Split cavlc out of h264.c.

Seems to speed the code up a little...
The placement of many generic functions between h264.c and h264.h is still open
Currently they are a little randomly placed between them.

Originally committed as revision 21178 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/Makefile

@@ -132,7 +132,7 @@ OBJS-$(CONFIG_H263_ENCODER)            += mpegvideo_enc.o mpeg4video.o mpeg4vide
                                           ratecontrol.o h263.o ituh263enc.o flvenc.o mpeg12data.o \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_H264_DECODER)            += h264.o h264idct.o h264pred.o h264_loopfilter.o h264_direct.o cabac.o \
-                                          h264_sei.o h264_ps.o h264_refs.o\
+                                          h264_sei.o h264_ps.o h264_refs.o h264_cavlc.o\
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_H264_ENCODER)            += h264enc.o h264dspenc.o
 OBJS-$(CONFIG_H264_VAAPI_HWACCEL)      += vaapi_h264.o
@@ -283,7 +283,7 @@ OBJS-$(CONFIG_SVQ1_ENCODER)            += svq1enc.o svq1.o    \
                                           motion_est.o h263.o \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_SVQ3_DECODER)            += h264.o svq3.o h264idct.o h264pred.o h264_loopfilter.o h264_direct.o  \
-                                          h264_sei.o h264_ps.o h264_refs.o\
+                                          h264_sei.o h264_ps.o h264_refs.o h264_cavlc.o\
                                           cabac.o          \
                                           mpegvideo.o error_resilience.o \
                                           svq1dec.o svq1.o h263.o

libavcodec/h264.h

@@ -32,6 +32,7 @@
 #include "cabac.h"
 #include "mpegvideo.h"
 #include "h264pred.h"
+#include "rectangle.h"
 
 #define interlaced_dct interlaced_dct_is_a_bad_name
 #define mb_intra mb_intra_is_not_initialized_see_mb_type
@@ -642,6 +643,13 @@ void ff_h264_hl_decode_mb(H264Context *h);
 int ff_h264_frame_start(H264Context *h);
 av_cold int ff_h264_decode_init(AVCodecContext *avctx);
 av_cold int ff_h264_decode_end(AVCodecContext *avctx);
+av_cold void ff_h264_decode_init_vlc(void);
+
+/**
+ * decodes a macroblock
+ * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
+ */
+int ff_h264_decode_mb_cavlc(H264Context *h);
 
 void ff_h264_direct_dist_scale_factor(H264Context * const h);
 void ff_h264_direct_ref_list_init(H264Context * const h);
@@ -694,4 +702,592 @@ static inline int get_chroma_qp(H264Context *h, int t, int qscale){
     return h->pps.chroma_qp_table[t][qscale];
 }
 
+static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my);
+
+static void fill_caches(H264Context *h, int mb_type, int for_deblock){
+    MpegEncContext * const s = &h->s;
+    const int mb_xy= h->mb_xy;
+    int topleft_xy, top_xy, topright_xy, left_xy[2];
+    int topleft_type, top_type, topright_type, left_type[2];
+    const uint8_t * left_block;
+    int topleft_partition= -1;
+    int i;
+    static const uint8_t left_block_options[4][8]={
+        {0,1,2,3,7,10,8,11},
+        {2,2,3,3,8,11,8,11},
+        {0,0,1,1,7,10,7,10},
+        {0,2,0,2,7,10,7,10}
+    };
+
+    top_xy     = mb_xy  - (s->mb_stride << FIELD_PICTURE);
+
+    //FIXME deblocking could skip the intra and nnz parts.
