Mentor Releases Customized VMM
Check Out Comments For 'Goering on Mentor/Cadence VMM Support'

Goering on Mentor/Cadence VMM Support

Wow... I guess you learn something new every day.  Richard Goering over at SCDSource has just written up a great article on the fact that both Mentor and Cadence now support the VMM.  When I wrote about this last week I'd only heard about the Mentor support.  One of the best parts of the article is Goering's description of a discussion he had with Karen Bartleson of Synopsys:

Bartleson said, however, that Synopsys has no intent to support OVM. "Our goal is to support the Accellera [VIP] standard, so we get some unification. For us to support OVM would mean more confusion in the industry."
Goering continued...
So why is Mentor's support for VMM a good thing for interoperability, while Synopsys support for OVM would sow confusion? "Because VMM has been out there for two to three years in production and OVM was only recently introduced," Bartleson said. "If Mentor supports a more widely accepted, de facto standard, that's good for the industry. If Synopsys supports OVM, which is a new and different implementation, it will fragment things more."
There are a couple of problems with this statement.  First, the stated short-term charter of the Accellera VIP committee is to create a mechanism to allow VIP written using the OVM and VMM to communicate.  Creating a unified standard has been repeatedly described within the committee as a long term goal.  The implication is that the OVM and VMM are both widely used standards that need to be considered by Accellera. The VMM is certainly not the de facto standard for Mentor and Cadence customers.  Therefore, it seems unlikely that Synopsys would fragment the industry any more than it already is by supporting both the VMM and the OVM.
 
Second, remember that the OVM is based on the AVM and URM, both of which have been out for quite some time.  That's especially true if you consider the eRM (upon which the URM and subsequently the OVM are based).  To say that the OVM hasn't been around for long is really just masking the reality on the ground that there is more than one widely used verification standard.
 
Now, there is an issue here that has not been discussed.  The Mentor and Cadence versions of the VMM presumably don't work on VCS (disclaimer - I have not tried this).  That means that any changes made by Synopsys in the future will not be reflected in the Mentor and Cadence versions.  The same thing would happen in reverse if Synopsys chooses to port the OVM to VCS.  Therefore, as I discussed in my presentation at DAC, it is also important to focus on the fact that the major SystemVerilog simulators remain incompatible with one another from a language support perspective.  What we really need here is the ability to run a single collection of SystemVerilog source on all three simulators.  Then we wouldn't need customized versions of the VMM and OVM, we could run the standard distribution everywhere.  However, as luck would have it we now have, in the form of the OVM and VMM, code that can be used to improve the language capabilities of VCS, IUS, and Questasim. 
 
My conclusion to all of this?  Don't hold your breath... We've got a long way to go before we see true interoperability between the major EDA vendors.

Comments