[vlc-devel] [PATCH 3/4] video_output: try to create a decoder device even if creating the window failed
Alexandre Janniaux
ajanni at videolabs.io
Tue Jan 14 00:35:31 CET 2020
Hi,
To be exact, the question would be: do d3d/dxva require an
enabled window or can this be made so?
Nice point, I would prefer that too.
Regards,
--
Alexandre Janniaux
Videolabs
On Mon, Jan 13, 2020 at 04:17:50PM +0100, Thomas Guillem wrote:
> Sorry I meant:
>
> Why not only requiring a disabled window from a decoder device ?
> The big question for me is: do d3d require the window to be enabled ? (because no other HW APIs need it)
>
> I know the problem, I already read it twice ;)
>
> On Mon, Jan 13, 2020, at 16:14, Steve Lhomme wrote:
> > On 2020-01-13 16:03, Thomas Guillem wrote:
> > >
> > >
> > > On Mon, Jan 13, 2020, at 15:59, Steve Lhomme wrote:
> > >>
> > >> On 2020-01-13 15:54, Thomas Guillem wrote:
> > >>>
> > >>> On Mon, Jan 13, 2020, at 15:47, Rémi Denis-Courmont wrote:
> > >>>> Hi,
> > >>>>
> > >>>> Does VA require an actual window? I'd think it only requires a
> > >>>> display, so a vout_window_t in disabled state should suffice.
> > >>>
> > >>> Indeed, we don't have any API that require an enabled window.
> > >>> VA only need:
> > >>> Display *x11dpy = XOpenDisplay(window->display.x11);
> > >>> VADisplay vadpy = *vadpyp = vaGetDisplay(x11dpy);
> > >>>
> > >>> So, I think we could simplify everything by not requiring an enabled
> > >>> window. I that case, you can load the window module, and create the
> > >>> decoder device without a valid size.
> > >>
> > >> Luckily that may be what this patch does. It's using
> > >> sys->display_cfg.window regarless of its state. It may be NULL or set
> > >> but not enabled.
> > >
> > > I disagree with this patch. It add one more complexity, different bugs because the decoder device was created with or without a window because the decoder probed it during Open() or when the size was known...
> > >
> > > What not requiring the window to be enabled ?
> >
> > Because...
> >
> > >>>>
> > >>>> Le 13 janvier 2020 15:56:43 GMT+02:00, Thomas Guillem <thomas at gllm.fr>
> > >>>> a écrit :
> > >>>>
> > >>>>
> > >>>> On Mon, Jan 13, 2020, at 13:35, Steve Lhomme wrote:
> > >>>>
> > >>>> "In some cases (nvdec, mmal, mediacodec ?) the actual size of
> > >>>> the video
> > >>>> to decode is known only after data have been processed, long
> > >>>> after the
> > >>>> device is setup to use with this API. There are hacks to try
> > >>>> to guess
> > >>>> the size early but in the end it's just a hack."
> >
> > ...of this...
> >
> >
> > >>>> What to we do when we have a hw decoder that doesn't know its
> > >>>> actual size but a window is required ? (vaapi).
> > >>>>
> > >>>> >
> > >>>>
> > >>>> This removes the need for these hacks for those using a
> > >>>> decoder device
> > >>>> that doesn't need a window.
> > >>>>
> > >>>> On 2020-01-13 13:27, Alexandre Janniaux wrote:
> > >>>>
> > >>>> Hi,
> > >>>>
> > >>>> This looks dangerous if it is expected to have an enabled
> > >>>> window at decoder device creation. vout_window handle
> > >>>> values will basically be undefined without any way for the
> > >>>> decoder device to know it. Why do you need this?
