I haven't done the binary math to figure it out so far, but I wonder if signed vs unsigned division is also just doing an internal sign-extension to a wider type before the actual 'sign-agnostic' operation (e.g. a special case of the general sign-extension to a wider type). But as I said, not completely sure if that's the case, or if there are two entirely different hardware blocks for signed vs unsigned division. Same for muls vs mulu btw.