| View previous topic :: View next topic |
| Author |
Message |
dynoboy
Junior
Joined: Sep 25, 2008
Posts: 5
Location: Texas
|
 Posted: Fri Sep 17, 2010 1:07 am Post subject: UVM challenges for VMM users! |
 |
|
Hi,
Have been reading a bit about UVM lately and looks like OVM extends to become UVM, but then what happens to VMM?
Is there an easy way to migrate VMM IP to UVM? Or does it mean entire re-write?
Also, can anyone comment on Synopsys roadmap, and where does UVM fit in.
Will VMM cease to exist or will there be some middle path they plan to choose.
For OVM users / supporters its not a big deal as UVM seems to be just an extension.
Comments! |
|
| Back to top |
|
 |
dave_59
Senior
Joined: Jun 22, 2004
Posts: 974
Location: Fremont, CA
|
| Posted: Fri Sep 17, 2010 2:50 am Post subject: |
|
|
The same committee that is developing the UVM earlier released an OVM/VMM interoperability kit. See http://ovmworld.org/contribution-detail/24916. This should preserve much of your VMM legacy.
Dave Rich
Mentor Graphics |
|
| Back to top |
|
|
JPK
Senior
Joined: Dec 28, 2007
Posts: 10
Location: Northeast U.S.
|
| Posted: Fri Sep 17, 2010 5:14 pm Post subject: |
|
|
I do not seen any problem using both. They are both open source now and it makes sense to incorporate the best of both in your work. In particular, VMM has far and away better real solutions where O/Uvm is still a concept that you must implement. This is coming from a person that has developed large testbenches in both. metrics upon request.
The vendor can comment for itself but I have seen a credible roadmap where all and any methodology you choose - or a hybrid of them - are supported.
JPK |
|
| Back to top |
|
|
sharanbr
Senior
Joined: Sep 27, 2004
Posts: 194
|
| Posted: Tue Sep 21, 2010 4:54 am Post subject: |
|
|
| JPK wrote: | I do not seen any problem using both. They are both open source now and it makes sense to incorporate the best of both in your work. In particular, VMM has far and away better real solutions where O/Uvm is still a concept that you must implement. This is coming from a person that has developed large testbenches in both. metrics upon request.
The vendor can comment for itself but I have seen a credible roadmap where all and any methodology you choose - or a hybrid of them - are supported.
JPK |
JPK,
I am under the impression that OVM is much more powerful than VMM.
I would like to hear from you why you think VMM is better.
Regards,
PS: I am just an user of these methodologies and the type of work I do, I don't even have a choice which methodology I can use |
|
| Back to top |
|
|
coryd
Newbie
Joined: Nov 17, 2010
Posts: 3
|
| Posted: Mon Mar 07, 2011 12:28 pm Post subject: Thoughts |
|
|
Pros: There is very nice support for the DUT reset and TB run phases. Also, the TLM 2.0 non-blocking transport support would have been nice on my last project, we had to construct the handling of multiple concurrent driver threads on our own.
Cons: I haven't looked much at the UCM 1.0 Prod Register Model yet, but it apparently is lacking some RAL features. It is disappointing that there is no implementation of the VMM Datastream Scoreboard. The uvm_scoreboard component is an empty shell... The first thing I would do if adopting UVM is find a way to use the vmm_sb_ds (maybe the OVM-VMM inter-operability toolkit would help?)
cheers |
|
| Back to top |
|
|
Janick
Site Admin
Joined: Nov 29, 2003
Posts: 1382
Location: Ottawa, ON Canada
|
| Posted: Mon Mar 07, 2011 8:45 pm Post subject: Re: Thoughts |
|
|
| coryd wrote: | | I haven't looked much at the UVM 1.0 Prod Register Model yet, but it apparently is lacking some RAL features. |
I don't see why: the UVM Register Library is based on RAL. The DUT integration had to be changed to fit UVM's mechanism, but the use model is the same. All capabilities have been preserved. |
|
| Back to top |
|
|
Janick
Site Admin
Joined: Nov 29, 2003
Posts: 1382
Location: Ottawa, ON Canada
|
| Posted: Mon Mar 07, 2011 8:50 pm Post subject: Re: Thoughts |
|
|
| coryd wrote: | | The uvm_scoreboard component is an empty shell... |
That's to be used as a base class for your scoreboard so UVM-aware tools would be able to identify it as the scoreboard.
You might want to look at the uvm_in_order_comparator in the pre-defined components section. Not as powerful as the VMM datastream scoreboard (single stream, one-to-one only, single comparison function) but it may satisfy your need. |
|
| Back to top |
|
|
|
Forum FAQ
Search
Usergroups
Profile
Digest
Log in
