yading@10: /*
yading@10:  * MSMPEG4 backend for encoder and decoder
yading@10:  * Copyright (c) 2001 Fabrice Bellard
yading@10:  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
yading@10:  *
yading@10:  * msmpeg4v1 & v2 stuff by Michael Niedermayer <michaelni@gmx.at>
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * MSMPEG4 backend for encoder and decoder
yading@10:  */
yading@10: 
yading@10: #include "avcodec.h"
yading@10: #include "dsputil.h"
yading@10: #include "mpegvideo.h"
yading@10: #include "msmpeg4.h"
yading@10: #include "libavutil/x86/asm.h"
yading@10: #include "h263.h"
yading@10: #include "mpeg4video.h"
yading@10: #include "msmpeg4data.h"
yading@10: #include "vc1data.h"
yading@10: #include "libavutil/imgutils.h"
yading@10: 
yading@10: /*
yading@10:  * You can also call this codec : MPEG4 with a twist !
yading@10:  *
yading@10:  * TODO:
yading@10:  *        - (encoding) select best mv table (two choices)
yading@10:  *        - (encoding) select best vlc/dc table
yading@10:  */
yading@10: //#define DEBUG
yading@10: 
yading@10: /* This table is practically identical to the one from h263
yading@10:  * except that it is inverted. */
yading@10: static av_cold void init_h263_dc_for_msmpeg4(void)
yading@10: {
yading@10:         int level, uni_code, uni_len;
yading@10: 
yading@10:         if(ff_v2_dc_chroma_table[255 + 256][1])
yading@10:             return;
yading@10: 
yading@10:         for(level=-256; level<256; level++){
yading@10:             int size, v, l;
yading@10:             /* find number of bits */
yading@10:             size = 0;
yading@10:             v = abs(level);
yading@10:             while (v) {
yading@10:                 v >>= 1;
yading@10:                     size++;
yading@10:             }
yading@10: 
yading@10:             if (level < 0)
yading@10:                 l= (-level) ^ ((1 << size) - 1);
yading@10:             else
yading@10:                 l= level;
yading@10: 
yading@10:             /* luminance h263 */
yading@10:             uni_code= ff_mpeg4_DCtab_lum[size][0];
yading@10:             uni_len = ff_mpeg4_DCtab_lum[size][1];
yading@10:             uni_code ^= (1<<uni_len)-1; //M$ does not like compatibility
yading@10: 
yading@10:             if (size > 0) {
yading@10:                 uni_code<<=size; uni_code|=l;
yading@10:                 uni_len+=size;
yading@10:                 if (size > 8){
yading@10:                     uni_code<<=1; uni_code|=1;
yading@10:                     uni_len++;
yading@10:                 }
yading@10:             }
yading@10:             ff_v2_dc_lum_table[level + 256][0] = uni_code;
yading@10:             ff_v2_dc_lum_table[level + 256][1] = uni_len;
yading@10: 
yading@10:             /* chrominance h263 */
yading@10:             uni_code= ff_mpeg4_DCtab_chrom[size][0];
yading@10:             uni_len = ff_mpeg4_DCtab_chrom[size][1];
yading@10:             uni_code ^= (1<<uni_len)-1; //M$ does not like compatibility
yading@10: 
yading@10:             if (size > 0) {
yading@10:                 uni_code<<=size; uni_code|=l;
yading@10:                 uni_len+=size;
yading@10:                 if (size > 8){
yading@10:                     uni_code<<=1; uni_code|=1;
yading@10:                     uni_len++;
yading@10:                 }
yading@10:             }
yading@10:             ff_v2_dc_chroma_table[level + 256][0] = uni_code;
yading@10:             ff_v2_dc_chroma_table[level + 256][1] = uni_len;
yading@10: 
yading@10:         }
yading@10: }
yading@10: 
yading@10: av_cold void ff_msmpeg4_common_init(MpegEncContext *s)
yading@10: {
yading@10:     switch(s->msmpeg4_version){
yading@10:     case 1:
yading@10:     case 2:
yading@10:         s->y_dc_scale_table=
yading@10:         s->c_dc_scale_table= ff_mpeg1_dc_scale_table;
yading@10:         break;
yading@10:     case 3:
yading@10:         if(s->workaround_bugs){
yading@10:             s->y_dc_scale_table= ff_old_ff_y_dc_scale_table;
yading@10:             s->c_dc_scale_table= ff_wmv1_c_dc_scale_table;
yading@10:         } else{
yading@10:             s->y_dc_scale_table= ff_mpeg4_y_dc_scale_table;
yading@10:             s->c_dc_scale_table= ff_mpeg4_c_dc_scale_table;
yading@10:         }
yading@10:         break;
yading@10:     case 4:
yading@10:     case 5:
yading@10:         s->y_dc_scale_table= ff_wmv1_y_dc_scale_table;
yading@10:         s->c_dc_scale_table= ff_wmv1_c_dc_scale_table;
yading@10:         break;