+    if(for_deblock && (h->slice_num == 1 || h->slice_table[mb_xy] == h->slice_table[top_xy]) && !FRAME_MBAFF)
+        return;
+
+    /* Wow, what a mess, why didn't they simplify the interlacing & intra
+     * stuff, I can't imagine that these complex rules are worth it. */
+
+    topleft_xy = top_xy - 1;
+    topright_xy= top_xy + 1;
+    left_xy[1] = left_xy[0] = mb_xy-1;
+    left_block = left_block_options[0];
+    if(FRAME_MBAFF){
+        const int pair_xy          = s->mb_x     + (s->mb_y & ~1)*s->mb_stride;
+        const int top_pair_xy      = pair_xy     - s->mb_stride;
+        const int topleft_pair_xy  = top_pair_xy - 1;
+        const int topright_pair_xy = top_pair_xy + 1;
+        const int topleft_mb_field_flag  = IS_INTERLACED(s->current_picture.mb_type[topleft_pair_xy]);
+        const int top_mb_field_flag      = IS_INTERLACED(s->current_picture.mb_type[top_pair_xy]);
+        const int topright_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[topright_pair_xy]);
+        const int left_mb_field_flag     = IS_INTERLACED(s->current_picture.mb_type[pair_xy-1]);
+        const int curr_mb_field_flag     = IS_INTERLACED(mb_type);
+        const int bottom = (s->mb_y & 1);
+        tprintf(s->avctx, "fill_caches: curr_mb_field_flag:%d, left_mb_field_flag:%d, topleft_mb_field_flag:%d, top_mb_field_flag:%d, topright_mb_field_flag:%d\n", curr_mb_field_flag, left_mb_field_flag, topleft_mb_field_flag, top_mb_field_flag, topright_mb_field_flag);
+
+        if (curr_mb_field_flag && (bottom || top_mb_field_flag)){
+            top_xy -= s->mb_stride;
+        }
+        if (curr_mb_field_flag && (bottom || topleft_mb_field_flag)){
+            topleft_xy -= s->mb_stride;
+        } else if(bottom && !curr_mb_field_flag && left_mb_field_flag) {
+            topleft_xy += s->mb_stride;
+            // take top left mv from the middle of the mb, as opposed to all other modes which use the bottom right partition
+            topleft_partition = 0;
+        }
+        if (curr_mb_field_flag && (bottom || topright_mb_field_flag)){
+            topright_xy -= s->mb_stride;
+        }
+        if (left_mb_field_flag != curr_mb_field_flag) {
+            left_xy[1] = left_xy[0] = pair_xy - 1;
+            if (curr_mb_field_flag) {
+                left_xy[1] += s->mb_stride;
+                left_block = left_block_options[3];
+            } else {
+                left_block= left_block_options[2 - bottom];
+            }
+        }
+    }
+
+    h->top_mb_xy = top_xy;
+    h->left_mb_xy[0] = left_xy[0];
+    h->left_mb_xy[1] = left_xy[1];
+    if(for_deblock){
+        topleft_type = 0;
+        topright_type = 0;
+        top_type     = h->slice_table[top_xy     ] < 0xFFFF ? s->current_picture.mb_type[top_xy]     : 0;
+        left_type[0] = h->slice_table[left_xy[0] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[0]] : 0;
+        left_type[1] = h->slice_table[left_xy[1] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[1]] : 0;
+
+        if(MB_MBAFF && !IS_INTRA(mb_type)){
+            int list;
+            for(list=0; list<h->list_count; list++){
+                //These values where changed for ease of performing MC, we need to change them back
+                //FIXME maybe we can make MC and loop filter use the same values or prevent
+                //the MC code from changing ref_cache and rather use a temporary array.
+                if(USES_LIST(mb_type,list)){
+                    int8_t *ref = &s->current_picture.ref_index[list][h->mb2b8_xy[mb_xy]];
+                    *(uint32_t*)&h->ref_cache[list][scan8[ 0]] =
+                    *(uint32_t*)&h->ref_cache[list][scan8[ 2]] = (pack16to32(ref[0],ref[1])&0x00FF00FF)*0x0101;
+                    ref += h->b8_stride;
+                    *(uint32_t*)&h->ref_cache[list][scan8[ 8]] =
+                    *(uint32_t*)&h->ref_cache[list][scan8[10]] = (pack16to32(ref[0],ref[1])&0x00FF00FF)*0x0101;
+                }
+            }
+        }
+    }else{
+        topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
+        top_type     = h->slice_table[top_xy     ] == h->slice_num ? s->current_picture.mb_type[top_xy]     : 0;
+        topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
+        left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
+        left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
+
+    if(IS_INTRA(mb_type)){
+        int type_mask= h->pps.constrained_intra_pred ? IS_INTRA(-1) : -1;
+        h->topleft_samples_available=
