snow: split snow in snowdec and snowenc

The common non inlined code goes in snow.c, the common inlined code in
snow.h, tables move in snowdata.h (included only by snow.c)
This commit is contained in:
Luca Barbato 2011-11-02 10:10:40 -07:00
parent 8b7412fe4e
commit 7f1b427018
6 changed files with 3268 additions and 3113 deletions

View File

@ -341,12 +341,12 @@ OBJS-$(CONFIG_SIPR_DECODER) += sipr.o acelp_pitch_delay.o \
OBJS-$(CONFIG_SMACKAUD_DECODER) += smacker.o
OBJS-$(CONFIG_SMACKER_DECODER) += smacker.o
OBJS-$(CONFIG_SMC_DECODER) += smc.o
OBJS-$(CONFIG_SNOW_DECODER) += snow.o rangecoder.o
OBJS-$(CONFIG_SNOW_ENCODER) += snow.o rangecoder.o motion_est.o \
ratecontrol.o h263.o \
mpegvideo.o error_resilience.o \
ituh263enc.o mpegvideo_enc.o \
mpeg12data.o
OBJS-$(CONFIG_SNOW_DECODER) += snowdec.o snow.o rangecoder.o
OBJS-$(CONFIG_SNOW_ENCODER) += snowenc.o snow.o rangecoder.o \
motion_est.o ratecontrol.o \
h263.o mpegvideo.o \
error_resilience.o ituh263enc.o \
mpegvideo_enc.o mpeg12data.o
OBJS-$(CONFIG_SOL_DPCM_DECODER) += dpcm.o
OBJS-$(CONFIG_SP5X_DECODER) += sp5xdec.o mjpegdec.o mjpeg.o
OBJS-$(CONFIG_SRT_DECODER) += srtdec.o ass.o

File diff suppressed because it is too large Load Diff

View File

@ -25,6 +25,10 @@
#include "dsputil.h"
#include "dwt.h"
#include "rangecoder.h"
#include "mathops.h"
#include "mpegvideo.h"
#define MID_STATE 128
#define MAX_PLANES 4
@ -36,6 +40,138 @@
#define LOG2_OBMC_MAX 8
#define OBMC_MAX (1<<(LOG2_OBMC_MAX))
typedef struct BlockNode{
int16_t mx;
int16_t my;
uint8_t ref;
uint8_t color[3];
uint8_t type;
//#define TYPE_SPLIT 1
#define BLOCK_INTRA 1
#define BLOCK_OPT 2
//#define TYPE_NOCOLOR 4
uint8_t level; //FIXME merge into type?
}BlockNode;
static const BlockNode null_block= { //FIXME add border maybe
.color= {128,128,128},
.mx= 0,
.my= 0,
.ref= 0,
.type= 0,
.level= 0,
};
#define LOG2_MB_SIZE 4
#define MB_SIZE (1<<LOG2_MB_SIZE)
#define ENCODER_EXTRA_BITS 4
#define HTAPS_MAX 8
typedef struct x_and_coeff{
int16_t x;
uint16_t coeff;
} x_and_coeff;
typedef struct SubBand{
int level;
int stride;
int width;
int height;
int qlog; ///< log(qscale)/log[2^(1/6)]
DWTELEM *buf;
IDWTELEM *ibuf;
int buf_x_offset;
int buf_y_offset;
int stride_line; ///< Stride measured in lines, not pixels.
x_and_coeff * x_coeff;
struct SubBand *parent;
uint8_t state[/*7*2*/ 7 + 512][32];
}SubBand;
typedef struct Plane{
int width;
int height;
SubBand band[MAX_DECOMPOSITIONS][4];
int htaps;
int8_t hcoeff[HTAPS_MAX/2];
int diag_mc;
int fast_mc;
int last_htaps;
int8_t last_hcoeff[HTAPS_MAX/2];
int last_diag_mc;
}Plane;
typedef struct SnowContext{
AVClass *class;
AVCodecContext *avctx;
RangeCoder c;
DSPContext dsp;
DWTContext dwt;
AVFrame new_picture;
AVFrame input_picture; ///< new_picture with the internal linesizes
AVFrame current_picture;
AVFrame last_picture[MAX_REF_FRAMES];
uint8_t *halfpel_plane[MAX_REF_FRAMES][4][4];
AVFrame mconly_picture;
// uint8_t q_context[16];
uint8_t header_state[32];
uint8_t block_state[128 + 32*128];
int keyframe;
int always_reset;
int version;
int spatial_decomposition_type;
int last_spatial_decomposition_type;
int temporal_decomposition_type;
int spatial_decomposition_count;
int last_spatial_decomposition_count;
int temporal_decomposition_count;
int max_ref_frames;
int ref_frames;
int16_t (*ref_mvs[MAX_REF_FRAMES])[2];
uint32_t *ref_scores[MAX_REF_FRAMES];
DWTELEM *spatial_dwt_buffer;
IDWTELEM *spatial_idwt_buffer;
int colorspace_type;
int chroma_h_shift;
int chroma_v_shift;
int spatial_scalability;
int qlog;
int last_qlog;
int lambda;
int lambda2;
int pass1_rc;
int mv_scale;
int last_mv_scale;
int qbias;
int last_qbias;
#define QBIAS_SHIFT 3
int b_width;
int b_height;
int block_max_depth;
int last_block_max_depth;
Plane plane[MAX_PLANES];
BlockNode *block;
#define ME_CACHE_SIZE 1024
int me_cache[ME_CACHE_SIZE];
int me_cache_generation;
slice_buffer sb;
int memc_only;
MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to eventually make the motion estimation independent of MpegEncContext, so this will be removed then (FIXME/XXX)
uint8_t *scratchbuf;
}SnowContext;
/* Tables */
extern const uint8_t * const obmc_tab[4];
#ifdef __sgi
// Avoid a name clash on SGI IRIX
#undef qexp
#endif
extern uint8_t qexp[QROOT];
extern int scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES];
/* C bits used by mmx/sse2/altivec */
@ -75,4 +211,482 @@ static av_always_inline void snow_horizontal_compose_liftS_lead_out(int i, IDWTE
}
}
/* common code */
int ff_snow_common_init(AVCodecContext *avctx);
int ff_snow_common_init_after_header(AVCodecContext *avctx);
