No hardware decode on Kubuntu 22.04 + GTX 1060 6GB

[Ilgaz]

Hello

Here is the output of the commands

vainfo --display drm
Trying display: drm
libva info: VA-API version 1.18.0
libva info: Trying to open /usr/lib64/dri/nouveau_drv_video.so
libva info: Found init function __vaDriverInit_1_18
libva info: va_openDriver() returns 0
vainfo: VA-API version: 1.18 (libva 2.18.1)
vainfo: Driver version: Mesa Gallium driver 23.0.2 for NVAC
vainfo: Supported profile and entrypoints
VAProfileMPEG2Simple            : VAEntrypointVLD
VAProfileMPEG2Main              : VAEntrypointVLD
VAProfileVC1Simple              : VAEntrypointVLD
VAProfileVC1Main                : VAEntrypointVLD
VAProfileVC1Advanced            : VAEntrypointVLD
VAProfileH264ConstrainedBaseline: VAEntrypointVLD
VAProfileH264Main               : VAEntrypointVLD
VAProfileH264High               : VAEntrypointVLD
VAProfileNone                   : VAEntrypointVideoProc

for DRVNAME in $(find /usr/lib -name “*drv_video.so” -exec basename {} \; | sed ‘s/_drv_video.so//
’); do for DEVICE in /dev/dri/* ; do LIBVA_DRIVER_NAME=$DRVNAME vainfo –display drm –device $DEVICE 1>/dev/null
2>/dev/null ; [ $? -eq 0 ] && echo “$DRVNAME $DEVICE” ; done ; done 2>/dev/null | grep -v dummy

Unfortunately, no output has been shown as a result of that command.

[Britgirl]

for DRVNAME in $(find /usr/lib -name "*drv_video.so" -exec basename {} \; | sed 's/_drv_video.so//'); do for DEVICE in /dev/dri/* ; do LIBVA_DRIVER_NAME=$DRVNAME vainfo --display drm --device $DEVICE 1>/dev/null 2>/dev/null ; [ $? -eq 0 ] && echo "$DRVNAME $DEVICE" ; done ; done 2>/dev/null | grep -v dummy

 

did you run as I have pasted above? We noticed that in your output shows -display drm -device... and not --display drm --device... (my fault maybe because I didn’t put my text in tags, since corrected)

If that does not produce any output, you can try with “find /usr/lib64......” instead of “find /usr/lib.....”

Provided there is some output you should then

1. edit /usr/NX/bin/nxplayer
2. add line

LIBVA_DRIVER_NAME=name

where ‘name’ is the driver that shows in the output. This instructs va-api decoder to use that specific driver. You should place it before the line:

exec "$NX_SYSTEM/bin/nxplayer.bin" "$@"`

3. Then start your NoMachine session. If it does not work, the ~/.nx/R-*/session directory will contain more info.

[Ilgaz]

Hello,  thanks again for caring.

When I issue the command with the lib64 path under openSUSE Tumbleweed:

nouveau /dev/dri/card0
nouveau /dev/dri/renderD128
r600 /dev/dri/card0
r600 /dev/dri/renderD128
radeonsi /dev/dri/card0
radeonsi /dev/dri/renderD128
virtio_gpu /dev/dri/card0
virtio_gpu /dev/dri/renderD128

So I added the line (I run nouveau/nvidia)

LIBVA_DRIVER_NAME=nouveau

When I launch the nxplayer.bin and attempt to connect a host it core dumps. I attached the dump file. There are some errors with libva reported in session (log) I am attaching it too.

Attachments:

1. nxplayercoredump.txt
2. session.zip

[Britgirl]

Logs show that the error seems to be in the GPU pipeline and with nouveau.

Please try disabling GPU rendering. You can do this by setting the following option key "Enable GPU rendering" value="true"   to falsein player.cfg.

[Ilgaz]

That setting fixes the crash (coredump) however I get hit by a familiar-looking glitch making the session impossible to interact with. It looks almost exactly the same as the video glitches I saw with Firefox/Wayland. I believe it got fixed with better multithreading on Nouveau, I am actively using video acceleration with Firefox.  https://bugzilla.mozilla.org/show_bug.cgi?id=1658772 . I will inform Nouveau developers about it. Anyway thank you for caring.

Attachments:

1. 

   casaba-Ubuntu-22.04.2-LTS-2023-04-30-16.58.11.png
   

[Britgirl]

Yes, it’s a case for the nouveau developers.

In the meantime, when connecting with NoMachine you’ll be able to use software decoding instead.

[SparkleCoder]

My my my…I wish I had seen this thread sooner, as I have some first-hand experience with almost this exact situation and learned a few things during the torturously long debugging of it. I don’t know if this will be of any assistance, but here’s hoping…

First, I am pleased to inform you that reports of the demise of the vdpau-va-driver package have been grossly exaggerated, haha. It’s actually been reincarnated under a new developer not associated with the Mesa project and is now named nvidia-vaapi-driver. I wish that was the key to solving this whole affair, but I’m afraid there’s more. The code changes that were apparently necessary to solve my issue were only just released in the latest version, which sadly hasn’t been a priority for anyone on the Debian/Ubuntu/KDE packaging teams to push out to the repos.

That leaves it to you to compile the driver yourself, if you’re comfortable with doing that, or if not then I’m happy to share the packages I built of that latest version for everything from Jammy all the way up to Mantic, which I just uploaded to my PPA in case that was your preference. Naturally, you’ll also want to add the official NVIDIA apt repository to your /etc/apt/sources.list.d directory by placing a file in it ending in .list with deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ / in it, then adding their signing key to /etc/apt/trusted.gpg.d with the command sudo apt-key --keyring /etc/apt/trusted.gpg.d/nvidia.gpg adv --keyserver keyserver.ubuntu.com --recv-keys EB693B3035CD5710E231E123A4B469963BF863CC. Polish it off with a sudo apt update and a sudo apt full-upgrade and if my experience is any guide, you’ll be NoMachine-ing all over God’s creation after a quick reboot. I’m sorry this advice comes so late and hopefully your problem is already long-solved, but perhaps in that case this’ll be of use to the next person to meet this fate. Best wishes to you.

P.S. The above directions were created with a baked-in assumption that you already have the apt-transport-https and apt-utils packages already installed on your system, as tends to be the case. If that happens not to be so, please install them with sudo apt install before following them. Toodles.

[Britgirl]

Thanks SparkleCoder for your contribution!

We’ll make sure we reference your post if others come with the same issue 🙂

[Author]

Posts