+        h->top_samples_available=
+        h->left_samples_available= 0xFFFF;
+        h->topright_samples_available= 0xEEEA;
+
+        if(!(top_type & type_mask)){
+            h->topleft_samples_available= 0xB3FF;
+            h->top_samples_available= 0x33FF;
+            h->topright_samples_available= 0x26EA;
+        }
+        if(IS_INTERLACED(mb_type) != IS_INTERLACED(left_type[0])){
+            if(IS_INTERLACED(mb_type)){
+                if(!(left_type[0] & type_mask)){
+                    h->topleft_samples_available&= 0xDFFF;
+                    h->left_samples_available&= 0x5FFF;
+                }
+                if(!(left_type[1] & type_mask)){
+                    h->topleft_samples_available&= 0xFF5F;
+                    h->left_samples_available&= 0xFF5F;
+                }
+            }else{
+                int left_typei = h->slice_table[left_xy[0] + s->mb_stride ] == h->slice_num
+                                ? s->current_picture.mb_type[left_xy[0] + s->mb_stride] : 0;
+                assert(left_xy[0] == left_xy[1]);
+                if(!((left_typei & type_mask) && (left_type[0] & type_mask))){
+                    h->topleft_samples_available&= 0xDF5F;
+                    h->left_samples_available&= 0x5F5F;
+                }
+            }
+        }else{
+            if(!(left_type[0] & type_mask)){
+                h->topleft_samples_available&= 0xDF5F;
+                h->left_samples_available&= 0x5F5F;
+            }
+        }
+
+        if(!(topleft_type & type_mask))
+            h->topleft_samples_available&= 0x7FFF;
+
+        if(!(topright_type & type_mask))
+            h->topright_samples_available&= 0xFBFF;
+
+        if(IS_INTRA4x4(mb_type)){
+            if(IS_INTRA4x4(top_type)){
+                h->intra4x4_pred_mode_cache[4+8*0]= h->intra4x4_pred_mode[top_xy][4];
+                h->intra4x4_pred_mode_cache[5+8*0]= h->intra4x4_pred_mode[top_xy][5];
+                h->intra4x4_pred_mode_cache[6+8*0]= h->intra4x4_pred_mode[top_xy][6];
+                h->intra4x4_pred_mode_cache[7+8*0]= h->intra4x4_pred_mode[top_xy][3];
+            }else{
+                int pred;
+                if(!(top_type & type_mask))
+                    pred= -1;
+                else{
+                    pred= 2;
+                }
+                h->intra4x4_pred_mode_cache[4+8*0]=
+                h->intra4x4_pred_mode_cache[5+8*0]=
+                h->intra4x4_pred_mode_cache[6+8*0]=
+                h->intra4x4_pred_mode_cache[7+8*0]= pred;
+            }
+            for(i=0; i<2; i++){
+                if(IS_INTRA4x4(left_type[i])){
+                    h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[0+2*i]];
+                    h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[1+2*i]];
+                }else{
+                    int pred;
+                    if(!(left_type[i] & type_mask))
+                        pred= -1;
+                    else{
+                        pred= 2;
+                    }
+                    h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
+                    h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= pred;
+                }
+            }
+        }
+    }
+    }
+
+
+/*
+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){
+        h->non_zero_count_cache[4+8*0]= h->non_zero_count[top_xy][4];
+        h->non_zero_count_cache[5+8*0]= h->non_zero_count[top_xy][5];
+        h->non_zero_count_cache[6+8*0]= h->non_zero_count[top_xy][6];
+        h->non_zero_count_cache[7+8*0]= h->non_zero_count[top_xy][3];
+
+        h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][9];
+        h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][8];
+
+        h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][12];
+        h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][11];
+
+    }else{
+        h->non_zero_count_cache[4+8*0]=
+        h->non_zero_count_cache[5+8*0]=
+        h->non_zero_count_cache[6+8*0]=
+        h->non_zero_count_cache[7+8*0]=
+
+        h->non_zero_count_cache[1+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]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+
+    }
+
+    for (i=0; i<2; i++) {
+        if(left_type[i]){
+            h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[0+2*i]];
+            h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[1+2*i]];
+            h->non_zero_count_cache[0+8*1 +   8*i]= h->non_zero_count[left_xy[i]][left_block[4+2*i]];
