On Tue, Aug 16, 2011 at 05:08:52PM -0700, Ronald S. Bultje wrote:
> Hi,
> 
> On Tue, Aug 16, 2011 at 5:24 AM, Alberto Delmás <adel...@gmail.com> wrote:
> > +            av_log(avctx, AV_LOG_DEBUG, " %d.%.3d",
> > +                   sd->coefs[sprite][i] / (1<<16),
> > +                   (sd->coefs[sprite][i] & 0xFFFF) * 1000 / (1<<16));
> 
> I don't believe this works for negative numbers. Same for
> effects_param1/2. Probably should be abs(sd->coefs[sprite][i]) in the
> last line.
> 
> > @@ -208,7 +208,12 @@ void ff_copy_picture(Picture *dst, Picture *src){
> >   */
> >  static void free_frame_buffer(MpegEncContext *s, Picture *pic)
> >  {
> > -    ff_thread_release_buffer(s->avctx, (AVFrame*)pic);
> > +    /* Windows Media Image codecs allocate internal buffers with different
> > +       dimensions; ignore user defined callbacks for these */
> > +    if (s->codec_id != CODEC_ID_WMV3IMAGE && s->codec_id != 
> > CODEC_ID_VC1IMAGE)
> > +        ff_thread_release_buffer(s->avctx, (AVFrame*)pic);
> > +    else
> > +        avcodec_default_release_buffer(s->avctx, (AVFrame*)pic);
> 
> So coming back to this old one - is there some way we can signal at
> the buffer request phase which buffer is to be "displayed" and at
> display height, and which isn't? That'd allow us to keep using
> ff_thread_release_buffer() at least for these, which can be a
> significant performance gain.

Probably by augmenting API.

> I'm gonna assume Kostya also did some review here. Patch basically
> looks good to me.

Yes, I've looked at it before it was sent to ML.
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to