So basically it'll work with a harness that tweaks parameters of a PCell/macro generator/drawing-script, mint a GDS for that "hard macro", then this script (well, a future version of it) juices a very fine-grained RC network (and of course "active" and less-active "special" devices) out of it, such that the resulting .spice netlist can be referenced from an application-specific testbench and run through e.g. Xyce's fully-intrusive PCE for efficiently (and yet accurately!) covering the entirety of PVT in order to provide quantified yield expectations under appropriately custom criteria (binning, clock de-rating, or even just accepting meaningfully reduced yield in exchange for more performance in the dies that make it).
For example, in my TX-only "SerDes" efforts I'd want to estimate the 2~3D distribution of "transmit eye opening size" (amplitude & phase!) vs. operating clock.
