Summary
The recipe boils down to three steps
This chapter consists of a recipe for fine tuning an existing AIMMS application such that AIMMS more efficiently executes the definitions and statements and efficiently generates the constraints. The recipe consists of the following three steps:
First, construct active subsets by removing all elements for which the variables are fixed in advance. These active subsets should then be used throughout your core model. This reduces the work each time, even in the evaluation of the index domain conditions to be constructed next.
Second, construct index domain conditions for the parameters, variables, and constraints of the core model. This will make several dense expressions seemingly execute more sparse, because only a limited number of elements are evaluated. Especially with variables and constraints this avoids the generation of columns for fixed variables and empty constraints. Thus the number of bottlenecks in your application is further reduced.
Finally, use the AIMMS profiler to pinpoint those assignment statements,
FORloops, and constraints that still absorb a considerable amount of computation time, and analyze and possible modify them. A checklist that can be used for this analysis has been presented in Analyzing and Tuning Statements.