Why Start Now?
Languages: Broadly accepted, object oriented, explicit concurrency, restricted memory models
Software Compiler Technology: Whole program analysis, profile-based compilation, complex microarchitecture capability
Hardware Compiler Technology: High-level compilers close to direct software hand-off
Some of the ideas mentioned in previous slides are not new. Some have been tried before and have met with mixed success. Why take on this research now? What makes it more likely to succeed?
Much of what we have described has to do with methodology--approaches to design. These are informed by the priorities of the time as well as the tools and technologies available as a base.
We believe the existence and active development of the technologies listed above makes such an endeavor much more likely to succeed than in the past. In particular, if we can solve the portable IP problem efficiently, for the first time our ability to integrate more than a billion transistors on a single piece of silicon will provide the medium needed to implement many of these ideas cost-effectively. It represents an entirely new canvas for research and development.