[FFmpeg-devel] [PATCH 5/9] hevc: derive partial merge list

Christophe Gisquet christophe.gisquet at gmail.com
Thu Jul 10 16:09:39 CEST 2014


Note: this is also used for skip blocks, so the speed improvement of
this patch should be around 1 or 2%. (amvp probably much less)

2014-07-10 8:47 GMT+02:00 Christophe Gisquet <christophe.gisquet at gmail.com>:
> The merge list only needs to be derived up to the merge index.
> ---
>  libavcodec/hevc_mvs.c | 35 ++++++++++++++++++++++++++---------
>  1 file changed, 26 insertions(+), 9 deletions(-)
>
> diff --git a/libavcodec/hevc_mvs.c b/libavcodec/hevc_mvs.c
> index 2b017dd..0302c88 100644
> --- a/libavcodec/hevc_mvs.c
> +++ b/libavcodec/hevc_mvs.c
> @@ -315,6 +315,7 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>                                              int nPbW, int nPbH,
>                                              int log2_cb_size,
>                                              int singleMCLFlag, int part_idx,
> +                                            int merge_idx,
>                                              struct MvField mergecandlist[])
>  {
>      HEVCLocalContext *lc   = s->HEVClc;
> @@ -383,8 +384,11 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>          is_available_a1 = 0;
>      }
>
> -    if (is_available_a1)
> -        mergecandlist[nb_merge_cand++] = TAB_MVF_PU(A1);
> +    if (is_available_a1) {
> +        mergecandlist[0] = TAB_MVF_PU(A1);
> +        if (merge_idx == 0) return;
> +        nb_merge_cand++;
> +    }
>
>      // above spatial merge candidate
>      is_available_b1 = AVAILABLE(cand_up, B1);
> @@ -415,8 +419,11 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>      if (is_available_b1 && is_available_b0)
>          check_MER = !COMPARE_MV_REFIDX(B0, B1);
>
> -    if (is_available_b0 && check_MER)
> -        mergecandlist[nb_merge_cand++] = TAB_MVF_PU(B0);
> +    if (is_available_b0 && check_MER) {
> +        mergecandlist[nb_merge_cand] = TAB_MVF_PU(B0);
> +        if (merge_idx == nb_merge_cand) return;
> +        nb_merge_cand++;
> +    }
>
>      // left bottom spatial merge candidate
>      check_MER = 1;
> @@ -430,8 +437,11 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>      if (is_available_a1 && is_available_a0)
>          check_MER = !COMPARE_MV_REFIDX(A0, A1);
>
> -    if (is_available_a0 && check_MER)
> -        mergecandlist[nb_merge_cand++] = TAB_MVF_PU(A0);
> +    if (is_available_a0 && check_MER) {
> +        mergecandlist[nb_merge_cand] = TAB_MVF_PU(A0);
> +        if (merge_idx == nb_merge_cand) return;
> +        nb_merge_cand++;
> +    }
>
>      // above left spatial merge candidate
>      check_MER = 1;
> @@ -447,8 +457,11 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>      if (is_available_b1 && is_available_b2)
>          check_MER_1 = !COMPARE_MV_REFIDX(B2, B1);
>
> -    if (is_available_b2 && check_MER && check_MER_1 && nb_merge_cand != 4)
> -        mergecandlist[nb_merge_cand++] = TAB_MVF_PU(B2);
> +    if (is_available_b2 && check_MER && check_MER_1 && nb_merge_cand != 4) {
> +        mergecandlist[nb_merge_cand] = TAB_MVF_PU(B2);
> +        if (merge_idx == nb_merge_cand) return;
> +        nb_merge_cand++;
> +    }
>
>      // temporal motion vector candidate
>      if (s->sh.slice_temporal_mvp_enabled_flag &&
> @@ -470,6 +483,7 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>                  mergecandlist[nb_merge_cand].mv[1]      = mv_l1_col;
>                  mergecandlist[nb_merge_cand].ref_idx[1] = 0;
>              }
> +            if (merge_idx == nb_merge_cand) return;
>              nb_merge_cand++;
>          }
>      }
> @@ -500,6 +514,7 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>                  mergecandlist[nb_merge_cand].mv[0].y      = l0_cand.mv[0].y;
>                  mergecandlist[nb_merge_cand].mv[1].x      = l1_cand.mv[1].x;
>                  mergecandlist[nb_merge_cand].mv[1].y      = l1_cand.mv[1].y;
> +                if (merge_idx == nb_merge_cand) return;
>                  nb_merge_cand++;
>              }
>          }
> @@ -515,6 +530,7 @@ static void derive_spatial_merge_candidates(HEVCContext *s, int x0, int y0,
>          mergecandlist[nb_merge_cand].ref_idx[0]   = zero_idx < nb_refs ? zero_idx : 0;
>          mergecandlist[nb_merge_cand].ref_idx[1]   = zero_idx < nb_refs ? zero_idx : 0;
>
> +        if (merge_idx == nb_merge_cand) return;
>          nb_merge_cand++;
>          zero_idx++;
>      }
> @@ -545,7 +561,8 @@ void ff_hevc_luma_mv_merge_mode(HEVCContext *s, int x0, int y0, int nPbW,
>
>      ff_hevc_set_neighbour_available(s, x0, y0, nPbW, nPbH);
>      derive_spatial_merge_candidates(s, x0, y0, nPbW, nPbH, log2_cb_size,
> -                                    singleMCLFlag, part_idx, mergecand_list);
> +                                    singleMCLFlag, part_idx,
> +                                    merge_idx, mergecand_list);
>
>      if (mergecand_list[merge_idx].pred_flag == PF_BI &&
>          (nPbW2 + nPbH2) == 12) {
> --
> 1.9.2.msysgit.0
>



-- 
Christophe


More information about the ffmpeg-devel mailing list