#include "h264data.h"
#include "golomb.h"
+#include "cabac.h"
+
#undef NDEBUG
#include <assert.h>
int profile_idc;
int level_idc;
- int multiple_slice_groups; ///< more_than_one_slice_group_allowed_flag
- int arbitrary_slice_order; ///< arbitrary_slice_order_allowed_flag
- int redundant_slices; ///< redundant_slices_allowed_flag
int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
int poc_type; ///< pic_order_cnt_type
int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
int offset_for_top_to_bottom_field;
int poc_cycle_length; ///< num_ref_frames_in_pic_order_cnt_cycle
int ref_frame_count; ///< num_ref_frames
- int required_frame_num_update_behaviour_flag;
+ int gaps_in_frame_num_allowed_flag;
int mb_width; ///< frame_width_in_mbs_minus1 + 1
int mb_height; ///< frame_height_in_mbs_minus1 + 1
int frame_mbs_only_flag;
int mb_aff; ///<mb_adaptive_frame_field_flag
int direct_8x8_inference_flag;
+ int crop; ///< frame_cropping_flag
+ int crop_left; ///< frame_cropping_rect_left_offset
+ int crop_right; ///< frame_cropping_rect_right_offset
+ int crop_top; ///< frame_cropping_rect_top_offset
+ int crop_bottom; ///< frame_cropping_rect_bottom_offset
int vui_parameters_present_flag;
- int sar_width;
- int sar_height;
+ AVRational sar;
+ int timing_info_present_flag;
+ uint32_t num_units_in_tick;
+ uint32_t time_scale;
+ int fixed_frame_rate_flag;
short offset_for_ref_frame[256]; //FIXME dyn aloc?
}SPS;
int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
int constrained_intra_pred; ///< constrained_intra_pred_flag
int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
- int crop; ///< frame_cropping_flag
- int crop_left; ///< frame_cropping_rect_left_offset
- int crop_right; ///< frame_cropping_rect_right_offset
- int crop_top; ///< frame_cropping_rect_top_offset
- int crop_bottom; ///< frame_cropping_rect_bottom_offset
}PPS;
/**
uint8_t *rbsp_buffer;
int rbsp_buffer_size;
- int mb_stride; ///< stride of some mb tables
+ /**
+ * Used to parse AVC variant of h264
+ */
+ int is_avc; ///< this flag is != 0 if codec is avc1
+ int got_avcC; ///< flag used to parse avcC data only once
+ int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
int chroma_qp; //QPc
unsigned int top_samples_available;
unsigned int topright_samples_available;
unsigned int left_samples_available;
+ uint8_t (*top_border)[16+2*8];
+ uint8_t left_border[17+2*9];
/**
* non zero coeff count cache.
uint16_t *mb2b_xy; //FIXME are these 4 a good idea?
uint16_t *mb2b8_xy;
- int b_stride;
+ int b_stride; //FIXME use s->b4_stride
int b8_stride;
+ int halfpel_flag;
+ int thirdpel_flag;
+
+ int unknown_svq3_flag;
+ int next_slice_index;
+
SPS sps_buffer[MAX_SPS_COUNT];
SPS sps; ///< current sps
int chroma_offset[2][16][2];
//deblock
- int disable_deblocking_filter_idc;
- int slice_alpha_c0_offset_div2;
- int slice_beta_offset_div2;
+ int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
+ int slice_alpha_c0_offset;
+ int slice_beta_offset;
int redundant_pic_count;
int direct_spatial_mv_pred;
+ int dist_scale_factor[16];
/**
* num_ref_idx_l0/1_active_minus1 + 1
GetBitContext *inter_gb_ptr;
DCTELEM mb[16*24] __align8;
+
+ /**
+ * Cabac
+ */
+ CABACContext cabac;
+ uint8_t cabac_state[399];
+ int cabac_init_idc;
+
+ /* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
+ uint16_t *cbp_table;
+ int top_cbp;
+ int left_cbp;
+ /* chroma_pred_mode for i4x4 or i16x16, else 0 */
+ uint8_t *chroma_pred_mode_table;
+ int last_qscale_diff;
+ int16_t (*mvd_table[2])[2];
+ int16_t mvd_cache[2][5*8][2];
+ uint8_t *direct_table;
+ uint8_t direct_cache[5*8];
+
}H264Context;
static VLC coeff_token_vlc[4];
static VLC run_vlc[6];
static VLC run7_vlc;
+static void svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
+static void svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
+static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr);
+
+static inline uint32_t pack16to32(int a, int b){
+#ifdef WORDS_BIGENDIAN
+ return (b&0xFFFF) + (a<<16);
+#else
+ return (a&0xFFFF) + (b<<16);
+#endif
+}
+
/**
* fill a rectangle.
- * @param h height of the recatangle, should be a constant
- * @param w width of the recatangle, should be a constant
+ * @param h height of the rectangle, should be a constant
+ * @param w width of the rectangle, should be a constant
* @param size the size of val (1 or 4), should be a constant
*/
-static inline void fill_rectangle(void *p, int w, int h, int stride, uint32_t val, int size){ //FIXME ensure this IS inlined
+static inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){ //FIXME ensure this IS inlined
+ uint8_t *p= (uint8_t*)vp;
assert(size==1 || size==4);
w *= size;
*(uint16_t*)(p + 1*stride)=
*(uint16_t*)(p + 2*stride)=
*(uint16_t*)(p + 3*stride)= size==4 ? val : val*0x0101;
+ }else if(w==4 && h==1){
+ *(uint32_t*)(p + 0*stride)= size==4 ? val : val*0x01010101;
}else if(w==4 && h==2){
*(uint32_t*)(p + 0*stride)=
*(uint32_t*)(p + 1*stride)= size==4 ? val : val*0x01010101;
static inline void fill_caches(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int topleft_xy, top_xy, topright_xy, left_xy[2];
int topleft_type, top_type, topright_type, left_type[2];
int left_block[4];
if(h->sps.mb_aff){
//FIXME
+ topleft_xy = 0; /* avoid warning */
+ top_xy = 0; /* avoid warning */
+ topright_xy = 0; /* avoid warning */
}else{
- topleft_xy = mb_xy-1 - h->mb_stride;
- top_xy = mb_xy - h->mb_stride;
- topright_xy= mb_xy+1 - h->mb_stride;
+ topleft_xy = mb_xy-1 - s->mb_stride;
+ top_xy = mb_xy - s->mb_stride;
+ topright_xy= mb_xy+1 - s->mb_stride;
left_xy[0] = mb_xy-1;
left_xy[1] = mb_xy-1;
left_block[0]= 0;
h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][10];
h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][11];
+
+ h->top_cbp= h->cbp_table[top_xy];
}else{
h->non_zero_count_cache[4+8*0]=
h->non_zero_count_cache[5+8*0]=
h->non_zero_count_cache[2+8*0]=
h->non_zero_count_cache[1+8*3]=
- h->non_zero_count_cache[2+8*3]= 64;
+ h->non_zero_count_cache[2+8*3]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+
+ if(IS_INTRA(mb_type)) h->top_cbp= 0x1C0;
+ else h->top_cbp= 0;
}
if(left_type[0]){
h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][5];
h->non_zero_count_cache[0+8*1]= h->non_zero_count[left_xy[0]][9]; //FIXME left_block
h->non_zero_count_cache[0+8*4]= h->non_zero_count[left_xy[0]][12];
+ h->left_cbp= h->cbp_table[left_xy[0]]; //FIXME interlacing
}else{
h->non_zero_count_cache[3+8*1]=
h->non_zero_count_cache[3+8*2]=
h->non_zero_count_cache[0+8*1]=
- h->non_zero_count_cache[0+8*4]= 64;
+ h->non_zero_count_cache[0+8*4]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+
+ if(IS_INTRA(mb_type)) h->left_cbp= 0x1C0;//FIXME interlacing
+ else h->left_cbp= 0;
}
if(left_type[1]){
h->non_zero_count_cache[3+8*3]=
h->non_zero_count_cache[3+8*4]=
h->non_zero_count_cache[0+8*2]=
- h->non_zero_count_cache[0+8*5]= 64;
+ h->non_zero_count_cache[0+8*5]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
}
#if 1
- if(IS_INTER(mb_type)){
+ //FIXME direct mb can skip much of this
+ if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
int list;
for(list=0; list<2; list++){
- if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list)){
+ if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list) && !IS_DIRECT(mb_type)){
/*if(!h->mv_cache_clean[list]){
memset(h->mv_cache [list], 0, 8*5*2*sizeof(int16_t)); //FIXME clean only input? clean at all?
memset(h->ref_cache[list], PART_NOT_AVAILABLE, 8*5*sizeof(int8_t));
h->mv_cache_clean[list]= 1;
}*/
- continue; //FIXME direct mode ...
+ continue;
}
h->mv_cache_clean[list]= 0;
*(uint32_t*)h->mv_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
*(uint32_t*)h->mv_cache [list][scan8[4 ]]=
*(uint32_t*)h->mv_cache [list][scan8[12]]= 0;
- }
-//FIXME
+ if( h->pps.cabac ) {
+ /* XXX beurk, Load mvd */
+ if(IS_INTER(topleft_type)){
+ const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy];
+ }else{
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 - 1*8]= 0;
+ }
+
+ if(IS_INTER(top_type)){
+ const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
+ *(uint32_t*)h->mvd_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 0];
+ *(uint32_t*)h->mvd_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 1];
+ *(uint32_t*)h->mvd_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 2];
+ *(uint32_t*)h->mvd_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 3];
+ }else{
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 0 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 1 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 2 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 3 - 1*8]= 0;
+ }
+ if(IS_INTER(left_type[0])){
+ const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[0]];
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[1]];
+ }else{
+ *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 0*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 1*8]= 0;
+ }
+ if(IS_INTER(left_type[1])){
+ const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[2]];
+ *(uint32_t*)h->mvd_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)h->mvd_table[list][b_xy + h->b_stride*left_block[3]];
+ }else{
+ *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 2*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] - 1 + 3*8]= 0;
+ }
+ *(uint32_t*)h->mvd_cache [list][scan8[5 ]+1]=
+ *(uint32_t*)h->mvd_cache [list][scan8[7 ]+1]=
+ *(uint32_t*)h->mvd_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
+ *(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
+ *(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;
+
+ if(h->slice_type == B_TYPE){
+ fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, 0, 1);
+
+ if(IS_DIRECT(top_type)){
+ *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0x01010101;
+ }else if(IS_8X8(top_type)){
+ int b8_xy = h->mb2b8_xy[top_xy] + h->b8_stride;
+ h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy];
+ h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 1];
+ }else{
+ *(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0;
+ }
+
+ //FIXME interlacing
+ if(IS_DIRECT(left_type[0])){
+ h->direct_cache[scan8[0] - 1 + 0*8]=
+ h->direct_cache[scan8[0] - 1 + 2*8]= 1;
+ }else if(IS_8X8(left_type[0])){
+ int b8_xy = h->mb2b8_xy[left_xy[0]] + 1;
+ h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[b8_xy];
+ h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[b8_xy + h->b8_stride];
+ }else{
+ h->direct_cache[scan8[0] - 1 + 0*8]=
+ h->direct_cache[scan8[0] - 1 + 2*8]= 0;
+ }
+ }
+ }
+ }
}
#endif
}
static inline void write_back_intra_pred_mode(H264Context *h){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
h->intra4x4_pred_mode[mb_xy][0]= h->intra4x4_pred_mode_cache[7+8*1];
h->intra4x4_pred_mode[mb_xy][1]= h->intra4x4_pred_mode_cache[7+8*2];
for(i=0; i<4; i++){
int status= top[ h->intra4x4_pred_mode_cache[scan8[0] + i] ];
if(status<0){
- fprintf(stderr, "top block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
return -1;
} else if(status){
h->intra4x4_pred_mode_cache[scan8[0] + i]= status;
for(i=0; i<4; i++){
int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ];
if(status<0){
- fprintf(stderr, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
return -1;
} else if(status){
h->intra4x4_pred_mode_cache[scan8[0] + 8*i]= status;
static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
+ if(mode < 0 || mode > 6) {
+ av_log(h->s.avctx, AV_LOG_ERROR, "out of range intra chroma pred mode at %d %d\n", s->mb_x, s->mb_y);
+ return -1;
+ }
+
if(!(h->top_samples_available&0x8000)){
mode= top[ mode ];
if(mode<0){
- fprintf(stderr, "top block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "top block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
}
if(!(h->left_samples_available&0x8000)){
mode= left[ mode ];
if(mode<0){
- fprintf(stderr, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
}
static inline void write_back_non_zero_count(H264Context *h){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[4+8*4];
h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[5+8*4];
* @param my the y component of the predicted motion vector
*/
static inline void pred_motion(H264Context * const h, int n, int part_width, int list, int ref, int * const mx, int * const my){
- MpegEncContext * const s = &h->s;
const int index8= scan8[n];
const int top_ref= h->ref_cache[list][ index8 - 8 ];
const int left_ref= h->ref_cache[list][ index8 - 1 ];
diagonal_ref= fetch_diagonal_mv(h, &C, index8, list, part_width);
match_count= (diagonal_ref==ref) + (top_ref==ref) + (left_ref==ref);
-
if(match_count > 1){ //most common
*mx= mid_pred(A[0], B[0], C[0]);
*my= mid_pred(A[1], B[1], C[1]);
}
}
- tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, s->mb_x, s->mb_y, n, list);
+ tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, h->s.mb_x, h->s.mb_y, n, list);
}
/**
* @param my the y component of the predicted motion vector
*/
static inline void pred_16x8_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
- MpegEncContext * const s = &h->s;
if(n==0){
const int top_ref= h->ref_cache[list][ scan8[0] - 8 ];
const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];
- tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", top_ref, B[0], B[1], s->mb_x, s->mb_y, n, list);
+ tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
if(top_ref == ref){
*mx= B[0];
const int left_ref= h->ref_cache[list][ scan8[8] - 1 ];
const int16_t * const A= h->mv_cache[list][ scan8[8] - 1 ];
- tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], s->mb_x, s->mb_y, n, list);
+ tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
if(left_ref == ref){
*mx= A[0];
* @param my the y component of the predicted motion vector
*/
static inline void pred_8x16_motion(H264Context * const h, int n, int list, int ref, int * const mx, int * const my){
- MpegEncContext * const s = &h->s;
if(n==0){
const int left_ref= h->ref_cache[list][ scan8[0] - 1 ];
const int16_t * const A= h->mv_cache[list][ scan8[0] - 1 ];
- tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], s->mb_x, s->mb_y, n, list);
+ tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
if(left_ref == ref){
*mx= A[0];
diagonal_ref= fetch_diagonal_mv(h, &C, scan8[4], list, 2);
- tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", diagonal_ref, C[0], C[1], s->mb_x, s->mb_y, n, list);
+ tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d", diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);
if(diagonal_ref == ref){
*mx= C[0];
}
static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my){
- MpegEncContext * const s = &h->s;
const int top_ref = h->ref_cache[0][ scan8[0] - 8 ];
const int left_ref= h->ref_cache[0][ scan8[0] - 1 ];
- tprintf("pred_pskip: (%d) (%d) at %2d %2d", top_ref, left_ref, s->mb_x, s->mb_y);
+ tprintf("pred_pskip: (%d) (%d) at %2d %2d\n", top_ref, left_ref, h->s.mb_x, h->s.mb_y);
if(top_ref == PART_NOT_AVAILABLE || left_ref == PART_NOT_AVAILABLE
|| (top_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 8 ] == 0)
return;
}
+static inline void direct_dist_scale_factor(H264Context * const h){
+ const int poc = h->s.current_picture_ptr->poc;
+ const int poc1 = h->ref_list[1][0].poc;
+ int i;
+ for(i=0; i<h->ref_count[0]; i++){
+ int poc0 = h->ref_list[0][i].poc;
+ int td = clip(poc1 - poc0, -128, 127);
+ if(td == 0 /* FIXME || pic0 is a long-term ref */){
+ h->dist_scale_factor[i] = 256;
+ }else{
+ int tb = clip(poc - poc0, -128, 127);
+ int tx = (16384 + (ABS(td) >> 1)) / td;
+ h->dist_scale_factor[i] = clip((tb*tx + 32) >> 6, -1024, 1023);
+ }
+ }
+}
+
+static inline void pred_direct_motion(H264Context * const h, int *mb_type){
+ MpegEncContext * const s = &h->s;
+ const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
+ const int b8_xy = 2*s->mb_x + 2*s->mb_y*h->b8_stride;
+ const int b4_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
+ const int mb_type_col = h->ref_list[1][0].mb_type[mb_xy];
+ const int16_t (*l1mv0)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[0][b4_xy];
+ const int8_t *l1ref0 = &h->ref_list[1][0].ref_index[0][b8_xy];
+ const int is_b8x8 = IS_8X8(*mb_type);
+ int sub_mb_type;
+ int i8, i4;
+
+ if(IS_8X8(mb_type_col) && !h->sps.direct_8x8_inference_flag){
+ /* FIXME save sub mb types from previous frames (or derive from MVs)
+ * so we know exactly what block size to use */
+ sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_4x4 */
+ *mb_type = MB_TYPE_8x8;
+ }else if(!is_b8x8 && (IS_16X16(mb_type_col) || IS_INTRA(mb_type_col))){
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_16x16 */
+ }else{
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type = MB_TYPE_8x8;
+ }
+ if(!is_b8x8)
+ *mb_type |= MB_TYPE_DIRECT2;
+
+ if(h->direct_spatial_mv_pred){
+ int ref[2];
+ int mv[2][2];
+ int list;
+
+ /* ref = min(neighbors) */
+ for(list=0; list<2; list++){
+ int refa = h->ref_cache[list][scan8[0] - 1];
+ int refb = h->ref_cache[list][scan8[0] - 8];
+ int refc = h->ref_cache[list][scan8[0] - 8 + 4];
+ if(refc == -2)
+ refc = h->ref_cache[list][scan8[0] - 8 - 1];
+ ref[list] = refa;
+ if(ref[list] < 0 || (refb < ref[list] && refb >= 0))
+ ref[list] = refb;
+ if(ref[list] < 0 || (refc < ref[list] && refc >= 0))
+ ref[list] = refc;
+ if(ref[list] < 0)
+ ref[list] = -1;
+ }
+
+ if(ref[0] < 0 && ref[1] < 0){
+ ref[0] = ref[1] = 0;
+ mv[0][0] = mv[0][1] =
+ mv[1][0] = mv[1][1] = 0;
+ }else{
+ for(list=0; list<2; list++){
+ if(ref[list] >= 0)
+ pred_motion(h, 0, 4, list, ref[list], &mv[list][0], &mv[list][1]);
+ else
+ mv[list][0] = mv[list][1] = 0;
+ }
+ }
+
+ if(ref[1] < 0){
+ *mb_type &= ~MB_TYPE_P0L1;
+ sub_mb_type &= ~MB_TYPE_P0L1;
+ }else if(ref[0] < 0){
+ *mb_type &= ~MB_TYPE_P0L0;
+ sub_mb_type &= ~MB_TYPE_P0L0;
+ }
+
+ if(IS_16X16(*mb_type)){
+ fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref[0], 1);
+ fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, ref[1], 1);
+ if(!IS_INTRA(mb_type_col) && l1ref0[0] == 0 &&
+ ABS(l1mv0[0][0]) <= 1 && ABS(l1mv0[0][1]) <= 1){
+ if(ref[0] > 0)
+ fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv[0][0],mv[0][1]), 4);
+ else
+ fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
+ if(ref[1] > 0)
+ fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv[1][0],mv[1][1]), 4);
+ else
+ fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
+ }else{
+ fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv[0][0],mv[0][1]), 4);
+ fill_rectangle(&h->mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv[1][0],mv[1][1]), 4);
+ }
+ }else{
+ for(i8=0; i8<4; i8++){
+ const int x8 = i8&1;
+ const int y8 = i8>>1;
+
+ if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+ continue;
+ h->sub_mb_type[i8] = sub_mb_type;
+
+ fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mv[0][0],mv[0][1]), 4);
+ fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mv[1][0],mv[1][1]), 4);
+ fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref[0], 1);
+ fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, ref[1], 1);
+
+ /* col_zero_flag */
+ if(!IS_INTRA(mb_type_col) && l1ref0[x8 + y8*h->b8_stride] == 0){
+ for(i4=0; i4<4; i4++){
+ const int16_t *mv_col = l1mv0[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
+ if(ABS(mv_col[0]) <= 1 && ABS(mv_col[1]) <= 1){
+ if(ref[0] == 0)
+ *(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
+ if(ref[1] == 0)
+ *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] = 0;
+ }
+ }
+ }
+ }
+ }
+ }else{ /* direct temporal mv pred */
+ /* FIXME assumes that L1ref0 used the same ref lists as current frame */
+ if(IS_16X16(*mb_type)){
+ fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
+ if(IS_INTRA(mb_type_col)){
+ fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
+ fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
+ fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
+ }else{
+ const int ref0 = l1ref0[0];
+ const int dist_scale_factor = h->dist_scale_factor[ref0];
+ const int16_t *mv_col = l1mv0[0];
+ int mv_l0[2];
+ mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
+ mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
+ fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref0, 1);
+ fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv_l0[0],mv_l0[1]), 4);
+ fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]), 4);
+ }
+ }else{
+ for(i8=0; i8<4; i8++){
+ const int x8 = i8&1;
+ const int y8 = i8>>1;
+ int ref0, dist_scale_factor;
+
+ if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+ continue;
+ h->sub_mb_type[i8] = sub_mb_type;
+ if(IS_INTRA(mb_type_col)){
+ fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
+ fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+ fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+ fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+ continue;
+ }
+
+ ref0 = l1ref0[x8 + y8*h->b8_stride];
+ dist_scale_factor = h->dist_scale_factor[ref0];
+
+ fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
+ fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+ for(i4=0; i4<4; i4++){
+ const int16_t *mv_col = l1mv0[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
+ int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
+ mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
+ mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
+ *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] =
+ pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
+ }
+ }
+ }
+ }
+}
+
static inline void write_back_motion(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
const int b_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
const int b8_xy= 2*s->mb_x + 2*s->mb_y*h->b8_stride;
int list;
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]=
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= 0;
}
+ if( h->pps.cabac ) {
+ /* FIXME needed ? */
+ for(y=0; y<4; y++){
+ *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]=
+ *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= 0;
+ }
+ }
for(y=0; y<2; y++){
- *(uint16_t*)s->current_picture.motion_val[list][b8_xy + y*h->b8_stride]= (LIST_NOT_USED&0xFF)*0x0101;
+ *(uint16_t*)&s->current_picture.ref_index[list][b8_xy + y*h->b8_stride]= (LIST_NOT_USED&0xFF)*0x0101;
}
}
- continue; //FIXME direct mode ...
