Advances in I/O Virtualization helped pave the way for the growing adoption of Virtualization Technologies in practical solution/service deployment. However, the complexity exists in virtualizing Graphics Processing Units (GPUs) still hinders the progress on supporting GPU accelerated tasks in a virtualized environment, from PC/laptop, to server, and also to mobile platforms. In this presentation, Kevin will review the challenges and existing solutions of virtualizing GPUs, and then provide a deep dive into the XenGT project, as a full GPU virtualization solution. A virtual GPU, same as the physical GPU, is presented to every VM, thus native graphics driver can run inside. High performance is achieved, through passing through performance critical GPU resources to multiple VMs. XenGT is an open source project on GitHub
The audience is anyone interested in GPU virtualization. Attendees can expect a detailed review of GPU virtualization challengies and different GPU virtualization techniques including pros and cons.
This presentation will help existing and new virtualziation developers better understand the value and complexity of GPU virtualization, and will hopefully encourage them to participate development/research based on our open source project.
Software Engineer, Intel
Zhiyuan comes from Intel Open Source Technology Center, Beijing site. He is one of the key developers of XenGT project. Before that, Zhiyuan used to be a compiler developer for seven years, working on C compiler for network processors.
Thursday March 27, 2014 2:00pm - 2:50pm
Attendance numbers do not account for private attendees. Get there early!