lavc/j2kenc: Support AV_PIX_FMT_PAL8.
[ffmpeg.git] / libavcodec / h264_cavlc.c
index af8a8b6..5e6a203 100644 (file)
@@ -21,7 +21,7 @@
 
 /**
  * @file
- * H.264 / AVC / MPEG4 part10 cavlc bitstream decoding.
+ * H.264 / AVC / MPEG-4 part10 cavlc bitstream decoding.
  * @author Michael Niedermayer <michaelni@gmx.at>
  */
 
@@ -30,9 +30,9 @@
 
 #include "internal.h"
 #include "avcodec.h"
-#include "h264.h"
-#include "h264data.h" // FIXME FIXME FIXME
+#include "h264dec.h"
 #include "h264_mvpred.h"
+#include "h264data.h"
 #include "golomb.h"
 #include "mpegutils.h"
 #include "libavutil/avassert.h"
@@ -248,19 +248,19 @@ static VLC chroma422_dc_coeff_token_vlc;
 static VLC_TYPE chroma422_dc_coeff_token_vlc_table[8192][2];
 static const int chroma422_dc_coeff_token_vlc_table_size = 8192;
 
-static VLC total_zeros_vlc[15];
+static VLC total_zeros_vlc[15+1];
 static VLC_TYPE total_zeros_vlc_tables[15][512][2];
 static const int total_zeros_vlc_tables_size = 512;
 
-static VLC chroma_dc_total_zeros_vlc[3];
+static VLC chroma_dc_total_zeros_vlc[3+1];
 static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2];
 static const int chroma_dc_total_zeros_vlc_tables_size = 8;
 
-static VLC chroma422_dc_total_zeros_vlc[7];
+static VLC chroma422_dc_total_zeros_vlc[7+1];
 static VLC_TYPE chroma422_dc_total_zeros_vlc_tables[7][32][2];
 static const int chroma422_dc_total_zeros_vlc_tables_size = 32;
 
-static VLC run_vlc[6];
+static VLC run_vlc[6+1];
 static VLC_TYPE run_vlc_tables[6][8][2];
 static const int run_vlc_tables_size = 8;
 
@@ -284,7 +284,7 @@ static int8_t cavlc_level_tab[7][1<<LEVEL_TAB_BITS][2];
  * Get the predicted number of non-zero coefficients.
  * @param n block index
  */
-static inline int pred_non_zero_count(H264Context *h, H264SliceContext *sl, int n)
+static inline int pred_non_zero_count(const H264Context *h, H264SliceContext *sl, int n)
 {
     const int index8= scan8[n];
     const int left = sl->non_zero_count_cache[index8 - 1];
@@ -293,7 +293,7 @@ static inline int pred_non_zero_count(H264Context *h, H264SliceContext *sl, int
 
     if(i<64) i= (i+1)>>1;
 
-    tprintf(h->avctx, "pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
+    ff_tlog(h->avctx, "pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
 
     return i&31;
 }
@@ -364,9 +364,9 @@ av_cold void ff_h264_decode_init_vlc(void){
         av_assert0(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
 
         for(i=0; i<3; i++){
-            chroma_dc_total_zeros_vlc[i].table = chroma_dc_total_zeros_vlc_tables[i];
-            chroma_dc_total_zeros_vlc[i].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
-            init_vlc(&chroma_dc_total_zeros_vlc[i],
+            chroma_dc_total_zeros_vlc[i+1].table = chroma_dc_total_zeros_vlc_tables[i];
+            chroma_dc_total_zeros_vlc[i+1].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
+            init_vlc(&chroma_dc_total_zeros_vlc[i+1],
                      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,
@@ -374,9 +374,9 @@ av_cold void ff_h264_decode_init_vlc(void){
         }
 
         for(i=0; i<7; i++){
-            chroma422_dc_total_zeros_vlc[i].table = chroma422_dc_total_zeros_vlc_tables[i];
-            chroma422_dc_total_zeros_vlc[i].table_allocated = chroma422_dc_total_zeros_vlc_tables_size;
-            init_vlc(&chroma422_dc_total_zeros_vlc[i],
+            chroma422_dc_total_zeros_vlc[i+1].table = chroma422_dc_total_zeros_vlc_tables[i];
+            chroma422_dc_total_zeros_vlc[i+1].table_allocated = chroma422_dc_total_zeros_vlc_tables_size;
+            init_vlc(&chroma422_dc_total_zeros_vlc[i+1],
                      CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 8,
                      &chroma422_dc_total_zeros_len [i][0], 1, 1,
                      &chroma422_dc_total_zeros_bits[i][0], 1, 1,
@@ -384,9 +384,9 @@ av_cold void ff_h264_decode_init_vlc(void){
         }
 