+ continue;
}
for(y=0; y<4; y++){
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+0 + 8*y];
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+2 + 8*y];
}
+ if( h->pps.cabac ) {
+ for(y=0; y<4; y++){
+ *(uint64_t*)h->mvd_table[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+0 + 8*y];
+ *(uint64_t*)h->mvd_table[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mvd_cache[list][scan8[0]+2 + 8*y];
+ }
+ }
for(y=0; y<2; y++){
s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+0 + 16*y];
s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+2 + 16*y];
}
}
+
+ if(h->slice_type == B_TYPE && h->pps.cabac){
+ if(IS_8X8(mb_type)){
+ h->direct_table[b8_xy+1+0*h->b8_stride] = IS_DIRECT(h->sub_mb_type[1]) ? 1 : 0;
+ h->direct_table[b8_xy+0+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[2]) ? 1 : 0;
+ h->direct_table[b8_xy+1+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[3]) ? 1 : 0;
+ }
+ }
}
/**
dst[di++]= 0;
dst[di++]= 0;
si+=3;
+ continue;
}else //next start code
break;
}
return dst;
}
+#if 0
/**
* @param src the data which should be escaped
* @param dst the target buffer, dst+1 == src is allowed as a special case
static void encode_rbsp_trailing(PutBitContext *pb){
int length;
put_bits(pb, 1, 1);
- length= (-get_bit_count(pb))&7;
+ length= (-put_bits_count(pb))&7;
if(length) put_bits(pb, length, 0);
}
+#endif
/**
* identifies the exact end of the bitstream
}
}
+#if 0
/**
* dct tranforms the 16 dc values.
* @param qp quantization parameter ??? FIXME
block[stride*10+offset]= (z0 - z3)>>1;
}
}
+#endif
+
#undef xStride
#undef stride
block[stride*1 + xStride*1]= ((e-b)*qmul + 0)>>1;
}
+#if 0
static void chroma_dc_dct_c(DCTELEM *block){
const int stride= 16*2;
const int xStride= 16;
block[stride*1 + xStride*0]= (a-c);
block[stride*1 + xStride*1]= (e-b);
}
+#endif
/**
* gets the chroma qp.
}
-/**
- *
- */
-static void h264_add_idct_c(uint8_t *dst, DCTELEM *block, int stride){
- int i;
- uint8_t *cm = cropTbl + MAX_NEG_CROP;
-
- block[0] += 32;
-#if 1
- for(i=0; i<4; i++){
- const int z0= block[i + 4*0] + block[i + 4*2];
- const int z1= block[i + 4*0] - block[i + 4*2];
- const int z2= (block[i + 4*1]>>1) - block[i + 4*3];
- const int z3= block[i + 4*1] + (block[i + 4*3]>>1);
-
- block[i + 4*0]= z0 + z3;
- block[i + 4*1]= z1 + z2;
- block[i + 4*2]= z1 - z2;
- block[i + 4*3]= z0 - z3;
- }
-
- for(i=0; i<4; i++){
- const int z0= block[0 + 4*i] + block[2 + 4*i];
- const int z1= block[0 + 4*i] - block[2 + 4*i];
- const int z2= (block[1 + 4*i]>>1) - block[3 + 4*i];
- const int z3= block[1 + 4*i] + (block[3 + 4*i]>>1);
-
- dst[0 + i*stride]= cm[ dst[0 + i*stride] + ((z0 + z3) >> 6) ];
- dst[1 + i*stride]= cm[ dst[1 + i*stride] + ((z1 + z2) >> 6) ];
- dst[2 + i*stride]= cm[ dst[2 + i*stride] + ((z1 - z2) >> 6) ];
- dst[3 + i*stride]= cm[ dst[3 + i*stride] + ((z0 - z3) >> 6) ];
- }
-#else
- for(i=0; i<4; i++){
- const int z0= block[0 + 4*i] + block[2 + 4*i];
- const int z1= block[0 + 4*i] - block[2 + 4*i];
- const int z2= (block[1 + 4*i]>>1) - block[3 + 4*i];
- const int z3= block[1 + 4*i] + (block[3 + 4*i]>>1);
-
- block[0 + 4*i]= z0 + z3;
- block[1 + 4*i]= z1 + z2;
- block[2 + 4*i]= z1 - z2;
- block[3 + 4*i]= z0 - z3;
- }
-
- for(i=0; i<4; i++){
- const int z0= block[i + 4*0] + block[i + 4*2];
- const int z1= block[i + 4*0] - block[i + 4*2];
- const int z2= (block[i + 4*1]>>1) - block[i + 4*3];
- const int z3= block[i + 4*1] + (block[i + 4*3]>>1);
-
- dst[i + 0*stride]= cm[ dst[i + 0*stride] + ((z0 + z3) >> 6) ];
- dst[i + 1*stride]= cm[ dst[i + 1*stride] + ((z1 + z2) >> 6) ];
- dst[i + 2*stride]= cm[ dst[i + 2*stride] + ((z1 - z2) >> 6) ];
- dst[i + 3*stride]= cm[ dst[i + 3*stride] + ((z0 - z3) >> 6) ];
- }
-#endif
-}
-
+#if 0
static void h264_diff_dct_c(DCTELEM *block, uint8_t *src1, uint8_t *src2, int stride){
int i;
//FIXME try int temp instead of block
block[3*4 + i]= z3 - 2*z2;
}
}
+#endif
//FIXME need to check that this doesnt overflow signed 32 bit for low qp, iam not sure, its very close
//FIXME check that gcc inlines this (and optimizes intra & seperate_dc stuff away)
src[2+0*stride]=
src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
-};
+}
static void pred4x4_down_left_c(uint8_t *src, uint8_t *topright, int stride){
LOAD_TOP_EDGE
src[3+2*stride]=
src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
-};
+}
static void pred4x4_vertical_right_c(uint8_t *src, uint8_t *topright, int stride){
const int lt= src[-1-1*stride];
src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
-};
+}
static void pred4x4_vertical_left_c(uint8_t *src, uint8_t *topright, int stride){
LOAD_TOP_EDGE
src[3+1*stride]=
src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
-};
+}
static void pred4x4_horizontal_up_c(uint8_t *src, uint8_t *topright, int stride){
LOAD_LEFT_EDGE
src[2+2*stride]=
src[2+3*stride]=
src[3+3*stride]=l3;
-};
+}
static void pred4x4_horizontal_down_c(uint8_t *src, uint8_t *topright, int stride){
const int lt= src[-1-1*stride];
src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
src[0+3*stride]=(l2 + l3 + 1)>>1;
src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
-};
+}
static void pred16x16_vertical_c(uint8_t *src, int stride){
int i;
}
}
-static void pred16x16_plane_c(uint8_t *src, int stride){
- uint8_t *cm = cropTbl + MAX_NEG_CROP;
- int i, dx, dy, dc;
- int temp[16];
-
- dc= 16*(src[15-stride] + src[-1+15*stride]);
-
- dx=dy=0;
- for(i=1; i<9; i++){
- dx += i*(src[7+i-stride] - src[7-i-stride]);
- dy += i*(src[-1+(7+i)*stride] - src[-1+(7-i)*stride]);
- }
- dx= (5*dx+32)>>6;
- dy= (5*dy+32)>>6;
-
- dc += 16;
+static inline void pred16x16_plane_compat_c(uint8_t *src, int stride, const int svq3){
+ int i, j, k;
+ int a;
+ uint8_t *cm = cropTbl + MAX_NEG_CROP;
+ const uint8_t * const src0 = src+7-stride;
+ const uint8_t *src1 = src+8*stride-1;
+ const uint8_t *src2 = src1-2*stride; // == src+6*stride-1;
+ int H = src0[1] - src0[-1];
+ int V = src1[0] - src2[ 0];
+ for(k=2; k<=8; ++k) {
+ src1 += stride; src2 -= stride;
+ H += k*(src0[k] - src0[-k]);
+ V += k*(src1[0] - src2[ 0]);
+ }
+ if(svq3){
+ H = ( 5*(H/4) ) / 16;
+ V = ( 5*(V/4) ) / 16;
+
+ /* required for 100% accuracy */
+ i = H; H = V; V = i;
+ }else{
+ H = ( 5*H+32 ) >> 6;
+ V = ( 5*V+32 ) >> 6;
+ }
+
+ a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
+ for(j=16; j>0; --j) {
+ int b = a;
+ a += V;
+ for(i=-16; i<0; i+=4) {
+ src[16+i] = cm[ (b ) >> 5 ];
+ src[17+i] = cm[ (b+ H) >> 5 ];
+ src[18+i] = cm[ (b+2*H) >> 5 ];
+ src[19+i] = cm[ (b+3*H) >> 5 ];
+ b += 4*H;
+ }
+ src += stride;
+ }
+}
- //FIXME modifiy dc,dx,dy to avoid -7
-
- for(i=0; i<16; i++)
- temp[i]= dx*(i-7) + dc;
-
- if( (dc - ABS(dx)*8 - ABS(dy)*8)>>5 < 0
- || (dc + ABS(dx)*8 + ABS(dy)*8)>>5 > 255){
-
- for(i=0; i<16; i++){
- int j;
- for(j=0; j<16; j++)
- src[j + i*stride]= cm[ (temp[j] + dy*(i-7))>>5 ];
- }
- }else{
- for(i=0; i<16; i++){
- int j;
- for(j=0; j<16; j++)
- src[j + i*stride]= (temp[j] + dy*(i-7))>>5;
- }
- }
+static void pred16x16_plane_c(uint8_t *src, int stride){
+ pred16x16_plane_compat_c(src, stride, 0);
}
static void pred8x8_vertical_c(uint8_t *src, int stride){
}
static void pred8x8_plane_c(uint8_t *src, int stride){
- uint8_t *cm = cropTbl + MAX_NEG_CROP;
- int i, dx, dy, dc;
- int temp[8];
-
- dc= 16*(src[7-stride] + src[-1+7*stride]);
-
- dx=dy=0;
- for(i=1; i<5; i++){
- dx += i*(src[3+i-stride] - src[3-i-stride]);
- dy += i*(src[-1+(3+i)*stride] - src[-1+(3-i)*stride]);
- }
- dx= (17*dx+16)>>5;
- dy= (17*dy+16)>>5;
-
- dc += 16;
-
- //FIXME modifiy dc,dx,dy to avoid -3
-
- for(i=0; i<8; i++)
- temp[i]= dx*(i-3) + dc;
-
- if( (dc - ABS(dx)*4 - ABS(dy)*4)>>5 < 0
- || (dc + ABS(dx)*4 + ABS(dy)*4)>>5 > 255){
-
- for(i=0; i<8; i++){
- int j;
- for(j=0; j<8; j++)
- src[j + i*stride]= cm[ (temp[j] + dy*(i-3))>>5 ];
- }
- }else{
- for(i=0; i<8; i++){
- int j;
- for(j=0; j<8; j++)
- src[j + i*stride]= (temp[j] + dy*(i-3))>>5;
- }
- }
+ int j, k;
+ int a;
+ uint8_t *cm = cropTbl + MAX_NEG_CROP;
+ const uint8_t * const src0 = src+3-stride;
+ const uint8_t *src1 = src+4*stride-1;
+ const uint8_t *src2 = src1-2*stride; // == src+2*stride-1;
+ int H = src0[1] - src0[-1];
+ int V = src1[0] - src2[ 0];
+ for(k=2; k<=4; ++k) {
+ src1 += stride; src2 -= stride;
+ H += k*(src0[k] - src0[-k]);
+ V += k*(src1[0] - src2[ 0]);
+ }
+ H = ( 17*H+16 ) >> 5;
+ V = ( 17*V+16 ) >> 5;
+
+ a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
+ for(j=8; j>0; --j) {
+ int b = a;
+ a += V;
+ src[0] = cm[ (b ) >> 5 ];
+ src[1] = cm[ (b+ H) >> 5 ];
+ src[2] = cm[ (b+2*H) >> 5 ];
+ src[3] = cm[ (b+3*H) >> 5 ];
+ src[4] = cm[ (b+4*H) >> 5 ];
+ src[5] = cm[ (b+5*H) >> 5 ];
+ src[6] = cm[ (b+6*H) >> 5 ];
+ src[7] = cm[ (b+7*H) >> 5 ];
+ src += stride;
+ }
}
static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, int chroma_height, int delta, int list,
|| full_my < 0-extra_height
|| full_mx + 16/*FIXME*/ > s->width + extra_width
|| full_my + 16/*FIXME*/ > s->height + extra_height){
- ff_emulated_edge_mc(s, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, s->width, s->height);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, s->width, s->height);
src_y= s->edge_emu_buffer + 2 + 2*s->linesize;
emu=1;
}
if(s->flags&CODEC_FLAG_GRAY) return;
if(emu){
- ff_emulated_edge_mc(s, src_cb, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_cb, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
src_cb= s->edge_emu_buffer;
}
chroma_op(dest_cb, src_cb, s->uvlinesize, chroma_height, mx&7, my&7);
if(emu){
- ff_emulated_edge_mc(s, src_cr, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_cr, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
src_cr= s->edge_emu_buffer;
}
chroma_op(dest_cr, src_cr, s->uvlinesize, chroma_height, mx&7, my&7);
qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg)){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
const int mb_type= s->current_picture.mb_type[mb_xy];
assert(IS_INTER(mb_type));
init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
&chroma_dc_coeff_token_len [0], 1, 1,
- &chroma_dc_coeff_token_bits[0], 1, 1);
+ &chroma_dc_coeff_token_bits[0], 1, 1, 1);
for(i=0; i<4; i++){
init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
&coeff_token_len [i][0], 1, 1,
- &coeff_token_bits[i][0], 1, 1);
+ &coeff_token_bits[i][0], 1, 1, 1);
}
for(i=0; i<3; i++){
init_vlc(&chroma_dc_total_zeros_vlc[i], CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
&chroma_dc_total_zeros_len [i][0], 1, 1,
- &chroma_dc_total_zeros_bits[i][0], 1, 1);
+ &chroma_dc_total_zeros_bits[i][0], 1, 1, 1);
}
for(i=0; i<15; i++){
init_vlc(&total_zeros_vlc[i], TOTAL_ZEROS_VLC_BITS, 16,
&total_zeros_len [i][0], 1, 1,
- &total_zeros_bits[i][0], 1, 1);
+ &total_zeros_bits[i][0], 1, 1, 1);
}
for(i=0; i<6; i++){
init_vlc(&run_vlc[i], RUN_VLC_BITS, 7,
&run_len [i][0], 1, 1,
- &run_bits[i][0], 1, 1);
+ &run_bits[i][0], 1, 1, 1);
}
init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
&run_len [6][0], 1, 1,
- &run_bits[6][0], 1, 1);
+ &run_bits[6][0], 1, 1, 1);
}
}
h->pred16x16[DC_128_PRED8x8 ]= pred16x16_128_dc_c;
}
-//FIXME factorize
-#define CHECKED_ALLOCZ(p, size)\
-{\
- p= av_mallocz(size);\
- if(p==NULL){\
- perror("malloc");\
- goto fail;\
- }\
-}
-
static void free_tables(H264Context *h){
- MpegEncContext * const s = &h->s;
-
av_freep(&h->intra4x4_pred_mode);
+ av_freep(&h->chroma_pred_mode_table);
+ av_freep(&h->cbp_table);
+ av_freep(&h->mvd_table[0]);
+ av_freep(&h->mvd_table[1]);
+ av_freep(&h->direct_table);
av_freep(&h->non_zero_count);
av_freep(&h->slice_table_base);
+ av_freep(&h->top_border);
h->slice_table= NULL;
-
+
av_freep(&h->mb2b_xy);
av_freep(&h->mb2b8_xy);
}
*/
static int alloc_tables(H264Context *h){
MpegEncContext * const s = &h->s;
- const int big_mb_num= h->mb_stride * (s->mb_height+1);
+ const int big_mb_num= s->mb_stride * (s->mb_height+1);
int x,y;
CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8 * sizeof(uint8_t))
+
CHECKED_ALLOCZ(h->non_zero_count , big_mb_num * 16 * sizeof(uint8_t))
CHECKED_ALLOCZ(h->slice_table_base , big_mb_num * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->top_border , s->mb_width * (16+8+8) * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))
+
+ if( h->pps.cabac ) {
+ CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t));
+ CHECKED_ALLOCZ(h->mvd_table[1], 32*big_mb_num * sizeof(uint16_t));
+ CHECKED_ALLOCZ(h->direct_table, 32*big_mb_num * sizeof(uint8_t));
+ }
memset(h->slice_table_base, -1, big_mb_num * sizeof(uint8_t));
- h->slice_table= h->slice_table_base + h->mb_stride + 1;
+ h->slice_table= h->slice_table_base + s->mb_stride + 1;
CHECKED_ALLOCZ(h->mb2b_xy , big_mb_num * sizeof(uint16_t));
CHECKED_ALLOCZ(h->mb2b8_xy , big_mb_num * sizeof(uint16_t));
for(y=0; y<s->mb_height; y++){
for(x=0; x<s->mb_width; x++){
- const int mb_xy= x + y*h->mb_stride;
+ const int mb_xy= x + y*s->mb_stride;
const int b_xy = 4*x + 4*y*h->b_stride;
const int b8_xy= 2*x + 2*y*h->b8_stride;
static void common_init(H264Context *h){
MpegEncContext * const s = &h->s;
- int i;
s->width = s->avctx->width;
s->height = s->avctx->height;
init_pred_ptrs(h);
+ s->unrestricted_mv=1;
s->decode=1; //FIXME
}
H264Context *h= avctx->priv_data;
