Different result with vmanager and standalone simulation, with same seed

There are lots of possible ways your results might end up being different. Are you using SV randomize() in your testbench? If so, you could try adding "-xceligen heartbeat" to the xrun command in both setups and diff the logs to see if the seed sequence changes somewhere during the test. There are other solver options that can help, if heartbeat makes it look like a seeding issue, have a look at this article: How to debug Random Instability across the different runs of same test

If you have waveforms for both tests, you could use the "simcompare" tool to diff the waveforms to identify the earliest point in time where the waves differ - it's easiest to do this inside SimVision using the "SimCompare Manager" under the "Tools" menu.

There are lots of possible ways your results might end up being different. Are you using SV randomize() in your testbench? If so, you could try adding "-xceligen heartbeat" to the xrun command in both setups and diff the logs to see if the seed sequence changes somewhere during the test. There are other solver options that can help, if heartbeat makes it look like a seeding issue, have a look at this article: How to debug Random Instability across the different runs of same test

If you have waveforms for both tests, you could use the "simcompare" tool to diff the waveforms to identify the earliest point in time where the waves differ - it's easiest to do this inside SimVision using the "SimCompare Manager" under the "Tools" menu.

Hi Stephen,

first of all, I want to thank you for fast response.

I am not sure if I am using randomize.I know that things that I drive from the test and testbench should be fixed. I know that I get different seed each time I run regression. But when I find that seed of a log in vmanager, and use it in standalone sim, I get that stand alone sim to pass, and it is failing in vmanager.

Maybe I need to mention that other tests are passing for me in regression, and they are passing each time I run them. This issue is only on couple of tests. This is verification using Specman. I am new to this, so I might miss some terminology.

What I was trying to say is that if you compare the logs for the same test+seed combination for vManager and stand-alone tests, you should be able to home in on the reason for the difference in result.

If you're using Specman instead of UVM-SV then you need to check the random stability / seed sequence a different way, but starting with a simple log diff would be best.

You should even take care to check that the exact same set of "e" files are loaded, etc.

Ok, I will do so and get back to you, thank you.

Hi All,

first I want to say thank you to StephenH. I found the issue with observing the log and looking into actual waveform of the sim. My problem was:
1. I was using generate statement in Instance generation, for different setups.

2. Generate function is elaborated and done only once in the begging. And it will take values from the initial, default setup in the model, as verification was using precompile script to speed up regressions, and everything after that was disregarded, in terms of generate setup.

Luckily, I managed to see the issue, after a lot of careful observation of waves and logs.

Conclusion: Generate function + precompile script was the issue.

Thank you StephenH, I really appreciate the time and advice.

Great, I'm glad you found the source of the problem and were able to fix it! Thanks for the feedback. :)