+            h->non_zero_count_cache[0+8*4 +   8*i]= h->non_zero_count[left_xy[i]][left_block[5+2*i]];
+        }else{
+            h->non_zero_count_cache[3+8*1 + 2*8*i]=
+            h->non_zero_count_cache[3+8*2 + 2*8*i]=
+            h->non_zero_count_cache[0+8*1 +   8*i]=
+            h->non_zero_count_cache[0+8*4 +   8*i]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+        }
+    }
+
+    if( h->pps.cabac ) {
+        // top_cbp
+        if(top_type) {
+            h->top_cbp = h->cbp_table[top_xy];
+        } else if(IS_INTRA(mb_type)) {
+            h->top_cbp = 0x1C0;
+        } else {
+            h->top_cbp = 0;
+        }
+        // left_cbp
+        if (left_type[0]) {
+            h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
+        } else if(IS_INTRA(mb_type)) {
+            h->left_cbp = 0x1C0;
+        } else {
+            h->left_cbp = 0;
+        }
+        if (left_type[0]) {
+            h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
+        }
+        if (left_type[1]) {
+            h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
+        }
+    }
+
+#if 1
+    if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
+        int list;
+        for(list=0; list<h->list_count; list++){
+            if(!USES_LIST(mb_type, list) && !IS_DIRECT(mb_type) && !h->deblocking_filter){
+                /*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;
+            }
+            h->mv_cache_clean[list]= 0;
+
+            if(USES_LIST(top_type, list)){
+                const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
+                const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
+                *(uint32_t*)h->mv_cache[list][scan8[0] + 0 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 0];
+                *(uint32_t*)h->mv_cache[list][scan8[0] + 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 1];
+                *(uint32_t*)h->mv_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 2];
+                *(uint32_t*)h->mv_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + 3];
+                h->ref_cache[list][scan8[0] + 0 - 1*8]=
+                h->ref_cache[list][scan8[0] + 1 - 1*8]= 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]= s->current_picture.ref_index[list][b8_xy + 1];
+            }else{
+                *(uint32_t*)h->mv_cache [list][scan8[0] + 0 - 1*8]=
+                *(uint32_t*)h->mv_cache [list][scan8[0] + 1 - 1*8]=
+                *(uint32_t*)h->mv_cache [list][scan8[0] + 2 - 1*8]=
+                *(uint32_t*)h->mv_cache [list][scan8[0] + 3 - 1*8]= 0;
+                *(uint32_t*)&h->ref_cache[list][scan8[0] + 0 - 1*8]= ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101;
+            }
+
+            for(i=0; i<2; i++){
+                int cache_idx = scan8[0] - 1 + i*2*8;
+                if(USES_LIST(left_type[i], list)){
+                    const int b_xy= h->mb2b_xy[left_xy[i]] + 3;
+                    const int b8_xy= h->mb2b8_xy[left_xy[i]] + 1;
+                    *(uint32_t*)h->mv_cache[list][cache_idx  ]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0+i*2]];
+                    *(uint32_t*)h->mv_cache[list][cache_idx+8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1+i*2]];
+                    h->ref_cache[list][cache_idx  ]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0+i*2]>>1)];
+                    h->ref_cache[list][cache_idx+8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[1+i*2]>>1)];
+                }else{
+                    *(uint32_t*)h->mv_cache [list][cache_idx  ]=
+                    *(uint32_t*)h->mv_cache [list][cache_idx+8]= 0;
+                    h->ref_cache[list][cache_idx  ]=
+                    h->ref_cache[list][cache_idx+8]= left_type[i] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+                }
+            }
+
+            if(for_deblock || ((IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred) && !FRAME_MBAFF))
+                continue;
+
+            if(USES_LIST(topleft_type, list)){
+                const int b_xy = h->mb2b_xy[topleft_xy] + 3 + h->b_stride + (topleft_partition & 2*h->b_stride);
+                const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + (topleft_partition & h->b8_stride);
+                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
+                h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
+            }else{
+                *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= 0;
