/*
* The simplest mpeg encoder (well, it was the simplest!)
* Copyright (c) 2000,2001 Fabrice Bellard.
+ * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* The simplest mpeg encoder (well, it was the simplest!).
*/
-#include <limits.h>
#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"
#include "faandct.h"
+#include <limits.h>
#ifdef USE_FASTMEMCPY
#include "fastmemcpy.h"
#ifdef CONFIG_ENCODERS
static void encode_picture(MpegEncContext *s, int picture_number);
#endif //CONFIG_ENCODERS
-static void dct_unquantize_mpeg1_c(MpegEncContext *s,
+static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
-static void dct_unquantize_mpeg2_c(MpegEncContext *s,
+static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
-static void dct_unquantize_h263_c(MpegEncContext *s,
+static void dct_unquantize_h263_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_h263_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_h261_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
+static void dct_unquantize_h261_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w);
#ifdef CONFIG_ENCODERS
static int dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow);
static int dct_quantize_trellis_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow);
+static int dct_quantize_refine(MpegEncContext *s, DCTELEM *block, int16_t *weight, DCTELEM *orig, int n, int qscale);
static int sse_mb(MpegEncContext *s);
+static void denoise_dct_c(MpegEncContext *s, DCTELEM *block);
#endif //CONFIG_ENCODERS
#ifdef HAVE_XVMC
/* init common dct for both encoder and decoder */
int DCT_common_init(MpegEncContext *s)
{
- s->dct_unquantize_h263 = dct_unquantize_h263_c;
- s->dct_unquantize_mpeg1 = dct_unquantize_mpeg1_c;
- s->dct_unquantize_mpeg2 = dct_unquantize_mpeg2_c;
+ s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_c;
+ s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_c;
+ s->dct_unquantize_h261_intra = dct_unquantize_h261_intra_c;
+ s->dct_unquantize_h261_inter = dct_unquantize_h261_inter_c;
+ s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
+ s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_c;
+ s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_c;
+ s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c;
#ifdef CONFIG_ENCODERS
s->dct_quantize= dct_quantize_c;
+ s->denoise_dct= denoise_dct_c;
#endif
#ifdef HAVE_MMX
/* load & permutate scantables
note: only wmv uses differnt ones
*/
- ff_init_scantable(s->dsp.idct_permutation, &s->inter_scantable , ff_zigzag_direct);
- ff_init_scantable(s->dsp.idct_permutation, &s->intra_scantable , ff_zigzag_direct);
+ if(s->alternate_scan){
+ ff_init_scantable(s->dsp.idct_permutation, &s->inter_scantable , ff_alternate_vertical_scan);
+ ff_init_scantable(s->dsp.idct_permutation, &s->intra_scantable , ff_alternate_vertical_scan);
+ }else{
+ ff_init_scantable(s->dsp.idct_permutation, &s->inter_scantable , ff_zigzag_direct);
+ ff_init_scantable(s->dsp.idct_permutation, &s->intra_scantable , ff_zigzag_direct);
+ }
ff_init_scantable(s->dsp.idct_permutation, &s->intra_h_scantable, ff_alternate_horizontal_scan);
ff_init_scantable(s->dsp.idct_permutation, &s->intra_v_scantable, ff_alternate_vertical_scan);
- s->picture_structure= PICT_FRAME;
-
return 0;
}
dst->type= FF_BUFFER_TYPE_COPY;
}
+static void copy_picture_attributes(MpegEncContext *s, AVFrame *dst, AVFrame *src){
+ int i;
+
+ dst->pict_type = src->pict_type;
+ dst->quality = src->quality;
+ dst->coded_picture_number = src->coded_picture_number;
+ dst->display_picture_number = src->display_picture_number;
+// dst->reference = src->reference;
+ dst->pts = src->pts;
+ dst->interlaced_frame = src->interlaced_frame;
+ dst->top_field_first = src->top_field_first;
+
+ if(s->avctx->me_threshold){
+ if(!src->motion_val[0])
+ av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_val not set!\n");
+ if(!src->mb_type)
+ av_log(s->avctx, AV_LOG_ERROR, "AVFrame.mb_type not set!\n");
+ if(!src->ref_index[0])
+ av_log(s->avctx, AV_LOG_ERROR, "AVFrame.ref_index not set!\n");
+ if(src->motion_subsample_log2 != dst->motion_subsample_log2)
+ av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_subsample_log2 doesnt match! (%d!=%d)\n",
+ src->motion_subsample_log2, dst->motion_subsample_log2);
+
+ memcpy(dst->mb_type, src->mb_type, s->mb_stride * s->mb_height * sizeof(dst->mb_type[0]));
+
+ for(i=0; i<2; i++){
+ int stride= ((16*s->mb_width )>>src->motion_subsample_log2) + 1;
+ int height= ((16*s->mb_height)>>src->motion_subsample_log2);
+
+ if(src->motion_val[i] && src->motion_val[i] != dst->motion_val[i]){
+ memcpy(dst->motion_val[i], src->motion_val[i], 2*stride*height*sizeof(int16_t));
+ }
+ if(src->ref_index[i] && src->ref_index[i] != dst->ref_index[i]){
+ memcpy(dst->ref_index[i], src->ref_index[i], s->b8_stride*2*s->mb_height*sizeof(int8_t));
+ }
+ }
+ }
+}
+
/**
* allocates a Picture
* The pixels are allocated/set by calling get_buffer() if shared=0
CHECKED_ALLOCZ(pic->mbskip_table , mb_array_size * sizeof(uint8_t)+2) //the +2 is for the slice end check
CHECKED_ALLOCZ(pic->qscale_table , mb_array_size * sizeof(uint8_t))
- CHECKED_ALLOCZ(pic->mb_type_base , big_mb_num * sizeof(int))
+ CHECKED_ALLOCZ(pic->mb_type_base , big_mb_num * sizeof(uint32_t))
pic->mb_type= pic->mb_type_base + s->mb_stride+1;
if(s->out_format == FMT_H264){
for(i=0; i<2; i++){
- CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b4_array_size+1) * sizeof(uint16_t))
- pic->motion_val[i]= pic->motion_val_base[i]+1;
- CHECKED_ALLOCZ(pic->ref_index[i] , b8_array_size * sizeof(uint8_t))
+ CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b4_array_size+2) * sizeof(int16_t))
+ pic->motion_val[i]= pic->motion_val_base[i]+2;
+ CHECKED_ALLOCZ(pic->ref_index[i], b8_array_size * sizeof(uint8_t))
}
pic->motion_subsample_log2= 2;
- }else if(s->out_format == FMT_H263 || s->encoding || (s->avctx->debug&(FF_DEBUG_VIS_MV|FF_DEBUG_MV))){
+ }else if(s->out_format == FMT_H263 || s->encoding || (s->avctx->debug&FF_DEBUG_MV) || (s->avctx->debug_mv)){
for(i=0; i<2; i++){
- CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b8_array_size+1) * sizeof(uint16_t)*2) //FIXME
- pic->motion_val[i]= pic->motion_val_base[i]+1;
+ CHECKED_ALLOCZ(pic->motion_val_base[i], 2 * (b8_array_size+2) * sizeof(int16_t))
+ pic->motion_val[i]= pic->motion_val_base[i]+2;
+ CHECKED_ALLOCZ(pic->ref_index[i], b8_array_size * sizeof(uint8_t))
}
pic->motion_subsample_log2= 3;
}
+ if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
+ CHECKED_ALLOCZ(pic->dct_coeff, 64 * mb_array_size * sizeof(DCTELEM)*6)
+ }
pic->qstride= s->mb_stride;
CHECKED_ALLOCZ(pic->pan_scan , 1 * sizeof(AVPanScan))
}
av_freep(&pic->mbskip_table);
av_freep(&pic->qscale_table);
av_freep(&pic->mb_type_base);
+ av_freep(&pic->dct_coeff);
av_freep(&pic->pan_scan);
pic->mb_type= NULL;
for(i=0; i<2; i++){
}
}
-/* init common structure for both encoder and decoder */
+static int init_duplicate_context(MpegEncContext *s, MpegEncContext *base){
+ int i;
+
+ // edge emu needs blocksize + filter length - 1 (=17x17 for halfpel / 21x21 for h264)
+ CHECKED_ALLOCZ(s->allocated_edge_emu_buffer, (s->width+64)*2*17*2); //(width + edge + align)*interlaced*MBsize*tolerance
+ s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*17;
+
+ //FIXME should be linesize instead of s->width*2 but that isnt known before get_buffer()
+ CHECKED_ALLOCZ(s->me.scratchpad, (s->width+64)*4*16*2*sizeof(uint8_t))
+ s->rd_scratchpad= s->me.scratchpad;
+ s->b_scratchpad= s->me.scratchpad;
+ s->obmc_scratchpad= s->me.scratchpad + 16;
+ if (s->encoding) {
+ CHECKED_ALLOCZ(s->me.map , ME_MAP_SIZE*sizeof(uint32_t))
+ CHECKED_ALLOCZ(s->me.score_map, ME_MAP_SIZE*sizeof(uint32_t))
+ if(s->avctx->noise_reduction){
+ CHECKED_ALLOCZ(s->dct_error_sum, 2 * 64 * sizeof(int))
+ }
+ }
+ CHECKED_ALLOCZ(s->blocks, 64*12*2 * sizeof(DCTELEM))
+ s->block= s->blocks[0];
+
+ for(i=0;i<12;i++){
+ s->pblocks[i] = (short *)(&s->block[i]);
+ }
+ return 0;
+fail:
+ return -1; //free() through MPV_common_end()
+}
+
+static void free_duplicate_context(MpegEncContext *s){
+ if(s==NULL) return;
+
+ av_freep(&s->allocated_edge_emu_buffer); s->edge_emu_buffer= NULL;
+ av_freep(&s->me.scratchpad);
+ s->rd_scratchpad=
+ s->b_scratchpad=
+ s->obmc_scratchpad= NULL;
+
+ av_freep(&s->dct_error_sum);
+ av_freep(&s->me.map);
+ av_freep(&s->me.score_map);
+ av_freep(&s->blocks);
+ s->block= NULL;
+}
+
+static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src){
+#define COPY(a) bak->a= src->a
+ COPY(allocated_edge_emu_buffer);
+ COPY(edge_emu_buffer);
+ COPY(me.scratchpad);
+ COPY(rd_scratchpad);
+ COPY(b_scratchpad);
+ COPY(obmc_scratchpad);
+ COPY(me.map);
+ COPY(me.score_map);
+ COPY(blocks);
+ COPY(block);
+ COPY(start_mb_y);
+ COPY(end_mb_y);
+ COPY(me.map_generation);
+ COPY(pb);
+ COPY(dct_error_sum);
+ COPY(dct_count[0]);
+ COPY(dct_count[1]);
+#undef COPY
+}
+
+void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src){
+ MpegEncContext bak;
+ int i;
+ //FIXME copy only needed parts
+//START_TIMER
+ backup_duplicate_context(&bak, dst);
+ memcpy(dst, src, sizeof(MpegEncContext));
+ backup_duplicate_context(dst, &bak);
+ for(i=0;i<12;i++){
+ dst->pblocks[i] = (short *)(&dst->block[i]);
+ }
+//STOP_TIMER("update_duplicate_context") //about 10k cycles / 0.01 sec for 1000frames on 1ghz with 2 threads
+}
+
+static void update_duplicate_context_after_me(MpegEncContext *dst, MpegEncContext *src){
+#define COPY(a) dst->a= src->a
+ COPY(pict_type);
+ COPY(current_picture);
+ COPY(f_code);
+ COPY(b_code);
+ COPY(qscale);
+ COPY(lambda);
+ COPY(lambda2);
+ COPY(picture_in_gop_number);
+ COPY(gop_picture_number);
+ COPY(frame_pred_frame_dct); //FIXME dont set in encode_header
+ COPY(progressive_frame); //FIXME dont set in encode_header
+ COPY(partitioned_frame); //FIXME dont set in encode_header
+#undef COPY
+}
+
+/**
+ * sets the given MpegEncContext to common defaults (same for encoding and decoding).
+ * the changed fields will not depend upon the prior state of the MpegEncContext.
+ */
+static void MPV_common_defaults(MpegEncContext *s){
+ s->y_dc_scale_table=
+ s->c_dc_scale_table= ff_mpeg1_dc_scale_table;
+ s->chroma_qscale_table= ff_default_chroma_qscale_table;
+ s->progressive_frame= 1;
+ s->progressive_sequence= 1;
+ s->picture_structure= PICT_FRAME;
+
+ s->coded_picture_number = 0;
+ s->picture_number = 0;
+ s->input_picture_number = 0;
+
+ s->picture_in_gop_number = 0;
+
+ s->f_code = 1;
+ s->b_code = 1;
+}
+
+/**
+ * sets the given MpegEncContext to defaults for decoding.
+ * the changed fields will not depend upon the prior state of the MpegEncContext.
+ */
+void MPV_decode_defaults(MpegEncContext *s){
+ MPV_common_defaults(s);
+}
+
+/**
+ * sets the given MpegEncContext to defaults for encoding.
+ * the changed fields will not depend upon the prior state of the MpegEncContext.
+ */
+
+#ifdef CONFIG_ENCODERS
+static void MPV_encode_defaults(MpegEncContext *s){
+ static int done=0;
+
+ MPV_common_defaults(s);
+
+ if(!done){
+ int i;
+ done=1;
+
+ default_mv_penalty= av_mallocz( sizeof(uint8_t)*(MAX_FCODE+1)*(2*MAX_MV+1) );
+ memset(default_mv_penalty, 0, sizeof(uint8_t)*(MAX_FCODE+1)*(2*MAX_MV+1));
+ memset(default_fcode_tab , 0, sizeof(uint8_t)*(2*MAX_MV+1));
+
+ for(i=-16; i<16; i++){
+ default_fcode_tab[i + MAX_MV]= 1;
+ }
+ }
+ s->me.mv_penalty= default_mv_penalty;
+ s->fcode_tab= default_fcode_tab;
+}
+#endif //CONFIG_ENCODERS
+
+/**
+ * init common structure for both encoder and decoder.
+ * this assumes that some variables like width/height are already set
+ */
int MPV_common_init(MpegEncContext *s)
{
- int y_size, c_size, yc_size, i, mb_array_size, x, y;
+ int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y;
+
+ if(s->avctx->thread_count > MAX_THREADS || (16*s->avctx->thread_count > s->height && s->height)){
+ av_log(s->avctx, AV_LOG_ERROR, "too many threads\n");
+ return -1;
+ }
dsputil_init(&s->dsp, s->avctx);
DCT_common_init(s);
s->flags= s->avctx->flags;
+ s->flags2= s->avctx->flags2;
s->mb_width = (s->width + 15) / 16;
s->mb_height = (s->height + 15) / 16;
s->b8_stride = s->mb_width*2 + 1;
s->b4_stride = s->mb_width*4 + 1;
mb_array_size= s->mb_height * s->mb_stride;
+ mv_table_size= (s->mb_height+2) * s->mb_stride + 1;
+
+ /* set chroma shifts */
+ avcodec_get_chroma_sub_sample(s->avctx->pix_fmt,&(s->chroma_x_shift),
+ &(s->chroma_y_shift) );
/* set default edge pos, will be overriden in decode_header if needed */
s->h_edge_pos= s->mb_width*16;
s->block_wrap[0]=
s->block_wrap[1]=
s->block_wrap[2]=
- s->block_wrap[3]= s->mb_width*2 + 2;
+ s->block_wrap[3]= s->b8_stride;
s->block_wrap[4]=
- s->block_wrap[5]= s->mb_width + 2;
-
- s->y_dc_scale_table=
- s->c_dc_scale_table= ff_mpeg1_dc_scale_table;
- s->chroma_qscale_table= ff_default_chroma_qscale_table;
- if (!s->encoding)
- s->progressive_sequence= 1;
- s->progressive_frame= 1;
-
- y_size = (2 * s->mb_width + 2) * (2 * s->mb_height + 2);
- c_size = (s->mb_width + 2) * (s->mb_height + 2);
+ s->block_wrap[5]= s->mb_stride;
+
+ y_size = s->b8_stride * (2 * s->mb_height + 1);
+ c_size = s->mb_stride * (s->mb_height + 1);
yc_size = y_size + 2 * c_size;
-
+
/* convert fourcc to upper case */
s->avctx->codec_tag= toupper( s->avctx->codec_tag &0xFF)
+ (toupper((s->avctx->codec_tag>>8 )&0xFF)<<8 )
+ (toupper((s->avctx->stream_codec_tag>>16)&0xFF)<<16)
+ (toupper((s->avctx->stream_codec_tag>>24)&0xFF)<<24);
- CHECKED_ALLOCZ(s->allocated_edge_emu_buffer, (s->width+64)*2*17*2); //(width + edge + align)*interlaced*MBsize*tolerance
- s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*17;
-
s->avctx->coded_frame= (AVFrame*)&s->current_picture;
CHECKED_ALLOCZ(s->mb_index2xy, (s->mb_num+1)*sizeof(int)) //error ressilience code looks cleaner with this
s->mb_index2xy[ s->mb_height*s->mb_width ] = (s->mb_height-1)*s->mb_stride + s->mb_width; //FIXME really needed?
