Hi! This testcase contains an out-of-bound access, which isn't necessary for the thing the testcase tests. I've verified that both the original and patched testcase hangs r233928 and succeeds with r233934 and committed to trunk as obvious.
2017-09-04 Jakub Jelinek <ja...@redhat.com> PR tree-optimization/70043 PR testsuite/82093 * gfortran.dg/vect/pr70043.f90 (fn1): Start loop from 1 instead of 0. --- gcc/testsuite/gfortran.dg/vect/pr70043.f90.jj 2016-03-03 15:28:39.000000000 +0100 +++ gcc/testsuite/gfortran.dg/vect/pr70043.f90 2017-09-04 09:49:20.261236667 +0200 @@ -1,4 +1,5 @@ -! { dg-do compile } +! PR tree-optimization/70043 +! { dg-do compile } ! { dg-additional-options "-Ofast -g" } ! { dg-additional-options "-march=haswell" { target i?86-*-* x86_64-*-* } } @@ -6,7 +7,7 @@ subroutine fn1(a, b) real(8), intent(in) :: b(100) real(8), intent(inout) :: a(100) real(8) c - do i=0,100 + do i=1,100 if( a(i) < 0.0 ) then c = a(i) * b(i) a(i) = a(i) - c / b(i) Jakub