+                h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+            }
+
+            if(USES_LIST(topright_type, list)){
+                const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
+                const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
+                *(uint32_t*)h->mv_cache[list][scan8[0] + 4 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
+                h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
+            }else{
+                *(uint32_t*)h->mv_cache [list][scan8[0] + 4 - 1*8]= 0;
+                h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+            }
+
+            if((IS_SKIP(mb_type) || IS_DIRECT(mb_type)) && !FRAME_MBAFF)
+                continue;
+
+            h->ref_cache[list][scan8[5 ]+1] =
+            h->ref_cache[list][scan8[7 ]+1] =
+            h->ref_cache[list][scan8[13]+1] =  //FIXME remove past 3 (init somewhere else)
+            h->ref_cache[list][scan8[4 ]] =
+            h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
+            *(uint32_t*)h->mv_cache [list][scan8[5 ]+1]=
+            *(uint32_t*)h->mv_cache [list][scan8[7 ]+1]=
+            *(uint32_t*)h->mv_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewhere else)
+            *(uint32_t*)h->mv_cache [list][scan8[4 ]]=
+            *(uint32_t*)h->mv_cache [list][scan8[12]]= 0;
+
+            if( h->pps.cabac ) {
+                /* XXX beurk, Load mvd */
+                if(USES_LIST(top_type, list)){
+                    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(USES_LIST(left_type[0], list)){
+                    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(USES_LIST(left_type[1], list)){
+                    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 somewhere else)
+                *(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
+                *(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;
+
+                if(h->slice_type_nos == FF_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;
+                    }
+
+                    if(IS_DIRECT(left_type[0]))
+                        h->direct_cache[scan8[0] - 1 + 0*8]= 1;
+                    else if(IS_8X8(left_type[0]))
+                        h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[h->mb2b8_xy[left_xy[0]] + 1 + h->b8_stride*(left_block[0]>>1)];
+                    else
+                        h->direct_cache[scan8[0] - 1 + 0*8]= 0;
+
+                    if(IS_DIRECT(left_type[1]))
+                        h->direct_cache[scan8[0] - 1 + 2*8]= 1;
+                    else if(IS_8X8(left_type[1]))
+                        h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[h->mb2b8_xy[left_xy[1]] + 1 + h->b8_stride*(left_block[2]>>1)];
+                    else
+                        h->direct_cache[scan8[0] - 1 + 2*8]= 0;
+                }
+            }
+
+            if(FRAME_MBAFF){
+#define MAP_MVS\
+                    MAP_F2F(scan8[0] - 1 - 1*8, topleft_type)\
+                    MAP_F2F(scan8[0] + 0 - 1*8, top_type)\
+                    MAP_F2F(scan8[0] + 1 - 1*8, top_type)\
+                    MAP_F2F(scan8[0] + 2 - 1*8, top_type)\
+                    MAP_F2F(scan8[0] + 3 - 1*8, top_type)\
+                    MAP_F2F(scan8[0] + 4 - 1*8, topright_type)\
+                    MAP_F2F(scan8[0] - 1 + 0*8, left_type[0])\
+                    MAP_F2F(scan8[0] - 1 + 1*8, left_type[0])\
+                    MAP_F2F(scan8[0] - 1 + 2*8, left_type[1])\
+                    MAP_F2F(scan8[0] - 1 + 3*8, left_type[1])
+                if(MB_FIELD){
+#define MAP_F2F(idx, mb_type)\
+                    if(!IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
+                        h->ref_cache[list][idx] <<= 1;\
+                        h->mv_cache[list][idx][1] /= 2;\
+                        h->mvd_cache[list][idx][1] /= 2;\
+                    }
+                    MAP_MVS
+#undef MAP_F2F
+                }else{
+#define MAP_F2F(idx, mb_type)\
+                    if(IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
+                        h->ref_cache[list][idx] >>= 1;\
+                        h->mv_cache[list][idx][1] <<= 1;\
+                        h->mvd_cache[list][idx][1] <<= 1;\
+                    }
+                    MAP_MVS
+#undef MAP_F2F
+                }
+            }
+        }
+    }
+#endif
+
+    h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
+}
+
+/**
+ * gets the predicted intra4x4 prediction mode.
