SetsulFor float that means the choice is between 0.199999988079071044921875 and 0.20000000298023223876953125. The latter is simply closer to the "true value" 0.2 than the former.
So who determines how rounding is done between IEEE and the CPU architecture maker?