MpegEncContext * const s = &h->s;
+ MPV_decode_defaults(s);
+
s->avctx = avctx;
common_init(h);
s->workaround_bugs= avctx->workaround_bugs;
// set defaults
- s->progressive_sequence=1;
// s->decode_mb= ff_h263_decode_mb;
s->low_delay= 1;
avctx->pix_fmt= PIX_FMT_YUV420P;
decode_init_vlc(h);
+ if(avctx->codec_tag != 0x31637661 && avctx->codec_tag != 0x31435641) // avc1
+ h->is_avc = 0;
+ else {
+ if((avctx->extradata_size == 0) || (avctx->extradata == NULL)) {
+ av_log(avctx, AV_LOG_ERROR, "AVC codec requires avcC data\n");
+ return -1;
+ }
+ h->is_avc = 1;
+ h->got_avcC = 0;
+ }
+
return 0;
}
// s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
}
+static inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize){
+ MpegEncContext * const s = &h->s;
+ int i;
+
+ src_y -= linesize;
+ src_cb -= uvlinesize;
+ src_cr -= uvlinesize;
+
+ h->left_border[0]= h->top_border[s->mb_x][15];
+ for(i=1; i<17; i++){
+ h->left_border[i]= src_y[15+i* linesize];
+ }
+
+ *(uint64_t*)(h->top_border[s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
+ *(uint64_t*)(h->top_border[s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);
+
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ h->left_border[17 ]= h->top_border[s->mb_x][16+7];
+ h->left_border[17+9]= h->top_border[s->mb_x][24+7];
+ for(i=1; i<9; i++){
+ h->left_border[i+17 ]= src_cb[7+i*uvlinesize];
+ h->left_border[i+17+9]= src_cr[7+i*uvlinesize];
+ }
+ *(uint64_t*)(h->top_border[s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
+ *(uint64_t*)(h->top_border[s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
+ }
+}
+
+static inline void xchg_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int xchg){
+ MpegEncContext * const s = &h->s;
+ int temp8, i;
+ uint64_t temp64;
+ int deblock_left = (s->mb_x > 0);
+ int deblock_top = (s->mb_y > 0);
+
+ src_y -= linesize + 1;
+ src_cb -= uvlinesize + 1;
+ src_cr -= uvlinesize + 1;
+
+#define XCHG(a,b,t,xchg)\
+t= a;\
+if(xchg)\
+ a= b;\
+b= t;
+
+ if(deblock_left){
+ for(i = !deblock_top; i<17; i++){
+ XCHG(h->left_border[i ], src_y [i* linesize], temp8, xchg);
+ }
+ }
+
+ if(deblock_top){
+ XCHG(*(uint64_t*)(h->top_border[s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
+ XCHG(*(uint64_t*)(h->top_border[s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
+ }
+
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ if(deblock_left){
+ for(i = !deblock_top; i<9; i++){
+ XCHG(h->left_border[i+17 ], src_cb[i*uvlinesize], temp8, xchg);
+ XCHG(h->left_border[i+17+9], src_cr[i*uvlinesize], temp8, xchg);
+ }
+ }
+ if(deblock_top){
+ XCHG(*(uint64_t*)(h->top_border[s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
+ XCHG(*(uint64_t*)(h->top_border[s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
+ }
+ }
+}
+
static void hl_decode_mb(H264Context *h){
MpegEncContext * const s = &h->s;
const int mb_x= s->mb_x;
const int mb_y= s->mb_y;
- const int mb_xy= mb_x + mb_y*h->mb_stride;
+ const int mb_xy= mb_x + mb_y*s->mb_stride;
const int mb_type= s->current_picture.mb_type[mb_xy];
uint8_t *dest_y, *dest_cb, *dest_cr;
int linesize, uvlinesize /*dct_offset*/;
}
if(IS_INTRA(mb_type)){
+ if(h->deblocking_filter)
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1);
+
if(!(s->flags&CODEC_FLAG_GRAY)){
h->pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize);
h->pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize);
if(!s->encoding){
for(i=0; i<16; i++){
uint8_t * const ptr= dest_y + h->block_offset[i];
- uint8_t *topright= ptr + 4 - linesize;
- const int topright_avail= (h->topright_samples_available<<i)&0x8000;
+ uint8_t *topright;
const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
int tr;
- if(!topright_avail){
- tr= ptr[3 - linesize]*0x01010101;
- topright= (uint8_t*) &tr;
- }
+ if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
+ const int topright_avail= (h->topright_samples_available<<i)&0x8000;
+ assert(mb_y || linesize <= h->block_offset[i]);
+ if(!topright_avail){
+ tr= ptr[3 - linesize]*0x01010101;
+ topright= (uint8_t*) &tr;
+ }else if(i==5 && h->deblocking_filter){
+ tr= *(uint32_t*)h->top_border[mb_x+1];
+ topright= (uint8_t*) &tr;
+ }else
+ topright= ptr + 4 - linesize;
+ }else
+ topright= NULL;
h->pred4x4[ dir ](ptr, topright, linesize);
- if(h->non_zero_count_cache[ scan8[i] ])
- h264_add_idct_c(ptr, h->mb + i*16, linesize);
+ if(h->non_zero_count_cache[ scan8[i] ]){
+ if(s->codec_id == CODEC_ID_H264)
+ s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
+ else
+ svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
+ }
}
}
}else{
h->pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize);
- h264_luma_dc_dequant_idct_c(h->mb, s->qscale);
+ if(s->codec_id == CODEC_ID_H264)
+ h264_luma_dc_dequant_idct_c(h->mb, s->qscale);
+ else
+ svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
}
- }else{
+ if(h->deblocking_filter)
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
+ }else if(s->codec_id == CODEC_ID_H264){
hl_motion(h, dest_y, dest_cb, dest_cr,
s->dsp.put_h264_qpel_pixels_tab, s->dsp.put_h264_chroma_pixels_tab,
s->dsp.avg_h264_qpel_pixels_tab, s->dsp.avg_h264_chroma_pixels_tab);
if(!IS_INTRA4x4(mb_type)){
- for(i=0; i<16; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
- uint8_t * const ptr= dest_y + h->block_offset[i];
- h264_add_idct_c(ptr, h->mb + i*16, linesize);
+ if(s->codec_id == CODEC_ID_H264){
+ for(i=0; i<16; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
+ uint8_t * const ptr= dest_y + h->block_offset[i];
+ s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
+ }
+ }
+ }else{
+ for(i=0; i<16; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
+ uint8_t * const ptr= dest_y + h->block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
+ }
}
}
}
if(!(s->flags&CODEC_FLAG_GRAY)){
chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp);
chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp);
- for(i=16; i<16+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cb + h->block_offset[i];
- h264_add_idct_c(ptr, h->mb + i*16, uvlinesize);
+ if(s->codec_id == CODEC_ID_H264){
+ for(i=16; i<16+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cb + h->block_offset[i];
+ s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
+ }
}
- }
- for(i=20; i<20+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cr + h->block_offset[i];
- h264_add_idct_c(ptr, h->mb + i*16, uvlinesize);
+ for(i=20; i<20+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cr + h->block_offset[i];
+ s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
+ }
+ }
+ }else{
+ for(i=16; i<16+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cb + h->block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ }
+ }
+ for(i=20; i<20+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cr + h->block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ }
}
}
}
-}
-
-static void decode_mb_cabac(H264Context *h){
-// MpegEncContext * const s = &h->s;
+ if(h->deblocking_filter) {
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
+ filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr);
+ }
}
/**
for(out_i=0; out_i<h->short_ref_count; out_i++){
int best_i=-1;
- int best_poc=-1;
+ int best_poc=INT_MAX;
for(i=0; i<h->short_ref_count; i++){
const int poc= h->short_ref[i]->poc;
int index=0;
for(i=0; i<h->short_ref_count && index < h->ref_count[list]; i++){
- const int i2= list ? h->short_ref_count - i - 1 : i;
+ const int i2= list ? i : h->short_ref_count - i - 1;
const int poc= sorted_short_ref[i2].poc;
if(sorted_short_ref[i2].reference != 3) continue; //FIXME refernce field shit
int pic_id;
int i;
+ if(reordering_of_pic_nums_idc==3)
+ break;
if(index >= h->ref_count[list]){
- fprintf(stderr, "reference count overflow\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference count overflow\n");
return -1;
}
const int abs_diff_pic_num= get_ue_golomb(&s->gb) + 1;
if(abs_diff_pic_num >= h->max_pic_num){
- fprintf(stderr, "abs_diff_pic_num overflow\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "abs_diff_pic_num overflow\n");
return -1;
}
}
if(i < index){
- fprintf(stderr, "reference picture missing during reorder\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference picture missing during reorder\n");
memset(&h->ref_list[list][index], 0, sizeof(Picture)); //FIXME
}else if(i > index){
Picture tmp= h->ref_list[list][i];
}
h->ref_list[list][index]= tmp;
}
- }else if(reordering_of_pic_nums_idc==3)
- break;
- else{
- fprintf(stderr, "illegal reordering_of_pic_nums_idc\n");
+ }else{
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal reordering_of_pic_nums_idc\n");
return -1;
}
}
if(h->slice_type!=B_TYPE) break;
}
+
+ if(h->slice_type==B_TYPE && !h->direct_spatial_mv_pred)
+ direct_dist_scale_factor(h);
return 0;
}
}
/**
- * instantaneos decoder refresh.
+ * instantaneous decoder refresh.
*/
static void idr(H264Context *h){
int i;
int i;
if(s->avctx->debug&FF_DEBUG_MMCO)
- printf("remove short %d count %d\n", frame_num, h->short_ref_count);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
for(i=0; i<h->short_ref_count; i++){
Picture *pic= h->short_ref[i];
if(s->avctx->debug&FF_DEBUG_MMCO)
- printf("%d %d %X\n", i, pic->frame_num, (int)pic);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "%d %d %p\n", i, pic->frame_num, pic);
if(pic->frame_num == frame_num){
h->short_ref[i]= NULL;
memmove(&h->short_ref[i], &h->short_ref[i+1], (h->short_ref_count - i - 1)*sizeof(Picture*));
Picture *pic;
if((s->avctx->debug&FF_DEBUG_MMCO) && mmco_count==0)
- printf("no mmco here\n");
+ av_log(h->s.avctx, AV_LOG_DEBUG, "no mmco here\n");
for(i=0; i<mmco_count; i++){
if(s->avctx->debug&FF_DEBUG_MMCO)
- printf("mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_frame_num, h->mmco[i].long_index);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_frame_num, h->mmco[i].long_index);
switch(mmco[i].opcode){
case MMCO_SHORT2UNUSED:
pic= remove_short(h, s->current_picture_ptr->frame_num);
if(pic){
pic->reference=0;
- fprintf(stderr, "illegal short term buffer state detected\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
}
if(h->short_ref_count)
if(opcode==MMCO_SHORT2LONG || opcode==MMCO_LONG2UNUSED || opcode==MMCO_LONG || opcode==MMCO_SET_MAX_LONG){
h->mmco[i].long_index= get_ue_golomb(&s->gb);
if(/*h->mmco[i].long_index >= h->long_ref_count || h->long_ref[ h->mmco[i].long_index ] == NULL*/ h->mmco[i].long_index >= 16){
- fprintf(stderr, "illegal long ref in memory management control operation %d\n", opcode);
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal long ref in memory management control operation %d\n", opcode);
return -1;
}
}
if(opcode > MMCO_LONG){
- fprintf(stderr, "illegal memory management control operation %d\n", opcode);
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal memory management control operation %d\n", opcode);
return -1;
}
+ if(opcode == MMCO_END)
+ break;
}
h->mmco_index= i;
}else{
int first_mb_in_slice, pps_id;
int num_ref_idx_active_override_flag;
static const uint8_t slice_type_map[5]= {P_TYPE, B_TYPE, I_TYPE, SP_TYPE, SI_TYPE};
- float new_aspect;
s->current_picture.reference= h->nal_ref_idc != 0;
h->slice_type= get_ue_golomb(&s->gb);
if(h->slice_type > 9){
- fprintf(stderr, "slice type too large (%d) at %d %d\n", h->slice_type, s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "slice type too large (%d) at %d %d\n", h->slice_type, s->mb_x, s->mb_y);
+ return -1;
}
if(h->slice_type > 4){
h->slice_type -= 5;
pps_id= get_ue_golomb(&s->gb);
if(pps_id>255){
- fprintf(stderr, "pps_id out of range\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
return -1;
}
h->pps= h->pps_buffer[pps_id];
+ if(h->pps.slice_group_count == 0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "non existing PPS referenced\n");
+ return -1;
+ }
+
h->sps= h->sps_buffer[ h->pps.sps_id ];
+ if(h->sps.log2_max_frame_num == 0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "non existing SPS referenced\n");
+ return -1;
+ }
s->mb_width= h->sps.mb_width;
s->mb_height= h->sps.mb_height;
- h->mb_stride= s->mb_width + 1;
- h->b_stride= s->mb_width*4;
- h->b8_stride= s->mb_width*2;
+ h->b_stride= s->mb_width*4 + 1;
+ h->b8_stride= s->mb_width*2 + 1;
- s->mb_x = first_mb_in_slice % s->mb_width;
- s->mb_y = first_mb_in_slice / s->mb_width; //FIXME AFFW
+ s->resync_mb_x = s->mb_x = first_mb_in_slice % s->mb_width;
+ s->resync_mb_y = s->mb_y = first_mb_in_slice / s->mb_width; //FIXME AFFW
- s->width = 16*s->mb_width - 2*(h->pps.crop_left + h->pps.crop_right );
+ s->width = 16*s->mb_width - 2*(h->sps.crop_left + h->sps.crop_right );
if(h->sps.frame_mbs_only_flag)
- s->height= 16*s->mb_height - 2*(h->pps.crop_top + h->pps.crop_bottom);
+ s->height= 16*s->mb_height - 2*(h->sps.crop_top + h->sps.crop_bottom);
else
- s->height= 16*s->mb_height - 4*(h->pps.crop_top + h->pps.crop_bottom); //FIXME recheck
+ s->height= 16*s->mb_height - 4*(h->sps.crop_top + h->sps.crop_bottom); //FIXME recheck
- if(h->pps.crop_left || h->pps.crop_top){
- fprintf(stderr, "insane croping not completly supported, this could look slightly wrong ...\n");
- }
-
- if(s->aspected_height) //FIXME emms at end of slice ?
