Hi Stijn,
I'm confused about what code you're looking at; in simulate.py, I see:
# Restore checkpoint (if any)
if ckpt_dir:
ckpt = internal.core.getCheckpoint(ckpt_dir)
internal.core.unserializeGlobals(ckpt);
for obj in root.descendants(): obj.loadState(ckpt)
need_resume.append(root)
else:
for obj in root.descendants(): obj.initState()
That said, there is an issue with restoring checkpoints to systems that
switch from simple to detailed CPUs due to some changes in how naming works,
and I'm working on that right now. The basic problem is that in the system
you restore to, the workload objects belong to e.g. switch_cpu0 instead of
cpu0, and because of that difference they don't get restored properly. You
should be able to checkk in your config.ini and see if that's what's
happening to you.
Steve
On Fri, Mar 11, 2011 at 2:30 AM, Stijn Eyerman
<[email protected]>wrote:
> Hi,
>
> I recently updated to the most recent version of M5(-dev). I use ALPHA_FS
> mode and checkpointing. After updating (and applying my changes), I am not
> able to restore from checkpoints anymore. If I use the checkpoint-restore
> options, it just starts simulating at cycle 0, without restoring.
>
> A closer look showed me the following:
> The checkpoint directory is set in configs/common/Simulation.py. I checked
> that and it's OK.
> Then m5.instantiate(checkpoint_dir) is called, which is in
> src/python/m5/simulate.py. However, in that function, the argument
> (ckpt_dir) is nowhere used. It just states:
> # Restore checkpoint (if any)
> for obj in root.descendants(): obj.initState()
>
> So why is ckpt_dir constructed, while it is not used later on? And how are
> checkpoints restored using the command above?
> Since I'm not able to recover from checkpoints, something must be wrong. It
> used to work with a version of M5 just before July 2010, but it seems a lot
> has changed since then regarding checkpointing.
> Has someone seen similar problems? Or better: does someone have solutions?
>
> Thanks,
> Stijn
>
> --
> dr. ir. Stijn Eyerman
>
> Ghent University
> Electronics and Information Systems Department
> Sint-Pietersnieuwstraat 41
> 9000 Gent
> Belgium
>
> t: <%2B32%209%20264%203456>+32 9 264 3456
> f: <%2B32%209%20264%203594>+32 9 264 3594
> e: [email protected]
> w: http://www.elis.UGent.be/~seyerman/
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
> _______________________________________________
> 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
