I'm pretty sure that what happens is that the GS card operates on everything above it and that they do "stack".
More specifically, whenever itNEC reads a GW
(or other stuffanother card that defines wires like GH
) it stores that geometry into an array in memory, and when it encounters a GS
card it simply scales the coordinates of everything already in memory; thememory. The GS
card is then "forgotten", so it doesn't have any effect on anything after it, and a later GS
card doesn't care whether whether or not some geometry has been scaled by an earlier GS
or not; if it has, it will get scaled again by the new one and their effects will multiply.
The GM
card operates in a similar way. The difference with GM
is that you can ask it to apply only to a subset of tag numbers — but it still only applies to what's above it, and the default is to apply to all tags. Cebik notes that relative ordering of GM and GS matters; if you have a GS
that scales to feet with a GM
above it, then the size of the GM transformation will be in feet (because the geometry will be unscaled when the GM gets to it, and then everything will get scaled down, including the transformation), but if you put the GS
before the GM
then the transformation will be in meters.
The main reason you don't see decks with more than one GS
is because it's pretty confusing — having a single GS
that defines "what your units are" and putting it at the end is by far the simplest thing.