- new_aspect= h->sps.sar_width*s->width / (float)(s->height*h->sps.sar_height);
- else
- new_aspect=0;
-
if (s->context_initialized
- && ( s->width != s->avctx->width || s->height != s->avctx->height
- || ABS(new_aspect - s->avctx->aspect_ratio) > 0.001)) {
+ && ( s->width != s->avctx->width || s->height != s->avctx->height)) {
free_tables(h);
MPV_common_end(s);
}
s->avctx->width = s->width;
s->avctx->height = s->height;
- s->avctx->aspect_ratio= new_aspect;
+ s->avctx->sample_aspect_ratio= h->sps.sar;
+
+ if(h->sps.timing_info_present_flag && h->sps.fixed_frame_rate_flag){
+ s->avctx->frame_rate = h->sps.time_scale;
+ s->avctx->frame_rate_base = h->sps.num_units_in_tick;
+ }
}
- if(first_mb_in_slice == 0){
+ if(h->slice_num == 0){
frame_start(h);
}
}
if(h->nal_unit_type == NAL_IDR_SLICE){
- int idr_pic_id= get_ue_golomb(&s->gb);
+ get_ue_golomb(&s->gb); /* idr_pic_id */
}
if(h->sps.poc_type==0){
h->ref_count[1]= get_ue_golomb(&s->gb) + 1;
if(h->ref_count[0] > 32 || h->ref_count[1] > 32){
- fprintf(stderr, "reference overflow\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
return -1;
}
}
}
- if(first_mb_in_slice == 0){
+ if(h->slice_num == 0){
fill_default_ref_list(h);
}
if(s->current_picture.reference)
decode_ref_pic_marking(h);
- //FIXME CABAC stuff
- s->qscale = h->pps.init_qp + get_se_golomb(&s->gb); //slice_qp_delta
+ if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE && h->pps.cabac )
+ h->cabac_init_idc = get_ue_golomb(&s->gb);
+
+ h->last_qscale_diff = 0;
+ s->qscale = h->pps.init_qp + get_se_golomb(&s->gb);
+ if(s->qscale<0 || s->qscale>51){
+ av_log(s->avctx, AV_LOG_ERROR, "QP %d out of range\n", s->qscale);
+ return -1;
+ }
//FIXME qscale / qp ... stuff
if(h->slice_type == SP_TYPE){
- int sp_for_switch_flag= get_bits1(&s->gb);
+ get_bits1(&s->gb); /* sp_for_switch_flag */
}
if(h->slice_type==SP_TYPE || h->slice_type == SI_TYPE){
- int slice_qs_delta= get_se_golomb(&s->gb);
+ get_se_golomb(&s->gb); /* slice_qs_delta */
}
+ h->deblocking_filter = 1;
+ h->slice_alpha_c0_offset = 0;
+ h->slice_beta_offset = 0;
if( h->pps.deblocking_filter_parameters_present ) {
- h->disable_deblocking_filter_idc= get_ue_golomb(&s->gb);
- if( h->disable_deblocking_filter_idc != 1 ) {
- h->slice_alpha_c0_offset_div2= get_se_golomb(&s->gb);
- h->slice_beta_offset_div2= get_se_golomb(&s->gb);
+ h->deblocking_filter= get_ue_golomb(&s->gb);
+ if(h->deblocking_filter < 2)
+ h->deblocking_filter^= 1; // 1<->0
+
+ if( h->deblocking_filter ) {
+ h->slice_alpha_c0_offset = get_se_golomb(&s->gb) << 1;
+ h->slice_beta_offset = get_se_golomb(&s->gb) << 1;
}
- }else
- h->disable_deblocking_filter_idc= 0;
+ }
#if 0 //FMO
if( h->pps.num_slice_groups > 1 && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5)
slice_group_change_cycle= get_bits(&s->gb, ?);
#endif
+ h->slice_num++;
+
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- printf("mb:%d %c pps:%d frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d\n",
- first_mb_in_slice,
- ff_get_pict_type_char(h->slice_type),
+ av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d mb:%d %c pps:%d frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d\n",
+ h->slice_num, first_mb_in_slice,
+ av_get_pict_type_char(h->slice_type),
pps_id, h->frame_num,
s->current_picture_ptr->field_poc[0], s->current_picture_ptr->field_poc[1],
h->ref_count[0], h->ref_count[1],
s->qscale,
- h->disable_deblocking_filter_idc
+ h->deblocking_filter
);
}
log= 32 - av_log2(buf);
#ifdef TRACE
print_bin(buf>>(32-log), log);
- printf("%5d %2d %3d lpr @%5d in %s get_level_prefix\n", buf>>(32-log), log, log-1, get_bits_count(gb), __FILE__);
+ av_log(NULL, AV_LOG_DEBUG, "%5d %2d %3d lpr @%5d in %s get_level_prefix\n", buf>>(32-log), log, log-1, get_bits_count(gb), __FILE__);
#endif
LAST_SKIP_BITS(re, gb, log);
level_code= (prefix<<suffix_length) + get_bits(gb, 12); //part
if(suffix_length==0) level_code+=15; //FIXME doesnt make (much)sense
}else{
- fprintf(stderr, "prefix too large at %d %d\n", s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "prefix too large at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
if(ABS(level[i]) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
#else
if((2+level_code)>>1) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
- ? == prefix > 2 or sth
+ /* ? == prefix > 2 or sth */
#endif
tprintf("level: %d suffix_length:%d\n", level[i], suffix_length);
}
}
if(zeros_left<0){
- fprintf(stderr, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
return 0;
}
+/**
+ * decodes a P_SKIP or B_SKIP macroblock
+ */
+static void decode_mb_skip(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
+ int mb_type;
+
+ memset(h->non_zero_count[mb_xy], 0, 16);
+ memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
+
+ if( h->slice_type == B_TYPE )
+ {
+ // just for fill_caches. pred_direct_motion will set the real mb_type
+ mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
+ //FIXME mbaff
+
+ fill_caches(h, mb_type); //FIXME check what is needed and what not ...
+ pred_direct_motion(h, &mb_type);
+ if(h->pps.cabac){
+ fill_rectangle(h->mvd_cache[0][scan8[0]], 4, 4, 8, 0, 4);
+ fill_rectangle(h->mvd_cache[1][scan8[0]], 4, 4, 8, 0, 4);
+ }
+ }
+ else
+ {
+ int mx, my;
+ mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;
+
+ if(h->sps.mb_aff && s->mb_skip_run==0 && (s->mb_y&1)==0){
+ h->mb_field_decoding_flag= get_bits1(&s->gb);
+ }
+ if(h->mb_field_decoding_flag)
+ mb_type|= MB_TYPE_INTERLACED;
+
+ fill_caches(h, mb_type); //FIXME check what is needed and what not ...
+ pred_pskip_motion(h, &mx, &my);
+ fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
+ fill_rectangle( h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mx,my), 4);
+ if(h->pps.cabac)
+ fill_rectangle(h->mvd_cache[0][scan8[0]], 4, 4, 8, 0, 4);
+ }
+
+ write_back_motion(h, mb_type);
+ s->current_picture.mb_type[mb_xy]= mb_type|MB_TYPE_SKIP;
+ s->current_picture.qscale_table[mb_xy]= s->qscale;
+ h->slice_table[ mb_xy ]= h->slice_num;
+ h->prev_mb_skiped= 1;
+}
+
/**
* decodes a macroblock
* @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
*/
-static int decode_mb(H264Context *h){
+static int decode_mb_cavlc(H264Context *h){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int mb_type, partition_count, cbp;
- memset(h->mb, 0, sizeof(int16_t)*24*16); //FIXME avoid if allready clear (move after skip handlong?
+ s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?
tprintf("pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
-
+ cbp = 0; /* avoid warning. FIXME: find a solution without slowing
+ down the code */
if(h->slice_type != I_TYPE && h->slice_type != SI_TYPE){
if(s->mb_skip_run==-1)
s->mb_skip_run= get_ue_golomb(&s->gb);
if (s->mb_skip_run--) {
- int i, mx, my;
- /* skip mb */
-#if 0 //FIXME
- for(i=0;i<6;i++)
- s->block_last_index[i] = -1;
- s->mv_type = MV_TYPE_16X16;
- /* if P type, zero motion vector is implied */
- s->mv_dir = MV_DIR_FORWARD;
- s->mb_skiped = 1;
-#endif
-//FIXME b frame
- mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0;
-
- memset(h->non_zero_count[mb_xy], 0, 16);
- memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
-
- if(h->sps.mb_aff && s->mb_skip_run==0 && (s->mb_y&1)==0){
- h->mb_field_decoding_flag= get_bits1(&s->gb);
- }
-
- if(h->mb_field_decoding_flag)
- mb_type|= MB_TYPE_INTERLACED;
-
- fill_caches(h, mb_type); //FIXME check what is needed and what not ...
- pred_pskip_motion(h, &mx, &my);
- fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
- fill_rectangle( h->mv_cache[0][scan8[0]], 4, 4, 8, (mx&0xFFFF)+(my<<16), 4);
- write_back_motion(h, mb_type);
-
- s->current_picture.mb_type[mb_xy]= mb_type; //FIXME SKIP type
- h->slice_table[ mb_xy ]= h->slice_num;
-
- h->prev_mb_skiped= 1;
+ decode_mb_skip(h);
return 0;
}
}
assert(h->slice_type == I_TYPE);
decode_intra_mb:
if(mb_type > 25){
- fprintf(stderr, "mb_type %d in %c slice to large at %d %d\n", mb_type, ff_get_pict_type_char(h->slice_type), s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice to large at %d %d\n", mb_type, av_get_pict_type_char(h->slice_type), s->mb_x, s->mb_y);
return -1;
}
partition_count=0;
if(IS_INTRA_PCM(mb_type)){
const uint8_t *ptr;
- int x, y, i;
+ int x, y;
// we assume these blocks are very rare so we dont optimize it
align_get_bits(&s->gb);
skip_bits(&s->gb, 384); //FIXME check /fix the bitstream readers
+ //FIXME deblock filter, non_zero_count_cache init ...
memset(h->non_zero_count[mb_xy], 16, 16);
+ s->current_picture.qscale_table[mb_xy]= s->qscale;
return 0;
}
for(i=0; i<4; i++){
h->sub_mb_type[i]= get_ue_golomb(&s->gb);
if(h->sub_mb_type[i] >=13){
- fprintf(stderr, "B sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "B sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
return -1;
}
sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
h->sub_mb_type[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
}
+ if( IS_DIRECT(h->sub_mb_type[0]) || IS_DIRECT(h->sub_mb_type[1])
+ || IS_DIRECT(h->sub_mb_type[2]) || IS_DIRECT(h->sub_mb_type[3]))
+ pred_direct_motion(h, &mb_type);
}else{
assert(h->slice_type == P_TYPE || h->slice_type == SP_TYPE); //FIXME SP correct ?
for(i=0; i<4; i++){
h->sub_mb_type[i]= get_ue_golomb(&s->gb);
if(h->sub_mb_type[i] >=4){
- fprintf(stderr, "P sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "P sub_mb_type %d out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
return -1;
}
sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
const int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
if(ref_count == 0) continue;
for(i=0; i<4; i++){
- if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
+ if(IS_DIRECT(h->sub_mb_type[i])) continue;
+ if(IS_DIR(h->sub_mb_type[i], 0, list)){
ref[list][i] = get_te0_golomb(&s->gb, ref_count); //FIXME init to 0 before and skip?
}else{
//FIXME
if(ref_count == 0) continue;
for(i=0; i<4; i++){
+ if(IS_DIRECT(h->sub_mb_type[i])) continue;
h->ref_cache[list][ scan8[4*i] ]=h->ref_cache[list][ scan8[4*i]+1 ]=
h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
- if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
+ if(IS_DIR(h->sub_mb_type[i], 0, list)){
const int sub_mb_type= h->sub_mb_type[i];
const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
for(j=0; j<sub_partition_count[i]; j++){
}
}
}
- }else if(!IS_DIRECT(mb_type)){
+ }else if(IS_DIRECT(mb_type)){
+ pred_direct_motion(h, &mb_type);
+ s->current_picture.mb_type[mb_xy]= mb_type;
+ }else{
int list, mx, my, i;
//FIXME we should set ref_idx_l? to 0 if we use that later ...
if(IS_16X16(mb_type)){
for(list=0; list<2; list++){
- if(h->ref_count[0]>0){
+ if(h->ref_count[list]>0){
if(IS_DIR(mb_type, 0, list)){
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1);
my += get_se_golomb(&s->gb);
tprintf("final mv:%d %d\n", mx, my);
- fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, (mx&0xFFFF) + (my<<16), 4);
+ fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
}
}
}
if(IS_DIR(mb_type, i, list)){
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 1);
- }
+ }else // needed only for mixed refs (e.g. B_L0_L1_16x8)
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, (LIST_NOT_USED&0xFF), 1);
}
}
}
my += get_se_golomb(&s->gb);
tprintf("final mv:%d %d\n", mx, my);
- fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, (mx&0xFFFF) + (my<<16), 4);
- }
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
+ }else
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, 0, 4);
}
}
}else{
if(IS_DIR(mb_type, i, list)){ //FIXME optimize
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, val, 1);
- }
+ }else // needed only for mixed refs
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, (LIST_NOT_USED&0xFF), 1);
}
}
}
my += get_se_golomb(&s->gb);
tprintf("final mv:%d %d\n", mx, my);
- fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, (mx&0xFFFF) + (my<<16), 4);
- }
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
+ }else
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, 0, 4);
}
}
}
if(!IS_INTRA16x16(mb_type)){
cbp= get_ue_golomb(&s->gb);
if(cbp > 47){
- fprintf(stderr, "cbp too large (%d) at %d %d\n", cbp, s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%d) at %d %d\n", cbp, s->mb_x, s->mb_y);
return -1;
}
dquant= get_se_golomb(&s->gb);
if( dquant > 25 || dquant < -26 ){
- fprintf(stderr, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
+ av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
return -1;
}
}
}
}else{
- memset(&h->non_zero_count_cache[8], 0, 8*4); //FIXME stupid & slow
+ fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
}
}else{
for(i8x8=0; i8x8<4; i8x8++){
nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
}
}else{
- memset(&h->non_zero_count_cache[8], 0, 8*5);
+ uint8_t * const nnz= &h->non_zero_count_cache[0];
+ fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1);
+ nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
+ nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
}
+ s->current_picture.qscale_table[mb_xy]= s->qscale;
write_back_non_zero_count(h);
return 0;
}
-static int decode_slice(H264Context *h){
- MpegEncContext * const s = &h->s;
- const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;
-
- s->mb_skip_run= -1;
-
-#if 1
- for(;;){
- int ret= decode_mb(h);
-
- hl_decode_mb(h);
-
- if(ret>=0 && h->sps.mb_aff){ //FIXME optimal? or let mb_decode decode 16x32 ?
