Re: [Simulavr-devel] Trigger external component on clock cycle

[Lars Noschinski]

Hello!

* Klaus Rudolph <address@hidden> [2005-11-23 21:23]:
> sorry for my late answer, I am again totally busy :-(

Me too, so no problem ;)

> > I want to trigger some input (port change) each thousand clock cycles or
> > so. I guess this must be possible, but my C++ knowledge tends towards
> > zero, so ...
>
>
> I hope that you can derive from some simple classes like the uart which
> gives you an idea how to implement a class which simulates a
> synchronous triggered device which has external connections. Please
> take a look inside this class first. For detailed questions I maybe
> could give a more detailed information.

Ok, I'll have a look at serial*.

> > If I read the examples and the source correct, I'll need to have an
> > object, which I connect to the systemclock with Add (or
> > AddAsyncMember?). Do I have to create a C++-Class which derives from
> > SimulationMember and recompile simulavr or can this task be done also
> > from a (Tcl or Python) simulation script?
>
>
> You could recompile the simulator, or the library if you want the hard way :-)
> But you also can write your own main routine and bind against the
> library which makes much more sense.

Hm, ok :) I totally forgot about the library ...

> If you have a clocked device, maybe every 1000 cycles you simple use
> the Add member function with the given time. All async members would be
> accessed after each!  step of any! core cycle, which consumes a lot of
> calculation time and is not what you would do :-)

Then, I'll need an async member, as I want to have some percent
deviance. Performance isn't really a problem, as I need this for
interactive debugging - and I'm slower than simulavr ;)

> You have to create a own c++ class derived from SimulationMember.
>
> It is not possible to derive a class on ITCL base, which comes not from
> design of simulavr but of the swig wrapper. This is not really nice,
> but it is the price of using wrappers :-)
>
> But you also can write a script which steps the engine "manually" and
> change the pin state after 1000 steps/cycles. So you need no new class
> and no new main nor a recompile.  If this solves your problem and is
> fast enough you could do this in this way.

I could have thought of this solution ...

Thank you for your hints. Problem with simulavrxx is, that it is much
more useful then it seems to be at the first glance; only after I
absolutely needed a simulator, I found out, that it is powerful ;)

When I've some more time again, I'll write some little howto, to make it
a bit more visible.

Greetings,
   Lars