if (s->encoding) {
- int mv_table_size= s->mb_stride * (s->mb_height+2) + 1;
-
/* Allocate MV tables */
CHECKED_ALLOCZ(s->p_mv_table_base , mv_table_size * 2 * sizeof(int16_t))
CHECKED_ALLOCZ(s->b_forw_mv_table_base , mv_table_size * 2 * sizeof(int16_t))
s->b_bidir_back_mv_table= s->b_bidir_back_mv_table_base + s->mb_stride + 1;
s->b_direct_mv_table = s->b_direct_mv_table_base + s->mb_stride + 1;
- //FIXME should be linesize instead of s->width*2 but that isnt known before get_buffer()
- CHECKED_ALLOCZ(s->me.scratchpad, s->width*2*16*3*sizeof(uint8_t))
-
- CHECKED_ALLOCZ(s->me.map , ME_MAP_SIZE*sizeof(uint32_t))
- CHECKED_ALLOCZ(s->me.score_map, ME_MAP_SIZE*sizeof(uint32_t))
-
- if(s->codec_id==CODEC_ID_MPEG4){
- CHECKED_ALLOCZ(s->tex_pb_buffer, PB_BUFFER_SIZE);
- CHECKED_ALLOCZ( s->pb2_buffer, PB_BUFFER_SIZE);
- }
-
if(s->msmpeg4_version){
CHECKED_ALLOCZ(s->ac_stats, 2*2*(MAX_LEVEL+1)*(MAX_RUN+1)*2*sizeof(int));
}
CHECKED_ALLOCZ(s->avctx->stats_out, 256);
/* Allocate MB type table */
- CHECKED_ALLOCZ(s->mb_type , mb_array_size * sizeof(uint8_t)) //needed for encoding
+ CHECKED_ALLOCZ(s->mb_type , mb_array_size * sizeof(uint16_t)) //needed for encoding
CHECKED_ALLOCZ(s->lambda_table, mb_array_size * sizeof(int))
CHECKED_ALLOCZ(s->reordered_input_picture, MAX_PICTURE_COUNT * sizeof(Picture*))
if(s->avctx->noise_reduction){
- CHECKED_ALLOCZ(s->dct_error_sum, 2 * 64 * sizeof(int))
CHECKED_ALLOCZ(s->dct_offset, 2 * 64 * sizeof(uint16_t))
}
}
- CHECKED_ALLOCZ(s->blocks, 64*6*2 * sizeof(DCTELEM))
-
CHECKED_ALLOCZ(s->picture, MAX_PICTURE_COUNT * sizeof(Picture))
CHECKED_ALLOCZ(s->error_status_table, mb_array_size*sizeof(uint8_t))
- if(s->codec_id==CODEC_ID_MPEG4){
+ if(s->codec_id==CODEC_ID_MPEG4 || (s->flags & CODEC_FLAG_INTERLACED_ME)){
/* interlaced direct mode decoding tables */
- CHECKED_ALLOCZ(s->field_mv_table, mb_array_size*2*2 * sizeof(int16_t))
- CHECKED_ALLOCZ(s->field_select_table, mb_array_size*2* sizeof(int8_t))
+ for(i=0; i<2; i++){
+ int j, k;
+ for(j=0; j<2; j++){
+ for(k=0; k<2; k++){
+ CHECKED_ALLOCZ(s->b_field_mv_table_base[i][j][k] , mv_table_size * 2 * sizeof(int16_t))
+ s->b_field_mv_table[i][j][k] = s->b_field_mv_table_base[i][j][k] + s->mb_stride + 1;
+ }
+ CHECKED_ALLOCZ(s->b_field_select_table[i][j] , mb_array_size * 2 * sizeof(uint8_t))
+ CHECKED_ALLOCZ(s->p_field_mv_table_base[i][j] , mv_table_size * 2 * sizeof(int16_t))
+ s->p_field_mv_table[i][j] = s->p_field_mv_table_base[i][j] + s->mb_stride + 1;
+ }
+ CHECKED_ALLOCZ(s->p_field_select_table[i] , mb_array_size * 2 * sizeof(uint8_t))
+ }
}
if (s->out_format == FMT_H263) {
/* ac values */
- CHECKED_ALLOCZ(s->ac_val[0], yc_size * sizeof(int16_t) * 16);
- s->ac_val[1] = s->ac_val[0] + y_size;
+ CHECKED_ALLOCZ(s->ac_val_base, yc_size * sizeof(int16_t) * 16);
+ s->ac_val[0] = s->ac_val_base + s->b8_stride + 1;
+ s->ac_val[1] = s->ac_val_base + y_size + s->mb_stride + 1;
s->ac_val[2] = s->ac_val[1] + c_size;
/* cbp values */
- CHECKED_ALLOCZ(s->coded_block, y_size);
+ CHECKED_ALLOCZ(s->coded_block_base, y_size);
+ s->coded_block= s->coded_block_base + s->b8_stride + 1;
/* divx501 bitstream reorder buffer */
CHECKED_ALLOCZ(s->bitstream_buffer, BITSTREAM_BUFFER_SIZE);
if (s->h263_pred || s->h263_plus || !s->encoding) {
/* dc values */
//MN: we need these for error resilience of intra-frames
- CHECKED_ALLOCZ(s->dc_val[0], yc_size * sizeof(int16_t));
- s->dc_val[1] = s->dc_val[0] + y_size;
+ CHECKED_ALLOCZ(s->dc_val_base, yc_size * sizeof(int16_t));
+ s->dc_val[0] = s->dc_val_base + s->b8_stride + 1;
+ s->dc_val[1] = s->dc_val_base + y_size + s->mb_stride + 1;
s->dc_val[2] = s->dc_val[1] + c_size;
for(i=0;i<yc_size;i++)
- s->dc_val[0][i] = 1024;
+ s->dc_val_base[i] = 1024;
}
/* which mb is a intra block */
CHECKED_ALLOCZ(s->mbintra_table, mb_array_size);
memset(s->mbintra_table, 1, mb_array_size);
- /* default structure is frame */
- s->picture_structure = PICT_FRAME;
-
/* init macroblock skip table */
CHECKED_ALLOCZ(s->mbskip_table, mb_array_size+2);
//Note the +1 is for a quicker mpeg4 slice_end detection
CHECKED_ALLOCZ(s->prev_pict_types, PREV_PICT_TYPES_BUFFER_SIZE);
- s->block= s->blocks[0];
+ s->parse_context.state= -1;
+ if((s->avctx->debug&(FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)) || (s->avctx->debug_mv)){
+ s->visualization_buffer[0] = av_malloc((s->mb_width*16 + 2*EDGE_WIDTH) * s->mb_height*16 + 2*EDGE_WIDTH);
+ s->visualization_buffer[1] = av_malloc((s->mb_width*8 + EDGE_WIDTH) * s->mb_height*8 + EDGE_WIDTH);
+ s->visualization_buffer[2] = av_malloc((s->mb_width*8 + EDGE_WIDTH) * s->mb_height*8 + EDGE_WIDTH);
+ }
- for(i=0;i<12;i++){
- s->pblocks[i] = (short *)(&s->block[i]);
+ s->context_initialized = 1;
+
+ s->thread_context[0]= s;
+ for(i=1; i<s->avctx->thread_count; i++){
+ s->thread_context[i]= av_malloc(sizeof(MpegEncContext));
+ memcpy(s->thread_context[i], s, sizeof(MpegEncContext));
}
- s->parse_context.state= -1;
+ for(i=0; i<s->avctx->thread_count; i++){
+ if(init_duplicate_context(s->thread_context[i], s) < 0)
+ goto fail;
+ s->thread_context[i]->start_mb_y= (s->mb_height*(i ) + s->avctx->thread_count/2) / s->avctx->thread_count;
+ s->thread_context[i]->end_mb_y = (s->mb_height*(i+1) + s->avctx->thread_count/2) / s->avctx->thread_count;
+ }
- s->context_initialized = 1;
return 0;
fail:
MPV_common_end(s);
return -1;
}
-
-//extern int sads;
-
/* init common structure for both encoder and decoder */
void MPV_common_end(MpegEncContext *s)
{
- int i;
+ int i, j, k;
+
+ for(i=0; i<s->avctx->thread_count; i++){
+ free_duplicate_context(s->thread_context[i]);
+ }
+ for(i=1; i<s->avctx->thread_count; i++){
+ av_freep(&s->thread_context[i]);
+ }
av_freep(&s->parse_context.buffer);
s->parse_context.buffer_size=0;
s->b_bidir_forw_mv_table= NULL;
s->b_bidir_back_mv_table= NULL;
s->b_direct_mv_table= NULL;
+ for(i=0; i<2; i++){
+ for(j=0; j<2; j++){
+ for(k=0; k<2; k++){
+ av_freep(&s->b_field_mv_table_base[i][j][k]);
+ s->b_field_mv_table[i][j][k]=NULL;
+ }
+ av_freep(&s->b_field_select_table[i][j]);
+ av_freep(&s->p_field_mv_table_base[i][j]);
+ s->p_field_mv_table[i][j]=NULL;
+ }
+ av_freep(&s->p_field_select_table[i]);
+ }
- av_freep(&s->dc_val[0]);
- av_freep(&s->ac_val[0]);
- av_freep(&s->coded_block);
+ av_freep(&s->dc_val_base);
+ av_freep(&s->ac_val_base);
+ av_freep(&s->coded_block_base);
av_freep(&s->mbintra_table);
av_freep(&s->cbp_table);
av_freep(&s->pred_dir_table);
- av_freep(&s->me.scratchpad);
- av_freep(&s->me.map);
- av_freep(&s->me.score_map);
av_freep(&s->mbskip_table);
av_freep(&s->prev_pict_types);
av_freep(&s->bitstream_buffer);
- av_freep(&s->tex_pb_buffer);
- av_freep(&s->pb2_buffer);
- av_freep(&s->allocated_edge_emu_buffer); s->edge_emu_buffer= NULL;
- av_freep(&s->field_mv_table);
- av_freep(&s->field_select_table);
av_freep(&s->avctx->stats_out);
av_freep(&s->ac_stats);
av_freep(&s->error_status_table);
av_freep(&s->q_inter_matrix);
av_freep(&s->q_intra_matrix16);
av_freep(&s->q_inter_matrix16);
- av_freep(&s->blocks);
av_freep(&s->input_picture);
av_freep(&s->reordered_input_picture);
- av_freep(&s->dct_error_sum);
av_freep(&s->dct_offset);
if(s->picture){
}
}
av_freep(&s->picture);
- avcodec_default_free_buffers(s->avctx);
s->context_initialized = 0;
s->last_picture_ptr=
s->next_picture_ptr=
s->current_picture_ptr= NULL;
+
+ for(i=0; i<3; i++)
+ av_freep(&s->visualization_buffer[i]);
}
#ifdef CONFIG_ENCODERS
MpegEncContext *s = avctx->priv_data;
int i, dummy;
int chroma_h_shift, chroma_v_shift;
+
+ MPV_encode_defaults(s);
avctx->pix_fmt = PIX_FMT_YUV420P; // FIXME
s->bit_rate = avctx->bit_rate;
- s->bit_rate_tolerance = avctx->bit_rate_tolerance;
s->width = avctx->width;
s->height = avctx->height;
if(avctx->gop_size > 600){
avctx->gop_size=600;
}
s->gop_size = avctx->gop_size;
- s->max_qdiff= avctx->max_qdiff;
- s->qcompress= avctx->qcompress;
- s->qblur= avctx->qblur;
s->avctx = avctx;
s->flags= avctx->flags;
+ s->flags2= avctx->flags2;
s->max_b_frames= avctx->max_b_frames;
- s->b_frame_strategy= avctx->b_frame_strategy;
s->codec_id= avctx->codec->id;
s->luma_elim_threshold = avctx->luma_elim_threshold;
s->chroma_elim_threshold= avctx->chroma_elim_threshold;
s->quarter_sample= (avctx->flags & CODEC_FLAG_QPEL)!=0;
s->mpeg_quant= avctx->mpeg_quant;
s->rtp_mode= !!avctx->rtp_payload_size;
+ s->intra_dc_precision= avctx->intra_dc_precision;
if (s->gop_size <= 1) {
s->intra_only = 1;
s->me_method = avctx->me_method;
/* Fixed QSCALE */
- s->fixed_qscale = (avctx->flags & CODEC_FLAG_QSCALE);
+ s->fixed_qscale = !!(avctx->flags & CODEC_FLAG_QSCALE);
s->adaptive_quant= ( s->avctx->lumi_masking
|| s->avctx->dark_masking
|| (s->flags&CODEC_FLAG_QP_RD))
&& !s->fixed_qscale;
- s->obmc= (s->flags & CODEC_FLAG_OBMC);
- s->loop_filter= (s->flags & CODEC_FLAG_LOOP_FILTER);
+ s->obmc= !!(s->flags & CODEC_FLAG_OBMC);
+ s->loop_filter= !!(s->flags & CODEC_FLAG_LOOP_FILTER);
+ s->alternate_scan= !!(s->flags & CODEC_FLAG_ALT_SCAN);
+
+ if(avctx->rc_max_rate && !avctx->rc_buffer_size){
+ av_log(avctx, AV_LOG_ERROR, "a vbv buffer size is needed, for encoding with a maximum bitrate\n");
+ return -1;
+ }
+ if(avctx->rc_min_rate && avctx->rc_max_rate != avctx->rc_min_rate){
+ av_log(avctx, AV_LOG_INFO, "Warning min_rate > 0 but min_rate != max_rate isnt recommanded!\n");
+ }
+
+ if(avctx->rc_min_rate && avctx->rc_min_rate > avctx->bit_rate){
+ av_log(avctx, AV_LOG_INFO, "bitrate below min bitrate\n");
+ return -1;
+ }
+
+ if(avctx->rc_max_rate && avctx->rc_max_rate < avctx->bit_rate){
+ av_log(avctx, AV_LOG_INFO, "bitrate above max bitrate\n");
+ return -1;
+ }
+
+ if( s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate
+ && (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO)
+ && 90000LL * (avctx->rc_buffer_size-1) > s->avctx->rc_max_rate*0xFFFFLL){
+
+ av_log(avctx, AV_LOG_INFO, "Warning vbv_delay will be set to 0xFFFF (=VBR) as the specified vbv buffer is too large for the given bitrate!\n");
+ }
+
if((s->flags & CODEC_FLAG_4MV) && s->codec_id != CODEC_ID_MPEG4
- && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P){
+ && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P && s->codec_id != CODEC_ID_FLV1){
av_log(avctx, AV_LOG_ERROR, "4MV not supported by codec\n");
return -1;
}
-
+
if(s->obmc && s->avctx->mb_decision != FF_MB_DECISION_SIMPLE){
av_log(avctx, AV_LOG_ERROR, "OBMC is only supported with simple mb decission\n");
return -1;
av_log(avctx, AV_LOG_ERROR, "b frames not supported by codec\n");
return -1;
}
-
+
+ if((s->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME|CODEC_FLAG_ALT_SCAN))
+ && s->codec_id != CODEC_ID_MPEG4 && s->codec_id != CODEC_ID_MPEG2VIDEO){
+ av_log(avctx, AV_LOG_ERROR, "interlacing not supported by codec\n");
+ return -1;
+ }
+
if(s->mpeg_quant && s->codec_id != CODEC_ID_MPEG4){ //FIXME mpeg2 uses that too
av_log(avctx, AV_LOG_ERROR, "mpeg2 style quantization not supporetd by codec\n");
return -1;
return -1;
}
+ if(s->avctx->scenechange_threshold < 1000000000 && (s->flags & CODEC_FLAG_CLOSED_GOP)){
+ av_log(avctx, AV_LOG_ERROR, "closed gop with scene change detection arent supported yet\n");
+ return -1;
+ }
+
+ if(s->avctx->thread_count > 1 && s->codec_id != CODEC_ID_MPEG4
+ && s->codec_id != CODEC_ID_MPEG1VIDEO && s->codec_id != CODEC_ID_MPEG2VIDEO
+ && (s->codec_id != CODEC_ID_H263P || !(s->flags & CODEC_FLAG_H263P_SLICE_STRUCT))){
+ av_log(avctx, AV_LOG_ERROR, "multi threaded encoding not supported by codec\n");
+ return -1;
+ }
+
+ if(s->avctx->thread_count > 1)
+ s->rtp_mode= 1;
+
+ i= ff_gcd(avctx->frame_rate, avctx->frame_rate_base);
+ if(i > 1){
+ av_log(avctx, AV_LOG_INFO, "removing common factors from framerate\n");
+ avctx->frame_rate /= i;
+ avctx->frame_rate_base /= i;
+// return -1;
+ }
+
if(s->codec_id==CODEC_ID_MJPEG){
s->intra_quant_bias= 1<<(QUANT_BIAS_SHIFT-1); //(a + x/2)/x
s->inter_quant_bias= 0;
return -1;
}
- { /* set up some save defaults, some codecs might override them later */
- static int done=0;
- if(!done){
- int i;
- done=1;
-
- default_mv_penalty= av_mallocz( sizeof(uint8_t)*(MAX_FCODE+1)*(2*MAX_MV+1) );
- memset(default_mv_penalty, 0, sizeof(uint8_t)*(MAX_FCODE+1)*(2*MAX_MV+1));
- memset(default_fcode_tab , 0, sizeof(uint8_t)*(2*MAX_MV+1));
-
- for(i=-16; i<16; i++){
- default_fcode_tab[i + MAX_MV]= 1;
- }
- }
- }
- s->me.mv_penalty= default_mv_penalty;
- s->fcode_tab= default_fcode_tab;
-
- /* dont use mv_penalty table for crap MV as it would be confused */
- //FIXME remove after fixing / removing old ME
- if (s->me_method < ME_EPZS) s->me.mv_penalty = default_mv_penalty;
+ avctx->has_b_frames= !s->low_delay;
s->encoding = 1;
if(s->modified_quant)
s->chroma_qscale_table= ff_h263_chroma_qscale_table;
s->progressive_frame=
- s->progressive_sequence= !(avctx->flags & CODEC_FLAG_INTERLACED_DCT);
+ s->progressive_sequence= !(avctx->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME));
+ s->quant_precision=5;
+
+ ff_set_cmp(&s->dsp, s->dsp.ildct_cmp, s->avctx->ildct_cmp);
- ff_init_me(s);
-
#ifdef CONFIG_ENCODERS
#ifdef CONFIG_RISKY
if (s->out_format == FMT_H263)
ff_mpeg1_encode_init(s);
#endif
- /* init default q matrix */
+ /* init q matrix */
for(i=0;i<64;i++) {
int j= s->dsp.idct_permutation[i];
#ifdef CONFIG_RISKY
if(ff_rate_control_init(s) < 0)
return -1;
-
- s->picture_number = 0;
- s->picture_in_gop_number = 0;
- s->fake_picture_number = 0;
- /* motion detector init */
- s->f_code = 1;
- s->b_code = 1;
-
+
return 0;
}
assert(s->last_picture_ptr==NULL || s->out_format != FMT_H264 || s->codec_id == CODEC_ID_SVQ3);
/* mark&release old frames */
- if (s->pict_type != B_TYPE && s->last_picture_ptr && s->last_picture_ptr->data[0]) {
+ if (s->pict_type != B_TYPE && s->last_picture_ptr && s->last_picture_ptr != s->next_picture_ptr && s->last_picture_ptr->data[0]) {
avctx->release_buffer(avctx, (AVFrame*)s->last_picture_ptr);
/* release forgotten pictures */
pic= (AVFrame*)&s->picture[i];
}
- pic->reference= s->pict_type != B_TYPE ? 3 : 0;
+ pic->reference= s->pict_type != B_TYPE && !s->dropable ? 3 : 0;
- if(s->current_picture_ptr) //FIXME broken, we need a coded_picture_number in MpegEncContext
- pic->coded_picture_number= s->current_picture_ptr->coded_picture_number+1;
+ pic->coded_picture_number= s->coded_picture_number++;
if( alloc_picture(s, (Picture*)pic, 0) < 0)
return -1;
if(s->out_format != FMT_H264 || s->codec_id == CODEC_ID_SVQ3){
if (s->pict_type != B_TYPE) {
s->last_picture_ptr= s->next_picture_ptr;
- s->next_picture_ptr= s->current_picture_ptr;
- }
+ if(!s->dropable)
+ s->next_picture_ptr= s->current_picture_ptr;
+ }
+/* av_log(s->avctx, AV_LOG_DEBUG, "L%p N%p C%p L%p N%p C%p type:%d drop:%d\n", s->last_picture_ptr, s->next_picture_ptr,s->current_picture_ptr,
+ s->last_picture_ptr ? s->last_picture_ptr->data[0] : NULL,
+ s->next_picture_ptr ? s->next_picture_ptr->data[0] : NULL,
+ s->current_picture_ptr ? s->current_picture_ptr->data[0] : NULL,
+ s->pict_type, s->dropable);*/
if(s->last_picture_ptr) copy_picture(&s->last_picture, s->last_picture_ptr);
if(s->next_picture_ptr) copy_picture(&s->next_picture, s->next_picture_ptr);
/* set dequantizer, we cant do it during init as it might change for mpeg4
and we cant do it in the header decode as init isnt called for mpeg4 there yet */
- if(s->mpeg_quant || s->codec_id == CODEC_ID_MPEG2VIDEO)
- s->dct_unquantize = s->dct_unquantize_mpeg2;
- else if(s->out_format == FMT_H263)
- s->dct_unquantize = s->dct_unquantize_h263;
- else
- s->dct_unquantize = s->dct_unquantize_mpeg1;
+ if(s->mpeg_quant || s->codec_id == CODEC_ID_MPEG2VIDEO){
+ s->dct_unquantize_intra = s->dct_unquantize_mpeg2_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_mpeg2_inter;
+ }else if(s->out_format == FMT_H263){
+ s->dct_unquantize_intra = s->dct_unquantize_h263_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_h263_inter;
+ }else if(s->out_format == FMT_H261){
+ s->dct_unquantize_intra = s->dct_unquantize_h261_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_h261_inter;
+ }else{
+ s->dct_unquantize_intra = s->dct_unquantize_mpeg1_intra;
+ s->dct_unquantize_inter = s->dct_unquantize_mpeg1_inter;
+ }
if(s->dct_error_sum){
assert(s->avctx->noise_reduction && s->encoding);
* @param color color of the arrow
*/
static void draw_line(uint8_t *buf, int sx, int sy, int ex, int ey, int w, int h, int stride, int color){
- int t, x, y, f;
+ int t, x, y, fr, f;
sx= clip(sx, 0, w-1);
sy= clip(sy, 0, h-1);
ex-= sx;
f= ((ey-sy)<<16)/ex;
for(x= 0; x <= ex; x++){
- y= ((x*f) + (1<<15))>>16;
- buf[y*stride + x]+= color;
+ y = (x*f)>>16;
+ fr= (x*f)&0xFFFF;
+ buf[ y *stride + x]+= (color*(0x10000-fr))>>16;
+ buf[(y+1)*stride + x]+= (color* fr )>>16;
}
}else{
if(sy > ey){
if(ey) f= ((ex-sx)<<16)/ey;
else f= 0;
for(y= 0; y <= ey; y++){
- x= ((y*f) + (1<<15))>>16;
- buf[y*stride + x]+= color;
+ x = (y*f)>>16;
+ fr= (y*f)&0xFFFF;
+ buf[y*stride + x ]+= (color*(0x10000-fr))>>16;;
+ buf[y*stride + x+1]+= (color* fr )>>16;;
}
}
}
/**
* prints debuging info for the given picture.