> > >>>>
> > >>>> Regards,
> > >>>> --
> > >>>> Alexandre Janniaux
> > >>>> Videolabs
> > >>>>
> > >>>> On Mon, Jan 13, 2020 at 12:07:33PM +0100, Steve Lhomme wrote:
> > >>>>
> > >>>> ------------------------------------------------------------------------
> > >>>> src/video_output/video_output.c | 20 +++++++++-----------
> > >>>> 1 file changed, 9 insertions(+), 11 deletions(-)
> > >>>>
> > >>>> diff --git a/src/video_output/video_output.c
> > >>>> b/src/video_output/video_output.c
> > >>>> index 5d583dbe0ca..9540707ab2b 100644
> > >>>> --- a/src/video_output/video_output.c
> > >>>> +++ b/src/video_output/video_output.c
> > >>>> @@ -2099,20 +2099,18 @@ vlc_decoder_device
> > >>>> *vout_GetDevice(const vout_device_configuration_t *cfg)
> > >>>> assert(cfg->fmt != NULL);
> > >>>> vout_thread_sys_t *sys = cfg->vout->p;
> > >>>>
> > >>>> - if (!VoutCheckFormat(cfg->fmt))
> > >>>> - return NULL;
> > >>>> -
> > >>>> - video_format_t original;
> > >>>> - VoutFixFormat(&original, cfg->fmt);
> > >>>> -
> > >>>> vlc_mutex_lock(&sys->window_lock);
> > >>>> - int res = EnableWindowLocked(cfg->vout, &original);
> > >>>> - if (res == 0 && sys->dec_device == NULL)
> > >>>> + if (VoutCheckFormat(cfg->fmt))
> > >>>> + {
> > >>>> + video_format_t original;
> > >>>> + VoutFixFormat(&original, cfg->fmt);
> > >>>> +
> > >>>> + EnableWindowLocked(cfg->vout, &original);
> > >>>> + video_format_Clean(&original);
> > >>>> + }
> > >>>> + if (sys->dec_device == NULL)
> > >>>> sys->dec_device =
> > >>>> vlc_decoder_device_Create(&cfg->vout->obj,
> > >>>> sys->display_cfg.window);
> > >>>> dec_device = sys->dec_device ?
> > >>>> vlc_decoder_device_Hold( sys->dec_device ) : NULL;
> > >>>> vlc_mutex_unlock(&sys->window_lock);
> > >>>> - video_format_Clean(&original);
> > >>>> - if (res != 0)
> > >>>> - return NULL;
> > >>>> return dec_device;
> > >>>> }
> > >>>> --
> > >>>> 2.17.1
> > >>>> ------------------------------------------------------------------------
> > >>>> vlc-devel mailing list
> > >>>> To unsubscribe or modify your subscription options:
> > >>>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>>
> > >>>> ------------------------------------------------------------------------
> > >>>> vlc-devel mailing list
> > >>>> To unsubscribe or modify your subscription options:
> > >>>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>>
> > >>>> ------------------------------------------------------------------------
> > >>>> vlc-devel mailing list
> > >>>> To unsubscribe or modify your subscription options:
> > >>>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>>
> > >>>> ------------------------------------------------------------------------
> > >>>> vlc-devel mailing list
> > >>>> To unsubscribe or modify your subscription options:
> > >>>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>>
> > >>>>
> > >>>> --
> > >>>> Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez
> > >>>> excuser ma brièveté.
> > >>>> _______________________________________________
> > >>>> vlc-devel mailing list
> > >>>> To unsubscribe or modify your subscription options:
> > >>>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>
> > >>>
> > >>> _______________________________________________
> > >>> vlc-devel mailing list
> > >>> To unsubscribe or modify your subscription options:
> > >>> https://mailman.videolan.org/listinfo/vlc-devel
> > >>>
> > >> _______________________________________________
> > >> vlc-devel mailing list
> > >> To unsubscribe or modify your subscription options:
> > >> https://mailman.videolan.org/listinfo/vlc-devel
> > > _______________________________________________
> > > vlc-devel mailing list
> > > To unsubscribe or modify your subscription options:
> > > https://mailman.videolan.org/listinfo/vlc-devel
> > >
> > _______________________________________________
> > vlc-devel mailing list
> > To unsubscribe or modify your subscription options:
> > https://mailman.videolan.org/listinfo/vlc-devel
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel
More information about the vlc-devel
mailing list