This talk is part of the WID-DOW Collaborative Presentation Series on Doing Optimization at Wisconsin. A note on the location: Anyone without WID access can use the special events elevator on WID's 1st floor (near Aldo's Café) to access room 3280 (third floor teaching lab).
In this talk, I will describe some recent work in my research group on building a memory access accelerator. Our work is motivated by observing that many accelerators being proposed integrate or interact with a high-performance microprocessor. In such cases, the microprocessor becomes a significant power hog, while a low performance processor limits the accelerator's effectiveness.
As cloud computing becomes increasingly popular, organizations face greater security threats. Public clouds have become a central point of attack and successful compromises can cause potentially billions of dollars of damage. Physical attacks on data center machines are very concerning because an attacker can gain full control of the machines and circumvent software protections.
Energy is increasingly a first-order concern in computer systems. Exploiting energy-accuracy trade-offs is an attractive choice in applications that can tolerate inaccuracies. A key challenge, though, is how to isolate parts of the program that must be precise from those that can be approximated so that a program functions correctly even as quality of service degrades. Addressing that challenge leads to opportunities for approximate computing across the entire system stack.