- s->mb_y++;
- ret= decode_mb(h);
-
- hl_decode_mb(h);
- s->mb_y--;
- }
+static int decode_cabac_intra_mb_type(H264Context *h, int ctx_base, int intra_slice) {
+ uint8_t *state= &h->cabac_state[ctx_base];
+ int mb_type;
+
+ if(intra_slice){
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
+ int ctx=0;
+ if( s->mb_x > 0 && !IS_INTRA4x4( s->current_picture.mb_type[mb_xy-1] ) )
+ ctx++;
+ if( s->mb_y > 0 && !IS_INTRA4x4( s->current_picture.mb_type[mb_xy-s->mb_stride] ) )
+ ctx++;
+ if( get_cabac( &h->cabac, &state[ctx] ) == 0 )
+ return 0; /* I4x4 */
+ state += 2;
+ }else{
+ if( get_cabac( &h->cabac, &state[0] ) == 0 )
+ return 0; /* I4x4 */
+ }
- if(ret<0){
- fprintf(stderr, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+ if( get_cabac_terminate( &h->cabac ) )
+ return 25; /* PCM */
- return -1;
- }
-
- if(++s->mb_x >= s->mb_width){
- s->mb_x=0;
- ff_draw_horiz_band(s, 16*s->mb_y, 16);
- if(++s->mb_y >= s->mb_height){
- tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
+ mb_type = 1; /* I16x16 */
+ if( get_cabac( &h->cabac, &state[1] ) )
+ mb_type += 12; /* cbp_luma != 0 */
- if(get_bits_count(&s->gb) == s->gb.size_in_bits){
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+ if( get_cabac( &h->cabac, &state[2] ) ) {
+ if( get_cabac( &h->cabac, &state[2+intra_slice] ) )
+ mb_type += 4 * 2; /* cbp_chroma == 2 */
+ else
+ mb_type += 4 * 1; /* cbp_chroma == 1 */
+ }
+ if( get_cabac( &h->cabac, &state[3+intra_slice] ) )
+ mb_type += 2;
+ if( get_cabac( &h->cabac, &state[3+2*intra_slice] ) )
+ mb_type += 1;
+ return mb_type;
+}
- return 0;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+static int decode_cabac_mb_type( H264Context *h ) {
+ MpegEncContext * const s = &h->s;
- return -1;
- }
+ if( h->slice_type == I_TYPE ) {
+ return decode_cabac_intra_mb_type(h, 3, 1);
+ } else if( h->slice_type == P_TYPE ) {
+ if( get_cabac( &h->cabac, &h->cabac_state[14] ) == 0 ) {
+ /* P-type */
+ if( get_cabac( &h->cabac, &h->cabac_state[15] ) == 0 ) {
+ if( get_cabac( &h->cabac, &h->cabac_state[16] ) == 0 )
+ return 0; /* P_L0_D16x16; */
+ else
+ return 3; /* P_8x8; */
+ } else {
+ if( get_cabac( &h->cabac, &h->cabac_state[17] ) == 0 )
+ return 2; /* P_L0_D8x16; */
+ else
+ return 1; /* P_L0_D16x8; */
}
+ } else {
+ return decode_cabac_intra_mb_type(h, 17, 0) + 5;
}
-
- if(get_bits_count(&s->gb) >= s->gb.size_in_bits && s->mb_skip_run<=0){
- if(get_bits_count(&s->gb) == s->gb.size_in_bits){
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
-
- return 0;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
-
- return -1;
- }
+ } else if( h->slice_type == B_TYPE ) {
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
+ int ctx = 0;
+ int bits;
+
+ if( s->mb_x > 0 && !IS_SKIP( s->current_picture.mb_type[mb_xy-1] )
+ && !IS_DIRECT( s->current_picture.mb_type[mb_xy-1] ) )
+ ctx++;
+ if( s->mb_y > 0 && !IS_SKIP( s->current_picture.mb_type[mb_xy-s->mb_stride] )
+ && !IS_DIRECT( s->current_picture.mb_type[mb_xy-s->mb_stride] ) )
+ ctx++;
+
+ if( !get_cabac( &h->cabac, &h->cabac_state[27+ctx] ) )
+ return 0; /* B_Direct_16x16 */
+
+ if( !get_cabac( &h->cabac, &h->cabac_state[27+3] ) ) {
+ return 1 + get_cabac( &h->cabac, &h->cabac_state[27+5] ); /* B_L[01]_16x16 */
}
+
+ bits = get_cabac( &h->cabac, &h->cabac_state[27+4] ) << 3;
+ bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] ) << 2;
+ bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] ) << 1;
+ bits|= get_cabac( &h->cabac, &h->cabac_state[27+5] );
+ if( bits < 8 )
+ return bits + 3; /* B_Bi_16x16 through B_L1_L0_16x8 */
+ else if( bits == 13 ) {
+ return decode_cabac_intra_mb_type(h, 32, 0) + 23;
+ } else if( bits == 14 )
+ return 11; /* B_L1_L0_8x16 */
+ else if( bits == 15 )
+ return 22; /* B_8x8 */
+
+ bits= ( bits<<1 ) | get_cabac( &h->cabac, &h->cabac_state[27+5] );
+ return bits - 4; /* B_L0_Bi_* through B_Bi_Bi_* */
+ } else {
+ /* TODO SI/SP frames? */
+ return -1;
}
-#endif
-#if 0
- for(;s->mb_y < s->mb_height; s->mb_y++){
- for(;s->mb_x < s->mb_width; s->mb_x++){
- int ret= decode_mb(h);
-
- hl_decode_mb(h);
+}
- if(ret<0){
- fprintf(stderr, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+static int decode_cabac_mb_skip( H264Context *h) {
+ MpegEncContext * const s = &h->s;
+ const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
+ const int mba_xy = mb_xy - 1;
+ const int mbb_xy = mb_xy - s->mb_stride;
+ int ctx = 0;
+
+ if( s->mb_x > 0 && !IS_SKIP( s->current_picture.mb_type[mba_xy] ) )
+ ctx++;
+ if( s->mb_y > 0 && !IS_SKIP( s->current_picture.mb_type[mbb_xy] ) )
+ ctx++;
+
+ if( h->slice_type == P_TYPE || h->slice_type == SP_TYPE)
+ return get_cabac( &h->cabac, &h->cabac_state[11+ctx] );
+ else /* B-frame */
+ return get_cabac( &h->cabac, &h->cabac_state[24+ctx] );
+}
- return -1;
- }
-
- if(++s->mb_x >= s->mb_width){
- s->mb_x=0;
- if(++s->mb_y >= s->mb_height){
- if(get_bits_count(s->gb) == s->gb.size_in_bits){
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+static int decode_cabac_mb_intra4x4_pred_mode( H264Context *h, int pred_mode ) {
+ int mode = 0;
- return 0;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+ if( get_cabac( &h->cabac, &h->cabac_state[68] ) )
+ return pred_mode;
- return -1;
- }
- }
- }
-
- if(get_bits_count(s->?gb) >= s->gb?.size_in_bits){
- if(get_bits_count(s->gb) == s->gb.size_in_bits){
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+ if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
+ mode += 1;
+ if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
+ mode += 2;
+ if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
+ mode += 4;
+ if( mode >= pred_mode )
+ return mode + 1;
+ else
+ return mode;
+}
- return 0;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+static int decode_cabac_mb_chroma_pre_mode( H264Context *h) {
+ MpegEncContext * const s = &h->s;
+ const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
+ const int mba_xy = mb_xy - 1;
+ const int mbb_xy = mb_xy - s->mb_stride;
- return -1;
- }
- }
- }
- s->mb_x=0;
- ff_draw_horiz_band(s, 16*s->mb_y, 16);
- }
-#endif
- return -1; //not reached
+ int ctx = 0;
+
+ /* No need to test for IS_INTRA4x4 and IS_INTRA16x16, as we set chroma_pred_mode_table to 0 */
+ if( s->mb_x > 0 && h->chroma_pred_mode_table[mba_xy] != 0 )
+ ctx++;
+
+ if( s->mb_y > 0 && h->chroma_pred_mode_table[mbb_xy] != 0 )
+ ctx++;
+
+ if( get_cabac( &h->cabac, &h->cabac_state[64+ctx] ) == 0 )
+ return 0;
+
+ if( get_cabac( &h->cabac, &h->cabac_state[64+3] ) == 0 )
+ return 1;
+ if( get_cabac( &h->cabac, &h->cabac_state[64+3] ) == 0 )
+ return 2;
+ else
+ return 3;
}
-static inline int decode_vui_parameters(H264Context *h, SPS *sps){
+static const uint8_t block_idx_x[16] = {
+ 0, 1, 0, 1, 2, 3, 2, 3, 0, 1, 0, 1, 2, 3, 2, 3
+};
+static const uint8_t block_idx_y[16] = {
+ 0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3
+};
+static const uint8_t block_idx_xy[4][4] = {
+ { 0, 2, 8, 10},
+ { 1, 3, 9, 11},
+ { 4, 6, 12, 14},
+ { 5, 7, 13, 15}
+};
+
+static int decode_cabac_mb_cbp_luma( H264Context *h) {
MpegEncContext * const s = &h->s;
- int aspect_ratio_info_present_flag, aspect_ratio_idc;
+ const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
- aspect_ratio_info_present_flag= get_bits1(&s->gb);
-
- if( aspect_ratio_info_present_flag ) {
- aspect_ratio_idc= get_bits(&s->gb, 8);
- if( aspect_ratio_idc == EXTENDED_SAR ) {
- sps->sar_width= get_bits(&s->gb, 16);
- sps->sar_height= get_bits(&s->gb, 16);
- }else if(aspect_ratio_idc < 16){
- sps->sar_width= pixel_aspect[aspect_ratio_idc][0];
- sps->sar_height= pixel_aspect[aspect_ratio_idc][1];
- }else{
- fprintf(stderr, "illegal aspect ratio\n");
- return -1;
+ int cbp = 0;
+ int i8x8;
+
+ h->cbp_table[mb_xy] = 0; /* FIXME aaahahahah beurk */
+
+ for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
+ int mba_xy = -1;
+ int mbb_xy = -1;
+ int x, y;
+ int ctx = 0;
+
+ x = block_idx_x[4*i8x8];
+ y = block_idx_y[4*i8x8];
+
+ if( x > 0 )
+ mba_xy = mb_xy;
+ else if( s->mb_x > 0 )
+ mba_xy = mb_xy - 1;
+
+ if( y > 0 )
+ mbb_xy = mb_xy;
+ else if( s->mb_y > 0 )
+ mbb_xy = mb_xy - s->mb_stride;
+
+ /* No need to test for skip as we put 0 for skip block */
+ if( mba_xy >= 0 ) {
+ int i8x8a = block_idx_xy[(x-1)&0x03][y]/4;
+ if( ((h->cbp_table[mba_xy] >> i8x8a)&0x01) == 0 )
+ ctx++;
+ }
+
+ if( mbb_xy >= 0 ) {
+ int i8x8b = block_idx_xy[x][(y-1)&0x03]/4;
+ if( ((h->cbp_table[mbb_xy] >> i8x8b)&0x01) == 0 )
+ ctx += 2;
+ }
+
+ if( get_cabac( &h->cabac, &h->cabac_state[73 + ctx] ) ) {
+ cbp |= 1 << i8x8;
+ h->cbp_table[mb_xy] = cbp; /* FIXME aaahahahah beurk */
}
- }else{
- sps->sar_width=
- sps->sar_height= 0;
}
-// s->avctx->aspect_ratio= sar_width*s->width / (float)(s->height*sar_height);
-#if 0
-| overscan_info_present_flag |0 |u(1) |
-| if( overscan_info_present_flag ) | | |
-| overscan_appropriate_flag |0 |u(1) |
-| video_signal_type_present_flag |0 |u(1) |
-| if( video_signal_type_present_flag ) { | | |
-| video_format |0 |u(3) |
-| video_full_range_flag |0 |u(1) |
-| colour_description_present_flag |0 |u(1) |
-| if( colour_description_present_flag ) { | | |
-| colour_primaries |0 |u(8) |
-| transfer_characteristics |0 |u(8) |
-| matrix_coefficients |0 |u(8) |
-| } | | |
-| } | | |
-| chroma_location_info_present_flag |0 |u(1) |
-| if ( chroma_location_info_present_flag ) { | | |
-| chroma_sample_location_type_top_field |0 |ue(v) |
-| chroma_sample_location_type_bottom_field |0 |ue(v) |
-| } | | |
-| timing_info_present_flag |0 |u(1) |
-| if( timing_info_present_flag ) { | | |
-| num_units_in_tick |0 |u(32) |
-| time_scale |0 |u(32) |
-| fixed_frame_rate_flag |0 |u(1) |
-| } | | |
-| nal_hrd_parameters_present_flag |0 |u(1) |
-| if( nal_hrd_parameters_present_flag = = 1) | | |
-| hrd_parameters( ) | | |
-| vcl_hrd_parameters_present_flag |0 |u(1) |
-| if( vcl_hrd_parameters_present_flag = = 1) | | |
-| hrd_parameters( ) | | |
-| if( ( nal_hrd_parameters_present_flag = = 1 | || | |
-| | | |
-|( vcl_hrd_parameters_present_flag = = 1 ) ) | | |
