Unsigned binary subtraction when numbers are too large to form two's complement

I am in computer architecture and my prof. hasn't been able to give a satisfactory answer.

Assuming we have a 32-bit processor, if we have two unsigned integers x and y, both of which are greater than 2³¹ (so that they are too large to be converted properly to two's complement form) how do we perform subtraction?

If the subtrahend was a constant I can see how the compiler 开发者_如何学运维could compensate for this in static analysis. But how do we deal with this possibility at runtime, especially in a statically-compiled language like C? Does this just throw an overflow error or do some architectures subu support special operations?

Everything I find for MIPS says that it just performs two's complement subtraction, but that shouldn't work in this case, should it?

The situation you pose is not a problem. Specific example:

  0xFFFFFFFF -  0xFFFF0000
= 0xFFFFFFFF + -0xFFFF0000
= 0xFFFFFFFF +  0x00010000
= 0x0000FFFF