Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
This lecture by the instructor Mark Silberstein discusses the concept of an accelerator-centric OS architecture for omni-programmable systems, rethinking the role of CPUs in modern computers. The presentation covers topics such as the challenges of traditional CPU-centric architectures, the emergence of X-Processing Units (XPUs) for near-data processing, and the implementation of GPU-centric services. It explores the benefits of CPU-less designs, GPU-centric virtual memory management, and the potential of GPUrdma for direct GPU access to RDMA. The lecture also delves into the importance of application offload to SmartNICs in data centers and the ongoing work on the Omnix project.