[Gems-users] MOESI_CMP_NUCA protocol and ruby

["Mishali Naik" <mishalinaik@xxxxxxxxx>]

Hi,

            I have been running splash benchmarks on  4
processors/chip (PT_TO_PT topology) with MOESI_CMP_NUCA protocol and
DNUCA predictor. But what I have been seeing is there is more activity
on processor - 0 (more L2 accesses) but less activity on processors
-1, 2, and 3 (less L2 accesses). I am wondering if MOESI_CMP_NUCA
protocol supports 4 processors/chip? Or if I need to make any changes
to ruby in order to see same amount of activity on all the processors.
           I appreciate all your help.

thanks,
Mishali