Explain dataflow computation model

An option to the von Neumann model of computation is a dataflow computation model. In a dataflow model the control is tied to the flow of data. The order of commands in the program has no role in the implementation order. Computations take place when all the data items needed for beginning execution of an instruction are accessible. Data is in constant flow independent of reusable memory cells and its accessibility initiates execution. Since data may be accessible for several instructions at the same mean time, these instructions can be implemented in parallel.

The potential for parallel computation is replicated by the dataflow graph, the nodes of which are the directions of the series and the edges of which characterize data dependency between instructions. The dataflow diagram for the instruction z = w × (x+y) is:-

Data movs on the boundaries of the graph in the variety of data tokens, which include data values and status information. A asynchronous parallel computation is resolute by the firing rule, which is uttered by means of tokens: a node of DFG can shoot if there is a token on each input boundary. If a node fires, it get through the input tokens, performs the connected operation and places result tokens on the output boundary. Graph nodes can be one instructions or tasks comprising many instructions.