Comment 7 for bug 1797335

I notice that your cpu flags list avx2; this suggests the possibility that the code path being used is one that's specific to avx2, not just sse2. I have attempted to reproduce this on CPUs both with and without avx2 and I cannot. I get no statistically-significant difference in runtime, on either the avx2 system or the non-avx2 system, with glibc from Ubuntu 16.04 vs. glibc from Ubuntu 18.04.

This could be a CPU bug or a bug in vagrant rather than in glibc; there's currently not enough information to tell. I think we would need a trace that shows which implementation is being selected in glibc in this environment first and foremost, to confirm whether it's the correct one.