PATCH RFC: Use explicit representation of infinity in VRP

Diego Novillo dnovillo@redhat.com
Fri Mar 2 15:22:00 GMT 2007


Ian Lance Taylor wrote on 03/02/07 10:14:

> No, wraps_on_overflow_p isn't right.  Floating point types don't wrap
> on overflow, but we don't want to use a special overflow indication
> for them, since they already have one.  To reverse the sense of
> uses_overflow_infinity we would need to test for "not an integral
> type" || "wraps on overflow".  I don't see an easy name for that.

Gah, you're right.

> How about simply needs_overflow_infinity?

Sure.



More information about the Gcc-patches mailing list