The idea is to evaluate the program at compile time and obtain a trace of the program that allows you to reconstruct a semantically equivalent program which has some desirable properties. In your case (and most cases in literature) it's done for optimization. Partial evaluation, which is a subset of supercompilation, is similar to what you are doing. See the wiki page: https://en.wikipedia.org/wiki/Partial_evaluation
Some literature:
https://dl.acm.org/doi/10.1145/5956.5957
https://ndmitchell.com/downloads/paper-rethinking_supercompi...