Principal Investigator Srinivas Devadas
Project Website http://bigdata.csail.mit.edu/?q=node/42
Big Data needs Big Processors, and Big Processors need Big Caches. Increasingly, however, power and thermal considerations dictate that many small processors and many small caches supplant the paradigm of few big processors and caches.
The Execution Migration Machine (EM²) project aims to find the best way of using these resources. Instead of relying on complex cache coherence protocols with their large directories to bring data to the core that needs it for computation, we migrate the computation itself to the core that holds the necessary data. Because execution contexts can be smaller than the data they operate on, bringing them to the data reduces on-chip network traffic and energy. And, since cache lines are never duplicated, EM² makes the most efficient use of available on-chip cache space, reducing references to off-chip memory and unlocking the potential for better performance within a given power envelope.