Excellent. To answer OPs question: on some hardware, you cannot distinguish between an overflow and some other fault. On systems without this limitation, you mi

Excellent. To answer OPs question: on some hardware, you cannot distinguish between an overflow and some other fault. On systems without this limitation, you might include a model-specific header and look at a particular bit in flags to find the carry. I presume there are architectures, maybe with a math coprocessor, where nabbing those flags requires an interrupt. Asking the language to define math around each of those is actually less portable.