On Thu, Nov 8, 2012 at 9:04 PM, Jakub Jelinek wrote:
> Hi!
>
> With -fwrapv the range test optimization miscompiles the following testcase
> (both inter-bb version in 4.8+ in first function and the pure fold-const.c
> one since 3.4). The problem is that with -fwrapv and signed type
> -x in [-,b]
Hi!
With -fwrapv the range test optimization miscompiles the following testcase
(both inter-bb version in 4.8+ in first function and the pure fold-const.c
one since 3.4). The problem is that with -fwrapv and signed type
-x in [-,b] doesn't imply x in [-b,-] and -x in [a,-] doesn't imply
x in [-,-