Material Input Data and Efficiency

Regarding efficiency while making materials,

When using Input Data style nodes, like ‘VertextNormalWS’ or ‘Absolute World Position’; - or even potentially other types of output-only nodes like ‘LandscapeCoords’ - does it make a difference if you use more than one node, each time you need to use the data?

For example, is this

More efficient than this? (Assuming they were connected to a larger network, etc.)

Or is there simply no difference at all?

Would not make difference, because it passes several stages of optimization before it becomes actual shader, and in the case you posted above, there code will be the same.