My gut instinct was that operations like sin, cos, tan, etc, were the kind that would get aggressively inlined by any modern compiler, and then after inlining t

My gut instinct was that operations like sin, cos, tan, etc, were the kind that would get aggressively inlined by any modern compiler, and then after inlining the redundant conversions between radian/turns would be canceled out by optimization passes. Of course, gut instinct can be wrong, so I checked in godbolt and was surprised to find out that even if I try to force lto, it still doesn't get inlined and remains a call.