void ff_snow_common_end(SnowContext *s);
void ff_snow_release_buffer(AVCodecContext *avctx);
void ff_snow_reset_contexts(SnowContext *s);
int ff_snow_alloc_blocks(SnowContext *s);
int ff_snow_frame_start(SnowContext *s);
void ff_snow_pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride,
int sx, int sy, int b_w, int b_h, BlockNode *block,
int plane_index, int w, int h);
/* common inline functions */
//XXX doublecheck all of them should stay inlined
static inline void snow_set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type){
const int w= s->b_width << s->block_max_depth;
const int rem_depth= s->block_max_depth - level;
const int index= (x + y*w) << rem_depth;
const int block_w= 1<<rem_depth;
BlockNode block;
int i,j;
block.color[0]= l;
block.color[1]= cb;
block.color[2]= cr;
block.mx= mx;
block.my= my;
block.ref= ref;
block.type= type;
block.level= level;
for(j=0; j<block_w; j++){
for(i=0; i<block_w; i++){
s->block[index + i + j*w]= block;
}
}
}
static inline void pred_mv(SnowContext *s, int *mx, int *my, int ref,
const BlockNode *left, const BlockNode *top, const BlockNode *tr){
if(s->ref_frames == 1){
*mx = mid_pred(left->mx, top->mx, tr->mx);
*my = mid_pred(left->my, top->my, tr->my);
}else{
const int *scale = scale_mv_ref[ref];
*mx = mid_pred((left->mx * scale[left->ref] + 128) >>8,
(top ->mx * scale[top ->ref] + 128) >>8,
(tr ->mx * scale[tr ->ref] + 128) >>8);
*my = mid_pred((left->my * scale[left->ref] + 128) >>8,
(top ->my * scale[top ->ref] + 128) >>8,
(tr ->my * scale[tr ->ref] + 128) >>8);
}
}
static av_always_inline int same_block(BlockNode *a, BlockNode *b){
if((a->type&BLOCK_INTRA) && (b->type&BLOCK_INTRA)){
return !((a->color[0] - b->color[0]) | (a->color[1] - b->color[1]) | (a->color[2] - b->color[2]));
}else{
return !((a->mx - b->mx) | (a->my - b->my) | (a->ref - b->ref) | ((a->type ^ b->type)&BLOCK_INTRA));
}
}
//FIXME name cleanup (b_w, block_w, b_width stuff)
//XXX should we really inline it?
static av_always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, IDWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index){
const int b_width = s->b_width << s->block_max_depth;
const int b_height= s->b_height << s->block_max_depth;
const int b_stride= b_width;
BlockNode *lt= &s->block[b_x + b_y*b_stride];
BlockNode *rt= lt+1;
BlockNode *lb= lt+b_stride;
BlockNode *rb= lb+1;
uint8_t *block[4];
int tmp_step= src_stride >= 7*MB_SIZE ? MB_SIZE : MB_SIZE*src_stride;
uint8_t *tmp = s->scratchbuf;
uint8_t *ptmp;
int x,y;
if(b_x<0){
lt= rt;
lb= rb;
}else if(b_x + 1 >= b_width){
rt= lt;
rb= lb;
}
if(b_y<0){
lt= lb;
rt= rb;
}else if(b_y + 1 >= b_height){
lb= lt;
rb= rt;
}
if(src_x<0){ //FIXME merge with prev & always round internal width up to *16
obmc -= src_x;
b_w += src_x;
if(!sliced && !offset_dst)
dst -= src_x;
src_x=0;
}else if(src_x + b_w > w){
b_w = w - src_x;
}
if(src_y<0){
obmc -= src_y*obmc_stride;
b_h += src_y;
if(!sliced && !offset_dst)
dst -= src_y*dst_stride;
src_y=0;
}else if(src_y + b_h> h){
b_h = h - src_y;
}
if(b_w<=0 || b_h<=0) return;
assert(src_stride > 2*MB_SIZE + 5);
if(!sliced && offset_dst)
dst += src_x + src_y*dst_stride;
dst8+= src_x + src_y*src_stride;
// src += src_x + src_y*src_stride;
ptmp= tmp + 3*tmp_step;
block[0]= ptmp;
ptmp+=tmp_step;
ff_snow_pred_block(s, block[0], tmp, src_stride, src_x, src_y, b_w, b_h, lt, plane_index, w, h);
if(same_block(lt, rt)){
block[1]= block[0];
}else{
block[1]= ptmp;
ptmp+=tmp_step;
ff_snow_pred_block(s, block[1], tmp, src_stride, src_x, src_y, b_w, b_h, rt, plane_index, w, h);
}
if(same_block(lt, lb)){
block[2]= block[0];
}else if(same_block(rt, lb)){
block[2]= block[1];
}else{
block[2]= ptmp;
ptmp+=tmp_step;
ff_snow_pred_block(s, block[2], tmp, src_stride, src_x, src_y, b_w, b_h, lb, plane_index, w, h);
}
if(same_block(lt, rb) ){
block[3]= block[0];
}else if(same_block(rt, rb)){
block[3]= block[1];
}else if(same_block(lb, rb)){
block[3]= block[2];
}else{
block[3]= ptmp;
ff_snow_pred_block(s, block[3], tmp, src_stride, src_x, src_y, b_w, b_h, rb, plane_index, w, h);
}
if(sliced){
s->dwt.inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
}else{
for(y=0; y<b_h; y++){
//FIXME ugly misuse of obmc_stride
const uint8_t *obmc1= obmc + y*obmc_stride;
const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
const uint8_t *obmc4= obmc3+ (obmc_stride>>1);
for(x=0; x<b_w; x++){
int v= obmc1[x] * block[3][x + y*src_stride]
+obmc2[x] * block[2][x + y*src_stride]
+obmc3[x] * block[1][x + y*src_stride]
+obmc4[x] * block[0][x + y*src_stride];
v <<= 8 - LOG2_OBMC_MAX;
if(FRAC_BITS != 8){