-| low_delay_hrd_flag |0 |u(1) |
-| bitstream_restriction_flag |0 |u(1) |
-| if( bitstream_restriction_flag ) { |0 |u(1) |
-| motion_vectors_over_pic_boundaries_flag |0 |u(1) |
-| max_bytes_per_pic_denom |0 |ue(v) |
-| max_bits_per_mb_denom |0 |ue(v) |
-| log2_max_mv_length_horizontal |0 |ue(v) |
-| log2_max_mv_length_vertical |0 |ue(v) |
-| num_reorder_frames |0 |ue(v) |
-| max_dec_frame_buffering |0 |ue(v) |
-| } | | |
-|} | | |
-#endif
- return 0;
+ return cbp;
}
+static int decode_cabac_mb_cbp_chroma( H264Context *h) {
+ int ctx;
+ int cbp_a, cbp_b;
-static inline int decode_seq_parameter_set(H264Context *h){
+ cbp_a = (h->left_cbp>>4)&0x03;
+ cbp_b = (h-> top_cbp>>4)&0x03;
+
+ ctx = 0;
+ if( cbp_a > 0 ) ctx++;
+ if( cbp_b > 0 ) ctx += 2;
+ if( get_cabac( &h->cabac, &h->cabac_state[77 + ctx] ) == 0 )
+ return 0;
+
+ ctx = 4;
+ if( cbp_a == 2 ) ctx++;
+ if( cbp_b == 2 ) ctx += 2;
+ return 1 + get_cabac( &h->cabac, &h->cabac_state[77 + ctx] );
+}
+static int decode_cabac_mb_dqp( H264Context *h) {
MpegEncContext * const s = &h->s;
- int profile_idc, level_idc, multiple_slice_groups, arbitrary_slice_order, redundant_slices;
- int sps_id, i;
- SPS *sps;
-
- profile_idc= get_bits(&s->gb, 8);
- level_idc= get_bits(&s->gb, 8);
- multiple_slice_groups= get_bits1(&s->gb);
- arbitrary_slice_order= get_bits1(&s->gb);
- redundant_slices= get_bits1(&s->gb);
-
- sps_id= get_ue_golomb(&s->gb);
-
- sps= &h->sps_buffer[ sps_id ];
-
- sps->profile_idc= profile_idc;
- sps->level_idc= level_idc;
- sps->multiple_slice_groups= multiple_slice_groups;
- sps->arbitrary_slice_order= arbitrary_slice_order;
- sps->redundant_slices= redundant_slices;
-
- sps->log2_max_frame_num= get_ue_golomb(&s->gb) + 4;
+ int mbn_xy;
+ int ctx = 0;
+ int val = 0;
- sps->poc_type= get_ue_golomb(&s->gb);
-
- if(sps->poc_type == 0){ //FIXME #define
- sps->log2_max_poc_lsb= get_ue_golomb(&s->gb) + 4;
- } else if(sps->poc_type == 1){//FIXME #define
- sps->delta_pic_order_always_zero_flag= get_bits1(&s->gb);
- sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
- sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
- sps->poc_cycle_length= get_ue_golomb(&s->gb);
-
- for(i=0; i<sps->poc_cycle_length; i++)
- sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
- }
- if(sps->poc_type > 2){
- fprintf(stderr, "illegal POC type %d\n", sps->poc_type);
- return -1;
+ if( s->mb_x > 0 )
+ mbn_xy = s->mb_x + s->mb_y*s->mb_stride - 1;
+ else
+ mbn_xy = s->mb_width - 1 + (s->mb_y-1)*s->mb_stride;
+
+ if( mbn_xy >= 0 && h->last_qscale_diff != 0 && ( IS_INTRA16x16(s->current_picture.mb_type[mbn_xy] ) || (h->cbp_table[mbn_xy]&0x3f) ) )
+ ctx++;
+
+ while( get_cabac( &h->cabac, &h->cabac_state[60 + ctx] ) ) {
+ if( ctx < 2 )
+ ctx = 2;
+ else
+ ctx = 3;
+ val++;
}
- sps->ref_frame_count= get_ue_golomb(&s->gb);
- sps->required_frame_num_update_behaviour_flag= get_bits1(&s->gb);
- sps->mb_width= get_ue_golomb(&s->gb) + 1;
- sps->mb_height= get_ue_golomb(&s->gb) + 1;
- sps->frame_mbs_only_flag= get_bits1(&s->gb);
- if(!sps->frame_mbs_only_flag)
- sps->mb_aff= get_bits1(&s->gb);
+ if( val&0x01 )
+ return (val + 1)/2;
else
- sps->mb_aff= 0;
+ return -(val + 1)/2;
+}
+static int decode_cabac_p_mb_sub_type( H264Context *h ) {
+ if( get_cabac( &h->cabac, &h->cabac_state[21] ) )
+ return 0; /* 8x8 */
+ if( !get_cabac( &h->cabac, &h->cabac_state[22] ) )
+ return 1; /* 8x4 */
+ if( get_cabac( &h->cabac, &h->cabac_state[23] ) )
+ return 2; /* 4x8 */
+ return 3; /* 4x4 */
+}
+static int decode_cabac_b_mb_sub_type( H264Context *h ) {
+ int type;
+ if( !get_cabac( &h->cabac, &h->cabac_state[36] ) )
+ return 0; /* B_Direct_8x8 */
+ if( !get_cabac( &h->cabac, &h->cabac_state[37] ) )
+ return 1 + get_cabac( &h->cabac, &h->cabac_state[39] ); /* B_L0_8x8, B_L1_8x8 */
+ type = 3;
+ if( get_cabac( &h->cabac, &h->cabac_state[38] ) ) {
+ if( get_cabac( &h->cabac, &h->cabac_state[39] ) )
+ return 11 + get_cabac( &h->cabac, &h->cabac_state[39] ); /* B_L1_4x4, B_Bi_4x4 */
+ type += 4;
+ }
+ type += 2*get_cabac( &h->cabac, &h->cabac_state[39] );
+ type += get_cabac( &h->cabac, &h->cabac_state[39] );
+ return type;
+}
- sps->direct_8x8_inference_flag= get_bits1(&s->gb);
+static int decode_cabac_mb_ref( H264Context *h, int list, int n ) {
+ int refa = h->ref_cache[list][scan8[n] - 1];
+ int refb = h->ref_cache[list][scan8[n] - 8];
+ int ref = 0;
+ int ctx = 0;
+
+ if( h->slice_type == B_TYPE) {
+ if( refa > 0 && !h->direct_cache[scan8[n] - 1] )
+ ctx++;
+ if( refb > 0 && !h->direct_cache[scan8[n] - 8] )
+ ctx += 2;
+ } else {
+ if( refa > 0 )
+ ctx++;
+ if( refb > 0 )
+ ctx += 2;
+ }
- sps->vui_parameters_present_flag= get_bits1(&s->gb);
- if( sps->vui_parameters_present_flag )
- decode_vui_parameters(h, sps);
-
- if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- printf("sps:%d profile:%d/%d poc:%d ref:%d %dx%d %s %s %s\n",
- sps_id, sps->profile_idc, sps->level_idc,
- sps->poc_type,
- sps->ref_frame_count,
- sps->mb_width, sps->mb_height,
- sps->frame_mbs_only_flag ? "FRM" : (sps->mb_aff ? "MB-AFF" : "PIC-AFF"),
- sps->direct_8x8_inference_flag ? "8B8" : "",
- sps->vui_parameters_present_flag ? "VUI" : ""
- );
+ while( get_cabac( &h->cabac, &h->cabac_state[54+ctx] ) ) {
+ ref++;
+ if( ctx < 4 )
+ ctx = 4;
+ else
+ ctx = 5;
}
- return 0;
+ return ref;
}
-static inline int decode_picture_parameter_set(H264Context *h){
- MpegEncContext * const s = &h->s;
- int pps_id= get_ue_golomb(&s->gb);
- PPS *pps= &h->pps_buffer[pps_id];
+static int decode_cabac_mb_mvd( H264Context *h, int list, int n, int l ) {
+ int amvd = abs( h->mvd_cache[list][scan8[n] - 1][l] ) +
+ abs( h->mvd_cache[list][scan8[n] - 8][l] );
+ int ctxbase = (l == 0) ? 40 : 47;
+ int ctx, mvd;
+
+ if( amvd < 3 )
+ ctx = 0;
+ else if( amvd > 32 )
+ ctx = 2;
+ else
+ ctx = 1;
+
+ if(!get_cabac(&h->cabac, &h->cabac_state[ctxbase+ctx]))
+ return 0;
+
+ mvd= 1;
+ ctx= 3;
+ while( mvd < 9 && get_cabac( &h->cabac, &h->cabac_state[ctxbase+ctx] ) ) {
+ mvd++;
+ if( ctx < 6 )
+ ctx++;
+ }
+
+ if( mvd >= 9 ) {
+ int k = 3;
+ while( get_cabac_bypass( &h->cabac ) ) {
+ mvd += 1 << k;
+ k++;
+ }
+ while( k-- ) {
+ if( get_cabac_bypass( &h->cabac ) )
+ mvd += 1 << k;
+ }
+ }
+ if( get_cabac_bypass( &h->cabac ) ) return -mvd;
+ else return mvd;
+}
+
+static int inline get_cabac_cbf_ctx( H264Context *h, int cat, int idx ) {
+ int nza, nzb;
+ int ctx = 0;
+
+ if( cat == 0 ) {
+ nza = h->left_cbp&0x100;
+ nzb = h-> top_cbp&0x100;
+ } else if( cat == 1 || cat == 2 ) {
+ nza = h->non_zero_count_cache[scan8[idx] - 1];
+ nzb = h->non_zero_count_cache[scan8[idx] - 8];
+ } else if( cat == 3 ) {
+ nza = (h->left_cbp>>(6+idx))&0x01;
+ nzb = (h-> top_cbp>>(6+idx))&0x01;
+ } else {
+ assert(cat == 4);
+ nza = h->non_zero_count_cache[scan8[16+idx] - 1];
+ nzb = h->non_zero_count_cache[scan8[16+idx] - 8];
+ }
+
+ if( nza > 0 )
+ ctx++;
+
+ if( nzb > 0 )
+ ctx += 2;
+
+ return ctx + 4 * cat;
+}
+
+static int inline decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int qp, int max_coeff) {
+ const int mb_xy = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
+ const uint16_t *qmul= dequant_coeff[qp];
+ static const int significant_coeff_flag_offset[5] = { 0, 15, 29, 44, 47 };
+ static const int coeff_abs_level_m1_offset[5] = {227+ 0, 227+10, 227+20, 227+30, 227+39 };
+
+ int index[16];
+
+ int i, last;
+ int coeff_count = 0;
+
+ int abslevel1 = 1;
+ int abslevelgt1 = 0;
+
+ /* cat: 0-> DC 16x16 n = 0
+ * 1-> AC 16x16 n = luma4x4idx
+ * 2-> Luma4x4 n = luma4x4idx
+ * 3-> DC Chroma n = iCbCr
+ * 4-> AC Chroma n = 4 * iCbCr + chroma4x4idx
+ */
+
+ /* read coded block flag */
+ if( get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n ) ] ) == 0 ) {
+ if( cat == 1 || cat == 2 )
+ h->non_zero_count_cache[scan8[n]] = 0;
+ else if( cat == 4 )
+ h->non_zero_count_cache[scan8[16+n]] = 0;
+
+ return 0;
+ }
+
+ for(last= 0; last < max_coeff - 1; last++) {
+ if( get_cabac( &h->cabac, &h->cabac_state[105+significant_coeff_flag_offset[cat]+last] )) {
+ index[coeff_count++] = last;
+ if( get_cabac( &h->cabac, &h->cabac_state[166+significant_coeff_flag_offset[cat]+last] ) ) {
+ last= max_coeff;
+ break;
+ }
+ }
+ }
+ if( last == max_coeff -1 ) {
+ index[coeff_count++] = last;
+ }
+ assert(coeff_count > 0);
+
+ if( cat == 0 )
+ h->cbp_table[mb_xy] |= 0x100;
+ else if( cat == 1 || cat == 2 )
+ h->non_zero_count_cache[scan8[n]] = coeff_count;
+ else if( cat == 3 )
+ h->cbp_table[mb_xy] |= 0x40 << n;
+ else {
+ assert( cat == 4 );
+ h->non_zero_count_cache[scan8[16+n]] = coeff_count;
+ }
+
+ for( i = coeff_count - 1; i >= 0; i-- ) {
+ int ctx = (abslevelgt1 != 0 ? 0 : FFMIN( 4, abslevel1 )) + coeff_abs_level_m1_offset[cat];
+ int j= scantable[index[i]];
+
+ if( get_cabac( &h->cabac, &h->cabac_state[ctx] ) == 0 ) {
+ if( cat == 0 || cat == 3 ) {
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = -1;
+ else block[j] = 1;
+ }else{
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = -qmul[j];
+ else block[j] = qmul[j];
+ }
+
+ abslevel1++;
+ } else {
+ int coeff_abs = 2;
+ ctx = 5 + FFMIN( 4, abslevelgt1 ) + coeff_abs_level_m1_offset[cat];
+ while( coeff_abs < 15 && get_cabac( &h->cabac, &h->cabac_state[ctx] ) ) {
+ coeff_abs++;
+ }
+
+ if( coeff_abs >= 15 ) {
+ int j = 0;
+ while( get_cabac_bypass( &h->cabac ) ) {
+ coeff_abs += 1 << j;
+ j++;
+ }
+
+ while( j-- ) {
+ if( get_cabac_bypass( &h->cabac ) )
+ coeff_abs += 1 << j ;
+ }
+ }
+
+ if( cat == 0 || cat == 3 ) {
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs;
+ else block[j] = coeff_abs;
+ }else{
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs * qmul[j];
+ else block[j] = coeff_abs * qmul[j];
+ }
+
+ abslevelgt1++;
+ }
+ }
+ return 0;
+}
+
+/**
+ * decodes a macroblock
+ * @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
+ */
+static int decode_mb_cabac(H264Context *h) {
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
+ int mb_type, partition_count, cbp = 0;
+
+ s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?)
