Nomadic Threads: A Runtime Approach for Managing Remote Memory Accesses in Multiprocessors

This paper describes a multithreaded runtime system for distributed memory multicomputers that significantly reduces the number of message transfers when compared to conventional " remote memory access " approaches. Instead of statically executing on its assigned processor and fetching data from remote storage, a Nomadic Thread transfers itself to the processor which contains the data it needs. This enables Nomadic Threads to take advantage of the spatial locality found in many algorithms that use arrays of data, because segments of arrays are made local by the migration of a thread to the proper node. By reducing the number of messages and taking advantage of locality, the Nomadic Threads approach allows programs to execute faster than conventional approaches while providing a simple runtime interface to compilers. Nomadic Threads are currently implemented using Active Messages [1] for the Thinking Machines Corp. Connection Machine 5 [2].