Designer’s Guide Community

mas_serg

New Member

Offline

Posts: 4

Segmentation Fault in the spectre simulation
Jun 28^th, 2004, 6:42pm

I very often get "Segmentation Fault" in transient spectre simulation after 20-30% completeness. I use IC5032 on Red Hat 7.2, 2.8G P4, 2G RAM. I simulate freqency devider (oscillator plus 12 flip-flops plus some analog blocks).
Why it happened?
For mixed mode simulation - no problem. But for final check in analog mode I always have to come back to solaris system and I always get good results. But it is not convenient to jump from one OS to another. By the way, linux is much faster. Solaris has cadence version 5.0... and 4G RAM. But I don't now, how to put so much memory to PC in order to check this reson. I heard, version IC5033 has the same problem.
May be I need change some spectre env. but I didn't do that in solaris.
Any ideas?
(I hope, Mr. Andrew Beckett and Mr. Ken Kundert have some opinion for this situation)

Thank you in advance.
Sam.
29.06.2004

Back to top

IP Logged

Andrew Beckett

Senior Fellow

Offline

Life, don't talk to
me about Life...

Posts: 1742
Bracknell, UK

Re: Segmentation Fault in the spectre simulation
Reply #1 - Jun 28^th, 2004, 10:19pm

I assume you're not running out of memory? The memory limits for 32 bit applications are lower on Linux than on Sun (2G as opposed to about 3.7G on Solaris), if my memory serves me correctly.

The seg fault could be almost anything. Are you running on any particular Linux distro?

Can you report it to customer support? (with the netlist and models which show the problem). If you can't use the standard support channel, send me a message on the messaging system in this forum with your email address, and we can communicate that way (note, this is not an open invitation to everyone to do the same!).

One wild idea is that you might have some discontinuities in the device models (caused by poor modelling, and insufficient capacitance to smooth these out) - try setting the cmin parameter on your tran analysis to 1e-15 (i.e. 1fF). This adds a small capacitor from every node to ground; a workaround, but I've seen this help in odd cases where you get crashes due to device modelling problems.

Regards,

Andrew.

Back to top

IP Logged

mas_serg New Member Offline Posts: 4	Re: Segmentation Fault in the spectre simulation Reply #2 - Jul 5^th, 2004, 6:59pm Thank you for answer, Mr. Andrew. I have sent to you a private message with my email, but didn't get answer. May me my message disappeared in some web place? Let me now, please. Thanks, best regards, Sam.
Back to top	IP Logged

Andrew Beckett Senior Fellow Offline Life, don't talk to me about Life... Posts: 1742 Bracknell, UK	Re: Segmentation Fault in the spectre simulation Reply #3 - Jul 6^th, 2004, 9:06pm I got it, just haven't had a chance to follow up yet. I'm rather busy at the moment. Andrew.
Back to top	IP Logged