*/
-void ff_print_debug_info(MpegEncContext *s, Picture *pict){
+void ff_print_debug_info(MpegEncContext *s, AVFrame *pict){
if(!pict || !pict->mb_type) return;
if(s->avctx->debug&(FF_DEBUG_SKIP | FF_DEBUG_QP | FF_DEBUG_MB_TYPE)){
int x,y;
-
+
+ av_log(s->avctx,AV_LOG_DEBUG,"New frame, type: ");
+ switch (pict->pict_type) {
+ case FF_I_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"I\n"); break;
+ case FF_P_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"P\n"); break;
+ case FF_B_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"B\n"); break;
+ case FF_S_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"S\n"); break;
+ case FF_SI_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"SI\n"); break;
+ case FF_SP_TYPE: av_log(s->avctx,AV_LOG_DEBUG,"SP\n"); break;
+ }
for(y=0; y<s->mb_height; y++){
for(x=0; x<s->mb_width; x++){
if(s->avctx->debug&FF_DEBUG_SKIP){
}
if(s->avctx->debug&FF_DEBUG_MB_TYPE){
int mb_type= pict->mb_type[x + y*s->mb_stride];
-
//Type & MV direction
if(IS_PCM(mb_type))
av_log(s->avctx, AV_LOG_DEBUG, "P");
}
}
- if(s->avctx->debug&(FF_DEBUG_VIS_MV|FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)){
+ if((s->avctx->debug&(FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)) || (s->avctx->debug_mv)){
const int shift= 1 + s->quarter_sample;
int mb_y;
- uint8_t *ptr= pict->data[0];
+ uint8_t *ptr;
+ int i;
+ int h_chroma_shift, v_chroma_shift;
s->low_delay=0; //needed to see the vectors without trashing the buffers
+ avcodec_get_chroma_sub_sample(s->avctx->pix_fmt, &h_chroma_shift, &v_chroma_shift);
+ for(i=0; i<3; i++){
+ memcpy(s->visualization_buffer[i], pict->data[i], (i==0) ? pict->linesize[i]*s->height:pict->linesize[i]*s->height >> v_chroma_shift);
+ pict->data[i]= s->visualization_buffer[i];
+ }
+ pict->type= FF_BUFFER_TYPE_COPY;
+ ptr= pict->data[0];
+
for(mb_y=0; mb_y<s->mb_height; mb_y++){
int mb_x;
for(mb_x=0; mb_x<s->mb_width; mb_x++){
const int mb_index= mb_x + mb_y*s->mb_stride;
- if((s->avctx->debug&FF_DEBUG_VIS_MV) && pict->motion_val){
- if(IS_8X8(pict->mb_type[mb_index])){
- int i;
- for(i=0; i<4; i++){
+ if((s->avctx->debug_mv) && pict->motion_val){
+ int type;
+ for(type=0; type<3; type++){
+ int direction = 0;
+ switch (type) {
+ case 0: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_P_FOR)) || (pict->pict_type!=FF_P_TYPE))
+ continue;
+ direction = 0;
+ break;
+ case 1: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_FOR)) || (pict->pict_type!=FF_B_TYPE))
+ continue;
+ direction = 0;
+ break;
+ case 2: if ((!(s->avctx->debug_mv&FF_DEBUG_VIS_MV_B_BACK)) || (pict->pict_type!=FF_B_TYPE))
+ continue;
+ direction = 1;
+ break;
+ }
+ if(!USES_LIST(pict->mb_type[mb_index], direction))
+ continue;
+
+ //FIXME for h264
+ if(IS_8X8(pict->mb_type[mb_index])){
+ int i;
+ for(i=0; i<4; i++){
int sx= mb_x*16 + 4 + 8*(i&1);
int sy= mb_y*16 + 4 + 8*(i>>1);
- int xy= 1 + mb_x*2 + (i&1) + (mb_y*2 + 1 + (i>>1))*(s->mb_width*2 + 2);
- int mx= (pict->motion_val[0][xy][0]>>shift) + sx;
- int my= (pict->motion_val[0][xy][1]>>shift) + sy;
+ int xy= mb_x*2 + (i&1) + (mb_y*2 + (i>>1))*s->b8_stride;
+ int mx= (pict->motion_val[direction][xy][0]>>shift) + sx;
+ int my= (pict->motion_val[direction][xy][1]>>shift) + sy;
draw_arrow(ptr, sx, sy, mx, my, s->width, s->height, s->linesize, 100);
- }
- }else if(IS_16X8(pict->mb_type[mb_index])){
- int i;
- for(i=0; i<2; i++){
+ }
+ }else if(IS_16X8(pict->mb_type[mb_index])){
+ int i;
+ for(i=0; i<2; i++){
int sx=mb_x*16 + 8;
int sy=mb_y*16 + 4 + 8*i;
- int xy=1 + mb_x*2 + (mb_y*2 + 1 + i)*(s->mb_width*2 + 2);
- int mx=(pict->motion_val[0][xy][0]>>shift) + sx;
- int my=(pict->motion_val[0][xy][1]>>shift) + sy;
- draw_arrow(ptr, sx, sy, mx, my, s->width, s->height, s->linesize, 100);
+ int xy= mb_x*2 + (mb_y*2 + i)*s->b8_stride;
+ int mx=(pict->motion_val[direction][xy][0]>>shift);
+ int my=(pict->motion_val[direction][xy][1]>>shift);
+
+ if(IS_INTERLACED(pict->mb_type[mb_index]))
+ my*=2;
+
+ draw_arrow(ptr, sx, sy, mx+sx, my+sy, s->width, s->height, s->linesize, 100);
+ }
+ }else{
+ int sx= mb_x*16 + 8;
+ int sy= mb_y*16 + 8;
+ int xy= mb_x*2 + mb_y*2*s->b8_stride;
+ int mx= (pict->motion_val[direction][xy][0]>>shift) + sx;
+ int my= (pict->motion_val[direction][xy][1]>>shift) + sy;
+ draw_arrow(ptr, sx, sy, mx, my, s->width, s->height, s->linesize, 100);
}
- }else{
- int sx= mb_x*16 + 8;
- int sy= mb_y*16 + 8;
- int xy= 1 + mb_x*2 + (mb_y*2 + 1)*(s->mb_width*2 + 2);
- int mx= (pict->motion_val[0][xy][0]>>shift) + sx;
- int my= (pict->motion_val[0][xy][1]>>shift) + sy;
- draw_arrow(ptr, sx, sy, mx, my, s->width, s->height, s->linesize, 100);
- }
+ }
}
if((s->avctx->debug&FF_DEBUG_VIS_QP) && pict->motion_val){
uint64_t c= (pict->qscale_table[mb_index]*128/31) * 0x0101010101010101ULL;
for(y=0; y<h; y+=16){
for(x=0; x<w; x+=16){
int offset= x + y*stride;
- int sad = s->dsp.pix_abs16x16(src + offset, ref + offset, stride);
+ int sad = s->dsp.sad[0](NULL, src + offset, ref + offset, stride, 16);
int mean= (s->dsp.pix_sum(src + offset, stride) + 128)>>8;
int sae = get_sae(src + offset, mean, stride);
}
}
}
- pic->quality= pic_arg->quality;
- pic->pict_type= pic_arg->pict_type;
- pic->pts = pic_arg->pts;
- pic->interlaced_frame = pic_arg->interlaced_frame;
- pic->top_field_first = pic_arg->top_field_first;
-
- if(s->input_picture[encoding_delay])
- pic->display_picture_number= s->input_picture[encoding_delay]->display_picture_number + 1;
+ copy_picture_attributes(s, pic, pic_arg);
+ pic->display_picture_number= s->input_picture_number++;
+ if(pic->pts != AV_NOPTS_VALUE){
+ s->user_specified_pts= pic->pts;
+ }else{
+ if(s->user_specified_pts){
+ pic->pts= s->user_specified_pts + AV_TIME_BASE*(int64_t)s->avctx->frame_rate_base / s->avctx->frame_rate;
+ av_log(s->avctx, AV_LOG_INFO, "Warning: AVFrame.pts=? trying to guess (%Ld)\n", pic->pts);
+ }else{
+ pic->pts= av_rescale(pic->display_picture_number*(int64_t)s->avctx->frame_rate_base, AV_TIME_BASE, s->avctx->frame_rate);
+ }
+ }
}
-
+
/* shift buffer entries */
for(i=1; i<MAX_PICTURE_COUNT /*s->encoding_delay+1*/; i++)
s->input_picture[i-1]= s->input_picture[i];
static void select_input_picture(MpegEncContext *s){
int i;
- int coded_pic_num=0;
-
- if(s->reordered_input_picture[0])
- coded_pic_num= s->reordered_input_picture[0]->coded_picture_number + 1;
for(i=1; i<MAX_PICTURE_COUNT; i++)
s->reordered_input_picture[i-1]= s->reordered_input_picture[i];
if(/*s->picture_in_gop_number >= s->gop_size ||*/ s->next_picture_ptr==NULL || s->intra_only){
s->reordered_input_picture[0]= s->input_picture[0];
s->reordered_input_picture[0]->pict_type= I_TYPE;
- s->reordered_input_picture[0]->coded_picture_number= coded_pic_num;
+ s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++;
}else{
int b_frames;
av_log(s->avctx, AV_LOG_ERROR, "warning, too many bframes in a row\n");
b_frames = s->max_b_frames;
}
- }else if(s->b_frame_strategy==0){
+ }else if(s->avctx->b_frame_strategy==0){
b_frames= s->max_b_frames;
while(b_frames && !s->input_picture[b_frames]) b_frames--;
- }else if(s->b_frame_strategy==1){
+ }else if(s->avctx->b_frame_strategy==1){
for(i=1; i<s->max_b_frames+1; i++){
if(s->input_picture[i] && s->input_picture[i]->b_frame_score==0){
s->input_picture[i]->b_frame_score=
//static int b_count=0;
//b_count+= b_frames;
//av_log(s->avctx, AV_LOG_DEBUG, "b_frames: %d\n", b_count);
-
+ if(s->picture_in_gop_number + b_frames >= s->gop_size){
+ if(s->flags & CODEC_FLAG_CLOSED_GOP)
+ b_frames=0;
+ s->input_picture[b_frames]->pict_type= I_TYPE;
+ }
+
+ if( (s->flags & CODEC_FLAG_CLOSED_GOP)
+ && b_frames
+ && s->input_picture[b_frames]->pict_type== I_TYPE)
+ b_frames--;
+
s->reordered_input_picture[0]= s->input_picture[b_frames];
- if( s->picture_in_gop_number + b_frames >= s->gop_size
- || s->reordered_input_picture[0]->pict_type== I_TYPE)
- s->reordered_input_picture[0]->pict_type= I_TYPE;
- else
+ if(s->reordered_input_picture[0]->pict_type != I_TYPE)
s->reordered_input_picture[0]->pict_type= P_TYPE;
- s->reordered_input_picture[0]->coded_picture_number= coded_pic_num;
+ s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++;
for(i=0; i<b_frames; i++){
- coded_pic_num++;
s->reordered_input_picture[i+1]= s->input_picture[i];
s->reordered_input_picture[i+1]->pict_type= B_TYPE;
- s->reordered_input_picture[i+1]->coded_picture_number= coded_pic_num;
+ s->reordered_input_picture[i+1]->coded_picture_number= s->coded_picture_number++;
}
}
}
s->reordered_input_picture[0]->data[i]= NULL;
s->reordered_input_picture[0]->type= 0;
- //FIXME bad, copy * except
- pic->pict_type = s->reordered_input_picture[0]->pict_type;
- pic->quality = s->reordered_input_picture[0]->quality;
- pic->coded_picture_number = s->reordered_input_picture[0]->coded_picture_number;
- pic->reference = s->reordered_input_picture[0]->reference;
- pic->pts = s->reordered_input_picture[0]->pts;
+ pic->reference = s->reordered_input_picture[0]->reference;
alloc_picture(s, pic, 0);
+ copy_picture_attributes(s, (AVFrame*)pic, (AVFrame*)s->reordered_input_picture[0]);
+
s->current_picture_ptr= pic;
}else{
// input is not a shared pix -> reuse buffer for current_pix
return -1;
}
- init_put_bits(&s->pb, buf, buf_size);
+ for(i=0; i<avctx->thread_count; i++){
+ int start_y= s->thread_context[i]->start_mb_y;
+ int end_y= s->thread_context[i]-> end_mb_y;
+ int h= s->mb_height;
+ uint8_t *start= buf + buf_size*start_y/h;
+ uint8_t *end = buf + buf_size* end_y/h;
+
+ init_put_bits(&s->thread_context[i]->pb, start, end - start);
+ }
s->picture_in_gop_number++;
/* output? */
if(s->new_picture.data[0]){
-
s->pict_type= s->new_picture.pict_type;
//emms_c();
//printf("qs:%f %f %d\n", s->new_picture.quality, s->current_picture.quality, s->qscale);
for(i=0; i<4; i++){
avctx->error[i] += s->current_picture_ptr->error[i];
}
- }
- s->input_picture_number++;
+ flush_put_bits(&s->pb);
+ s->frame_bits = put_bits_count(&s->pb);
+
+ stuffing_count= ff_vbv_update(s, s->frame_bits);
+ if(stuffing_count){
+ switch(s->codec_id){
+ case CODEC_ID_MPEG1VIDEO:
+ case CODEC_ID_MPEG2VIDEO:
+ while(stuffing_count--){
+ put_bits(&s->pb, 8, 0);
+ }
+ break;
+ case CODEC_ID_MPEG4:
+ put_bits(&s->pb, 16, 0);
+ put_bits(&s->pb, 16, 0x1C3);
+ stuffing_count -= 4;
+ while(stuffing_count--){
+ put_bits(&s->pb, 8, 0xFF);
+ }
+ break;
+ default:
+ av_log(s->avctx, AV_LOG_ERROR, "vbv buffer overflow\n");
+ }
+ flush_put_bits(&s->pb);
+ s->frame_bits = put_bits_count(&s->pb);
+ }
- flush_put_bits(&s->pb);
- s->frame_bits = (pbBufPtr(&s->pb) - s->pb.buf) * 8;
+ /* update mpeg1/2 vbv_delay for CBR */
+ if(s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate && s->out_format == FMT_MPEG1
+ && 90000LL * (avctx->rc_buffer_size-1) <= s->avctx->rc_max_rate*0xFFFFLL){
+ int vbv_delay;
- stuffing_count= ff_vbv_update(s, s->frame_bits);
- if(stuffing_count){
- switch(s->codec_id){
- case CODEC_ID_MPEG1VIDEO:
- case CODEC_ID_MPEG2VIDEO:
- while(stuffing_count--){
- put_bits(&s->pb, 8, 0);
- }
- break;
- case CODEC_ID_MPEG4:
- put_bits(&s->pb, 16, 0);
- put_bits(&s->pb, 16, 0x1C3);
- stuffing_count -= 4;
- while(stuffing_count--){
- put_bits(&s->pb, 8, 0xFF);
- }
- break;
- default:
- av_log(s->avctx, AV_LOG_ERROR, "vbv buffer overflow\n");
+ assert(s->repeat_first_field==0);
+
+ vbv_delay= lrintf(90000 * s->rc_context.buffer_index / s->avctx->rc_max_rate);
+ assert(vbv_delay < 0xFFFF);
+
+ s->vbv_delay_ptr[0] &= 0xF8;
+ s->vbv_delay_ptr[0] |= vbv_delay>>13;
+ s->vbv_delay_ptr[1] = vbv_delay>>5;
+ s->vbv_delay_ptr[2] &= 0x07;
+ s->vbv_delay_ptr[2] |= vbv_delay<<3;
}
- flush_put_bits(&s->pb);
- s->frame_bits = (pbBufPtr(&s->pb) - s->pb.buf) * 8;
+ s->total_bits += s->frame_bits;
+ avctx->frame_bits = s->frame_bits;
+ }else{
+ assert((pbBufPtr(&s->pb) == s->pb.buf));
+ s->frame_bits=0;
}
-
- s->total_bits += s->frame_bits;
- avctx->frame_bits = s->frame_bits;
+ assert((s->frame_bits&7)==0);
return s->frame_bits/8;
}
static inline void gmc1_motion(MpegEncContext *s,
uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int dest_offset,
- uint8_t **ref_picture, int src_offset)
+ uint8_t **ref_picture)
{
uint8_t *ptr;
int offset, src_x, src_y, linesize, uvlinesize;
linesize = s->linesize;
uvlinesize = s->uvlinesize;
- ptr = ref_picture[0] + (src_y * linesize) + src_x + src_offset;
+ ptr = ref_picture[0] + (src_y * linesize) + src_x;
- dest_y+=dest_offset;
if(s->flags&CODEC_FLAG_EMU_EDGE){
if( (unsigned)src_x >= s->h_edge_pos - 17
|| (unsigned)src_y >= s->v_edge_pos - 17){
if (src_y == s->height>>1)
motion_y =0;
- offset = (src_y * uvlinesize) + src_x + (src_offset>>1);
+ offset = (src_y * uvlinesize) + src_x;
ptr = ref_picture[1] + offset;
if(s->flags&CODEC_FLAG_EMU_EDGE){
if( (unsigned)src_x >= (s->h_edge_pos>>1) - 9
emu=1;
}
}
- s->dsp.gmc1(dest_cb + (dest_offset>>1), ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);
+ s->dsp.gmc1(dest_cb, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);
ptr = ref_picture[2] + offset;
if(emu){
ff_emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1);
ptr= s->edge_emu_buffer;
}
- s->dsp.gmc1(dest_cr + (dest_offset>>1), ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);
+ s->dsp.gmc1(dest_cr, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding);
return;
}
static inline void gmc_motion(MpegEncContext *s,
uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int dest_offset,
- uint8_t **ref_picture, int src_offset)
+ uint8_t **ref_picture)
{
uint8_t *ptr;
int linesize, uvlinesize;
linesize = s->linesize;
uvlinesize = s->uvlinesize;
- ptr = ref_picture[0] + src_offset;
+ ptr = ref_picture[0];
- dest_y+=dest_offset;
-
ox= s->sprite_offset[0][0] + s->sprite_delta[0][0]*s->mb_x*16 + s->sprite_delta[0][1]*s->mb_y*16;
oy= s->sprite_offset[0][1] + s->sprite_delta[1][0]*s->mb_x*16 + s->sprite_delta[1][1]*s->mb_y*16;
if(s->flags&CODEC_FLAG_GRAY) return;
-
- dest_cb+=dest_offset>>1;
- dest_cr+=dest_offset>>1;
-
ox= s->sprite_offset[1][0] + s->sprite_delta[0][0]*s->mb_x*8 + s->sprite_delta[0][1]*s->mb_y*8;
oy= s->sprite_offset[1][1] + s->sprite_delta[1][0]*s->mb_x*8 + s->sprite_delta[1][1]*s->mb_y*8;
- ptr = ref_picture[1] + (src_offset>>1);
+ ptr = ref_picture[1];
s->dsp.gmc(dest_cb, ptr, uvlinesize, 8,
ox,
oy,
a+1, (1<<(2*a+1)) - s->no_rounding,
s->h_edge_pos>>1, s->v_edge_pos>>1);
- ptr = ref_picture[2] + (src_offset>>1);
+ ptr = ref_picture[2];
s->dsp.gmc(dest_cr, ptr, uvlinesize, 8,
ox,
oy,
}
static inline int hpel_motion(MpegEncContext *s,
- uint8_t *dest, uint8_t *src,
+ uint8_t *dest, uint8_t *src,
+ int field_based, int field_select,
int src_x, int src_y,
int width, int height, int stride,
int h_edge_pos, int v_edge_pos,
if(s->unrestricted_mv && (s->flags&CODEC_FLAG_EMU_EDGE)){
if( (unsigned)src_x > h_edge_pos - (motion_x&1) - w
|| (unsigned)src_y > v_edge_pos - (motion_y&1) - h){
- ff_emulated_edge_mc(s->edge_emu_buffer, src, stride, w+1, h+1,
- src_x, src_y, h_edge_pos, v_edge_pos);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<<field_based,
+ src_x, src_y<<field_based, h_edge_pos, s->v_edge_pos);
src= s->edge_emu_buffer;
emu=1;
}
}
+ if(field_select)
+ src += s->linesize;
pix_op[dxy](dest, src, stride, h);
return emu;
}
/* apply one mpeg motion vector to the three components */
-static inline void mpeg_motion(MpegEncContext *s,
+static always_inline void mpeg_motion(MpegEncContext *s,
uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int dest_offset,
- uint8_t **ref_picture, int src_offset,
- int field_based, op_pixels_func (*pix_op)[4],
+ int field_based, int bottom_field, int field_select,
+ uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
int motion_x, int motion_y, int h)
{
- uint8_t *ptr;
- int dxy, offset, mx, my, src_x, src_y, height, v_edge_pos, uvlinesize;
- int emu=0;
+ uint8_t *ptr_y, *ptr_cb, *ptr_cr;
+ int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, uvlinesize, linesize;
+
#if 0
if(s->quarter_sample)
{
}
#endif
- height = s->height >> field_based;
v_edge_pos = s->v_edge_pos >> field_based;
+ linesize = s->current_picture.linesize[0] << field_based;
uvlinesize = s->current_picture.linesize[1] << field_based;
- emu= hpel_motion(s,
- dest_y + dest_offset, ref_picture[0] + src_offset,
- s->mb_x * 16, s->mb_y * (16 >> field_based),
- s->width, height, s->current_picture.linesize[0] << field_based,
- s->h_edge_pos, v_edge_pos,
- 16, h, pix_op[0],
- motion_x, motion_y);
-
-
- if(s->flags&CODEC_FLAG_GRAY) return;
+ dxy = ((motion_y & 1) << 1) | (motion_x & 1);
+ src_x = s->mb_x* 16 + (motion_x >> 1);
+ src_y =(s->mb_y<<(4-field_based)) + (motion_y >> 1);
if (s->out_format == FMT_H263) {
- dxy = 0;
- if ((motion_x & 3) != 0)
- dxy |= 1;
- if ((motion_y & 3) != 0)
- dxy |= 2;
- mx = motion_x >> 2;
- my = motion_y >> 2;
+ if((s->workaround_bugs & FF_BUG_HPEL_CHROMA) && field_based){
+ mx = (motion_x>>1)|(motion_x&1);
+ my = motion_y >>1;
+ uvdxy = ((my & 1) << 1) | (mx & 1);
+ uvsrc_x = s->mb_x* 8 + (mx >> 1);
+ uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1);
+ }else{
+ uvdxy = dxy | (motion_y & 2) | ((motion_x & 2) >> 1);
+ uvsrc_x = src_x>>1;
+ uvsrc_y = src_y>>1;
+ }
+ }else if(s->out_format == FMT_H261){//even chroma mv's are full pel in H261
+ mx = motion_x / 4;
+ my = motion_y / 4;
+ uvdxy = 0;
+ uvsrc_x = s->mb_x*8 + mx;
+ uvsrc_y = s->mb_y*8 + my;
} else {
- mx = motion_x / 2;
- my = motion_y / 2;
- dxy = ((my & 1) << 1) | (mx & 1);
- mx >>= 1;
- my >>= 1;
+ if(s->chroma_y_shift){
+ mx = motion_x / 2;
+ my = motion_y / 2;
+ uvdxy = ((my & 1) << 1) | (mx & 1);
+ uvsrc_x = s->mb_x* 8 + (mx >> 1);
+ uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1);
+ } else {
+ if(s->chroma_x_shift){
+ //Chroma422
+ mx = motion_x / 2;
+ uvdxy = ((motion_y & 1) << 1) | (mx & 1);
+ uvsrc_x = s->mb_x* 8 + (mx >> 1);
+ uvsrc_y = src_y;
+ } else {
+ //Chroma444
+ uvdxy = dxy;
+ uvsrc_x = src_x;
+ uvsrc_y = src_y;
+ }
+ }
}
-
- src_x = s->mb_x * 8 + mx;
- src_y = s->mb_y * (8 >> field_based) + my;
- src_x = clip(src_x, -8, s->width >> 1);
- if (src_x == (s->width >> 1))
- dxy &= ~1;
- src_y = clip(src_y, -8, height >> 1);
- if (src_y == (height >> 1))
- dxy &= ~2;
- offset = (src_y * uvlinesize) + src_x + (src_offset >> 1);
- ptr = ref_picture[1] + offset;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr - (src_offset >> 1), s->uvlinesize, 9, 9+field_based,
- src_x, src_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
- ptr= s->edge_emu_buffer + (src_offset >> 1);
+
+ ptr_y = ref_picture[0] + src_y * linesize + src_x;
+ ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x;
+ ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x;
+
+ if( (unsigned)src_x > s->h_edge_pos - (motion_x&1) - 16
+ || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){
+ if(s->codec_id == CODEC_ID_MPEG2VIDEO ||
+ s->codec_id == CODEC_ID_MPEG1VIDEO){
+ av_log(s->avctx,AV_LOG_DEBUG,"MPEG motion vector out of boundary\n");
+ return ;
+ }
+ ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based,
+ src_x, src_y<<field_based, s->h_edge_pos, s->v_edge_pos);
+ ptr_y = s->edge_emu_buffer;
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize;
+ ff_emulated_edge_mc(uvbuf , ptr_cb, s->uvlinesize, 9, 9+field_based,
+ uvsrc_x, uvsrc_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
+ ff_emulated_edge_mc(uvbuf+16, ptr_cr, s->uvlinesize, 9, 9+field_based,
+ uvsrc_x, uvsrc_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
+ ptr_cb= uvbuf;
+ ptr_cr= uvbuf+16;
+ }
}
- pix_op[1][dxy](dest_cb + (dest_offset >> 1), ptr, uvlinesize, h >> 1);
- ptr = ref_picture[2] + offset;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr - (src_offset >> 1), s->uvlinesize, 9, 9+field_based,
- src_x, src_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
- ptr= s->edge_emu_buffer + (src_offset >> 1);
+ if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data
+ dest_y += s->linesize;
+ dest_cb+= s->uvlinesize;
+ dest_cr+= s->uvlinesize;
+ }
+
+ if(field_select){
+ ptr_y += s->linesize;
+ ptr_cb+= s->uvlinesize;
+ ptr_cr+= s->uvlinesize;
+ }
+
+ pix_op[0][dxy](dest_y, ptr_y, linesize, h);
+
