*/
/**
- * @file libavcodec/h264.c
+ * @file
* H.264 / AVC / MPEG4 part10 codec.
* @author Michael Niedermayer <michaelni@gmx.at>
*/
#include "vdpau_internal.h"
#include "cabac.h"
-#if ARCH_X86
-#include "x86/h264_i386.h"
-#endif
//#undef NDEBUG
#include <assert.h>
-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 const uint8_t rem6[52]={
0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3,
};
0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8,
};
+static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = {
+ PIX_FMT_DXVA2_VLD,
+ PIX_FMT_VAAPI_VLD,
+ PIX_FMT_YUVJ420P,
+ PIX_FMT_NONE
+};
+
void ff_h264_write_back_intra_pred_mode(H264Context *h){
- const int mb_xy= h->mb_xy;
+ int8_t *mode= h->intra4x4_pred_mode + h->mb2br_xy[h->mb_xy];
- 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];
- h->intra4x4_pred_mode[mb_xy][2]= h->intra4x4_pred_mode_cache[7+8*3];
- h->intra4x4_pred_mode[mb_xy][3]= h->intra4x4_pred_mode_cache[7+8*4];
- h->intra4x4_pred_mode[mb_xy][4]= h->intra4x4_pred_mode_cache[4+8*4];
- h->intra4x4_pred_mode[mb_xy][5]= h->intra4x4_pred_mode_cache[5+8*4];
- h->intra4x4_pred_mode[mb_xy][6]= h->intra4x4_pred_mode_cache[6+8*4];
+ AV_COPY32(mode, h->intra4x4_pred_mode_cache + 4 + 8*4);
+ mode[4]= h->intra4x4_pred_mode_cache[7+8*3];
+ mode[5]= h->intra4x4_pred_mode_cache[7+8*2];
+ mode[6]= h->intra4x4_pred_mode_cache[7+8*1];
}
/**
# if HAVE_FAST_64BIT
# define RS 7
for(i=0; i+1<length; i+=9){
- if(!((~*(const uint64_t*)(src+i) & (*(const uint64_t*)(src+i) - 0x0100010001000101ULL)) & 0x8000800080008080ULL))
+ if(!((~AV_RN64A(src+i) & (AV_RN64A(src+i) - 0x0100010001000101ULL)) & 0x8000800080008080ULL))
# else
# define RS 3
for(i=0; i+1<length; i+=5){
- if(!((~*(const uint32_t*)(src+i) & (*(const uint32_t*)(src+i) - 0x01000101U)) & 0x80008080U))
+ if(!((~AV_RN32A(src+i) & (AV_RN32A(src+i) - 0x01000101U)) & 0x80008080U))
# endif
continue;
if(i>0 && !src[i]) i--;
x_offset, y_offset, qpix_put, chroma_put);
if(h->use_weight == 2){
- int weight0 = h->implicit_weight[refn0][refn1];
+ int weight0 = h->implicit_weight[refn0][refn1][s->mb_y&1];
int weight1 = 64 - weight0;
luma_weight_avg( dest_y, tmp_y, h-> mb_linesize, 5, weight0, weight1, 0);
chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, 5, weight0, weight1, 0);
chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, 5, weight0, weight1, 0);
}else{
luma_weight_avg(dest_y, tmp_y, h->mb_linesize, h->luma_log2_weight_denom,
- h->luma_weight[0][refn0], h->luma_weight[1][refn1],
- h->luma_offset[0][refn0] + h->luma_offset[1][refn1]);
+ h->luma_weight[refn0][0][0] , h->luma_weight[refn1][1][0],
+ h->luma_weight[refn0][0][1] + h->luma_weight[refn1][1][1]);
chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[0][refn0][0], h->chroma_weight[1][refn1][0],
- h->chroma_offset[0][refn0][0] + h->chroma_offset[1][refn1][0]);
+ h->chroma_weight[refn0][0][0][0] , h->chroma_weight[refn1][1][0][0],
+ h->chroma_weight[refn0][0][0][1] + h->chroma_weight[refn1][1][0][1]);
chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[0][refn0][1], h->chroma_weight[1][refn1][1],
- h->chroma_offset[0][refn0][1] + h->chroma_offset[1][refn1][1]);
+ h->chroma_weight[refn0][0][1][0] , h->chroma_weight[refn1][1][1][0],
+ h->chroma_weight[refn0][0][1][1] + h->chroma_weight[refn1][1][1][1]);
}
}else{
int list = list1 ? 1 : 0;
qpix_put, chroma_put);
luma_weight_op(dest_y, h->mb_linesize, h->luma_log2_weight_denom,
- h->luma_weight[list][refn], h->luma_offset[list][refn]);
+ h->luma_weight[refn][list][0], h->luma_weight[refn][list][1]);
if(h->use_weight_chroma){
chroma_weight_op(dest_cb, h->mb_uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[list][refn][0], h->chroma_offset[list][refn][0]);
+ h->chroma_weight[refn][list][0][0], h->chroma_weight[refn][list][0][1]);
chroma_weight_op(dest_cr, h->mb_uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[list][refn][1], h->chroma_offset[list][refn][1]);
+ h->chroma_weight[refn][list][1][0], h->chroma_weight[refn][list][1][1]);
}
}
}
h264_weight_func *weight_op, h264_biweight_func *weight_avg,
int list0, int list1){
if((h->use_weight==2 && list0 && list1
- && (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ] != 32))
+ && (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ][h->s.mb_y&1] != 32))
|| h->use_weight==1)
mc_part_weighted(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr,
x_offset, y_offset, qpix_put, chroma_put,
if(IS_16X16(mb_type)){
mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
- &weight_op[0], &weight_avg[0],
+ weight_op, weight_avg,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1));
}else if(IS_16X8(mb_type)){
mc_part(h, 0, 0, 4, 8, dest_y, dest_cb, dest_cr, 0, 0,
av_freep(&h->list_counts);
av_freep(&h->mb2b_xy);
- av_freep(&h->mb2b8_xy);
+ av_freep(&h->mb2br_xy);
for(i = 0; i < MAX_THREADS; i++) {
hx = h->thread_context[i];
static void init_dequant8_coeff_table(H264Context *h){
int i,q,x;
- const int transpose = (h->s.dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly
+ const int transpose = (h->h264dsp.h264_idct8_add != ff_h264_idct8_add_c); //FIXME ugly
h->dequant8_coeff[0] = h->dequant8_buffer[0];
h->dequant8_coeff[1] = h->dequant8_buffer[1];
static void init_dequant4_coeff_table(H264Context *h){
int i,j,q,x;
- const int transpose = (h->s.dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
+ const int transpose = (h->h264dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
for(i=0; i<6; i++ ){
h->dequant4_coeff[i] = h->dequant4_buffer[i];
for(j=0; j<i; j++){
int ff_h264_alloc_tables(H264Context *h){
MpegEncContext * const s = &h->s;
