[Gems-users] Is Ruby Cycles a good performance metric for CC protocol performance?

[Edward Lee <edwl202@xxxxxxxxx>]

Hi all,

I am investigating some optimizations in CC protocols. I am using only Ruby so far for my simulations. I was wondering while comparing two protocols of same base like MOESI_SMP_directory (default) versus a modified version of the same protocol, would it be fair to compare the Ruby Cycles performance metric to determine the affect on application execution time. Or should I use Opal to get more accurate results? I am already looking for cache misses especially coherence misses and network utilization parameters including the power model from GARNET.

Any suggestion is appreciated.

Regards,

Ed