FFmpeg: libavcodec/vp3dsp.c Source File

00001 /*
00002  * Copyright (C) 2004 the ffmpeg project
00003  *
00004  * This file is part of FFmpeg.
00005  *
00006  * FFmpeg is free software; you can redistribute it and/or
00007  * modify it under the terms of the GNU Lesser General Public
00008  * License as published by the Free Software Foundation; either
00009  * version 2.1 of the License, or (at your option) any later version.
00010  *
00011  * FFmpeg is distributed in the hope that it will be useful,
00012  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00013  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00014  * Lesser General Public License for more details.
00015  *
00016  * You should have received a copy of the GNU Lesser General Public
00017  * License along with FFmpeg; if not, write to the Free Software
00018  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00019  */
00020 
00027 #include "avcodec.h"
00028 #include "dsputil.h"
00029 
00030 #define IdctAdjustBeforeShift 8
00031 #define xC1S7 64277
00032 #define xC2S6 60547
00033 #define xC3S5 54491
00034 #define xC4S4 46341
00035 #define xC5S3 36410
00036 #define xC6S2 25080
00037 #define xC7S1 12785
00038 
00039 #define M(a,b) (((a) * (b))>>16)
00040 
00041 static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
00042 {
00043     int16_t *ip = input;
00044 
00045     int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
00046     int Ed, Gd, Add, Bdd, Fd, Hd;
00047 
00048     int i;
00049 
00050     /* Inverse DCT on the rows now */
00051     for (i = 0; i < 8; i++) {
00052         /* Check for non-zero values */
00053         if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
00054             A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
00055             B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
00056             C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
00057             D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
00058 
00059             Ad = M(xC4S4, (A - C));
00060             Bd = M(xC4S4, (B - D));
00061 
00062             Cd = A + C;
00063             Dd = B + D;
00064 
00065             E = M(xC4S4, (ip[0] + ip[4]));
00066             F = M(xC4S4, (ip[0] - ip[4]));
00067 
00068             G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
00069             H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
00070 
00071             Ed = E - G;
00072             Gd = E + G;
00073 
00074             Add = F + Ad;
00075             Bdd = Bd - H;
00076 
00077             Fd = F - Ad;
00078             Hd = Bd + H;
00079 
00080             /*  Final sequence of operations over-write original inputs. */
00081             ip[0] = Gd + Cd ;
00082             ip[7] = Gd - Cd ;
00083 
00084             ip[1] = Add + Hd;
00085             ip[2] = Add - Hd;
00086 
00087             ip[3] = Ed + Dd ;
00088             ip[4] = Ed - Dd ;
00089 
00090             ip[5] = Fd + Bdd;
00091             ip[6] = Fd - Bdd;
00092         }
00093 
00094         ip += 8;            /* next row */
00095     }
00096 
00097     ip = input;
00098 
00099     for ( i = 0; i < 8; i++) {
00100         /* Check for non-zero values (bitwise or faster than ||) */
00101         if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
00102              ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
00103 
00104             A = M(xC1S7, ip[1*8]) + M(xC7S1, ip[7*8]);
00105             B = M(xC7S1, ip[1*8]) - M(xC1S7, ip[7*8]);
00106             C = M(xC3S5, ip[3*8]) + M(xC5S3, ip[5*8]);
00107             D = M(xC3S5, ip[5*8]) - M(xC5S3, ip[3*8]);
00108 
00109             Ad = M(xC4S4, (A - C));
00110             Bd = M(xC4S4, (B - D));
00111 
00112             Cd = A + C;
00113             Dd = B + D;
00114 
00115             E = M(xC4S4, (ip[0*8] + ip[4*8])) + 8;
00116             F = M(xC4S4, (ip[0*8] - ip[4*8])) + 8;
00117 
00118             if(type==1){  //HACK
00119                 E += 16*128;
00120                 F += 16*128;
00121             }
00122 
00123             G = M(xC2S6, ip[2*8]) + M(xC6S2, ip[6*8]);
00124             H = M(xC6S2, ip[2*8]) - M(xC2S6, ip[6*8]);
00125 
00126             Ed = E - G;
00127             Gd = E + G;
00128 
00129             Add = F + Ad;
00130             Bdd = Bd - H;
00131 
00132             Fd = F - Ad;
00133             Hd = Bd + H;
00134 
00135             /* Final sequence of operations over-write original inputs. */
00136             if(type==0){
00137                 ip[0*8] = (Gd + Cd )  >> 4;
00138                 ip[7*8] = (Gd - Cd )  >> 4;
00139 
00140                 ip[1*8] = (Add + Hd ) >> 4;
00141                 ip[2*8] = (Add - Hd ) >> 4;
00142 
00143                 ip[3*8] = (Ed + Dd )  >> 4;
00144                 ip[4*8] = (Ed - Dd )  >> 4;
00145 
00146                 ip[5*8] = (Fd + Bdd ) >> 4;
00147                 ip[6*8] = (Fd - Bdd ) >> 4;
00148             }else if(type==1){
00149                 dst[0*stride] = av_clip_uint8((Gd + Cd )  >> 4);
00150                 dst[7*stride] = av_clip_uint8((Gd - Cd )  >> 4);
00151 
