Quoting Janne Grunau (2016-05-29 11:13:34)
> On 2016-05-16 10:25:54 +0200, Anton Khirnov wrote:
> > Currently it's done in the code that initialises the ref list for
> > MBAFF, which is not a logical place for it. Move it to the function that
> > parses the pred table from the bitstream, which is analogous to what is
> > done for the implicit weight table as well.
> > ---
> > libavcodec/h264_parse.c | 10 +++++++++-
> > libavcodec/h264_refs.c | 7 -------
> > 2 files changed, 9 insertions(+), 8 deletions(-)
> >
> > diff --git a/libavcodec/h264_parse.c b/libavcodec/h264_parse.c
> > index 7211c9d..052e8e4 100644
> > --- a/libavcodec/h264_parse.c
> > +++ b/libavcodec/h264_parse.c
> > @@ -26,7 +26,7 @@ int ff_h264_pred_weight_table(GetBitContext *gb, const
> > SPS *sps,
> > const int *ref_count, int slice_type_nos,
> > H264PredWeightTable *pwt)
> > {
> > - int list, i;
> > + int list, i, j;
> > int luma_def, chroma_def;
> >
> > pwt->use_weight = 0;
> > @@ -78,6 +78,14 @@ int ff_h264_pred_weight_table(GetBitContext *gb, const
> > SPS *sps,
> > }
> > }
> > }
> > +
> > + // for MBAFF
> > + pwt->luma_weight[16 + 2 * i][list][0] = pwt->luma_weight[16 +
> > 2 * i + 1][list][0] = pwt->luma_weight[i][list][0];
> > + pwt->luma_weight[16 + 2 * i][list][1] = pwt->luma_weight[16 +
> > 2 * i + 1][list][1] = pwt->luma_weight[i][list][1];
> > + for (j = 0; j < 2; j++) {
> > + pwt->chroma_weight[16 + 2 * i][list][j][0] =
> > pwt->chroma_weight[16 + 2 * i + 1][list][j][0] =
> > pwt->chroma_weight[i][list][j][0];
> > + pwt->chroma_weight[16 + 2 * i][list][j][1] =
> > pwt->chroma_weight[16 + 2 * i + 1][list][j][1] =
> > pwt->chroma_weight[i][list][j][1];
> > + }
>
> maybe pass "FRAME_MBAFF(h)" as flag to ff_h264_pred_weight_table() and
> make this block conditional. Probably won't be noticeable though
The context-global variables might not always be set at this point, so
I'd prefer not to depend on them. Especially since, as you say, it's not
likely to be noticeable.
--
Anton Khirnov
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
