SBG Vertical sorting

[Kalashnikovni]

I'll implement a new version to generate a causalized model that is used afterwards to generate simulation code. The main changes are the following:

• Add guess vertices that are entry points of each SCC.
• Add residual vertices that are the exit points of each SCC.
  This change allows the distinction between models that have many SCCs from those that have a large SCC. For example, for a model with many SCCs the output will be something like:

for i in 1:N loop
  guessA = ...;
  eq1;
  ...
  eqk;
  resA = ...;
end for;

In contrast, if a model has a large SCC the output will:

guessA = ...;
eq1;
for i in 1:N loop
  eq2;
end for;
...
resA = ...;

I'll keep iss-94 branch which has an older version of this algorithm just in case, if I need to reuse it.

[Kalashnikovni]

I started improving the ModelicaSBG structure, so that all information relating to the analyzed model is kept inside that structure. Also, I modified set-edges so the domain and translation members work as in the case of set-vertices, instead of keeping the final value directly in the domain.

• iss-95 commit

[Kalashnikovni]

Next, I rewrote the output of the horizontal sorting. Now it is a pair of (AST::Equation, AST::ExpList), because in the final result, if we have arrays of algebraic loops we must indicate the matching inside a loop, so a list of expressions is needed to represent that situation. Then I improved the code to generate Modelica code indicating the matching as a comment next to each equality.

• iss-95 commit

[Kalashnikovni]

The next step was to update the algebraic loops detection conversion to Modelica code. I only print equations now to indicate the algebraic loops.

• iss-95 commit

[Kalashnikovni]

Afterwards, I also applied the same changes to the tearing algorithm.

• iss-95 commit

[Kalashnikovni]

Finally, I reached the vertical sorting stage. First, I modified the VerticalSortingResult::toModelicaFormat function, to traverse first a map ordering SCCs, and then it goes to analyze the order inside each SCC.

It was also necessary to add guess and residual vertices to the ModelicaSBG structure. Since Modelica arrays are "rectangular", for each array variable a, I create variables guess_a and res_a with the same size of a. If the guess and residual are not accessed in all of their ranges, in the future I might need to balance the system of equations with fillers such as guess_a[i] = -1;.

• iss-95 commit

[Kalashnikovni]

To close this issue I also added guess and residual variable declarations to generate valid Modelica code after causalization.

• modelicacc-dev commit