Get email delivery of the Cadence blog featured here
In the numerous tweets, blog posts, and online forum discussions on the upcoming Universal Verification Methodology (UVM) standard from Accellera, I have seen a couple of references along the lines of "UVM=OVM+VMM" and that really concerns me. It concerns me because it's not accurate, but it concerns me even more because the very idea doesn't make sense.
Note that the Accellera decision was specifically to make the OVM the base for the UVM, with the expectation that additional features would be added on top of this base for future UVM releases. It is quite possible that some of the additional features may be derived from VMM as well as from OVM World contributions, but the UVM as it exists and is defined today is the OVM, not OVM+VMM.
So am I just splitting hairs here? I don't think so. There is a lot of overlap between the baseline features of the OVM and VMM, and any attempt to try to create some sort of "OVM+VMM" Frankensteinian hybrid would likely be a disaster. Either functionality would be duplicated to try to make it familiar to users of both methodologies or it would end up as a third methodology, satisfying no one.
Accellera did the right thing by choosing the OVM as the base for the UVM. I urge the committee to release a UVM 1.0 as soon as possible, fully compatible with OVM 2.1, and to maintain backwards compatibility in future UVM releases as new features are added. This will ensure that the 10K registered users at OVM World and all their colleagues have a painless transition to the UVM.
Keeping the UVM fully backwards-compatible with the OVM benefits VMM users as well. Recall that the first stage of the Accellera VIP effort was to define a library for OVM-VMM interoperability. This library provides VMM users a clear path to the OVM without having to recode all their VIP, so ensuring that UVM=OVM will provide VMM users a clear path to the UVM standard as well.
The truth is out there...sometimes it's in a blog.