[FFmpeg-devel] [PATCH] vp9: fix bug in updating of coef probabilities with parallelmode=1.

[Clément Bœsch]

On Mon, Dec 23, 2013 at 09:41:35PM -0500, Ronald S. Bultje wrote:
> Fixes trac 3228.
> ---
>  libavcodec/vp9.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/libavcodec/vp9.c b/libavcodec/vp9.c
> index a52924c..88c5bc6 100644
> --- a/libavcodec/vp9.c
> +++ b/libavcodec/vp9.c
> @@ -3547,13 +3547,16 @@ static int vp9_decode_frame(AVCodecContext *ctx, void *frame,
>      if (s->refreshctx && s->parallelmode) {
>          int j, k, l, m;
>  
> -        for (i = 0; i < 4; i++)
> +        for (i = 0; i < 4; i++) {
>              for (j = 0; j < 2; j++)
>                  for (k = 0; k < 2; k++)
>                      for (l = 0; l < 6; l++)
>                          for (m = 0; m < 6; m++)
>                              memcpy(s->prob_ctx[s->framectxid].coef[i][j][k][l][m],
>                                     s->prob.coef[i][j][k][l][m], 3);
> +            if (s->txfmmode == i)
> +                break;
> +        }

I tested with some other samples, output looks ok, no valgrind complain
(if we strip out the current weird stack bug).

little nit: we could probably just change the loop into
-        for (i = 0; i < 4; i++)
+        for (i = 0; i < 4 && i <= s->txfmmode; i++)

...but I'm assuming you're just getting consistent with the coeff updates
code.

Patch is fine anyway.

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20131224/06ea7831/attachment.asc>