Sweet. Looks like you can make thousands of asynchronous kernels and it doesn't kill perf (ran testsuite with qthreads..had to build that, looks like a threading library from Sandia national labs). Pretty cool. Seems like far better oversubscription handling than TBB or OpenMP.
Cool to see this project again. I looked at it awhile back, and used it for a few small projects. Looks like the dev has been up to a lot of work in the interim. Any idea how long before we'll see a full beta release (assuming dev reads this?).