My (admittedly unsophisticated) mental model is that int32 and the like are simple primitives that can be stored in registers and use the processor's native ADD

My (admittedly unsophisticated) mental model is that int32 and the like are simple primitives that can be stored in registers and use the processor's native ADD/MUL codes, while BigInt is going to be some boxed structure with its own implementation of arithmetic operations. For as common as number munging is (especially in somewhat lower level code that Go shoots for), it would carry a huge performance penalty, wouldn't it? Or are there more sophisticated ways of offering a "true integer" type that I'm not aware of?