hevc: add wavefront parallel processing

cherry picked from commit b971f2c8fdc60f8bab605a6e8060492eb548a53a
cherry picked from commit e57b0a2c915ce6b8a9d57b8292f6581f0680842e

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

9 files changed, 415 insertions, 121 deletions

diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index cdf0237c23..58a95ad131 100644
--- a/libavcodec/hevc.c
+++ b/libavcodec/hevc.c
@@ -23,6 +23,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include "libavutil/atomic.h"
 #include "libavutil/attributes.h"
 #include "libavutil/common.h"
 #include "libavutil/internal.h"
@@ -70,6 +71,10 @@ static void pic_arrays_free(HEVCContext *s)
     av_freep(&s->horizontal_bs);
     av_freep(&s->vertical_bs);
 
+    av_freep(&s->sh.entry_point_offset);
+    av_freep(&s->sh.size);
+    av_freep(&s->sh.offset);
+
     av_buffer_pool_uninit(&s->tab_mvf_pool);
     av_buffer_pool_uninit(&s->rpl_tab_pool);
 }
@@ -284,9 +289,9 @@ static int decode_lt_rps(HEVCContext *s, LongTermRPS *rps, GetBitContext *gb)
 
 static int hls_slice_header(HEVCContext *s)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     SliceHeader   *sh = &s->sh;
-    int i, ret;
+    int i, j, ret;
 
     // Coded parameters
     sh->first_slice_in_pic_flag = get_bits1(gb);
@@ -590,10 +595,33 @@ static int hls_slice_header(HEVCContext *s)
         sh->num_entry_point_offsets = get_ue_golomb_long(gb);
         if (sh->num_entry_point_offsets > 0) {
             int offset_len = get_ue_golomb_long(gb) + 1;
-
-            for (i = 0; i < sh->num_entry_point_offsets; i++)
-                skip_bits(gb, offset_len);
-        }
+            int segments = offset_len >> 4;
+            int rest = (offset_len & 15);
+            av_freep(&sh->entry_point_offset);
+            av_freep(&sh->offset);
+            av_freep(&sh->size);
+            sh->entry_point_offset = av_malloc(sh->num_entry_point_offsets * sizeof(int));
+            sh->offset = av_malloc(sh->num_entry_point_offsets * sizeof(int));
+            sh->size = av_malloc(sh->num_entry_point_offsets * sizeof(int));
+            for (i = 0; i < sh->num_entry_point_offsets; i++) {
+                int val = 0;
+                for (j = 0; j < segments; j++) {
+                    val <<= 16;
+                    val += get_bits(gb, 16);
+                }
+                if (rest) {
+                    val <<= rest;
+                    val += get_bits(gb, rest);
+                }
+                sh->entry_point_offset[i] = val + 1; // +1; // +1 to get the size
+            }
+            if (s->threads_number > 1 && (s->pps->num_tile_rows > 1 || s->pps->num_tile_columns > 1)) {
+                s->enable_parallel_tiles = 0; // TODO: you can enable tiles in parallel here
+                s->threads_number = 1;
+            } else
+                s->enable_parallel_tiles = 0;
+        } else
+            s->enable_parallel_tiles = 0;
     }
 
     if (s->pps->slice_header_extension_present_flag) {
@@ -606,10 +634,10 @@ static int hls_slice_header(HEVCContext *s)
     sh->slice_qp = 26 + s->pps->pic_init_qp_minus26 + sh->slice_qp_delta;
     sh->slice_ctb_addr_rs = sh->slice_segment_addr;
 
-    s->HEVClc.first_qp_group = !s->sh.dependent_slice_segment_flag;
+    s->HEVClc->first_qp_group = !s->sh.dependent_slice_segment_flag;
 
     if (!s->pps->cu_qp_delta_enabled_flag)
-        s->HEVClc.qp_y = ((s->sh.slice_qp + 52 + 2 * s->sps->qp_bd_offset) %
+        s->HEVClc->qp_y = ((s->sh.slice_qp + 52 + 2 * s->sps->qp_bd_offset) %
                           (52 + s->sps->qp_bd_offset)) - s->sps->qp_bd_offset;
 
     s->slice_initialized = 1;
@@ -633,7 +661,7 @@ do {                                                    \
 
 static void hls_sao_param(HEVCContext *s, int rx, int ry)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int sao_merge_left_flag = 0;
     int sao_merge_up_flag   = 0;
     int shift = s->sps->bit_depth - FFMIN(s->sps->bit_depth, 10);
@@ -705,7 +733,7 @@ static void hls_sao_param(HEVCContext *s, int rx, int ry)
 static void hls_transform_unit(HEVCContext *s, int x0, int  y0, int xBase, int yBase, int cb_xBase, int cb_yBase,
                                int log2_cb_size, int log2_trafo_size, int trafo_depth, int blk_idx)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int scan_idx = SCAN_DIAG;
     int scan_idx_c = SCAN_DIAG;
 
@@ -791,7 +819,7 @@ static void set_deblocking_bypass(HEVCContext *s, int x0, int y0, int log2_cb_si
 static void hls_transform_tree(HEVCContext *s, int x0, int y0, int xBase, int yBase, int cb_xBase, int cb_yBase,
                                int log2_cb_size, int log2_trafo_size, int trafo_depth, int blk_idx)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     uint8_t split_transform_flag;
 
     if (trafo_depth > 0 && log2_trafo_size == 2) {
@@ -889,7 +917,7 @@ static void hls_transform_tree(HEVCContext *s, int x0, int y0, int xBase, int yB
 static int hls_pcm_sample(HEVCContext *s, int x0, int y0, int log2_cb_size)
 {
     //TODO: non-4:2:0 support
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     GetBitContext gb;
     int cb_size = 1 << log2_cb_size;
     int    stride0 = s->frame->linesize[0];
@@ -900,7 +928,7 @@ static int hls_pcm_sample(HEVCContext *s, int x0, int y0, int log2_cb_size)
     uint8_t *dst2 = &s->frame->data[2][(y0 >> s->sps->vshift[2]) * stride2 + ((x0 >> s->sps->hshift[2]) << s->sps->pixel_shift)];
 
     int length = cb_size * cb_size * s->sps->pcm.bit_depth + ((cb_size * cb_size) >> 1) * s->sps->pcm.bit_depth;
-    const uint8_t *pcm = skip_bytes(&s->HEVClc.cc, (length + 7) >> 3);
+    const uint8_t *pcm = skip_bytes(&s->HEVClc->cc, (length + 7) >> 3);
     int ret;
 
     ff_hevc_deblocking_boundary_strengths(s, x0, y0, log2_cb_size,
@@ -933,7 +961,7 @@ static int hls_pcm_sample(HEVCContext *s, int x0, int y0, int log2_cb_size)
 static void luma_mc(HEVCContext *s, int16_t *dst, ptrdiff_t dststride, AVFrame *ref,
                     const Mv *mv, int x_off, int y_off, int block_w, int block_h)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     uint8_t *src = ref->data[0];
     ptrdiff_t srcstride = ref->linesize[0];
     int pic_width = s->sps->width;
@@ -979,7 +1007,7 @@ static void luma_mc(HEVCContext *s, int16_t *dst, ptrdiff_t dststride, AVFrame *
 static void chroma_mc(HEVCContext *s, int16_t *dst1, int16_t *dst2, ptrdiff_t dststride, AVFrame *ref,
                       const Mv *mv, int x_off, int y_off, int block_w, int block_h)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     uint8_t *src1 = ref->data[1];
     uint8_t *src2 = ref->data[2];
     ptrdiff_t src1stride = ref->linesize[1];
@@ -1030,7 +1058,8 @@ static void hevc_await_progress(HEVCContext *s, HEVCFrame *ref,
     int y = (mv->y >> 2) + y0;
 
