If you use a compiled language, you should periodically look at Godbolt and see what your code is doing and what changes to your code will do in the compiled output.
In this case a positively insane way of calculating squares and cubes generates 311 lines of ARM assembler output that will swallow your memory. With even something as simple as -O1 on the command line it's replaced by one or two multiplications respectively. With -fwhole-program it removes the functions entirely and interlaces them into the loop in main().
Know your tools. It makes huge differences!
"I mean outside of the areas that are documented by people that work, it is impossible."