How smart is the Compiler with math.h

[ToniAz]

Hello everyone!

I have hit the speed wall and it seems that floating point operations is holding the application back from early completion, though I doubt there is any way around. My question is:

How smart is the compiler when it comes to some common functions in math.h, namely pow. When the "exponent" is a small integer, for example pow(x,3), does the function return x*x*x?

Are there some smart shortcuts supported too by math.h to anyone\s knowledge?

Thanks!

[helios]

Compilers are generally not clever enough to to outright eliminate function calls. Clang is probably the cleverest, and I doubt it can do it.

I'm not sure if this is standard behavior, but VC++ defines a pow(double,int) overload that uses a loop instead of calling the actual pow(). When both parameters are double, though, like in pow(3.141592,2.0), it still does call pow().

[modoran]

VC++ defines a pow(double,int) overload that uses a loop instead of calling the actual pow()

I doubt it works if you use math.h, I think only works with cmath.