v >>= 8 - FRAC_BITS;
}
if(add){
v += dst[x + y*dst_stride];
v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS;
if(v&(~255)) v= ~(v>>31);
dst8[x + y*src_stride] = v;
}else{
dst[x + y*dst_stride] -= v;
}
}
}
}
}
static av_always_inline void predict_slice(SnowContext *s, IDWTELEM *buf, int plane_index, int add, int mb_y){
Plane *p= &s->plane[plane_index];
const int mb_w= s->b_width << s->block_max_depth;
const int mb_h= s->b_height << s->block_max_depth;
int x, y, mb_x;
int block_size = MB_SIZE >> s->block_max_depth;
int block_w = plane_index ? block_size/2 : block_size;
const uint8_t *obmc = plane_index ? obmc_tab[s->block_max_depth+1] : obmc_tab[s->block_max_depth];
const int obmc_stride= plane_index ? block_size : 2*block_size;
int ref_stride= s->current_picture.linesize[plane_index];
uint8_t *dst8= s->current_picture.data[plane_index];
int w= p->width;
int h= p->height;
if(s->keyframe || (s->avctx->debug&512)){
if(mb_y==mb_h)
return;
if(add){
for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
for(x=0; x<w; x++){
int v= buf[x + y*w] + (128<<FRAC_BITS) + (1<<(FRAC_BITS-1));
v >>= FRAC_BITS;
if(v&(~255)) v= ~(v>>31);
dst8[x + y*ref_stride]= v;
}
}
}else{
for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
for(x=0; x<w; x++){
buf[x + y*w]-= 128<<FRAC_BITS;
}
}
}
return;
}
for(mb_x=0; mb_x<=mb_w; mb_x++){
add_yblock(s, 0, NULL, buf, dst8, obmc,
block_w*mb_x - block_w/2,
block_w*mb_y - block_w/2,
block_w, block_w,
w, h,
w, ref_stride, obmc_stride,
mb_x - 1, mb_y - 1,
add, 1, plane_index);
}
}
static av_always_inline void predict_plane(SnowContext *s, IDWTELEM *buf, int plane_index, int add){
const int mb_h= s->b_height << s->block_max_depth;
int mb_y;
for(mb_y=0; mb_y<=mb_h; mb_y++)
predict_slice(s, buf, plane_index, add, mb_y);
}
static inline void set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type){
const int w= s->b_width << s->block_max_depth;
const int rem_depth= s->block_max_depth - level;
const int index= (x + y*w) << rem_depth;
const int block_w= 1<<rem_depth;
BlockNode block;
int i,j;
block.color[0]= l;
block.color[1]= cb;
block.color[2]= cr;
block.mx= mx;
block.my= my;
block.ref= ref;
block.type= type;
block.level= level;
for(j=0; j<block_w; j++){
for(i=0; i<block_w; i++){
s->block[index + i + j*w]= block;
}
}
}
static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
const int offset[3]= {
y*c-> stride + x,
((y*c->uvstride + x)>>1),
((y*c->uvstride + x)>>1),
};
int i;
for(i=0; i<3; i++){
c->src[0][i]= src [i];
c->ref[0][i]= ref [i] + offset[i];
}
assert(!ref_index);
}
/* bitstream functions */
extern const int8_t quant3bA[256];
#define QEXPSHIFT (7-FRAC_BITS+8) //FIXME try to change this to 0
static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed){
int i;
if(v){
const int a= FFABS(v);
const int e= av_log2(a);
const int el= FFMIN(e, 10);
put_rac(c, state+0, 0);
for(i=0; i<el; i++){
put_rac(c, state+1+i, 1); //1..10
}
for(; i<e; i++){
put_rac(c, state+1+9, 1); //1..10
}
put_rac(c, state+1+FFMIN(i,9), 0);
for(i=e-1; i>=el; i--){
put_rac(c, state+22+9, (a>>i)&1); //22..31
}
for(; i>=0; i--){
put_rac(c, state+22+i, (a>>i)&1); //22..31
}
if(is_signed)
put_rac(c, state+11 + el, v < 0); //11..21
}else{
put_rac(c, state+0, 1);
}
}
static inline int get_symbol(RangeCoder *c, uint8_t *state, int is_signed){
if(get_rac(c, state+0))
return 0;
else{
int i, e, a;
e= 0;
while(get_rac(c, state+1 + FFMIN(e,9))){ //1..10
e++;
}
a= 1;
for(i=e-1; i>=0; i--){
a += a + get_rac(c, state+22 + FFMIN(i,9)); //22..31
}
e= -(is_signed && get_rac(c, state+11 + FFMIN(e,10))); //11..21
return (a^e)-e;
}
}
static inline void put_symbol2(RangeCoder *c, uint8_t *state, int v, int log2){
int i;
int r= log2>=0 ? 1<<log2 : 1;
assert(v>=0);
assert(log2>=-4);
while(v >= r){
put_rac(c, state+4+log2, 1);
v -= r;
log2++;
if(log2>0) r+=r;
}
put_rac(c, state+4+log2, 0);
for(i=log2-1; i>=0; i--){
put_rac(c, state+31-i, (v>>i)&1);
}
}
static inline int get_symbol2(RangeCoder *c, uint8_t *state, int log2){
int i;
int r= log2>=0 ? 1<<log2 : 1;
int v=0;
assert(log2>=-4);
while(get_rac(c, state+4+log2)){
v+= r;
log2++;
if(log2>0) r+=r;
}
for(i=log2-1; i>=0; i--){
v+= get_rac(c, state+31-i)<<i;
}
return v;
}
static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, int orientation){
const int w= b->width;
const int h= b->height;
int x,y;
int run, runs;
x_and_coeff *xc= b->x_coeff;
x_and_coeff *prev_xc= NULL;
x_and_coeff *prev2_xc= xc;
x_and_coeff *parent_xc= parent ? parent->x_coeff : NULL;
x_and_coeff *prev_parent_xc= parent_xc;
runs= get_symbol2(&s->c, b->state[30], 0);
