So inaction on overflow checking in hardware is regularly justified by claims that overflow checking in software imposes only a minor performance burden (viz. O

So inaction on overflow checking in hardware is regularly justified by claims that overflow checking in software imposes only a minor performance burden (viz. OP), while inaction on overflow checking in software is justified by claims that that it imposes an unacceptable performance burden. I feel that something doesn't add up here.