Spectre simulation ADE vs command line

I'm able to successfully run a spectre simulation via ADE with version
4.4.6. But if I use Simulation -> Netlist -> Recreate and save the
resulting netlist to file and try to simulate it with >spectre
my_netlist.scs I get a completely different simulation result.

I'm a bit puzzled as to why this would be. Surely spectre is using
the exact same netlist in either case.

Any ideas?

Thanks,

Michael

 

Michael:
Look in the netlist directory for a file called runSimulation. Do you
get different results if you run runSimulation?

 

Thanks Erik.

If I simulate via ADE the simulation completes and I get reasonable
looking output.

If I use Simulation -> Netlist -> Recreate and save the resulting
netlist to file and try to simulate it with >spectre my_netlist.scs,
Spectre immediately finds fatal errors (models complaining about
divide by zero).

Now if I use runSimulation I will get identical output to that when I
run in ADE until spectre reports that it is entering remote command
mode, at which point spectre has another error.

So this leads me to believe that one of the command line options in
runSimulation is causing a different execution. My run simulation
contains the command:
spectre -env artist4.4.6 +log ../psf/spectre.out +inter=mpsc
+mpssession=spectre0_280 -format psfbin -raw ../psf input.scs

Now I've issolated these options and I'll always get that divide by
zero error unless I'm using the interactive options.

It's a bit disturbing that the batch mode result is different than the
interactive mode result.

Any explanation or other things I should try?

Thanks,

Michael

 

That's exactly right Erik. If I run:
spectre -env artist4.4.6 +log ../psf/spectre.out -format psfbin -raw
.../psf input.scs

I'll get fatal errors where the model complains about divide by zero.
If I simulate in ADE no errors are produced and I get a reasonable
simulation result.

Michael

 

I don't think that I'm using compiled models. In either case
CMI_CONFIG is nil. If I create the ocean script and run it as you
instruct it simulates without error.

If I run >spectre input.scs from the command line the error message is
something like this:
Error found by spectr in model:instance during hierarchy flattening.
model_file: model: parameter 'w': Attempt to divide by zero.

If I run from the command line with spectre 5.0 it also runs without
error. The error only seems to occur with spectre 4.4.6.

Michael