On Sept. 10, 2015, 4:20 p.m., Abdul Mutaal Ahmad wrote:
> > As you already mentioned, restoring from a checkpoint probably doesn't
> > work. Looks good other than that. I'm Ok with this change if you add a
> > warning when restoring from checkpoints, but please fix checkpointing in a
> > separate commit in that case.
>
> Andrew Bardsley wrote:
> Andreas' suggestions seems to work for the C++ config example (I've been
> freshening it too, expect a patch soon). I inserted:
>
> DrainManager::instance().preCheckpointRestore();
>
> Before checkpoint restoring and that seemed to work.
>
> Andreas Hansson wrote:
> Abdul, we pushed a fix a few weeks ago. Can you double check if any of
> the functionality in this patch is still needed, and if so re-post?
>
> Abdul Mutaal Ahmad wrote:
> Andreas, can you send me patch link?
>
> Andrew Bardsley wrote:
> The patch Andreas H is referring to was mine, Abdul, and it seems to have
> been committed already.
>
> Unfortunately, I didn't fix the problem in the SystemC example in util as
> this patch, 3101, already seemed to be addressing that example (I assumed
> Andreas S's suggested changes would be integrated into 3101 in time).
>
> Anyway, my patch fixes checkpointing for the plain util/cxx_config
> example (as noone else seemed to be updating it alongside the SystemC
> example). As the SystemC example is derived from the cxx_config example you
> should be able to apply its use of DrainManager::preCheckpointRestore() to
> the SystemC example.
>
> Andrew Bardsley wrote:
> OK, I've had a play and I can get an FS checkpoint from gem5.opt to
> restore if I add:
>
> if (checkpoint_restore) {
> std::cerr << "Restoring checkpoint\n";
>
> CheckpointIn *checkpoint = new CheckpointIn(checkpoint_dir,
> config_manager->getSimObjectResolver());
>
> + DrainManager::instance().preCheckpointRestore();
> + Serializable::unserializeGlobals(*checkpoint);
>
> /* Catch SystemC up with gem5 after checkpoint restore.
> * Note that gem5 leading SystemC is always a violation of
> the
> * required relationship between the two, hence this careful
>
> You'll notice the same additions have already been made to
> util/cxx_config/main.cc. Be sure to insert the unserializeGlobals *before*
> the time catchup code that follows.
>
> Abdul Mutaal Ahmad wrote:
> yeah, got it. I am working on it now. you will see a patch soon.
Andreas Sandberg, Andrew Bardsley: can either of you review this patch?
I'll commit it after I see at least one of you marking it as ship it.
- Nilay
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3101/#review7144
-----------------------------------------------------------
On Nov. 26, 2015, 12:28 p.m., Abdul Mutaal Ahmad wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3101/
> -----------------------------------------------------------
>
> (Updated Nov. 26, 2015, 12:28 p.m.)
>
>
> Review request for Default.
>
>
> Repository: gem5
>
>
> Description
> -------
>
> Due to changes in Drain Manager, the lastest systemc coupling doesn't work.
> The code for handling Checkpoint has been modified and it is now compatiable
> with new drain manager.
>
>
> Diffs
> -----
>
> util/systemc/main.cc UNKNOWN
>
> Diff: http://reviews.gem5.org/r/3101/diff/
>
>
> Testing
> -------
>
> Testing is being done on systemC coupling. It needs more testing to verify
> checkpointing feature.
>
>
> Thanks,
>
> Abdul Mutaal Ahmad
>
>
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
