It seems like an outlandish claim, but the brainiacs at MIT and the Texas Advanced Computing Center insist that they've created an Android app that is genuinely capable of doing supercomputing on a phone. Here's how they did it:
First, the researchers performed a complex series of calculations on the Ranger supercomputer—calculations that only a real, take-up-a-whole-room supercomputer could handle—and then generated a reduced model of the results. They turned that model into an Android app, loaded it on a Nexus One, and then were able to adjust the parameters of the complex data set and visualize the results in seconds, from just about anywhere. Using a reduced model of the real supercomputer's work seems sort of like cheating to me, but then again who am I to question the guys at MIT.
The researchers claim the app is doing real-deal supercomputing because it accounts for the error introduced by the reduced model, giving the user a range of solutions for their chosen parameters. One of the researchers explains, "We have a bound on how much accuracy we're losing with our reduced model, so we can say with rigor that we're doing supercomputing on a phone."
The team thinks this type of system will almost certainly have real world applications, such as landmine detection, architectural modeling, or performance optimization for cars or planes. One researcher suggested that the reduced model app could receive input data from sensors and dictate small optimizations, say to an aircraft's ailerons or flaps, on the fly.
But simply showing how the smartphones we carry can be used to tackle such complex computational tasks is impressive in and of itself. As one of the TACC researchers said, "It's demonstrating that with a small processor, you can still get a meaningful answer to a big problem." [TACC via Slashdot]