Fixed-point math is a case where you'd really benefit a lot from a rich type system. Instead of "This is an integer," or "This is a floating-point number," you'd like to be able to say, "This is a 32-bit fixed-point number with 2 decimal digits of fraction," or perhaps more commonly, "This is a 16-bit fixed-point number with 8 bits of fraction". This allows you to do something like floating-point math, but with all the manipulation of the exponent done at compile time, leaving only mantissas to run-time. Especially for addition, subtraction, and comparisons, this is a huge savings on processors without floating-point hardware, such a huge savings that it's often worth it to do it by hand, without compiler support.
Type systems like C's can't cope with types like that. C++'s template system can do surprising things sometimes, but I don't think it can do it either. But you could certainly imagine a language whose type system was rich enough to let you do it.