yading@10: #if CONFIG_VC1_DECODER
yading@10:     case 6:
yading@10:         s->y_dc_scale_table= ff_wmv3_dc_scale_table;
yading@10:         s->c_dc_scale_table= ff_wmv3_dc_scale_table;
yading@10:         break;
yading@10: #endif
yading@10: 
yading@10:     }
yading@10: 
yading@10: 
yading@10:     if(s->msmpeg4_version>=4){
yading@10:         ff_init_scantable(s->dsp.idct_permutation, &s->intra_scantable  , ff_wmv1_scantable[1]);
yading@10:         ff_init_scantable(s->dsp.idct_permutation, &s->intra_h_scantable, ff_wmv1_scantable[2]);
yading@10:         ff_init_scantable(s->dsp.idct_permutation, &s->intra_v_scantable, ff_wmv1_scantable[3]);
yading@10:         ff_init_scantable(s->dsp.idct_permutation, &s->inter_scantable  , ff_wmv1_scantable[0]);
yading@10:     }
yading@10:     //Note the default tables are set in common_init in mpegvideo.c
yading@10: 
yading@10:     init_h263_dc_for_msmpeg4();
yading@10: }
yading@10: 
yading@10: /* predict coded block */
yading@10: int ff_msmpeg4_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)
yading@10: {
yading@10:     int xy, wrap, pred, a, b, c;
yading@10: 
yading@10:     xy = s->block_index[n];
yading@10:     wrap = s->b8_stride;
yading@10: 
yading@10:     /* B C
yading@10:      * A X
yading@10:      */
yading@10:     a = s->coded_block[xy - 1       ];
yading@10:     b = s->coded_block[xy - 1 - wrap];
yading@10:     c = s->coded_block[xy     - wrap];
yading@10: 
yading@10:     if (b == c) {
yading@10:         pred = a;
yading@10:     } else {
yading@10:         pred = c;
yading@10:     }
yading@10: 
yading@10:     /* store value */
yading@10:     *coded_block_ptr = &s->coded_block[xy];
yading@10: 
yading@10:     return pred;
yading@10: }
yading@10: 
yading@10: static int get_dc(uint8_t *src, int stride, int scale)
yading@10: {
yading@10:     int y;
yading@10:     int sum=0;
yading@10:     for(y=0; y<8; y++){
yading@10:         int x;
yading@10:         for(x=0; x<8; x++){
yading@10:             sum+=src[x + y*stride];
yading@10:         }
yading@10:     }
yading@10:     return FASTDIV((sum + (scale>>1)), scale);
yading@10: }
yading@10: 
yading@10: /* dir = 0: left, dir = 1: top prediction */
yading@10: int ff_msmpeg4_pred_dc(MpegEncContext *s, int n,
yading@10:                        int16_t **dc_val_ptr, int *dir_ptr)
yading@10: {
yading@10:     int a, b, c, wrap, pred, scale;
yading@10:     int16_t *dc_val;
yading@10: 
yading@10:     /* find prediction */
yading@10:     if (n < 4) {
yading@10:         scale = s->y_dc_scale;
yading@10:     } else {
yading@10:         scale = s->c_dc_scale;
yading@10:     }
yading@10: 
yading@10:     wrap = s->block_wrap[n];
yading@10:     dc_val= s->dc_val[0] + s->block_index[n];
yading@10: 
yading@10:     /* B C
yading@10:      * A X
yading@10:      */
yading@10:     a = dc_val[ - 1];
yading@10:     b = dc_val[ - 1 - wrap];
yading@10:     c = dc_val[ - wrap];
yading@10: 
yading@10:     if(s->first_slice_line && (n&2)==0 && s->msmpeg4_version<4){
yading@10:         b=c=1024;
yading@10:     }
yading@10: 
yading@10:     /* XXX: the following solution consumes divisions, but it does not
yading@10:        necessitate to modify mpegvideo.c. The problem comes from the
yading@10:        fact they decided to store the quantized DC (which would lead
yading@10:        to problems if Q could vary !) */
yading@10: #if ARCH_X86 && HAVE_7REGS && HAVE_EBX_AVAILABLE
yading@10:     __asm__ volatile(
yading@10:         "movl %3, %%eax         \n\t"
yading@10:         "shrl $1, %%eax         \n\t"
yading@10:         "addl %%eax, %2         \n\t"
yading@10:         "addl %%eax, %1         \n\t"
yading@10:         "addl %0, %%eax         \n\t"
yading@10:         "imull %4               \n\t"
yading@10:         "movl %%edx, %0         \n\t"
yading@10:         "movl %1, %%eax         \n\t"
yading@10:         "imull %4               \n\t"
yading@10:         "movl %%edx, %1         \n\t"
yading@10:         "movl %2, %%eax         \n\t"
yading@10:         "imull %4               \n\t"
yading@10:         "movl %%edx, %2         \n\t"
yading@10:         : "+b" (a), "+c" (b), "+D" (c)
yading@10:         : "g" (scale), "S" (ff_inverse[scale])
yading@10:         : "%eax", "%edx"
yading@10:     );
yading@10: #else
yading@10:     /* #elif ARCH_ALPHA */
yading@10:     /* Divisions are extremely costly on Alpha; optimize the most
yading@10:        common case. But they are costly everywhere...
