CUDA Persistent Threads
========================

A style of using CUDA which sizes work to 
just fit the physical SMs and pulls new
work from a queue.  Contrary to the usual
approach of launching more blocks than could 
possibly be operated on simultaeously.

* http://stackoverflow.com/questions/14821029/persistent-threads-in-opencl-and-cuda

* http://on-demand-gtc.gputechconf.com/gtcnew/on-demand-gtc.php?sessionTopic=&searchByKeyword=S0157&submit=&select=+&sessionEvent=&sessionYear=2012&sessionFormat=