if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
else run= INT_MAX;
for(y=0; y<h; y++){
int v=0;
int lt=0, t=0, rt=0;
if(y && prev_xc->x == 0){
rt= prev_xc->coeff;
}
for(x=0; x<w; x++){
int p=0;
const int l= v;
lt= t; t= rt;
if(y){
if(prev_xc->x <= x)
prev_xc++;
if(prev_xc->x == x + 1)
rt= prev_xc->coeff;
else
rt=0;
}
if(parent_xc){
if(x>>1 > parent_xc->x){
parent_xc++;
}
if(x>>1 == parent_xc->x){
p= parent_xc->coeff;
}
}
if(/*ll|*/l|lt|t|rt|p){
int context= av_log2(/*FFABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1));
v=get_rac(&s->c, &b->state[0][context]);
if(v){
v= 2*(get_symbol2(&s->c, b->state[context + 2], context-4) + 1);
v+=get_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l&0xFF] + 3*quant3bA[t&0xFF]]);
xc->x=x;
(xc++)->coeff= v;
}
}else{
if(!run){
if(runs-- > 0) run= get_symbol2(&s->c, b->state[1], 3);
else run= INT_MAX;
v= 2*(get_symbol2(&s->c, b->state[0 + 2], 0-4) + 1);
v+=get_rac(&s->c, &b->state[0][16 + 1 + 3]);
xc->x=x;
(xc++)->coeff= v;
}else{
int max_run;
run--;
v=0;
if(y) max_run= FFMIN(run, prev_xc->x - x - 2);
else max_run= FFMIN(run, w-x-1);
if(parent_xc)
max_run= FFMIN(max_run, 2*parent_xc->x - x - 1);
x+= max_run;
run-= max_run;
}
}
}
(xc++)->x= w+1; //end marker
prev_xc= prev2_xc;
prev2_xc= xc;
if(parent_xc){
if(y&1){
while(parent_xc->x != parent->width+1)
parent_xc++;
parent_xc++;
prev_parent_xc= parent_xc;
}else{
parent_xc= prev_parent_xc;
}
}
}
(xc++)->x= w+1; //end marker
}
#endif /* AVCODEC_SNOW_H */

132
libavcodec/snowdata.h Normal file
View File

@ -0,0 +1,132 @@
/*
* Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
* Copyright (C) 2006 Robert Edele <yartrebo@earthlink.net>
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_SNOWDATA_H
#define AVCODEC_SNOWDATA_H
#include "snow.h"
static const uint8_t obmc32[1024]={
0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0,
0, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 16, 20, 20, 20, 24, 24, 20, 20, 20, 16, 16, 16, 12, 12, 8, 8, 8, 4, 4, 4, 0,
0, 4, 8, 8, 12, 12, 16, 20, 20, 24, 28, 28, 32, 32, 36, 40, 40, 36, 32, 32, 28, 28, 24, 20, 20, 16, 12, 12, 8, 8, 4, 0,
0, 4, 8, 12, 16, 20, 24, 28, 28, 32, 36, 40, 44, 48, 52, 56, 56, 52, 48, 44, 40, 36, 32, 28, 28, 24, 20, 16, 12, 8, 4, 0,
4, 8, 12, 16, 20, 24, 28, 32, 40, 44, 48, 52, 56, 60, 64, 68, 68, 64, 60, 56, 52, 48, 44, 40, 32, 28, 24, 20, 16, 12, 8, 4,
4, 8, 12, 20, 24, 32, 36, 40, 48, 52, 56, 64, 68, 76, 80, 84, 84, 80, 76, 68, 64, 56, 52, 48, 40, 36, 32, 24, 20, 12, 8, 4,
4, 8, 16, 24, 28, 36, 44, 48, 56, 60, 68, 76, 80, 88, 96,100,100, 96, 88, 80, 76, 68, 60, 56, 48, 44, 36, 28, 24, 16, 8, 4,
4, 12, 20, 28, 32, 40, 48, 56, 64, 72, 80, 88, 92,100,108,116,116,108,100, 92, 88, 80, 72, 64, 56, 48, 40, 32, 28, 20, 12, 4,
4, 12, 20, 28, 40, 48, 56, 64, 72, 80, 88, 96,108,116,124,132,132,124,116,108, 96, 88, 80, 72, 64, 56, 48, 40, 28, 20, 12, 4,
4, 16, 24, 32, 44, 52, 60, 72, 80, 92,100,108,120,128,136,148,148,136,128,120,108,100, 92, 80, 72, 60, 52, 44, 32, 24, 16, 4,
4, 16, 28, 36, 48, 56, 68, 80, 88,100,112,120,132,140,152,164,164,152,140,132,120,112,100, 88, 80, 68, 56, 48, 36, 28, 16, 4,
4, 16, 28, 40, 52, 64, 76, 88, 96,108,120,132,144,156,168,180,180,168,156,144,132,120,108, 96, 88, 76, 64, 52, 40, 28, 16, 4,
8, 20, 32, 44, 56, 68, 80, 92,108,120,132,144,156,168,180,192,192,180,168,156,144,132,120,108, 92, 80, 68, 56, 44, 32, 20, 8,
8, 20, 32, 48, 60, 76, 88,100,116,128,140,156,168,184,196,208,208,196,184,168,156,140,128,116,100, 88, 76, 60, 48, 32, 20, 8,
8, 20, 36, 52, 64, 80, 96,108,124,136,152,168,180,196,212,224,224,212,196,180,168,152,136,124,108, 96, 80, 64, 52, 36, 20, 8,
8, 24, 40, 56, 68, 84,100,116,132,148,164,180,192,208,224,240,240,224,208,192,180,164,148,132,116,100, 84, 68, 56, 40, 24, 8,
8, 24, 40, 56, 68, 84,100,116,132,148,164,180,192,208,224,240,240,224,208,192,180,164,148,132,116,100, 84, 68, 56, 40, 24, 8,
8, 20, 36, 52, 64, 80, 96,108,124,136,152,168,180,196,212,224,224,212,196,180,168,152,136,124,108, 96, 80, 64, 52, 36, 20, 8,
8, 20, 32, 48, 60, 76, 88,100,116,128,140,156,168,184,196,208,208,196,184,168,156,140,128,116,100, 88, 76, 60, 48, 32, 20, 8,
8, 20, 32, 44, 56, 68, 80, 92,108,120,132,144,156,168,180,192,192,180,168,156,144,132,120,108, 92, 80, 68, 56, 44, 32, 20, 8,
4, 16, 28, 40, 52, 64, 76, 88, 96,108,120,132,144,156,168,180,180,168,156,144,132,120,108, 96, 88, 76, 64, 52, 40, 28, 16, 4,
4, 16, 28, 36, 48, 56, 68, 80, 88,100,112,120,132,140,152,164,164,152,140,132,120,112,100, 88, 80, 68, 56, 48, 36, 28, 16, 4,
