This is a log of a calculator which the author maintains. The machine conditions are as follows:
Since Tesla C2075 is very new GPU, it could not be detected by Fedora9 which was installed in the computer. I changed to the CentOS-6.2 for the aspect of security and convenience of the use. Now both GPUs are detected by the OS.
Virtual-GL is a system which send Open-GL application images from a host computer to cliants' machines.
Just using X-window system, we don't have to install.
However, in the case of using graphical-rich process, namely, application with Open-GL from other computer, the X can't send images.
In this case, we have to install another graphic transportation process, and hence we use Virtual-GL.
We put setting log here.
Virtual-GL can be downloaded from the Virtual-GL web page.
There are documentation and you'll get a manual.
We need softwares both host (logined machine) and cliants (login machine).
There are no setting in the cliants, just using /opt/VirtualGL/bin/vglconnect to login to the host computer (OSX, Linux).
There are some settings in the hots.
We have to do following things to set machine configuration (see user's guide above):
Normaly, it will be enough to do.
As libglx.so is not newest in our case, we sysmbolic linked /usr/lib64/xorg/modules/extensions/libglx.so to libglx.so.295.53.
Additionally, we only installed CUDA device driver on the host, and hence we had to install normal device driver to run Open-GL.
In the case of processing a software with VirtualGL, type as follows:
$/opt/VirtualGL/bin/vglrun "application name"