     //ff_thread_await_progress(&ref->tf, FFMIN(s->height, y), 0);
-    ff_thread_await_progress(&ref->tf, INT_MAX, 0);
+    if (s->threads_type == FF_THREAD_FRAME )
+        ff_thread_await_progress(&ref->tf, INT_MAX, 0);
 }
 
 static void hls_prediction_unit(HEVCContext *s, int x0, int y0, int nPbW, int nPbH, int log2_cb_size, int partIdx)
@@ -1038,7 +1067,7 @@ static void hls_prediction_unit(HEVCContext *s, int x0, int y0, int nPbW, int nP
 #define POS(c_idx, x, y)                                                              \
     &s->frame->data[c_idx][((y) >> s->sps->vshift[c_idx]) * s->frame->linesize[c_idx] + \
                            (((x) >> s->sps->hshift[c_idx]) << s->sps->pixel_shift)]
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int merge_idx = 0;
     enum InterPredIdc inter_pred_idc = PRED_L0;
     struct MvField current_mv = {{{ 0 }}};
@@ -1288,7 +1317,7 @@ static void hls_prediction_unit(HEVCContext *s, int x0, int y0, int nPbW, int nP
 static int luma_intra_pred_mode(HEVCContext *s, int x0, int y0, int pu_size,
                                 int prev_intra_luma_pred_flag)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int x_pu = x0 >> s->sps->log2_min_pu_size;
     int y_pu = y0 >> s->sps->log2_min_pu_size;
     int pic_width_in_min_pu = s->sps->width >> s->sps->log2_min_pu_size;
@@ -1386,7 +1415,7 @@ static av_always_inline void set_ct_depth(HEVCContext *s, int x0, int y0,
 
 static void intra_prediction_unit(HEVCContext *s, int x0, int y0, int log2_cb_size)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     static const uint8_t intra_chroma_table[4] = {0, 26, 10, 1};
     uint8_t prev_intra_luma_pred_flag[4];
     int split   = lc->cu.part_mode == PART_NxN;
@@ -1425,7 +1454,7 @@ static void intra_prediction_unit(HEVCContext *s, int x0, int y0, int log2_cb_si
 
 static void intra_prediction_unit_default_value(HEVCContext *s, int x0, int y0, int log2_cb_size)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int pb_size = 1 << log2_cb_size;
     int size_in_pus = pb_size >> s->sps->log2_min_pu_size;
     int pic_width_in_min_pu = s->sps->width >> s->sps->log2_min_pu_size;
@@ -1446,7 +1475,7 @@ static void intra_prediction_unit_default_value(HEVCContext *s, int x0, int y0,
 static int hls_coding_unit(HEVCContext *s, int x0, int y0, int log2_cb_size)
 {
     int cb_size          = 1 << log2_cb_size;
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int log2_min_cb_size = s->sps->log2_min_coding_block_size;
     int length           = cb_size >> log2_min_cb_size;
     int pic_width_in_ctb = s->sps->width >> log2_min_cb_size;
@@ -1595,7 +1624,7 @@ static int hls_coding_unit(HEVCContext *s, int x0, int y0, int log2_cb_size)
 
 static int hls_coding_quadtree(HEVCContext *s, int x0, int y0, int log2_cb_size, int cb_depth)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int ret;
 
     lc->ct.depth = cb_depth;
@@ -1663,7 +1692,7 @@ static int hls_coding_quadtree(HEVCContext *s, int x0, int y0, int log2_cb_size,
 
 static void hls_decode_neighbour(HEVCContext *s, int x_ctb, int y_ctb, int ctb_addr_ts)
 {
-    HEVCLocalContext *lc  = &s->HEVClc;
+    HEVCLocalContext *lc  = s->HEVClc;
     int ctb_size          = 1 << s->sps->log2_ctb_size;
     int ctb_addr_rs       = s->pps->ctb_addr_ts_to_rs[ctb_addr_ts];
     int ctb_addr_in_slice = ctb_addr_rs - s->sh.slice_addr;
@@ -1716,8 +1745,9 @@ static void hls_decode_neighbour(HEVCContext *s, int x_ctb, int y_ctb, int ctb_a
     lc->ctb_up_left_flag = ((x_ctb > 0) && (y_ctb > 0)  && (ctb_addr_in_slice-1 >= s->sps->ctb_width) && (s->pps->tile_id[ctb_addr_ts] == s->pps->tile_id[s->pps->ctb_addr_rs_to_ts[ctb_addr_rs-1 - s->sps->ctb_width]]));
 }
 
-static int hls_slice_data(HEVCContext *s)
+static int hls_decode_entry(AVCodecContext *avctxt, void *isFilterThread)
 {
+    HEVCContext *s  = avctxt->priv_data;
     int ctb_size    = 1 << s->sps->log2_ctb_size;
     int more_data   = 1;
     int x_ctb       = 0;
@@ -1755,13 +1785,175 @@ static int hls_slice_data(HEVCContext *s)
     return ctb_addr_ts;
 }
 
