* Joel Falcou has this intriguing Quaff project to build up compile-time skeletons of data flow for parallelized algorithms and IIRC we talked at BoostCon about how it would be neat to generalize skeletons to compile-time graphs. Anywhere you know the whole contents of the graph at compile time - in that case, the contents of some expressions - you might want to consider MPL.Graph for paying the abstraction penalty upfront instead of passing it on to your users. A lot is going on about Quaf at themoment, including a deep reflexion on
On 16/01/11 09:10, Gordon Woodhull wrote: proper way to capture the data flow itself and how ot capture the target architecture graph placement. Petri Nets ar ebeing considered atm. I have to ressurect this project once nt2 is on its way to release.