const int big_mb_num= s->mb_stride * (s->mb_height+1);
+ const int row_mb_num= 2*s->mb_stride*s->avctx->thread_count;
int x,y;
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, big_mb_num * 8 * sizeof(uint8_t), fail)
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, row_mb_num * 8 * sizeof(uint8_t), fail)
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->non_zero_count , big_mb_num * 32 * sizeof(uint8_t), fail)
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->slice_table_base , (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base), fail)
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->cbp_table, big_mb_num * sizeof(uint16_t), fail)
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t), fail)
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[0], 32*big_mb_num * sizeof(uint16_t), fail);
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[1], 32*big_mb_num * sizeof(uint16_t), fail);
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->direct_table, 32*big_mb_num * sizeof(uint8_t) , fail);
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[0], 16*row_mb_num * sizeof(uint8_t), fail);
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mvd_table[1], 16*row_mb_num * sizeof(uint8_t), fail);
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->direct_table, 4*big_mb_num * sizeof(uint8_t) , fail);
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->list_counts, big_mb_num * sizeof(uint8_t), fail)
memset(h->slice_table_base, -1, (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base));
h->slice_table= h->slice_table_base + s->mb_stride*2 + 1;
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2b_xy , big_mb_num * sizeof(uint32_t), fail);
- FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2b8_xy , big_mb_num * sizeof(uint32_t), fail);
+ FF_ALLOCZ_OR_GOTO(h->s.avctx, h->mb2br_xy , big_mb_num * sizeof(uint32_t), fail);
for(y=0; y<s->mb_height; y++){
for(x=0; x<s->mb_width; x++){
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;
h->mb2b_xy [mb_xy]= b_xy;
- h->mb2b8_xy[mb_xy]= b8_xy;
+ h->mb2br_xy[mb_xy]= 8*(FMO ? mb_xy : (mb_xy % (2*s->mb_stride)));
}
}
/**
* Mimic alloc_tables(), but for every context thread.
*/
-static void clone_tables(H264Context *dst, H264Context *src){
- dst->intra4x4_pred_mode = src->intra4x4_pred_mode;
+static void clone_tables(H264Context *dst, H264Context *src, int i){
+ MpegEncContext * const s = &src->s;
+ dst->intra4x4_pred_mode = src->intra4x4_pred_mode + i*8*2*s->mb_stride;
dst->non_zero_count = src->non_zero_count;
dst->slice_table = src->slice_table;
dst->cbp_table = src->cbp_table;
dst->mb2b_xy = src->mb2b_xy;
- dst->mb2b8_xy = src->mb2b8_xy;
+ dst->mb2br_xy = src->mb2br_xy;
dst->chroma_pred_mode_table = src->chroma_pred_mode_table;
- dst->mvd_table[0] = src->mvd_table[0];
- dst->mvd_table[1] = src->mvd_table[1];
+ dst->mvd_table[0] = src->mvd_table[0] + i*8*2*s->mb_stride;
+ dst->mvd_table[1] = src->mvd_table[1] + i*8*2*s->mb_stride;
dst->direct_table = src->direct_table;
+ dst->list_counts = src->list_counts;
dst->s.obmc_scratchpad = NULL;
ff_h264_pred_init(&dst->hpc, src->s.codec_id);
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[0], h->s.mb_width * (16+8+8) * sizeof(uint8_t), fail)
FF_ALLOCZ_OR_GOTO(h->s.avctx, h->top_borders[1], h->s.mb_width * (16+8+8) * sizeof(uint8_t), fail)
+ h->ref_cache[0][scan8[5 ]+1] = h->ref_cache[0][scan8[7 ]+1] = h->ref_cache[0][scan8[13]+1] =
+ h->ref_cache[1][scan8[5 ]+1] = h->ref_cache[1][scan8[7 ]+1] = h->ref_cache[1][scan8[13]+1] = PART_NOT_AVAILABLE;
+
return 0;
fail:
return -1; // free_tables will clean up for us
}
+static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size);
+
static av_cold void common_init(H264Context *h){
MpegEncContext * const s = &h->s;
s->height = s->avctx->height;
s->codec_id= s->avctx->codec->id;
+ ff_h264dsp_init(&h->h264dsp);
ff_h264_pred_init(&h->hpc, s->codec_id);
h->dequant_coeff_pps= -1;
memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t));
}
+int ff_h264_decode_extradata(H264Context *h)
+{
+ AVCodecContext *avctx = h->s.avctx;
+
+ if(*(char *)avctx->extradata == 1){
+ int i, cnt, nalsize;
+ unsigned char *p = avctx->extradata;
+
+ h->is_avc = 1;
+
+ if(avctx->extradata_size < 7) {
+ av_log(avctx, AV_LOG_ERROR, "avcC too short\n");
+ 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 = AV_RB16(p) + 2;
+ if(decode_nal_units(h, p, nalsize) < 0) {
+ 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 = AV_RB16(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;
+ } else {
+ h->is_avc = 0;
+ if(decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0)
+ return -1;
+ }
+ return 0;
+}
+
av_cold int ff_h264_decode_init(AVCodecContext *avctx){
H264Context *h= avctx->priv_data;
MpegEncContext * const s = &h->s;
ff_h264_decode_init_vlc();
- if(avctx->extradata_size > 0 && avctx->extradata &&
- *(char *)avctx->extradata == 1){
- h->is_avc = 1;
- h->got_avcC = 0;
- } else {
- h->is_avc = 0;
- }
-
h->thread_context[0] = h;
h->outputed_poc = INT_MIN;
h->prev_poc_msb= 1<<16;
+ h->x264_build = -1;
ff_h264_reset_sei(h);
if(avctx->codec_id == CODEC_ID_H264){
if(avctx->ticks_per_frame == 1){
}
avctx->ticks_per_frame = 2;
}
+
+ if(avctx->extradata_size > 0 && avctx->extradata &&
+ ff_h264_decode_extradata(h))
+ return -1;
+
+ if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){
+ s->avctx->has_b_frames = h->sps.num_reorder_frames;
+ s->low_delay = 0;
+ }
+
return 0;
}
if(!h->thread_context[i]->s.obmc_scratchpad)
h->thread_context[i]->s.obmc_scratchpad = av_malloc(16*2*s->linesize + 8*2*s->uvlinesize);
