Performance: Model Construction and Solver Interfaces

This chapter documents tools for profiling model construction and improving the performance of both model construction and interaction with solvers. We begin by discussing various profiling tools which can be used to help identify performance bottlenecks. Pyomo has built-in profiling capabilities, but there are also Python packages, such as cProfile and line profiler, dedicated to performance profiling. Section 9.2 discusses the LinearExpression class, which can be used to drastically improve model construction time for some applications. Section 9.3 describes how persistent solver interfaces can be used to repeatedly solve models with small changes very efficiently. Finally, Section 9.4 discusses sparse index sets.