editing input.scs not working

Rob,

I thought I'd sent this yesterday, but obviously neglected to hit the Post button...

Anyway, this is not new - the spectre interface in ADE has always worked this way. The idea is that (by default) spectre is run from ADE in "interactive" mode. What this means is that when ADE launches the simulation, spectre is launched and then reads the netlist. It then goes "interactive" and control is passed to ADE. ADE can then ask it to run analyses or change parameters. The idea is that you can run a simulation, change a parameter in memory, run simulation again, and so on - all the time spectre keeps running, and there's no need for the additional overhead of starting simulation, checking out licenses each time, re-reading the netlist - when most of that work has been done.

If the change is too significant (i.e. the schematic has been updated), then ADE tells spectre to exit and it is re-started, re-reading the new input.scs. 

So what this means is that if spectre is already running, spectre will ignore any changes to the input.scs.

You can solve this two ways:

• Hit the stop button in ADE. This will force spectre to exit, and so the next run button will definitely re-read the input.scs
• Set envSetVal("spectre.envOpts" "controlMode" 'string "batch") prior to starting ADE. This will mean that spectre's running in batch rather than "interactive" mode ("interactive" is the other available value). Each time spectre will start, read the netlist and checkout licenses, simulate, and exit. The downside of this is that with parametric analysis or quick design variable changes, you won't get the performance benefit. Note that if using APS, it always is in batch mode anyway.

Hope that helps!

Regards,

Andrew.

 You are the best Andrew! I'll give those a try.

If I understand your requirement correctly, you need to netlist, hack the netlist, then run simulation, and come back to ADE to continue other stuff, etc. Simulate won't help since it would renetlist either way. So there is another simple way. 

1) Do your convetional netlisting 

2) Open the netlist and modify it the way you want

3) Now 'cd' to the netlist directory, open the runSimulation file, and execute the single whole line in that file on the command prompt as-is. That's what goes to simulation. For e.g. for me it is:

a) cd /tmp/test/simulation/my_cell/spectre/schematic/netlist/

b) vi runSimulation

c) spectre  input.scs  +escchars +log ../psf/spectre.out  -format psfxl -raw ../psf   +aps +lqtimeout 900 -maxw 5 -maxn 5 -escchars  

d) New psf data is generated. Now come back to ADE and continue stuff (for e.g. --> Plot after simulation: refresh, or open Waveform viewer, etc.)

Hope this helps.

- Tito 

 Thanks. I wondered if I could just execute that runSimulation file.

Note that this (using runSimulation) won't actually work unless you take into account what I said earlier. If in "interactive" mode (which tito80 wasn't, because aps was being used, the script says:

# runSimulation scripts generated for spectre interactive are not intended to be used for command line.
spectre  input.scs +escchars +log ../psf/spectre.out +inter=mpsc +mpssession=spectre0_5417_1  -format psfxl -raw ../psf
-I/home/username/demos/OCEAN/Models +lqtimeout 900 -maxw 5 -maxn 5

You'd find that if you ran this, it would hang - because it is waiting to be controlled by ADE. You'd have to either generate the runSimulation script when in "batch" mode, or edit out the  +inter=mpsc +mpssession=spectre0_5417_1 part.

So it wouldn't have worked for Rob unless the advice I gave had been taken into account.

Regards,

Andrew.

Ok, Andrew, I tried your way (Stopping and rerunning, setting it to batch). But I am facing some problems. In the original scs, there are some parameters defined which I believe are picked up from spectre.ocn files in the netlist/ directory. However, when I execute "Run" for my modified scs (after killing Spectre in the previous run, and then modify orig.scs to modified.scs), it complains:

ERROR (SFE-1997): "input.modified.scs" 3538: I78: parameter `vhigh': The parameter `vcc' is used but not set.

And in my modified.scs, I have for the instance and paramter list:

I78 (ctlesel\<2\> ctlesel\<1\> ctlesel\<0\>) bus3 setval=ctle vhigh=vcc vlow=0 

...

parameters .... vcc ....

Any pointer on how to solve this parameter unset problem? 

Tito,

That's because you presumably did not set a value for the design variable in ADE, and so you've just got:

parameters ... vcc ...

rather than:

parameters ... vcc=1.2 ...

Spectre would not read the "spectre.ocn" file at all; that's just a record of what got changed - it's not really used by anything.

Regards,

Andrew.