[mythtv] render2019 comments
mark.kendall at gmail.com
Thu Oct 3 08:53:03 UTC 2019
On Wed, 2 Oct 2019 at 12:53, John Pilkington <johnpilk222 at gmail.com> wrote:
> I now have b449f5e running under el7 on this Core2duo Intel Q33 2.8 GHz
> box with no extra video card and a single monitor at 75 Hz. It seems
> fine for DVB-T SD recordings.
> In my earlier tests I had only tried changing the 'Current Video
> Playback Profile' and saw no change in the OSD playback_data. Now both
> boxes have that set to Normal, and define Decoder, Max CPUs etc in the
> screens that follow it; NVDEC (where present) and deinterlacing type
> are now shown.
Sorry - I meant to get back to you earlier. I was going to suggest
that you probably needed to edit the profile itself to ensure you were
> On the el7 box I get this sort of report on startup and with
> mythpreviewgen and mythcommflag --rebuild. It works, and says what it
> would like to do. :-)
> running: ionice -c3 mythpreviewgen --chanid 10001 --starttime
> 20191001223800 -q
> Failed to open VDPAU backend libvdpau_va_gl.so: cannot open shared
> object file: No such file or directory
> libva info: VA-API version 0.40.0
> libva info: va_getDriverName() returns 0
> libva info: Trying to open /usr/lib64/dri/i915_drv_video.so
> libva info: va_openDriver() returns -1
> [AVHWDeviceContext @ 0x250ed80] Failed to initialise VAAPI connection:
> -1 (unknown libva error).
> Preview created.
For what it's worth - most of those messages are driver library
messages (not mythtv logging). They are harmless.
The reason I added functionality checks on startup (in
AVFormatDecoder) was to ensure that we didn't clutter the video
settings pages with decoders that were not useable. On most Intel
systems, for example, you will normally get VAAPI, VDPAU, NVDec and
V4L2 codecs support compiled in (VDPAU headers are normally present,
NVDEC is largely a runtime check in FFmpeg and V4L2 will always be
available but usually not functioning) - but only VAAPI is viable.
The one problem at the moment is that I get intermittent NVDEC
functionality check failures - this also happens at seemingly random
times when starting playback.
Furthermore, I've not made any attempt to remove those checks when not
using mythfrontend or mythavtest. This is because a roughly 2 line
patch would enable hardware accelerated decoding for commercial
flagging - which may be of use/interest to some. It would also require
new settinga but does have complications around how to manage the
number of concurrent GPU hardware contexts (most systems will only
manage one), prioritising playback over other operations, commflagging
operations would in most cases need to be running with a display (so
headless operation would not work - with the current exception of MMAL
for the Pi) and given that we farm out a lot of this to separate
processes, how to coordinate between them.
- there will also, I'm sure, be someone who has a dual GPU setup who
wants to use their discreet card for video playback and integrated GPU
for commercial flagging...
Thanks and regards
More information about the mythtv-dev