Trade accounts receivable turnover
What you want to do is multiply the two dimensions together, then divide by two, and only then finally convert the result to separate digits for printing. So what you should be doing is something like this. Of course, a modular multiplicative inverse is faster than either, and is always an option since AAM only supports an imm8 constant divisor (which doesn't have to be 10; that's just the default when used with no operand ). On Haswell: AAM is 8 uops, 21c latency, one per 8c throughput.