https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125717
--- Comment #7 from Harald Anlauf <anlauf at gcc dot gnu.org> --- (In reply to Dhruv Chawla from comment #6) > (In reply to Harald Anlauf from comment #5) > > BTW: I just tried the variant with annot_expr_parallel_kind on the testcase > > > > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83064#c0 > > > > and it fails with -O2 -ftree-parallelize-loops=2 . > > So there is more to do. > > Sorry, not very familiar with Fortran - what do you mean by failure here? > From a quick glance, it seems like the testcase was doing float operations > in parallel (whose results will inherently depend on the order of > execution). So this testcase seems invalid to me? By failure I mean that one gets a completely wrong (incorrect) result. The above printed for me: PI -1.2097982560039648 PI 3.14159274 whereas without -ftree-parallelize-loops: PI 3.1415927131954575 PI 3.14159274
