2010/04/19 Gregory Crosswhite
Thanks for the link; my ultimate interest, though, is in an architecture that could scale to multiple machines rather than multiple cores with shared memory on a single machine. Has there been any interest and/or progress in making DPH run on multiple machines and other NUMA architectures?
I wonder what it would take to do this. One approach is some compiler "magic" that provides you with an RTS that can communicate with other RTSen over TCP and chunks the computation "appropriately". Or maybe you give it a chunk size and it gives you some number of executables that find one another using Bonjour. Values not on "this node" are found via some hashing scheme (the same scheme used to chunk in the first place). There is a lot to know about this problem area. It would be a great alternative to OpenMPI. -- Jason Dusek