+
+ if( h->sps.mb_aff ) {
+ av_log( h->s.avctx, AV_LOG_ERROR, "Fields not supported with CABAC\n" );
+ return -1;
+ }
+
+ if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE ) {
+ /* read skip flags */
+ if( decode_cabac_mb_skip( h ) ) {
+ decode_mb_skip(h);
+
+ h->cbp_table[mb_xy] = 0;
+ h->chroma_pred_mode_table[mb_xy] = 0;
+ h->last_qscale_diff = 0;
+
+ return 0;
+
+ }
+ }
+ h->prev_mb_skiped = 0;
+
+ if( ( mb_type = decode_cabac_mb_type( h ) ) < 0 ) {
+ av_log( h->s.avctx, AV_LOG_ERROR, "decode_cabac_mb_type failed\n" );
+ return -1;
+ }
+
+ if( h->slice_type == B_TYPE ) {
+ if( mb_type < 23 ){
+ partition_count= b_mb_type_info[mb_type].partition_count;
+ mb_type= b_mb_type_info[mb_type].type;
+ }else{
+ mb_type -= 23;
+ goto decode_intra_mb;
+ }
+ } else if( h->slice_type == P_TYPE ) {
+ if( mb_type < 5) {
+ partition_count= p_mb_type_info[mb_type].partition_count;
+ mb_type= p_mb_type_info[mb_type].type;
+ } else {
+ mb_type -= 5;
+ goto decode_intra_mb;
+ }
+ } else {
+ assert(h->slice_type == I_TYPE);
+decode_intra_mb:
+ partition_count = 0;
+ cbp= i_mb_type_info[mb_type].cbp;
+ h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
+ mb_type= i_mb_type_info[mb_type].type;
+ }
+#if 0
+ if(h->mb_field_decoding_flag)
+ mb_type |= MB_TYPE_INTERLACED;
+#endif
+
+ s->current_picture.mb_type[mb_xy]= mb_type;
+ h->slice_table[ mb_xy ]= h->slice_num;
+
+ if(IS_INTRA_PCM(mb_type)) {
+ /* TODO */
+ assert(0);
+ h->cbp_table[mb_xy] = 0xf +4*2; //FIXME ?!
+ h->cbp_table[mb_xy] |= 0x1C0;
+ h->chroma_pred_mode_table[mb_xy] = 0;
+ s->current_picture.qscale_table[mb_xy]= s->qscale;
+ return -1;
+ }
+
+ fill_caches(h, mb_type);
+
+ if( IS_INTRA( mb_type ) ) {
+ if( IS_INTRA4x4( mb_type ) ) {
+ int i;
+ for( i = 0; i < 16; i++ ) {
+ int pred = pred_intra_mode( h, i );
+ h->intra4x4_pred_mode_cache[ scan8[i] ] = decode_cabac_mb_intra4x4_pred_mode( h, pred );
+
+ //av_log( s->avctx, AV_LOG_ERROR, "i4x4 pred=%d mode=%d\n", pred, h->intra4x4_pred_mode_cache[ scan8[i] ] );
+ }
+ write_back_intra_pred_mode(h);
+ if( check_intra4x4_pred_mode(h) < 0 ) return -1;
+ } else {
+ h->intra16x16_pred_mode= check_intra_pred_mode( h, h->intra16x16_pred_mode );
+ if( h->intra16x16_pred_mode < 0 ) return -1;
+ }
+ h->chroma_pred_mode_table[mb_xy] =
+ h->chroma_pred_mode = decode_cabac_mb_chroma_pre_mode( h );
+
+ h->chroma_pred_mode= check_intra_pred_mode( h, h->chroma_pred_mode );
+ if( h->chroma_pred_mode < 0 ) return -1;
+ } else if( partition_count == 4 ) {
+ int i, j, sub_partition_count[4], list, ref[2][4];
+
+ if( h->slice_type == B_TYPE ) {
+ for( i = 0; i < 4; i++ ) {
+ h->sub_mb_type[i] = decode_cabac_b_mb_sub_type( h );
+ sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
+ h->sub_mb_type[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
+ }
+ if( IS_DIRECT(h->sub_mb_type[0]) || IS_DIRECT(h->sub_mb_type[1])
+ || IS_DIRECT(h->sub_mb_type[2]) || IS_DIRECT(h->sub_mb_type[3])) {
+ pred_direct_motion(h, &mb_type);
+ if( h->ref_count[0] > 1 || h->ref_count[1] > 1 ) {
+ for( i = 0; i < 4; i++ )
+ if( IS_DIRECT(h->sub_mb_type[i]) )
+ fill_rectangle( &h->direct_cache[scan8[4*i]], 2, 2, 8, 1, 1 );
+ }
+ }
+ } else {
+ for( i = 0; i < 4; i++ ) {
+ h->sub_mb_type[i] = decode_cabac_p_mb_sub_type( h );
+ sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
+ h->sub_mb_type[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
+ }
+ }
+
+ for( list = 0; list < 2; list++ ) {
+ if( h->ref_count[list] > 0 ) {
+ for( i = 0; i < 4; i++ ) {
+ if(IS_DIRECT(h->sub_mb_type[i])) continue;
+ if(IS_DIR(h->sub_mb_type[i], 0, list)){
+ if( h->ref_count[list] > 1 )
+ ref[list][i] = decode_cabac_mb_ref( h, list, 4*i );
+ else
+ ref[list][i] = 0;
+ } else {
+ ref[list][i] = -1;
+ }
+ h->ref_cache[list][ scan8[4*i]+1 ]=
+ h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
+ }
+ }
+ }
+
+ for(list=0; list<2; list++){
+ for(i=0; i<4; i++){
+ if(IS_DIRECT(h->sub_mb_type[i])){
+ fill_rectangle(h->mvd_cache[list][scan8[4*i]], 2, 2, 8, 0, 4);
+ continue;
+ }
+ h->ref_cache[list][ scan8[4*i] ]=h->ref_cache[list][ scan8[4*i]+1 ];
+
+ if(IS_DIR(h->sub_mb_type[i], 0, list) && !IS_DIRECT(h->sub_mb_type[i])){
+ const int sub_mb_type= h->sub_mb_type[i];
+ const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
+ for(j=0; j<sub_partition_count[i]; j++){
+ int mpx, mpy;
+ int mx, my;
+ const int index= 4*i + block_width*j;
+ int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
+ int16_t (* mvd_cache)[2]= &h->mvd_cache[list][ scan8[index] ];
+ pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mpx, &mpy);
+
+ mx = mpx + decode_cabac_mb_mvd( h, list, index, 0 );
+ my = mpy + decode_cabac_mb_mvd( h, list, index, 1 );
+ tprintf("final mv:%d %d\n", mx, my);
+
+ if(IS_SUB_8X8(sub_mb_type)){
+ mv_cache[ 0 ][0]= mv_cache[ 1 ][0]=
+ mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
+ mv_cache[ 0 ][1]= mv_cache[ 1 ][1]=
+ mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;
+
+ mvd_cache[ 0 ][0]= mvd_cache[ 1 ][0]=
+ mvd_cache[ 8 ][0]= mvd_cache[ 9 ][0]= mx - mpx;
+ mvd_cache[ 0 ][1]= mvd_cache[ 1 ][1]=
+ mvd_cache[ 8 ][1]= mvd_cache[ 9 ][1]= my - mpy;
+ }else if(IS_SUB_8X4(sub_mb_type)){
+ mv_cache[ 0 ][0]= mv_cache[ 1 ][0]= mx;
+ mv_cache[ 0 ][1]= mv_cache[ 1 ][1]= my;
+
+ mvd_cache[ 0 ][0]= mvd_cache[ 1 ][0]= mx- mpx;
+ mvd_cache[ 0 ][1]= mvd_cache[ 1 ][1]= my - mpy;
+ }else if(IS_SUB_4X8(sub_mb_type)){
+ mv_cache[ 0 ][0]= mv_cache[ 8 ][0]= mx;
+ mv_cache[ 0 ][1]= mv_cache[ 8 ][1]= my;
+
+ mvd_cache[ 0 ][0]= mvd_cache[ 8 ][0]= mx - mpx;
+ mvd_cache[ 0 ][1]= mvd_cache[ 8 ][1]= my - mpy;
+ }else{
+ assert(IS_SUB_4X4(sub_mb_type));
+ mv_cache[ 0 ][0]= mx;
+ mv_cache[ 0 ][1]= my;
+
+ mvd_cache[ 0 ][0]= mx - mpx;
+ mvd_cache[ 0 ][1]= my - mpy;
+ }
+ }
+ }else{
+ uint32_t *p= (uint32_t *)&h->mv_cache[list][ scan8[4*i] ][0];
+ uint32_t *pd= (uint32_t *)&h->mvd_cache[list][ scan8[4*i] ][0];
+ p[0] = p[1] = p[8] = p[9] = 0;
+ pd[0]= pd[1]= pd[8]= pd[9]= 0;
+ }
+ }
+ }
+ } else if( IS_DIRECT(mb_type) ) {
+ pred_direct_motion(h, &mb_type);
+ s->current_picture.mb_type[mb_xy]= mb_type;
+ fill_rectangle(h->mvd_cache[0][scan8[0]], 4, 4, 8, 0, 4);
+ fill_rectangle(h->mvd_cache[1][scan8[0]], 4, 4, 8, 0, 4);
+ } else {
+ int list, mx, my, i, mpx, mpy;
+ if(IS_16X16(mb_type)){
+ for(list=0; list<2; list++){
+ if(IS_DIR(mb_type, 0, list)){
+ if(h->ref_count[list] > 0 ){
+ const int ref = h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 0 ) : 0;
+ fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, ref, 1);
+ }
+ }
+ }
+ for(list=0; list<2; list++){
+ if(IS_DIR(mb_type, 0, list)){
+ pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mpx, &mpy);
+
+ mx = mpx + decode_cabac_mb_mvd( h, list, 0, 0 );
+ my = mpy + decode_cabac_mb_mvd( h, list, 0, 1 );
+ tprintf("final mv:%d %d\n", mx, my);
+
+ fill_rectangle(h->mvd_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
+ fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
+ }
+ }
+ }
+ else if(IS_16X8(mb_type)){
+ for(list=0; list<2; list++){
+ if(h->ref_count[list]>0){
+ for(i=0; i<2; i++){
+ if(IS_DIR(mb_type, i, list)){
+ const int ref= h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 8*i ) : 0;
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, ref, 1);
+ }else
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, (LIST_NOT_USED&0xFF), 1);
+ }
+ }
+ }
+ for(list=0; list<2; list++){
+ for(i=0; i<2; i++){
+ if(IS_DIR(mb_type, i, list)){
+ pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mpx, &mpy);
+ mx = mpx + decode_cabac_mb_mvd( h, list, 8*i, 0 );
+ my = mpy + decode_cabac_mb_mvd( h, list, 8*i, 1 );
+ tprintf("final mv:%d %d\n", mx, my);
+
+ fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx-mpx,my-mpy), 4);
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
+ }else{ // needed only for mixed refs
+ fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, 0, 4);
+ fill_rectangle(h-> mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, 0, 4);
+ }
+ }
+ }
+ }else{
+ assert(IS_8X16(mb_type));
+ for(list=0; list<2; list++){
+ if(h->ref_count[list]>0){
+ for(i=0; i<2; i++){
+ if(IS_DIR(mb_type, i, list)){ //FIXME optimize
+ const int ref= h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 4*i ) : 0;
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, ref, 1);
+ }else
+ fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, (LIST_NOT_USED&0xFF), 1);
+ }
+ }
+ }
+ for(list=0; list<2; list++){
+ for(i=0; i<2; i++){
+ if(IS_DIR(mb_type, i, list)){
+ pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mpx, &mpy);
+ mx = mpx + decode_cabac_mb_mvd( h, list, 4*i, 0 );
+ my = mpy + decode_cabac_mb_mvd( h, list, 4*i, 1 );
+
+ tprintf("final mv:%d %d\n", mx, my);
+ fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
+ fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
+ }else{ // needed only for mixed refs
+ fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, 0, 4);
+ fill_rectangle(h-> mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, 0, 4);
+ }
+ }
+ }
+ }
+ }
+
+ if( IS_INTER( mb_type ) ) {
+ h->chroma_pred_mode_table[mb_xy] = 0;
+ write_back_motion( h, mb_type );
+ }
+
+ if( !IS_INTRA16x16( mb_type ) ) {
+ cbp = decode_cabac_mb_cbp_luma( h );
+ cbp |= decode_cabac_mb_cbp_chroma( h ) << 4;
+ }
+
+ h->cbp_table[mb_xy] = cbp;
+
+ if( cbp || IS_INTRA16x16( mb_type ) ) {
+ const uint8_t *scan, *dc_scan;
+ int dqp;
+
+ if(IS_INTERLACED(mb_type)){
+ scan= field_scan;
+ dc_scan= luma_dc_field_scan;
+ }else{
+ scan= zigzag_scan;
+ dc_scan= luma_dc_zigzag_scan;
+ }
+
+ h->last_qscale_diff = dqp = decode_cabac_mb_dqp( h );
+ s->qscale += dqp;
+ if(((unsigned)s->qscale) > 51){
+ if(s->qscale<0) s->qscale+= 52;
+ else s->qscale-= 52;
+ }
+ h->chroma_qp = get_chroma_qp(h, s->qscale);
+
+ if( IS_INTRA16x16( mb_type ) ) {
+ int i;
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" );
+ if( decode_cabac_residual( h, h->mb, 0, 0, dc_scan, s->qscale, 16) < 0)
+ return -1;
+ if( cbp&15 ) {
+ for( i = 0; i < 16; i++ ) {
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i );
+ if( decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, s->qscale, 15) < 0 )
+ return -1;
+ }
+ } else {
+ fill_rectangle(&h->non_zero_count_cache[scan8[0]], 4, 4, 8, 0, 1);
+ }
+ } else {
+ int i8x8, i4x4;
+ for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
+ if( cbp & (1<<i8x8) ) {
+ for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
+ const int index = 4*i8x8 + i4x4;
+ //av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index );
+ if( decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, s->qscale, 16) < 0 )
+ return -1;
+ }
+ } else {
+ uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
+ nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
+ }
+ }
+ }
+
+ if( cbp&0x30 ){
+ int c;
+ for( c = 0; c < 2; c++ ) {
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
+ if( decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, h->chroma_qp, 4) < 0)
+ return -1;
+ }
+ }
+
+ if( cbp&0x20 ) {
+ int c, i;
+ for( c = 0; c < 2; c++ ) {
+ for( i = 0; i < 4; i++ ) {
+ const int index = 16 + 4 * c + i;
+ //av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
+ if( decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, h->chroma_qp, 15) < 0)
+ return -1;
+ }
+ }
+ } else {
+ uint8_t * const nnz= &h->non_zero_count_cache[0];
+ nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
+ nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
+ }
+ } else {
+ uint8_t * const nnz= &h->non_zero_count_cache[0];
+ fill_rectangle(&nnz[scan8[0]], 4, 4, 8, 0, 1);
+ nnz[ scan8[16]+0 ] = nnz[ scan8[16]+1 ] =nnz[ scan8[16]+8 ] =nnz[ scan8[16]+9 ] =
+ nnz[ scan8[20]+0 ] = nnz[ scan8[20]+1 ] =nnz[ scan8[20]+8 ] =nnz[ scan8[20]+9 ] = 0;
+ }
+
+ s->current_picture.qscale_table[mb_xy]= s->qscale;
+ write_back_non_zero_count(h);
+
+ return 0;
+}
+
+
+static void filter_mb_edgev( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
+ int i, d;
+ const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
+
+ for( i = 0; i < 4; i++ ) {
+ if( bS[i] == 0 ) {
+ pix += 4 * stride;
+ continue;
+ }
+
+ if( bS[i] < 4 ) {
+ const int tc0 = tc0_table[index_a][bS[i] - 1];
+ /* 4px edge length */
+ for( d = 0; d < 4; d++ ) {
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int p2 = pix[-3];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+ const int q2 = pix[2];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+ int tc = tc0;
+ int i_delta;
+
+ if( ABS( p2 - p0 ) < beta ) {
+ pix[-2] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
+ }
+ if( ABS( q2 - q0 ) < beta ) {
+ pix[1] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
+ }
+
+ i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+ pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ }
+ pix += stride;
+ }
+ }else{
+ /* 4px edge length */
+ for( d = 0; d < 4; d++ ) {
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int p2 = pix[-3];
+
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+ const int q2 = pix[2];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ if(ABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
+ if( ABS( p2 - p0 ) < beta)
+ {
+ const int p3 = pix[-4];
+ /* p0', p1', p2' */
+ pix[-1] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
+ pix[-2] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
+ pix[-3] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
+ } else {
+ /* p0' */
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ }
+ if( ABS( q2 - q0 ) < beta)
+ {
+ const int q3 = pix[3];
+ /* q0', q1', q2' */
+ pix[0] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
+ pix[1] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
+ pix[2] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
+ } else {
+ /* q0' */
+ pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ }else{
+ /* p0', q0' */
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ pix[ 0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ }
+ pix += stride;
+ }
+ }
+ }
+}
+static void filter_mb_edgecv( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
+ int i, d;
+ const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
+
+ for( i = 0; i < 4; i++ ) {
+ if( bS[i] == 0 ) {
+ pix += 2 * stride;
+ continue;
+ }
+
+ if( bS[i] < 4 ) {
+ const int tc = tc0_table[index_a][bS[i] - 1] + 1;
+ /* 2px edge length (because we use same bS than the one for luma) */
+ for( d = 0; d < 2; d++ ){
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+ const int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+
+ pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ }
+ pix += stride;
+ }
+ }else{
+ /* 2px edge length (because we use same bS than the one for luma) */
+ for( d = 0; d < 2; d++ ){
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
+ pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
+ }
+ pix += stride;
+ }
+ }
+ }
+}
+
+static void filter_mb_edgeh( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
+ int i, d;
+ const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
+ const int pix_next = stride;
+
+ for( i = 0; i < 4; i++ ) {
+ if( bS[i] == 0 ) {
+ pix += 4;
+ continue;
+ }
+
+ if( bS[i] < 4 ) {
+ const int tc0 = tc0_table[index_a][bS[i] - 1];
+ /* 4px edge length */
+ for( d = 0; d < 4; d++ ) {
+ const int p0 = pix[-1*pix_next];
+ const int p1 = pix[-2*pix_next];
+ const int p2 = pix[-3*pix_next];
+ const int q0 = pix[0];
+ const int q1 = pix[1*pix_next];
+ const int q2 = pix[2*pix_next];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ int tc = tc0;
+ int i_delta;
+
+ if( ABS( p2 - p0 ) < beta ) {
+ pix[-2*pix_next] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
+ }
+ if( ABS( q2 - q0 ) < beta ) {
+ pix[pix_next] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
+ }
+
+ i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+ pix[-pix_next] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ }
+ pix++;
+ }
+ }else{
+ /* 4px edge length */
+ for( d = 0; d < 4; d++ ) {
+ const int p0 = pix[-1*pix_next];
+ const int p1 = pix[-2*pix_next];
+ const int p2 = pix[-3*pix_next];
+ const int q0 = pix[0];
+ const int q1 = pix[1*pix_next];
+ const int q2 = pix[2*pix_next];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ const int p3 = pix[-4*pix_next];
+ const int q3 = pix[ 3*pix_next];
+
+ if(ABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
+ if( ABS( p2 - p0 ) < beta) {
+ /* p0', p1', p2' */
+ pix[-1*pix_next] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
+ pix[-2*pix_next] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
+ pix[-3*pix_next] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
+ } else {
+ /* p0' */
+ pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ }
+ if( ABS( q2 - q0 ) < beta) {
+ /* q0', q1', q2' */
+ pix[0*pix_next] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
+ pix[1*pix_next] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
+ pix[2*pix_next] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
+ } else {
+ /* q0' */
+ pix[0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ }else{
+ /* p0', q0' */
+ pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ pix[ 0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ }
+ pix++;
+ }
+ }
+ }
+}
+
+static void filter_mb_edgech( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
+ int i, d;
+ const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
+ const int pix_next = stride;
+
+ for( i = 0; i < 4; i++ )
+ {
+ if( bS[i] == 0 ) {
+ pix += 2;
+ continue;
+ }
+
+ if( bS[i] < 4 ) {
+ int tc = tc0_table[index_a][bS[i] - 1] + 1;
+ /* 2px edge length (see deblocking_filter_edgecv) */
+ for( d = 0; d < 2; d++ ) {
+ const int p0 = pix[-1*pix_next];
+ const int p1 = pix[-2*pix_next];
+ const int q0 = pix[0];
+ const int q1 = pix[1*pix_next];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+
+ pix[-pix_next] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ }
+ pix++;
+ }
+ }else{
+ /* 2px edge length (see deblocking_filter_edgecv) */
+ for( d = 0; d < 2; d++ ) {
+ const int p0 = pix[-1*pix_next];
+ const int p1 = pix[-2*pix_next];
+ const int q0 = pix[0];
+ const int q1 = pix[1*pix_next];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ pix[-pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
+ pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
+ }
+ pix++;
+ }
+ }
+ }
+}
+
+static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr) {
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= mb_x + mb_y*s->mb_stride;
+ int linesize, uvlinesize;
+ int dir;
+
+ /* FIXME Implement deblocking filter for field MB */
+ if( h->sps.mb_aff ) {
+ return;
+ }
+ linesize = s->linesize;
+ uvlinesize = s->uvlinesize;
+
+ /* dir : 0 -> vertical edge, 1 -> horizontal edge */
+ for( dir = 0; dir < 2; dir++ )
+ {
+ int start = 0;
+ int edge;
+
+ /* test picture boundary */
+ if( ( dir == 0 && mb_x == 0 ) || ( dir == 1 && mb_y == 0 ) ) {
+ start = 1;
+ }
+ /* FIXME test slice boundary */
+ if( h->deblocking_filter == 2 ) {
+ }
+
+ /* Calculate bS */
+ for( edge = start; edge < 4; edge++ ) {
+ /* mbn_xy: neighbour macroblock (how that works for field ?) */
+ int mbn_xy = edge > 0 ? mb_xy : ( dir == 0 ? mb_xy -1 : mb_xy - s->mb_stride );
+ int bS[4];
+ int qp;
+
+ if( IS_INTRA( s->current_picture.mb_type[mb_xy] ) ||
+ IS_INTRA( s->current_picture.mb_type[mbn_xy] ) ) {
+ bS[0] = bS[1] = bS[2] = bS[3] = ( edge == 0 ? 4 : 3 );
+ } else {
+ int i;
+ for( i = 0; i < 4; i++ ) {
+ int x = dir == 0 ? edge : i;
+ int y = dir == 0 ? i : edge;
+ int b_idx= 8 + 4 + x + 8*y;
+ int bn_idx= b_idx - (dir ? 8:1);
+
+ if( h->non_zero_count_cache[b_idx] != 0 ||
+ h->non_zero_count_cache[bn_idx] != 0 ) {
+ bS[i] = 2;
+ }
+ else if( h->slice_type == P_TYPE ) {
+ if( h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx] ||
+ ABS( h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] ) >= 4 ||
+ ABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= 4 )
+ bS[i] = 1;
+ else
+ bS[i] = 0;
+ }
+ else {
+ /* FIXME Add support for B frame */
+ return;
+ }
+ }
+
+ if(bS[0]+bS[1]+bS[2]+bS[3] == 0)
+ continue;
+ }
+
+ /* Filter edge */
+ qp = ( s->qscale + s->current_picture.qscale_table[mbn_xy] + 1 ) >> 1;
+ if( dir == 0 ) {
+ filter_mb_edgev( h, &img_y[4*edge], linesize, bS, qp );
+ if( (edge&1) == 0 ) {
+ int chroma_qp = ( h->chroma_qp +
+ get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ filter_mb_edgecv( h, &img_cb[2*edge], uvlinesize, bS, chroma_qp );
+ filter_mb_edgecv( h, &img_cr[2*edge], uvlinesize, bS, chroma_qp );
+ }
+ } else {
+ filter_mb_edgeh( h, &img_y[4*edge*linesize], linesize, bS, qp );
+ if( (edge&1) == 0 ) {
+ int chroma_qp = ( h->chroma_qp +
+ get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ filter_mb_edgech( h, &img_cb[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
+ filter_mb_edgech( h, &img_cr[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
+ }
+ }
+ }
+ }
+}
+
+static int decode_slice(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;
+
+ s->mb_skip_run= -1;
+
+ if( h->pps.cabac ) {
+ int i;
+
+ /* realign */
+ align_get_bits( &s->gb );
+
+ /* init cabac */
+ ff_init_cabac_states( &h->cabac, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64 );
+ ff_init_cabac_decoder( &h->cabac,
+ s->gb.buffer + get_bits_count(&s->gb)/8,
+ ( s->gb.size_in_bits - get_bits_count(&s->gb) + 7)/8);
+ /* calculate pre-state */
+ for( i= 0; i < 399; i++ ) {
+ int pre;
+ if( h->slice_type == I_TYPE )
+ pre = clip( ((cabac_context_init_I[i][0] * s->qscale) >>4 ) + cabac_context_init_I[i][1], 1, 126 );
+ else
+ pre = clip( ((cabac_context_init_PB[h->cabac_init_idc][i][0] * s->qscale) >>4 ) + cabac_context_init_PB[h->cabac_init_idc][i][1], 1, 126 );
+
+ if( pre <= 63 )
+ h->cabac_state[i] = 2 * ( 63 - pre ) + 0;
+ else
+ h->cabac_state[i] = 2 * ( pre - 64 ) + 1;
+ }
+
+ for(;;){
+ int ret = decode_mb_cabac(h);
+ int eos = get_cabac_terminate( &h->cabac ); /* End of Slice flag */
+
+ if(ret>=0) hl_decode_mb(h);
+
+ /* XXX: useless as decode_mb_cabac it doesn't support that ... */
+ if( ret >= 0 && h->sps.mb_aff ) { //FIXME optimal? or let mb_decode decode 16x32 ?
+ s->mb_y++;
+
+ if(ret>=0) ret = decode_mb_cabac(h);
+ eos = get_cabac_terminate( &h->cabac );
+
+ hl_decode_mb(h);
+ s->mb_y--;
+ }
+
+ if( ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 1) {
+ av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+ return -1;
+ }
+
+ if( ++s->mb_x >= s->mb_width ) {
+ s->mb_x = 0;
+ ff_draw_horiz_band(s, 16*s->mb_y, 16);
+ ++s->mb_y;
+ }
+
+ if( eos || s->mb_y >= s->mb_height ) {
+ tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+ return 0;
+ }
+#if 0
+ /* TODO test over-reading in cabac code */
+ else if( read too much in h->cabac ) {
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+ return -1;
+ }
+#endif
+ }
+
+ } else {
+ for(;;){
+ int ret = decode_mb_cavlc(h);
+
+ if(ret>=0) hl_decode_mb(h);
+
+ if(ret>=0 && h->sps.mb_aff){ //FIXME optimal? or let mb_decode decode 16x32 ?