4, 16, 24, 32, 44, 52, 60, 72, 80, 92,100,108,120,128,136,148,148,136,128,120,108,100, 92, 80, 72, 60, 52, 44, 32, 24, 16, 4,
4, 12, 20, 28, 40, 48, 56, 64, 72, 80, 88, 96,108,116,124,132,132,124,116,108, 96, 88, 80, 72, 64, 56, 48, 40, 28, 20, 12, 4,
4, 12, 20, 28, 32, 40, 48, 56, 64, 72, 80, 88, 92,100,108,116,116,108,100, 92, 88, 80, 72, 64, 56, 48, 40, 32, 28, 20, 12, 4,
4, 8, 16, 24, 28, 36, 44, 48, 56, 60, 68, 76, 80, 88, 96,100,100, 96, 88, 80, 76, 68, 60, 56, 48, 44, 36, 28, 24, 16, 8, 4,
4, 8, 12, 20, 24, 32, 36, 40, 48, 52, 56, 64, 68, 76, 80, 84, 84, 80, 76, 68, 64, 56, 52, 48, 40, 36, 32, 24, 20, 12, 8, 4,
4, 8, 12, 16, 20, 24, 28, 32, 40, 44, 48, 52, 56, 60, 64, 68, 68, 64, 60, 56, 52, 48, 44, 40, 32, 28, 24, 20, 16, 12, 8, 4,
0, 4, 8, 12, 16, 20, 24, 28, 28, 32, 36, 40, 44, 48, 52, 56, 56, 52, 48, 44, 40, 36, 32, 28, 28, 24, 20, 16, 12, 8, 4, 0,
0, 4, 8, 8, 12, 12, 16, 20, 20, 24, 28, 28, 32, 32, 36, 40, 40, 36, 32, 32, 28, 28, 24, 20, 20, 16, 12, 12, 8, 8, 4, 0,
0, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 16, 20, 20, 20, 24, 24, 20, 20, 20, 16, 16, 16, 12, 12, 8, 8, 8, 4, 4, 4, 0,
0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0,
//error:0.000020
};
static const uint8_t obmc16[256]={
0, 4, 4, 8, 8, 12, 12, 16, 16, 12, 12, 8, 8, 4, 4, 0,
4, 8, 16, 20, 28, 32, 40, 44, 44, 40, 32, 28, 20, 16, 8, 4,
4, 16, 24, 36, 44, 56, 64, 76, 76, 64, 56, 44, 36, 24, 16, 4,
8, 20, 36, 48, 64, 76, 92,104,104, 92, 76, 64, 48, 36, 20, 8,
8, 28, 44, 64, 80,100,116,136,136,116,100, 80, 64, 44, 28, 8,
12, 32, 56, 76,100,120,144,164,164,144,120,100, 76, 56, 32, 12,
12, 40, 64, 92,116,144,168,196,196,168,144,116, 92, 64, 40, 12,
16, 44, 76,104,136,164,196,224,224,196,164,136,104, 76, 44, 16,
16, 44, 76,104,136,164,196,224,224,196,164,136,104, 76, 44, 16,
12, 40, 64, 92,116,144,168,196,196,168,144,116, 92, 64, 40, 12,
12, 32, 56, 76,100,120,144,164,164,144,120,100, 76, 56, 32, 12,
8, 28, 44, 64, 80,100,116,136,136,116,100, 80, 64, 44, 28, 8,
8, 20, 36, 48, 64, 76, 92,104,104, 92, 76, 64, 48, 36, 20, 8,
4, 16, 24, 36, 44, 56, 64, 76, 76, 64, 56, 44, 36, 24, 16, 4,
4, 8, 16, 20, 28, 32, 40, 44, 44, 40, 32, 28, 20, 16, 8, 4,
0, 4, 4, 8, 8, 12, 12, 16, 16, 12, 12, 8, 8, 4, 4, 0,
//error:0.000015
};
//linear *64
static const uint8_t obmc8[64]={
4, 12, 20, 28, 28, 20, 12, 4,
12, 36, 60, 84, 84, 60, 36, 12,
20, 60,100,140,140,100, 60, 20,
28, 84,140,196,196,140, 84, 28,
28, 84,140,196,196,140, 84, 28,
20, 60,100,140,140,100, 60, 20,
12, 36, 60, 84, 84, 60, 36, 12,
4, 12, 20, 28, 28, 20, 12, 4,
//error:0.000000
};
//linear *64
static const uint8_t obmc4[16]={
16, 48, 48, 16,
48,144,144, 48,
48,144,144, 48,
16, 48, 48, 16,
//error:0.000000
};
const int8_t quant3bA[256]={
0, 0, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
};
const uint8_t * const obmc_tab[4]= {
obmc32, obmc16, obmc8, obmc4
};
/* runtime generated tables */
uint8_t qexp[QROOT];
int scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES];
#endif /* AVCODEC_SNOW_H */

546
libavcodec/snowdec.c Normal file
View File

@ -0,0 +1,546 @@
/*
* Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/intmath.h"
#include "libavutil/log.h"
#include "libavutil/opt.h"
#include "avcodec.h"
#include "dsputil.h"
#include "dwt.h"
#include "snow.h"
#include "rangecoder.h"
#include "mathops.h"
#include "mpegvideo.h"
#include "h263.h"
#undef NDEBUG
#include <assert.h>
static av_always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * sb, IDWTELEM * old_buffer, int plane_index, int add, int mb_y){
Plane *p= &s->plane[plane_index];
const int mb_w= s->b_width << s->block_max_depth;
const int mb_h= s->b_height << s->block_max_depth;
int x, y, mb_x;
int block_size = MB_SIZE >> s->block_max_depth;
int block_w = plane_index ? block_size/2 : block_size;
const uint8_t *obmc = plane_index ? obmc_tab[s->block_max_depth+1] : obmc_tab[s->block_max_depth];
int obmc_stride= plane_index ? block_size : 2*block_size;
int ref_stride= s->current_picture.linesize[plane_index];
uint8_t *dst8= s->current_picture.data[plane_index];
int w= p->width;
int h= p->height;
if(s->keyframe || (s->avctx->debug&512)){
if(mb_y==mb_h)
return;
if(add){
for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
// DWTELEM * line = slice_buffer_get_line(sb, y);
IDWTELEM * line = sb->line[y];
for(x=0; x<w; x++){
// int v= buf[x + y*w] + (128<<FRAC_BITS) + (1<<(FRAC_BITS-1));
int v= line[x] + (128<<FRAC_BITS) + (1<<(FRAC_BITS-1));
v >>= FRAC_BITS;
if(v&(~255)) v= ~(v>>31);
dst8[x + y*ref_stride]= v;
}
}
}else{
for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
// DWTELEM * line = slice_buffer_get_line(sb, y);
