Questions about using terminal server product for VirtualGL applications

[willm]

Hello

Sorry , I’m not familiar with the graphics application environment and i need some explanations.

We evaluate the NoMachine Linux terminal server production (workstation-evaluation_5.1.62)

on a PowerEdge T630 ( 2 * 20 cores ) with GeForce GTX 1070 GPU cards .

OS : Centos7

Nvidia proprietary driver: 375.20

Scientist researchers want to use GPUs for computing and visualisation 3D with virtual GL for OpenGL applications ; We only do remote display, we cant put physical screens on the cards.

I try two types of configuration but i have some questions

1) a server X run on a GPU card with the option « AllowEmptyInitialConfiguration »

I connect remotely to the physical desktop ; but the resolution is 640×480 by default and i cant modify it ; glxinfo is correct ; but the screen is not usable.

how we can modify the size and resolution of the screen ? Is it possible ?

2) with a virtual desktop connexion , i have a KDE desktop ; i have configured virtualgl according to https://www.nomachine.com/AR11K00737&an=virtualgl

but in this environment how can we run openGL applications without the command vglrun ?

Thank you for your answers

[graywolf]

I connect remotely to the physical desktop ; but the resolution is 640×480 by default and i cant modify it ; glxinfo is correct ; but the screen is not usable.

how we can modify the size and resolution of the screen ? Is it possible ?

Available resolutions when attaching to the physical display don’t depend on NoMachine, but on what is allowed by the graphics card and/or the X server configuration.

By the way, it doesn’t matter. VirtualGL needs that to offload the rendering, but you should never access like it was a usual display because its X authority is relaxed or disabled. See here: Why VirtualGL requires access to the display :0.

I have configured virtualgl according to AR11K00737 but in this environment how can we run openGL applications without the command vglrun?

The same article you mentioned suggests to change key EnableVirtualGLSupport to enable VirtualGL for all application. If I understand well, you aren’t goint to enable VirtualGL for all, but you need it for some specific program and you’d like to avoid typing “vglrun”. In that case I’d suggest to create aliases in the user’s shell, so that command “my-3d-program”  actually launches “/usr/NX/scripts/vgl/vglrun my-3d-program”.

[Author]

Posts