         for(i=0; i<15; i++){
-            total_zeros_vlc[i].table = total_zeros_vlc_tables[i];
-            total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size;
-            init_vlc(&total_zeros_vlc[i],
+            total_zeros_vlc[i+1].table = total_zeros_vlc_tables[i];
+            total_zeros_vlc[i+1].table_allocated = total_zeros_vlc_tables_size;
+            init_vlc(&total_zeros_vlc[i+1],
                      TOTAL_ZEROS_VLC_BITS, 16,
                      &total_zeros_len [i][0], 1, 1,
                      &total_zeros_bits[i][0], 1, 1,
@@ -394,9 +394,9 @@ av_cold void ff_h264_decode_init_vlc(void){
         }
 
         for(i=0; i<6; i++){
-            run_vlc[i].table = run_vlc_tables[i];
-            run_vlc[i].table_allocated = run_vlc_tables_size;
-            init_vlc(&run_vlc[i],
+            run_vlc[i+1].table = run_vlc_tables[i];
+            run_vlc[i+1].table_allocated = run_vlc_tables_size;
+            init_vlc(&run_vlc[i+1],
                      RUN_VLC_BITS, 7,
                      &run_len [i][0], 1, 1,
                      &run_bits[i][0], 1, 1,
@@ -413,9 +413,6 @@ av_cold void ff_h264_decode_init_vlc(void){
     }
 }
 
-/**
- *
- */
 static inline int get_level_prefix(GetBitContext *gb){
     unsigned int buf;
     int log;
@@ -425,10 +422,6 @@ static inline int get_level_prefix(GetBitContext *gb){
     buf=GET_CACHE(re, gb);
 
     log= 32 - av_log2(buf);
-#ifdef TRACE
-    print_bin(buf>>(32-log), log);
-    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);
     CLOSE_READER(re, gb);
@@ -443,7 +436,7 @@ static inline int get_level_prefix(GetBitContext *gb){
  * @param max_coeff number of coefficients in the block
  * @return <0 if an error occurred
  */
-static int decode_residual(H264Context *h, H264SliceContext *sl,
+static int decode_residual(const H264Context *h, H264SliceContext *sl,
                            GetBitContext *gb, int16_t *block, int n,
                            const uint8_t *scantable, const uint32_t *qmul,
                            int max_coeff)
@@ -478,12 +471,12 @@ static int decode_residual(H264Context *h, H264SliceContext *sl,
     if(total_coeff==0)
         return 0;
     if(total_coeff > (unsigned)max_coeff) {
-        av_log(h->avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", h->mb_x, h->mb_y, total_coeff);
+        av_log(h->avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", sl->mb_x, sl->mb_y, total_coeff);
         return -1;
     }
 
     trailing_ones= coeff_token&3;
-    tprintf(h->avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
+    ff_tlog(h->avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
     av_assert2(total_coeff<=16);
 
     i = show_bits(gb, 3);
@@ -577,13 +570,13 @@ static int decode_residual(H264Context *h, H264SliceContext *sl,
     else{
         if (max_coeff <= 8) {
             if (max_coeff == 4)
-                zeros_left = get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[total_coeff].table,
+                zeros_left = get_vlc2(gb, chroma_dc_total_zeros_vlc[total_coeff].table,
                                       CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
             else
-                zeros_left = get_vlc2(gb, (chroma422_dc_total_zeros_vlc-1)[total_coeff].table,
+                zeros_left = get_vlc2(gb, chroma422_dc_total_zeros_vlc[total_coeff].table,
                                       CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1);
         } else {
-            zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
+            zeros_left= get_vlc2(gb, total_zeros_vlc[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
         }
     }
 
