Inside The Discovery Cloud: Everything You Learned About Computing Is Wrong
For decades, computer scientists have coasted on the momentum of Moore's law, confident that the steadily increasing number of transistors on a microchip would eventually solve all challenges. But as more and more experts predict the encroaching end of Moore's, much of the inherited computing wisdom of the past may soon expire as well. How to write efficient algorithms, build effective computer architectures, and strike the optimal balance between cost and performance -- the traditional thinking on all of these pillars of computer science could soon be rewritten as the field looks towards the exascale and alternative computing strategies.
In their provocative tag-team talk for the April edition of the Inside The Discovery Cloud Speaker Series, CI Senior Fellows Pete Beckman and Andrew Chien took turns chopping down these CS evergreens and offering glimpses of future computing. The researchers explain how shifts from sequential to parallel algorithms, from general-use to task-customized hardware, and from a focus on operations to optimization of data movement and scheduling will drive the next generation of computers, from supercomputers such as the recently-announced Aurora to consumer laptops.