I can hardly believe it. I’ve been in the EDA business since 1980 when I joined TI’s Design Automation Department after graduating from Cal Poly with my BSEE. Since 1995, much of my attention has been focused on EDA standards. I reached a moment of truth this year when I admitted, albeit reluctantly, that I could be called a standards-lifer. So, I decided it’s time to share my perspectives on what’s going on in the standards arena. Welcome to my blog - I can’t wait to hear from you! OVM and VMM back in the news
Posted by Karen B on October 23rd, 2008
I read an article today titled “OVM vs. VMM: What’s Next?” Overall, the article describes the emergence and current state of these verification methodologies, and it talks about a “battleground” – a word that is sure to excite readers. In the world of the standards game, I don’t expect articles to always be 100% perfect, and I’m certain the author wrote his unbiased view, yet I noticed some inaccuracies and misconceptions that I’d like to comment on.Â
-
VMM from ARM and Synopsys was introduced in 2005, not 2007. Two additional years of use in our fast-paced industry brings significant experience and maturity to a methodology.
-
Mentor and Cadence did not “throw their technology into the open source world with Accellera”. It was Synopsys that donated to Accellera our complete VMM implementation. Mentor and Cadence were vocally opposed to the verification methodology standardization work in Accellera. Their open source introduction of OVM was separate from Accellera. Fortunately, they are now active participants in Accellera’s verification IP technical subcommittee.
-
Synopsys has not “kept the applications that run on top of VMM in-house”. On the contrary, these are included in Synopsys’ VMM open source offering.Â
-
VMM is not proprietary. It is freely available to everyone for download, without registration requirements – unlike OVM, I should add – at VMM Central.
I know that users want a single verification methodology standard, developed by a well-respected organization like Accellera. Let’s bring VMM and OVM base class libraries together under a single, Accellera-sanctioned standard and enable EDA vendors to compete on superior methodology applications and automation.
Â
EDA standards blog The Standards Game
VMM Verification Methodology Manual open verification methodology OVM
Accellera VIP TSC
RSS
Hi Karen,
It’s great that Synsopys is working through Accellera to standardize VMM, which is now available for download and use under the Apache license without registration. The vision of bringing VMM and OVM base classes together is excellent. This would cut costs by allow verification engineers around the world to speak the same verification terminology. Something else that would make IP more reusable and verification skills more portable would be a common register abstraction. Are VMM RAL, RALF and ralgen part of the complete VMM implementation that was donated to Accellera?
Thanks,
Jeremy
Hi Jeremy,
Thanks for your reply. Yes, VMM RAL, RALF were part of the Accellera donation. RALGEN is a small SystemVerilog-generating utility that is free for download at http://www.vmmcentral.org. Typically, Accellera doesn’t ask for donations of utilities, but we wanted to be sure the VMM utilities were available to the industry anyway.
Good luck in your endeavors,
Karen
Karen,
you’re being a little bit disingenuous with this snippet: “Mentor and Cadence did not “throw their technology into the open source world with Accellera”. It was Synopsys that donated to Accellera our complete VMM implementation.”
Donating something to accellera does not make it open source. Open source is s type of license: GPL, Apache, LGPL etc. that specifies the rights the copyright owner of the code grants to users of that code. Open source licenses are permissive rather than restrictive – that’s what makes open source software “free” – not necessarily free “as in beer” but “free as in speech”.
As for registering in order to download the OVM source, I beleive this was to enforce the Apache licensing terms – but I prefer the method Synopsys have for downloading the VMM in that the Apache license is presented as a click-thru before the download proceeds.
As for users wanting a single standard, I doubt that will ever happen. After all, verilog and VHDL still co-exist
Really, the “methodology” wrappers on top of SV (in other words the VMM and OVM base class libraries) are really at a pretty low level in terms of the overall effort required in verification. The hard part is figuring out what your DUT is supposed to do and being confident you’ve verified enough of it that it is at a point where it can be taped out. Whether your environment is “plumbed” in a VMM or OVM style doesn’t make much difference. Pretty much all 3rd party VIPs are not written in SV anyway and just have an adaptation “gasket” that allows them to connect to a VMM env or an OVM env.
Paul.
Hi Karen,
Do you know when VMM-LP would be released to public? I can hardly wait to hear the news.
Thanks in advance.
BR,
Tony Hsu
Hi Tony,
It will be soon – January, 2009 is the target. Final peer reviews are wrapping up. If you’re not already signed up, you can sign up to be automatically notified via VmmCentral when it comes out:
http://www.vmmcentral.org/cgi-bin/vmmlp/reg1.cgi
Thanks for your interest and Happy Holidays,
Karen
Hi again, Tony -
The VMM-LP is out now. Details at: http://www.vmmcentral.org/vmmlp/vmmlp.html
Best regards,
Karen
I think it is funny that I stumbled across this site when looking for UVM/OVM vs VMM comparisons. The blogger makes such a big deal out of not needing to register to get VMM…well it seems they have changed their tune and now you do need to register.
At least I can still download UVM and OVL without registering.
Hi Joe -
Well, it took me a while, but I finally got to the bottom of this. I didn’t realize that registration had become a requirement – it certainly wasn’t at the beginning. After we researched it, we realized that the registration requirement should be removed.
So we did! Registration is no longer required (but you do have to agree to the license, as with other standards).
I want to say thanks for pointing this out to me. Otherwise I wouldn’t have been able to correct it.