Re:Asynchronous CPU, MIPS
by kelnos on Saturday January 23, 2010 @07:53PM
(#1072722)
This stuff is really cool. One of my college professors, Rajit Manohar, was one of the guys who did a lot of the major research into async VLSI in the 90s. He and a few of his colleagues founded a company in San Jose (Achronix) that designs and builds async FPGAs.
It's an amazingly different way of thinking about computing. Generally async designs take up more physical die space (because of things like dual-rail logic), but the power savings can be pretty significant since you're not running a clock... basically the logic runs as fast or as slow as it needs to, and doesn't do much of anything when idle.
I took a class where we designed and did a little layout for some CPU-subset blocks. It was fun, but often very challenging. At least at the time, there weren't many (any?) good test/validation tools. A friend of mine has been working on one for the past 8 years or so (http://www.csl.cornell.edu/~fang/hackt/), first as a part of his PhD thesis, later as a part of his job.
It's been years since I've touched this stuff... if I'd continued down the EE path (like many others, I switched to software post-college) I'd (ideally) be doing async stuff now.
It's an amazingly different way of thinking about computing. Generally async designs take up more physical die space (because of things like dual-rail logic), but the power savings can be pretty significant since you're not running a clock... basically the logic runs as fast or as slow as it needs to, and doesn't do much of anything when idle.
I took a class where we designed and did a little layout for some CPU-subset blocks. It was fun, but often very challenging. At least at the time, there weren't many (any?) good test/validation tools. A friend of mine has been working on one for the past 8 years or so (http://www.csl.cornell.edu/~fang/hackt/), first as a part of his PhD thesis, later as a part of his job.
It's been years since I've touched this stuff... if I'd continued down the EE path (like many others, I switched to software post-college) I'd (ideally) be doing async stuff now.