[Bug middle-end/55863] New: Fails to fold (i + 2) - (i + 1) to 1

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55863



             Bug #: 55863

           Summary: Fails to fold (i + 2) - (i + 1) to 1

    Classification: Unclassified

           Product: gcc

           Version: 4.8.0

            Status: UNCONFIRMED

          Keywords: missed-optimization

          Severity: normal

          Priority: P3

         Component: middle-end

        AssignedTo: unassig...@gcc.gnu.org

        ReportedBy: rgue...@gcc.gnu.org





For



(analyze_overlapping_iterations

  (chrec_a = {pretmp_1529 + 1, +, 1}_20)

  (chrec_b = {pretmp_1529 + 2, +, 1}_20)



The SIV test currently fails because when computing the difference

between the CHRECs we compute pretmp_1529 + 2 + ~pretmp15_29.  This

is because when computing A - B we check if B is easy to negate - which

X + 1 is - and compute it as A + ~X which association then fails to

optimize.



This makes dependence tests of this kind fail.