When you hear 'supercomputer' and 'small' in the same sentence, you likely wonder how small a room are they talking about. However, when the folks at MIT and Texas Advanced Computing Center say small, they mean small as in pocket-sized.
A collaboration between MIT and the Texas Advanced Computing Center has led to an Android application that the two institutions claim can do honest-to-goodness supercomputing.
TACC explains that the team at MIT performed a series of expensive high-fidelity simulations on TACC's Ranger supercomputer and then generated a smaller, reduced model, which was used to create an Android application for a Nexus One.
"You don’t need to have a high-powered computer on hand," insists David Knezevic, a post-doctoral associate in mechanical engineering at MIT. "Once you've created the reduced model, you can do all the computations on a phone."
Though this kind of model reduction has been done before, TACC says the MIT system's real advantage is its rigorous error bounds, which tell the user the range of possible solutions, and provide a metric of whether an answer is accurate or not.
Knezevic goes on to say that using a reduced scale model also results in faster computations.
"The payoff for model reduction is larger when you can go from an expensive supercomputer solution to a calculation that takes a couple of seconds on a smart phone," he said. "That’s a speed up of orders of magnitude."
Check out MIT for a video of the process in action, or TACC to learn more about how they did it.
