Summary of omc scaling issues

[Francesco Casella]

This ticket collects all the known issues with omc scaling poorly with the model size:

• #1451 Backend scaling for simple array-equation
• #3695 removeSimpleEquations is slow on large models
• #3781 Fix scaling of detectJacobianSparsePattern
• #3816 Matching and sorting scales as O(N³) in models with large numbers of discrete variables
• #3905 findZeroCrossings scales as O(N²)
• #4005 EncapsulateWhenConditions scales as O(N³) in models where a when clause is triggered by a large vector of conditions
• #4006 SimCode scales as O(N²) in models with many when clauses and events
• #4388 collectInitialBindings scales as O(N²)
• #4398 comSubExp scales as O(N²)
• #4460 resolveLoops scales as O(N³)
• #4552 EvaluateParameters scales as O(N²) with large parameter arrays
• #4590 SimCode execution times scale as O(N⁴) with large parameter arrays
• #5107 SymbolicJacobian.generateSparsePattern scales as O(N²⁾
• #5108 Models with large arrays of parameters cause the runtime to segfault
• #5382 Model with large arrays and events causes simCode to scale badly

[Francesco Casella]

It seems we no longer have methods scaling as O(N³), which is a significant achievement.

There are still some O(N²) methods that should be addressed.