Dataflow architectures can tolerate long unpredictable communication delays and support generation and coordination of parallel activities directly in hardware by exploiting fine grain parallelism. But it has several drawbacks which must be overcome in order to be a massively parallel machine.
This thesis investigates the parallelism characteristics of tagged token dataflow architecture and shows the effects of control driven scheduling on it. Variable Grain Tagged Token Dataflow Architecture (VGTTDA) is devised for simulations. VGTTDA schedules dynamically variable grain macro operators within which instructions are executed in conventional RISC-like pipeline. By executing several test programs on the VGTTDA, it is shown that control driven scheduling mechanism improves the performance of dataflow computer.
