No subject

alike), that used locks to implement work-stealing, should be a fairly
tractable target. Once you've done that (with appropriate abstract),
changing to wait-free shouldn't be too much work. The emphasis there
is on getting the right abstraction, so changing the bits to make.

So, if it was me, I would work using CCSP as reference (borrow the
algorithms and ideas, maybe a little bit of code), but do a clean
start architecting a _work-stealing_ scheduler in the TVM.
The work is in the software engineering, understanding and
architecting the TVM... in doing that I think you'll realise the most
suitable algorithmic fit for the TVM.
Going straight from the start line all the way to wait-free or copying
and attempting to port all the existing code will probably lead to you
running out of steam half-way through.

That's my opinion, and is it just that, an opinion on what I would do
in the same situation :-)
Feel free to keep sending questions, the more specific they are, the
easy they are for me to answer (i.e. the quicker the response).

Cheers,

Carl