Right-size precision

04 Aug 2014

The need to right-size precision to save energy, bandwidth, storage, and electrical power

The representation of real numbers within computers leaves a lot to be desired. Every programmer is bitten at some point by floating point precision issues. Open a JavaScript console and type 0.1 + 0.2 for the canonical example. This fascinating paper frames the problem in whole new perspective and proposes a couple of innovative solutions. As the author says…

Let’s make computer math much smarter than what it was in 1914!


comments powered by Disqus