2 Replies Latest reply on Nov 28, 2018 11:11 AM by inderjeet.kaur

    LVS comparison hyperscaling


      What are the benefits of using -hyper cmp option for LVS run? This was used to see any benefits for run time reduction for full chip LVS. The runtime increases by 5.5 hrs as compare to no cmp run.

        • 1. Re: LVS comparison hyperscaling

          Set up properly, the benefit is speed. It sounds like you may not have had enough memory on your master. (I'm not sure how much LVS generally uses -- I mostly work with OPC, which can be huge.) 


          The Calibre Verification User's Manual warns about -hyper cmp "...recommended when a machine has at least four cpus. Hyperscaling uses the same licensing environment as MT and Calibre MTflex modes. No additional licenses are required. There is increased memory usage on the master host when using ‑hyper."  (Underlines mine.)  Post-tapeout operations regularly overload the memory when runs try to use the master's CPUs as remotes, and the big clue is that the run slows to a crawl or even freezes while the machine frantically swaps memory.  Do you also see the performance hit when using other machines for the remotes?


          Hope that helps-



          • 2. Re: LVS comparison hyperscaling

            Thanks for the answer, Sam. My question is for single machine usage only and not for master+remote setup. LVS comparison stage uses 1 CPU (-hyper means hyperscaling to be used for extraction stage only). Specifying cmp with -hyper option means that hyperscaling should be used for comparison also. This should ideally reduce the run time for comparison stage. But I am not seeing the decrease in run time.