- /* some macroblocks will be accessed before they're available */
- if(FRAME_MBAFF || s->avctx->thread_count > 1)
- memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(*h->slice_table));
+ /* some macroblocks can be accessed before they're available in case of lost slices, mbaff or threading*/
+ memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(*h->slice_table));
// 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, int simple){
MpegEncContext * const s = &h->s;
+ uint8_t *top_border;
int top_idx = 1;
src_y -= linesize;
if(!simple && FRAME_MBAFF){
if(s->mb_y&1){
if(!MB_MBAFF){
- *(uint64_t*)(h->top_borders[0][s->mb_x]+ 0)= *(uint64_t*)(src_y + 15*linesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+ 8)= *(uint64_t*)(src_y +8+15*linesize);
+ top_border = h->top_borders[0][s->mb_x];
+ AV_COPY128(top_border, src_y + 15*linesize);
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+7*uvlinesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+7*uvlinesize);
+ AV_COPY64(top_border+16, src_cb+7*uvlinesize);
+ AV_COPY64(top_border+24, src_cr+7*uvlinesize);
}
}
}else if(MB_MBAFF){
return;
}
+ top_border = h->top_borders[top_idx][s->mb_x];
// There are two lines saved, the line above the the top macroblock of a pair,
// and the line above the bottom macroblock
-
- *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
- *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);
+ AV_COPY128(top_border, src_y + 16*linesize);
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
- *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
+ AV_COPY64(top_border+16, src_cb+8*uvlinesize);
+ AV_COPY64(top_border+24, 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, int simple){
MpegEncContext * const s = &h->s;
- int temp8, i;
- uint64_t temp64;
int deblock_left;
int deblock_top;
- int mb_xy;
int top_idx = 1;
+ uint8_t *top_border_m1;
+ uint8_t *top_border;
if(!simple && FRAME_MBAFF){
if(s->mb_y&1){
}
if(h->deblocking_filter == 2) {
- mb_xy = h->mb_xy;
- deblock_left = h->slice_table[mb_xy] == h->slice_table[mb_xy - 1];
- deblock_top = h->slice_table[mb_xy] == h->slice_table[h->top_mb_xy];
+ deblock_left = h->left_type[0];
+ deblock_top = h->top_type;
} else {
deblock_left = (s->mb_x > 0);
deblock_top = (s->mb_y > !!MB_FIELD);
src_cb -= uvlinesize + 1;
src_cr -= uvlinesize + 1;
-#define XCHG(a,b,t,xchg)\
-t= a;\
-if(xchg)\
- a= b;\
-b= t;
+ top_border_m1 = h->top_borders[top_idx][s->mb_x-1];
+ top_border = h->top_borders[top_idx][s->mb_x];
+
+#define XCHG(a,b,xchg)\
+if (xchg) AV_SWAP64(b,a);\
+else AV_COPY64(b,a);
if(deblock_top){
if(deblock_left){
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x-1]+8), *(uint64_t*)(src_y -7), temp64, 1);
+ XCHG(top_border_m1+8, src_y -7, 1);
}
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
+ XCHG(top_border+0, src_y +1, xchg);
+ XCHG(top_border+8, src_y +9, 1);
if(s->mb_x+1 < s->mb_width){
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x+1]), *(uint64_t*)(src_y +17), temp64, 1);
+ XCHG(h->top_borders[top_idx][s->mb_x+1], src_y +17, 1);
}
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
if(deblock_top){
if(deblock_left){
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x-1]+16), *(uint64_t*)(src_cb -7), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x-1]+24), *(uint64_t*)(src_cr -7), temp64, 1);
+ XCHG(top_border_m1+16, src_cb -7, 1);
+ XCHG(top_border_m1+24, src_cr -7, 1);
}
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
+ XCHG(top_border+16, src_cb+1, 1);
+ XCHG(top_border+24, src_cr+1, 1);
}
}
}
idct_dc_add =
idct_add = s->dsp.add_pixels8;
}else{
- idct_dc_add = s->dsp.h264_idct8_dc_add;
- idct_add = s->dsp.h264_idct8_add;
+ idct_dc_add = h->h264dsp.h264_idct8_dc_add;
+ idct_add = h->h264dsp.h264_idct8_add;
}
for(i=0; i<16; i+=4){
uint8_t * const ptr= dest_y + block_offset[i];
idct_dc_add =
idct_add = s->dsp.add_pixels4;
}else{
- idct_dc_add = s->dsp.h264_idct_dc_add;
- idct_add = s->dsp.h264_idct_add;
+ idct_dc_add = h->h264dsp.h264_idct_dc_add;
+ idct_add = h->h264dsp.h264_idct_add;
}
for(i=0; i<16; i++){
uint8_t * const ptr= dest_y + block_offset[i];
else
idct_add (ptr, h->mb + i*16, linesize);
}else
- svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
+ ff_svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
}
}
}
if(!transform_bypass)
h264_luma_dc_dequant_idct_c(h->mb, s->qscale, h->dequant4_coeff[0][s->qscale][0]);
}else
- svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
+ ff_svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
}
if(h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, simple);
hl_motion(h, dest_y, dest_cb, dest_cr,
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
- s->dsp.weight_h264_pixels_tab, s->dsp.biweight_h264_pixels_tab);
+ h->h264dsp.weight_h264_pixels_tab, h->h264dsp.biweight_h264_pixels_tab);
}
}
}
}else{
- s->dsp.h264_idct_add16intra(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+ h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
}
}else if(h->cbp&15){
if(transform_bypass){
}
}else{
if(IS_8x8DCT(mb_type)){
- s->dsp.h264_idct8_add4(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+ h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
}else{
- s->dsp.h264_idct_add16(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
+ h->h264dsp.h264_idct_add16(dest_y, block_offset, h->mb, linesize, h->non_zero_count_cache);
}
}
}
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 + block_offset[i];
- svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
+ ff_svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
}
}
}
chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp[0], h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp[1], h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
if(is_h264){
- idct_add = s->dsp.h264_idct_add;
- idct_dc_add = s->dsp.h264_idct_dc_add;
+ idct_add = h->h264dsp.h264_idct_add;
+ idct_dc_add = h->h264dsp.h264_idct_dc_add;
for(i=16; i<16+8; i++){
if(h->non_zero_count_cache[ scan8[i] ])
idct_add (dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize);
for(i=16; i<16+8; i++){
if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
uint8_t * const ptr= dest[(i&4)>>2] + block_offset[i];
- svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[s->qscale + 12] - 12, 2);
+ ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[s->qscale + 12] - 12, 2);
}
}
}
}
if(h->cbp || IS_INTRA(mb_type))
s->dsp.clear_blocks(h->mb);
-
- if(h->deblocking_filter && 0) {
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, simple);
- fill_caches(h, mb_type, 1); //FIXME don't fill stuff which isn't used by filter_mb
- h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy]);
- h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy]);
- if (!simple && FRAME_MBAFF) {
- ff_h264_filter_mb (h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
- } else {
- ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
- }
- }
}
/**
luma_weight_flag= get_bits1(&s->gb);
if(luma_weight_flag){
- h->luma_weight[list][i]= get_se_golomb(&s->gb);
- h->luma_offset[list][i]= get_se_golomb(&s->gb);
- if( h->luma_weight[list][i] != luma_def
- || h->luma_offset[list][i] != 0) {
+ h->luma_weight[i][list][0]= get_se_golomb(&s->gb);
+ h->luma_weight[i][list][1]= get_se_golomb(&s->gb);
+ if( h->luma_weight[i][list][0] != luma_def
+ || h->luma_weight[i][list][1] != 0) {
h->use_weight= 1;
h->luma_weight_flag[list]= 1;
}
}else{
- h->luma_weight[list][i]= luma_def;
- h->luma_offset[list][i]= 0;
+ h->luma_weight[i][list][0]= luma_def;
+ h->luma_weight[i][list][1]= 0;
}
if(CHROMA){
if(chroma_weight_flag){
int j;
for(j=0; j<2; j++){
- h->chroma_weight[list][i][j]= get_se_golomb(&s->gb);
- h->chroma_offset[list][i][j]= get_se_golomb(&s->gb);
- if( h->chroma_weight[list][i][j] != chroma_def
- || h->chroma_offset[list][i][j] != 0) {
+ h->chroma_weight[i][list][j][0]= get_se_golomb(&s->gb);
+ h->chroma_weight[i][list][j][1]= get_se_golomb(&s->gb);
+ if( h->chroma_weight[i][list][j][0] != chroma_def
+ || h->chroma_weight[i][list][j][1] != 0) {
h->use_weight_chroma= 1;
h->chroma_weight_flag[list]= 1;
}
}else{
int j;
for(j=0; j<2; j++){
- h->chroma_weight[list][i][j]= chroma_def;
- h->chroma_offset[list][i][j]= 0;
+ h->chroma_weight[i][list][j][0]= chroma_def;
+ h->chroma_weight[i][list][j][1]= 0;
}
}
}
return 0;
}
-static void implicit_weight_table(H264Context *h){
+/**
+ * Initialize implicit_weight table.
+ * @param field 0/1 initialize the weight for interlaced MBAFF
+ * -1 initializes the rest
+ */
+static void implicit_weight_table(H264Context *h, int field){
MpegEncContext * const s = &h->s;
- int ref0, ref1, i;
- int cur_poc = s->current_picture_ptr->poc;
+ int ref0, ref1, i, cur_poc, ref_start, ref_count0, ref_count1;
for (i = 0; i < 2; i++) {
h->luma_weight_flag[i] = 0;
h->chroma_weight_flag[i] = 0;
}
- if( h->ref_count[0] == 1 && h->ref_count[1] == 1
+ if(field < 0){
+ cur_poc = s->current_picture_ptr->poc;
+ if( h->ref_count[0] == 1 && h->ref_count[1] == 1 && !FRAME_MBAFF
&& h->ref_list[0][0].poc + h->ref_list[1][0].poc == 2*cur_poc){
h->use_weight= 0;
h->use_weight_chroma= 0;
return;
}
+ ref_start= 0;
+ ref_count0= h->ref_count[0];
+ ref_count1= h->ref_count[1];
+ }else{
+ cur_poc = s->current_picture_ptr->field_poc[field];
+ ref_start= 16;
+ ref_count0= 16+2*h->ref_count[0];
+ ref_count1= 16+2*h->ref_count[1];
+ }
h->use_weight= 2;
h->use_weight_chroma= 2;
h->luma_log2_weight_denom= 5;
h->chroma_log2_weight_denom= 5;
- for(ref0=0; ref0 < h->ref_count[0]; ref0++){
+ for(ref0=ref_start; ref0 < ref_count0; ref0++){
int poc0 = h->ref_list[0][ref0].poc;
- for(ref1=0; ref1 < h->ref_count[1]; ref1++){
+ for(ref1=ref_start; ref1 < ref_count1; ref1++){
int poc1 = h->ref_list[1][ref1].poc;
int td = av_clip(poc1 - poc0, -128, 127);
+ int w= 32;
if(td){
int tb = av_clip(cur_poc - poc0, -128, 127);
int tx = (16384 + (FFABS(td) >> 1)) / td;
- int dist_scale_factor = av_clip((tb*tx + 32) >> 6, -1024, 1023) >> 2;
- if(dist_scale_factor < -64 || dist_scale_factor > 128)
- h->implicit_weight[ref0][ref1] = 32;
- else
- h->implicit_weight[ref0][ref1] = 64 - dist_scale_factor;
- }else
- h->implicit_weight[ref0][ref1] = 32;
+ int dist_scale_factor = (tb*tx + 32) >> 8;
+ if(dist_scale_factor >= -64 && dist_scale_factor <= 128)
+ w = 64 - dist_scale_factor;
+ }
+ if(field<0){
+ h->implicit_weight[ref0][ref1][0]=
+ h->implicit_weight[ref0][ref1][1]= w;
+ }else{
+ h->implicit_weight[ref0][ref1][field]=w;
+ }
}
}
}
* initialize scan tables
*/
static void init_scan_tables(H264Context *h){
- MpegEncContext * const s = &h->s;
int i;
- if(s->dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
+ if(h->h264dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
memcpy(h->zigzag_scan, zigzag_scan, 16*sizeof(uint8_t));
memcpy(h-> field_scan, field_scan, 16*sizeof(uint8_t));
}else{
#undef T
}
}
- if(s->dsp.h264_idct8_add == ff_h264_idct8_add_c){
+ if(h->h264dsp.h264_idct8_add == ff_h264_idct8_add_c){
memcpy(h->zigzag_scan8x8, ff_zigzag_direct, 64*sizeof(uint8_t));
memcpy(h->zigzag_scan8x8_cavlc, zigzag_scan8x8_cavlc, 64*sizeof(uint8_t));
memcpy(h->field_scan8x8, field_scan8x8, 64*sizeof(uint8_t));
}
/**
- * Replicates H264 "master" context to thread contexts.