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ pix_op[s->chroma_x_shift][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift);
+ pix_op[s->chroma_x_shift][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift);
}
- pix_op[1][dxy](dest_cr + (dest_offset >> 1), ptr, uvlinesize, h >> 1);
}
//FIXME move to dsputil, avg variant, 16x16 version
static inline void put_obmc(uint8_t *dst, uint8_t *src[5], int stride){
if(i && mv[i][0]==mv[MID][0] && mv[i][1]==mv[MID][1]){
ptr[i]= ptr[MID];
}else{
- ptr[i]= s->edge_emu_buffer + 16 + 8*(i&1) + s->linesize*8*(i>>1);
- hpel_motion(s, ptr[i], src,
+ ptr[i]= s->obmc_scratchpad + 8*(i&1) + s->linesize*8*(i>>1);
+ hpel_motion(s, ptr[i], src, 0, 0,
src_x, src_y,
s->width, s->height, s->linesize,
s->h_edge_pos, s->v_edge_pos,
static inline void qpel_motion(MpegEncContext *s,
uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int dest_offset,
- uint8_t **ref_picture, int src_offset,
- int field_based, op_pixels_func (*pix_op)[4],
+ int field_based, int bottom_field, int field_select,
+ uint8_t **ref_picture, op_pixels_func (*pix_op)[4],
qpel_mc_func (*qpix_op)[16],
int motion_x, int motion_y, int h)
{
- uint8_t *ptr;
- int dxy, offset, mx, my, src_x, src_y, height, v_edge_pos, linesize, uvlinesize;
- int emu=0;
+ uint8_t *ptr_y, *ptr_cb, *ptr_cr;
+ int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, linesize, uvlinesize;
dxy = ((motion_y & 3) << 2) | (motion_x & 3);
- src_x = s->mb_x * 16 + (motion_x >> 2);
+ src_x = s->mb_x * 16 + (motion_x >> 2);
src_y = s->mb_y * (16 >> field_based) + (motion_y >> 2);
- height = s->height >> field_based;
v_edge_pos = s->v_edge_pos >> field_based;
- src_x = clip(src_x, -16, s->width);
- if (src_x == s->width)
- dxy &= ~3;
- src_y = clip(src_y, -16, height);
- if (src_y == height)
- dxy &= ~12;
linesize = s->linesize << field_based;
uvlinesize = s->uvlinesize << field_based;
- ptr = ref_picture[0] + (src_y * linesize) + src_x + src_offset;
- dest_y += dest_offset;
-//printf("%d %d %d\n", src_x, src_y, dxy);
- if(s->flags&CODEC_FLAG_EMU_EDGE){
- if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16
- || (unsigned)src_y > v_edge_pos - (motion_y&3) - h ){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr - src_offset, s->linesize, 17, 17+field_based,
- src_x, src_y<<field_based, s->h_edge_pos, s->v_edge_pos);
- ptr= s->edge_emu_buffer + src_offset;
- emu=1;
- }
- }
- if(!field_based)
- qpix_op[0][dxy](dest_y, ptr, linesize);
- else{
- //damn interlaced mode
- //FIXME boundary mirroring is not exactly correct here
- qpix_op[1][dxy](dest_y , ptr , linesize);
- qpix_op[1][dxy](dest_y+8, ptr+8, linesize);
- }
-
- if(s->flags&CODEC_FLAG_GRAY) return;
-
if(field_based){
mx= motion_x/2;
my= motion_y>>1;
mx= (mx>>1)|(mx&1);
my= (my>>1)|(my&1);
- dxy= (mx&1) | ((my&1)<<1);
+ uvdxy= (mx&1) | ((my&1)<<1);
mx>>=1;
my>>=1;
- src_x = s->mb_x * 8 + mx;
- src_y = s->mb_y * (8 >> field_based) + my;
- src_x = clip(src_x, -8, s->width >> 1);
- if (src_x == (s->width >> 1))
- dxy &= ~1;
- src_y = clip(src_y, -8, height >> 1);
- if (src_y == (height >> 1))
- dxy &= ~2;
+ uvsrc_x = s->mb_x * 8 + mx;
+ uvsrc_y = s->mb_y * (8 >> field_based) + my;
+
+ ptr_y = ref_picture[0] + src_y * linesize + src_x;
+ ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x;
+ ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x;
+
+ if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16
+ || (unsigned)src_y > v_edge_pos - (motion_y&3) - h ){
+ ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based,
+ src_x, src_y<<field_based, s->h_edge_pos, s->v_edge_pos);
+ ptr_y= s->edge_emu_buffer;
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize;
+ ff_emulated_edge_mc(uvbuf, ptr_cb, s->uvlinesize, 9, 9 + field_based,
+ uvsrc_x, uvsrc_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
+ ff_emulated_edge_mc(uvbuf + 16, ptr_cr, s->uvlinesize, 9, 9 + field_based,
+ uvsrc_x, uvsrc_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
+ ptr_cb= uvbuf;
+ ptr_cr= uvbuf + 16;
+ }
+ }
- offset = (src_y * uvlinesize) + src_x + (src_offset >> 1);
- ptr = ref_picture[1] + offset;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr - (src_offset >> 1), s->uvlinesize, 9, 9 + field_based,
- src_x, src_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
- ptr= s->edge_emu_buffer + (src_offset >> 1);
+ if(!field_based)
+ qpix_op[0][dxy](dest_y, ptr_y, linesize);
+ else{
+ if(bottom_field){
+ dest_y += s->linesize;
+ dest_cb+= s->uvlinesize;
+ dest_cr+= s->uvlinesize;
+ }
+
+ if(field_select){
+ ptr_y += s->linesize;
+ ptr_cb += s->uvlinesize;
+ ptr_cr += s->uvlinesize;
+ }
+ //damn interlaced mode
+ //FIXME boundary mirroring is not exactly correct here
+ qpix_op[1][dxy](dest_y , ptr_y , linesize);
+ qpix_op[1][dxy](dest_y+8, ptr_y+8, linesize);
}
- pix_op[1][dxy](dest_cb + (dest_offset >> 1), ptr, uvlinesize, h >> 1);
-
- ptr = ref_picture[2] + offset;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr - (src_offset >> 1), s->uvlinesize, 9, 9 + field_based,
- src_x, src_y<<field_based, s->h_edge_pos>>1, s->v_edge_pos>>1);
- ptr= s->edge_emu_buffer + (src_offset >> 1);
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ pix_op[1][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> 1);
+ pix_op[1][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> 1);
}
- pix_op[1][dxy](dest_cr + (dest_offset >> 1), ptr, uvlinesize, h >> 1);
}
inline int ff_h263_round_chroma(int x){
if(s->obmc && s->pict_type != B_TYPE){
int16_t mv_cache[4][4][2];
const int xy= s->mb_x + s->mb_y*s->mb_stride;
- const int mot_stride= s->mb_width*2 + 2;
- const int mot_xy= 1 + mb_x*2 + (mb_y*2 + 1)*mot_stride;
+ const int mot_stride= s->b8_stride;
+ const int mot_xy= mb_x*2 + mb_y*2*mot_stride;
assert(!s->mb_skiped);
#ifdef CONFIG_RISKY
if(s->mcsel){
if(s->real_sprite_warping_points==1){
- gmc1_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, 0);
+ gmc1_motion(s, dest_y, dest_cb, dest_cr,
+ ref_picture);
}else{
- gmc_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, 0);
+ gmc_motion(s, dest_y, dest_cb, dest_cr,
+ ref_picture);
}
}else if(s->quarter_sample){
- qpel_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, 0,
- 0, pix_op, qpix_op,
+ qpel_motion(s, dest_y, dest_cb, dest_cr,
+ 0, 0, 0,
+ ref_picture, pix_op, qpix_op,
s->mv[dir][0][0], s->mv[dir][0][1], 16);
}else if(s->mspel){
ff_mspel_motion(s, dest_y, dest_cb, dest_cr,
}else
#endif
{
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, 0,
- 0, pix_op,
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 0, 0, 0,
+ ref_picture, pix_op,
s->mv[dir][0][0], s->mv[dir][0][1], 16);
}
break;
}else{
for(i=0;i<4;i++) {
hpel_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize,
- ref_picture[0],
+ ref_picture[0], 0, 0,
mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8,
s->width, s->height, s->linesize,
s->h_edge_pos, s->v_edge_pos,
case MV_TYPE_FIELD:
if (s->picture_structure == PICT_FRAME) {
if(s->quarter_sample){
- /* top field */
- qpel_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, s->field_select[dir][0] ? s->linesize : 0,
- 1, pix_op, qpix_op,
- s->mv[dir][0][0], s->mv[dir][0][1], 8);
- /* bottom field */
- qpel_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
- ref_picture, s->field_select[dir][1] ? s->linesize : 0,
- 1, pix_op, qpix_op,
- s->mv[dir][1][0], s->mv[dir][1][1], 8);
+ for(i=0; i<2; i++){
+ qpel_motion(s, dest_y, dest_cb, dest_cr,
+ 1, i, s->field_select[dir][i],
+ ref_picture, pix_op, qpix_op,
+ s->mv[dir][i][0], s->mv[dir][i][1], 8);
+ }
}else{
/* top field */
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, s->field_select[dir][0] ? s->linesize : 0,
- 1, pix_op,
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 1, 0, s->field_select[dir][0],
+ ref_picture, pix_op,
s->mv[dir][0][0], s->mv[dir][0][1], 8);
/* bottom field */
- mpeg_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
- ref_picture, s->field_select[dir][1] ? s->linesize : 0,
- 1, pix_op,
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 1, 1, s->field_select[dir][1],
+ ref_picture, pix_op,
s->mv[dir][1][0], s->mv[dir][1][1], 8);
}
} else {
- int offset;
- if(s->picture_structure == s->field_select[dir][0] + 1 || s->pict_type == B_TYPE || s->first_field){
- offset= s->field_select[dir][0] ? s->linesize : 0;
- }else{
- ref_picture= s->current_picture.data;
- offset= s->field_select[dir][0] ? s->linesize : -s->linesize;
+ if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != B_TYPE && !s->first_field){
+ ref_picture= s->current_picture_ptr->data;
}
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, offset,
- 0, pix_op,
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 0, 0, s->field_select[dir][0],
+ ref_picture, pix_op,
s->mv[dir][0][0], s->mv[dir][0][1], 16);
}
break;
- case MV_TYPE_16X8:{
- int offset;
- uint8_t ** ref2picture;
+ case MV_TYPE_16X8:
+ for(i=0; i<2; i++){
+ uint8_t ** ref2picture;
- if(s->picture_structure == s->field_select[dir][0] + 1 || s->pict_type == B_TYPE || s->first_field){
+ if(s->picture_structure == s->field_select[dir][i] + 1 || s->pict_type == B_TYPE || s->first_field){
ref2picture= ref_picture;
- offset= s->field_select[dir][0] ? s->linesize : 0;
}else{
- ref2picture= s->current_picture.data;
- offset= s->field_select[dir][0] ? s->linesize : -s->linesize;
+ ref2picture= s->current_picture_ptr->data;
}
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref2picture, offset,
- 0, pix_op,
- s->mv[dir][0][0], s->mv[dir][0][1], 8);
-
-
- if(s->picture_structure == s->field_select[dir][1] + 1 || s->pict_type == B_TYPE || s->first_field){
- ref2picture= ref_picture;
- offset= s->field_select[dir][1] ? s->linesize : 0;
- }else{
- ref2picture= s->current_picture.data;
- offset= s->field_select[dir][1] ? s->linesize : -s->linesize;
- }
- // I know it is ugly but this is the only way to fool emu_edge without rewrite mpeg_motion
- mpeg_motion(s, dest_y+16*s->linesize, dest_cb+8*s->uvlinesize, dest_cr+8*s->uvlinesize,
- 0,
- ref2picture, offset,
- 0, pix_op,
- s->mv[dir][1][0], s->mv[dir][1][1]+16, 8);
- }
-
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 0, 0, s->field_select[dir][i],
+ ref2picture, pix_op,
+ s->mv[dir][i][0], s->mv[dir][i][1] + 16*i, 8);
+
+ dest_y += 16*s->linesize;
+ dest_cb+= (16>>s->chroma_y_shift)*s->uvlinesize;
+ dest_cr+= (16>>s->chroma_y_shift)*s->uvlinesize;
+ }
break;
case MV_TYPE_DMV:
- {
- op_pixels_func (*dmv_pix_op)[4];
- int offset;
-
- dmv_pix_op = s->dsp.put_pixels_tab;
-
if(s->picture_structure == PICT_FRAME){
- //put top field from top field
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, 0,
- 1, dmv_pix_op,
- s->mv[dir][0][0], s->mv[dir][0][1], 8);
- //put bottom field from bottom field
- mpeg_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
- ref_picture, s->linesize,
- 1, dmv_pix_op,
- s->mv[dir][0][0], s->mv[dir][0][1], 8);
-
- dmv_pix_op = s->dsp.avg_pixels_tab;
-
- //avg top field from bottom field
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture, s->linesize,
- 1, dmv_pix_op,
- s->mv[dir][2][0], s->mv[dir][2][1], 8);
- //avg bottom field from top field
- mpeg_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
- ref_picture, 0,
- 1, dmv_pix_op,
- s->mv[dir][3][0], s->mv[dir][3][1], 8);
-
+ for(i=0; i<2; i++){
+ int j;
+ for(j=0; j<2; j++){
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 1, j, j^i,
+ ref_picture, pix_op,
+ s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], 8);
+ }
+ pix_op = s->dsp.avg_pixels_tab;
+ }
}else{
- offset=(s->picture_structure == PICT_BOTTOM_FIELD)?
- s->linesize : 0;
-
- //put field from the same parity
- //same parity is never in the same frame
- mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
- ref_picture,offset,
- 0,dmv_pix_op,
- s->mv[dir][0][0],s->mv[dir][0][1],16);
-
- // after put we make avg of the same block
- dmv_pix_op=s->dsp.avg_pixels_tab;
-
- //opposite parity is always in the same frame if this is second field
- if(!s->first_field){
- ref_picture = s->current_picture.data;
- //top field is one linesize from frame beginig
- offset=(s->picture_structure == PICT_BOTTOM_FIELD)?
- -s->linesize : s->linesize;
- }else
- offset=(s->picture_structure == PICT_BOTTOM_FIELD)?
- 0 : s->linesize;
-
- //avg field from the opposite parity
- mpeg_motion(s, dest_y, dest_cb, dest_cr,0,
- ref_picture, offset,
- 0,dmv_pix_op,
- s->mv[dir][2][0],s->mv[dir][2][1],16);
+ for(i=0; i<2; i++){
+ mpeg_motion(s, dest_y, dest_cb, dest_cr,
+ 0, 0, s->picture_structure != i+1,
+ ref_picture, pix_op,
+ s->mv[dir][2*i][0],s->mv[dir][2*i][1],16);
+
+ // after put we make avg of the same block
+ pix_op=s->dsp.avg_pixels_tab;
+
+ //opposite parity is always in the same frame if this is second field
+ if(!s->first_field){
+ ref_picture = s->current_picture_ptr->data;
+ }
+ }
}
- }
break;
default: assert(0);
}
static inline void put_dct(MpegEncContext *s,
DCTELEM *block, int i, uint8_t *dest, int line_size, int qscale)
{
- s->dct_unquantize(s, block, i, qscale);
+ s->dct_unquantize_intra(s, block, i, qscale);
s->dsp.idct_put (dest, line_size, block);
}
DCTELEM *block, int i, uint8_t *dest, int line_size, int qscale)
{
if (s->block_last_index[i] >= 0) {
- s->dct_unquantize(s, block, i, qscale);
+ s->dct_unquantize_inter(s, block, i, qscale);
s->dsp.idct_add (dest, line_size, block);
}
*/
void ff_clean_intra_table_entries(MpegEncContext *s)
{
- int wrap = s->block_wrap[0];
+ int wrap = s->b8_stride;
int xy = s->block_index[0];
s->dc_val[0][xy ] =
s->coded_block[xy + 1 + wrap] = 0;
}
/* chroma */
- wrap = s->block_wrap[4];
- xy = s->mb_x + 1 + (s->mb_y + 1) * wrap;
+ wrap = s->mb_stride;
+ xy = s->mb_x + s->mb_y * wrap;
s->dc_val[1][xy] =
s->dc_val[2][xy] = 1024;
/* ac pred */
memset(s->ac_val[1][xy], 0, 16 * sizeof(int16_t));
memset(s->ac_val[2][xy], 0, 16 * sizeof(int16_t));
- s->mbintra_table[s->mb_x + s->mb_y*s->mb_stride]= 0;
+ s->mbintra_table[xy]= 0;
}
/* generic function called after a macroblock has been parsed by the
s->mv : motion vector
s->interlaced_dct : true if interlaced dct used (mpeg2)
*/
-void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64])
+void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64])
{
int mb_x, mb_y;
const int mb_xy = s->mb_y * s->mb_stride + s->mb_x;
mb_x = s->mb_x;
mb_y = s->mb_y;
+ if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
+ /* save DCT coefficients */
+ int i,j;
+ DCTELEM *dct = &s->current_picture.dct_coeff[mb_xy*64*6];
+ for(i=0; i<6; i++)
+ for(j=0; j<64; j++)
+ *dct++ = block[i][s->dsp.idct_permutation[j]];
+ }
+
s->current_picture.qscale_table[mb_xy]= s->qscale;
/* update DC predictors for P macroblocks */
}
}
- if (s->interlaced_dct) {
- dct_linesize = linesize * 2;
- dct_offset = linesize;
- } else {
- dct_linesize = linesize;
- dct_offset = linesize * 8;
- }
+ dct_linesize = linesize << s->interlaced_dct;
+ dct_offset =(s->interlaced_dct)? linesize : linesize*8;
+
if(readable){
dest_y= s->dest[0];
dest_cb= s->dest[1];
dest_cr= s->dest[2];
}else{
- dest_y = s->edge_emu_buffer+32; //FIXME cleanup scratchpad pointers
- dest_cb= s->edge_emu_buffer+48;
- dest_cr= s->edge_emu_buffer+56;
+ dest_y = s->b_scratchpad;
+ dest_cb= s->b_scratchpad+16*linesize;
+ dest_cr= s->b_scratchpad+32*linesize;
}
if (!s->mb_intra) {
/* motion handling */
add_dct(s, block[3], 3, dest_y + dct_offset + 8, dct_linesize);
if(!(s->flags&CODEC_FLAG_GRAY)){
- add_dct(s, block[4], 4, dest_cb, uvlinesize);
- add_dct(s, block[5], 5, dest_cr, uvlinesize);
- }
- }
+ if(s->chroma_y_shift){//Chroma420
+ add_dct(s, block[4], 4, dest_cb, uvlinesize);
+ add_dct(s, block[5], 5, dest_cr, uvlinesize);
+ }else{
+ //chroma422
+ dct_linesize = uvlinesize << s->interlaced_dct;
+ dct_offset =(s->interlaced_dct)? uvlinesize : uvlinesize*8;
+
+ add_dct(s, block[4], 4, dest_cb, dct_linesize);
+ add_dct(s, block[5], 5, dest_cr, dct_linesize);
+ add_dct(s, block[6], 6, dest_cb+dct_offset, dct_linesize);
+ add_dct(s, block[7], 7, dest_cr+dct_offset, dct_linesize);
+ if(!s->chroma_x_shift){//Chroma444
+ add_dct(s, block[8], 8, dest_cb+8, dct_linesize);
+ add_dct(s, block[9], 9, dest_cr+8, dct_linesize);
+ add_dct(s, block[10], 10, dest_cb+8+dct_offset, dct_linesize);
+ add_dct(s, block[11], 11, dest_cr+8+dct_offset, dct_linesize);
+ }
+ }
+ }//fi gray
+ }
#ifdef CONFIG_RISKY
else{
ff_wmv2_add_mb(s, block, dest_y, dest_cb, dest_cr);
s->dsp.idct_put(dest_y + dct_offset + 8, dct_linesize, block[3]);
if(!(s->flags&CODEC_FLAG_GRAY)){
- s->dsp.idct_put(dest_cb, uvlinesize, block[4]);
- s->dsp.idct_put(dest_cr, uvlinesize, block[5]);
- }
+ if(s->chroma_y_shift){
+ s->dsp.idct_put(dest_cb, uvlinesize, block[4]);
+ s->dsp.idct_put(dest_cr, uvlinesize, block[5]);
+ }else{
+
+ dct_linesize = uvlinesize << s->interlaced_dct;
+ dct_offset =(s->interlaced_dct)? uvlinesize : uvlinesize*8;
+
+ s->dsp.idct_put(dest_cb, dct_linesize, block[4]);
+ s->dsp.idct_put(dest_cr, dct_linesize, block[5]);
+ s->dsp.idct_put(dest_cb + dct_offset, dct_linesize, block[6]);
+ s->dsp.idct_put(dest_cr + dct_offset, dct_linesize, block[7]);
+ if(!s->chroma_x_shift){//Chroma444
+ s->dsp.idct_put(dest_cb + 8, dct_linesize, block[8]);
+ s->dsp.idct_put(dest_cr + 8, dct_linesize, block[9]);
+ s->dsp.idct_put(dest_cb + 8 + dct_offset, dct_linesize, block[10]);
+ s->dsp.idct_put(dest_cr + 8 + dct_offset, dct_linesize, block[11]);
+ }
+ }
+ }//gray
}
}
if(!readable){
s->dsp.put_pixels_tab[0][0](s->dest[0], dest_y , linesize,16);
- s->dsp.put_pixels_tab[1][0](s->dest[1], dest_cb, uvlinesize, 8);
- s->dsp.put_pixels_tab[1][0](s->dest[2], dest_cr, uvlinesize, 8);
+ s->dsp.put_pixels_tab[s->chroma_x_shift][0](s->dest[1], dest_cb, uvlinesize,16 >> s->chroma_y_shift);
+ s->dsp.put_pixels_tab[s->chroma_x_shift][0](s->dest[2], dest_cr, uvlinesize,16 >> s->chroma_y_shift);
}
}
}
int i;
const int maxlevel= s->max_qcoeff;
const int minlevel= s->min_qcoeff;
+ int overflow=0;
if(s->mb_intra){
i=1; //skip clipping of intra dc
const int j= s->intra_scantable.permutated[i];
int level = block[j];
- if (level>maxlevel) level=maxlevel;
- else if(level<minlevel) level=minlevel;
-
- block[j]= level;
- }
-}
-
-#if 0
-static int pix_vcmp16x8(uint8_t *s, int stride){ //FIXME move to dsputil & optimize
- int score=0;
- int x,y;
-
- for(y=0; y<7; y++){
- for(x=0; x<16; x+=4){
- score+= ABS(s[x ] - s[x +stride]) + ABS(s[x+1] - s[x+1+stride])
- +ABS(s[x+2] - s[x+2+stride]) + ABS(s[x+3] - s[x+3+stride]);
- }
- s+= stride;
- }
-
- return score;
-}
-
-static int pix_diff_vcmp16x8(uint8_t *s1, uint8_t*s2, int stride){ //FIXME move to dsputil & optimize
- int score=0;
- int x,y;
-
- for(y=0; y<7; y++){
- for(x=0; x<16; x++){
- score+= ABS(s1[x ] - s2[x ] - s1[x +stride] + s2[x +stride]);
- }
- s1+= stride;
- s2+= stride;
- }
-
- return score;
-}
-#else
-#define SQ(a) ((a)*(a))
-
-static int pix_vcmp16x8(uint8_t *s, int stride){ //FIXME move to dsputil & optimize
- int score=0;
- int x,y;
-
- for(y=0; y<7; y++){
- for(x=0; x<16; x+=4){
- score+= SQ(s[x ] - s[x +stride]) + SQ(s[x+1] - s[x+1+stride])
- +SQ(s[x+2] - s[x+2+stride]) + SQ(s[x+3] - s[x+3+stride]);
- }
- s+= stride;
- }
-
- return score;
-}
-
-static int pix_diff_vcmp16x8(uint8_t *s1, uint8_t*s2, int stride){ //FIXME move to dsputil & optimize
- int score=0;
- int x,y;
-
- for(y=0; y<7; y++){
- for(x=0; x<16; x++){
- score+= SQ(s1[x ] - s2[x ] - s1[x +stride] + s2[x +stride]);
+ if (level>maxlevel){
+ level=maxlevel;
+ overflow++;
+ }else if(level<minlevel){
+ level=minlevel;
+ overflow++;
}
- s1+= stride;
- s2+= stride;
+
+ block[j]= level;
}
- return score;
+ if(overflow && s->avctx->mb_decision == FF_MB_DECISION_SIMPLE)
+ av_log(s->avctx, AV_LOG_INFO, "warning, cliping %d dct coefficents to %d..%d\n", overflow, minlevel, maxlevel);
}
-#endif
-
#endif //CONFIG_ENCODERS
/**
}else{
offset[0]= y * s->linesize;;
offset[1]=
- offset[2]= (y>>1) * s->uvlinesize;;
+ offset[2]= (y >> s->chroma_y_shift) * s->uvlinesize;
offset[3]= 0;
}
const int linesize= s->current_picture.linesize[0]; //not s->linesize as this woulnd be wrong for field pics
const int uvlinesize= s->current_picture.linesize[1];
- s->block_index[0]= s->block_wrap[0]*(s->mb_y*2 + 1) - 1 + s->mb_x*2;