IDWTELEM * line = sb->line[y];
for(x=0; x<w; x++){
line[x] -= 128 << FRAC_BITS;
// buf[x + y*w]-= 128<<FRAC_BITS;
}
}
}
return;
}
for(mb_x=0; mb_x<=mb_w; mb_x++){
add_yblock(s, 1, sb, old_buffer, dst8, obmc,
block_w*mb_x - block_w/2,
block_w*mb_y - block_w/2,
block_w, block_w,
w, h,
w, ref_stride, obmc_stride,
mb_x - 1, mb_y - 1,
add, 0, plane_index);
}
}
static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, slice_buffer * sb, int start_y, int h, int save_state[1]){
const int w= b->width;
int y;
const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
int new_index = 0;
if(b->ibuf == s->spatial_idwt_buffer || s->qlog == LOSSLESS_QLOG){
qadd= 0;
qmul= 1<<QEXPSHIFT;
}
/* If we are on the second or later slice, restore our index. */
if (start_y != 0)
new_index = save_state[0];
for(y=start_y; y<h; y++){
int x = 0;
int v;
IDWTELEM * line = slice_buffer_get_line(sb, y * b->stride_line + b->buf_y_offset) + b->buf_x_offset;
memset(line, 0, b->width*sizeof(IDWTELEM));
v = b->x_coeff[new_index].coeff;
x = b->x_coeff[new_index++].x;
while(x < w){
register int t= ( (v>>1)*qmul + qadd)>>QEXPSHIFT;
register int u= -(v&1);
line[x] = (t^u) - u;
v = b->x_coeff[new_index].coeff;
x = b->x_coeff[new_index++].x;
}
}
/* Save our variables for the next slice. */
save_state[0] = new_index;
return;
}
static void decode_q_branch(SnowContext *s, int level, int x, int y){
const int w= s->b_width << s->block_max_depth;
const int rem_depth= s->block_max_depth - level;
const int index= (x + y*w) << rem_depth;
int trx= (x+1)<<rem_depth;
const BlockNode *left = x ? &s->block[index-1] : &null_block;
const BlockNode *top = y ? &s->block[index-w] : &null_block;
const BlockNode *tl = y && x ? &s->block[index-w-1] : left;
const BlockNode *tr = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
if(s->keyframe){
set_blocks(s, level, x, y, null_block.color[0], null_block.color[1], null_block.color[2], null_block.mx, null_block.my, null_block.ref, BLOCK_INTRA);
return;
}
if(level==s->block_max_depth || get_rac(&s->c, &s->block_state[4 + s_context])){
int type, mx, my;
int l = left->color[0];
int cb= left->color[1];
int cr= left->color[2];
int ref = 0;
int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 0*av_log2(2*FFABS(tr->mx - top->mx));
int my_context= av_log2(2*FFABS(left->my - top->my)) + 0*av_log2(2*FFABS(tr->my - top->my));
type= get_rac(&s->c, &s->block_state[1 + left->type + top->type]) ? BLOCK_INTRA : 0;
if(type){
pred_mv(s, &mx, &my, 0, left, top, tr);
l += get_symbol(&s->c, &s->block_state[32], 1);
cb+= get_symbol(&s->c, &s->block_state[64], 1);
cr+= get_symbol(&s->c, &s->block_state[96], 1);
}else{
if(s->ref_frames > 1)
ref= get_symbol(&s->c, &s->block_state[128 + 1024 + 32*ref_context], 0);
pred_mv(s, &mx, &my, ref, left, top, tr);
mx+= get_symbol(&s->c, &s->block_state[128 + 32*(mx_context + 16*!!ref)], 1);
my+= get_symbol(&s->c, &s->block_state[128 + 32*(my_context + 16*!!ref)], 1);
}
set_blocks(s, level, x, y, l, cb, cr, mx, my, ref, type);
}else{
decode_q_branch(s, level+1, 2*x+0, 2*y+0);
decode_q_branch(s, level+1, 2*x+1, 2*y+0);
decode_q_branch(s, level+1, 2*x+0, 2*y+1);
decode_q_branch(s, level+1, 2*x+1, 2*y+1);
}
}
static void dequantize_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int start_y, int end_y){
const int w= b->width;
const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
int x,y;
if(s->qlog == LOSSLESS_QLOG) return;
for(y=start_y; y<end_y; y++){
// DWTELEM * line = slice_buffer_get_line_from_address(sb, src + (y * stride));
IDWTELEM * line = slice_buffer_get_line(sb, (y * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
for(x=0; x<w; x++){
int i= line[x];
if(i<0){
line[x]= -((-i*qmul + qadd)>>(QEXPSHIFT)); //FIXME try different bias
}else if(i>0){
line[x]= (( i*qmul + qadd)>>(QEXPSHIFT));
}
}
}
}
static void correlate_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median, int start_y, int end_y){
const int w= b->width;
int x,y;
IDWTELEM * line=0; // silence silly "could be used without having been initialized" warning
IDWTELEM * prev;
if (start_y != 0)
line = slice_buffer_get_line(sb, ((start_y - 1) * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
for(y=start_y; y<end_y; y++){
prev = line;
// line = slice_buffer_get_line_from_address(sb, src + (y * stride));
line = slice_buffer_get_line(sb, (y * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
for(x=0; x<w; x++){
if(x){
if(use_median){
if(y && x+1<w) line[x] += mid_pred(line[x - 1], prev[x], prev[x + 1]);
else line[x] += line[x - 1];
}else{
if(y) line[x] += mid_pred(line[x - 1], prev[x], line[x - 1] + prev[x] - prev[x - 1]);