+ * Replicate H264 "master" context to thread contexts.
*/
static void clone_slice(H264Context *dst, H264Context *src)
{
h->slice_type_nos= slice_type & 3;
s->pict_type= h->slice_type; // to make a few old functions happy, it's wrong though
- if (s->pict_type == FF_B_TYPE && s0->last_picture_ptr == NULL) {
- av_log(h->s.avctx, AV_LOG_ERROR,
- "B picture before any references, skipping\n");
- return -1;
- }
pps_id= get_ue_golomb(&s->gb);
if(pps_id>=MAX_PPS_COUNT){
}
h->sps = *h0->sps_buffers[h->pps.sps_id];
+ s->avctx->profile = h->sps.profile_idc;
+ s->avctx->level = h->sps.level_idc;
+ s->avctx->refs = h->sps.ref_frame_count;
+
if(h == h0 && h->dequant_coeff_pps != pps_id){
h->dequant_coeff_pps = pps_id;
init_dequant_tables(h);
s->mb_height= h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag);
h->b_stride= s->mb_width*4;
- h->b8_stride= s->mb_width*2;
s->width = 16*s->mb_width - 2*FFMIN(h->sps.crop_right, 7);
if(h->sps.frame_mbs_only_flag)
s->height= 16*s->mb_height - 4*FFMIN(h->sps.crop_bottom, 3);
if (s->context_initialized
- && ( s->width != s->avctx->width || s->height != s->avctx->height)) {
+ && ( s->width != s->avctx->width || s->height != s->avctx->height
+ || av_cmp_q(h->sps.sar, s->avctx->sample_aspect_ratio))) {
if(h != h0)
return -1; // width / height changed during parallelized decoding
free_tables(h);
}
if(h->sps.timing_info_present_flag){
- s->avctx->time_base= (AVRational){h->sps.num_units_in_tick, h->sps.time_scale};
- if(h->x264_build > 0 && h->x264_build < 44)
- s->avctx->time_base.den *= 2;
+ int64_t den= h->sps.time_scale;
+ if(h->x264_build < 44U)
+ den *= 2;
av_reduce(&s->avctx->time_base.num, &s->avctx->time_base.den,
- s->avctx->time_base.num, s->avctx->time_base.den, 1<<30);
- }
- s->avctx->pix_fmt = s->avctx->get_format(s->avctx, s->avctx->codec->pix_fmts);
+ h->sps.num_units_in_tick, den, 1<<30);
+ }
+ s->avctx->pix_fmt = s->avctx->get_format(s->avctx,
+ s->avctx->codec->pix_fmts ?
+ s->avctx->codec->pix_fmts :
+ s->avctx->color_range == AVCOL_RANGE_JPEG ?
+ hwaccel_pixfmt_list_h264_jpeg_420 :
+ ff_hwaccel_pixfmt_list_420);
s->avctx->hwaccel = ff_find_hwaccel(s->avctx->codec->id, s->avctx->pix_fmt);
if (MPV_common_init(s) < 0)
c = h->thread_context[i] = av_malloc(sizeof(H264Context));
memcpy(c, h->s.thread_context[i], sizeof(MpegEncContext));
memset(&c->s + 1, 0, sizeof(H264Context) - sizeof(MpegEncContext));
+ c->h264dsp = h->h264dsp;
c->sps = h->sps;
c->pps = h->pps;
init_scan_tables(c);
- clone_tables(c, h);
+ clone_tables(c, h, i);
}
for(i = 0; i < s->avctx->thread_count; i++)
if(h0->current_slice == 0){
while(h->frame_num != h->prev_frame_num &&
h->frame_num != (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num)){
- av_log(NULL, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->frame_num, h->prev_frame_num);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->frame_num, h->prev_frame_num);
if (ff_h264_frame_start(h) < 0)
return -1;
h->prev_frame_num++;
h->prev_frame_num %= 1<<h->sps.log2_max_frame_num;
s->current_picture_ptr->frame_num= h->prev_frame_num;
- ff_h264_execute_ref_pic_marking(h, NULL, 0);
+ ff_generate_sliding_window_mmcos(h);
+ ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
}
/* See if we have a decoded first field looking for a pair... */
if( (h->pps.weighted_pred && h->slice_type_nos == FF_P_TYPE )
|| (h->pps.weighted_bipred_idc==1 && h->slice_type_nos== FF_B_TYPE ) )
pred_weight_table(h);
- else if(h->pps.weighted_bipred_idc==2 && h->slice_type_nos== FF_B_TYPE)
- implicit_weight_table(h);
- else {
+ else if(h->pps.weighted_bipred_idc==2 && h->slice_type_nos== FF_B_TYPE){
+ implicit_weight_table(h, -1);
+ }else {
h->use_weight = 0;
for (i = 0; i < 2; i++) {
h->luma_weight_flag[i] = 0;
if(h->nal_ref_idc)
ff_h264_decode_ref_pic_marking(h0, &s->gb);
- if(FRAME_MBAFF)
+ if(FRAME_MBAFF){
ff_h264_fill_mbaff_ref_list(h);
+ if(h->pps.weighted_bipred_idc==2 && h->slice_type_nos== FF_B_TYPE){
+ implicit_weight_table(h, 0);
+ implicit_weight_table(h, 1);
+ }
+ }
+
if(h->slice_type_nos==FF_B_TYPE && !h->direct_spatial_mv_pred)
ff_h264_direct_dist_scale_factor(h);
ff_h264_direct_ref_list_init(h);
}
h->deblocking_filter = 1;
- h->slice_alpha_c0_offset = 0;
- h->slice_beta_offset = 0;
+ h->slice_alpha_c0_offset = 52;
+ h->slice_beta_offset = 52;
if( h->pps.deblocking_filter_parameters_present ) {
tmp= get_ue_golomb_31(&s->gb);
if(tmp > 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;
+ h->slice_alpha_c0_offset += get_se_golomb(&s->gb) << 1;
+ h->slice_beta_offset += get_se_golomb(&s->gb) << 1;
+ if( h->slice_alpha_c0_offset > 104U
+ || h->slice_beta_offset > 104U){
+ av_log(s->avctx, AV_LOG_ERROR, "deblocking filter parameters %d %d out of range\n", h->slice_alpha_c0_offset, h->slice_beta_offset);
+ return -1;
+ }
}
}
return 1; // deblocking switched inside frame
}
}
- h->qp_thresh= 15 - FFMIN(h->slice_alpha_c0_offset, h->slice_beta_offset) - FFMAX3(0, h->pps.chroma_qp_index_offset[0], h->pps.chroma_qp_index_offset[1]);
+ h->qp_thresh= 15 + 52 - FFMIN(h->slice_alpha_c0_offset, h->slice_beta_offset) - FFMAX3(0, h->pps.chroma_qp_index_offset[0], h->pps.chroma_qp_index_offset[1]);
