The VirtualGL module may be used on Open OnDemand desktop sessions to provide hardware accelerated OpenGL functionality to Linux desktop applications. This allows applications to provide 3D rendering/display functionality when running on a GPU node.
If you do not need hardware accelerated 3D / OpenGL, then you may instead use MESA OpenGL, which is available on all Open OnDemand partition types.
Allocating GPU Cards
To make use of hardware Nvidia OpenGL you must run your session in a GPU partition and allocate at least one GPU card. If you do not do this then VirtualGL will fallback to MESA software OpenGL.
The VirtualGL software is available in one module:
$ module avail VirtualGL
---------------------------------------- /opt/software/easybuild/modules/all ----------------------------------------
VirtualGL/3.1.1-GCC-13.3.0 VirtualGL/3.1.3-GCC-13.3.0 (D)
$ module load VirtualGL
$ module load VirtualGL
$ vglrun glxgears
...
Automatically Loading VirtualGL
If you choose one of the Open OnDemand sessions which end in (GPU), for example VNC Desktop Session (GPU) then VirtualGL is already loaded for you and you do not need to do anything to take advantage of it.
If you choose an Open OnDemand session without the (GPU) suffix, then you will need start a terminal in that session, do a module load VirtualGL and prefix all commands with vglrun as outlined above.
The (GPU) sessions are provided as a convenience for this reason.
You must be on an Open OnDemand session on a GPU partition, with at least one GPU card allocated. In your Open OnDemand desktop session open a terminal and run nvidia-smi, this should show you at least one Nvidia card available.