintf(os07 + od07));
00111             dest[7*stride + i]= av_clip_uint8(((int)dest[7*stride + i]) + lrintf(os07 - od07));
00112             dest[1*stride + i]= av_clip_uint8(((int)dest[1*stride + i]) + lrintf(os16 + od16));
00113             dest[6*stride + i]= av_clip_uint8(((int)dest[6*stride + i]) + lrintf(os16 - od16));
00114             dest[2*stride + i]= av_clip_uint8(((int)dest[2*stride + i]) + lrintf(os25 + od25));
00115             dest[5*stride + i]= av_clip_uint8(((int)dest[5*stride + i]) + lrintf(os25 - od25));
00116             dest[3*stride + i]= av_clip_uint8(((int)dest[3*stride + i]) + lrintf(os34 - od34));
00117             dest[4*stride + i]= av_clip_uint8(((int)dest[4*stride + i]) + lrintf(os34 + od34));
00118         }else{
00119             dest[0*stride + i]= av_clip_uint8(lrintf(os07 + od07));
00120             dest[7*stride + i]= av_clip_uint8(lrintf(os07 - od07));
00121             dest[1*stride + i]= av_clip_uint8(lrintf(os16 + od16));
00122             dest[6*stride + i]= av_clip_uint8(lrintf(os16 - od16));
00123             dest[2*stride + i]= av_clip_uint8(lrintf(os25 + od25));
00124             dest[5*stride + i]= av_clip_uint8(lrintf(os25 - od25));
00125             dest[3*stride + i]= av_clip_uint8(lrintf(os34 - od34));
00126             dest[4*stride + i]= av_clip_uint8(lrintf(os34 + od34));
00127         }
00128     }
00129 }
00130 
00131 void ff_faanidct(DCTELEM block[64]){
00132     FLOAT temp[64];
00133     int i;
00134 
00135     emms_c();
00136 
00137     for(i=0; i<64; i++)
00138         temp[i] = block[i] * prescale[i];
00139 
00140     p8idct(block, temp, NULL, 0, 1, 8, 0);
00141     p8idct(block, temp, NULL, 0, 8, 1, 1);
00142 }
00143 
00144 void ff_faanidct_add(uint8_t *dest, int line_size, DCTELEM block[64]){
00145     FLOAT temp[64];
00146     int i;
00147 
00148     emms_c();
00149 
00150     for(i=0; i<64; i++)
00151         temp[i] = block[i] * prescale[i];
00152 
00153     p8idct(block, temp, NULL,         0, 1, 8, 0);
00154     p8idct(NULL , temp, dest, line_size, 8, 1, 2);
00155 }
00156 
00157 void ff_faanidct_put(uint8_t *dest, int line_size, DCTELEM block[64]){
00158     FLOAT temp[64];
00159     int i;
00160 
00161     emms_c();
00162 
00163     for(i=0; i<64; i++)
00164         temp[i] = block[i] * prescale[i];
00165 
00166     p8idct(block, temp, NULL,         0, 1, 8, 0);
00167     p8idct(NULL , temp, dest, line_size, 8, 1, 3);
00168 }