Developing Microprocessors

Computer architectures have evolved immensely since the first comercial CPU was released 50 years ago, mainly driven by the computing consumer space. Extremely complex techniques have been researched to increase processing performance, not only with regards to instruction throughput, but also in terms of power and energy efficiency, increasingly important with the expansion and evolution witnessed in the mobile and embedded markets. These advances lead to a state of maturity without precedents. Nonetheless, advances in other fields of information technologies during the last 10 years have created the environment conditions for different types of more specialized processing architectures to appear and evolve. Namely, GPUs, driven by the requirements of the gaming industry and more recently big data and financial market analysis. Today, a new era of computer architecture is in the verge of reaching us, where more than significantly improving cost, performance, and energy, we will also witness a strong focus on two fronts: security and reliability driven by the IoT and automotive markets, and domain specific hardware accelerators specialized for neural networks driven by data mining.
This presentation will cover some of these aspects in perspective with what is currently taught at the university and will give insights into some of the main concerns of the main players in this fields.

Orador

Frederico Pratas (IPS & Synopsis)

Computer architect with PhD in Electrical and Computer Engineering and experience on research and development of hardware architectures, reconfigurable hardware design, and acceleration of scientific applications for high performance computing on different platforms.