As I recall there was some known reason why it wouldn't work - I remember us
saying at a tutorial that "in theory" you could do something SMARTS-like but
that we didn't have the capability to return back to an old cpu yet.  I
don't remember what the reason was, I was assuming Kevin would
remember....Kevin?

Lisa

On Tue, Nov 4, 2008 at 11:53 AM, Steve Reinhardt <[EMAIL PROTECTED]> wrote:

> As far as I know there's no reason why switching back to the original
> CPU shouldn't work.  If it's not working I'd say that's a bug.  If the
> symptom is that nothing happens, it sounds like the tick events on the
> CPUs aren't getting scheduled, or maybe the thread context is not
> properly transitioning to the active state.  The first thing I'd check
> is what state the thread context of an old CPU is left in after you do
> a takeOverFrom() on it... if that's not the same state as a newly
> created idle CPU then that's probably at least part of the problem.
>
> Steve
>
> On Mon, Nov 3, 2008 at 2:18 PM, Clint Smullen <[EMAIL PROTECTED]>
> wrote:
> > I ran a test that tried to switch back to a a set of timing simple
> > CPUs after running with the O3 cpus for a while, but the simulator
> > just goes off into space. Using a fresh set of simple CPUs solves the
> > problem, but I was curious if someone might know of an inherent reason
> > why switching back did not work. If a user wanted to switch back and
> > forth multiple times (perhaps executing a set of simpoints with O3,
> > and the rest with simple timing to get accurate cache modeling),
> > having to have all new sets of cores for each part would be annoying
> > (though still doable). For a scheme where the switching is more
> > dynamic, this solution would not be feasible.
> >
> >        - Clint Smullen
> > _______________________________________________
> > m5-users mailing list
> > [email protected]
> > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> >
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
>
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to