else line[x] += line[x - 1];
}
}else{
if(y) line[x] += prev[x];
}
}
}
}
static void decode_qlogs(SnowContext *s){
int plane_index, level, orientation;
for(plane_index=0; plane_index<3; plane_index++){
for(level=0; level<s->spatial_decomposition_count; level++){
for(orientation=level ? 1:0; orientation<4; orientation++){
int q;
if (plane_index==2) q= s->plane[1].band[level][orientation].qlog;
else if(orientation==2) q= s->plane[plane_index].band[level][1].qlog;
else q= get_symbol(&s->c, s->header_state, 1);
s->plane[plane_index].band[level][orientation].qlog= q;
}
}
}
}
#define GET_S(dst, check) \
tmp= get_symbol(&s->c, s->header_state, 0);\
if(!(check)){\
av_log(s->avctx, AV_LOG_ERROR, "Error " #dst " is %d\n", tmp);\
return -1;\
}\
dst= tmp;
static int decode_header(SnowContext *s){
int plane_index, tmp;
uint8_t kstate[32];
memset(kstate, MID_STATE, sizeof(kstate));
s->keyframe= get_rac(&s->c, kstate);
if(s->keyframe || s->always_reset){
ff_snow_reset_contexts(s);
s->spatial_decomposition_type=
s->qlog=
s->qbias=
s->mv_scale=
s->block_max_depth= 0;
}
if(s->keyframe){
GET_S(s->version, tmp <= 0U)
s->always_reset= get_rac(&s->c, s->header_state);
s->temporal_decomposition_type= get_symbol(&s->c, s->header_state, 0);
s->temporal_decomposition_count= get_symbol(&s->c, s->header_state, 0);
GET_S(s->spatial_decomposition_count, 0 < tmp && tmp <= MAX_DECOMPOSITIONS)
s->colorspace_type= get_symbol(&s->c, s->header_state, 0);
s->chroma_h_shift= get_symbol(&s->c, s->header_state, 0);
s->chroma_v_shift= get_symbol(&s->c, s->header_state, 0);
s->spatial_scalability= get_rac(&s->c, s->header_state);
// s->rate_scalability= get_rac(&s->c, s->header_state);
GET_S(s->max_ref_frames, tmp < (unsigned)MAX_REF_FRAMES)
s->max_ref_frames++;
decode_qlogs(s);
}
if(!s->keyframe){
if(get_rac(&s->c, s->header_state)){
for(plane_index=0; plane_index<2; plane_index++){
int htaps, i, sum=0;
Plane *p= &s->plane[plane_index];
p->diag_mc= get_rac(&s->c, s->header_state);
htaps= get_symbol(&s->c, s->header_state, 0)*2 + 2;
if((unsigned)htaps > HTAPS_MAX || htaps==0)
return -1;
p->htaps= htaps;
for(i= htaps/2; i; i--){
p->hcoeff[i]= get_symbol(&s->c, s->header_state, 0) * (1-2*(i&1));
sum += p->hcoeff[i];
}
p->hcoeff[0]= 32-sum;
}
s->plane[2].diag_mc= s->plane[1].diag_mc;
s->plane[2].htaps = s->plane[1].htaps;
memcpy(s->plane[2].hcoeff, s->plane[1].hcoeff, sizeof(s->plane[1].hcoeff));
}
if(get_rac(&s->c, s->header_state)){
GET_S(s->spatial_decomposition_count, 0 < tmp && tmp <= MAX_DECOMPOSITIONS)
decode_qlogs(s);
}
}
s->spatial_decomposition_type+= get_symbol(&s->c, s->header_state, 1);
if(s->spatial_decomposition_type > 1U){
av_log(s->avctx, AV_LOG_ERROR, "spatial_decomposition_type %d not supported", s->spatial_decomposition_type);
return -1;
}
if(FFMIN(s->avctx-> width>>s->chroma_h_shift,
s->avctx->height>>s->chroma_v_shift) >> (s->spatial_decomposition_count-1) <= 0){
av_log(s->avctx, AV_LOG_ERROR, "spatial_decomposition_count %d too large for size", s->spatial_decomposition_count);
return -1;
}
s->qlog += get_symbol(&s->c, s->header_state, 1);
s->mv_scale += get_symbol(&s->c, s->header_state, 1);
s->qbias += get_symbol(&s->c, s->header_state, 1);
s->block_max_depth+= get_symbol(&s->c, s->header_state, 1);
if(s->block_max_depth > 1 || s->block_max_depth < 0){
av_log(s->avctx, AV_LOG_ERROR, "block_max_depth= %d is too large", s->block_max_depth);
s->block_max_depth= 0;
return -1;
}
return 0;
}
static av_cold int decode_init(AVCodecContext *avctx)
{
avctx->pix_fmt= PIX_FMT_YUV420P;
ff_snow_common_init(avctx);
return 0;
}
static void decode_blocks(SnowContext *s){
int x, y;
int w= s->b_width;
int h= s->b_height;
for(y=0; y<h; y++){
for(x=0; x<w; x++){
decode_q_branch(s, 0, x, y);
}
}
}
static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt){
const uint8_t *buf = avpkt->data;
int buf_size = avpkt->size;
SnowContext *s = avctx->priv_data;
RangeCoder * const c= &s->c;
int bytes_read;
AVFrame *picture = data;
int level, orientation, plane_index;
ff_init_range_decoder(c, buf, buf_size);
ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
s->current_picture.pict_type= AV_PICTURE_TYPE_I; //FIXME I vs. P
if(decode_header(s)<0)
return -1;
ff_snow_common_init_after_header(avctx);
// realloc slice buffer for the case that spatial_decomposition_count changed
ff_slice_buffer_destroy(&s->sb);
ff_slice_buffer_init(&s->sb, s->plane[0].height, (MB_SIZE >> s->block_max_depth) + s->spatial_decomposition_count * 8 + 1, s->plane[0].width, s->spatial_idwt_buffer);
