On 18/04/2021 01:55, NightStrike via Gcc-help wrote: > I hope I'm using the term "constant folding" here correctly. If not, > I apologize. > > Is there a limit to the amount of mathematical subexpressions that GCC > can remove at compile time? > > Assuming the following: > > int x = 2 * 3; > > Presumably with the right compiler options, that will compile as: > > int x = 6; > > But if that is more complicated: > > int x = 2 * 3 * 4 / 6 / 7 + 8 - 9; > > how far will GCC go before it hits some internal limit where it can no > longer do the calculation at compile time? > I think in cases like this, there it is limited by memory. Some analysis and compile-time calculations have limits (such as for constexpr functions or nested templates in C++) which can be adjusted by compiler command line options. But usually you have to be doing something /very/ strange to hit the limits. You can test your code to see, using an online compiler such as: <https://godbolt.org/z/jsojsePGj> Look at the generated assembly - even if you are not familiar with assembly, you should be able to see quickly if the results of the expression are calculated at compile-time or run-time. (Don't forget to enable optimisation.) You can also use "__builtin_constant_p(x)" to determine if the compiler sees "x" as a compile-time constant or not.