#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)
}
for(j=0; j<2; j++){
+ int id_list[16];
int *ref2frm= h->ref2frm[h->slice_num&(MAX_SLICES-1)][j];
+ for(i=0; i<16; i++){
+ id_list[i]= 60;
+ if(h->ref_list[j][i].data[0]){
+ int k;
+ uint8_t *base= h->ref_list[j][i].base[0];
+ for(k=0; k<h->short_ref_count; k++)
+ if(h->short_ref[k]->base[0] == base){
+ id_list[i]= k;
+ break;
+ }
+ for(k=0; k<h->long_ref_count; k++)
+ if(h->long_ref[k] && h->long_ref[k]->base[0] == base){
+ id_list[i]= h->short_ref_count + k;
+ break;
+ }
+ }
+ }
+
ref2frm[0]=
ref2frm[1]= -1;
for(i=0; i<16; i++)
- ref2frm[i+2]= 4*h->ref_list[j][i].frame_num
+ ref2frm[i+2]= 4*id_list[i]
+(h->ref_list[j][i].reference&3);
ref2frm[18+0]=
ref2frm[18+1]= -1;
for(i=16; i<48; i++)
- ref2frm[i+4]= 4*h->ref_list[j][i].frame_num
+ ref2frm[i+4]= 4*id_list[(i-16)>>1]
+(h->ref_list[j][i].reference&3);
}
h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
h->emu_edge_height= (FRAME_MBAFF || FIELD_PICTURE) ? 0 : h->emu_edge_width;
- s->avctx->refs= h->sps.ref_frame_count;
-
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n",
h->slice_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->deblocking_filter, h->slice_alpha_c0_offset/2, h->slice_beta_offset/2,
+ h->deblocking_filter, h->slice_alpha_c0_offset/2-26, h->slice_beta_offset/2-26,
h->use_weight,
h->use_weight==1 && h->use_weight_chroma ? "c" : "",
h->slice_type == FF_B_TYPE ? (h->direct_spatial_mv_pred ? "SPAT" : "TEMP") : ""
return 0;
}
-int ff_h264_get_slice_type(H264Context *h)
+int ff_h264_get_slice_type(const H264Context *h)
{
switch (h->slice_type) {
case FF_P_TYPE: return 0;
}
}
+/**
+ *
+ * @return non zero if the loop filter can be skiped
+ */
+static int fill_filter_caches(H264Context *h, int mb_type){
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= h->mb_xy;
+ int top_xy, left_xy[2];
+ int top_type, left_type[2];
+
+ top_xy = mb_xy - (s->mb_stride << MB_FIELD);
+
+ //FIXME deblocking could skip the intra and nnz parts.
+
+ /* Wow, what a mess, why didn't they simplify the interlacing & intra
+ * stuff, I can't imagine that these complex rules are worth it. */
+
+ left_xy[1] = left_xy[0] = mb_xy-1;
+ if(FRAME_MBAFF){
+ const int left_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
+ const int curr_mb_field_flag = IS_INTERLACED(mb_type);
+ if(s->mb_y&1){
+ if (left_mb_field_flag != curr_mb_field_flag) {
+ left_xy[0] -= s->mb_stride;
+ }
+ }else{
+ if(curr_mb_field_flag){
+ top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
+ }
+ if (left_mb_field_flag != curr_mb_field_flag) {
+ left_xy[1] += s->mb_stride;
+ }
+ }
+ }
+
+ h->top_mb_xy = top_xy;
+ h->left_mb_xy[0] = left_xy[0];
+ h->left_mb_xy[1] = left_xy[1];
+ {
+ //for sufficiently low qp, filtering wouldn't do anything
+ //this is a conservative estimate: could also check beta_offset and more accurate chroma_qp
+ int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice
+ int qp = s->current_picture.qscale_table[mb_xy];
+ if(qp <= qp_thresh
+ && (left_xy[0]<0 || ((qp + s->current_picture.qscale_table[left_xy[0]] + 1)>>1) <= qp_thresh)
+ && (top_xy < 0 || ((qp + s->current_picture.qscale_table[top_xy ] + 1)>>1) <= qp_thresh)){
+ if(!FRAME_MBAFF)
+ return 1;
+ if( (left_xy[0]< 0 || ((qp + s->current_picture.qscale_table[left_xy[1] ] + 1)>>1) <= qp_thresh)
+ && (top_xy < s->mb_stride || ((qp + s->current_picture.qscale_table[top_xy -s->mb_stride] + 1)>>1) <= qp_thresh))
+ return 1;
+ }
+ }
+
+ top_type = s->current_picture.mb_type[top_xy] ;
+ left_type[0] = s->current_picture.mb_type[left_xy[0]];
+ left_type[1] = s->current_picture.mb_type[left_xy[1]];
+ if(h->deblocking_filter == 2){
+ if(h->slice_table[top_xy ] != h->slice_num) top_type= 0;
+ if(h->slice_table[left_xy[0] ] != h->slice_num) left_type[0]= left_type[1]= 0;
+ }else{
+ if(h->slice_table[top_xy ] == 0xFFFF) top_type= 0;
+ if(h->slice_table[left_xy[0] ] == 0xFFFF) left_type[0]= left_type[1] =0;
+ }
+ h->top_type = top_type ;
+ h->left_type[0]= left_type[0];
+ h->left_type[1]= left_type[1];
+
+ if(IS_INTRA(mb_type))
+ return 0;
+
+ AV_COPY64(&h->non_zero_count_cache[0+8*1], &h->non_zero_count[mb_xy][ 0]);
+ AV_COPY64(&h->non_zero_count_cache[0+8*2], &h->non_zero_count[mb_xy][ 8]);
+ AV_COPY32(&h->non_zero_count_cache[0+8*5], &h->non_zero_count[mb_xy][16]);
+ AV_COPY32(&h->non_zero_count_cache[4+8*3], &h->non_zero_count[mb_xy][20]);
+ AV_COPY64(&h->non_zero_count_cache[0+8*4], &h->non_zero_count[mb_xy][24]);
+
+ h->cbp= h->cbp_table[mb_xy];
+
+ {
+ int list;
+ for(list=0; list<h->list_count; list++){
+ int8_t *ref;
+ int y, b_stride;
+ int16_t (*mv_dst)[2];
+ int16_t (*mv_src)[2];
+
+ if(!USES_LIST(mb_type, list)){
+ fill_rectangle( h->mv_cache[list][scan8[0]], 4, 4, 8, pack16to32(0,0), 4);
+ AV_WN32A(&h->ref_cache[list][scan8[ 0]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
+ AV_WN32A(&h->ref_cache[list][scan8[ 2]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
+ AV_WN32A(&h->ref_cache[list][scan8[ 8]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
+ AV_WN32A(&h->ref_cache[list][scan8[10]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