- s->block_index[1]= s->block_wrap[0]*(s->mb_y*2 + 1) + s->mb_x*2;
- s->block_index[2]= s->block_wrap[0]*(s->mb_y*2 + 2) - 1 + s->mb_x*2;
- s->block_index[3]= s->block_wrap[0]*(s->mb_y*2 + 2) + s->mb_x*2;
- s->block_index[4]= s->block_wrap[4]*(s->mb_y + 1) + s->block_wrap[0]*(s->mb_height*2 + 2) + s->mb_x;
- s->block_index[5]= s->block_wrap[4]*(s->mb_y + 1 + s->mb_height + 2) + s->block_wrap[0]*(s->mb_height*2 + 2) + s->mb_x;
-
- if(s->pict_type==B_TYPE && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME){
- s->dest[0] = s->current_picture.data[0] + s->mb_x * 16 - 16;
- s->dest[1] = s->current_picture.data[1] + s->mb_x * 8 - 8;
- s->dest[2] = s->current_picture.data[2] + s->mb_x * 8 - 8;
- }else{
- s->dest[0] = s->current_picture.data[0] + (s->mb_y * 16* linesize ) + s->mb_x * 16 - 16;
- s->dest[1] = s->current_picture.data[1] + (s->mb_y * 8 * uvlinesize) + s->mb_x * 8 - 8;
- s->dest[2] = s->current_picture.data[2] + (s->mb_y * 8 * uvlinesize) + s->mb_x * 8 - 8;
- }
+ s->block_index[0]= s->b8_stride*(s->mb_y*2 ) - 2 + s->mb_x*2;
+ s->block_index[1]= s->b8_stride*(s->mb_y*2 ) - 1 + s->mb_x*2;
+ s->block_index[2]= s->b8_stride*(s->mb_y*2 + 1) - 2 + s->mb_x*2;
+ s->block_index[3]= s->b8_stride*(s->mb_y*2 + 1) - 1 + s->mb_x*2;
+ s->block_index[4]= s->mb_stride*(s->mb_y + 1) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
+ s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
+ //block_index is not used by mpeg2, so it is not affected by chroma_format
+
+ s->dest[0] = s->current_picture.data[0] + (s->mb_x - 1)*16;
+ s->dest[1] = s->current_picture.data[1] + (s->mb_x - 1)*(16 >> s->chroma_x_shift);
+ s->dest[2] = s->current_picture.data[2] + (s->mb_x - 1)*(16 >> s->chroma_x_shift);
+
+ if(!(s->pict_type==B_TYPE && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME))
+ {
+ s->dest[0] += s->mb_y * linesize * 16;
+ s->dest[1] += s->mb_y * uvlinesize * (16 >> s->chroma_y_shift);
+ s->dest[2] += s->mb_y * uvlinesize * (16 >> s->chroma_y_shift);
+ }
}
#ifdef CONFIG_ENCODERS
+static void get_vissual_weight(int16_t *weight, uint8_t *ptr, int stride){
+ int x, y;
+//FIXME optimize
+ for(y=0; y<8; y++){
+ for(x=0; x<8; x++){
+ int x2, y2;
+ int sum=0;
+ int sqr=0;
+ int count=0;
+
+ for(y2= FFMAX(y-1, 0); y2 < FFMIN(8, y+2); y2++){
+ for(x2= FFMAX(x-1, 0); x2 < FFMIN(8, x+2); x2++){
+ int v= ptr[x2 + y2*stride];
+ sum += v;
+ sqr += v*v;
+ count++;
+ }
+ }
+ weight[x + 8*y]= (36*ff_sqrt(count*sqr - sum*sum)) / count;
+ }
+ }
+}
+
static void encode_mb(MpegEncContext *s, int motion_x, int motion_y)
{
+ int16_t weight[6][64];
+ DCTELEM orig[6][64];
const int mb_x= s->mb_x;
const int mb_y= s->mb_y;
int i;
int skip_dct[6];
int dct_offset = s->linesize*8; //default for progressive frames
+ uint8_t *ptr_y, *ptr_cb, *ptr_cr;
+ int wrap_y, wrap_c;
for(i=0; i<6; i++) skip_dct[i]=0;
if(!(s->flags&CODEC_FLAG_QP_RD)){
s->dquant= s->qscale - last_qp;
- if(s->out_format==FMT_H263)
+ if(s->out_format==FMT_H263){
s->dquant= clip(s->dquant, -2, 2); //FIXME RD
- if(s->codec_id==CODEC_ID_MPEG4){
- if(!s->mb_intra){
- if((s->mv_dir&MV_DIRECT) || s->mv_type==MV_TYPE_8X8)
- s->dquant=0;
+ if(s->codec_id==CODEC_ID_MPEG4){
+ if(!s->mb_intra){
+ if(s->pict_type == B_TYPE){
+ if(s->dquant&1)
+ s->dquant= (s->dquant/2)*2;
+ if(s->mv_dir&MV_DIRECT)
+ s->dquant= 0;
+ }
+ if(s->mv_type==MV_TYPE_8X8)
+ s->dquant=0;
+ }
}
}
}
ff_set_qscale(s, last_qp + s->dquant);
- }
+ }else if(s->flags&CODEC_FLAG_QP_RD)
+ ff_set_qscale(s, s->qscale + s->dquant);
- if (s->mb_intra) {
- uint8_t *ptr;
- int wrap_y;
- int emu=0;
+ wrap_y = s->linesize;
+ wrap_c = s->uvlinesize;
+ ptr_y = s->new_picture.data[0] + (mb_y * 16 * wrap_y) + mb_x * 16;
+ ptr_cb = s->new_picture.data[1] + (mb_y * 8 * wrap_c) + mb_x * 8;
+ ptr_cr = s->new_picture.data[2] + (mb_y * 8 * wrap_c) + mb_x * 8;
- wrap_y = s->linesize;
- ptr = s->new_picture.data[0] + (mb_y * 16 * wrap_y) + mb_x * 16;
+ if(mb_x*16+16 > s->width || mb_y*16+16 > s->height){
+ ff_emulated_edge_mc(s->edge_emu_buffer , ptr_y , wrap_y,16,16,mb_x*16,mb_y*16, s->width , s->height);
+ ptr_y= s->edge_emu_buffer;
+ ff_emulated_edge_mc(s->edge_emu_buffer+18*wrap_y , ptr_cb, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
+ ptr_cb= s->edge_emu_buffer+18*wrap_y;
+ ff_emulated_edge_mc(s->edge_emu_buffer+18*wrap_y+9, ptr_cr, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
+ ptr_cr= s->edge_emu_buffer+18*wrap_y+9;
+ }
- if(mb_x*16+16 > s->width || mb_y*16+16 > s->height){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr, wrap_y, 16, 16, mb_x*16, mb_y*16, s->width, s->height);
- ptr= s->edge_emu_buffer;
- emu=1;
- }
-
+ if (s->mb_intra) {
if(s->flags&CODEC_FLAG_INTERLACED_DCT){
int progressive_score, interlaced_score;
+
+ s->interlaced_dct=0;
+ progressive_score= s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y, 8)
+ +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y*8, NULL, wrap_y, 8) - 400;
+
+ if(progressive_score > 0){
+ interlaced_score = s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y*2, 8)
+ +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y , NULL, wrap_y*2, 8);
+ if(progressive_score > interlaced_score){
+ s->interlaced_dct=1;
- progressive_score= pix_vcmp16x8(ptr, wrap_y ) + pix_vcmp16x8(ptr + wrap_y*8, wrap_y );
- interlaced_score = pix_vcmp16x8(ptr, wrap_y*2) + pix_vcmp16x8(ptr + wrap_y , wrap_y*2);
-
- if(progressive_score > interlaced_score + 100){
- s->interlaced_dct=1;
-
- dct_offset= wrap_y;
- wrap_y<<=1;
- }else
- s->interlaced_dct=0;
+ dct_offset= wrap_y;
+ wrap_y<<=1;
+ }
+ }
}
- s->dsp.get_pixels(s->block[0], ptr , wrap_y);
- s->dsp.get_pixels(s->block[1], ptr + 8, wrap_y);
- s->dsp.get_pixels(s->block[2], ptr + dct_offset , wrap_y);
- s->dsp.get_pixels(s->block[3], ptr + dct_offset + 8, wrap_y);
+ s->dsp.get_pixels(s->block[0], ptr_y , wrap_y);
+ s->dsp.get_pixels(s->block[1], ptr_y + 8, wrap_y);
+ s->dsp.get_pixels(s->block[2], ptr_y + dct_offset , wrap_y);
+ s->dsp.get_pixels(s->block[3], ptr_y + dct_offset + 8, wrap_y);
if(s->flags&CODEC_FLAG_GRAY){
skip_dct[4]= 1;
skip_dct[5]= 1;
}else{
- int wrap_c = s->uvlinesize;
- ptr = s->new_picture.data[1] + (mb_y * 8 * wrap_c) + mb_x * 8;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
- ptr= s->edge_emu_buffer;
- }
- s->dsp.get_pixels(s->block[4], ptr, wrap_c);
-
- ptr = s->new_picture.data[2] + (mb_y * 8 * wrap_c) + mb_x * 8;
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
- ptr= s->edge_emu_buffer;
- }
- s->dsp.get_pixels(s->block[5], ptr, wrap_c);
+ s->dsp.get_pixels(s->block[4], ptr_cb, wrap_c);
+ s->dsp.get_pixels(s->block[5], ptr_cr, wrap_c);
}
}else{
op_pixels_func (*op_pix)[4];
qpel_mc_func (*op_qpix)[16];
uint8_t *dest_y, *dest_cb, *dest_cr;
- uint8_t *ptr_y, *ptr_cb, *ptr_cr;
- int wrap_y, wrap_c;
- int emu=0;
dest_y = s->dest[0];
dest_cb = s->dest[1];
dest_cr = s->dest[2];
- wrap_y = s->linesize;
- wrap_c = s->uvlinesize;
- ptr_y = s->new_picture.data[0] + (mb_y * 16 * wrap_y) + mb_x * 16;
- ptr_cb = s->new_picture.data[1] + (mb_y * 8 * wrap_c) + mb_x * 8;
- ptr_cr = s->new_picture.data[2] + (mb_y * 8 * wrap_c) + mb_x * 8;
if ((!s->no_rounding) || s->pict_type==B_TYPE){
op_pix = s->dsp.put_pixels_tab;
MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix, op_qpix);
}
- if(mb_x*16+16 > s->width || mb_y*16+16 > s->height){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, wrap_y, 16, 16, mb_x*16, mb_y*16, s->width, s->height);
- ptr_y= s->edge_emu_buffer;
- emu=1;
- }
-
if(s->flags&CODEC_FLAG_INTERLACED_DCT){
int progressive_score, interlaced_score;
+
+ s->interlaced_dct=0;
+ progressive_score= s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y, 8)
+ +s->dsp.ildct_cmp[0](s, dest_y + wrap_y*8, ptr_y + wrap_y*8, wrap_y, 8) - 400;
- progressive_score= pix_diff_vcmp16x8(ptr_y , dest_y , wrap_y )
- + pix_diff_vcmp16x8(ptr_y + wrap_y*8, dest_y + wrap_y*8, wrap_y );
- interlaced_score = pix_diff_vcmp16x8(ptr_y , dest_y , wrap_y*2)
- + pix_diff_vcmp16x8(ptr_y + wrap_y , dest_y + wrap_y , wrap_y*2);
+ if(s->avctx->ildct_cmp == FF_CMP_VSSE) progressive_score -= 400;
+
+ if(progressive_score>0){
+ interlaced_score = s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y*2, 8)
+ +s->dsp.ildct_cmp[0](s, dest_y + wrap_y , ptr_y + wrap_y , wrap_y*2, 8);
- if(progressive_score > interlaced_score + 600){
- s->interlaced_dct=1;
+ if(progressive_score > interlaced_score){
+ s->interlaced_dct=1;
- dct_offset= wrap_y;
- wrap_y<<=1;
- }else
- s->interlaced_dct=0;
+ dct_offset= wrap_y;
+ wrap_y<<=1;
+ }
+ }
}
s->dsp.diff_pixels(s->block[0], ptr_y , dest_y , wrap_y);
skip_dct[4]= 1;
skip_dct[5]= 1;
}else{
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr_cb, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
- ptr_cb= s->edge_emu_buffer;
- }
s->dsp.diff_pixels(s->block[4], ptr_cb, dest_cb, wrap_c);
- if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, ptr_cr, wrap_c, 8, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
- ptr_cr= s->edge_emu_buffer;
- }
s->dsp.diff_pixels(s->block[5], ptr_cr, dest_cr, wrap_c);
}
/* pre quantization */
if(s->current_picture.mc_mb_var[s->mb_stride*mb_y+ mb_x]<2*s->qscale*s->qscale){
//FIXME optimize
- if(s->dsp.pix_abs8x8(ptr_y , dest_y , wrap_y) < 20*s->qscale) skip_dct[0]= 1;
- if(s->dsp.pix_abs8x8(ptr_y + 8, dest_y + 8, wrap_y) < 20*s->qscale) skip_dct[1]= 1;
- if(s->dsp.pix_abs8x8(ptr_y +dct_offset , dest_y +dct_offset , wrap_y) < 20*s->qscale) skip_dct[2]= 1;
- if(s->dsp.pix_abs8x8(ptr_y +dct_offset+ 8, dest_y +dct_offset+ 8, wrap_y) < 20*s->qscale) skip_dct[3]= 1;
- if(s->dsp.pix_abs8x8(ptr_cb , dest_cb , wrap_c) < 20*s->qscale) skip_dct[4]= 1;
- if(s->dsp.pix_abs8x8(ptr_cr , dest_cr , wrap_c) < 20*s->qscale) skip_dct[5]= 1;
-#if 0
-{
- static int stat[7];
- int num=0;
- for(i=0; i<6; i++)
- if(skip_dct[i]) num++;
- stat[num]++;
-
- if(s->mb_x==0 && s->mb_y==0){
- for(i=0; i<7; i++){
- printf("%6d %1d\n", stat[i], i);
- }
- }
-}
-#endif
+ if(s->dsp.sad[1](NULL, ptr_y , dest_y , wrap_y, 8) < 20*s->qscale) skip_dct[0]= 1;
+ if(s->dsp.sad[1](NULL, ptr_y + 8, dest_y + 8, wrap_y, 8) < 20*s->qscale) skip_dct[1]= 1;
+ if(s->dsp.sad[1](NULL, ptr_y +dct_offset , dest_y +dct_offset , wrap_y, 8) < 20*s->qscale) skip_dct[2]= 1;
+ if(s->dsp.sad[1](NULL, ptr_y +dct_offset+ 8, dest_y +dct_offset+ 8, wrap_y, 8) < 20*s->qscale) skip_dct[3]= 1;
+ if(s->dsp.sad[1](NULL, ptr_cb , dest_cb , wrap_c, 8) < 20*s->qscale) skip_dct[4]= 1;
+ if(s->dsp.sad[1](NULL, ptr_cr , dest_cr , wrap_c, 8) < 20*s->qscale) skip_dct[5]= 1;
}
+ }
+ if(s->avctx->quantizer_noise_shaping){
+ if(!skip_dct[0]) get_vissual_weight(weight[0], ptr_y , wrap_y);
+ if(!skip_dct[1]) get_vissual_weight(weight[1], ptr_y + 8, wrap_y);
+ if(!skip_dct[2]) get_vissual_weight(weight[2], ptr_y + dct_offset , wrap_y);
+ if(!skip_dct[3]) get_vissual_weight(weight[3], ptr_y + dct_offset + 8, wrap_y);
+ if(!skip_dct[4]) get_vissual_weight(weight[4], ptr_cb , wrap_c);
+ if(!skip_dct[5]) get_vissual_weight(weight[5], ptr_cr , wrap_c);
+ memcpy(orig[0], s->block[0], sizeof(DCTELEM)*64*6);
}
/* DCT & quantize */
- if(s->out_format==FMT_MJPEG){
- for(i=0;i<6;i++) {
- int overflow;
- s->block_last_index[i] = s->dct_quantize(s, s->block[i], i, 8, &overflow);
- if (overflow) clip_coeffs(s, s->block[i], s->block_last_index[i]);
- }
- }else{
+ assert(s->out_format!=FMT_MJPEG || s->qscale==8);
+ {
for(i=0;i<6;i++) {
if(!skip_dct[i]){
int overflow;
}else
s->block_last_index[i]= -1;
}
+ if(s->avctx->quantizer_noise_shaping){
+ for(i=0;i<6;i++) {
+ if(!skip_dct[i]){
+ s->block_last_index[i] = dct_quantize_refine(s, s->block[i], weight[i], orig[i], i, s->qscale);
+ }
+ }
+ }
if(s->luma_elim_threshold && !s->mb_intra)
for(i=0; i<4; i++)
s->block[5][0]= (1024 + s->c_dc_scale/2)/ s->c_dc_scale;
}
+ //non c quantize code returns incorrect block_last_index FIXME
+ if(s->alternate_scan && s->dct_quantize != dct_quantize_c){
+ for(i=0; i<6; i++){
+ int j;
+ if(s->block_last_index[i]>0){
+ for(j=63; j>0; j--){
+ if(s->block[i][ s->intra_scantable.permutated[j] ]) break;
+ }
+ s->block_last_index[i]= j;
+ }
+ }
+ }
+
/* huffman encode */
switch(s->codec_id){ //FIXME funct ptr could be slightly faster
case CODEC_ID_MPEG1VIDEO:
#endif //CONFIG_ENCODERS
-/**
- * combines the (truncated) bitstream to a complete frame
- * @returns -1 if no complete frame could be created
- */
-int ff_combine_frame( MpegEncContext *s, int next, uint8_t **buf, int *buf_size){
- ParseContext *pc= &s->parse_context;
-
-#if 0
- if(pc->overread){
- printf("overread %d, state:%X next:%d index:%d o_index:%d\n", pc->overread, pc->state, next, pc->index, pc->overread_index);
- printf("%X %X %X %X\n", (*buf)[0], (*buf)[1],(*buf)[2],(*buf)[3]);
- }
-#endif
-
- /* copy overreaded byes from last frame into buffer */
- for(; pc->overread>0; pc->overread--){
- pc->buffer[pc->index++]= pc->buffer[pc->overread_index++];
- }
-
- pc->last_index= pc->index;
-
- /* copy into buffer end return */
- if(next == END_NOT_FOUND){
- pc->buffer= av_fast_realloc(pc->buffer, &pc->buffer_size, (*buf_size) + pc->index + FF_INPUT_BUFFER_PADDING_SIZE);
-
- memcpy(&pc->buffer[pc->index], *buf, *buf_size);
- pc->index += *buf_size;
- return -1;
- }
-
- *buf_size=
- pc->overread_index= pc->index + next;
-
- /* append to buffer */
- if(pc->index){
- pc->buffer= av_fast_realloc(pc->buffer, &pc->buffer_size, next + pc->index + FF_INPUT_BUFFER_PADDING_SIZE);
-
- memcpy(&pc->buffer[pc->index], *buf, next + FF_INPUT_BUFFER_PADDING_SIZE );
- pc->index = 0;
- *buf= pc->buffer;
- }
-
- /* store overread bytes */
- for(;next < 0; next++){
- pc->state = (pc->state<<8) | pc->buffer[pc->last_index + next];
- pc->overread++;
- }
-
-#if 0
- if(pc->overread){
- printf("overread %d, state:%X next:%d index:%d o_index:%d\n", pc->overread, pc->state, next, pc->index, pc->overread_index);
- printf("%X %X %X %X\n", (*buf)[0], (*buf)[1],(*buf)[2],(*buf)[3]);
- }
-#endif
-
- return 0;
-}
-
void ff_mpeg_flush(AVCodecContext *avctx){
int i;
MpegEncContext *s = avctx->priv_data;
+ if(s==NULL || s->picture==NULL)
+ return;
+
for(i=0; i<MAX_PICTURE_COUNT; i++){
if(s->picture[i].data[0] && ( s->picture[i].type == FF_BUFFER_TYPE_INTERNAL
|| s->picture[i].type == FF_BUFFER_TYPE_USER))
s->parse_context.overread_index= 0;
s->parse_context.index= 0;
s->parse_context.last_index= 0;
+ s->bitstream_buffer_size=0;
}
#ifdef CONFIG_ENCODERS
void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length)
{
- int bytes= length>>4;
+ const uint16_t *srcw= (uint16_t*)src;
+ int words= length>>4;
int bits= length&15;
int i;
if(length==0) return;
-
- for(i=0; i<bytes; i++) put_bits(pb, 16, be2me_16(((uint16_t*)src)[i]));
- put_bits(pb, bits, be2me_16(((uint16_t*)src)[i])>>(16-bits));
-}
-
+
+ if(words < 16){
+ for(i=0; i<words; i++) put_bits(pb, 16, be2me_16(srcw[i]));
+ }else if(put_bits_count(pb)&7){
+ for(i=0; i<words; i++) put_bits(pb, 16, be2me_16(srcw[i]));
+ }else{
+ for(i=0; put_bits_count(pb)&31; i++)
+ put_bits(pb, 8, src[i]);
+ flush_put_bits(pb);
+ memcpy(pbBufPtr(pb), src+i, 2*words-i);
+ skip_put_bytes(pb, 2*words-i);
+ }
+
+ put_bits(pb, bits, be2me_16(srcw[words])>>(16-bits));
+}
+
static inline void copy_context_before_encode(MpegEncContext *d, MpegEncContext *s, int type){
int i;
if(*next_block){
memcpy(dest_backup, s->dest, sizeof(s->dest));
- s->dest[0] = s->me.scratchpad;
- s->dest[1] = s->me.scratchpad + 16;
- s->dest[2] = s->me.scratchpad + 16 + 8;
- assert(2*s->uvlinesize == s->linesize); //should be no prob for encoding
- assert(s->linesize >= 64); //FIXME
+ s->dest[0] = s->rd_scratchpad;
+ s->dest[1] = s->rd_scratchpad + 16*s->linesize;
+ s->dest[2] = s->rd_scratchpad + 16*s->linesize + 8;
+ assert(s->linesize >= 32); //FIXME
}
encode_mb(s, motion_x, motion_y);
- score= get_bit_count(&s->pb);
+ score= put_bits_count(&s->pb);
if(s->data_partitioning){
- score+= get_bit_count(&s->pb2);
- score+= get_bit_count(&s->tex_pb);
+ score+= put_bits_count(&s->pb2);
+ score+= put_bits_count(&s->tex_pb);
}
if(s->avctx->mb_decision == FF_MB_DECISION_RD){
int x,y;
if(w==16 && h==16)
- return s->dsp.sse[0](NULL, src1, src2, stride);
+ return s->dsp.sse[0](NULL, src1, src2, stride, 16);
else if(w==8 && h==8)
- return s->dsp.sse[1](NULL, src1, src2, stride);
+ return s->dsp.sse[1](NULL, src1, src2, stride, 8);
for(y=0; y<h; y++){
for(x=0; x<w; x++){
if(s->mb_y*16 + 16 > s->height) h= s->height- s->mb_y*16;
if(w==16 && h==16)
- return s->dsp.sse[0](NULL, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize)
- +s->dsp.sse[1](NULL, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize)
- +s->dsp.sse[1](NULL, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize);
+ if(s->avctx->mb_cmp == FF_CMP_NSSE){
+ return s->dsp.nsse[0](s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16)
+ +s->dsp.nsse[1](s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8)
+ +s->dsp.nsse[1](s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8);
+ }else{
+ return s->dsp.sse[0](NULL, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16)
+ +s->dsp.sse[1](NULL, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8)
+ +s->dsp.sse[1](NULL, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8);
+ }
else
return sse(s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], w, h, s->linesize)
+sse(s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], w>>1, h>>1, s->uvlinesize)
+sse(s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], w>>1, h>>1, s->uvlinesize);
}
-static void encode_picture(MpegEncContext *s, int picture_number)
-{
- int mb_x, mb_y, pdif = 0;
- int i;
- int bits;
- MpegEncContext best_s, backup_s;
- uint8_t bit_buf[2][3000];
- uint8_t bit_buf2[2][3000];
- uint8_t bit_buf_tex[2][3000];
- PutBitContext pb[2], pb2[2], tex_pb[2];
-
- for(i=0; i<2; i++){
- init_put_bits(&pb [i], bit_buf [i], 3000);
- init_put_bits(&pb2 [i], bit_buf2 [i], 3000);
- init_put_bits(&tex_pb[i], bit_buf_tex[i], 3000);
- }
-
- s->picture_number = picture_number;
-
- /* Reset the average MB variance */
- s->current_picture.mb_var_sum = 0;
- s->current_picture.mc_mb_var_sum = 0;
+static int pre_estimate_motion_thread(AVCodecContext *c, void *arg){
+ MpegEncContext *s= arg;
-#ifdef CONFIG_RISKY
- /* we need to initialize some time vars before we can encode b-frames */
- // RAL: Condition added for MPEG1VIDEO
- if (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO || (s->h263_pred && !s->h263_msmpeg4))
- ff_set_mpeg4_time(s, s->picture_number);
-#endif
-
- s->scene_change_score=0;
- s->lambda= s->current_picture_ptr->quality; //FIXME qscale / ... stuff for ME ratedistoration
-
- if(s->pict_type==I_TYPE){
- if(s->msmpeg4_version >= 3) s->no_rounding=1;
- else s->no_rounding=0;
- }else if(s->pict_type!=B_TYPE){
- if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4)
- s->no_rounding ^= 1;
- }
-
- /* Estimate motion for every MB */
- s->mb_intra=0; //for the rate distoration & bit compare functions
- if(s->pict_type != I_TYPE){
- if(s->pict_type != B_TYPE){
- if((s->avctx->pre_me && s->last_non_b_pict_type==I_TYPE) || s->avctx->pre_me==2){
- s->me.pre_pass=1;
- s->me.dia_size= s->avctx->pre_dia_size;
-
- for(mb_y=s->mb_height-1; mb_y >=0 ; mb_y--) {
- for(mb_x=s->mb_width-1; mb_x >=0 ; mb_x--) {
- s->mb_x = mb_x;
- s->mb_y = mb_y;
- ff_pre_estimate_p_frame_motion(s, mb_x, mb_y);
- }
- }
- s->me.pre_pass=0;
- }
- }
-
- s->me.dia_size= s->avctx->dia_size;
- for(mb_y=0; mb_y < s->mb_height; mb_y++) {
- s->block_index[0]= s->block_wrap[0]*(mb_y*2 + 1) - 1;
- s->block_index[1]= s->block_wrap[0]*(mb_y*2 + 1);
- s->block_index[2]= s->block_wrap[0]*(mb_y*2 + 2) - 1;
- s->block_index[3]= s->block_wrap[0]*(mb_y*2 + 2);
- for(mb_x=0; mb_x < s->mb_width; mb_x++) {
- s->mb_x = mb_x;
- s->mb_y = mb_y;
- s->block_index[0]+=2;
- s->block_index[1]+=2;
- s->block_index[2]+=2;
- s->block_index[3]+=2;
-
- /* compute motion vector & mb_type and store in context */
- if(s->pict_type==B_TYPE)
- ff_estimate_b_frame_motion(s, mb_x, mb_y);
- else
- ff_estimate_p_frame_motion(s, mb_x, mb_y);
- }
- }
- }else /* if(s->pict_type == I_TYPE) */{
- /* I-Frame */
- //FIXME do we need to zero them?
