Flang's code is somewhat slower than gfortran's from comparing the geometric bottom line from the Polyhedron benchmarks on SKX with profile-directed optimizatio

Flang's code is somewhat slower than gfortran's from comparing the geometric bottom line from the Polyhedron benchmarks on SKX with profile-directed optimization. I won't probably argue argue with arguments about whether that's a reasonable comparison.