The interesting change deals with the way nets inherit names from their connections. A net is an electrical node that results from the analysis of a cell definition (considering the definitions of its components, but none of its instances). For example, if cell type A contains three rectangles (r1, r2, and r3) that do not touch each other, and an instance (named b) of cell type B, and nothing else, and if cell type B contains nothing but one rectangle rb, and in A b.rb touches r1 and r2 but not r3, we get the following output: cell type B contains one net, corresponding to rb, and cell type A contains two nets, one corresponding to r1 and r2, and the other corresponding to r3. Nets may be contrasted to another concept, which we'll call signals here. A signal is like an electrical node on the silicon; it is all of what is eventually connected. A net is the part of a signal that is localized to a cell definition. Both kinds of Spinifex output (Thyme and Structural) name nets, not signals.

Spinifex recognizes that nets may have multiple names. The Thyme style output allows only one name; Spinifex chooses one to use, and puts in a comment giving the others.

If the designer gives names directly to a net (with the $SignalName property on an object comprising the net), those are all the names the net will have.

If the designer gives no names directly to a net, it inherits names from its connections. Each net is, in general, connected to nets in subcells. An otherwise unnamed net will try to inherit the names of the subcell nets it is connected to. If a name collision occurs (i.e., two different nets (in the same parent, of course) are both trying to have a certain name), they try instead using the subcell net name qualified with the subcell instance name. The qualified name is instanceName.subnetName. If there is still a name collision, the names are disambiguated by appending -1, -2, &etc. to the colliding nets. For purposes of naming (and too many other things as well), transistors are not considered subcells; they do not contribute anything to the naming issue. The rules in this paragraph are what's different from the old Spinifex; it didn't try qualifying the names.

If a designer gives no names directly to a net, and it is not connected to anything, it will get a simply generated name (e.g., n42).

For an example of qualified name inheritence, suppose cell type A contains a net, and an instance (named bOfA) of cell type B, and that cell type B contains a net named nb, and that the net in A is connected to the net nb of bOfA. Spinifex will try the following names in order for the net in A: nb, bOfA.nb, bOfA.nb-I (for some I).

Other changes in naming:

Spinifex will now output some summary geometric information in the Structural style: bounding boxes and instantiation transformations. See Geometry.Tioga in [Indigo]<Rosemary>Compare0.2>Compare.DF for details. This can be disabled by setting SpinifexOutput.outputGeometricInfo to FALSE.