+static int hls_slice_data(HEVCContext *s)
+{
+    int arg[2];
+    int ret[2];
+
+    arg[0] = 0;
+    arg[1] = 1;
+
+    s->avctx->execute(s->avctx, hls_decode_entry, arg, ret , 1, sizeof(int));
+    return ret[0];
+}
+static int hls_decode_entry_wpp(AVCodecContext *avctxt, void *input_ctb_row, int job, int self_id)
+{
+    HEVCContext *s1  = avctxt->priv_data, *s;
+    HEVCLocalContext *lc;
+    int ctb_size    = 1<< s1->sps->log2_ctb_size;
+    int more_data   = 1;
+    int *ctb_row_p    = input_ctb_row;
+    int ctb_row = ctb_row_p[job];
+    int ctb_addr_rs = s1->sh.slice_ctb_addr_rs + ctb_row * ((s1->sps->width + ctb_size - 1) >> s1->sps->log2_ctb_size);
+    int ctb_addr_ts = s1->pps->ctb_addr_rs_to_ts[ctb_addr_rs];
+    int thread = ctb_row % s1->threads_number;
+    int ret;
+
+    s = s1->sList[self_id];
+    lc = s->HEVClc;
+
+    if(ctb_row) {
+        ret = init_get_bits8(&lc->gb, s->data + s->sh.offset[ctb_row - 1], s->sh.size[ctb_row - 1]);
+
+        if (ret < 0)
+            return ret;
+        ff_init_cabac_decoder(&lc->cc, s->data + s->sh.offset[(ctb_row)-1], s->sh.size[ctb_row - 1]);
+    }
+
+    while(more_data && ctb_addr_ts < s->sps->ctb_size) {
+        int x_ctb = (ctb_addr_rs % s->sps->ctb_width) << s->sps->log2_ctb_size;
+        int y_ctb = (ctb_addr_rs / s->sps->ctb_width) << s->sps->log2_ctb_size;
+
+        hls_decode_neighbour(s, x_ctb, y_ctb, ctb_addr_ts);
+
+        ff_thread_await_progress2(s->avctx, ctb_row, thread, SHIFT_CTB_WPP);
+
+        if (avpriv_atomic_int_get(&s1->wpp_err)){
+            ff_thread_report_progress2(s->avctx, ctb_row , thread, SHIFT_CTB_WPP);
+            return 0;
+        }
+
+        ff_hevc_cabac_init(s, ctb_addr_ts);
+        hls_sao_param(s, x_ctb >> s->sps->log2_ctb_size, y_ctb >> s->sps->log2_ctb_size);
+        more_data = hls_coding_quadtree(s, x_ctb, y_ctb, s->sps->log2_ctb_size, 0);
+
+        if (more_data < 0)
+            return more_data;
+
+        ctb_addr_ts++;
+
+        ff_hevc_save_states(s, ctb_addr_ts);
+        ff_thread_report_progress2(s->avctx, ctb_row, thread, 1);
+        ff_hevc_hls_filters(s, x_ctb, y_ctb, ctb_size);
+
+        if (!more_data && (x_ctb+ctb_size) < s->sps->width && ctb_row != s->sh.num_entry_point_offsets) {
+            avpriv_atomic_int_set(&s1->wpp_err,  1);
+            ff_thread_report_progress2(s->avctx, ctb_row ,thread, SHIFT_CTB_WPP);
+            return 0;
+        }
+
+        if ((x_ctb+ctb_size) >= s->sps->width && (y_ctb+ctb_size) >= s->sps->height ) {
+            ff_hevc_hls_filter(s, x_ctb, y_ctb);
+            ff_thread_report_progress2(s->avctx, ctb_row , thread, SHIFT_CTB_WPP);
+            return ctb_addr_ts;
+        }
+        ctb_addr_rs       = s->pps->ctb_addr_ts_to_rs[ctb_addr_ts];
+        x_ctb+=ctb_size;
+
+        if(x_ctb >= s->sps->width) {
+            break;
+        }
+    }
+    ff_thread_report_progress2(s->avctx, ctb_row ,thread, SHIFT_CTB_WPP);
+
+    return 0;
+}
+
+static int hls_slice_data_wpp(HEVCContext *s, const uint8_t *nal, int length)
+{
+    HEVCLocalContext *lc = s->HEVClc;
+    int *ret = av_malloc((s->sh.num_entry_point_offsets + 1) * sizeof(int));
+    int *arg = av_malloc((s->sh.num_entry_point_offsets + 1) * sizeof(int));
+    int offset;
+    int startheader, cmpt = 0;
+    int i, j, res = 0;
+
+
+    if (!s->sList[1]) {
+        ff_alloc_entries(s->avctx, s->sh.num_entry_point_offsets + 1);
+
+
+        for (i = 1; i < s->threads_number; i++) {
+            s->sList[i] = av_malloc(sizeof(HEVCContext));
+            memcpy(s->sList[i], s, sizeof(HEVCContext));
+            s->HEVClcList[i] = av_malloc(sizeof(HEVCLocalContext));
+            s->HEVClcList[i]->edge_emu_buffer = av_malloc((MAX_PB_SIZE + 7) * s->frame->linesize[0]);
+            s->sList[i]->HEVClc = s->HEVClcList[i];
+        }
+    }
+
+    offset = (lc->gb.index >> 3);
+
+    for (j = 0, cmpt = 0, startheader = offset + s->sh.entry_point_offset[0]; j < s->skipped_bytes; j++) {
+        if (s->skipped_bytes_pos[j] >= offset && s->skipped_bytes_pos[j] < startheader) {
+            startheader--;
+            cmpt++;
+        }
+    }
+
+    for (i = 1; i < s->sh.num_entry_point_offsets; i++) {
+        offset += (s->sh.entry_point_offset[i - 1] - cmpt);
+        for (j = 0, cmpt = 0, startheader = offset
+             + s->sh.entry_point_offset[i]; j < s->skipped_bytes; j++) {
+            if (s->skipped_bytes_pos[j] >= offset && s->skipped_bytes_pos[j] < startheader) {
+                startheader--;
+                cmpt++;
+            }
+        }
+        s->sh.size[i - 1] = s->sh.entry_point_offset[i] - cmpt;
+        s->sh.offset[i - 1] = offset;
+
+    }
+    if (s->sh.num_entry_point_offsets != 0) {
+        offset += s->sh.entry_point_offset[s->sh.num_entry_point_offsets - 1] - cmpt;
+        s->sh.size[s->sh.num_entry_point_offsets - 1] = length - offset;
+        s->sh.offset[s->sh.num_entry_point_offsets - 1] = offset;
+
+    }
+    s->data = nal;
+
+    for (i = 1; i < s->threads_number; i++) {
+        s->sList[i]->HEVClc->first_qp_group = 1;
+        s->sList[i]->HEVClc->qp_y = s->sList[0]->HEVClc->qp_y;
+        memcpy(s->sList[i], s, sizeof(HEVCContext));
+        s->sList[i]->HEVClc = s->HEVClcList[i];
+    }
+
+    avpriv_atomic_int_set(&s->wpp_err, 0);
+    ff_reset_entries(s->avctx);
+
+    for (i = 0; i <= s->sh.num_entry_point_offsets; i++) {
+        arg[i] = i;
+        ret[i] = 0;
+    }
+
+    if (s->pps->entropy_coding_sync_enabled_flag)
+        s->avctx->execute2(s->avctx, (void *) hls_decode_entry_wpp, arg, ret, s->sh.num_entry_point_offsets + 1);
+
+    for (i = 0; i <= s->sh.num_entry_point_offsets; i++)
+        res += ret[i];
+    av_free(ret);
+    av_free(arg);
+    return res;
+}
+
 /**
  * @return AVERROR_INVALIDDATA if the packet is not a valid NAL unit,
  * 0 if the unit should be skipped, 1 otherwise
  */
 static int hls_nal_unit(HEVCContext *s)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     int nuh_layer_id;
 
     if (get_bits1(gb) != 0)
@@ -1812,7 +2004,7 @@ static void restore_tqb_pixels(HEVCContext *s)
 
 static int hevc_frame_start(HEVCContext *s)
 {
-    HEVCLocalContext *lc     = &s->HEVClc;
+    HEVCLocalContext *lc     = s->HEVClc;
     int pic_width_in_min_pu  = s->sps->width  >> s->sps->log2_min_pu_size;
     int pic_height_in_min_pu = s->sps->height >> s->sps->log2_min_pu_size;
     int pic_width_in_min_tu  = s->sps->width  >> s->sps->log2_min_transform_block_size;
@@ -1857,7 +2049,7 @@ static int hevc_frame_start(HEVCContext *s)
 
     return 0;
 fail:
-    if (s->ref)
+    if (s->ref && s->threads_type == FF_THREAD_FRAME)
         ff_thread_report_progress(&s->ref->tf, INT_MAX, 0);
     s->ref = NULL;
     return ret;
@@ -1865,7 +2057,7 @@ fail:
 
 static int decode_nal_unit(HEVCContext *s, const uint8_t *nal, int length)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     GetBitContext *gb = &lc->gb;
     int ctb_addr_ts;
     int ret;
@@ -1967,7 +2159,11 @@ static int decode_nal_unit(HEVCContext *s, const uint8_t *nal, int length)
             }
         }
 
-        ctb_addr_ts = hls_slice_data(s);
+        if (s->threads_number > 1 && s->sh.num_entry_point_offsets > 0)
+            ctb_addr_ts = hls_slice_data_wpp(s, nal, length);
+        else
+            ctb_addr_ts = hls_slice_data(s);
+
         if (ctb_addr_ts >= (s->sps->ctb_width * s->sps->ctb_height)) {
             s->is_decoded = 1;
             if ((s->pps->transquant_bypass_enable_flag ||
@@ -1996,12 +2192,13 @@ static int decode_nal_unit(HEVCContext *s, const uint8_t *nal, int length)
 
 /* FIXME: This is adapted from ff_h264_decode_nal, avoiding duplication
    between these functions would be nice. */
-static int extract_rbsp(const uint8_t *src, int length,
+static int extract_rbsp(HEVCContext *s, const uint8_t *src, int length,
                         HEVCNAL *nal)
 {
     int i, si, di;
     uint8_t *dst;
 
+    s->skipped_bytes = 0;
 #define STARTCODE_TEST                                                  \
         if (i + 2 < length && src[i + 1] == 0 && src[i + 2] <= 3) {     \
             if (src[i + 2] != 3) {                                      \
@@ -2073,6 +2270,16 @@ static int extract_rbsp(const uint8_t *src, int length,
                 dst[di++]  = 0;
                 si        += 3;
 
