On Tue, Nov 08, 2011 at 03:47:32PM +0600, Mashiat Sarker Shakkhar wrote:
> This is at odd with the spec but complies with the reference decoder.
> ---
> libavcodec/vc1dec.c | 5 +----
> 1 files changed, 1 insertions(+), 4 deletions(-)
>
> diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
> index 3253292..a7ea3ab 100644
> --- a/libavcodec/vc1dec.c
> +++ b/libavcodec/vc1dec.c
> @@ -1660,10 +1660,7 @@ static inline void vc1_pred_mv(VC1Context *v, int n,
> int dmv_x, int dmv_y,
>
> if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {
> /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also
> 10.3.5.4.3.5) */
> - if (v->field_mode && !s->quarter_sample)
> - hybridmv_thresh = 16;
> - else
> - hybridmv_thresh = 32;
> + hybridmv_thresh = 32;
> if (a_valid && c_valid) {
> if (is_intra[xy - wrap])
> sum = FFABS(px) + FFABS(py);
> --
It makes sense if you always operate on qpel-precision MVs (even when you got
hpel you multiply it by two).
So if you do so patch is OK.
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