+ */
+static inline int pred_intra_mode(H264Context *h, int n){
+    const int index8= scan8[n];
+    const int left= h->intra4x4_pred_mode_cache[index8 - 1];
+    const int top = h->intra4x4_pred_mode_cache[index8 - 8];
+    const int min= FFMIN(left, top);
+
+    tprintf(h->s.avctx, "mode:%d %d min:%d\n", left ,top, min);
+
+    if(min<0) return DC_PRED;
+    else      return min;
+}
+
+static inline void write_back_non_zero_count(H264Context *h){
+    const int mb_xy= h->mb_xy;
+
+    h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[7+8*1];
+    h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[7+8*2];
+    h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[7+8*3];
+    h->non_zero_count[mb_xy][3]= h->non_zero_count_cache[7+8*4];
+    h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[4+8*4];
+    h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[5+8*4];
+    h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[6+8*4];
+
+    h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[1+8*2];
+    h->non_zero_count[mb_xy][8]= h->non_zero_count_cache[2+8*2];
+    h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[2+8*1];
+
+    h->non_zero_count[mb_xy][12]=h->non_zero_count_cache[1+8*5];
+    h->non_zero_count[mb_xy][11]=h->non_zero_count_cache[2+8*5];
+    h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[2+8*4];
+}
+
+static inline void write_back_motion(H264Context *h, int mb_type){
+    MpegEncContext * const s = &h->s;
+    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;
+
+    if(!USES_LIST(mb_type, 0))
+        fill_rectangle(&s->current_picture.ref_index[0][b8_xy], 2, 2, h->b8_stride, (uint8_t)LIST_NOT_USED, 1);
+
+    for(list=0; list<h->list_count; list++){
+        int y;
+        if(!USES_LIST(mb_type, list))
+            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 ) {
+            if(IS_SKIP(mb_type))
+                fill_rectangle(h->mvd_table[list][b_xy], 4, 4, h->b_stride, 0, 4);
+            else
+            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];
+            }
+        }
+
+        {
+            int8_t *ref_index = &s->current_picture.ref_index[list][b8_xy];
+            ref_index[0+0*h->b8_stride]= h->ref_cache[list][scan8[0]];
+            ref_index[1+0*h->b8_stride]= h->ref_cache[list][scan8[4]];
+            ref_index[0+1*h->b8_stride]= h->ref_cache[list][scan8[8]];
+            ref_index[1+1*h->b8_stride]= h->ref_cache[list][scan8[12]];
+        }
+    }
+
+    if(h->slice_type_nos == FF_B_TYPE && h->pps.cabac){
+        if(IS_8X8(mb_type)){
+            uint8_t *direct_table = &h->direct_table[b8_xy];
+            direct_table[1+0*h->b8_stride] = IS_DIRECT(h->sub_mb_type[1]) ? 1 : 0;
+            direct_table[0+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[2]) ? 1 : 0;
+            direct_table[1+1*h->b8_stride] = IS_DIRECT(h->sub_mb_type[3]) ? 1 : 0;
+        }
+    }
+}
+
+static inline int get_dct8x8_allowed(H264Context *h){
+    if(h->sps.direct_8x8_inference_flag)
+        return !(*(uint64_t*)h->sub_mb_type & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8                )*0x0001000100010001ULL));
+    else
+        return !(*(uint64_t*)h->sub_mb_type & ((MB_TYPE_16x8|MB_TYPE_8x16|MB_TYPE_8x8|MB_TYPE_DIRECT2)*0x0001000100010001ULL));
+}
+
+static void predict_field_decoding_flag(H264Context *h){
+    MpegEncContext * const s = &h->s;
+    const int mb_xy= h->mb_xy;
+    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;
+}
+
+/**
+ * decodes a P_SKIP or B_SKIP macroblock
+ */
+static void decode_mb_skip(H264Context *h){
+    MpegEncContext * const s = &h->s;
+    const int mb_xy= h->mb_xy;
+    int mb_type=0;
+
+    memset(h->non_zero_count[mb_xy], 0, 16);
+    memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
+
+    if(MB_FIELD)
+        mb_type|= MB_TYPE_INTERLACED;
+
+    if( h->slice_type_nos == FF_B_TYPE )
+    {
+        // just for fill_caches. pred_direct_motion will set the real mb_type
+        mb_type|= MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
+
+        fill_caches(h, mb_type, 0); //FIXME check what is needed and what not ...
