Towards Refinement of Unbounded Parallelism in ASMs Using Concurrency and Reflection.

The BSP bridging model can be exploited to support MapReduce processing. This article describes how this can be realised using a work-stealing approach, where an idle processor can autonomously grab a thread from a partially ordered pool of open threads and execute it. It is further outlined that this can be generalised for the refinement of an unboundedly parallel ASM by a concurrent, reflective BSP-ASM, i.e. the individual agents are associated with reflective ASMs, i.e. they can adapt their own program.