+                s->skipped_bytes++;
+                if (s->skipped_bytes_pos_size < s->skipped_bytes) {
+                    s->skipped_bytes_pos_size *= 2;
+                    av_reallocp_array(&s->skipped_bytes_pos,
+                            s->skipped_bytes_pos_size,
+                            sizeof(*s->skipped_bytes_pos));
+                    if (!s->skipped_bytes_pos)
+                        return NULL;
+                }
+                s->skipped_bytes_pos[s->skipped_bytes-1] = di - 1;
                 continue;
             } else // next start code
                 goto nsc;
@@ -2144,17 +2351,30 @@ static int decode_nal_units(HEVCContext *s, const uint8_t *buf, int length)
             }
             s->nals = tmp;
             memset(s->nals + s->nals_allocated, 0, (new_size - s->nals_allocated) * sizeof(*tmp));
+            av_reallocp_array(&s->skipped_bytes_nal, new_size, sizeof(*s->skipped_bytes_nal));
+            av_reallocp_array(&s->skipped_bytes_pos_size_nal, new_size, sizeof(*s->skipped_bytes_pos_size_nal));
+            av_reallocp_array(&s->skipped_bytes_pos_nal, new_size, sizeof(*s->skipped_bytes_pos_nal));
+            s->skipped_bytes_pos_size_nal[s->nals_allocated] = 1024; // initial buffer size
+            s->skipped_bytes_pos_nal[s->nals_allocated] = av_malloc_array(s->skipped_bytes_pos_size_nal[s->nals_allocated], sizeof(*s->skipped_bytes_pos));
             s->nals_allocated = new_size;
         }
-        nal = &s->nals[s->nb_nals++];
+        s->skipped_bytes_pos_size = s->skipped_bytes_pos_size_nal[s->nb_nals];
+        s->skipped_bytes_pos = s->skipped_bytes_pos_nal[s->nb_nals];
+        nal = &s->nals[s->nb_nals];
+
+        consumed = extract_rbsp(s, buf, extract_length, nal);
+
+        s->skipped_bytes_nal[s->nb_nals] = s->skipped_bytes;
+        s->skipped_bytes_pos_size_nal[s->nb_nals] = s->skipped_bytes_pos_size;
+        s->skipped_bytes_pos_nal[s->nb_nals++] = s->skipped_bytes_pos;
+
 
-        consumed = extract_rbsp(buf, extract_length, nal);
         if (consumed < 0) {
             ret = consumed;
             goto fail;
         }
 
-        ret = init_get_bits8(&s->HEVClc.gb, nal->data, nal->size);
+        ret = init_get_bits8(&s->HEVClc->gb, nal->data, nal->size);
         if (ret < 0)
             goto fail;
         hls_nal_unit(s);
@@ -2168,7 +2388,11 @@ static int decode_nal_units(HEVCContext *s, const uint8_t *buf, int length)
 
     /* parse the NAL units */
     for (i = 0; i < s->nb_nals; i++) {
-        int ret = decode_nal_unit(s, s->nals[i].data, s->nals[i].size);
+        int ret;
+        s->skipped_bytes = s->skipped_bytes_nal[i];
+        s->skipped_bytes_pos = s->skipped_bytes_pos_nal[i];
+
+        ret = decode_nal_unit(s, s->nals[i].data, s->nals[i].size);
         if (ret < 0) {
             av_log(s->avctx, AV_LOG_WARNING, "Error parsing NAL unit #%d.\n", i);
             if (s->avctx->err_recognition & AV_EF_EXPLODE)
@@ -2177,7 +2401,7 @@ static int decode_nal_units(HEVCContext *s, const uint8_t *buf, int length)
     }
 
 fail:
-    if (s->ref)
+    if (s->ref && s->threads_type == FF_THREAD_FRAME)
         ff_thread_report_progress(&s->ref->tf, INT_MAX, 0);
 
     return ret;
@@ -2337,7 +2561,7 @@ fail:
 static av_cold int hevc_decode_free(AVCodecContext *avctx)
 {
     HEVCContext       *s = avctx->priv_data;
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int i;
 
     pic_arrays_free(s);
@@ -2345,6 +2569,15 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
     av_freep(&lc->edge_emu_buffer);
     av_freep(&s->md5_ctx);
 
+    for(i=0; i < s->nals_allocated; i++) {
+        av_freep(&s->skipped_bytes_pos_nal[i]);
+    }
+    av_freep(&s->skipped_bytes_pos_size_nal);
+    av_freep(&s->skipped_bytes_nal);
+    av_freep(&s->skipped_bytes_pos_nal);
+
+    av_freep(&s->cabac_state);
+
     av_frame_free(&s->tmp_frame);
     av_frame_free(&s->output_frame);
 
@@ -2360,6 +2593,21 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
     for (i = 0; i < FF_ARRAY_ELEMS(s->pps_list); i++)
         av_buffer_unref(&s->pps_list[i]);
 
+    av_freep(&s->sh.entry_point_offset);
+    av_freep(&s->sh.offset);
+    av_freep(&s->sh.size);
+
+    for (i = 1; i < s->threads_number; i++) {
+        lc = s->HEVClcList[i];
+        if (lc) {
+            av_freep(&lc->edge_emu_buffer);
+
+            av_freep(&s->HEVClcList[i]);
+            av_freep(&s->sList[i]);
+        }
+    }
+    av_freep(&s->HEVClcList[0]);
+
     for (i = 0; i < s->nals_allocated; i++)
         av_freep(&s->nals[i].rbsp_buffer);
     av_freep(&s->nals);
@@ -2375,6 +2623,16 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
 
     s->avctx = avctx;
 
+    s->HEVClc = av_mallocz(sizeof(HEVCLocalContext));
+    if (!s->HEVClc)
+        goto fail;
+    s->HEVClcList[0] = s->HEVClc;
+    s->sList[0] = s;
+
+    s->cabac_state = av_malloc(HEVC_CONTEXTS);
+    if (!s->cabac_state)
+        goto fail;
+
     s->tmp_frame = av_frame_alloc();
     if (!s->tmp_frame)
         goto fail;
@@ -2454,6 +2712,8 @@ static int hevc_update_thread_context(AVCodecContext *dst,
     s->is_nalff        = s0->is_nalff;
     s->nal_length_size = s0->nal_length_size;
 
+    s->threads_number      = s0->threads_number;
+    s->threads_type        = s0->threads_type;
 
     if (s0->eos) {
         s->seq_decode = (s->seq_decode + 1) & 0xff;
@@ -2538,6 +2798,13 @@ static av_cold int hevc_decode_init(AVCodecContext *avctx)
     if (ret < 0)
         return ret;
 
+    s->enable_parallel_tiles = 0;
+
+    if(avctx->active_thread_type & FF_THREAD_SLICE)
+        s->threads_number = avctx->thread_count;
+    else
+        s->threads_number = 1;
+
     if (avctx->extradata_size > 0 && avctx->extradata) {
         ret = hevc_decode_extradata(s);
         if (ret < 0) {
@@ -2546,6 +2813,11 @@ static av_cold int hevc_decode_init(AVCodecContext *avctx)
         }
     }
 
+    if((avctx->active_thread_type & FF_THREAD_FRAME) && avctx->thread_count > 1)
+            s->threads_type = FF_THREAD_FRAME;
+        else
+            s->threads_type = FF_THREAD_SLICE;
+
     return 0;
 }
 
@@ -2598,5 +2870,5 @@ AVCodec ff_hevc_decoder = {
     .flush                 = hevc_decode_flush,
     .update_thread_context = hevc_update_thread_context,
     .init_thread_copy      = hevc_init_thread_copy,
-    .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_FRAME_THREADS | CODEC_CAP_EXPERIMENTAL,
+    .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_SLICE_THREADS | CODEC_CAP_FRAME_THREADS | CODEC_CAP_EXPERIMENTAL,
 };
diff --git a/libavcodec/hevc.h b/libavcodec/hevc.h
index 000ccd46d6..01db69b74b 100644
--- a/libavcodec/hevc.h
+++ b/libavcodec/hevc.h
@@ -39,6 +39,9 @@
 #define MAX_DPB_SIZE 16 // A.4.1
 #define MAX_REFS 16
 
