Engineering a Physical Synthesis System

P. Groeneveld

Abstract

I will focus on the engineering job of building a huge EDA software system. I'll descibe and justify our choices for the fundamental algorithms we chose at Magma for the synthesis, placement and routing of huge systems. The overall goal was to build a system that is robust and as close as possible to 'correct-by-construction'. Also, I'll illustrate why - with such large systems - algorithmic optimality is often pretty bad. We'll look at the use of hierarchy for assembling large systems. Finally, I'll try to describe the ins and outs of working for a starup EDA company.