Hi guys,
Recently at SPAA 2012 we presented a paper you might be interested in. The title is "High-Performance RMA-Based Broadcast on the Intel SCC". We discuss how to efficiently use one-sided communication (i.e. MPBs) to implement the collective operations (broadcast in this case).
Here is the link dl.acm.org/citation.cfm?id=2312029
If you have any questions, feel free to ask here.