00152                 dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
00153                 dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
00154 
00155                 dst[3*stride] = av_clip_uint8((Ed + Dd )  >> 4);
00156                 dst[4*stride] = av_clip_uint8((Ed - Dd )  >> 4);
00157 
00158                 dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
00159                 dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
00160             }else{
00161                 dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd )  >> 4));
00162                 dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd )  >> 4));
00163 
00164                 dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
00165                 dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
00166 
00167                 dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd )  >> 4));
00168                 dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd )  >> 4));
00169 
00170                 dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
00171                 dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
00172             }
00173 
00174         } else {
00175             if(type==0){
00176                 ip[0*8] =
00177                 ip[1*8] =
00178                 ip[2*8] =
00179                 ip[3*8] =
00180                 ip[4*8] =
00181                 ip[5*8] =
00182                 ip[6*8] =
00183                 ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
00184             }else if(type==1){
00185                 dst[0*stride]=
00186                 dst[1*stride]=
00187                 dst[2*stride]=
00188                 dst[3*stride]=
00189                 dst[4*stride]=
00190                 dst[5*stride]=
00191                 dst[6*stride]=
00192                 dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20));
00193             }else{
00194                 if(ip[0*8]){
00195                     int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
00196                     dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
00197                     dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
00198                     dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
00199                     dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
00200                     dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
00201                     dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
00202                     dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
00203                     dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
00204                 }
00205             }
00206         }
00207 
00208         ip++;            /* next column */
00209         dst++;
00210     }
00211 }
00212 
00213 void ff_vp3_idct_c(DCTELEM *block/* align 16*/){
00214     idct(NULL, 0, block, 0);
00215 }
00216 
00217 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
00218     idct(dest, line_size, block, 1);
00219 }
00220 
00221 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
00222     idct(dest, line_size, block, 2);
00223 }
00224 
00225 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/){
00226     int i, dc = (block[0] + 15) >> 5;
00227 
00228     for(i = 0; i < 8; i++){
00229         dest[0] = av_clip_uint8(dest[0] + dc);
00230         dest[1] = av_clip_uint8(dest[1] + dc);
00231         dest[2] = av_clip_uint8(dest[2] + dc);
00232         dest[3] = av_clip_uint8(dest[3] + dc);
00233         dest[4] = av_clip_uint8(dest[4] + dc);
00234         dest[5] = av_clip_uint8(dest[5] + dc);
00235         dest[6] = av_clip_uint8(dest[6] + dc);
00236         dest[7] = av_clip_uint8(dest[7] + dc);
00237         dest += line_size;
00238     }
00239 }
00240 
00241 void ff_vp3_v_loop_filter_c(uint8_t *first_pixel, int stride, int *bounding_values)
00242 {
00243     unsigned char *end;
00244     int filter_value;
00245     const int nstride= -stride;
00246 
00247     for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
00248         filter_value =
00249             (first_pixel[2 * nstride] - first_pixel[ stride])
00250          +3*(first_pixel[0          ] - first_pixel[nstride]);
00251         filter_value = bounding_values[(filter_value + 4) >> 3];
00252         first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
00253         first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
00254     }
00255 }
00256 
00257 void ff_vp3_h_loop_filter_c(uint8_t *first_pixel, int stride, int *bounding_values)
00258 {
00259     unsigned char *end;
00260     int filter_value;
00261 
00262     for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
00263         filter_value =
00264             (first_pixel[-2] - first_pixel[ 1])
00265          +3*(first_pixel[ 0] - first_pixel[-1]);
00266         filter_value = bounding_values[(filter_value + 4) >> 3];
00267         first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
00268         first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
00269     }
00270 }