+ continue;
+ }
+
+ ref = &s->current_picture.ref_index[list][4*mb_xy];
+ {
+ int (*ref2frm)[64] = h->ref2frm[ h->slice_num&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
+ AV_WN32A(&h->ref_cache[list][scan8[ 0]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
+ AV_WN32A(&h->ref_cache[list][scan8[ 2]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
+ ref += 2;
+ AV_WN32A(&h->ref_cache[list][scan8[ 8]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
+ AV_WN32A(&h->ref_cache[list][scan8[10]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
+ }
+
+ b_stride = h->b_stride;
+ mv_dst = &h->mv_cache[list][scan8[0]];
+ mv_src = &s->current_picture.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride];
+ for(y=0; y<4; y++){
+ AV_COPY128(mv_dst + 8*y, mv_src + y*b_stride);
+ }
+
+ }
+ }
+
+
+/*
+0 . T T. T T T T
+1 L . .L . . . .
+2 L . .L . . . .
+3 . T TL . . . .
+4 L . .L . . . .
+5 L . .. . . . .
+*/
+//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
+ if(top_type){
+ AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
+ }
+
+ if(left_type[0]){
+ h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][7+0*8];
+ h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][7+1*8];
+ h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[0]][7+2*8];
+ h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[0]][7+3*8];
+ }
+
+ // CAVLC 8x8dct requires NNZ values for residual decoding that differ from what the loop filter needs
+ if(!CABAC && h->pps.transform_8x8_mode){
+ if(IS_8x8DCT(top_type)){
+ h->non_zero_count_cache[4+8*0]=
+ h->non_zero_count_cache[5+8*0]= h->cbp_table[top_xy] & 4;
+ h->non_zero_count_cache[6+8*0]=
+ h->non_zero_count_cache[7+8*0]= h->cbp_table[top_xy] & 8;
+ }
+ if(IS_8x8DCT(left_type[0])){
+ h->non_zero_count_cache[3+8*1]=
+ h->non_zero_count_cache[3+8*2]= h->cbp_table[left_xy[0]]&2; //FIXME check MBAFF
+ }
+ if(IS_8x8DCT(left_type[1])){
+ h->non_zero_count_cache[3+8*3]=
+ h->non_zero_count_cache[3+8*4]= h->cbp_table[left_xy[1]]&8; //FIXME check MBAFF
+ }
+
+ if(IS_8x8DCT(mb_type)){
+ h->non_zero_count_cache[scan8[0 ]]= h->non_zero_count_cache[scan8[1 ]]=
+ h->non_zero_count_cache[scan8[2 ]]= h->non_zero_count_cache[scan8[3 ]]= h->cbp & 1;
+
+ h->non_zero_count_cache[scan8[0+ 4]]= h->non_zero_count_cache[scan8[1+ 4]]=
+ h->non_zero_count_cache[scan8[2+ 4]]= h->non_zero_count_cache[scan8[3+ 4]]= h->cbp & 2;
+
+ h->non_zero_count_cache[scan8[0+ 8]]= h->non_zero_count_cache[scan8[1+ 8]]=
+ h->non_zero_count_cache[scan8[2+ 8]]= h->non_zero_count_cache[scan8[3+ 8]]= h->cbp & 4;
+
+ h->non_zero_count_cache[scan8[0+12]]= h->non_zero_count_cache[scan8[1+12]]=
+ h->non_zero_count_cache[scan8[2+12]]= h->non_zero_count_cache[scan8[3+12]]= h->cbp & 8;
+ }
+ }
+
+ if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
+ int list;
+ for(list=0; list<h->list_count; list++){
+ if(USES_LIST(top_type, list)){
+ const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
+ const int b8_xy= 4*top_xy + 2;
+ int (*ref2frm)[64] = h->ref2frm[ h->slice_table[top_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
+ AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
+ h->ref_cache[list][scan8[0] + 0 - 1*8]=
+ h->ref_cache[list][scan8[0] + 1 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 0]];
+ h->ref_cache[list][scan8[0] + 2 - 1*8]=
+ h->ref_cache[list][scan8[0] + 3 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 1]];
+ }else{
+ AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
+ AV_WN32A(&h->ref_cache[list][scan8[0] + 0 - 1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u);
+ }
+
+ if(!IS_INTERLACED(mb_type^left_type[0])){
+ if(USES_LIST(left_type[0], list)){
+ const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
+ const int b8_xy= 4*left_xy[0] + 1;
+ int (*ref2frm)[64] = h->ref2frm[ h->slice_table[left_xy[0]]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 0 ], s->current_picture.motion_val[list][b_xy + h->b_stride*0]);
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 8 ], s->current_picture.motion_val[list][b_xy + h->b_stride*1]);
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1 +16 ], s->current_picture.motion_val[list][b_xy + h->b_stride*2]);
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1 +24 ], s->current_picture.motion_val[list][b_xy + h->b_stride*3]);
+ h->ref_cache[list][scan8[0] - 1 + 0 ]=
+ h->ref_cache[list][scan8[0] - 1 + 8 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*0]];
+ h->ref_cache[list][scan8[0] - 1 +16 ]=
+ h->ref_cache[list][scan8[0] - 1 +24 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*1]];
+ }else{
+ AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 0 ]);
+ AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 8 ]);
+ AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +16 ]);
+ AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +24 ]);
+ h->ref_cache[list][scan8[0] - 1 + 0 ]=
+ h->ref_cache[list][scan8[0] - 1 + 8 ]=
+ h->ref_cache[list][scan8[0] - 1 + 16 ]=
+ h->ref_cache[list][scan8[0] - 1 + 24 ]= LIST_NOT_USED;
+ }
+ }
+ }
+ }
+
+ return 0;
+}
+
