The optimiser uses the syntax tree generated by the parser and prunes redundant parts of it. This greatly reduces the size of the program to be compiled, while achieving the same result.
In this compiler, the optimisation runs before the target code generation, so the generated code is optimised. The implemented optimisations include:
-
Replacing constant variables with constants – this is done by checking if a variable is ever reassigned. If it is not reassigned (i.e. it is a constant), all references to the variable are replaced with the constant instead.
-
Replacing operations on two constants with the result – this is done by checking arithmetic expressions for having constants as both of the variables. If so, the operation gets performed during compilation-time.
-
Removing unused variables – this is done by checking if a variable is declared but never used. If this is the case, the declaration is removed.
-
Reducing the size of the tree – this is done by reducing the number of sequence nodes by merging sequence nodes with only one child.
This is done by recursively optimising the syntax tree repeatedly until no more optimisations are made during a single run. As a result of this amount of heavy optimisation, all of the sample programs are compiled down to a single print statement.
The optimised version of a previous example can be seen in Figure 19.
Figure 19. Optimised Version of Tree in Figure 13 (A1.2. number-addition.ts)
If the target code generator is given the optimised syntax tree above, the code in Figure 20 is generated.
#include <stdio.h>
int main() {
printf("%d", 3);
return 0;
}Figure 20. Code Generated from Figure 19 (A1.2. number-addition.ts)
Finally, running the Mocha test suite on the optimised compiler reveals that all of the tests defined in the Test Cases section pass, as seen in Figure 21.
Figure 21. Final Run of Mocha Test Suite (from Test Cases)