- memset(s->current_picture.motion_val[0][0], 0, sizeof(int16_t)*(s->mb_width*2 + 2)*(s->mb_height*2 + 2)*2);
- memset(s->p_mv_table , 0, sizeof(int16_t)*(s->mb_stride)*s->mb_height*2);
- memset(s->mb_type , MB_TYPE_INTRA, sizeof(uint8_t)*s->mb_stride*s->mb_height);
-
- if(!s->fixed_qscale){
- /* finding spatial complexity for I-frame rate control */
- for(mb_y=0; mb_y < s->mb_height; mb_y++) {
- for(mb_x=0; mb_x < s->mb_width; mb_x++) {
- int xx = mb_x * 16;
- int yy = mb_y * 16;
- uint8_t *pix = s->new_picture.data[0] + (yy * s->linesize) + xx;
- int varc;
- int sum = s->dsp.pix_sum(pix, s->linesize);
-
- varc = (s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500 + 128)>>8;
-
- s->current_picture.mb_var [s->mb_stride * mb_y + mb_x] = varc;
- s->current_picture.mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
- s->current_picture.mb_var_sum += varc;
- }
- }
+ s->me.pre_pass=1;
+ s->me.dia_size= s->avctx->pre_dia_size;
+ s->first_slice_line=1;
+ for(s->mb_y= s->end_mb_y-1; s->mb_y >= s->start_mb_y; s->mb_y--) {
+ for(s->mb_x=s->mb_width-1; s->mb_x >=0 ;s->mb_x--) {
+ ff_pre_estimate_p_frame_motion(s, s->mb_x, s->mb_y);
}
+ s->first_slice_line=0;
}
- emms_c();
+
+ s->me.pre_pass=0;
+
+ return 0;
+}
- if(s->scene_change_score > s->avctx->scenechange_threshold && s->pict_type == P_TYPE){
- s->pict_type= I_TYPE;
- memset(s->mb_type , MB_TYPE_INTRA, sizeof(uint8_t)*s->mb_stride*s->mb_height);
-//printf("Scene change detected, encoding as I Frame %d %d\n", s->current_picture.mb_var_sum, s->current_picture.mc_mb_var_sum);
- }
+static int estimate_motion_thread(AVCodecContext *c, void *arg){
+ MpegEncContext *s= arg;
- if(!s->umvplus){
- if(s->pict_type==P_TYPE || s->pict_type==S_TYPE) {
- s->f_code= ff_get_best_fcode(s, s->p_mv_table, MB_TYPE_INTER);
-
- ff_fix_long_p_mvs(s);
+ s->me.dia_size= s->avctx->dia_size;
+ s->first_slice_line=1;
+ for(s->mb_y= s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {
+ s->mb_x=0; //for block init below
+ ff_init_block_index(s);
+ for(s->mb_x=0; s->mb_x < s->mb_width; s->mb_x++) {
+ s->block_index[0]+=2;
+ s->block_index[1]+=2;
+ s->block_index[2]+=2;
+ s->block_index[3]+=2;
+
+ /* compute motion vector & mb_type and store in context */
+ if(s->pict_type==B_TYPE)
+ ff_estimate_b_frame_motion(s, s->mb_x, s->mb_y);
+ else
+ ff_estimate_p_frame_motion(s, s->mb_x, s->mb_y);
}
+ s->first_slice_line=0;
+ }
+ return 0;
+}
- if(s->pict_type==B_TYPE){
- int a, b;
-
- a = ff_get_best_fcode(s, s->b_forw_mv_table, MB_TYPE_FORWARD);
- b = ff_get_best_fcode(s, s->b_bidir_forw_mv_table, MB_TYPE_BIDIR);
- s->f_code = FFMAX(a, b);
+static int mb_var_thread(AVCodecContext *c, void *arg){
+ MpegEncContext *s= arg;
+ int mb_x, mb_y;
- a = ff_get_best_fcode(s, s->b_back_mv_table, MB_TYPE_BACKWARD);
- b = ff_get_best_fcode(s, s->b_bidir_back_mv_table, MB_TYPE_BIDIR);
- s->b_code = FFMAX(a, b);
+ for(mb_y=s->start_mb_y; mb_y < s->end_mb_y; mb_y++) {
+ for(mb_x=0; mb_x < s->mb_width; mb_x++) {
+ int xx = mb_x * 16;
+ int yy = mb_y * 16;
+ uint8_t *pix = s->new_picture.data[0] + (yy * s->linesize) + xx;
+ int varc;
+ int sum = s->dsp.pix_sum(pix, s->linesize);
+
+ varc = (s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500 + 128)>>8;
- ff_fix_long_b_mvs(s, s->b_forw_mv_table, s->f_code, MB_TYPE_FORWARD);
- ff_fix_long_b_mvs(s, s->b_back_mv_table, s->b_code, MB_TYPE_BACKWARD);
- ff_fix_long_b_mvs(s, s->b_bidir_forw_mv_table, s->f_code, MB_TYPE_BIDIR);
- ff_fix_long_b_mvs(s, s->b_bidir_back_mv_table, s->b_code, MB_TYPE_BIDIR);
+ s->current_picture.mb_var [s->mb_stride * mb_y + mb_x] = varc;
+ s->current_picture.mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
+ s->me.mb_var_sum_temp += varc;
}
}
-
- if (!s->fixed_qscale)
- s->current_picture.quality = ff_rate_estimate_qscale(s);
+ return 0;
+}
- if(s->adaptive_quant){
-#ifdef CONFIG_RISKY
- switch(s->codec_id){
- case CODEC_ID_MPEG4:
- ff_clean_mpeg4_qscales(s);
- break;
- case CODEC_ID_H263:
- case CODEC_ID_H263P:
- case CODEC_ID_FLV1:
- ff_clean_h263_qscales(s);
- break;
+static void write_slice_end(MpegEncContext *s){
+ if(s->codec_id==CODEC_ID_MPEG4){
+ if(s->partitioned_frame){
+ ff_mpeg4_merge_partitions(s);
}
-#endif
-
- s->lambda= s->lambda_table[0];
- //FIXME broken
- }else
- s->lambda= s->current_picture.quality;
-//printf("%d %d\n", s->avctx->global_quality, s->current_picture.quality);
- update_qscale(s);
- if(s->qscale < 3 && s->max_qcoeff<=128 && s->pict_type==I_TYPE && !(s->flags & CODEC_FLAG_QSCALE))
- s->qscale= 3; //reduce cliping problems
-
- if (s->out_format == FMT_MJPEG) {
- /* for mjpeg, we do include qscale in the matrix */
- s->intra_matrix[0] = ff_mpeg1_default_intra_matrix[0];
- for(i=1;i<64;i++){
- int j= s->dsp.idct_permutation[i];
-
- s->intra_matrix[j] = CLAMP_TO_8BIT((ff_mpeg1_default_intra_matrix[i] * s->qscale) >> 3);
- }
- convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16,
- s->intra_matrix, s->intra_quant_bias, 8, 8);
+ ff_mpeg4_stuffing(&s->pb);
+ }else if(s->out_format == FMT_MJPEG){
+ ff_mjpeg_stuffing(&s->pb);
}
-
- //FIXME var duplication
- s->current_picture.key_frame= s->pict_type == I_TYPE;
- s->current_picture.pict_type= s->pict_type;
- if(s->current_picture.key_frame)
- s->picture_in_gop_number=0;
+ align_put_bits(&s->pb);
+ flush_put_bits(&s->pb);
+}
- s->last_bits= get_bit_count(&s->pb);
- switch(s->out_format) {
- case FMT_MJPEG:
- mjpeg_picture_header(s);
- break;
-#ifdef CONFIG_RISKY
- case FMT_H263:
- if (s->codec_id == CODEC_ID_WMV2)
- ff_wmv2_encode_picture_header(s, picture_number);
- else if (s->h263_msmpeg4)
- msmpeg4_encode_picture_header(s, picture_number);
- else if (s->h263_pred)
- mpeg4_encode_picture_header(s, picture_number);
- else if (s->codec_id == CODEC_ID_RV10)
- rv10_encode_picture_header(s, picture_number);
- else if (s->codec_id == CODEC_ID_FLV1)
- ff_flv_encode_picture_header(s, picture_number);
- else
- h263_encode_picture_header(s, picture_number);
- break;
-#endif
- case FMT_MPEG1:
- mpeg1_encode_picture_header(s, picture_number);
- break;
- case FMT_H264:
- break;
+static int encode_thread(AVCodecContext *c, void *arg){
+ MpegEncContext *s= arg;
+ int mb_x, mb_y, pdif = 0;
+ int i, j;
+ MpegEncContext best_s, backup_s;
+ uint8_t bit_buf[2][3000];
+ uint8_t bit_buf2[2][3000];
+ uint8_t bit_buf_tex[2][3000];
+ PutBitContext pb[2], pb2[2], tex_pb[2];
+//printf("%d->%d\n", s->resync_mb_y, s->end_mb_y);
+
+ for(i=0; i<2; i++){
+ init_put_bits(&pb [i], bit_buf [i], 3000);
+ init_put_bits(&pb2 [i], bit_buf2 [i], 3000);
+ init_put_bits(&tex_pb[i], bit_buf_tex[i], 3000);
}
- bits= get_bit_count(&s->pb);
- s->header_bits= bits - s->last_bits;
- s->last_bits= bits;
+
+ s->last_bits= put_bits_count(&s->pb);
s->mv_bits=0;
s->misc_bits=0;
s->i_tex_bits=0;
for(i=0; i<3; i++){
/* init last dc values */
/* note: quant matrix value (8) is implied here */
- s->last_dc[i] = 128;
+ s->last_dc[i] = 128 << s->intra_dc_precision;
s->current_picture_ptr->error[i] = 0;
}
s->mb_skip_run = 0;
- s->last_mv[0][0][0] = 0;
- s->last_mv[0][0][1] = 0;
- s->last_mv[1][0][0] = 0;
- s->last_mv[1][0][1] = 0;
+ memset(s->last_mv, 0, sizeof(s->last_mv));
s->last_mv_dir = 0;
#endif
s->resync_mb_x=0;
- s->resync_mb_y=0;
+ s->resync_mb_y=0;
s->first_slice_line = 1;
s->ptr_lastgob = s->pb.buf;
- for(mb_y=0; mb_y < s->mb_height; mb_y++) {
+ for(mb_y= s->start_mb_y; mb_y < s->end_mb_y; mb_y++) {
+// printf("row %d at %X\n", s->mb_y, (int)s);
s->mb_x=0;
s->mb_y= mb_y;
int mb_type= s->mb_type[xy];
// int d;
int dmin= INT_MAX;
+ int dir;
s->mb_x = mb_x;
ff_update_block_index(s);
if(s->rtp_mode){
int current_packet_size, is_gob_start;
- current_packet_size= pbBufPtr(&s->pb) - s->ptr_lastgob;
+ current_packet_size= ((put_bits_count(&s->pb)+7)>>3) - (s->ptr_lastgob - s->pb.buf);
is_gob_start= s->avctx->rtp_payload_size && current_packet_size >= s->avctx->rtp_payload_size && mb_y + mb_x>0;
+ if(s->start_mb_y == mb_y && mb_y > 0 && mb_x==0) is_gob_start=1;
+
switch(s->codec_id){
case CODEC_ID_H263:
case CODEC_ID_H263P:
if(s->mb_skip_run) is_gob_start=0;
break;
}
-
+
if(is_gob_start){
- if(s->codec_id==CODEC_ID_MPEG4 && s->partitioned_frame){
- ff_mpeg4_merge_partitions(s);
- ff_mpeg4_init_partitions(s);
+ if(s->start_mb_y != mb_y || mb_x!=0){
+ write_slice_end(s);
+
+ if(s->codec_id==CODEC_ID_MPEG4 && s->partitioned_frame){
+ ff_mpeg4_init_partitions(s);
+ }
}
- if(s->codec_id==CODEC_ID_MPEG4)
- ff_mpeg4_stuffing(&s->pb);
-
- align_put_bits(&s->pb);
- flush_put_bits(&s->pb);
-
- assert((get_bit_count(&s->pb)&7) == 0);
+ assert((put_bits_count(&s->pb)&7) == 0);
current_packet_size= pbBufPtr(&s->pb) - s->ptr_lastgob;
+
+ if(s->avctx->error_rate && s->resync_mb_x + s->resync_mb_y > 0){
+ int r= put_bits_count(&s->pb)/8 + s->picture_number + s->codec_id + s->mb_x + s->mb_y;
+ int d= 100 / s->avctx->error_rate;
+ if(r % d == 0){
+ current_packet_size=0;
+#ifndef ALT_BITSTREAM_WRITER
+ s->pb.buf_ptr= s->ptr_lastgob;
+#endif
+ assert(pbBufPtr(&s->pb) == s->ptr_lastgob);
+ }
+ }
if (s->avctx->rtp_callback)
- s->avctx->rtp_callback(s->ptr_lastgob, current_packet_size, 0);
+ s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, current_packet_size, 0);
switch(s->codec_id){
case CODEC_ID_MPEG4:
}
if(s->flags&CODEC_FLAG_PASS1){
- int bits= get_bit_count(&s->pb);
+ int bits= put_bits_count(&s->pb);
s->misc_bits+= bits - s->last_bits;
s->last_bits= bits;
}
backup_s.tex_pb= s->tex_pb;
}
- if(mb_type&MB_TYPE_INTER){
+ if(mb_type&CANDIDATE_MB_TYPE_INTER){
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 0;
s->mv[0][0][0] = s->p_mv_table[xy][0];
s->mv[0][0][1] = s->p_mv_table[xy][1];
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_INTER, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]);
}
- if(mb_type&MB_TYPE_SKIPED){
+ if(mb_type&CANDIDATE_MB_TYPE_INTER_I){
+ s->mv_dir = MV_DIR_FORWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[0][i] = s->p_field_select_table[i][xy];
+ s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0];
+ s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1];
+ }
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER_I, pb, pb2, tex_pb,
+ &dmin, &next_block, 0, 0);
+ }
+ if(mb_type&CANDIDATE_MB_TYPE_SKIPED){
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 0;
s->mv[0][0][0] = 0;
s->mv[0][0][1] = 0;
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_SKIPED, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_SKIPED, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]);
}
- if(mb_type&MB_TYPE_INTER4V){
+ if(mb_type&CANDIDATE_MB_TYPE_INTER4V){
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_8X8;
s->mb_intra= 0;
s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0];
s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1];
}
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_INTER4V, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER4V, pb, pb2, tex_pb,
&dmin, &next_block, 0, 0);
}
- if(mb_type&MB_TYPE_FORWARD){
+ if(mb_type&CANDIDATE_MB_TYPE_FORWARD){
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 0;
s->mv[0][0][0] = s->b_forw_mv_table[xy][0];
s->mv[0][0][1] = s->b_forw_mv_table[xy][1];
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_FORWARD, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]);
}
- if(mb_type&MB_TYPE_BACKWARD){
+ if(mb_type&CANDIDATE_MB_TYPE_BACKWARD){
s->mv_dir = MV_DIR_BACKWARD;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 0;
s->mv[1][0][0] = s->b_back_mv_table[xy][0];
s->mv[1][0][1] = s->b_back_mv_table[xy][1];
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_BACKWARD, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[1][0][0], s->mv[1][0][1]);
}
- if(mb_type&MB_TYPE_BIDIR){
+ if(mb_type&CANDIDATE_MB_TYPE_BIDIR){
s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 0;
s->mv[0][0][1] = s->b_bidir_forw_mv_table[xy][1];
s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0];
s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1];
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_BIDIR, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR, pb, pb2, tex_pb,
&dmin, &next_block, 0, 0);
}
- if(mb_type&MB_TYPE_DIRECT){
+ if(mb_type&CANDIDATE_MB_TYPE_DIRECT){
int mx= s->b_direct_mv_table[xy][0];
int my= s->b_direct_mv_table[xy][1];
#ifdef CONFIG_RISKY
ff_mpeg4_set_direct_mv(s, mx, my);
#endif
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_DIRECT, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_DIRECT, pb, pb2, tex_pb,
&dmin, &next_block, mx, my);
}
- if(mb_type&MB_TYPE_INTRA){
+ if(mb_type&CANDIDATE_MB_TYPE_FORWARD_I){
+ s->mv_dir = MV_DIR_FORWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[0][i] = s->b_field_select_table[0][i][xy];
+ s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0];
+ s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1];
+ }
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD_I, pb, pb2, tex_pb,
+ &dmin, &next_block, 0, 0);
+ }
+ if(mb_type&CANDIDATE_MB_TYPE_BACKWARD_I){
+ s->mv_dir = MV_DIR_BACKWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[1][i] = s->b_field_select_table[1][i][xy];
+ s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0];
+ s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1];
+ }
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD_I, pb, pb2, tex_pb,
+ &dmin, &next_block, 0, 0);
+ }
+ if(mb_type&CANDIDATE_MB_TYPE_BIDIR_I){
+ s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(dir=0; dir<2; dir++){
+ for(i=0; i<2; i++){
+ j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy];
+ s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0];
+ s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1];
+ }
+ }
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR_I, pb, pb2, tex_pb,
+ &dmin, &next_block, 0, 0);
+ }
+ if(mb_type&CANDIDATE_MB_TYPE_INTRA){
s->mv_dir = 0;
s->mv_type = MV_TYPE_16X16;
s->mb_intra= 1;
s->mv[0][0][0] = 0;
s->mv[0][0][1] = 0;
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_INTRA, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTRA, pb, pb2, tex_pb,
&dmin, &next_block, 0, 0);
if(s->h263_pred || s->h263_aic){
if(best_s.mb_intra)
}
}
- encode_mb_hq(s, &backup_s, &best_s, MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb,
+ encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb,
&dmin, &next_block, s->mv[mvdir][0][0], s->mv[mvdir][0][1]);
if(best_s.qscale != qp){
if(s->mb_intra){
copy_context_after_encode(s, &best_s, -1);
- pb_bits_count= get_bit_count(&s->pb);
+ pb_bits_count= put_bits_count(&s->pb);
flush_put_bits(&s->pb);
ff_copy_bits(&backup_s.pb, bit_buf[next_block^1], pb_bits_count);
s->pb= backup_s.pb;
if(s->data_partitioning){
- pb2_bits_count= get_bit_count(&s->pb2);
+ pb2_bits_count= put_bits_count(&s->pb2);
flush_put_bits(&s->pb2);
ff_copy_bits(&backup_s.pb2, bit_buf2[next_block^1], pb2_bits_count);
s->pb2= backup_s.pb2;
- tex_pb_bits_count= get_bit_count(&s->tex_pb);
+ tex_pb_bits_count= put_bits_count(&s->tex_pb);
flush_put_bits(&s->tex_pb);
ff_copy_bits(&backup_s.tex_pb, bit_buf_tex[next_block^1], tex_pb_bits_count);
s->tex_pb= backup_s.tex_pb;
}
- s->last_bits= get_bit_count(&s->pb);
+ s->last_bits= put_bits_count(&s->pb);
#ifdef CONFIG_RISKY
if (s->out_format == FMT_H263 && s->pict_type!=B_TYPE)
ff_h263_update_motion_val(s);
#endif
- if(next_block==0){
- s->dsp.put_pixels_tab[0][0](s->dest[0], s->me.scratchpad , s->linesize ,16);
- s->dsp.put_pixels_tab[1][0](s->dest[1], s->me.scratchpad + 16, s->uvlinesize, 8);
- s->dsp.put_pixels_tab[1][0](s->dest[2], s->me.scratchpad + 24, s->uvlinesize, 8);
+ if(next_block==0){ //FIXME 16 vs linesize16
+ s->dsp.put_pixels_tab[0][0](s->dest[0], s->rd_scratchpad , s->linesize ,16);
+ s->dsp.put_pixels_tab[1][0](s->dest[1], s->rd_scratchpad + 16*s->linesize , s->uvlinesize, 8);
+ s->dsp.put_pixels_tab[1][0](s->dest[2], s->rd_scratchpad + 16*s->linesize + 8, s->uvlinesize, 8);
}
if(s->avctx->mb_decision == FF_MB_DECISION_BITS)
// only one MB-Type possible
switch(mb_type){
- case MB_TYPE_INTRA:
+ case CANDIDATE_MB_TYPE_INTRA:
s->mv_dir = 0;
s->mb_intra= 1;
motion_x= s->mv[0][0][0] = 0;
motion_y= s->mv[0][0][1] = 0;
break;
- case MB_TYPE_INTER:
+ case CANDIDATE_MB_TYPE_INTER:
s->mv_dir = MV_DIR_FORWARD;
s->mb_intra= 0;
motion_x= s->mv[0][0][0] = s->p_mv_table[xy][0];
motion_y= s->mv[0][0][1] = s->p_mv_table[xy][1];
break;
- case MB_TYPE_INTER4V:
+ case CANDIDATE_MB_TYPE_INTER_I:
+ s->mv_dir = MV_DIR_FORWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[0][i] = s->p_field_select_table[i][xy];
+ s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0];
+ s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1];
+ }
+ motion_x = motion_y = 0;
+ break;
+ case CANDIDATE_MB_TYPE_INTER4V:
s->mv_dir = MV_DIR_FORWARD;
s->mv_type = MV_TYPE_8X8;
s->mb_intra= 0;
}
motion_x= motion_y= 0;
break;
- case MB_TYPE_DIRECT:
+ case CANDIDATE_MB_TYPE_DIRECT:
s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT;
s->mb_intra= 0;
motion_x=s->b_direct_mv_table[xy][0];
ff_mpeg4_set_direct_mv(s, motion_x, motion_y);
#endif
break;
- case MB_TYPE_BIDIR:
+ case CANDIDATE_MB_TYPE_BIDIR:
s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD;
s->mb_intra= 0;
motion_x=0;
s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0];
s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1];
break;
- case MB_TYPE_BACKWARD:
+ case CANDIDATE_MB_TYPE_BACKWARD:
s->mv_dir = MV_DIR_BACKWARD;
s->mb_intra= 0;
motion_x= s->mv[1][0][0] = s->b_back_mv_table[xy][0];
motion_y= s->mv[1][0][1] = s->b_back_mv_table[xy][1];
break;
- case MB_TYPE_FORWARD:
+ case CANDIDATE_MB_TYPE_FORWARD:
s->mv_dir = MV_DIR_FORWARD;
s->mb_intra= 0;
motion_x= s->mv[0][0][0] = s->b_forw_mv_table[xy][0];
motion_y= s->mv[0][0][1] = s->b_forw_mv_table[xy][1];
// printf(" %d %d ", motion_x, motion_y);
break;
