Skip to content

2019 01 19

Jump to bottom
Matthias Köfferlein edited this page Jan 20, 2019 · 2 revisions

Persistence and some Reworks

There is little use in extracting the netlist if you cannot do something with it. Plus, as extraction takes a while you want to save the results. This is what persistence is for.

Essentially the scheme is

iter = ... # the RecursiveShapeIterator for your layout hierarchy
l2n = RBA::LayoutToNetlist::new(iter)

... extract devices, nets ...

l2n.write("myfile.l2n")

# to read the data back
l2n = RBA::LayoutToNetlist::new
l2n.read("myfile.l2n")

To cover this requirement, I had to introduce two new features:

  • Device abstracts: A device abstract describes a device in terms of shapes. Device abstracts sit between the device class (a generic type of device) and the device itself (the actual instance). Device abstracts carry the shapes of the device terminals and correspond to a layout cell.

  • Named layers: Layers derived from LayoutToNetlist have to be named so they can be put into the file. Named layers carry an extended meaning: only named layers are kept within the LayoutToNetlist object. All other layers will be discarded when no longer used. This is in particular useful for computed layers. Only named layers can participate in the connectivity.

The dump file format

The dump file is a text file containing the ingredients of a LayoutToNetlist object:

  • General information (database unit etc.)
  • Layer definitions (what layers are there)
  • Connectivity (what layer connect to which other layers)
  • Global connections (what layers connect to which global net)
  • Device abstracts (device cells with the terminal shapes)
  • Circuits

Circuits are made from these components:

  • Nets (electrical connections with the shapes making up the connections)
  • Devices (the devices and what nets connect to their terminals)
  • Pin (the connections the circuit makes to the outside)
  • Subcircuits (other circuits being embedded into this circuit)
Clone this wiki locally