+#define MAX_NB_THREADS 16
+#define SHIFT_CTB_WPP 2
+
 /**
  * 7.4.2.1
  */
@@ -563,6 +566,9 @@ typedef struct SliceHeader {
 
     uint8_t slice_loop_filter_across_slices_enabled_flag;
 
+    int *entry_point_offset;
+    int * offset;
+    int * size;
     int num_entry_point_offsets;
 
     uint8_t luma_log2_weight_denom;
@@ -705,14 +711,6 @@ typedef struct HEVCFrame {
     AVBufferRef *rpl_buf;
 } HEVCFrame;
 
-typedef struct FilterData {
-    int x;
-    int y;
-    int size;
-    int slice_or_tiles_left_boundary;
-    int slice_or_tiles_up_boundary;
-} FilterData;
-
 typedef struct HEVCNAL {
     uint8_t *rbsp_buffer;
     int rbsp_buffer_size;
@@ -745,20 +743,24 @@ typedef struct HEVCLocalContext {
     PredictionUnit pu;
     NeighbourAvailable na;
     DECLARE_ALIGNED(16, int16_t, mc_buffer[(MAX_PB_SIZE + 7) * MAX_PB_SIZE]);
-    FilterData *save_boundary_strengths;
-    int nb_saved;
 } HEVCLocalContext;
 
 typedef struct HEVCContext {
     const AVClass *c;  // needed by private avoptions
     AVCodecContext      *avctx;
 
-    HEVCLocalContext     HEVClc;
+    struct HEVCContext  *sList[MAX_NB_THREADS];
+
+    HEVCLocalContext    *HEVClcList[MAX_NB_THREADS];
+    HEVCLocalContext    *HEVClc;
+
+    uint8_t             threads_type;
+    uint8_t             threads_number;
 
     int                 width;
     int                 height;
 
-    uint8_t cabac_state[HEVC_CONTEXTS];
+    uint8_t *cabac_state;
 
     AVFrame *frame;
     AVFrame *sao_frame;
@@ -826,6 +828,18 @@ typedef struct HEVCContext {
     uint16_t seq_decode;
     uint16_t seq_output;
 
+    int enable_parallel_tiles;
+    int wpp_err;
+    int skipped_bytes;
+    int *skipped_bytes_pos;
+    int skipped_bytes_pos_size;
+
+    int *skipped_bytes_nal;
+    int **skipped_bytes_pos_nal;
+    int *skipped_bytes_pos_size_nal;
+
+    uint8_t *data;
+
     HEVCNAL *nals;
     int nb_nals;
     int nals_allocated;
diff --git a/libavcodec/hevc_cabac.c b/libavcodec/hevc_cabac.c
index e6cbf1a4ce..2f6d0addde 100644
--- a/libavcodec/hevc_cabac.c
+++ b/libavcodec/hevc_cabac.c
@@ -507,13 +507,13 @@ void ff_hevc_save_states(HEVCContext *s, int ctb_addr_ts)
         ((ctb_addr_ts % s->sps->ctb_width) == 2 ||
          (s->sps->ctb_width == 2 &&
           (ctb_addr_ts % s->sps->ctb_width) == 0))) {
-        memcpy(s->cabac_state, s->HEVClc.cabac_state, HEVC_CONTEXTS);
+        memcpy(s->cabac_state, s->HEVClc->cabac_state, HEVC_CONTEXTS);
     }
 }
 
 static void load_states(HEVCContext *s)
 {
-    memcpy(s->HEVClc.cabac_state, s->cabac_state, HEVC_CONTEXTS);
+    memcpy(s->HEVClc->cabac_state, s->cabac_state, HEVC_CONTEXTS);
 }
 
 static void cabac_reinit(HEVCLocalContext *lc)
@@ -523,10 +523,10 @@ static void cabac_reinit(HEVCLocalContext *lc)
 
 static void cabac_init_decoder(HEVCContext *s)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     skip_bits(gb, 1);
     align_get_bits(gb);
-    ff_init_cabac_decoder(&s->HEVClc.cc,
+    ff_init_cabac_decoder(&s->HEVClc->cc,
                           gb->buffer + get_bits_count(gb) / 8,
                           (get_bits_left(gb) + 7) / 8);
 }
@@ -548,7 +548,7 @@ static void cabac_init_state(HEVCContext *s)
         pre ^= pre >> 31;
         if (pre > 124)
             pre = 124 + (pre & 1);
-        s->HEVClc.cabac_state[i] =  pre;
+        s->HEVClc->cabac_state[i] =  pre;
     }
 }
 
@@ -572,13 +572,19 @@ void ff_hevc_cabac_init(HEVCContext *s, int ctb_addr_ts)
     } else {
         if (s->pps->tiles_enabled_flag &&
             (s->pps->tile_id[ctb_addr_ts] != s->pps->tile_id[ctb_addr_ts - 1])) {
-            cabac_reinit(&s->HEVClc);
+            if (s->threads_number == 1)
+                cabac_reinit(s->HEVClc);
+            else
+                cabac_init_decoder(s);
             cabac_init_state(s);
         }
         if (s->pps->entropy_coding_sync_enabled_flag) {
             if ((ctb_addr_ts % s->sps->ctb_width) == 0) {
-                get_cabac_terminate(&s->HEVClc.cc);
-                cabac_reinit(&s->HEVClc);
+                get_cabac_terminate(&s->HEVClc->cc);
+                if (s->threads_number == 1)
+                    cabac_reinit(s->HEVClc);
+                else
+                    cabac_init_decoder(s);
 
                 if (s->sps->ctb_width == 1)
                     cabac_init_state(s);
@@ -589,7 +595,7 @@ void ff_hevc_cabac_init(HEVCContext *s, int ctb_addr_ts)
     }
 }
 
-#define GET_CABAC(ctx) get_cabac(&s->HEVClc.cc, &s->HEVClc.cabac_state[ctx])
+#define GET_CABAC(ctx) get_cabac(&s->HEVClc->cc, &s->HEVClc->cabac_state[ctx])
 
 int ff_hevc_sao_merge_flag_decode(HEVCContext *s)
 {
@@ -601,7 +607,7 @@ int ff_hevc_sao_type_idx_decode(HEVCContext *s)
     if (!GET_CABAC(elem_offset[SAO_TYPE_IDX]))
         return 0;
 
-    if (!get_cabac_bypass(&s->HEVClc.cc))
+    if (!get_cabac_bypass(&s->HEVClc->cc))
         return SAO_BAND;
     return SAO_EDGE;
 }
@@ -609,10 +615,10 @@ int ff_hevc_sao_type_idx_decode(HEVCContext *s)
 int ff_hevc_sao_band_position_decode(HEVCContext *s)
 {
     int i;
-    int value = get_cabac_bypass(&s->HEVClc.cc);
+    int value = get_cabac_bypass(&s->HEVClc->cc);
 
     for (i = 0; i < 4; i++)
-        value = (value << 1) | get_cabac_bypass(&s->HEVClc.cc);
+        value = (value << 1) | get_cabac_bypass(&s->HEVClc->cc);
     return value;
 }
 
@@ -621,26 +627,26 @@ int ff_hevc_sao_offset_abs_decode(HEVCContext *s)
     int i = 0;
     int length = (1 << (FFMIN(s->sps->bit_depth, 10) - 5)) - 1;
 
