Abstract: A significant part of the software out there performs mathematical calculations and often close-to-optimal performance is needed. Creating such optimized programs is very difficult and often leads to processor-specific code. I argue that, at least for basic mathematical functionality, computers should write (and rewrite for new processor generations) such highly optimized code. With Spiral we have built such a program generator for a small, but important mathematical domain. In this talk I discuss the main research challenges in expanding this work including recent progress.
Bio: Markus Püschel is a Professor of Computer Science at ETH Zurich, Switzerland, which he joined in 2010. Before, he was a Professor of Electrical and Computer Engineering at Carnegie Mellon University, where he still has an adjunct status. One of his major research topics is the automatic synthesis of high performance programs for mathematical functionality (www.spiral.net).
More information: http://www.inf.ethz.ch/~markusp