Explain the trip from a StarRC multicorner GPD to a single-corner PrimeTime annotation, and why multicorner extraction is worth it.
From PDVerse PrimeTime STA Interview Guide · pdVerse Mentor Guide
Ten-second Interview Answer
StarRC with SIMULTANEOUS_MULTI_CORNER: YES extracts several corners in one run into a multicorner GPD; PrimeTime picks one with set_app_var parasitic_corner_name <corner> then reads it. It pays off because the expensive geometric processing is done once for all corners.
Complete Technical Explanation
On the extraction side, StarRC run with SIMULTANEOUS_MULTI_CORNER: YES extracts several parasitic corners at once and stores them together as a single multicorner GPD. On the PrimeTime side you choose which one you want with set_app_var parasitic_corner_name <corner> and then read it with read_parasitics -format gpd -keep_capacitive_coupling <dir>; the read report echoes the annotated corner name, which is your confirmation that you got the corner you asked for. The economics are what make it worth doing. MCMM signoff needs timing at many RC corners - cbest, cworst, rcworst, typ and so on - and if you extract each corner in its own StarRC run you repeat the geometric processing every time, which is the expensive part of extraction. A single multicorner extraction does the geometry once and emits all the corners from it, which cuts extraction turnaround dramatically. PrimeTime then cheaply annotates whichever corner a given scenario needs out of that one database. If you enable the Parasitic Explorer before reading, you can even query across corners using -parasitic_corners . Structurally it is the parasitic-side analogue of cross-library scaling: do the costly work once and reuse it across corners rather than paying for it per corner.
In Pd / Signoff
This shapes how the extraction request is written for an MCMM signoff: one multicorner StarRC run rather than one run per RC corner, with each PrimeTime scenario setting parasitic_corner_name to the corner it needs before reading the shared GPD. The corner echo in the read report is the artefact you check per scenario, since a scenario silently annotated at the wrong RC corner produces credible but wrong timing.
Common Trap
Reading the multicorner GPD without setting parasitic_corner_name for the scenario, or not checking the corner name echoed by the read report. The annotation succeeds and the timing looks reasonable, but the scenario is signed off against the wrong RC corner - a failure no completeness check will catch.
Expect Next
Why does the geometric processing dominate the cost of running StarRC once per corner instead of once for all corners?
Topic And Primetime Commands
Back Annotation / SPEF & Parasitics. set_app_var, read_parasitics, parasitic_corner_name, parasitic_corners
Keep building interview depth
Continue practising STA
Continue learning free
Get a practical low-power chapter
Receive the existing “Low Power and Multivoltage Fundamentals” PDF chapter and its download link by email.
Continue practising