yading@10:      */
yading@10:     if (scale == 8) {
yading@10:         a = (a + (8 >> 1)) / 8;
yading@10:         b = (b + (8 >> 1)) / 8;
yading@10:         c = (c + (8 >> 1)) / 8;
yading@10:     } else {
yading@10:         a = FASTDIV((a + (scale >> 1)), scale);
yading@10:         b = FASTDIV((b + (scale >> 1)), scale);
yading@10:         c = FASTDIV((c + (scale >> 1)), scale);
yading@10:     }
yading@10: #endif
yading@10:     /* XXX: WARNING: they did not choose the same test as MPEG4. This
yading@10:        is very important ! */
yading@10:     if(s->msmpeg4_version>3){
yading@10:         if(s->inter_intra_pred){
yading@10:             uint8_t *dest;
yading@10:             int wrap;
yading@10: 
yading@10:             if(n==1){
yading@10:                 pred=a;
yading@10:                 *dir_ptr = 0;
yading@10:             }else if(n==2){
yading@10:                 pred=c;
yading@10:                 *dir_ptr = 1;
yading@10:             }else if(n==3){
yading@10:                 if (abs(a - b) < abs(b - c)) {
yading@10:                     pred = c;
yading@10:                     *dir_ptr = 1;
yading@10:                 } else {
yading@10:                     pred = a;
yading@10:                     *dir_ptr = 0;
yading@10:                 }
yading@10:             }else{
yading@10:                 if(n<4){
yading@10:                     wrap= s->linesize;
yading@10:                     dest= s->current_picture.f.data[0] + (((n >> 1) + 2*s->mb_y) * 8*  wrap ) + ((n & 1) + 2*s->mb_x) * 8;
yading@10:                 }else{
yading@10:                     wrap= s->uvlinesize;
yading@10:                     dest= s->current_picture.f.data[n - 3] + (s->mb_y * 8 * wrap) + s->mb_x * 8;
yading@10:                 }
yading@10:                 if(s->mb_x==0) a= (1024 + (scale>>1))/scale;
yading@10:                 else           a= get_dc(dest-8, wrap, scale*8);
yading@10:                 if(s->mb_y==0) c= (1024 + (scale>>1))/scale;
yading@10:                 else           c= get_dc(dest-8*wrap, wrap, scale*8);
yading@10: 
yading@10:                 if (s->h263_aic_dir==0) {
yading@10:                     pred= a;
yading@10:                     *dir_ptr = 0;
yading@10:                 }else if (s->h263_aic_dir==1) {
yading@10:                     if(n==0){
yading@10:                         pred= c;
yading@10:                         *dir_ptr = 1;
yading@10:                     }else{
yading@10:                         pred= a;
yading@10:                         *dir_ptr = 0;
yading@10:                     }
yading@10:                 }else if (s->h263_aic_dir==2) {
yading@10:                     if(n==0){
yading@10:                         pred= a;
yading@10:                         *dir_ptr = 0;
yading@10:                     }else{
yading@10:                         pred= c;
yading@10:                         *dir_ptr = 1;
yading@10:                     }
yading@10:                 } else {
yading@10:                     pred= c;
yading@10:                     *dir_ptr = 1;
yading@10:                 }
yading@10:             }
yading@10:         }else{
yading@10:             if (abs(a - b) < abs(b - c)) {
yading@10:                 pred = c;
yading@10:                 *dir_ptr = 1;
yading@10:             } else {
yading@10:                 pred = a;
yading@10:                 *dir_ptr = 0;
yading@10:             }
yading@10:         }
yading@10:     }else{
yading@10:         if (abs(a - b) <= abs(b - c)) {
yading@10:             pred = c;
yading@10:             *dir_ptr = 1;
yading@10:         } else {
yading@10:             pred = a;
yading@10:             *dir_ptr = 0;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     /* update predictor */
yading@10:     *dc_val_ptr = &dc_val[0];
yading@10:     return pred;
yading@10: }
yading@10: