[sv-ac] Possible definition of $past

Hello,

to satisfy simulation and synthesis requirements for $past, would the
following solution be acceptable?

Best regards,
ed

----------

Definition of the behavior of $past:

In the following I try to provide a definition of the behavior of $past
that depends on the calling context. This definition is based on recent
discussions in the SV-EC committee regarding the structure of scheduling
regions. Namely, it appears that the structure may be changed so that
there is symmetry between the "active" and the "reactive" regions. This
would create an NBA-reactive region, similar to the NBA region in the
"active" space. NBA-like assignments in programs would thus be scheduled
to take place in the NBA-reactive region. 

The definition of $past that I propose uses either of the NBA regions to
update the value of $past depending on whether the function is called
from a module or from an interface. Recall that $sampled(x) returns the
value of x from the preponed region of the current time step.

If $past(x,,,@clk) is called from a module or interface (executing in
the active region), the equivalent RTL implementation that defines the
behavior of the function is

always @clk past_variable <= $sampled(x);

If $past(x,,,@clk) is called from a program (executing in the reactive
region), the equivalent behavioral implementation that defines the
behavior of the function is

initial forever @clk past_variable <= $sampled(x);

In assertions, it uses implicitly the $sampled value of the active form
of $past. 

---