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 "libavutil/attributes.h"
00028 #include "libavutil/common.h"
00029 #include "avcodec.h"
00030 #include "dsputil.h"
00031 #include "vp3dsp.h"
00032 
00033 #define IdctAdjustBeforeShift 8
00034 #define xC1S7 64277
00035 #define xC2S6 60547
00036 #define xC3S5 54491
00037 #define xC4S4 46341
00038 #define xC5S3 36410
00039 #define xC6S2 25080
00040 #define xC7S1 12785
00041 
00042 #define M(a,b) (((a) * (b))>>16)
00043 
00044 static av_always_inline void idct(uint8_t *dst, int stride, int16_t *input, int type)
00045 {
00046     int16_t *ip = input;
00047 
00048     int A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;
00049     int Ed, Gd, Add, Bdd, Fd, Hd;
00050 
00051     int i;
00052 
00053     /* Inverse DCT on the rows now */
00054     for (i = 0; i < 8; i++) {
00055         /* Check for non-zero values */
00056         if ( ip[0] | ip[1] | ip[2] | ip[3] | ip[4] | ip[5] | ip[6] | ip[7] ) {
00057             A = M(xC1S7, ip[1]) + M(xC7S1, ip[7]);
00058             B = M(xC7S1, ip[1]) - M(xC1S7, ip[7]);
00059             C = M(xC3S5, ip[3]) + M(xC5S3, ip[5]);
00060             D = M(xC3S5, ip[5]) - M(xC5S3, ip[3]);
00061 
00062             Ad = M(xC4S4, (A - C));
00063             Bd = M(xC4S4, (B - D));
00064 
00065             Cd = A + C;
00066             Dd = B + D;
00067 
00068             E = M(xC4S4, (ip[0] + ip[4]));
00069             F = M(xC4S4, (ip[0] - ip[4]));
00070 
00071             G = M(xC2S6, ip[2]) + M(xC6S2, ip[6]);
00072             H = M(xC6S2, ip[2]) - M(xC2S6, ip[6]);
00073 
00074             Ed = E - G;
00075             Gd = E + G;
00076 
00077             Add = F + Ad;
00078             Bdd = Bd - H;
00079 
00080             Fd = F - Ad;
00081             Hd = Bd + H;
00082 
00083             /*  Final sequence of operations over-write original inputs. */
00084             ip[0] = Gd + Cd ;
00085             ip[7] = Gd - Cd ;
00086 
00087             ip[1] = Add + Hd;
00088             ip[2] = Add - Hd;
00089 
00090             ip[3] = Ed + Dd ;
00091             ip[4] = Ed - Dd ;
00092 
00093             ip[5] = Fd + Bdd;
00094             ip[6] = Fd - Bdd;
00095         }
00096 
00097         ip += 8;            /* next row */
00098     }
00099 
00100     ip = input;
00101 
00102     for ( i = 0; i < 8; i++) {
00103         /* Check for non-zero values (bitwise or faster than ||) */
00104         if ( ip[1 * 8] | ip[2 * 8] | ip[3 * 8] |
00105              ip[4 * 8] | ip[5 * 8] | ip[6 * 8] | ip[7 * 8] ) {
00106 
00107             A = M(xC1S7, ip[1*8]) + M(xC7S1, ip[7*8]);
00108             B = M(xC7S1, ip[1*8]) - M(xC1S7, ip[7*8]);
00109             C = M(xC3S5, ip[3*8]) + M(xC5S3, ip[5*8]);
00110             D = M(xC3S5, ip[5*8]) - M(xC5S3, ip[3*8]);
00111 
00112             Ad = M(xC4S4, (A - C));
00113             Bd = M(xC4S4, (B - D));
00114 
00115             Cd = A + C;
00116             Dd = B + D;
00117 
00118             E = M(xC4S4, (ip[0*8] + ip[4*8])) + 8;
00119             F = M(xC4S4, (ip[0*8] - ip[4*8])) + 8;
00120 
00121             if(type==1){  //HACK
00122                 E += 16*128;
00123                 F += 16*128;
00124             }
00125 
00126             G = M(xC2S6, ip[2*8]) + M(xC6S2, ip[6*8]);
00127             H = M(xC6S2, ip[2*8]) - M(xC2S6, ip[6*8]);
00128 
00129             Ed = E - G;
00130             Gd = E + G;
00131 
00132             Add = F + Ad;
00133             Bdd = Bd - H;
00134 
00135             Fd = F - Ad;
00136             Hd = Bd + H;
00137 
00138             /* Final sequence of operations over-write original inputs. */
00139             if(type==0){
00140                 ip[0*8] = (Gd + Cd )  >> 4;
00141                 ip[7*8] = (Gd - Cd )  >> 4;
00142 
00143                 ip[1*8] = (Add + Hd ) >> 4;
00144                 ip[2*8] = (Add - Hd ) >> 4;
00145 
00146                 ip[3*8] = (Ed + Dd )  >> 4;
00147                 ip[4*8] = (Ed - Dd )  >> 4;
00148 
00149                 ip[5*8] = (Fd + Bdd ) >> 4;
00150                 ip[6*8] = (Fd - Bdd ) >> 4;
00151             }else if(type==1){
00152                 dst[0*stride] = av_clip_uint8((Gd + Cd )  >> 4);
00153                 dst[7*stride] = av_clip_uint8((Gd - Cd )  >> 4);
00154 
00155                 dst[1*stride] = av_clip_uint8((Add + Hd ) >> 4);
00156                 dst[2*stride] = av_clip_uint8((Add - Hd ) >> 4);
00157 
00158                 dst[3*stride] = av_clip_uint8((Ed + Dd )  >> 4);
00159                 dst[4*stride] = av_clip_uint8((Ed - Dd )  >> 4);
00160 
00161                 dst[5*stride] = av_clip_uint8((Fd + Bdd ) >> 4);
00162                 dst[6*stride] = av_clip_uint8((Fd - Bdd ) >> 4);
00163             }else{
00164                 dst[0*stride] = av_clip_uint8(dst[0*stride] + ((Gd + Cd )  >> 4));