-    while (i < length && get_cabac_bypass(&s->HEVClc.cc))
+    while (i < length && get_cabac_bypass(&s->HEVClc->cc))
         i++;
     return i;
 }
 
 int ff_hevc_sao_offset_sign_decode(HEVCContext *s)
 {
-    return get_cabac_bypass(&s->HEVClc.cc);
+    return get_cabac_bypass(&s->HEVClc->cc);
 }
 
 int ff_hevc_sao_eo_class_decode(HEVCContext *s)
 {
-    int ret = (get_cabac_bypass(&s->HEVClc.cc) << 1);
-    ret    |=  get_cabac_bypass(&s->HEVClc.cc);
+    int ret = (get_cabac_bypass(&s->HEVClc->cc) << 1);
+    ret    |=  get_cabac_bypass(&s->HEVClc->cc);
     return ret;
 }
 
 int ff_hevc_end_of_slice_flag_decode(HEVCContext *s)
 {
-    return get_cabac_terminate(&s->HEVClc.cc);
+    return get_cabac_terminate(&s->HEVClc->cc);
 }
 
 int ff_hevc_cu_transquant_bypass_flag_decode(HEVCContext *s)
@@ -655,9 +661,9 @@ int ff_hevc_skip_flag_decode(HEVCContext *s, int x0, int y0, int x_cb, int y_cb)
     int x0b = x0 & ((1 << s->sps->log2_ctb_size) - 1);
     int y0b = y0 & ((1 << s->sps->log2_ctb_size) - 1);
 
-    if (s->HEVClc.ctb_left_flag || x0b)
+    if (s->HEVClc->ctb_left_flag || x0b)
         inc = SAMPLE_CTB(s->skip_flag, x_cb-1, y_cb);
-    if (s->HEVClc.ctb_up_flag || y0b)
+    if (s->HEVClc->ctb_up_flag || y0b)
         inc += SAMPLE_CTB(s->skip_flag, x_cb, y_cb-1);
 
     return GET_CABAC(elem_offset[SKIP_FLAG] + inc);
@@ -675,7 +681,7 @@ int ff_hevc_cu_qp_delta_abs(HEVCContext *s)
     }
     if (prefix_val >= 5) {
         int k = 0;
-        while (k < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc.cc)) {
+        while (k < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc->cc)) {
             suffix_val += 1 << k;
             k++;
         }
@@ -683,14 +689,14 @@ int ff_hevc_cu_qp_delta_abs(HEVCContext *s)
             av_log(s->avctx, AV_LOG_ERROR, "CABAC_MAX_BIN : %d\n", k);
 
         while (k--)
-            suffix_val += get_cabac_bypass(&s->HEVClc.cc) << k;
+            suffix_val += get_cabac_bypass(&s->HEVClc->cc) << k;
     }
     return prefix_val + suffix_val;
 }
 
 int ff_hevc_cu_qp_delta_sign_flag(HEVCContext *s)
 {
-    return get_cabac_bypass(&s->HEVClc.cc);
+    return get_cabac_bypass(&s->HEVClc->cc);
 }
 
 int ff_hevc_pred_mode_decode(HEVCContext *s)
@@ -706,9 +712,9 @@ int ff_hevc_split_coding_unit_flag_decode(HEVCContext *s, int ct_depth, int x0,
     int x_cb = x0 >> s->sps->log2_min_coding_block_size;
     int y_cb = y0 >> s->sps->log2_min_coding_block_size;
 
-    if (s->HEVClc.ctb_left_flag || x0b)
+    if (s->HEVClc->ctb_left_flag || x0b)
         depth_left = s->tab_ct_depth[(y_cb)*s->sps->min_cb_width + x_cb-1];
-    if (s->HEVClc.ctb_up_flag || y0b)
+    if (s->HEVClc->ctb_up_flag || y0b)
         depth_top = s->tab_ct_depth[(y_cb-1)*s->sps->min_cb_width + x_cb];
 
     inc += (depth_left > ct_depth);
@@ -721,7 +727,7 @@ int ff_hevc_part_mode_decode(HEVCContext *s, int log2_cb_size)
     if (GET_CABAC(elem_offset[PART_MODE])) // 1
         return PART_2Nx2N;
     if (log2_cb_size == s->sps->log2_min_coding_block_size) {
-        if (s->HEVClc.cu.pred_mode == MODE_INTRA) // 0
+        if (s->HEVClc->cu.pred_mode == MODE_INTRA) // 0
             return PART_NxN;
         if (GET_CABAC(elem_offset[PART_MODE] + 1)) // 01
             return PART_2NxN;
@@ -741,21 +747,21 @@ int ff_hevc_part_mode_decode(HEVCContext *s, int log2_cb_size)
     if (GET_CABAC(elem_offset[PART_MODE] + 1)) { // 01X, 01XX
         if (GET_CABAC(elem_offset[PART_MODE] + 3)) // 011
             return PART_2NxN;
-        if (get_cabac_bypass(&s->HEVClc.cc)) // 0101
+        if (get_cabac_bypass(&s->HEVClc->cc)) // 0101
             return PART_2NxnD;
         return PART_2NxnU; // 0100
     }
 
     if (GET_CABAC(elem_offset[PART_MODE] + 3)) // 001
         return PART_Nx2N;
-    if (get_cabac_bypass(&s->HEVClc.cc)) // 0001
+    if (get_cabac_bypass(&s->HEVClc->cc)) // 0001
         return PART_nRx2N;
     return  PART_nLx2N; // 0000
 }
 
 int ff_hevc_pcm_flag_decode(HEVCContext *s)
 {
-    return get_cabac_terminate(&s->HEVClc.cc);
+    return get_cabac_terminate(&s->HEVClc->cc);
 }
 
 int ff_hevc_prev_intra_luma_pred_flag_decode(HEVCContext *s)
@@ -766,7 +772,7 @@ int ff_hevc_prev_intra_luma_pred_flag_decode(HEVCContext *s)
 int ff_hevc_mpm_idx_decode(HEVCContext *s)
 {
     int i = 0;
-    while (i < 2 && get_cabac_bypass(&s->HEVClc.cc))
+    while (i < 2 && get_cabac_bypass(&s->HEVClc->cc))
         i++;
     return i;
 }
@@ -774,10 +780,10 @@ int ff_hevc_mpm_idx_decode(HEVCContext *s)
 int ff_hevc_rem_intra_luma_pred_mode_decode(HEVCContext *s)
 {
     int i;
-    int value = get_cabac_bypass(&s->HEVClc.cc);
+    int value = get_cabac_bypass(&s->HEVClc->cc);
 
     for (i = 0; i < 4; i++)
-        value = (value << 1) | get_cabac_bypass(&s->HEVClc.cc);
+        value = (value << 1) | get_cabac_bypass(&s->HEVClc->cc);
     return value;
 }
 
@@ -787,8 +793,8 @@ int ff_hevc_intra_chroma_pred_mode_decode(HEVCContext *s)
     if (!GET_CABAC(elem_offset[INTRA_CHROMA_PRED_MODE]))
         return 4;
 
-    ret  = (get_cabac_bypass(&s->HEVClc.cc) << 1);
-    ret |=  get_cabac_bypass(&s->HEVClc.cc);
+    ret  = (get_cabac_bypass(&s->HEVClc->cc) << 1);
+    ret |=  get_cabac_bypass(&s->HEVClc->cc);
     return ret;
 }
 
@@ -797,7 +803,7 @@ int ff_hevc_merge_idx_decode(HEVCContext *s)
     int i = GET_CABAC(elem_offset[MERGE_IDX]);
 
     if (i != 0) {
-        while (i < s->sh.max_num_merge_cand-1 && get_cabac_bypass(&s->HEVClc.cc))
+        while (i < s->sh.max_num_merge_cand-1 && get_cabac_bypass(&s->HEVClc->cc))
             i++;
     }
     return i;
@@ -812,7 +818,7 @@ int ff_hevc_inter_pred_idc_decode(HEVCContext *s, int nPbW, int nPbH)
 {
     if (nPbW + nPbH == 12)
         return GET_CABAC(elem_offset[INTER_PRED_IDC] + 4);
-    if (GET_CABAC(elem_offset[INTER_PRED_IDC] + s->HEVClc.ct.depth))
+    if (GET_CABAC(elem_offset[INTER_PRED_IDC] + s->HEVClc->ct.depth))
         return PRED_BI;
 
     return GET_CABAC(elem_offset[INTER_PRED_IDC] + 4);
@@ -827,7 +833,7 @@ int ff_hevc_ref_idx_lx_decode(HEVCContext *s, int num_ref_idx_lx)
     while (i < max_ctx && GET_CABAC(elem_offset[REF_IDX_L0] + i))
         i++;
     if (i == 2) {
-        while (i < max && get_cabac_bypass(&s->HEVClc.cc))
+        while (i < max && get_cabac_bypass(&s->HEVClc->cc))
             i++;
     }
 
