Here is a semi-proposed patch for private futex support, one of the
items on the roadmap for the next release cycle. However I've been
unable to demonstrate it yielding better performance than non-private
futex, and since it adds (small, but nonzero) complexity and worse
behavior on outdated (pre-2.6.22) kernels, I'm a bit hesitant to
actually commit it. If anyone is interested in this feature, please
see if you can find some examples that demonstrate that it measurably
improves performance.

Rich