@@ -593,7 +586,7 @@ static int decode_residual(H264Context *h, H264SliceContext *sl,
         ((type*)block)[*scantable] = level[0]; \
         for(i=1;i<total_coeff && zeros_left > 0;i++) { \
             if(zeros_left < 7) \
-                run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \
+                run_before= get_vlc2(gb, run_vlc[zeros_left].table, RUN_VLC_BITS, 1); \
             else \
                 run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
             zeros_left -= run_before; \
@@ -608,7 +601,7 @@ static int decode_residual(H264Context *h, H264SliceContext *sl,
         ((type*)block)[*scantable] = ((int)(level[0] * qmul[*scantable] + 32))>>6; \
         for(i=1;i<total_coeff && zeros_left > 0;i++) { \
             if(zeros_left < 7) \
-                run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \
+                run_before= get_vlc2(gb, run_vlc[zeros_left].table, RUN_VLC_BITS, 1); \
             else \
                 run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
             zeros_left -= run_before; \
@@ -628,14 +621,19 @@ static int decode_residual(H264Context *h, H264SliceContext *sl,
     }
 
     if(zeros_left<0){
-        av_log(h->avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", h->mb_x, h->mb_y);
+        av_log(h->avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", sl->mb_x, sl->mb_y);
         return -1;
     }
 
     return 0;
 }
 
-static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContext *sl, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){
+static av_always_inline
+int decode_luma_residual(const H264Context *h, H264SliceContext *sl,
+                         GetBitContext *gb, const uint8_t *scan,
+                         const uint8_t *scan8x8, int pixel_shift,
+                         int mb_type, int cbp, int p)
+{
     int i4x4, i8x8;
     int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1];
     if(IS_INTRA16x16(mb_type)){
@@ -643,7 +641,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
         AV_ZERO128(sl->mb_luma_dc[p]+8);
         AV_ZERO128(sl->mb_luma_dc[p]+16);
         AV_ZERO128(sl->mb_luma_dc[p]+24);
-        if( decode_residual(h, sl, h->intra_gb_ptr, sl->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
+        if (decode_residual(h, sl, gb, sl->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX + p, scan, NULL, 16) < 0) {
             return -1; //FIXME continue if partitioned and other return -1 too
         }
 
@@ -653,8 +651,8 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
             for(i8x8=0; i8x8<4; i8x8++){
                 for(i4x4=0; i4x4<4; i4x4++){
                     const int index= i4x4 + 4*i8x8 + p*16;
-                    if( decode_residual(h, sl, h->intra_gb_ptr, sl->mb + (16*index << pixel_shift),
-                        index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
+                    if( decode_residual(h, sl, gb, sl->mb + (16*index << pixel_shift),
+                        index, scan + 1, h->ps.pps->dequant4_coeff[p][qscale], 15) < 0 ){
                         return -1;
                     }
                 }
@@ -676,7 +674,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
                     for(i4x4=0; i4x4<4; i4x4++){
                         const int index= i4x4 + 4*i8x8 + p*16;
                         if( decode_residual(h, sl, gb, buf, index, scan8x8+16*i4x4,
-                                            h->dequant8_coeff[cqm][qscale], 16) < 0 )
+                                            h->ps.pps->dequant8_coeff[cqm][qscale], 16) < 0 )
                             return -1;
                     }
                     nnz = &sl->non_zero_count_cache[scan8[4 * i8x8 + p * 16]];
@@ -686,7 +684,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
                     for(i4x4=0; i4x4<4; i4x4++){
                         const int index= i4x4 + 4*i8x8 + p*16;
                         if( decode_residual(h, sl, gb, sl->mb + (16*index << pixel_shift), index,
-                                            scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
+                                            scan, h->ps.pps->dequant4_coeff[cqm][qscale], 16) < 0 ){
                             return -1;
                         }
                         new_cbp |= sl->non_zero_count_cache[scan8[index]] << i8x8;
@@ -701,54 +699,53 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
     }
 }
 
-int ff_h264_decode_mb_cavlc(H264Context *h, H264SliceContext *sl)
+int ff_h264_decode_mb_cavlc(const H264Context *h, H264SliceContext *sl)
 {
     int mb_xy;
     int partition_count;
     unsigned int mb_type, cbp;
-    int dct8x8_allowed= h->pps.transform_8x8_mode;
-    int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2;
+    int dct8x8_allowed = h->ps.pps->transform_8x8_mode;
+    const int decode_chroma = h->ps.sps->chroma_format_idc == 1 || h->ps.sps->chroma_format_idc == 2;
     const int pixel_shift = h->pixel_shift;
-    unsigned local_ref_count[2];
 
-    mb_xy = h->mb_xy = h->mb_x + h->mb_y*h->mb_stride;
+    mb_xy = sl->mb_xy = sl->mb_x + sl->mb_y*h->mb_stride;
 