@@ -859,20 +865,20 @@ static av_always_inline int mvd_decode(HEVCContext *s)
     int ret = 2;
     int k = 1;
 
-    while (k < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc.cc)) {
+    while (k < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc->cc)) {
         ret += 1 << k;
         k++;
     }
     if (k == CABAC_MAX_BIN)
         av_log(s->avctx, AV_LOG_ERROR, "CABAC_MAX_BIN : %d\n", k);
     while (k--)
-        ret += get_cabac_bypass(&s->HEVClc.cc) << k;
-    return get_cabac_bypass_sign(&s->HEVClc.cc, -ret);
+        ret += get_cabac_bypass(&s->HEVClc->cc) << k;
+    return get_cabac_bypass_sign(&s->HEVClc->cc, -ret);
 }
 
 static av_always_inline int mvd_sign_flag_decode(HEVCContext *s)
 {
-    return get_cabac_bypass_sign(&s->HEVClc.cc, -1);
+    return get_cabac_bypass_sign(&s->HEVClc->cc, -1);
 }
 
 int ff_hevc_split_transform_flag_decode(HEVCContext *s, int log2_trafo_size)
@@ -929,10 +935,10 @@ static av_always_inline int last_significant_coeff_suffix_decode(HEVCContext *s,
 {
     int i;
     int length = (last_significant_coeff_prefix >> 1) - 1;
-    int value = get_cabac_bypass(&s->HEVClc.cc);
+    int value = get_cabac_bypass(&s->HEVClc->cc);
 
     for (i = 1; i < length; i++)
-        value = (value << 1) | get_cabac_bypass(&s->HEVClc.cc);
+        value = (value << 1) | get_cabac_bypass(&s->HEVClc->cc);
     return value;
 }
 
@@ -1021,18 +1027,18 @@ static av_always_inline int coeff_abs_level_remaining_decode(HEVCContext *s, int
     int last_coeff_abs_level_remaining;
     int i;
 
-    while (prefix < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc.cc))
+    while (prefix < CABAC_MAX_BIN && get_cabac_bypass(&s->HEVClc->cc))
         prefix++;
     if (prefix == CABAC_MAX_BIN)
         av_log(s->avctx, AV_LOG_ERROR, "CABAC_MAX_BIN : %d\n", prefix);
     if (prefix < 3) {
         for (i = 0; i < rc_rice_param; i++)
-            suffix = (suffix << 1) | get_cabac_bypass(&s->HEVClc.cc);
+            suffix = (suffix << 1) | get_cabac_bypass(&s->HEVClc->cc);
         last_coeff_abs_level_remaining = (prefix << rc_rice_param) + suffix;
     } else {
         int prefix_minus3 = prefix - 3;
         for (i = 0; i < prefix_minus3 + rc_rice_param; i++)
-            suffix = (suffix << 1) | get_cabac_bypass(&s->HEVClc.cc);
+            suffix = (suffix << 1) | get_cabac_bypass(&s->HEVClc->cc);
         last_coeff_abs_level_remaining = (((1 << prefix_minus3) + 3 - 1)
                                               << rc_rice_param) + suffix;
     }
@@ -1045,7 +1051,7 @@ static av_always_inline int coeff_sign_flag_decode(HEVCContext *s, uint8_t nb)
     int ret = 0;
 
     for (i = 0; i < nb; i++)
-        ret = (ret << 1) | get_cabac_bypass(&s->HEVClc.cc);
+        ret = (ret << 1) | get_cabac_bypass(&s->HEVClc->cc);
     return ret;
 }
 
@@ -1058,7 +1064,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
         x_c = (scan_x_cg[offset >> 4] << 2) + scan_x_off[n];    \
         y_c = (scan_y_cg[offset >> 4] << 2) + scan_y_off[n];    \
     } while (0)
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int transform_skip_flag = 0;
 
     int last_significant_coeff_x, last_significant_coeff_y;
@@ -1394,7 +1400,7 @@ void ff_hevc_hls_residual_coding(HEVCContext *s, int x0, int y0,
 
 void ff_hevc_hls_mvd_coding(HEVCContext *s, int x0, int y0, int log2_cb_size)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int x = abs_mvd_greater0_flag_decode(s);
     int y = abs_mvd_greater0_flag_decode(s);
 
diff --git a/libavcodec/hevc_filter.c b/libavcodec/hevc_filter.c
index d67cdd5cba..84e21dcaa9 100644
--- a/libavcodec/hevc_filter.c
+++ b/libavcodec/hevc_filter.c
@@ -73,7 +73,7 @@ static int chroma_tc(HEVCContext *s, int qp_y, int c_idx, int tc_offset)
 
 static int get_qPy_pred(HEVCContext *s, int xC, int yC, int xBase, int yBase, int log2_cb_size)
 {
-    HEVCLocalContext *lc     = &s->HEVClc;
+    HEVCLocalContext *lc     = s->HEVClc;
     int ctb_size_mask        = (1 << s->sps->log2_ctb_size) - 1;
     int MinCuQpDeltaSizeMask = (1 << (s->sps->log2_ctb_size - s->pps->diff_cu_qp_delta_depth)) - 1;
     int xQgBase              = xBase - ( xBase & MinCuQpDeltaSizeMask );
@@ -153,11 +153,11 @@ void ff_hevc_set_qPy(HEVCContext *s, int xC, int yC, int xBase, int yBase, int l
 {
     int qp_y = get_qPy_pred(s, xC, yC, xBase, yBase, log2_cb_size);
 
-    if (s->HEVClc.tu.cu_qp_delta != 0) {
+    if (s->HEVClc->tu.cu_qp_delta != 0) {
         int off = s->sps->qp_bd_offset;
-        s->HEVClc.qp_y = ((qp_y + s->HEVClc.tu.cu_qp_delta + 52 + 2 * off) % (52 + off)) - off;
+        s->HEVClc->qp_y = ((qp_y + s->HEVClc->tu.cu_qp_delta + 52 + 2 * off) % (52 + off)) - off;
     } else
-        s->HEVClc.qp_y = qp_y;
+        s->HEVClc->qp_y = qp_y;
 }
 
 static int get_qPy(HEVCContext *s, int xC, int yC)
diff --git a/libavcodec/hevc_mvs.c b/libavcodec/hevc_mvs.c
index 701510903b..0b9286bd7b 100644
--- a/libavcodec/hevc_mvs.c
+++ b/libavcodec/hevc_mvs.c
@@ -40,7 +40,7 @@ static const uint8_t l0_l1_cand_idx[12][2] = {
 
 void ff_hevc_set_neighbour_available(HEVCContext *s, int x0, int y0, int nPbW, int nPbH)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int x0b = x0 & ((1 << s->sps->log2_ctb_size) - 1);
     int y0b = y0 & ((1 << s->sps->log2_ctb_size) - 1);
 
@@ -98,7 +98,7 @@ static int check_prediction_block_available(HEVCContext *s, int log2_cb_size,
                                             int x0, int y0, int nPbW, int nPbH,
                                             int xA1, int yA1, int partIdx)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
 
     if (lc->cu.x < xA1 && lc->cu.y < yA1 &&
         (lc->cu.x + (1 << log2_cb_size)) > xA1 &&
@@ -272,7 +272,8 @@ static int temporal_luma_motion_vector(HEVCContext *s, int x0, int y0,
     xPRb = x0 + nPbW;
     yPRb = y0 + nPbH;
 
