[Bug middle-end/82004] [8 Regression] SPEC CPU2017 628.pop2_s miscompare

[andrey.y.guskov at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82004

--- Comment #4 from Andrey Guskov <andrey.y.guskov at intel dot com> ---
Okay, finally I`ve got a minimal reproducer.
All this miscompare boils down to the following:

program r628
real(8) :: l = -3
print *, merge("PASSED", "FAILED", 1D-3 .le. 10**l)
end program r628

Normally, 10^-3 should equal 10^-3, but not after the optimization in question.

$ gfortran -m64 -O3 -o r628 r628.f90 && ./r628
 PASSED
$ gfortran -m64 -Ofast -o r628 r628.f90 && ./r628
 FAILED