Greetings all,

Just wanted to run this by everyone to see if people thought it was a good
idea.

Right now the when one creates a checkpoint it has a plain old name.
However, if one tries to run a checkpoint that has a different number of
cores or a different memory size than what the checkpoint was created with,
there would be a failure. I'm wondering if it would be a good idea to
simply store the number of cores and the memory size as a suffix to the
checkpoint string.

So instead of naming the checkpoint "facesim", it'd be something like
"facesim__4C_4096M".

The easiest way to do this would be to simply modify create_checkpoint.py
and run_bench.py to save this metadata upon checkpoint creation and to
check that it matches the current configuration before running from a
checkpoint. The only question becomes how do we read the number of cores
out of the marss binary (as opposed to the number being user-specified
which could lead to a case where things are even more confusing). Perhaps
an extra runtime flag that will simply print out the number of cores and
exit?

I haven't looked at the source code for the actual checkpointing functions,
but this could probably also be done directly in qemu/marss, but that might
be overly complicated?

Does anyone think that this would this be worthwhile to implement?
Thanks,
Paul
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel

Reply via email to