After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 761452 - glvideomixer: does not work with full HD resolution on raspberrypi
glvideomixer: does not work with full HD resolution on raspberrypi
Status: RESOLVED INCOMPLETE
Product: GStreamer
Classification: Platform
Component: gst-plugins-bad
git master
Other Linux
: Normal normal
: git master
Assigned To: GStreamer Maintainers
GStreamer Maintainers
: 759359 (view as bug list)
Depends on: 760916
Blocks:
 
 
Reported: 2016-02-02 11:35 UTC by Brijesh Singh
Modified: 2017-05-20 13:25 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description Brijesh Singh 2016-02-02 11:35:58 UTC
I am trying glvideomixer on rpi.

The following pipeline (1080P scaling) does not work, it does not report any error but screen is blank

gst-launch-1.0 glvideomixer name=mixer ! fpsdisplaysink sync=false videotestsrc ! video/x-raw,width=1920,height=1080 ! mixer.


But if i change width and height to a smaller resolution then it works flawlessly.

-Brijesh
Comment 1 Nicolas Dufresne (ndufresne) 2016-02-02 17:29:45 UTC
If this *can work*, I believe this is dependent on bug # 760916 . I believe the PI GL stack can barely handle that, but who knows, removing the extra copy may make it work.
Comment 2 Matthew Waters (ystreet00) 2016-02-02 22:43:17 UTC
I don't think the RPi in it's current state can handle 1080p textures at all.  Most of the RPi player's mention avoiding using GL for decoding into 1080p textures.
Comment 3 Matthew Waters (ystreet00) 2016-02-02 22:51:05 UTC
Be sure to have gpu_mem set high enough (at least 192, probably more like 256).
Comment 4 Brijesh Singh 2016-02-03 21:48:57 UTC
fyi,  gpu setting didn't help for me.
Comment 5 Matthew Waters (ystreet00) 2016-05-04 08:42:33 UTC
Testing this, I can get ~10 fps from a 1080p videotestsrc and comparing with and without glvideomixer is roughly ~1 fps difference.

bug 760916 shouldn't have made a difference to whether this is possible or not.

Most likely, you're limited by videotestsrc generating buffers.  gltestsrc (available for the rpi in latest master) will perform better than videotestsrc and gives me ~30fps without a glvideomixer.  With gltestsrc ! glvideomixer ! fpsdisplaysink gives my ~13fps for a 1080p stream.
Comment 6 Sam Hurst 2016-09-13 09:00:08 UTC
Apologies for bumping an old thread, but I'm just enquiring about what Matthew said about the current state of the Raspberry Pi:

I have a build based on the 1.7.90 branch of GStreamer that can manage to decode 720p50 and 1080p25 video without dropping frames, using gl for display; however it seems that somewhere between then and now the performance has taken quite a hit in the gl state. Is this possibly just needing some optimisation of GstGLMemory(EGL) for the Raspberry Pi use case? Or given what you were saying in comment 2, that it's better to find a different way of displaying content on the screen? A different image sink perhaps?
Comment 7 Matthew Waters (ystreet00) 2016-09-15 02:48:41 UTC
(In reply to Sam Hurst from comment #6)
> Apologies for bumping an old thread, but I'm just enquiring about what
> Matthew said about the current state of the Raspberry Pi:

Please avoid polluting Bugzilla with support questions.

That being said, GstGLMemoryEGL reduced the CPU consumption of hardware accelerated video decoding with glimagesink on the RPi.  We need more information on your exact pipeline to know whether it's an issue or not.
Comment 8 Matthew Waters (ystreet00) 2016-09-26 04:43:17 UTC
*** Bug 759359 has been marked as a duplicate of this bug. ***
Comment 9 Matthew Waters (ystreet00) 2016-11-15 07:45:40 UTC
Brijesh, what versions of gst-plugins-bad do you have?
Comment 10 Matthew Waters (ystreet00) 2017-05-20 13:25:43 UTC
Closing this bug report as no further information has been provided. Please feel free to reopen this bug report if you can provide the information that was asked for in a previous comment.
Thanks!