+ case CANDIDATE_MB_TYPE_FORWARD_I:
+ s->mv_dir = MV_DIR_FORWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[0][i] = s->b_field_select_table[0][i][xy];
+ s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0];
+ s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1];
+ }
+ motion_x=motion_y=0;
+ break;
+ case CANDIDATE_MB_TYPE_BACKWARD_I:
+ s->mv_dir = MV_DIR_BACKWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(i=0; i<2; i++){
+ j= s->field_select[1][i] = s->b_field_select_table[1][i][xy];
+ s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0];
+ s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1];
+ }
+ motion_x=motion_y=0;
+ break;
+ case CANDIDATE_MB_TYPE_BIDIR_I:
+ s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD;
+ s->mv_type = MV_TYPE_FIELD;
+ s->mb_intra= 0;
+ for(dir=0; dir<2; dir++){
+ for(i=0; i<2; i++){
+ j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy];
+ s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0];
+ s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1];
+ }
+ }
+ motion_x=motion_y=0;
+ break;
default:
motion_x=motion_y=0; //gcc warning fix
av_log(s->avctx, AV_LOG_ERROR, "illegal MB type\n");
}
if(s->loop_filter)
ff_h263_loop_filter(s);
-//printf("MB %d %d bits\n", s->mb_x+s->mb_y*s->mb_stride, get_bit_count(&s->pb));
+//printf("MB %d %d bits\n", s->mb_x+s->mb_y*s->mb_stride, put_bits_count(&s->pb));
}
}
- emms_c();
#ifdef CONFIG_RISKY
- if(s->codec_id==CODEC_ID_MPEG4 && s->partitioned_frame)
- ff_mpeg4_merge_partitions(s);
-
+ //not beautifull here but we must write it before flushing so it has to be here
if (s->msmpeg4_version && s->msmpeg4_version<4 && s->pict_type == I_TYPE)
msmpeg4_encode_ext_header(s);
-
- if(s->codec_id==CODEC_ID_MPEG4)
- ff_mpeg4_stuffing(&s->pb);
#endif
+ write_slice_end(s);
+
/* Send the last GOB if RTP */
if (s->avctx->rtp_callback) {
- flush_put_bits(&s->pb);
pdif = pbBufPtr(&s->pb) - s->ptr_lastgob;
/* Call the RTP callback to send the last GOB */
- s->avctx->rtp_callback(s->ptr_lastgob, pdif, 0);
+ emms_c();
+ s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, pdif, 0);
}
+
+ return 0;
}
-#endif //CONFIG_ENCODERS
+#define MERGE(field) dst->field += src->field; src->field=0
+static void merge_context_after_me(MpegEncContext *dst, MpegEncContext *src){
+ MERGE(me.scene_change_score);
+ MERGE(me.mc_mb_var_sum_temp);
+ MERGE(me.mb_var_sum_temp);
+}
-void ff_denoise_dct(MpegEncContext *s, DCTELEM *block){
- const int intra= s->mb_intra;
+static void merge_context_after_encode(MpegEncContext *dst, MpegEncContext *src){
int i;
- s->dct_count[intra]++;
-
- for(i=0; i<64; i++){
- int level= block[i];
-
- if(level){
- if(level>0){
- s->dct_error_sum[intra][i] += level;
- level -= s->dct_offset[intra][i];
- if(level<0) level=0;
- }else{
- s->dct_error_sum[intra][i] -= level;
- level += s->dct_offset[intra][i];
- if(level>0) level=0;
- }
- block[i]= level;
+ MERGE(dct_count[0]); //note, the other dct vars are not part of the context
+ MERGE(dct_count[1]);
+ MERGE(mv_bits);
+ MERGE(i_tex_bits);
+ MERGE(p_tex_bits);
+ MERGE(i_count);
+ MERGE(f_count);
+ MERGE(b_count);
+ MERGE(skip_count);
+ MERGE(misc_bits);
+ MERGE(error_count);
+ MERGE(padding_bug_score);
+
+ if(dst->avctx->noise_reduction){
+ for(i=0; i<64; i++){
+ MERGE(dct_error_sum[0][i]);
+ MERGE(dct_error_sum[1][i]);
}
}
+
+ assert(put_bits_count(&src->pb) % 8 ==0);
+ assert(put_bits_count(&dst->pb) % 8 ==0);
+ ff_copy_bits(&dst->pb, src->pb.buf, put_bits_count(&src->pb));
+ flush_put_bits(&dst->pb);
}
-#ifdef CONFIG_ENCODERS
-
+static void encode_picture(MpegEncContext *s, int picture_number)
+{
+ int i;
+ int bits;
+
+ s->picture_number = picture_number;
+
+ /* Reset the average MB variance */
+ s->me.mb_var_sum_temp =
+ s->me.mc_mb_var_sum_temp = 0;
+
+#ifdef CONFIG_RISKY
+ /* we need to initialize some time vars before we can encode b-frames */
+ // RAL: Condition added for MPEG1VIDEO
+ if (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO || (s->h263_pred && !s->h263_msmpeg4))
+ ff_set_mpeg4_time(s, s->picture_number); //FIXME rename and use has_b_frames or similar
+#endif
+
+ s->me.scene_change_score=0;
+
+// s->lambda= s->current_picture_ptr->quality; //FIXME qscale / ... stuff for ME ratedistoration
+
+ if(s->pict_type==I_TYPE){
+ if(s->msmpeg4_version >= 3) s->no_rounding=1;
+ else s->no_rounding=0;
+ }else if(s->pict_type!=B_TYPE){
+ if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4)
+ s->no_rounding ^= 1;
+ }
+
+ s->mb_intra=0; //for the rate distoration & bit compare functions
+ for(i=1; i<s->avctx->thread_count; i++){
+ ff_update_duplicate_context(s->thread_context[i], s);
+ }
+
+ ff_init_me(s);
+
+ /* Estimate motion for every MB */
+ if(s->pict_type != I_TYPE){
+ if(s->pict_type != B_TYPE && s->avctx->me_threshold==0){
+ if((s->avctx->pre_me && s->last_non_b_pict_type==I_TYPE) || s->avctx->pre_me==2){
+ s->avctx->execute(s->avctx, pre_estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count);
+ }
+ }
+
+ s->avctx->execute(s->avctx, estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count);
+ }else /* if(s->pict_type == I_TYPE) */{
+ /* I-Frame */
+ for(i=0; i<s->mb_stride*s->mb_height; i++)
+ s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA;
+
+ if(!s->fixed_qscale){
+ /* finding spatial complexity for I-frame rate control */
+ s->avctx->execute(s->avctx, mb_var_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count);
+ }
+ }
+ for(i=1; i<s->avctx->thread_count; i++){
+ merge_context_after_me(s, s->thread_context[i]);
+ }
+ s->current_picture.mc_mb_var_sum= s->current_picture_ptr->mc_mb_var_sum= s->me.mc_mb_var_sum_temp;
+ s->current_picture. mb_var_sum= s->current_picture_ptr-> mb_var_sum= s->me. mb_var_sum_temp;
+ emms_c();
+
+ if(s->me.scene_change_score > s->avctx->scenechange_threshold && s->pict_type == P_TYPE){
+ s->pict_type= I_TYPE;
+ for(i=0; i<s->mb_stride*s->mb_height; i++)
+ s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA;
+//printf("Scene change detected, encoding as I Frame %d %d\n", s->current_picture.mb_var_sum, s->current_picture.mc_mb_var_sum);
+ }
+
+ if(!s->umvplus){
+ if(s->pict_type==P_TYPE || s->pict_type==S_TYPE) {
+ s->f_code= ff_get_best_fcode(s, s->p_mv_table, CANDIDATE_MB_TYPE_INTER);
+
+ if(s->flags & CODEC_FLAG_INTERLACED_ME){
+ int a,b;
+ a= ff_get_best_fcode(s, s->p_field_mv_table[0][0], CANDIDATE_MB_TYPE_INTER_I); //FIXME field_select
+ b= ff_get_best_fcode(s, s->p_field_mv_table[1][1], CANDIDATE_MB_TYPE_INTER_I);
+ s->f_code= FFMAX(s->f_code, FFMAX(a,b));
+ }
+
+ ff_fix_long_p_mvs(s);
+ ff_fix_long_mvs(s, NULL, 0, s->p_mv_table, s->f_code, CANDIDATE_MB_TYPE_INTER, 0);
+ if(s->flags & CODEC_FLAG_INTERLACED_ME){
+ int j;
+ for(i=0; i<2; i++){
+ for(j=0; j<2; j++)
+ ff_fix_long_mvs(s, s->p_field_select_table[i], j,
+ s->p_field_mv_table[i][j], s->f_code, CANDIDATE_MB_TYPE_INTER_I, 0);
+ }
+ }
+ }
+
+ if(s->pict_type==B_TYPE){
+ int a, b;
+
+ a = ff_get_best_fcode(s, s->b_forw_mv_table, CANDIDATE_MB_TYPE_FORWARD);
+ b = ff_get_best_fcode(s, s->b_bidir_forw_mv_table, CANDIDATE_MB_TYPE_BIDIR);
+ s->f_code = FFMAX(a, b);
+
+ a = ff_get_best_fcode(s, s->b_back_mv_table, CANDIDATE_MB_TYPE_BACKWARD);
+ b = ff_get_best_fcode(s, s->b_bidir_back_mv_table, CANDIDATE_MB_TYPE_BIDIR);
+ s->b_code = FFMAX(a, b);
+
+ ff_fix_long_mvs(s, NULL, 0, s->b_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_FORWARD, 1);
+ ff_fix_long_mvs(s, NULL, 0, s->b_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BACKWARD, 1);
+ ff_fix_long_mvs(s, NULL, 0, s->b_bidir_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_BIDIR, 1);
+ ff_fix_long_mvs(s, NULL, 0, s->b_bidir_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BIDIR, 1);
+ if(s->flags & CODEC_FLAG_INTERLACED_ME){
+ int dir, j;
+ for(dir=0; dir<2; dir++){
+ for(i=0; i<2; i++){
+ for(j=0; j<2; j++){
+ int type= dir ? (CANDIDATE_MB_TYPE_BACKWARD_I|CANDIDATE_MB_TYPE_BIDIR_I)
+ : (CANDIDATE_MB_TYPE_FORWARD_I |CANDIDATE_MB_TYPE_BIDIR_I);
+ ff_fix_long_mvs(s, s->b_field_select_table[dir][i], j,
+ s->b_field_mv_table[dir][i][j], dir ? s->b_code : s->f_code, type, 1);
+ }
+ }
+ }
+ }
+ }
+ }
+
+ if (!s->fixed_qscale)
+ s->current_picture.quality = ff_rate_estimate_qscale(s); //FIXME pic_ptr
+
+ if(s->adaptive_quant){
+#ifdef CONFIG_RISKY
+ switch(s->codec_id){
+ case CODEC_ID_MPEG4:
+ ff_clean_mpeg4_qscales(s);
+ break;
+ case CODEC_ID_H263:
+ case CODEC_ID_H263P:
+ case CODEC_ID_FLV1:
+ ff_clean_h263_qscales(s);
+ break;
+ }
+#endif
+
+ s->lambda= s->lambda_table[0];
+ //FIXME broken
+ }else
+ s->lambda= s->current_picture.quality;
+//printf("%d %d\n", s->avctx->global_quality, s->current_picture.quality);
+ update_qscale(s);
+
+ if(s->qscale < 3 && s->max_qcoeff<=128 && s->pict_type==I_TYPE && !(s->flags & CODEC_FLAG_QSCALE))
+ s->qscale= 3; //reduce cliping problems
+
+ if (s->out_format == FMT_MJPEG) {
+ /* for mjpeg, we do include qscale in the matrix */
+ s->intra_matrix[0] = ff_mpeg1_default_intra_matrix[0];
+ for(i=1;i<64;i++){
+ int j= s->dsp.idct_permutation[i];
+
+ s->intra_matrix[j] = CLAMP_TO_8BIT((ff_mpeg1_default_intra_matrix[i] * s->qscale) >> 3);
+ }
+ convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16,
+ s->intra_matrix, s->intra_quant_bias, 8, 8);
+ s->qscale= 8;
+ }
+
+ //FIXME var duplication
+ s->current_picture.key_frame= s->pict_type == I_TYPE; //FIXME pic_ptr
+ s->current_picture.pict_type= s->pict_type;
+
+ if(s->current_picture.key_frame)
+ s->picture_in_gop_number=0;
+
+ s->last_bits= put_bits_count(&s->pb);
+ switch(s->out_format) {
+ case FMT_MJPEG:
+ mjpeg_picture_header(s);
+ break;
+#ifdef CONFIG_RISKY
+ case FMT_H263:
+ if (s->codec_id == CODEC_ID_WMV2)
+ ff_wmv2_encode_picture_header(s, picture_number);
+ else if (s->h263_msmpeg4)
+ msmpeg4_encode_picture_header(s, picture_number);
+ else if (s->h263_pred)
+ mpeg4_encode_picture_header(s, picture_number);
+ else if (s->codec_id == CODEC_ID_RV10)
+ rv10_encode_picture_header(s, picture_number);
+ else if (s->codec_id == CODEC_ID_FLV1)
+ ff_flv_encode_picture_header(s, picture_number);
+ else
+ h263_encode_picture_header(s, picture_number);
+ break;
+#endif
+ case FMT_MPEG1:
+ mpeg1_encode_picture_header(s, picture_number);
+ break;
+ case FMT_H264:
+ break;
+ default:
+ assert(0);
+ }
+ bits= put_bits_count(&s->pb);
+ s->header_bits= bits - s->last_bits;
+
+ for(i=1; i<s->avctx->thread_count; i++){
+ update_duplicate_context_after_me(s->thread_context[i], s);
+ }
+ s->avctx->execute(s->avctx, encode_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count);
+ for(i=1; i<s->avctx->thread_count; i++){
+ merge_context_after_encode(s, s->thread_context[i]);
+ }
+ emms_c();
+}
+
+#endif //CONFIG_ENCODERS
+
+static void denoise_dct_c(MpegEncContext *s, DCTELEM *block){
+ const int intra= s->mb_intra;
+ int i;
+
+ s->dct_count[intra]++;
+
+ for(i=0; i<64; i++){
+ int level= block[i];
+
+ if(level){
+ if(level>0){
+ s->dct_error_sum[intra][i] += level;
+ level -= s->dct_offset[intra][i];
+ if(level<0) level=0;
+ }else{
+ s->dct_error_sum[intra][i] -= level;
+ level += s->dct_offset[intra][i];
+ if(level>0) level=0;
+ }
+ block[i]= level;
+ }
+ }
+}
+
+#ifdef CONFIG_ENCODERS
+
static int dct_quantize_trellis_c(MpegEncContext *s,
DCTELEM *block, int n,
int qscale, int *overflow){
const int *qmat;
const uint8_t *scantable= s->intra_scantable.scantable;
+ const uint8_t *perm_scantable= s->intra_scantable.permutated;
int max=0;
unsigned int threshold1, threshold2;
int bias=0;
int run_tab[65];
int level_tab[65];
int score_tab[65];
+ int survivor[65];
+ int survivor_count;
int last_run=0;
int last_level=0;
int last_score= 0;
- int last_i= 0;
- int not_coded_score= 0;
- int coeff[3][64];
+ int last_i;
+ int coeff[2][64];
int coeff_count[64];
int qmul, qadd, start_i, last_non_zero, i, dc;
const int esc_length= s->ac_esc_length;
uint8_t * length;
uint8_t * last_length;
- int score_limit=0;
- int left_limit= 0;
const int lambda= s->lambda2 >> (FF_LAMBDA_SHIFT - 6);
- const int patch_table= s->out_format == FMT_MPEG1 && !s->mb_intra;
s->dsp.fdct (block);
if(s->dct_error_sum)
- ff_denoise_dct(s, block);
-
+ s->denoise_dct(s, block);
qmul= qscale*16;
qadd= ((qscale-1)|1)*8;
length = s->inter_ac_vlc_length;
last_length= s->inter_ac_vlc_last_length;
}
+ last_i= start_i;
threshold1= (1<<QMAT_SHIFT) - bias - 1;
threshold2= (threshold1<<1);
- for(i=start_i; i<64; i++) {
+ for(i=63; i>=start_i; i--) {
const int j = scantable[i];
- const int k= i-start_i;
- int level = block[j];
- level = level * qmat[j];
+ int level = block[j] * qmat[j];
+
+ if(((unsigned)(level+threshold1))>threshold2){
+ last_non_zero = i;
+ break;
+ }
+ }
+
+ for(i=start_i; i<=last_non_zero; i++) {
+ const int j = scantable[i];
+ int level = block[j] * qmat[j];
// if( bias+level >= (1<<(QMAT_SHIFT - 3))
// || bias-level >= (1<<(QMAT_SHIFT - 3))){
if(((unsigned)(level+threshold1))>threshold2){
if(level>0){
level= (bias + level)>>QMAT_SHIFT;
- coeff[0][k]= level;
- coeff[1][k]= level-1;
+ coeff[0][i]= level;
+ coeff[1][i]= level-1;
// coeff[2][k]= level-2;
}else{
level= (bias - level)>>QMAT_SHIFT;
- coeff[0][k]= -level;
- coeff[1][k]= -level+1;
+ coeff[0][i]= -level;
+ coeff[1][i]= -level+1;
// coeff[2][k]= -level+2;
}
- coeff_count[k]= FFMIN(level, 2);
- assert(coeff_count[k]);
+ coeff_count[i]= FFMIN(level, 2);
+ assert(coeff_count[i]);
max |=level;
- last_non_zero = i;
}else{
- coeff[0][k]= (level>>31)|1;
- coeff_count[k]= 1;
+ coeff[0][i]= (level>>31)|1;
+ coeff_count[i]= 1;
}
}
return last_non_zero;
}
- score_tab[0]= 0;
+ score_tab[start_i]= 0;
+ survivor[0]= start_i;
+ survivor_count= 1;
- if(patch_table){
-// length[UNI_AC_ENC_INDEX(0, 63)]=
-// length[UNI_AC_ENC_INDEX(0, 65)]= 2;
- }
-
- for(i=0; i<=last_non_zero - start_i; i++){
- int level_index, run, j;
- const int dct_coeff= block[ scantable[i + start_i] ];
+ for(i=start_i; i<=last_non_zero; i++){
+ int level_index, j;
+ const int dct_coeff= ABS(block[ scantable[i] ]);
const int zero_distoration= dct_coeff*dct_coeff;
int best_score=256*256*256*120;
-
- last_score += zero_distoration;
- not_coded_score += zero_distoration;
for(level_index=0; level_index < coeff_count[i]; level_index++){
int distoration;
int level= coeff[level_index][i];
+ const int alevel= ABS(level);
int unquant_coeff;
assert(level);
if(s->out_format == FMT_H263){
- if(level>0){
- unquant_coeff= level*qmul + qadd;
- }else{
- unquant_coeff= level*qmul - qadd;
- }
+ unquant_coeff= alevel*qmul + qadd;
}else{ //MPEG1
- j= s->dsp.idct_permutation[ scantable[i + start_i] ]; //FIXME optimize
+ j= s->dsp.idct_permutation[ scantable[i] ]; //FIXME optimize
if(s->mb_intra){
- if (level < 0) {
- unquant_coeff = (int)((-level) * qscale * s->intra_matrix[j]) >> 3;
- unquant_coeff = -((unquant_coeff - 1) | 1);
- } else {
- unquant_coeff = (int)( level * qscale * s->intra_matrix[j]) >> 3;
+ unquant_coeff = (int)( alevel * qscale * s->intra_matrix[j]) >> 3;
unquant_coeff = (unquant_coeff - 1) | 1;
- }
}else{
- if (level < 0) {
- unquant_coeff = ((((-level) << 1) + 1) * qscale * ((int) s->inter_matrix[j])) >> 4;
- unquant_coeff = -((unquant_coeff - 1) | 1);
- } else {
- unquant_coeff = ((( level << 1) + 1) * qscale * ((int) s->inter_matrix[j])) >> 4;
+ unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[j])) >> 4;
unquant_coeff = (unquant_coeff - 1) | 1;
- }
}
unquant_coeff<<= 3;
}
- distoration= (unquant_coeff - dct_coeff) * (unquant_coeff - dct_coeff);
+ distoration= (unquant_coeff - dct_coeff) * (unquant_coeff - dct_coeff) - zero_distoration;
level+=64;
if((level&(~127)) == 0){
- for(run=0; run<=i - left_limit; run++){
+ for(j=survivor_count-1; j>=0; j--){
+ int run= i - survivor[j];
int score= distoration + length[UNI_AC_ENC_INDEX(run, level)]*lambda;
score += score_tab[i-run];
if(score < best_score){
- best_score=
- score_tab[i+1]= score;
+ best_score= score;
run_tab[i+1]= run;
level_tab[i+1]= level-64;
}
}
if(s->out_format == FMT_H263){
- for(run=0; run<=i - left_limit; run++){
+ for(j=survivor_count-1; j>=0; j--){
+ int run= i - survivor[j];
int score= distoration + last_length[UNI_AC_ENC_INDEX(run, level)]*lambda;
score += score_tab[i-run];
if(score < last_score){
}
}else{
distoration += esc_length*lambda;
- for(run=0; run<=i - left_limit; run++){
+ for(j=survivor_count-1; j>=0; j--){
+ int run= i - survivor[j];
int score= distoration + score_tab[i-run];
if(score < best_score){
- best_score=
- score_tab[i+1]= score;
+ best_score= score;
run_tab[i+1]= run;
level_tab[i+1]= level-64;
}
}
if(s->out_format == FMT_H263){
- for(run=0; run<=i - left_limit; run++){
+ for(j=survivor_count-1; j>=0; j--){
+ int run= i - survivor[j];
int score= distoration + score_tab[i-run];
if(score < last_score){
last_score= score;
}
}
}
-
- for(j=left_limit; j<=i; j++){
- score_tab[j] += zero_distoration;
- }
- score_limit+= zero_distoration;
- if(score_tab[i+1] < score_limit)
- score_limit= score_tab[i+1];
+ score_tab[i+1]= best_score;
+
//Note: there is a vlc code in mpeg4 which is 1 bit shorter then another one with a shorter run and the same level
- while(score_tab[ left_limit ] > score_limit + lambda) left_limit++;
-
- if(patch_table){
-// length[UNI_AC_ENC_INDEX(0, 63)]=
-// length[UNI_AC_ENC_INDEX(0, 65)]= 3;
+ if(last_non_zero <= 27){
+ for(; survivor_count; survivor_count--){
+ if(score_tab[ survivor[survivor_count-1] ] <= best_score)
+ break;
+ }
+ }else{
+ for(; survivor_count; survivor_count--){
+ if(score_tab[ survivor[survivor_count-1] ] <= best_score + lambda)
+ break;
+ }
}
+
+ survivor[ survivor_count++ ]= i+1;
}
if(s->out_format != FMT_H263){
last_score= 256*256*256*120;
- for(i= left_limit; i<=last_non_zero - start_i + 1; i++){
+ for(i= survivor[0]; i<=last_non_zero + 1; i++){
int score= score_tab[i];
if(i) score += lambda*2; //FIXME exacter?