static void loop_filter(H264Context *h){
MpegEncContext * const s = &h->s;
uint8_t *dest_y, *dest_cb, *dest_cr;
if(h->deblocking_filter) {
for(mb_x= 0; mb_x<s->mb_width; mb_x++){
for(mb_y=end_mb_y - FRAME_MBAFF; mb_y<= end_mb_y; mb_y++){
- int list, mb_xy, mb_type, is_complex;
+ int mb_xy, mb_type;
mb_xy = h->mb_xy = mb_x + mb_y*s->mb_stride;
h->slice_num= h->slice_table[mb_xy];
mb_type= s->current_picture.mb_type[mb_xy];
h->list_count= h->list_counts[mb_xy];
- if(h->list_count==2){
- h->slice_type= h->slice_type_nos= FF_B_TYPE;
- }else if(h->list_count==1){
- h->slice_type= h->slice_type_nos= FF_P_TYPE;
- }else
- h->slice_type= h->slice_type_nos= FF_I_TYPE;
if(FRAME_MBAFF)
h->mb_mbaff = h->mb_field_decoding_flag = !!IS_INTERLACED(mb_type);
- is_complex = CONFIG_SMALL || h->is_complex || IS_INTRA_PCM(mb_type) || s->qscale == 0; //FIXME qscale might be wrong
-
s->mb_x= mb_x;
s->mb_y= mb_y;
dest_y = s->current_picture.data[0] + (mb_x + mb_y * s->linesize ) * 16;
linesize = h->mb_linesize = s->linesize;
uvlinesize = h->mb_uvlinesize = s->uvlinesize;
}
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, !is_complex);
- fill_caches(h, mb_type, 1); //FIXME don't fill stuff which isn't used by filter_mb
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
+ if(fill_filter_caches(h, mb_type))
+ continue;
h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy]);
h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy]);
- if (is_complex && FRAME_MBAFF) {
+ if (FRAME_MBAFF) {
ff_h264_filter_mb (h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
} else {
ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
h->slice_type= old_slice_type;
s->mb_x= 0;
s->mb_y= end_mb_y - FRAME_MBAFF;
+ h->chroma_qp[0] = get_chroma_qp(h, 0, s->qscale);
+ h->chroma_qp[1] = get_chroma_qp(h, 1, s->qscale);
+}
+
+static void predict_field_decoding_flag(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
+ int mb_type = (h->slice_table[mb_xy-1] == h->slice_num)
+ ? s->current_picture.mb_type[mb_xy-1]
+ : (h->slice_table[mb_xy-s->mb_stride] == h->slice_num)
+ ? s->current_picture.mb_type[mb_xy-s->mb_stride]
+ : 0;
+ h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0;
}
static int decode_slice(struct AVCodecContext *avctx, void *arg){
}
eos = get_cabac_terminate( &h->cabac );
+ if((s->workaround_bugs & FF_BUG_TRUNCATED) && h->cabac.bytestream > h->cabac.bytestream_end + 2){
+ 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( ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) {
av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d, bytestream (%td)\n", s->mb_x, s->mb_y, h->cabac.bytestream_end - h->cabac.bytestream);
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);
++s->mb_y;
if(FIELD_OR_MBAFF_PICTURE) {
++s->mb_y;
+ if(FRAME_MBAFF && s->mb_y < s->mb_height)
+ predict_field_decoding_flag(h);
}
}
++s->mb_y;
if(FIELD_OR_MBAFF_PICTURE) {
++s->mb_y;
+ if(FRAME_MBAFF && s->mb_y < s->mb_height)
+ predict_field_decoding_flag(h);
}
if(s->mb_y >= s->mb_height){
tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
if (ptr==NULL || dst_length < 0){
return -1;
}
+ i= buf_index + consumed;
+ if((s->workaround_bugs & FF_BUG_AUTODETECT) && i+3<next_avc &&
+ buf[i]==0x00 && buf[i+1]==0x00 && buf[i+2]==0x01 && buf[i+3]==0xE0)
+ s->workaround_bugs |= FF_BUG_TRUNCATED;
+
+ if(!(s->workaround_bugs & FF_BUG_TRUNCATED)){
while(ptr[dst_length - 1] == 0 && dst_length > 0)
dst_length--;
+ }
bit_length= !dst_length ? 0 : (8*dst_length - ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1));
if(s->avctx->debug&FF_DEBUG_STARTCODE){
if((err = decode_slice_header(hx, h)))
break;
- if (s->avctx->hwaccel && h->current_slice == 1) {
- if (s->avctx->hwaccel->start_frame(s->avctx, NULL, 0) < 0)
+ if (h->current_slice == 1) {
+ if (s->avctx->hwaccel && s->avctx->hwaccel->start_frame(s->avctx, NULL, 0) < 0)
return -1;
+ if(CONFIG_H264_VDPAU_DECODER && s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
+ ff_vdpau_h264_picture_start(s);
}
s->current_picture_ptr->key_frame |=
return 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 = AV_RB16(p) + 2;
- if(decode_nal_units(h, p, nalsize) < 0) {
- 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 = AV_RB16(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->got_avcC && !h->is_avc && s->avctx->extradata_size){
- if(decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) < 0)
- return -1;
- h->got_avcC = 1;
- }
-
buf_index=decode_nal_units(h, buf, buf_size);
if(buf_index < 0)
return -1;
}
// printf("\n");
- s->dsp.h264_idct_add(ref, block, 4);
+ h->h264dsp.h264_idct_add(ref, block, 4);
/* for(j=0; j<16; j++){
printf("%d ", ref[j]);
}
AVCodec h264_decoder = {
"h264",
- CODEC_TYPE_VIDEO,
+ AVMEDIA_TYPE_VIDEO,
CODEC_ID_H264,
sizeof(H264Context),
ff_h264_decode_init,
/*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY,
.flush= flush_dpb,
.long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
- .pix_fmts= ff_hwaccel_pixfmt_list_420,
};
#if CONFIG_H264_VDPAU_DECODER
AVCodec h264_vdpau_decoder = {
"h264_vdpau",
- CODEC_TYPE_VIDEO,
+ AVMEDIA_TYPE_VIDEO,
CODEC_ID_H264,
sizeof(H264Context),
ff_h264_decode_init,