The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Problems with the artil process
https://designers-guide.org/forum/YaBB.pl?num=1139920033

Message started by timsu on Feb 14^th, 2006, 4:27am

Title: Problems with the artil process
Post by timsu on Feb 14^th, 2006, 4:27am

When I am trying to run a Monte Carlo simulation the following error always occur:

"Fatal error found by spectre during Monte Carlo analysis 'mc1'.
cannot communicate with artil process."

The only reference of this 'artil process' i've manage to found is in the spectre reference manual:

"For each iteration of each Monte Carlo child analyses, Spectre (through Artil) will write a line to this ASCII file which contains scalar data..."

By skipping the nominal run i can 'bypass' this error and make a full run although since the problem still exist no scalar data is available after the analysis, only the process data.

Does anyone know what could cause this problem or how i can solve it?

Thanks
// Tim

Title: Re: Problems with the artil process
Post by Andrew Beckett on Feb 14^th, 2006, 9:32pm

In the netlist directory (either that or the psf dir), there should be a file "artil.log" which might reveal why this is failing.

artil is a process which is started by monte carlo in order to do the calculations of the scalar expressions.

What version are you using? I found a sourcelink solution indicating a problem in IC5033, which was fixed in the 5.0.33.500.3.25 subversion.

Regards,

Andrew.

Title: Re: Problems with the artil process
Post by timsu on Feb 15^th, 2006, 6:39am

The log file is always empty after the runs so that won't tell me much. We are using the version 5.0.33.500.3.24 so the problem might be solved by the new version, unfortunately our group doesnt have access to the support (and hence sourcelink?) so I could not investigate the solution you are mentioning.

Would it be possible to create a one-time process and mismatch information to use in normal transient simulations? It is the mismatch information that i'm mostly interested in so a corner analysis wont help me that much (and dcmatch is only for dc...).

Otherwise i might just have to hope that our next version upgrade will solve this problem...

// Tim

Title: Re: Problems with the artil process
Post by Andrew Beckett on Feb 18^th, 2006, 6:31am

Hi Tim,

Well, you could always not save any scalar expressions, tell it to save the family waveforms, and then do just one run - you can then plot the results from the transient run as normal. This would include mismatch. You could then vary the start run number to get different points in the sequence.

Regards,

Andrew.

Title: Re: Problems with the artil process
Post by timsu on Feb 21^st, 2006, 2:08am

After patching to a newer version this problem now went away and we are able to run the monte carlo analysis without any problems.

Thanks for your help Andrew.

//Tim