[Bug 757941] vaapiencode: Allow to set profiles via capsfilters

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=757941

--- Comment #24 from Víctor Manuel Jáquez Leal <vjaquez at igalia.com> ---
(In reply to Hyunjun Ko from comment #22)
> (In reply to Víctor Manuel Jáquez Leal from comment #21)
> > Comment on attachment 353053 [details] [review] [review]
> > libs: encoder: h264: set profile via capsfilter
> > 
> > Hyunjun,
> > 
> > Please rebase this patch with the patches I posted.
> > 
> > Keep in mind that
> > 
> > if no max_profile_idc is set, then the encoder will have to choose one
> > accordingly with the properties (as is right now)
> > 
> > if max_profile_idc is set and it is the same as the profile_idc, do nothing
> > 
> > if max_profile_idc is "bigger" than the profile_idc, modify the profile_idc
> > and adjust the properties.
> > 
> Okay by here.
> 
> > if max_profile_idc is "lower" that the profile_idc... perhaps we should use
> > the profile_idc
> 
> There are some confusing cases here.
> 
> 1. vaapih264enc dct8x8=true ! video/x-h264,profile=Main
> 2. vaapih264enc ! video/x-h264,profile=Main
> 
> These cases are same internally since dct8x8 is TRUE by default.
> I thought in the case 1 we should error out.
> And for the case 2, we should use Main profile and get it working.

I agree.

> 
> If you agree with my thought, I think we'd better return default of
> dct8x8,num_b_frame,use_cabac to the previous, so that we could handle cases
> easily.
> 
> What do you think?

If I understand correctly, what are you saying is to revert 4aec5bdd and then
handle these cases???

One more thing: when need to test the stereo profile, in order to see if we
don't have a regression there. I haven't tested it.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.