Cilkplus is a valid approach if you are interested in using a task and data parallel language extension available on Intel's compiler, newer variants of gcc and Intel's fork of clang. As mentioned in the article a more comparable library for standard C++ is Intel's Threaded Building Blocks.

I used to work at Intel and was a technical lead for games developer architecture feedback and ran the multicore gaming initiative for a while. The feedback on cilkplus was that it didn't allow sufficient control, and as a non-standard language feature posed issues for porting to some systems. Intel's TBB however has been used by some games developers, and its API has evolved features in response to game developer feedback. Most games developers however prefer to roll their own.

Note that I didn't downvote this, but if you wanted to implement a task scheduler (the subject of the post) using cilkplus wouldn't be an obvious starting point.

Perhaps you're being a bit dismissive, but Cilk is an option people should explore. For those interested in how such things work, there are extensive papers from the Cilk group at MIT: http://supertech.csail.mit.edu/papers.html

In particular, I recommend "Cilk: An Efficient Multithreaded Runtime System" (http://supertech.csail.mit.edu/papers/PPoPP95.pdf) and "The Implementation of the Cilk-5 Multithreaded Language" (http://supertech.csail.mit.edu/papers/cilk5.pdf).