-    tprintf(h->avctx, "pic:%d mb:%d/%d\n", h->frame_num, h->mb_x, h->mb_y);
+    ff_tlog(h->avctx, "pic:%d mb:%d/%d\n", h->poc.frame_num, sl->mb_x, sl->mb_y);
     cbp = 0; /* avoid warning. FIXME: find a solution without slowing
                 down the code */
     if (sl->slice_type_nos != AV_PICTURE_TYPE_I) {
-        if(h->mb_skip_run==-1)
-            h->mb_skip_run= get_ue_golomb_long(&h->gb);
+        if (sl->mb_skip_run == -1)
+            sl->mb_skip_run = get_ue_golomb_long(&sl->gb);
 
-        if (h->mb_skip_run--) {
-            if(FRAME_MBAFF(h) && (h->mb_y&1) == 0){
-                if(h->mb_skip_run==0)
-                    h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&h->gb);
+        if (sl->mb_skip_run--) {
+            if (FRAME_MBAFF(h) && (sl->mb_y & 1) == 0) {
+                if (sl->mb_skip_run == 0)
+                    sl->mb_mbaff = sl->mb_field_decoding_flag = get_bits1(&sl->gb);
             }
             decode_mb_skip(h, sl);
             return 0;
         }
     }
     if (FRAME_MBAFF(h)) {
-        if( (h->mb_y&1) == 0 )
-            h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&h->gb);
+        if ((sl->mb_y & 1) == 0)
+            sl->mb_mbaff = sl->mb_field_decoding_flag = get_bits1(&sl->gb);
     }
 
     sl->prev_mb_skipped = 0;
 
-    mb_type= get_ue_golomb(&h->gb);
+    mb_type= get_ue_golomb(&sl->gb);
     if (sl->slice_type_nos == AV_PICTURE_TYPE_B) {
         if(mb_type < 23){
-            partition_countb_mb_type_info[mb_type].partition_count;
-            mb_type=         b_mb_type_info[mb_type].type;
+            partition_count = ff_h264_b_mb_type_info[mb_type].partition_count;
+            mb_type         = ff_h264_b_mb_type_info[mb_type].type;
         }else{
             mb_type -= 23;
             goto decode_intra_mb;
         }
     } else if (sl->slice_type_nos == AV_PICTURE_TYPE_P) {
         if(mb_type < 5){
-            partition_countp_mb_type_info[mb_type].partition_count;
-            mb_type=         p_mb_type_info[mb_type].type;
+            partition_count = ff_h264_p_mb_type_info[mb_type].partition_count;
+            mb_type         = ff_h264_p_mb_type_info[mb_type].type;
         }else{
             mb_type -= 5;
             goto decode_intra_mb;
@@ -759,31 +756,31 @@ int ff_h264_decode_mb_cavlc(H264Context *h, H264SliceContext *sl)
             mb_type--;
 decode_intra_mb:
         if(mb_type > 25){
-            av_log(h->avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(sl->slice_type), h->mb_x, h->mb_y);
+            av_log(h->avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(sl->slice_type), sl->mb_x, sl->mb_y);
             return -1;
         }
         partition_count=0;
-        cbpi_mb_type_info[mb_type].cbp;
-        sl->intra16x16_pred_mode = i_mb_type_info[mb_type].pred_mode;
-        mb_typei_mb_type_info[mb_type].type;
+        cbp                      = ff_h264_i_mb_type_info[mb_type].cbp;
+        sl->intra16x16_pred_mode = ff_h264_i_mb_type_info[mb_type].pred_mode;
+        mb_type                  = ff_h264_i_mb_type_info[mb_type].type;
     }
 
-    if(MB_FIELD(h))
+    if (MB_FIELD(sl))
         mb_type |= MB_TYPE_INTERLACED;
 
     h->slice_table[mb_xy] = sl->slice_num;
 
     if(IS_INTRA_PCM(mb_type)){
-        const int mb_size = ff_h264_mb_sizes[h->sps.chroma_format_idc] *
-                            h->sps.bit_depth_luma;
+        const int mb_size = ff_h264_mb_sizes[h->ps.sps->chroma_format_idc] *
+                            h->ps.sps->bit_depth_luma;
 
         // We assume these blocks are very rare so we do not optimize it.
-        sl->intra_pcm_ptr = align_get_bits(&h->gb);
-        if (get_bits_left(&h->gb) < mb_size) {
+        sl->intra_pcm_ptr = align_get_bits(&sl->gb);
+        if (get_bits_left(&sl->gb) < mb_size) {
             av_log(h->avctx, AV_LOG_ERROR, "Not enough data for an intra PCM block.\n");
             return AVERROR_INVALIDDATA;
         }
-        skip_bits_long(&h->gb, mb_size);
+        skip_bits_long(&sl->gb, mb_size);
 
