Connecting Nodes Together

There are two ways to connect two nodes in a graph (i.e., to draw edges between nodes):

  1. With Graph.connect_by_name()

  2. With Graph.connect(), Node.o(), and Node.i()

Graph.connect_by_name() tries to automatically connect outputs of one node to inputs of the other node if the files have the same name. Multiple edges can be drawn with a single call.

Graph.connect() explicitly connects a single output of a node (specified by Node.o()) to a single input of another node (specified by Node.i()). Only a single edge is drawn.

Graph building is generally cleaner when connecting by name. The more explicit connect API is useful when names do not match and it is inconvenient to adjust node configurations to make them match.

Automatic Connection by Name

Here is a simple graph with two nodes, one for the design RTL and the other which runs synthesis. We would like to connect the design RTL to the synthesis node.

_images/connect-1.png 
g = Graph()

rtl = Node( ...  get rtl  ... )
dc  = Node( ... get synth ... )

g.add_node( rtl )
g.add_node( dc  )

The RTL node has an output “design.v” and the synthesis node takes an input “design.v”. Since these names match, we can use Graph.connect_by_name() to simply connect these nodes with an edge like this:

g.connect_by_name( rtl, dc )

We get this graph with automatic connection:

_images/connect-2.png

Explicit Connections

Here is another simple graph with the same two nodes for design RTL and for synthesis. However, the names no longer match. We would still like to make this connection.

_images/connect-3.png

We want to connect the RTL node’s output “GcdUnit.v” to the synthesis node’s input “design.v”. Since these names do not match, connecting by name will not automatically draw any edges.

We can connect explicitly using Graph.connect(), Node.o(), and Node.i():

g.connect( rtl.o( 'GcdUnit.v' ), dc.i( 'design.v' ) )

We can read this clearly: Connect rtl output “GcdUnit.v” to dc input “design.v”.

Node.o() and Node.i() are helper methods necessary for differentiating outputs from inputs within a node. For example, the synthesis node has an input file “design.v” (i.e., the RTL design) and also an output file “design.v” (i.e., the synthesized netlist) with the same name.

We get this graph with our explicit connection:

_images/connect-4.png