[FFmpeg-devel] [PATCH 2/8] cbs_h265: Use helper macro for maximum values of fixed-width elements

[Jun Zhao]

On 2018/3/12 20:47, Mark Thompson wrote:
> On 12/03/18 00:41, Jun Zhao wrote:
>> On 2018/3/12 2:30, Mark Thompson wrote:
>>> Apply the same logic as the previous patch to H.265.  There are no cases
>>> which currently overflow here, but this is still more consistent.
>>> ---
>>>  libavcodec/cbs_h265_syntax_template.c | 16 ++++++++--------
>>>  1 file changed, 8 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/libavcodec/cbs_h265_syntax_template.c b/libavcodec/cbs_h265_syntax_template.c
>>> index dae7f2dd46..140c827c9d 100644
>>> --- a/libavcodec/cbs_h265_syntax_template.c
>>> +++ b/libavcodec/cbs_h265_syntax_template.c
>>> @@ -665,7 +665,7 @@ static int FUNC(sps_scc_extension)(CodedBitstreamContext *ctx, RWContext *rw,
>>>                                            : current->bit_depth_chroma_minus8 + 8;
>>>                  for (i = 0; i <= current->sps_num_palette_predictor_initializer_minus1; i++)
>>>                      u(bit_depth, sps_palette_predictor_initializers[comp][i],
>>> -                      0, (1 << bit_depth) - 1);
>>> +                      0, MAX_UINT_BITS(bit_depth));
>>>              }
>>>          }
>>>      }
>>> @@ -827,7 +827,7 @@ static int FUNC(sps)(CodedBitstreamContext *ctx, RWContext *rw,
>>>          for (i = 0; i < current->num_long_term_ref_pics_sps; i++) {
>>>              u(current->log2_max_pic_order_cnt_lsb_minus4 + 4,
>>>                lt_ref_pic_poc_lsb_sps[i],
>>> -              0, (1 << (current->log2_max_pic_order_cnt_lsb_minus4 + 4)) - 1);
>>> +              0, MAX_UINT_BITS(current->log2_max_pic_order_cnt_lsb_minus4 + 4));
>>>              flag(used_by_curr_pic_lt_sps_flag[i]);
>>>          }
>>>      }
>>> @@ -845,7 +845,7 @@ static int FUNC(sps)(CodedBitstreamContext *ctx, RWContext *rw,
>>>          flag(sps_multilayer_extension_flag);
>>>          flag(sps_3d_extension_flag);
>>>          flag(sps_scc_extension_flag);
>>> -        u(4, sps_extension_4bits, 0, (1 << 4) - 1);
>>> +        u(4, sps_extension_4bits, 0, MAX_UINT_BITS(4));
>>>      }
>>>  
>>>      if (current->sps_range_extension_flag)
>>> @@ -925,7 +925,7 @@ static int FUNC(pps_scc_extension)(CodedBitstreamContext *ctx, RWContext *rw,
>>>                                            : current->chroma_bit_depth_entry_minus8 + 8;
>>>                  for (i = 0; i < current->pps_num_palette_predictor_initializer; i++)
>>>                      u(bit_depth, pps_palette_predictor_initializers[comp][i],
>>> -                      0, (1 << bit_depth) - 1);
>>> +                      0, MAX_UINT_BITS(bit_depth));
>>>              }
>>>          }
>>>      }
>>> @@ -1038,7 +1038,7 @@ static int FUNC(pps)(CodedBitstreamContext *ctx, RWContext *rw,
>>>          flag(pps_multilayer_extension_flag);
>>>          flag(pps_3d_extension_flag);
>>>          flag(pps_scc_extension_flag);
>>> -        u(4, pps_extension_4bits, 0, (1 << 4) - 1);
>>> +        u(4, pps_extension_4bits, 0, MAX_UINT_BITS(4));
>>>      }
>>>      if (current->pps_range_extension_flag)
>>>          CHECK(FUNC(pps_range_extension)(ctx, rw, current));
>>> @@ -1274,7 +1274,7 @@ static int FUNC(slice_segment_header)(CodedBitstreamContext *ctx, RWContext *rw,
>>>              const H265RawSTRefPicSet *rps;
>>>  
>>>              u(sps->log2_max_pic_order_cnt_lsb_minus4 + 4, slice_pic_order_cnt_lsb,
>>> -              0, (1 << (sps->log2_max_pic_order_cnt_lsb_minus4 + 4)) - 1);
>>> +              0, MAX_UINT_BITS(sps->log2_max_pic_order_cnt_lsb_minus4 + 4));
>>>  
>>>              flag(short_term_ref_pic_set_sps_flag);
>>>              if (!current->short_term_ref_pic_set_sps_flag) {
>>> @@ -1321,7 +1321,7 @@ static int FUNC(slice_segment_header)(CodedBitstreamContext *ctx, RWContext *rw,
>>>                              ++num_pic_total_curr;
>>>                      } else {
>>>                          u(sps->log2_max_pic_order_cnt_lsb_minus4 + 4, poc_lsb_lt[i],
>>> -                          0, (1 << (sps->log2_max_pic_order_cnt_lsb_minus4 + 4)) - 1);
>>> +                          0, MAX_UINT_BITS(sps->log2_max_pic_order_cnt_lsb_minus4 + 4));
>>>                          flag(used_by_curr_pic_lt_flag[i]);
>>>                          if (current->used_by_curr_pic_lt_flag[i])
>>>                              ++num_pic_total_curr;
>>> @@ -1487,7 +1487,7 @@ static int FUNC(slice_segment_header)(CodedBitstreamContext *ctx, RWContext *rw,
>>>              ue(offset_len_minus1, 0, 31);
>>>              for (i = 0; i < current->num_entry_point_offsets; i++)
>>>                  u(current->offset_len_minus1 + 1, entry_point_offset_minus1[i],
>>> -                  0, (1 << (current->offset_len_minus1 + 1)) - 1);
>>> +                  0, MAX_UINT_BITS(current->offset_len_minus1 + 1));
>>>          }
>>>      }
>>>  
>> Other thing about cbs_265, now cbs_264 can support SEI, but cbs_265 not,
>> is it will support SEI in cbs_265?
> It's on my list of stuff which should be done at some point, but not a priority due to lack of any immediate use-case.  Do you have some reason for wanting it?
I got some HEVC 10 bit test clip with HDR (High-dynamic-range) 
information, when I try to dump this information, I found cbs_265 didn't
support SEI.

The other thing is, I think h265_vaapi encoder can't support SEI at the
same time. 
>
> - Mark
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel