A quick update. I'm excited to tell you that we have GPU sharing support in VizStack now. The most important application of this perhaps is multiple remote users per GPU using VirtualGL/TurboVNC.
I'll be cutting out a new release 1.1 on Monday. VizStack has seen many other changes since 1.0-2. The key ones would be
- Multiple Remote users per GPU using VirtualGL/TurboVNC
- Support for bezels on tiled displays
- Much improved documentation. The older manual is now split into a user guide, an admin guide and a dev guide. The user guide documents how users would typically use the utilities provided by VizStack. The dev guide shows how to program using the vizstack API, and explains several details. Note that the documentation is still a work-in-progress.
- Some more examples of using the python API to automate tasks (running SPECViewPerf and CUDA bandwidth benchmarks).
- Sample script that lets you run Equalizer applications
- Automatic detection of unknown GPUs and display devices in the configure script.
- The template files for GPUs, displays, etc are needed only on the master node. This ensures you don't need to copy the files to other nodes.
- Many fixes !
If you want to live life on the bleeding edge, then checkout the "shree" branch from SVN. Those who want to play it safe, monday isn't that far away!