On 3/2/20 3:16 PM, BALATON Zoltan wrote: >> (2) IEEE has a number of implementation choices for corner cases, and we need >> to implement the target's choices, not the host's choices. > > But how is that related to inexact flag and float_round_nearest_even rounding > mode which are the only two things can_use_fpu() function checks for?
float_round_nearest_even is the default rounding mode, and is what the host fpu has been set for. In order to use hardfloat for a different rounding mode, we would need to change the host fpu control register, which is slow. Inexact is the only flag that we cannot compute easily from the inputs and output. We check for NaN and Inf as inputs and outputs, which is the clue that we may have to raise Invalid or Overflow. r~