+        ff_h264_pred_direct_motion(h, &mb_type);
+        mb_type|= MB_TYPE_SKIP;
+    }
+    else
+    {
+        int mx, my;
+        mb_type|= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;
+
+        fill_caches(h, mb_type, 0); //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);
+    }
+
+    write_back_motion(h, mb_type);
+    s->current_picture.mb_type[mb_xy]= mb_type;
+    s->current_picture.qscale_table[mb_xy]= s->qscale;
+    h->slice_table[ mb_xy ]= h->slice_num;
+    h->prev_mb_skipped= 1;
+}
+
 #endif /* AVCODEC_H264_H */

libavcodec/h264data.h

@@ -38,173 +38,6 @@
 static const uint8_t golomb_to_pict_type[5]=
 {FF_P_TYPE, FF_B_TYPE, FF_I_TYPE, FF_SP_TYPE, FF_SI_TYPE};
 
-static const uint8_t golomb_to_intra4x4_cbp[48]={
- 47, 31, 15,  0, 23, 27, 29, 30,  7, 11, 13, 14, 39, 43, 45, 46,
- 16,  3,  5, 10, 12, 19, 21, 26, 28, 35, 37, 42, 44,  1,  2,  4,
-  8, 17, 18, 20, 24,  6,  9, 22, 25, 32, 33, 34, 36, 40, 38, 41
-};
-
-static const uint8_t golomb_to_inter_cbp[48]={
-  0, 16,  1,  2,  4,  8, 32,  3,  5, 10, 12, 15, 47,  7, 11, 13,
- 14,  6,  9, 31, 35, 37, 42, 44, 33, 34, 36, 40, 39, 43, 45, 46,
- 17, 18, 20, 24, 19, 21, 26, 28, 23, 27, 29, 30, 22, 25, 38, 41
-};
-
-static const uint8_t golomb_to_inter_cbp_gray[16]={
- 0, 1, 2, 4, 8, 3, 5,10,12,15, 7,11,13,14, 6, 9,
-};
-
-static const uint8_t golomb_to_intra4x4_cbp_gray[16]={
-15, 0, 7,11,13,14, 3, 5,10,12, 1, 2, 4, 8, 6, 9,
-};
-
-static const uint8_t chroma_dc_coeff_token_len[4*5]={
- 2, 0, 0, 0,
- 6, 1, 0, 0,
- 6, 6, 3, 0,
- 6, 7, 7, 6,
- 6, 8, 8, 7,
-};
-
-static const uint8_t chroma_dc_coeff_token_bits[4*5]={
- 1, 0, 0, 0,
- 7, 1, 0, 0,
- 4, 6, 1, 0,
- 3, 3, 2, 5,
- 2, 3, 2, 0,
-};
-
-static const uint8_t coeff_token_len[4][4*17]={
-{
-     1, 0, 0, 0,
-     6, 2, 0, 0,     8, 6, 3, 0,     9, 8, 7, 5,    10, 9, 8, 6,
-    11,10, 9, 7,    13,11,10, 8,    13,13,11, 9,    13,13,13,10,
-    14,14,13,11,    14,14,14,13,    15,15,14,14,    15,15,15,14,
-    16,15,15,15,    16,16,16,15,    16,16,16,16,    16,16,16,16,
-},
-{
-     2, 0, 0, 0,
-     6, 2, 0, 0,     6, 5, 3, 0,     7, 6, 6, 4,     8, 6, 6, 4,
-     8, 7, 7, 5,     9, 8, 8, 6,    11, 9, 9, 6,    11,11,11, 7,
-    12,11,11, 9,    12,12,12,11,    12,12,12,11,    13,13,13,12,
-    13,13,13,13,    13,14,13,13,    14,14,14,13,    14,14,14,14,
-},
-{
-     4, 0, 0, 0,
-     6, 4, 0, 0,     6, 5, 4, 0,     6, 5, 5, 4,     7, 5, 5, 4,
-     7, 5, 5, 4,     7, 6, 6, 4,     7, 6, 6, 4,     8, 7, 7, 5,
-     8, 8, 7, 6,     9, 8, 8, 7,     9, 9, 8, 8,     9, 9, 9, 8,
-    10, 9, 9, 9,    10,10,10,10,    10,10,10,10,    10,10,10,10,
-},
-{
-     6, 0, 0, 0,
-     6, 6, 0, 0,     6, 6, 6, 0,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-}
-};
-
-static const uint8_t coeff_token_bits[4][4*17]={
-{
-     1, 0, 0, 0,
