[shadow-support] clarification on arguments to enable deterministic simulation

Giang Nguyen cauthu at gmail.com
Wed Oct 12 18:45:40 CDT 2016


> > 2) Seems like I also should use "--cpu-threshold" with a negative value to get determinism? If so which value is reasonable? Any negative value?


Oh sorry, in this second question I'd meant to refer to
"--cpu-precision". Wondering if you caught that...



> I know we had determinism working at one point, but I have not verified it recently. The strip_log_for_compare.py script in the tools directory is meant to help you compare two shadow log files from separate but identical experiments to see if the output is indeed identical.

That's good to know. Thanks!


More information about the shadow-support mailing list