}
}
- s->coded_score[n] = last_score - not_coded_score;
+ s->coded_score[n] = last_score;
- dc= block[0];
- last_non_zero= last_i - 1 + start_i;
+ dc= ABS(block[0]);
+ last_non_zero= last_i - 1;
memset(block + start_i, 0, (64-start_i)*sizeof(DCTELEM));
if(last_non_zero < start_i)
for(i=0; i<coeff_count[0]; i++){
int level= coeff[i][0];
- int unquant_coeff, score, distoration;
+ int alevel= ABS(level);
+ int unquant_coeff, score, distortion;
if(s->out_format == FMT_H263){
- if(level>0){
- unquant_coeff= (level*qmul + qadd)>>3;
- }else{
- unquant_coeff= (level*qmul - qadd)>>3;
- }
+ unquant_coeff= (alevel*qmul + qadd)>>3;
}else{ //MPEG1
- if (level < 0) {
- unquant_coeff = ((((-level) << 1) + 1) * qscale * ((int) s->inter_matrix[0])) >> 4;
- unquant_coeff = -((unquant_coeff - 1) | 1);
- } else {
- unquant_coeff = ((( level << 1) + 1) * qscale * ((int) s->inter_matrix[0])) >> 4;
- unquant_coeff = (unquant_coeff - 1) | 1;
- }
+ unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[0])) >> 4;
+ unquant_coeff = (unquant_coeff - 1) | 1;
}
unquant_coeff = (unquant_coeff + 4) >> 3;
unquant_coeff<<= 3 + 3;
- distoration= (unquant_coeff - dc) * (unquant_coeff - dc);
+ distortion= (unquant_coeff - dc) * (unquant_coeff - dc);
level+=64;
- if((level&(~127)) == 0)
- score= distoration + last_length[UNI_AC_ENC_INDEX(0, level)]*lambda;
- else
- score= distoration + esc_length*lambda;
+ if((level&(~127)) == 0) score= distortion + last_length[UNI_AC_ENC_INDEX(0, level)]*lambda;
+ else score= distortion + esc_length*lambda;
if(score < best_score){
best_score= score;
i= last_i;
assert(last_level);
-//FIXME use permutated scantable
- block[ s->dsp.idct_permutation[ scantable[last_non_zero] ] ]= last_level;
+
+ block[ perm_scantable[last_non_zero] ]= last_level;
i -= last_run + 1;
- for(;i>0 ; i -= run_tab[i] + 1){
- const int j= s->dsp.idct_permutation[ scantable[i - 1 + start_i] ];
+ for(; i>start_i; i -= run_tab[i] + 1){
+ block[ perm_scantable[i-1] ]= level_tab[i];
+ }
+
+ return last_non_zero;
+}
+
+//#define REFINE_STATS 1
+static int16_t basis[64][64];
+
+static void build_basis(uint8_t *perm){
+ int i, j, x, y;
+ emms_c();
+ for(i=0; i<8; i++){
+ for(j=0; j<8; j++){
+ for(y=0; y<8; y++){
+ for(x=0; x<8; x++){
+ double s= 0.25*(1<<BASIS_SHIFT);
+ int index= 8*i + j;
+ int perm_index= perm[index];
+ if(i==0) s*= sqrt(0.5);
+ if(j==0) s*= sqrt(0.5);
+ basis[perm_index][8*x + y]= lrintf(s * cos((M_PI/8.0)*i*(x+0.5)) * cos((M_PI/8.0)*j*(y+0.5)));
+ }
+ }
+ }
+ }
+}
+
+static int dct_quantize_refine(MpegEncContext *s, //FIXME breaks denoise?
+ DCTELEM *block, int16_t *weight, DCTELEM *orig,
+ int n, int qscale){
+ int16_t rem[64];
+ DCTELEM d1[64];
+ const int *qmat;
+ const uint8_t *scantable= s->intra_scantable.scantable;
+ const uint8_t *perm_scantable= s->intra_scantable.permutated;
+// unsigned int threshold1, threshold2;
+// int bias=0;
+ int run_tab[65];
+ int prev_run=0;
+ int prev_level=0;
+ int qmul, qadd, start_i, last_non_zero, i, dc;
+ uint8_t * length;
+ uint8_t * last_length;
+ int lambda;
+ int rle_index, run, q, sum;
+#ifdef REFINE_STATS
+static int count=0;
+static int after_last=0;
+static int to_zero=0;
+static int from_zero=0;
+static int raise=0;
+static int lower=0;
+static int messed_sign=0;
+#endif
+
+ if(basis[0][0] == 0)
+ build_basis(s->dsp.idct_permutation);
- block[j]= level_tab[i];
- assert(block[j]);
+ qmul= qscale*2;
+ qadd= (qscale-1)|1;
+ if (s->mb_intra) {
+ if (!s->h263_aic) {
+ if (n < 4)
+ q = s->y_dc_scale;
+ else
+ q = s->c_dc_scale;
+ } else{
+ /* For AIC we skip quant/dequant of INTRADC */
+ q = 1;
+ qadd=0;
+ }
+ q <<= RECON_SHIFT-3;
+ /* note: block[0] is assumed to be positive */
+ dc= block[0]*q;
+// block[0] = (block[0] + (q >> 1)) / q;
+ start_i = 1;
+ qmat = s->q_intra_matrix[qscale];
+// if(s->mpeg_quant || s->out_format == FMT_MPEG1)
+// bias= 1<<(QMAT_SHIFT-1);
+ length = s->intra_ac_vlc_length;
+ last_length= s->intra_ac_vlc_last_length;
+ } else {
+ dc= 0;
+ start_i = 0;
+ qmat = s->q_inter_matrix[qscale];
+ length = s->inter_ac_vlc_length;
+ last_length= s->inter_ac_vlc_last_length;
+ }
+ last_non_zero = s->block_last_index[n];
+
+#ifdef REFINE_STATS
+{START_TIMER
+#endif
+ dc += (1<<(RECON_SHIFT-1));
+ for(i=0; i<64; i++){
+ rem[i]= dc - (orig[i]<<RECON_SHIFT); //FIXME use orig dirrectly insteadof copying to rem[]
+ }
+#ifdef REFINE_STATS
+STOP_TIMER("memset rem[]")}
+#endif
+ sum=0;
+ for(i=0; i<64; i++){
+ int one= 36;
+ int qns=4;
+ int w;
+
+ w= ABS(weight[i]) + qns*one;
+ w= 15 + (48*qns*one + w/2)/w; // 16 .. 63
+
+ weight[i] = w;
+// w=weight[i] = (63*qns + (w/2)) / w;
+
+ assert(w>0);
+ assert(w<(1<<6));
+ sum += w*w;
+ }
+ lambda= sum*(uint64_t)s->lambda2 >> (FF_LAMBDA_SHIFT - 6 + 6 + 6 + 6);
+#ifdef REFINE_STATS
+{START_TIMER
+#endif
+ run=0;
+ rle_index=0;
+ for(i=start_i; i<=last_non_zero; i++){
+ int j= perm_scantable[i];
+ const int level= block[j];
+ int coeff;
+
+ if(level){
+ if(level<0) coeff= qmul*level - qadd;
+ else coeff= qmul*level + qadd;
+ run_tab[rle_index++]=run;
+ run=0;
+
+ s->dsp.add_8x8basis(rem, basis[j], coeff);
+ }else{
+ run++;
+ }
}
+#ifdef REFINE_STATS
+if(last_non_zero>0){
+STOP_TIMER("init rem[]")
+}
+}
+
+{START_TIMER
+#endif
+ for(;;){
+ int best_score=s->dsp.try_8x8basis(rem, weight, basis[0], 0);
+ int best_coeff=0;
+ int best_change=0;
+ int run2, best_unquant_change=0, analyze_gradient;
+#ifdef REFINE_STATS
+{START_TIMER
+#endif
+ analyze_gradient = last_non_zero > 2 || s->avctx->quantizer_noise_shaping >= 3;
+
+ if(analyze_gradient){
+#ifdef REFINE_STATS
+{START_TIMER
+#endif
+ for(i=0; i<64; i++){
+ int w= weight[i];
+
+ d1[i] = (rem[i]*w*w + (1<<(RECON_SHIFT+12-1)))>>(RECON_SHIFT+12);
+ }
+#ifdef REFINE_STATS
+STOP_TIMER("rem*w*w")}
+{START_TIMER
+#endif
+ s->dsp.fdct(d1);
+#ifdef REFINE_STATS
+STOP_TIMER("dct")}
+#endif
+ }
+
+ if(start_i){
+ const int level= block[0];
+ int change, old_coeff;
+
+ assert(s->mb_intra);
+
+ old_coeff= q*level;
+
+ for(change=-1; change<=1; change+=2){
+ int new_level= level + change;
+ int score, new_coeff;
+
+ new_coeff= q*new_level;
+ if(new_coeff >= 2048 || new_coeff < 0)
+ continue;
+
+ score= s->dsp.try_8x8basis(rem, weight, basis[0], new_coeff - old_coeff);
+ if(score<best_score){
+ best_score= score;
+ best_coeff= 0;
+ best_change= change;
+ best_unquant_change= new_coeff - old_coeff;
+ }
+ }
+ }
+
+ run=0;
+ rle_index=0;
+ run2= run_tab[rle_index++];
+ prev_level=0;
+ prev_run=0;
+
+ for(i=start_i; i<64; i++){
+ int j= perm_scantable[i];
+ const int level= block[j];
+ int change, old_coeff;
+
+ if(s->avctx->quantizer_noise_shaping < 3 && i > last_non_zero + 1)
+ break;
+
+ if(level){
+ if(level<0) old_coeff= qmul*level - qadd;
+ else old_coeff= qmul*level + qadd;
+ run2= run_tab[rle_index++]; //FIXME ! maybe after last
+ }else{
+ old_coeff=0;
+ run2--;
+ assert(run2>=0 || i >= last_non_zero );
+ }
+
+ for(change=-1; change<=1; change+=2){
+ int new_level= level + change;
+ int score, new_coeff, unquant_change;
+
+ score=0;
+ if(s->avctx->quantizer_noise_shaping < 2 && ABS(new_level) > ABS(level))
+ continue;
+
+ if(new_level){
+ if(new_level<0) new_coeff= qmul*new_level - qadd;
+ else new_coeff= qmul*new_level + qadd;
+ if(new_coeff >= 2048 || new_coeff <= -2048)
+ continue;
+ //FIXME check for overflow
+
+ if(level){
+ if(level < 63 && level > -63){
+ if(i < last_non_zero)
+ score += length[UNI_AC_ENC_INDEX(run, new_level+64)]
+ - length[UNI_AC_ENC_INDEX(run, level+64)];
+ else
+ score += last_length[UNI_AC_ENC_INDEX(run, new_level+64)]
+ - last_length[UNI_AC_ENC_INDEX(run, level+64)];
+ }
+ }else{
+ assert(ABS(new_level)==1);
+
+ if(analyze_gradient){
+ int g= d1[ scantable[i] ];
+ if(g && (g^new_level) >= 0)
+ continue;
+ }
+
+ if(i < last_non_zero){
+ int next_i= i + run2 + 1;
+ int next_level= block[ perm_scantable[next_i] ] + 64;
+
+ if(next_level&(~127))
+ next_level= 0;
+
+ if(next_i < last_non_zero)
+ score += length[UNI_AC_ENC_INDEX(run, 65)]
+ + length[UNI_AC_ENC_INDEX(run2, next_level)]
+ - length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)];
+ else
+ score += length[UNI_AC_ENC_INDEX(run, 65)]
+ + last_length[UNI_AC_ENC_INDEX(run2, next_level)]
+ - last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)];
+ }else{
+ score += last_length[UNI_AC_ENC_INDEX(run, 65)];
+ if(prev_level){
+ score += length[UNI_AC_ENC_INDEX(prev_run, prev_level)]
+ - last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)];
+ }
+ }
+ }
+ }else{
+ new_coeff=0;
+ assert(ABS(level)==1);
+
+ if(i < last_non_zero){
+ int next_i= i + run2 + 1;
+ int next_level= block[ perm_scantable[next_i] ] + 64;
+
+ if(next_level&(~127))
+ next_level= 0;
+
+ if(next_i < last_non_zero)
+ score += length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]
+ - length[UNI_AC_ENC_INDEX(run2, next_level)]
+ - length[UNI_AC_ENC_INDEX(run, 65)];
+ else
+ score += last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]
+ - last_length[UNI_AC_ENC_INDEX(run2, next_level)]
+ - length[UNI_AC_ENC_INDEX(run, 65)];
+ }else{
+ score += -last_length[UNI_AC_ENC_INDEX(run, 65)];
+ if(prev_level){
+ score += last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)]
+ - length[UNI_AC_ENC_INDEX(prev_run, prev_level)];
+ }
+ }
+ }
+
+ score *= lambda;
+
+ unquant_change= new_coeff - old_coeff;
+ assert((score < 100*lambda && score > -100*lambda) || lambda==0);
+
+ score+= s->dsp.try_8x8basis(rem, weight, basis[j], unquant_change);
+ if(score<best_score){
+ best_score= score;
+ best_coeff= i;
+ best_change= change;
+ best_unquant_change= unquant_change;
+ }
+ }
+ if(level){
+ prev_level= level + 64;
+ if(prev_level&(~127))
+ prev_level= 0;
+ prev_run= run;
+ run=0;
+ }else{
+ run++;
+ }
+ }
+#ifdef REFINE_STATS
+STOP_TIMER("iterative step")}
+#endif
+
+ if(best_change){
+ int j= perm_scantable[ best_coeff ];
+
+ block[j] += best_change;
+
+ if(best_coeff > last_non_zero){
+ last_non_zero= best_coeff;
+ assert(block[j]);
+#ifdef REFINE_STATS
+after_last++;
+#endif
+ }else{
+#ifdef REFINE_STATS
+if(block[j]){
+ if(block[j] - best_change){
+ if(ABS(block[j]) > ABS(block[j] - best_change)){
+ raise++;
+ }else{
+ lower++;
+ }
+ }else{
+ from_zero++;
+ }
+}else{
+ to_zero++;
+}
+#endif
+ for(; last_non_zero>=start_i; last_non_zero--){
+ if(block[perm_scantable[last_non_zero]])
+ break;
+ }
+ }
+#ifdef REFINE_STATS
+count++;
+if(256*256*256*64 % count == 0){
+ printf("after_last:%d to_zero:%d from_zero:%d raise:%d lower:%d sign:%d xyp:%d/%d/%d\n", after_last, to_zero, from_zero, raise, lower, messed_sign, s->mb_x, s->mb_y, s->picture_number);
+}
+#endif
+ run=0;
+ rle_index=0;
+ for(i=start_i; i<=last_non_zero; i++){
+ int j= perm_scantable[i];
+ const int level= block[j];
+
+ if(level){
+ run_tab[rle_index++]=run;
+ run=0;
+ }else{
+ run++;
+ }
+ }
+
+ s->dsp.add_8x8basis(rem, basis[j], best_unquant_change);
+ }else{
+ break;
+ }
+ }
+#ifdef REFINE_STATS
+if(last_non_zero>0){
+STOP_TIMER("iterative search")
+}
+}
+#endif
return last_non_zero;
}
DCTELEM *block, int n,
int qscale, int *overflow)
{
- int i, j, level, last_non_zero, q;
+ int i, j, level, last_non_zero, q, start_i;
const int *qmat;
const uint8_t *scantable= s->intra_scantable.scantable;
int bias;
s->dsp.fdct (block);
if(s->dct_error_sum)
- ff_denoise_dct(s, block);
+ s->denoise_dct(s, block);
if (s->mb_intra) {
if (!s->h263_aic) {
/* note: block[0] is assumed to be positive */
block[0] = (block[0] + (q >> 1)) / q;
- i = 1;
+ start_i = 1;
last_non_zero = 0;
qmat = s->q_intra_matrix[qscale];
bias= s->intra_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT);
} else {
- i = 0;
+ start_i = 0;
last_non_zero = -1;
qmat = s->q_inter_matrix[qscale];
bias= s->inter_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT);
}
threshold1= (1<<QMAT_SHIFT) - bias - 1;
threshold2= (threshold1<<1);
+ for(i=63;i>=start_i;i--) {
+ j = scantable[i];
+ level = block[j] * qmat[j];
- for(;i<64;i++) {
+ if(((unsigned)(level+threshold1))>threshold2){
+ last_non_zero = i;
+ break;
+ }else{
+ block[j]=0;
+ }
+ }
+ for(i=start_i; i<=last_non_zero; i++) {
j = scantable[i];
- level = block[j];
- level = level * qmat[j];
+ level = block[j] * qmat[j];
// if( bias+level >= (1<<QMAT_SHIFT)
// || bias-level >= (1<<QMAT_SHIFT)){
block[j]= -level;
}
max |=level;
- last_non_zero = i;
}else{
block[j]=0;
}
#endif //CONFIG_ENCODERS
-static void dct_unquantize_mpeg1_c(MpegEncContext *s,
+static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, nCoeffs;
nCoeffs= s->block_last_index[n];
- if (s->mb_intra) {
- if (n < 4)
- block[0] = block[0] * s->y_dc_scale;
- else
- block[0] = block[0] * s->c_dc_scale;
- /* XXX: only mpeg1 */
- quant_matrix = s->intra_matrix;
- for(i=1;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = (level - 1) | 1;
- level = -level;
- } else {
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = (level - 1) | 1;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ /* XXX: only mpeg1 */
+ quant_matrix = s->intra_matrix;
+ for(i=1;i<=nCoeffs;i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = (level - 1) | 1;
+ level = -level;
+ } else {
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = (level - 1) | 1;
}
+ block[j] = level;
}
- } else {
- i = 0;
- quant_matrix = s->inter_matrix;
- for(;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = (level - 1) | 1;
- level = -level;
- } else {
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = (level - 1) | 1;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
+ }
+}
+
+static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, nCoeffs;
+ const uint16_t *quant_matrix;
+
+ nCoeffs= s->block_last_index[n];
+
+ quant_matrix = s->inter_matrix;
+ for(i=0; i<=nCoeffs; i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = (level - 1) | 1;
+ level = -level;
+ } else {
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = (level - 1) | 1;
}
+ block[j] = level;
}
}
}
-static void dct_unquantize_mpeg2_c(MpegEncContext *s,
+static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, nCoeffs;
if(s->alternate_scan) nCoeffs= 63;
else nCoeffs= s->block_last_index[n];
- if (s->mb_intra) {
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ quant_matrix = s->intra_matrix;
+ for(i=1;i<=nCoeffs;i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = -level;
+ } else {
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ }
+ block[j] = level;
+ }
+ }
+}
+
+static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, nCoeffs;
+ const uint16_t *quant_matrix;
+ int sum=-1;
+
+ if(s->alternate_scan) nCoeffs= 63;
+ else nCoeffs= s->block_last_index[n];
+
+ quant_matrix = s->inter_matrix;
+ for(i=0; i<=nCoeffs; i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ level = -level;
+ } else {
+ level = (((level << 1) + 1) * qscale *
+ ((int) (quant_matrix[j]))) >> 4;
+ }
+ block[j] = level;
+ sum+=level;
+ }
+ }
+ block[63]^=sum&1;
+}
+
+static void dct_unquantize_h263_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, qmul, qadd;
+ int nCoeffs;
+
+ assert(s->block_last_index[n]>=0);
+
+ qmul = qscale << 1;
+
+ if (!s->h263_aic) {
if (n < 4)
block[0] = block[0] * s->y_dc_scale;
else
block[0] = block[0] * s->c_dc_scale;
- quant_matrix = s->intra_matrix;
- for(i=1;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- level = -level;
- } else {
- level = (int)(level * qscale * quant_matrix[j]) >> 3;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
- }
- }
- } else {
- int sum=-1;
- i = 0;
- quant_matrix = s->inter_matrix;
- for(;i<=nCoeffs;i++) {
- int j= s->intra_scantable.permutated[i];
- level = block[j];
- if (level) {
- if (level < 0) {
- level = -level;
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- level = -level;
- } else {
- level = (((level << 1) + 1) * qscale *
- ((int) (quant_matrix[j]))) >> 4;
- }
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
- block[j] = level;
- sum+=level;
+ qadd = (qscale - 1) | 1;
+ }else{
+ qadd = 0;
+ }
+ if(s->ac_pred)
+ nCoeffs=63;
+ else
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
+
+ for(i=1; i<=nCoeffs; i++) {
+ level = block[i];
+ if (level) {
+ if (level < 0) {
+ level = level * qmul - qadd;
+ } else {
+ level = level * qmul + qadd;
}
+ block[i] = level;
}
- block[63]^=sum&1;
}
}
-
-static void dct_unquantize_h263_c(MpegEncContext *s,
+static void dct_unquantize_h263_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, qmul, qadd;
qadd = (qscale - 1) | 1;
qmul = qscale << 1;
- if (s->mb_intra) {
- if (!s->h263_aic) {
- if (n < 4)
- block[0] = block[0] * s->y_dc_scale;
- else
- block[0] = block[0] * s->c_dc_scale;
- }else
- qadd = 0;
- i = 1;
- nCoeffs= 63; //does not allways use zigzag table
- } else {
- i = 0;
- nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
- }
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
- for(;i<=nCoeffs;i++) {
+ for(i=0; i<=nCoeffs; i++) {
level = block[i];
if (level) {
if (level < 0) {
} else {
level = level * qmul + qadd;
}
-#ifdef PARANOID
- if (level < -2048 || level > 2047)
- fprintf(stderr, "unquant error %d %d\n", i, level);
-#endif
block[i] = level;
}
}
}
+static void dct_unquantize_h261_intra_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, even;
+ int nCoeffs;
+
+ assert(s->block_last_index[n]>=0);
+
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ even = (qscale & 1)^1;
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
+
+ for(i=1; i<=nCoeffs; i++){
+ level = block[i];
+ if (level){
+ if (level < 0){
+ level = qscale * ((level << 1) - 1) + even;
+ }else{
+ level = qscale * ((level << 1) + 1) - even;
+ }
+ }
+ block[i] = level;
+ }
+}
+
+static void dct_unquantize_h261_inter_c(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, even;
+ int nCoeffs;
+
+ assert(s->block_last_index[n]>=0);
+
+ even = (qscale & 1)^1;
+
+ nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
+
+ for(i=0; i<=nCoeffs; i++){
+ level = block[i];
+ if (level){
+ if (level < 0){
+ level = qscale * ((level << 1) - 1) + even;
+ }else{
+ level = qscale * ((level << 1) + 1) - even;
+ }
+ }
+ block[i] = level;
+ }
+}
static const AVOption mpeg4_options[] =
{
};
#ifdef CONFIG_ENCODERS
-
-AVCodec mpeg1video_encoder = {
- "mpeg1video",
- CODEC_TYPE_VIDEO,
- CODEC_ID_MPEG1VIDEO,
- sizeof(MpegEncContext),
- MPV_encode_init,
- MPV_encode_picture,
- MPV_encode_end,
-};
-
#ifdef CONFIG_RISKY
-
-AVCodec mpeg2video_encoder = {
- "mpeg2video",
- CODEC_TYPE_VIDEO,
- CODEC_ID_MPEG2VIDEO,
- sizeof(MpegEncContext),
- MPV_encode_init,
- MPV_encode_picture,
- MPV_encode_end,
-};
-
AVCodec h263_encoder = {
"h263",
CODEC_TYPE_VIDEO,
MPV_encode_picture,
MPV_encode_end,
.options = mpeg4_options,
+ .capabilities= CODEC_CAP_DELAY,
};
AVCodec msmpeg4v1_encoder = {
projects / ffmpeg / blobdiff