Dataflow computations are represented by a class of marked graphs called process marked graphs. A general algorithm is presented for implementing process marked graph synchronization with multiple threads that communicate using only reads and writes of shared memory.

PDF file

Details