         // In deblocking, the quantizer is 0
         h->cur_pic.qscale_table[mb_xy] = 0;
@@ -794,9 +791,6 @@ decode_intra_mb:
         return 0;
     }
 
-    local_ref_count[0] = sl->ref_count[0] << MB_MBAFF(h);
-    local_ref_count[1] = sl->ref_count[1] << MB_MBAFF(h);
-
     fill_decode_neighbors(h, sl, mb_type);
     fill_decode_caches(h, sl, mb_type);
 
@@ -807,7 +801,7 @@ decode_intra_mb:
         if(IS_INTRA4x4(mb_type)){
             int i;
             int di = 1;
-            if(dct8x8_allowed && get_bits1(&h->gb)){
+            if(dct8x8_allowed && get_bits1(&sl->gb)){
                 mb_type |= MB_TYPE_8x8DCT;
                 di = 4;
             }
@@ -816,8 +810,8 @@ decode_intra_mb:
             for(i=0; i<16; i+=di){
                 int mode = pred_intra_mode(h, sl, i);
 
-                if(!get_bits1(&h->gb)){
-                    const int rem_mode= get_bits(&h->gb, 3);
+                if(!get_bits1(&sl->gb)){
+                    const int rem_mode= get_bits(&sl->gb, 3);
                     mode = rem_mode + (rem_mode >= mode);
                 }
 
@@ -827,15 +821,18 @@ decode_intra_mb:
                     sl->intra4x4_pred_mode_cache[scan8[i]] = mode;
             }
             write_back_intra_pred_mode(h, sl);
-            if (ff_h264_check_intra4x4_pred_mode(h, sl) < 0)
+            if (ff_h264_check_intra4x4_pred_mode(sl->intra4x4_pred_mode_cache, h->avctx,
+                                                 sl->top_samples_available, sl->left_samples_available) < 0)
                 return -1;
         }else{
-            sl->intra16x16_pred_mode = ff_h264_check_intra_pred_mode(h, sl, sl->intra16x16_pred_mode, 0);
+            sl->intra16x16_pred_mode = ff_h264_check_intra_pred_mode(h->avctx, sl->top_samples_available,
+                                                                     sl->left_samples_available, sl->intra16x16_pred_mode, 0);
             if (sl->intra16x16_pred_mode < 0)
                 return -1;
         }
         if(decode_chroma){
-            pred_mode= ff_h264_check_intra_pred_mode(h, sl, get_ue_golomb_31(&h->gb), 1);
+            pred_mode= ff_h264_check_intra_pred_mode(h->avctx, sl->top_samples_available,
+                                                     sl->left_samples_available, get_ue_golomb_31(&sl->gb), 1);
             if(pred_mode < 0)
                 return -1;
             sl->chroma_pred_mode = pred_mode;
@@ -847,13 +844,13 @@ decode_intra_mb:
 