+ s->mb_y++;
+ ret = decode_mb_cavlc(h);
+
+ if(ret>=0) hl_decode_mb(h);
+ s->mb_y--;
+ }
+
+ if(ret<0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+
+ return -1;
+ }
+
+ if(++s->mb_x >= s->mb_width){
+ s->mb_x=0;
+ ff_draw_horiz_band(s, 16*s->mb_y, 16);
+ if(++s->mb_y >= s->mb_height){
+ tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
+
+ if(get_bits_count(&s->gb) == s->gb.size_in_bits ) {
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return 0;
+ }else{
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return -1;
+ }
+ }
+ }
+
+ if(get_bits_count(&s->gb) >= s->gb.size_in_bits && s->mb_skip_run<=0){
+ tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
+ if(get_bits_count(&s->gb) == s->gb.size_in_bits ){
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return 0;
+ }else{
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+
+ return -1;
+ }
+ }
+ }
+ }
+
+#if 0
+ for(;s->mb_y < s->mb_height; s->mb_y++){
+ for(;s->mb_x < s->mb_width; s->mb_x++){
+ int ret= decode_mb(h);
+
+ hl_decode_mb(h);
+
+ if(ret<0){
+ fprintf(stderr, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+
+ return -1;
+ }
+
+ if(++s->mb_x >= s->mb_width){
+ s->mb_x=0;
+ if(++s->mb_y >= s->mb_height){
+ if(get_bits_count(s->gb) == s->gb.size_in_bits){
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return 0;
+ }else{
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return -1;
+ }
+ }
+ }
+
+ if(get_bits_count(s->?gb) >= s->gb?.size_in_bits){
+ if(get_bits_count(s->gb) == s->gb.size_in_bits){
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
+
+ return 0;
+ }else{
+ ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
+
+ return -1;
+ }
+ }
+ }
+ s->mb_x=0;
+ ff_draw_horiz_band(s, 16*s->mb_y, 16);
+ }
+#endif
+ return -1; //not reached
+}
+
+static inline int decode_vui_parameters(H264Context *h, SPS *sps){
+ MpegEncContext * const s = &h->s;
+ int aspect_ratio_info_present_flag, aspect_ratio_idc;
+
+ aspect_ratio_info_present_flag= get_bits1(&s->gb);
+
+ if( aspect_ratio_info_present_flag ) {
+ aspect_ratio_idc= get_bits(&s->gb, 8);
+ if( aspect_ratio_idc == EXTENDED_SAR ) {
+ sps->sar.num= get_bits(&s->gb, 16);
+ sps->sar.den= get_bits(&s->gb, 16);
+ }else if(aspect_ratio_idc < 16){
+ sps->sar= pixel_aspect[aspect_ratio_idc];
+ }else{
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal aspect ratio\n");
+ return -1;
+ }
+ }else{
+ sps->sar.num=
+ sps->sar.den= 0;
+ }
+// s->avctx->aspect_ratio= sar_width*s->width / (float)(s->height*sar_height);
+
+ if(get_bits1(&s->gb)){ /* overscan_info_present_flag */
+ get_bits1(&s->gb); /* overscan_appropriate_flag */
+ }
+
+ if(get_bits1(&s->gb)){ /* video_signal_type_present_flag */
+ get_bits(&s->gb, 3); /* video_format */
+ get_bits1(&s->gb); /* video_full_range_flag */
+ if(get_bits1(&s->gb)){ /* colour_description_present_flag */
+ get_bits(&s->gb, 8); /* colour_primaries */
+ get_bits(&s->gb, 8); /* transfer_characteristics */
+ get_bits(&s->gb, 8); /* matrix_coefficients */
+ }
+ }
+
+ if(get_bits1(&s->gb)){ /* chroma_location_info_present_flag */
+ get_ue_golomb(&s->gb); /* chroma_sample_location_type_top_field */
+ get_ue_golomb(&s->gb); /* chroma_sample_location_type_bottom_field */
+ }
+
+ sps->timing_info_present_flag = get_bits1(&s->gb);
+ if(sps->timing_info_present_flag){
+ sps->num_units_in_tick = get_bits_long(&s->gb, 32);
+ sps->time_scale = get_bits_long(&s->gb, 32);
+ sps->fixed_frame_rate_flag = get_bits1(&s->gb);
+ }
+
+#if 0
+| nal_hrd_parameters_present_flag |0 |u(1) |
+| if( nal_hrd_parameters_present_flag = = 1) | | |
+| hrd_parameters( ) | | |
+| vcl_hrd_parameters_present_flag |0 |u(1) |
+| if( vcl_hrd_parameters_present_flag = = 1) | | |
+| hrd_parameters( ) | | |
+| if( ( nal_hrd_parameters_present_flag = = 1 | || | |
+| | | |
+|( vcl_hrd_parameters_present_flag = = 1 ) ) | | |
+| low_delay_hrd_flag |0 |u(1) |
+| bitstream_restriction_flag |0 |u(1) |
+| if( bitstream_restriction_flag ) { |0 |u(1) |
+| motion_vectors_over_pic_boundaries_flag |0 |u(1) |
+| max_bytes_per_pic_denom |0 |ue(v) |
+| max_bits_per_mb_denom |0 |ue(v) |
+| log2_max_mv_length_horizontal |0 |ue(v) |
+| log2_max_mv_length_vertical |0 |ue(v) |
+| num_reorder_frames |0 |ue(v) |
+| max_dec_frame_buffering |0 |ue(v) |
+| } | | |
+|} | | |
+#endif
+ return 0;
+}
+
+static inline int decode_seq_parameter_set(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ int profile_idc, level_idc;
+ int sps_id, i;
+ SPS *sps;
+
+ profile_idc= get_bits(&s->gb, 8);
+ get_bits1(&s->gb); //constraint_set0_flag
+ get_bits1(&s->gb); //constraint_set1_flag
+ get_bits1(&s->gb); //constraint_set2_flag
+ get_bits1(&s->gb); //constraint_set3_flag
+ get_bits(&s->gb, 4); // reserved
+ level_idc= get_bits(&s->gb, 8);
+ sps_id= get_ue_golomb(&s->gb);
+
+ sps= &h->sps_buffer[ sps_id ];
+ sps->profile_idc= profile_idc;
+ sps->level_idc= level_idc;
+
+ sps->log2_max_frame_num= get_ue_golomb(&s->gb) + 4;
+ sps->poc_type= get_ue_golomb(&s->gb);
+
+ if(sps->poc_type == 0){ //FIXME #define
+ sps->log2_max_poc_lsb= get_ue_golomb(&s->gb) + 4;
+ } else if(sps->poc_type == 1){//FIXME #define
+ sps->delta_pic_order_always_zero_flag= get_bits1(&s->gb);
+ sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
+ sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
+ sps->poc_cycle_length= get_ue_golomb(&s->gb);
+
+ for(i=0; i<sps->poc_cycle_length; i++)
+ sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
+ }
+ if(sps->poc_type > 2){
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal POC type %d\n", sps->poc_type);
+ return -1;
+ }
+
+ sps->ref_frame_count= get_ue_golomb(&s->gb);
+ if(sps->ref_frame_count > MAX_PICTURE_COUNT-2){
+ av_log(h->s.avctx, AV_LOG_ERROR, "too many reference frames\n");
+ }
+ sps->gaps_in_frame_num_allowed_flag= get_bits1(&s->gb);
+ sps->mb_width= get_ue_golomb(&s->gb) + 1;
+ sps->mb_height= get_ue_golomb(&s->gb) + 1;
+ sps->frame_mbs_only_flag= get_bits1(&s->gb);
+ if(!sps->frame_mbs_only_flag)
+ sps->mb_aff= get_bits1(&s->gb);
+ else
+ sps->mb_aff= 0;
+
+ sps->direct_8x8_inference_flag= get_bits1(&s->gb);
+
+ sps->crop= get_bits1(&s->gb);
+ if(sps->crop){
+ sps->crop_left = get_ue_golomb(&s->gb);
+ sps->crop_right = get_ue_golomb(&s->gb);
+ sps->crop_top = get_ue_golomb(&s->gb);
+ sps->crop_bottom= get_ue_golomb(&s->gb);
+ if(sps->crop_left || sps->crop_top){
+ av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completly supported, this could look slightly wrong ...\n");
+ }
+ }else{
+ sps->crop_left =
+ sps->crop_right =
+ sps->crop_top =
+ sps->crop_bottom= 0;
+ }
+
+ sps->vui_parameters_present_flag= get_bits1(&s->gb);
+ if( sps->vui_parameters_present_flag )
+ decode_vui_parameters(h, sps);
+
+ if(s->avctx->debug&FF_DEBUG_PICT_INFO){
+ av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%d profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s\n",
+ sps_id, sps->profile_idc, sps->level_idc,
+ sps->poc_type,
+ sps->ref_frame_count,
+ sps->mb_width, sps->mb_height,
+ sps->frame_mbs_only_flag ? "FRM" : (sps->mb_aff ? "MB-AFF" : "PIC-AFF"),
+ sps->direct_8x8_inference_flag ? "8B8" : "",
+ sps->crop_left, sps->crop_right,
+ sps->crop_top, sps->crop_bottom,
+ sps->vui_parameters_present_flag ? "VUI" : ""
+ );
+ }
+ return 0;
+}
+
+static inline int decode_picture_parameter_set(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ int pps_id= get_ue_golomb(&s->gb);
+ PPS *pps= &h->pps_buffer[pps_id];
pps->sps_id= get_ue_golomb(&s->gb);
pps->cabac= get_bits1(&s->gb);
pps->slice_group_count= get_ue_golomb(&s->gb) + 1;
if(pps->slice_group_count > 1 ){
pps->mb_slice_group_map_type= get_ue_golomb(&s->gb);
-fprintf(stderr, "FMO not supported\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "FMO not supported\n");
switch(pps->mb_slice_group_map_type){
case 0:
#if 0
|) | | |
| slice_group_id[ i ] |1 |u(v) |
#endif
+ break;
}
}
pps->ref_count[0]= get_ue_golomb(&s->gb) + 1;
pps->ref_count[1]= get_ue_golomb(&s->gb) + 1;
if(pps->ref_count[0] > 32 || pps->ref_count[1] > 32){
- fprintf(stderr, "reference overflow (pps)\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
return -1;
}
pps->deblocking_filter_parameters_present= get_bits1(&s->gb);
pps->constrained_intra_pred= get_bits1(&s->gb);
pps->redundant_pic_cnt_present = get_bits1(&s->gb);
- pps->crop= get_bits1(&s->gb);
- if(pps->crop){
- pps->crop_left = get_ue_golomb(&s->gb);
- pps->crop_right = get_ue_golomb(&s->gb);
- pps->crop_top = get_ue_golomb(&s->gb);
- pps->crop_bottom= get_ue_golomb(&s->gb);
- }else{
- pps->crop_left =
- pps->crop_right =
- pps->crop_top =
- pps->crop_bottom= 0;
- }
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- printf("pps:%d sps:%d %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d %s %s %s crop:%d/%d/%d/%d\n",
+ av_log(h->s.avctx, AV_LOG_DEBUG, "pps:%d sps:%d %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d %s %s %s\n",
pps_id, pps->sps_id,
pps->cabac ? "CABAC" : "CAVLC",
pps->slice_group_count,
pps->init_qp, pps->init_qs, pps->chroma_qp_index_offset,
pps->deblocking_filter_parameters_present ? "LPAR" : "",
pps->constrained_intra_pred ? "CONSTR" : "",
- pps->redundant_pic_cnt_present ? "REDU" : "",
- pps->crop_left, pps->crop_right,
- pps->crop_top, pps->crop_bottom
+ pps->redundant_pic_cnt_present ? "REDU" : ""
);
}
* finds the end of the current frame in the bitstream.
* @return the position of the first byte of the next frame, or -1
*/
-static int find_frame_end(MpegEncContext *s, uint8_t *buf, int buf_size){
- ParseContext *pc= &s->parse_context;
- int last_addr, i;
+static int find_frame_end(H264Context *h, const uint8_t *buf, int buf_size){
+ int i;
uint32_t state;
+ ParseContext *pc = &(h->s.parse_context);
//printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]);
// mb_addr= pc->mb_addr - 1;
state= pc->state;
- //FIXME this will fail with slices
- for(i=0; i<buf_size; i++){
- state= (state<<8) | buf[i];
+ for(i=0; i<=buf_size; i++){
if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
+ tprintf("find_frame_end new startcode = %08x, frame_start_found = %d, pos = %d\n", state, pc->frame_start_found, i);
if(pc->frame_start_found){
- pc->state=-1;
- pc->frame_start_found= 0;
- return i-3;
+ // If there isn't one more byte in the buffer
+ // the test on first_mb_in_slice cannot be done yet
+ // do it at next call.
+ if (i >= buf_size) break;
+ if (buf[i] & 0x80) {
+ // first_mb_in_slice is 0, probably the first nal of a new
+ // slice
+ tprintf("find_frame_end frame_end_found, state = %08x, pos = %d\n", state, i);
+ pc->state=-1;
+ pc->frame_start_found= 0;
+ return i-4;
+ }
}
- pc->frame_start_found= 1;
+ pc->frame_start_found = 1;
}
+ if (i<buf_size)
+ state= (state<<8) | buf[i];
}
pc->state= state;
- return -1;
+ return END_NOT_FOUND;
+}
+
+static int h264_parse(AVCodecParserContext *s,
+ AVCodecContext *avctx,
+ uint8_t **poutbuf, int *poutbuf_size,
+ const uint8_t *buf, int buf_size)
+{
+ H264Context *h = s->priv_data;
+ ParseContext *pc = &h->s.parse_context;
+ int next;
+
+ next= find_frame_end(h, buf, buf_size);
+
+ if (ff_combine_frame(pc, next, (uint8_t **)&buf, &buf_size) < 0) {
+ *poutbuf = NULL;
+ *poutbuf_size = 0;
+ return buf_size;
+ }
+
+ *poutbuf = (uint8_t *)buf;
+ *poutbuf_size = buf_size;
+ return next;
}
static int decode_nal_units(H264Context *h, uint8_t *buf, int buf_size){
MpegEncContext * const s = &h->s;
AVCodecContext * const avctx= s->avctx;
int buf_index=0;
+#if 0
int i;
-#if 0
for(i=0; i<32; i++){
printf("%X ", buf[i]);
}
#endif
+ h->slice_num = 0;
for(;;){
int consumed;
int dst_length;
int bit_length;
uint8_t *ptr;
+ int i, nalsize = 0;
+ if(h->is_avc) {
+ if(buf_index >= buf_size) break;
+ nalsize = 0;
+ for(i = 0; i < h->nal_length_size; i++)
+ nalsize = (nalsize << 8) | buf[buf_index++];
+ } else {
// start code prefix search
for(; buf_index + 3 < buf_size; buf_index++){
// this should allways succeed in the first iteration
if(buf_index+3 >= buf_size) break;
buf_index+=3;
+ }
- ptr= decode_nal(h, buf + buf_index, &dst_length, &consumed, buf_size - buf_index);
+ ptr= decode_nal(h, buf + buf_index, &dst_length, &consumed, h->is_avc ? nalsize : buf_size - buf_index);
if(ptr[dst_length - 1] == 0) dst_length--;
bit_length= 8*dst_length - decode_rbsp_trailing(ptr + dst_length - 1);
if(s->avctx->debug&FF_DEBUG_STARTCODE){
- printf("NAL %d at %d length %d\n", h->nal_unit_type, buf_index, dst_length);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d length %d\n", h->nal_unit_type, buf_index, dst_length);
}
+ if (h->is_avc && (nalsize != consumed))
+ av_log(h->s.avctx, AV_LOG_ERROR, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize);
+
buf_index += consumed;
- if(h->nal_ref_idc < s->hurry_up)
+ if( s->hurry_up == 1 && h->nal_ref_idc == 0 )
continue;
switch(h->nal_unit_type){
s->data_partitioning = 0;
if(decode_slice_header(h) < 0) return -1;
- if(h->redundant_pic_count==0)
+ if(h->redundant_pic_count==0 && s->hurry_up < 5 )
decode_slice(h);
break;
case NAL_DPA:
case NAL_DPC:
init_get_bits(&h->inter_gb, ptr, bit_length);
h->inter_gb_ptr= &h->inter_gb;
-
- if(h->redundant_pic_count==0)
+
+ if(h->redundant_pic_count==0 && h->intra_gb_ptr && s->data_partitioning && s->hurry_up < 5 )
decode_slice(h);
break;
case NAL_SEI:
break;
case NAL_FILTER_DATA:
break;
+ default:
+ av_log(avctx, AV_LOG_ERROR, "Unknown NAL code: %d\n", h->nal_unit_type);
}
//FIXME move after where irt is set
s->current_picture.key_frame= s->pict_type == I_TYPE;
}
+ if(!s->current_picture_ptr) return buf_index; //no frame
+
h->prev_frame_num_offset= h->frame_num_offset;
h->prev_frame_num= h->frame_num;
if(s->current_picture_ptr->reference){
assert(h->mmco_index==0);
ff_er_frame_end(s);
+
MPV_frame_end(s);
return buf_index;
H264Context *h = avctx->priv_data;
MpegEncContext *s = &h->s;
AVFrame *pict = data;
- float new_aspect;
int buf_index;
s->flags= avctx->flags;
+ s->flags2= avctx->flags2;
- *data_size = 0;
-
/* no supplementary picture */
if (buf_size == 0) {
return 0;
}
if(s->flags&CODEC_FLAG_TRUNCATED){
- int next= find_frame_end(s, buf, buf_size);
+ int next= find_frame_end(h, buf, buf_size);
- if( ff_combine_frame(s, next, &buf, &buf_size) < 0 )
+ if( ff_combine_frame(&s->parse_context, next, &buf, &buf_size) < 0 )
return buf_size;
//printf("next:%d buf_size:%d last_index:%d\n", next, buf_size, s->parse_context.last_index);
}
- if(s->avctx->extradata_size && s->picture_number==0){
+ if(h->is_avc && !h->got_avcC) {
+ int i, cnt, nalsize;
+ unsigned char *p = avctx->extradata;
+ if(avctx->extradata_size < 7) {
+ av_log(avctx, AV_LOG_ERROR, "avcC too short\n");
+ return -1;
+ }
+ if(*p != 1) {
+ av_log(avctx, AV_LOG_ERROR, "Unknown avcC version %d\n", *p);
+ return -1;
+ }
+ /* sps and pps in the avcC always have length coded with 2 bytes,
+ so put a fake nal_length_size = 2 while parsing them */
+ h->nal_length_size = 2;
+ // Decode sps from avcC
+ cnt = *(p+5) & 0x1f; // Number of sps
+ p += 6;
+ for (i = 0; i < cnt; i++) {
+ nalsize = BE_16(p) + 2;
+ if(decode_nal_units(h, p, nalsize) != nalsize) {
+ av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i);
+ return -1;
+ }
+ p += nalsize;
+ }
+ // Decode pps from avcC
+ cnt = *(p++); // Number of pps
+ for (i = 0; i < cnt; i++) {
+ nalsize = BE_16(p) + 2;
+ if(decode_nal_units(h, p, nalsize) != nalsize) {
+ av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i);
+ return -1;
+ }
+ p += nalsize;
+ }
+ // Now store right nal length size, that will be use to parse all other nals
+ h->nal_length_size = ((*(((char*)(avctx->extradata))+4))&0x03)+1;
+ // Do not reparse avcC
+ h->got_avcC = 1;
+ }
+
+ if(!h->is_avc && s->avctx->extradata_size && s->picture_number==0){
if(0 < decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) )
return -1;
}
*pict= *(AVFrame*)&s->last_picture;
}
#endif
+ if(!s->current_picture_ptr){
+ av_log(h->s.avctx, AV_LOG_DEBUG, "error, NO frame\n");
+ return -1;
+ }
+
*pict= *(AVFrame*)&s->current_picture; //FIXME
+ ff_print_debug_info(s, pict);
assert(pict->data[0]);
//printf("out %d\n", (int)pict->data[0]);
- if(avctx->debug&FF_DEBUG_QP){
- int8_t *qtab= pict->qscale_table;
- int x,y;
-
- for(y=0; y<s->mb_height; y++){
- for(x=0; x<s->mb_width; x++){
- printf("%2d ", qtab[x + y*s->mb_width]);
- }
- printf("\n");
- }
- printf("\n");
- }
#if 0 //?
/* Return the Picture timestamp as the frame number */
#if 0
/* dont output the last pic after seeking */
if(s->last_picture_ptr || s->low_delay)
- //Note this isnt a issue as a IDR pic should flush teh buffers
+ //Note this isnt a issue as a IDR pic should flush the buffers
#endif
*data_size = sizeof(AVFrame);
return get_consumed_bytes(s, buf_index, buf_size);
#if 0
static inline void fill_mb_avail(H264Context *h){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*h->mb_stride;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
if(s->mb_y){
- h->mb_avail[0]= s->mb_x && h->slice_table[mb_xy - h->mb_stride - 1] == h->slice_num;
- h->mb_avail[1]= h->slice_table[mb_xy - h->mb_stride ] == h->slice_num;
- h->mb_avail[2]= s->mb_x+1 < s->mb_width && h->slice_table[mb_xy - h->mb_stride + 1] == h->slice_num;
+ h->mb_avail[0]= s->mb_x && h->slice_table[mb_xy - s->mb_stride - 1] == h->slice_num;
+ h->mb_avail[1]= h->slice_table[mb_xy - s->mb_stride ] == h->slice_num;
+ h->mb_avail[2]= s->mb_x+1 < s->mb_width && h->slice_table[mb_xy - s->mb_stride + 1] == h->slice_num;
}else{
h->mb_avail[0]=
h->mb_avail[1]=
dsputil_init(&dsp, &avctx);
- init_put_bits(&pb, temp, SIZE, NULL, NULL);
+ init_put_bits(&pb, temp, SIZE);
printf("testing unsigned exp golomb\n");
for(i=0; i<COUNT; i++){
START_TIMER
}
- init_put_bits(&pb, temp, SIZE, NULL, NULL);
+ init_put_bits(&pb, temp, SIZE);
printf("testing signed exp golomb\n");
for(i=0; i<COUNT; i++){
START_TIMER
}
// printf("\n");
- h264_add_idct_c(ref, block, 4);
+ s->dsp.h264_idct_add(ref, block, 4);
/* for(j=0; j<16; j++){
printf("%d ", ref[j]);
}
NULL,
decode_end,
decode_frame,
- /*CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | */CODEC_CAP_TRUNCATED,
+ /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED,
+};
+
+AVCodecParser h264_parser = {
+ { CODEC_ID_H264 },
+ sizeof(H264Context),
+ NULL,
+ h264_parse,
+ ff_parse_close,
};
+#include "svq3.c"
projects / ffmpeg / blobdiff