Re: gEDA-user: Save/Restart Features in Verilog 2001?

[Paul Michael Hartke <phartke@Stanford.EDU>]

> phartke@Stanford.EDU said:
> > However, I see that annex is prefaced with "The system tasks and
> > functions described in this annex are for informative purposes only
> > and are not part of the IEEE standard Verilog HDL."
> 
> Ah, I missed that whole section:-)
> 
> Still, $save/$incsave/$restart are pretty unlikely to be addressed
> by me unless there is a pretty major clamor over it. I have other
> more pressing issues to deal with.

	I certainly respect that and recognize all the great work you do
on this project.  I figured as much which is why I asked for this info
from the other folks on the list in the original email.

> Independent of the standard, checkpoints would need to contain the
> state of all the running threads, the state and structure of
> all the netlist functors, and the contents of the event queue.
> This is theoretically quite possible. The threads would in fact be
> easy to checkpoint. The hard part would be the functors of the netlist.
> Functors would be tricky because they form a rather complex web of data
> structures.

	I have gotten spoiled by Java serialization!  But perhaps I
wouldn't be too happy with the performance of a Java Verilog simulator...
:-)

> The states of PLI applications, however, are completely out of
> reach of the simulation engine, and *that* glitch would sink most
> ships.

	This is what prompted my inquiry.  I am looking at several PLI
applications and was wondering if/how such checkpointing support would
work with Icarus Verilog.  It seems that all the VPI primitives are in the
standard--but perhaps only in the optional part.

> phartke@Stanford.EDU said:
> > cbStartOfSave/cbEndOfSave, cbStartOfRestart/cdEndOfRestart
> These are also described as "feature" callbacks, and are optional.

	I think I am getting a lesson in "standardise"--I never though to 
check which parts were optional or not!