[sv-ec] Why must coverpoints be declared in a covergroup?

Is there a reason why coverpoints can only be declared in a covergroup?

It would be more useful if they could be declared outside a covergroup.  It
would be more consistant with other aspects of the language.  Much like
sequences can be declared separately from properties in SVA.

I cover modes and configurations my DUT has operated in.  A coverage class
setup specifically for that purpose samples the state whenever the mode
changes.  Elsewhere, another coverage model  is tracking transactions
flowing through the system, sampling everytime a transaction flows through.
It can be advantageous to cross the mode with the type of transaction.
Currently, if the mode is a simple variable, the cross is trivial.  If
instead you want to cross against a complex coverpoint or another cross
definition defined elsewhere, you are out of luck.  It would seem the only
existing solution is to recreate the coverpoint definition in two places,
or "unroll" the coverpoint/cross declaration by creating an enumerated
variable and procedurally mapping values to the enum so it can be sampled
as a simple variable by multiple covergroups.

Regards,
Ryan

Ryan Warner
Seagate Technology
ryan.s.warner@seagate.com