         if (sl->slice_type_nos == AV_PICTURE_TYPE_B) {
             for(i=0; i<4; i++){
-                sl->sub_mb_type[i]= get_ue_golomb_31(&h->gb);
+                sl->sub_mb_type[i]= get_ue_golomb_31(&sl->gb);
                 if(sl->sub_mb_type[i] >=13){
-                    av_log(h->avctx, AV_LOG_ERROR, "B sub_mb_type %u out of range at %d %d\n", sl->sub_mb_type[i], h->mb_x, h->mb_y);
+                    av_log(h->avctx, AV_LOG_ERROR, "B sub_mb_type %u out of range at %d %d\n", sl->sub_mb_type[i], sl->mb_x, sl->mb_y);
                     return -1;
                 }
-                sub_partition_count[i]= b_sub_mb_type_info[ sl->sub_mb_type[i] ].partition_count;
-                sl->sub_mb_type[i]=      b_sub_mb_type_info[ sl->sub_mb_type[i] ].type;
+                sub_partition_count[i] = ff_h264_b_sub_mb_type_info[sl->sub_mb_type[i]].partition_count;
+                sl->sub_mb_type[i]     = ff_h264_b_sub_mb_type_info[sl->sub_mb_type[i]].type;
             }
             if( IS_DIRECT(sl->sub_mb_type[0]|sl->sub_mb_type[1]|sl->sub_mb_type[2]|sl->sub_mb_type[3])) {
                 ff_h264_pred_direct_motion(h, sl, &mb_type);
@@ -865,18 +862,18 @@ decode_intra_mb:
         }else{
             av_assert2(sl->slice_type_nos == AV_PICTURE_TYPE_P); //FIXME SP correct ?
             for(i=0; i<4; i++){
-                sl->sub_mb_type[i]= get_ue_golomb_31(&h->gb);
+                sl->sub_mb_type[i]= get_ue_golomb_31(&sl->gb);
                 if(sl->sub_mb_type[i] >=4){
-                    av_log(h->avctx, AV_LOG_ERROR, "P sub_mb_type %u out of range at %d %d\n", sl->sub_mb_type[i], h->mb_x, h->mb_y);
+                    av_log(h->avctx, AV_LOG_ERROR, "P sub_mb_type %u out of range at %d %d\n", sl->sub_mb_type[i], sl->mb_x, sl->mb_y);
                     return -1;
                 }
-                sub_partition_count[i]= p_sub_mb_type_info[ sl->sub_mb_type[i] ].partition_count;
-                sl->sub_mb_type[i]=      p_sub_mb_type_info[ sl->sub_mb_type[i] ].type;
+                sub_partition_count[i] = ff_h264_p_sub_mb_type_info[sl->sub_mb_type[i]].partition_count;
+                sl->sub_mb_type[i]     = ff_h264_p_sub_mb_type_info[sl->sub_mb_type[i]].type;
             }
         }
 
         for (list = 0; list < sl->list_count; list++) {
-            int ref_count = IS_REF0(mb_type) ? 1 : local_ref_count[list];
+            int ref_count = IS_REF0(mb_type) ? 1 : sl->ref_count[list] << MB_MBAFF(sl);
             for(i=0; i<4; i++){
                 if(IS_DIRECT(sl->sub_mb_type[i])) continue;
                 if(IS_DIR(sl->sub_mb_type[i], 0, list)){
@@ -884,9 +881,9 @@ decode_intra_mb:
                     if(ref_count == 1){
                         tmp= 0;
                     }else if(ref_count == 2){
-                        tmp= get_bits1(&h->gb)^1;
+                        tmp= get_bits1(&sl->gb)^1;
                     }else{
-                        tmp= get_ue_golomb_31(&h->gb);
+                        tmp= get_ue_golomb_31(&sl->gb);
                         if(tmp>=ref_count){
                             av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", tmp);
                             return -1;
@@ -920,9 +917,9 @@ decode_intra_mb:
                         const int index= 4*i + block_width*j;
                         int16_t (* mv_cache)[2]= &sl->mv_cache[list][ scan8[index] ];
                         pred_motion(h, sl, index, block_width, list, sl->ref_cache[list][ scan8[index] ], &mx, &my);
-                        mx += get_se_golomb(&h->gb);
-                        my += get_se_golomb(&h->gb);
-                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
+                        mx += get_se_golomb(&sl->gb);
+                        my += get_se_golomb(&sl->gb);
+                        ff_tlog(h->avctx, "final mv:%d %d\n", mx, my);
 
                         if(IS_SUB_8X8(sub_mb_type)){
                             mv_cache[ 1 ][0]=
@@ -948,7 +945,7 @@ decode_intra_mb:
         }
     }else if(IS_DIRECT(mb_type)){
         ff_h264_pred_direct_motion(h, sl, &mb_type);
-        dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
+        dct8x8_allowed &= h->ps.sps->direct_8x8_inference_flag;
     }else{
         int list, mx, my, i;
          //FIXME we should set ref_idx_l? to 0 if we use that later ...
@@ -956,13 +953,14 @@ decode_intra_mb:
             for (list = 0; list < sl->list_count; list++) {
                     unsigned int val;
                     if(IS_DIR(mb_type, 0, list)){
-                        if(local_ref_count[list]==1){
+                        unsigned rc = sl->ref_count[list] << MB_MBAFF(sl);
+                        if (rc == 1) {
                             val= 0;
-                        } else if(local_ref_count[list]==2){
-                            val= get_bits1(&h->gb)^1;
+                        } else if (rc == 2) {
+                            val= get_bits1(&sl->gb)^1;
                         }else{
-                            val= get_ue_golomb_31(&h->gb);
-                            if (val >= local_ref_count[list]){
+                            val= get_ue_golomb_31(&sl->gb);
+                            if (val >= rc) {
                                 av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                 return -1;
                             }
@@ -973,9 +971,9 @@ decode_intra_mb:
             for (list = 0; list < sl->list_count; list++) {
                 if(IS_DIR(mb_type, 0, list)){
                     pred_motion(h, sl, 0, 4, list, sl->ref_cache[list][ scan8[0] ], &mx, &my);
-                    mx += get_se_golomb(&h->gb);
-                    my += get_se_golomb(&h->gb);
-                    tprintf(h->avctx, "final mv:%d %d\n", mx, my);
+                    mx += get_se_golomb(&sl->gb);
+                    my += get_se_golomb(&sl->gb);
+                    ff_tlog(h->avctx, "final mv:%d %d\n", mx, my);
 
