Is there a way to increase the number of cores being used by the ams simulator? I see that no matter what I select spectre/aps++ multithreading etc, I just see 100% CPU in the top command. How do I enable higher CPU usage? The simulator used is ams.

Hi Andrew,
The thing is that my simulations are fast enough without any transistors or such pdk elements. Even verilogA, AnalogLib, ahdllib components don't slow down. The moment I put two inverters, the simulation becomes like hundred times slower. I would like to know if the problem is genuine and if there's a way out. To speed up. What could be the cause?

Could you please tell me as to a information you require?

Simulation time is proportional to the product of the number of equations being solved and the number of time points used. If things are slowing down dramatically, one of those two things must have increased significantly. You should figure out which. Then you should try to understand why it is changed. The addition of the inverters may result in trapezoidal rule ringing, which could result in many more time points. In this case the solution is simply to switch the transient option 'method=gear2only'. Or it may be that you have placed the inverters on the most active nodes, and the simulator requires more time points to accurately render the transitions (by adding inverters you are adding capacitors, and simulators try to control error on capacitors in order to compute the capacitive currents accurately). To help understand the time points, it is helpful to set the waveform tool to show you the actual time points used.

Oh, and most of these kinds of problems are caused by people setting simulator settings inappropriately. I have noticed and increasing tendency for designers to use ridiculously tight tolerances. So the first thing you should do is unset any accuracy related setting you set (reltol, abstols, errpreset, etc.) Just leave them all blank. Then, only tighten tolerances when there is a demonstrated need to. Do not do it prophylacticly.

-Ken