-     5, 1, 0, 0,     7, 4, 1, 0,     7, 6, 5, 3,     7, 6, 5, 3,
-     7, 6, 5, 4,    15, 6, 5, 4,    11,14, 5, 4,     8,10,13, 4,
-    15,14, 9, 4,    11,10,13,12,    15,14, 9,12,    11,10,13, 8,
-    15, 1, 9,12,    11,14,13, 8,     7,10, 9,12,     4, 6, 5, 8,
-},
-{
-     3, 0, 0, 0,
-    11, 2, 0, 0,     7, 7, 3, 0,     7,10, 9, 5,     7, 6, 5, 4,
-     4, 6, 5, 6,     7, 6, 5, 8,    15, 6, 5, 4,    11,14,13, 4,
-    15,10, 9, 4,    11,14,13,12,     8,10, 9, 8,    15,14,13,12,
-    11,10, 9,12,     7,11, 6, 8,     9, 8,10, 1,     7, 6, 5, 4,
-},
-{
-    15, 0, 0, 0,
-    15,14, 0, 0,    11,15,13, 0,     8,12,14,12,    15,10,11,11,
-    11, 8, 9,10,     9,14,13, 9,     8,10, 9, 8,    15,14,13,13,
-    11,14,10,12,    15,10,13,12,    11,14, 9,12,     8,10,13, 8,
-    13, 7, 9,12,     9,12,11,10,     5, 8, 7, 6,     1, 4, 3, 2,
-},
-{
-     3, 0, 0, 0,
-     0, 1, 0, 0,     4, 5, 6, 0,     8, 9,10,11,    12,13,14,15,
-    16,17,18,19,    20,21,22,23,    24,25,26,27,    28,29,30,31,
-    32,33,34,35,    36,37,38,39,    40,41,42,43,    44,45,46,47,
-    48,49,50,51,    52,53,54,55,    56,57,58,59,    60,61,62,63,
-}
-};
-
-static const uint8_t total_zeros_len[16][16]= {
-    {1,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9},
-    {3,3,3,3,3,4,4,4,4,5,5,6,6,6,6},
-    {4,3,3,3,4,4,3,3,4,5,5,6,5,6},
-    {5,3,4,4,3,3,3,4,3,4,5,5,5},
-    {4,4,4,3,3,3,3,3,4,5,4,5},
-    {6,5,3,3,3,3,3,3,4,3,6},
-    {6,5,3,3,3,2,3,4,3,6},
-    {6,4,5,3,2,2,3,3,6},
-    {6,6,4,2,2,3,2,5},
-    {5,5,3,2,2,2,4},
-    {4,4,3,3,1,3},
-    {4,4,2,1,3},
-    {3,3,1,2},
-    {2,2,1},
-    {1,1},
-};
-
-static const uint8_t total_zeros_bits[16][16]= {
-    {1,3,2,3,2,3,2,3,2,3,2,3,2,3,2,1},
-    {7,6,5,4,3,5,4,3,2,3,2,3,2,1,0},
-    {5,7,6,5,4,3,4,3,2,3,2,1,1,0},
-    {3,7,5,4,6,5,4,3,3,2,2,1,0},
-    {5,4,3,7,6,5,4,3,2,1,1,0},
-    {1,1,7,6,5,4,3,2,1,1,0},
-    {1,1,5,4,3,3,2,1,1,0},
-    {1,1,1,3,3,2,2,1,0},
-    {1,0,1,3,2,1,1,1},
-    {1,0,1,3,2,1,1},
-    {0,1,1,2,1,3},
-    {0,1,1,1,1},
-    {0,1,1,1},
-    {0,1,1},
-    {0,1},
-};
-
-static const uint8_t chroma_dc_total_zeros_len[3][4]= {
-    { 1, 2, 3, 3,},
-    { 1, 2, 2, 0,},
-    { 1, 1, 0, 0,},
-};
-
-static const uint8_t chroma_dc_total_zeros_bits[3][4]= {
-    { 1, 1, 1, 0,},
-    { 1, 1, 0, 0,},
-    { 1, 0, 0, 0,},
-};
-
-static const uint8_t run_len[7][16]={
-    {1,1},
-    {1,2,2},
-    {2,2,2,2},
-    {2,2,2,3,3},
-    {2,2,3,3,3,3},
-    {2,3,3,3,3,3,3},
-    {3,3,3,3,3,3,3,4,5,6,7,8,9,10,11},
-};
-
-static const uint8_t run_bits[7][16]={
-    {1,0},
-    {1,1,0},
-    {3,2,1,0},
-    {3,2,1,1,0},
-    {3,2,3,2,1,0},
-    {3,0,1,3,2,5,4},
-    {7,6,5,4,3,2,1,1,1,1,1,1,1,1,1},
-};
-
-
 static const uint8_t zigzag_scan[16]={
  0+0*4, 1+0*4, 0+1*4, 0+2*4,
  1+1*4, 2+0*4, 3+0*4, 2+1*4,