On 08/21/2018 03:46 AM, Richard Biener wrote: > > + /* If we're definitely dividing by zero, there's nothing to do. */ > + if (wide_int_range_zero_p (divisor_min, divisor_max, prec)) > + return false; > > I know we didn't do this before but for x / 0 we should compute UNDEFINED > as range. With the current interfacing this special case would require > handling > in the non-wide-int caller. No strong opinions here, just a couple notes.
If we set the result for this to VR_UNDEFINED, then it'll naturally be handled optimistically at PHI nodes, COND_EXPRs, etc. We'd normally get that effect via path isolation when we turn the x / 0 into a trap and simplify the CFG appropriately. However, using VR_UNDEFINED for division by zero in VRP captures the effect earlier which is generally a good thing. Jeff
