I've tested this on a FreeBSD VM (Skylake running VMware) through simply
building the kernel and updating the system from source, but it's not a
great environment in which to measure performance.
A similar patch (but against our 10.x derived system, which is much different
in the details) has run through much of our regression tests. I'm planning
to run performance tests on a variety of server systems from Ivy Bridge
through Broadwell with up to 32 cores or so, but focused on overall systems
reliability & performance.
If we have general many-core FreeBSD scaling tests, that would be useful.