A History of Placement Programming and Optimization

摘要

Since the start of SMT production, the optimization of machine programs has been one of the most interesting challenges in software engineering. With the amount of flexibility available from today's machines, optimization is so complex that even the latest supercomputer would take months to find all possible sequences of placement. SMT program optimization must complete in minutes or even seconds to be practical and useable, and so a much more intelligent approach needs to be taken. Let's look under the hood at SMT placement optimization algorithm strategies, building up in layers from the basic "random walk" exercise, through common feeder setups and line balancing, to finally consider where it has all gone wrong and look at a completely fresh approach.