It's kind of interesting, Linux is perhaps the penultimate example of standard "critical section" based kernel design. Thanks to brute force from the likes of

It's kind of interesting, Linux is perhaps the penultimate example of standard "critical section" based kernel design. Thanks to brute force from the likes of SGI and IBM this has been proven to around 4096 thread system images. I wonder how hard it will be to get to the next level. I wonder if it will continue to be logical structures on top of critical sections (I think RCU kind of fits this characterization), or if currently academic solutions of radically different design pay enough dividends that they start to make sense. And if so, what are some of those possible solutions? Perhaps there's inspiration from HW disciplines that can serve as inspiration?