Lawbreaking - In a P.E.R.F.E.C.T. way !



In 2009 The DARPA Red Balloon Challenge made headlines, but I didn't take a lot of notice at the time.  I didn't see any red balloons either.

However...

In Oct 2012 : MilitaryAerospace' Editor John Keller reported on the  advent of PERFECT.

The DARPA Microsystems Technology Office (MTO) last week awarded a $6.3 million contract to SRI International in Princeton, N.J.; and an $8.7 million contract to Reservoir Labs Inc. in New York for the Power Efficiency Revolution For Embedded Computing Technologies (PERFECT) program, which seeks to overcome power efficiency barriers that limit the capabilities of military embedded systems.

Scaling the current level of 1Gflop/Watt up to 75Gflop/Watt.  Most embedded computing for RTUs and SCADA systems rely for their power on either solar, batteries, or at best, rather tenuous power supplies.  Getting remote supervision happening with higher amounds of local computing at lower power drain is a continual challenge for researchers. Actually - having higher Flop performance per watt could also have it's advantages particularly in networked local computing in remote SCADA centres.

Some years back at the 8th Annual HPEC 20004 conference, the Workshop at the MIT Lincoln lab stated as it's theme,

"the End of Moore's law and it's Impact on Embedded Computing."

 

So just how do you benchmark High Performance Embedded Computing ? Fortunate there's a web reference for that... HPEC Benchmarks

I'm interested in these benchmarks because, once a better faster lighter, more energy efficient Flop is available, we should be able to do a higher order computation of local requirements.  It is unlikely that our community would be interested in the higher order military applications, so I left these out of the list, however the following applications could be of interest..

1. Time-Domain Finite Impulse Response Filter Bank - TDFIR


2. Frequency-Domain Finite Impulse Response Filter Bank - FDFIR


3. QR factorization - QR


4. Singular Value Decomposition - SVD


5. Constant False-Alarm Rate Detection - CFAR


6. Pattern Matching - PM

Many SCADA applications are relegated to a far distant host due to limitations in local computing horsepower.  With the advent of HPEC computing in the next few years, the above benchmarks could herald a time of doing much more interesting automation at the remote site.

 

 

 

The PERFECT project's timeline is around 4 years beginning 2013.  Let's see if DARPA has created a lawbreaker... Moore's law that is.

 

 

 

What do you think of Moore's law ? Is it time to break it ? Will the DARPA 'Red Balloon' of the PERFECT project succeed ?