-    ff_thread_await_progress(&ref->tf, INT_MAX, 0);
+    if (s->threads_type == FF_THREAD_FRAME )
+        ff_thread_await_progress(&ref->tf, INT_MAX, 0);
     if (tab_mvf &&
         y0 >> s->sps->log2_ctb_size == yPRb >> s->sps->log2_ctb_size &&
         yPRb < s->sps->height &&
@@ -322,7 +323,7 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
                                             int singleMCLFlag, int part_idx,
                                             struct MvField mergecandlist[])
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     RefPicList *refPicList = s->ref->refPicList;
     MvField *tab_mvf = s->ref->tab_mvf;
 
@@ -707,7 +708,7 @@ void ff_hevc_luma_mv_merge_mode(HEVCContext *s, int x0, int y0, int nPbW,
     struct MvField mergecand_list[MRG_MAX_NUM_CANDS] = { { { { 0 } } } };
     int nPbW2 = nPbW;
     int nPbH2 = nPbH;
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
 
     if (s->pps->log2_parallel_merge_level > 2 && nCS == 8) {
         singleMCLFlag = 1;
@@ -794,7 +795,7 @@ void ff_hevc_luma_mv_mvp_mode(HEVCContext *s, int x0, int y0, int nPbW,
                               int merge_idx, MvField *mv,
                               int mvp_lx_flag, int LX)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     MvField *tab_mvf = s->ref->tab_mvf;
     int isScaledFlag_L0 = 0;
     int availableFlagLXA0 = 0;
diff --git a/libavcodec/hevc_ps.c b/libavcodec/hevc_ps.c
index 20cb4eb4c3..3c13411f25 100644
--- a/libavcodec/hevc_ps.c
+++ b/libavcodec/hevc_ps.c
@@ -72,7 +72,7 @@ static const AVRational vui_sar[] = {
 int ff_hevc_decode_short_term_rps(HEVCContext *s, ShortTermRPS *rps,
                                   const HEVCSPS *sps, int is_slice_header)
 {
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     uint8_t rps_predict = 0;
     int delta_poc;
     int k0 = 0;
@@ -253,7 +253,7 @@ static void decode_hrd(HEVCContext *s)
 int ff_hevc_decode_nal_vps(HEVCContext *s)
 {
     int i,j;
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     int vps_id = 0;
     VPS *vps;
 
@@ -289,7 +289,7 @@ int ff_hevc_decode_nal_vps(HEVCContext *s)
         goto err;
     }
 
-    if (decode_profile_tier_level(&s->HEVClc, &vps->ptl, vps->vps_max_sub_layers) < 0) {
+    if (decode_profile_tier_level(s->HEVClc, &vps->ptl, vps->vps_max_sub_layers) < 0) {
         av_log(s->avctx, AV_LOG_ERROR, "Error decoding profile tier level.\n");
         goto err;
     }
@@ -347,7 +347,7 @@ err:
 static void decode_vui(HEVCContext *s, HEVCSPS *sps)
 {
     VUI *vui = &sps->vui;
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     int sar_present;
 
     av_log(s->avctx, AV_LOG_DEBUG, "Decoding VUI\n");
@@ -469,7 +469,7 @@ static void set_default_scaling_list_data(ScalingList *sl)
 
 static int scaling_list_data(HEVCContext *s, ScalingList *sl)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     uint8_t scaling_list_pred_mode_flag[4][6];
     int32_t scaling_list_dc_coef[2][6];
 
@@ -525,7 +525,7 @@ static int scaling_list_data(HEVCContext *s, ScalingList *sl)
 int ff_hevc_decode_nal_sps(HEVCContext *s)
 {
     const AVPixFmtDescriptor *desc;
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     int ret    = 0;
     int sps_id = 0;
     int log2_diff_max_min_transform_block_size;
@@ -559,7 +559,7 @@ int ff_hevc_decode_nal_sps(HEVCContext *s)
     }
 
     skip_bits1(gb); // temporal_id_nesting_flag
-    if (decode_profile_tier_level(&s->HEVClc, &sps->ptl, sps->max_sub_layers) < 0) {
+    if (decode_profile_tier_level(s->HEVClc, &sps->ptl, sps->max_sub_layers) < 0) {
         av_log(s->avctx, AV_LOG_ERROR, "error decoding profile tier level\n");
         ret = AVERROR_INVALIDDATA;
         goto err;
@@ -888,7 +888,7 @@ static void hevc_pps_free(void *opaque, uint8_t *data)
 
 int ff_hevc_decode_nal_pps(HEVCContext *s)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     HEVCSPS      *sps = NULL;
     int pic_area_in_ctbs, pic_area_in_min_cbs, pic_area_in_min_tbs;
     int log2_diff_ctb_min_tb_size;
diff --git a/libavcodec/hevc_refs.c b/libavcodec/hevc_refs.c
index c4a3b461c1..ed17d2ac92 100644
--- a/libavcodec/hevc_refs.c
+++ b/libavcodec/hevc_refs.c
@@ -351,7 +351,8 @@ static HEVCFrame *generate_missing_ref(HEVCContext *s, int poc)
     frame->sequence = s->seq_decode;
     frame->flags    = 0;
 
-    ff_thread_report_progress(&frame->tf, INT_MAX, 0);
+    if (s->threads_type == FF_THREAD_FRAME)
+        ff_thread_report_progress(&frame->tf, INT_MAX, 0);
 
     return frame;
 }
diff --git a/libavcodec/hevc_sei.c b/libavcodec/hevc_sei.c
index abdbc001e1..1a964477de 100644
--- a/libavcodec/hevc_sei.c
+++ b/libavcodec/hevc_sei.c
@@ -31,7 +31,7 @@ static void decode_nal_sei_decoded_picture_hash(HEVCContext *s, int payload_size
     uint8_t hash_type;
     //uint16_t picture_crc;
     //uint32_t picture_checksum;
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
     hash_type = get_bits(gb, 8);
 
 
@@ -79,7 +79,7 @@ static void decode_nal_sei_frame_packing_arrangement(HEVCLocalContext *lc)
 
 static int decode_nal_sei_message(HEVCContext *s)
 {
-    GetBitContext *gb = &s->HEVClc.gb;
+    GetBitContext *gb = &s->HEVClc->gb;
 
     int payload_type = 0;
     int payload_size = 0;
@@ -99,7 +99,7 @@ static int decode_nal_sei_message(HEVCContext *s)
         if (payload_type == 256 /*&& s->decode_checksum_sei*/)
             decode_nal_sei_decoded_picture_hash(s, payload_size);
         else if (payload_type == 45)
-            decode_nal_sei_frame_packing_arrangement(&s->HEVClc);
+            decode_nal_sei_frame_packing_arrangement(s->HEVClc);
         else {
             av_log(s->avctx, AV_LOG_DEBUG, "Skipped PREFIX SEI %d\n", payload_type);
             skip_bits(gb, 8*payload_size);
@@ -124,6 +124,6 @@ int ff_hevc_decode_nal_sei(HEVCContext *s)
 {
     do {
         decode_nal_sei_message(s);
-    } while (more_rbsp_data(&s->HEVClc.gb));
+    } while (more_rbsp_data(&s->HEVClc->gb));
     return 0;
 }
diff --git a/libavcodec/hevcpred_template.c b/libavcodec/hevcpred_template.c
index 30f7a51181..969750675c 100644
--- a/libavcodec/hevcpred_template.c
+++ b/libavcodec/hevcpred_template.c
@@ -65,7 +65,7 @@ static void FUNC(intra_pred)(HEVCContext *s, int x0, int y0, int log2_size, int
         for (i = (start); i < (start) + (length); i++) \
             if (!IS_INTRA(-1, i)) \
                 ptr[i] = ptr[i - 1]
-    HEVCLocalContext *lc = &s->HEVClc;
+    HEVCLocalContext *lc = s->HEVClc;
     int i;
     int hshift = s->sps->hshift[c_idx];
     int vshift = s->sps->vshift[c_idx];