                     fill_rectangle(sl->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
                 }
@@ -986,13 +984,14 @@ decode_intra_mb:
                     for(i=0; i<2; i++){
                         unsigned int val;
                         if(IS_DIR(mb_type, i, list)){
-                            if(local_ref_count[list] == 1) {
+                            unsigned rc = sl->ref_count[list] << MB_MBAFF(sl);
+                            if (rc == 1) {
                                 val= 0;
-                            } else if(local_ref_count[list] == 2) {
-                                val= get_bits1(&h->gb)^1;
+                            } else if (rc == 2) {
+                                val= get_bits1(&sl->gb)^1;
                             }else{
-                                val= get_ue_golomb_31(&h->gb);
-                                if (val >= local_ref_count[list]){
+                                val= get_ue_golomb_31(&sl->gb);
+                                if (val >= rc) {
                                     av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                     return -1;
                                 }
@@ -1007,9 +1006,9 @@ decode_intra_mb:
                     unsigned int val;
                     if(IS_DIR(mb_type, i, list)){
                         pred_16x8_motion(h, sl, 8*i, list, sl->ref_cache[list][scan8[0] + 16*i], &mx, &my);
-                        mx += get_se_golomb(&h->gb);
-                        my += get_se_golomb(&h->gb);
-                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
+                        mx += get_se_golomb(&sl->gb);
+                        my += get_se_golomb(&sl->gb);
+                        ff_tlog(h->avctx, "final mv:%d %d\n", mx, my);
 
                         val= pack16to32(mx,my);
                     }else
@@ -1023,13 +1022,14 @@ decode_intra_mb:
                     for(i=0; i<2; i++){
                         unsigned int val;
                         if(IS_DIR(mb_type, i, list)){ //FIXME optimize
-                            if(local_ref_count[list]==1){
+                            unsigned rc = sl->ref_count[list] << MB_MBAFF(sl);
+                            if (rc == 1) {
                                 val= 0;
-                            } else if(local_ref_count[list]==2){
-                                val= get_bits1(&h->gb)^1;
+                            } else if (rc == 2) {
+                                val= get_bits1(&sl->gb)^1;
                             }else{
-                                val= get_ue_golomb_31(&h->gb);
-                                if (val >= local_ref_count[list]){
+                                val= get_ue_golomb_31(&sl->gb);
+                                if (val >= rc) {
                                     av_log(h->avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
                                     return -1;
                                 }
@@ -1044,9 +1044,9 @@ decode_intra_mb:
                     unsigned int val;
                     if(IS_DIR(mb_type, i, list)){
                         pred_8x16_motion(h, sl, i*4, list, sl->ref_cache[list][ scan8[0] + 2*i ], &mx, &my);
-                        mx += get_se_golomb(&h->gb);
-                        my += get_se_golomb(&h->gb);
-                        tprintf(h->avctx, "final mv:%d %d\n", mx, my);
+                        mx += get_se_golomb(&sl->gb);
+                        my += get_se_golomb(&sl->gb);
+                        ff_tlog(h->avctx, "final mv:%d %d\n", mx, my);
 