for(plane_index=0; plane_index<3; plane_index++){
Plane *p= &s->plane[plane_index];
p->fast_mc= p->diag_mc && p->htaps==6 && p->hcoeff[0]==40
&& p->hcoeff[1]==-10
&& p->hcoeff[2]==2;
}
ff_snow_alloc_blocks(s);
if(ff_snow_frame_start(s) < 0)
return -1;
//keyframe flag duplication mess FIXME
if(avctx->debug&FF_DEBUG_PICT_INFO)
av_log(avctx, AV_LOG_ERROR, "keyframe:%d qlog:%d\n", s->keyframe, s->qlog);
decode_blocks(s);
for(plane_index=0; plane_index<3; plane_index++){
Plane *p= &s->plane[plane_index];
int w= p->width;
int h= p->height;
int x, y;
int decode_state[MAX_DECOMPOSITIONS][4][1]; /* Stored state info for unpack_coeffs. 1 variable per instance. */
if(s->avctx->debug&2048){
memset(s->spatial_dwt_buffer, 0, sizeof(DWTELEM)*w*h);
predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
for(y=0; y<h; y++){
for(x=0; x<w; x++){
int v= s->current_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x];
s->mconly_picture.data[plane_index][y*s->mconly_picture.linesize[plane_index] + x]= v;
}
}
}
{
for(level=0; level<s->spatial_decomposition_count; level++){
for(orientation=level ? 1 : 0; orientation<4; orientation++){
SubBand *b= &p->band[level][orientation];
unpack_coeffs(s, b, b->parent, orientation);
}
}
}
{
const int mb_h= s->b_height << s->block_max_depth;
const int block_size = MB_SIZE >> s->block_max_depth;
const int block_w = plane_index ? block_size/2 : block_size;
int mb_y;
DWTCompose cs[MAX_DECOMPOSITIONS];
int yd=0, yq=0;
int y;
int end_y;
ff_spatial_idwt_buffered_init(cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count);
for(mb_y=0; mb_y<=mb_h; mb_y++){
int slice_starty = block_w*mb_y;
int slice_h = block_w*(mb_y+1);
if (!(s->keyframe || s->avctx->debug&512)){
slice_starty = FFMAX(0, slice_starty - (block_w >> 1));
slice_h -= (block_w >> 1);
}
for(level=0; level<s->spatial_decomposition_count; level++){
for(orientation=level ? 1 : 0; orientation<4; orientation++){
SubBand *b= &p->band[level][orientation];
int start_y;
int end_y;
int our_mb_start = mb_y;
int our_mb_end = (mb_y + 1);
const int extra= 3;
start_y = (mb_y ? ((block_w * our_mb_start) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra: 0);
end_y = (((block_w * our_mb_end) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra);
if (!(s->keyframe || s->avctx->debug&512)){
start_y = FFMAX(0, start_y - (block_w >> (1+s->spatial_decomposition_count - level)));
end_y = FFMAX(0, end_y - (block_w >> (1+s->spatial_decomposition_count - level)));
}
start_y = FFMIN(b->height, start_y);
end_y = FFMIN(b->height, end_y);
if (start_y != end_y){
if (orientation == 0){
SubBand * correlate_band = &p->band[0][0];
int correlate_end_y = FFMIN(b->height, end_y + 1);
int correlate_start_y = FFMIN(b->height, (start_y ? start_y + 1 : 0));
decode_subband_slice_buffered(s, correlate_band, &s->sb, correlate_start_y, correlate_end_y, decode_state[0][0]);
correlate_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, 1, 0, correlate_start_y, correlate_end_y);
dequantize_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, start_y, end_y);
}
else
decode_subband_slice_buffered(s, b, &s->sb, start_y, end_y, decode_state[level][orientation]);
}
}
}
for(; yd<slice_h; yd+=4){
ff_spatial_idwt_buffered_slice(&s->dwt, cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count, yd);
}
if(s->qlog == LOSSLESS_QLOG){
for(; yq<slice_h && yq<h; yq++){
IDWTELEM * line = slice_buffer_get_line(&s->sb, yq);
for(x=0; x<w; x++){
line[x] <<= FRAC_BITS;
}
}
}
predict_slice_buffered(s, &s->sb, s->spatial_idwt_buffer, plane_index, 1, mb_y);
y = FFMIN(p->height, slice_starty);
end_y = FFMIN(p->height, slice_h);
while(y < end_y)
ff_slice_buffer_release(&s->sb, y++);
}
ff_slice_buffer_flush(&s->sb);
}
}
emms_c();
ff_snow_release_buffer(avctx);
if(!(s->avctx->debug&2048))
*picture= s->current_picture;
else
*picture= s->mconly_picture;
*data_size = sizeof(AVFrame);
bytes_read= c->bytestream - c->bytestream_start;
if(bytes_read ==0) av_log(s->avctx, AV_LOG_ERROR, "error at end of frame\n"); //FIXME
return bytes_read;
}
static av_cold int decode_end(AVCodecContext *avctx)
{
SnowContext *s = avctx->priv_data;
ff_slice_buffer_destroy(&s->sb);
ff_snow_common_end(s);
return 0;
}
AVCodec ff_snow_decoder = {
.name = "snow",
.type = AVMEDIA_TYPE_VIDEO,
.id = CODEC_ID_SNOW,
.priv_data_size = sizeof(SnowContext),
.init = decode_init,
.close = decode_end,
.decode = decode_frame,
.capabilities = CODEC_CAP_DR1 /*| CODEC_CAP_DRAW_HORIZ_BAND*/,
.long_name = NULL_IF_CONFIG_SMALL("Snow"),
};

1915
libavcodec/snowenc.c Normal file

File diff suppressed because it is too large Load Diff