> On Sep 17, 2018, at 2:51 PM, Peter Maydell <peter.mayd...@linaro.org> wrote: > > On 17 September 2018 at 18:56, Programmingkid <programmingk...@gmail.com> > wrote: >> >>> On Sep 17, 2018, at 1:52 PM, Peter Maydell <peter.mayd...@linaro.org> wrote: >>> >>> On 17 September 2018 at 18:48, Programmingkid <programmingk...@gmail.com> >>> wrote: >>>> What I hope to do is implement a PowerPC CPU flag that indicates if an >>>> rounded result is inexact. >>> >>> How is this different from the float_flag_inexact flag that softfloat >>> already >>> supports ? >> >> Sorry I meant to say if a rounded result is different from the intermediate >> result. This is what the FPSCR[FR] flag indicates on the PowerPC. > > Looking at the PPC spec that seems to be "inexact but not overflow" so you > should be able to calculate it via suitable handling of our existing > float flags.
So if an inexact exception takes place without the overflow bit set, it means the fraction rounded bit should be set? I don't believe that would work for PowerPC.