                         val= pack16to32(mx,my);
                     }else
@@ -1061,18 +1061,20 @@ decode_intra_mb:
         write_back_motion(h, sl, mb_type);
 
     if(!IS_INTRA16x16(mb_type)){
-        cbp= get_ue_golomb(&h->gb);
+        cbp= get_ue_golomb(&sl->gb);
 
         if(decode_chroma){
             if(cbp > 47){
-                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, h->mb_x, h->mb_y);
+                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, sl->mb_x, sl->mb_y);
                 return -1;
             }
-            if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp];
-            else                     cbp= golomb_to_inter_cbp   [cbp];
+            if (IS_INTRA4x4(mb_type))
+                cbp = ff_h264_golomb_to_intra4x4_cbp[cbp];
+            else
+                cbp = ff_h264_golomb_to_inter_cbp[cbp];
         }else{
             if(cbp > 15){
-                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, h->mb_x, h->mb_y);
+                av_log(h->avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, sl->mb_x, sl->mb_y);
                 return -1;
             }
             if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp];
@@ -1086,9 +1088,9 @@ decode_intra_mb:
     }
 
     if(dct8x8_allowed && (cbp&15) && !IS_INTRA(mb_type)){
-        mb_type |= MB_TYPE_8x8DCT*get_bits1(&h->gb);
+        mb_type |= MB_TYPE_8x8DCT*get_bits1(&sl->gb);
     }
-    h->cbp=
+    sl->cbp=
     h->cbp_table[mb_xy]= cbp;
     h->cur_pic.mb_type[mb_xy] = mb_type;
 
@@ -1096,33 +1098,34 @@ decode_intra_mb:
         int i4x4, i8x8, chroma_idx;
         int dquant;
         int ret;
-        GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
+        GetBitContext *gb = &sl->gb;
         const uint8_t *scan, *scan8x8;
-        const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8);
-
-        if(IS_INTERLACED(mb_type)){
-            scan8x8 = sl->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0;
-            scan    = sl->qscale ? h->field_scan : h->field_scan_q0;
-        }else{
-            scan8x8 = sl->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0;
-            scan    = sl->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
-        }
+        const int max_qp = 51 + 6 * (h->ps.sps->bit_depth_luma - 8);
 
-        dquant= get_se_golomb(&h->gb);
+        dquant= get_se_golomb(&sl->gb);
 
-        sl->qscale += dquant;
+        sl->qscale += (unsigned)dquant;
 
         if (((unsigned)sl->qscale) > max_qp){
             if (sl->qscale < 0) sl->qscale += max_qp + 1;
             else                sl->qscale -= max_qp+1;
             if (((unsigned)sl->qscale) > max_qp){
-                av_log(h->avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, h->mb_x, h->mb_y);
+                av_log(h->avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, sl->mb_x, sl->mb_y);
+                sl->qscale = max_qp;
                 return -1;
             }
         }
 
-        sl->chroma_qp[0] = get_chroma_qp(h, 0, sl->qscale);
-        sl->chroma_qp[1] = get_chroma_qp(h, 1, sl->qscale);
+        sl->chroma_qp[0] = get_chroma_qp(h->ps.pps, 0, sl->qscale);
+        sl->chroma_qp[1] = get_chroma_qp(h->ps.pps, 1, sl->qscale);
+
+        if(IS_INTERLACED(mb_type)){
+            scan8x8 = sl->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0;
+            scan    = sl->qscale ? h->field_scan : h->field_scan_q0;
+        }else{
+            scan8x8 = sl->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0;
+            scan    = sl->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
+        }
 
         if ((ret = decode_luma_residual(h, sl, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ) {
             return -1;
@@ -1136,21 +1139,21 @@ decode_intra_mb:
                 return -1;
             }
         } else {
-            const int num_c8x8 = h->sps.chroma_format_idc;
+            const int num_c8x8 = h->ps.sps->chroma_format_idc;
 
             if(cbp&0x30){
                 for(chroma_idx=0; chroma_idx<2; chroma_idx++)
                     if (decode_residual(h, sl, gb, sl->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
-                                        CHROMA_DC_BLOCK_INDEX+chroma_idx,
-                                        CHROMA422(h) ? chroma422_dc_scan : chroma_dc_scan,
-                                        NULL, 4*num_c8x8) < 0) {
+                                        CHROMA_DC_BLOCK_INDEX + chroma_idx,
+                                        CHROMA422(h) ? ff_h264_chroma422_dc_scan : ff_h264_chroma_dc_scan,
+                                        NULL, 4 * num_c8x8) < 0) {
                         return -1;
                     }
             }
 
             if(cbp&0x20){
                 for(chroma_idx=0; chroma_idx<2; chroma_idx++){
-                    const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]];
+                    const uint32_t *qmul = h->ps.pps->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]];
                     int16_t *mb = sl->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
                     for (i8x8 = 0; i8x8<num_c8x8; i8x8++) {
                         for (i4x4 = 0; i4x4 < 4; i4x4++) {