vglconnect

Overview

vglconnect is part of the VirtualGL suite of tools and is used to connect to GPU nodes in incline. It allows the efficient transport of the OpenGL rendered image from the GPU node back to the application being displayed on the virtual desktop.

Restrictions on Use

There are no restrictions on use, however, you must be logged in to the nyx3 virtual desktop from which you can then access the GPU node (jabberwock).

If you require access to the iCSF GPU node please email its-ri-team@manchester.ac.uk.

Set up procedure

Connect to the nyx3 virtual desktop using your X2GO client (see our X2GO setup guides for how to install an X2GO client on your local Windows, Mac or linux desktop/laptop).

Running the application

From the nyx3 virtual desktop, connect to the GPU node jabberwock by running:

vglconnect -s jabberwock

This will ask for your iCSF password twice.

Running GPU (OpenGL) GUI applications on jabberwock

Once logged in to jabberwock, the GPU node, you should set up your GUI applications as you normally do on the iCSF – load the appropriate modulefile for that application.

Then run the application using vglrun which allows the application to use the GPUs in jabberwock. For example:

vglrun myapp.exe optional args

This will run the application myapp.exe with any optional args you normally pass to the application. The application will use GPU 0 in jabberwock.

Selecting a Specific GPU

There are three GPUs in jabberwock and the GPU used by your application can be selected using one of the following:

vglrun -d :0.0 myapp.exe
vglrun -d :0.1 myapp.exe
vglrun -d :0.2 myapp.exe

If no GPU is specified then GPU 0 (:0.0) will be used by default. Having more than one application use the same GPU will reduce performance of that GPU.

Further information

Updates

None.

Last modified on April 14, 2014 at 2:13 pm by Site Admin