Home
People
Publications
Multiprocessor Architecture Watch
Concurrency Abstraction Watch
Laboratory

Andrew Mihal


Bio:

Andrew Mihal is a Ph.D. candidate in Electrical Engineering at the University of California, Berkeley. He earned his B.S. at Carnegie Mellon University in 1999. When not in the office, Andrew is an avid hiker and photographer.

Abstract:

Programmable platform architectures are a compelling alternative to ASICs for embedded systems, but they are notoriously difficult to program. The reason for this is the large discrepancy between the concurrency requirements of the embedded applications and the concurrency capabilities of the platforms: the "concurrency implementation gap". Domain-specific languages are a good way to capture application concurrency, but there are still many ways in which current platform architectures are a suboptimal match for application requirements. If programmable elements are to be the basic blocks of future architectures, one must reconsider what kinds of programmable elements should be designed in addition to how they should be programmed. The Cairn project takes a three-part approach to this challenge. Designers use separate abstractions for modeling applications, capturing the capabilities of architectures, and mapping applications onto architectures. Models of computation and actor-oriented design provide a mechanism for correctly modeling concurrent applications. The Sub-RISC paradigm provides efficient, cost-effective architectures and exports their capabilities. A disciplined mapping methodology leads to correct implementations and enables effective design space exploration. With the Cairn approach, designers can defeat the concurrency implementation gap and realize the benefits of programmable platforms over traditional ASICs.

Links:

Send feedback to mescal@gigascale.org
Contact 
©2002-2018 U.C. Regents