00165                 dst[7*stride] = av_clip_uint8(dst[7*stride] + ((Gd - Cd )  >> 4));
00166 
00167                 dst[1*stride] = av_clip_uint8(dst[1*stride] + ((Add + Hd ) >> 4));
00168                 dst[2*stride] = av_clip_uint8(dst[2*stride] + ((Add - Hd ) >> 4));
00169 
00170                 dst[3*stride] = av_clip_uint8(dst[3*stride] + ((Ed + Dd )  >> 4));
00171                 dst[4*stride] = av_clip_uint8(dst[4*stride] + ((Ed - Dd )  >> 4));
00172 
00173                 dst[5*stride] = av_clip_uint8(dst[5*stride] + ((Fd + Bdd ) >> 4));
00174                 dst[6*stride] = av_clip_uint8(dst[6*stride] + ((Fd - Bdd ) >> 4));
00175             }
00176 
00177         } else {
00178             if(type==0){
00179                 ip[0*8] =
00180                 ip[1*8] =
00181                 ip[2*8] =
00182                 ip[3*8] =
00183                 ip[4*8] =
00184                 ip[5*8] =
00185                 ip[6*8] =
00186                 ip[7*8] = ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
00187             }else if(type==1){
00188                 dst[0*stride]=
00189                 dst[1*stride]=
00190                 dst[2*stride]=
00191                 dst[3*stride]=
00192                 dst[4*stride]=
00193                 dst[5*stride]=
00194                 dst[6*stride]=
00195                 dst[7*stride]= av_clip_uint8(128 + ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20));
00196             }else{
00197                 if(ip[0*8]){
00198                     int v= ((xC4S4 * ip[0*8] + (IdctAdjustBeforeShift<<16))>>20);
00199                     dst[0*stride] = av_clip_uint8(dst[0*stride] + v);
00200                     dst[1*stride] = av_clip_uint8(dst[1*stride] + v);
00201                     dst[2*stride] = av_clip_uint8(dst[2*stride] + v);
00202                     dst[3*stride] = av_clip_uint8(dst[3*stride] + v);
00203                     dst[4*stride] = av_clip_uint8(dst[4*stride] + v);
00204                     dst[5*stride] = av_clip_uint8(dst[5*stride] + v);
00205                     dst[6*stride] = av_clip_uint8(dst[6*stride] + v);
00206                     dst[7*stride] = av_clip_uint8(dst[7*stride] + v);
00207                 }
00208             }
00209         }
00210 
00211         ip++;            /* next column */
00212         dst++;
00213     }
00214 }
00215 
00216 static void vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
00217     idct(dest, line_size, block, 1);
00218 }
00219 
00220 static void vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/){
00221     idct(dest, line_size, block, 2);
00222 }
00223 
00224 static void vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size,
00225                               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 static void vp3_v_loop_filter_c(uint8_t *first_pixel, int stride,
00242                                 int *bounding_values)
00243 {
00244     unsigned char *end;
00245     int filter_value;
00246     const int nstride= -stride;
00247 
00248     for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
00249         filter_value =
00250             (first_pixel[2 * nstride] - first_pixel[ stride])
00251          +3*(first_pixel[0          ] - first_pixel[nstride]);
00252         filter_value = bounding_values[(filter_value + 4) >> 3];
00253         first_pixel[nstride] = av_clip_uint8(first_pixel[nstride] + filter_value);
00254         first_pixel[0] = av_clip_uint8(first_pixel[0] - filter_value);
00255     }
00256 }
00257 
00258 static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
00259                                 int *bounding_values)
00260 {
00261     unsigned char *end;
00262     int filter_value;
00263 
00264     for (end= first_pixel + 8*stride; first_pixel != end; first_pixel += stride) {
00265         filter_value =
00266             (first_pixel[-2] - first_pixel[ 1])
00267          +3*(first_pixel[ 0] - first_pixel[-1]);
00268         filter_value = bounding_values[(filter_value + 4) >> 3];
00269         first_pixel[-1] = av_clip_uint8(first_pixel[-1] + filter_value);
00270         first_pixel[ 0] = av_clip_uint8(first_pixel[ 0] - filter_value);
00271     }
00272 }
00273 
00274 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
00275 {
00276     c->idct_put      = vp3_idct_put_c;
00277     c->idct_add      = vp3_idct_add_c;
00278     c->idct_dc_add   = vp3_idct_dc_add_c;
00279     c->v_loop_filter = vp3_v_loop_filter_c;
00280     c->h_loop_filter = vp3_h_loop_filter_c;
00281 
00282     c->idct_perm = FF_NO_IDCT_PERM;
00283 
00284     if (ARCH_ARM)
00285         ff_vp3dsp_init_arm(c, flags);
00286     if (ARCH_PPC)
00287         ff_vp3dsp_init_ppc(c, flags);
00288     if (ARCH_X86)
00289         ff_vp3dsp_init_x86(c, flags);
00290 }