Architecture designs are becoming more and more complicated. From uni-core processor to multi-/many-core SoCs, we are facing a rapidly growing design space. Computer architects tend to build models at all levels to efficiently explore and prune the design space. However, designing a system in an era of rapidly evolving application behaviors and significant technology shifts involves all kinds of uncertainties. While some do back of the envelope estimations here and there during design, uncertainty has been overlooked by most system designers.
In this talk, I will first examine high-level architecture models from classic Amdahl's Law to state-of-the-art CMP models. I will show how uncertainty is a long-missed concern in building and evaluating such models. Then I will explore techniques used in other fields inside and outside computer science to deal with such uncertainties. At last, I will end the talk with a brief discussion about our on-going research effort and future works.