It's not the first time I see this method popping up in profilers. It seems to be doing a lot of work (in a loop!) every jit compilation and can be cached (ISAs don't change). Here is a flamegraph from the bing service:

Another (less important) problem with that HasInstructionSet that MSVC doesn't want to inline GetRelativeBitMask here.

no matter if native PGO is applied or not, according to MSVC inliner's log it's "callgraph decision not to inline it" - looks like the reason it does that is that we compile coreclr with -Ox, it's inlined just fine with -O2