Why is the ~item~ member of uvm_simple_sequence qualified rand?

[JonathanBromley]

hi,

In uvm_sequence_builtin.svh we find that uvm_simple_sequence has

protected rand uvm_sequence_item item;

The fact that item is rand causes VCS to issue a warning when a simple_sequence is randomized, because at that time (in method uvm_sequence_base::do_sequence_kind) the item is null. Incidentally, this same issue was reported on the OVM forum a while back in this thread by someone using OVM with VCS, but the root cause was not identified in that discussion.

I suspect the rand qualifier is superfluous. uvm_simple_sequence::body() will randomize the item by an explicit call to item.randomize. Generally, when a sequence has members that are items or subsequences it's correct for them not to be rand, and to allow the randomization to be handled by uvm_do calls in the body method.

Is there some special reason why the simple_sequence's item member is rand? If not, can we log this as a (minor) bug and get it changed?

Thanks

Edited June 24, 2010 by JonathanBromley

[David Long]

Hi Jonathan,

We fixed this problem by adding

item = new;

to the uvm_simple_sequence constructor in the UVM library that we use. My personal opinion is that it can be a good idea to create sequence items in sequence constructors rather than by calling `uvm_do in the sequence body (which would use the other sequence macros to re-randomize and send the sequence item). However, in this case, I think that your proposal is a better fix.

Best Regards,

Dave

[JonathanBromley]

Dave,

Thanks for the sanity check! We'll patch our local copy for now.

[Erling]

Hi Jonathan,

We fixed this problem by adding

item = new;

to the uvm_simple_sequence constructor in the UVM library that we use. My personal opinion is that it can be a good idea to create sequence items in sequence constructors rather than by calling `uvm_do in the sequence body (which would use the other sequence macros to re-randomize and send the sequence item). However, in this case, I think that your proposal is a better fix.

Best Regards,

Dave

But why is <item> a member of uvm_simple_sequence in the first place (and not a local variable in the body method)? Declaring <item> as a protected member seems to indicate that it exists for use by a derived class, but is this indeed the intention and what advantage does this bring?

Regards,

Erling

[john_fowler]

Hello Jonathan,

This rand is unnecessary as you point out. It was coded in OVM and did not result in a complaint from Questa or IUS so it slipped through.

There is no reason to keep the rand here.

John

[JonathanBromley]

Is there some special reason why the simple_sequence's item member is rand? If not, can we log this as a (minor) bug and get it changed?

Logged as bug #3142 in the Mantis database.