[Gems-users] link power in orion 2.0


Date: Tue, 6 Sep 2011 10:18:47 +0800
From: Qi Zuo <bitsunnyday@xxxxxxxxx>
Subject: [Gems-users] link power in orion 2.0
Hi,
       I tried to use orion 2.0 to get more accurate link power and link area. The leakage and dynamic link power is calculated as following:

       Pleakage = LinkLeakagePowerPerMeter(link_len_in_m, Vdd) * link_len_in_m * data_width;
 
       Pdynamic = 0.5 * load * LinkDynamicEnergyPerBitPerMeter(link_len_in_m, Vdd) * freq * link_len_in_m * (double)data_width;

       where the link_len_in_m represents the link length (in meter). I think the load should be passed by Garnet and I calculated the load as following:

       double sim_cycles = (double) (g_eventQueue_ptr->getTime() - m_net_ptr->getRubyStartTime());
       load = (double) (m_link_utilized)/ sim_cycles;

       where the m_link_utilized is a member of class Link_d from garnet which I think represents link utilisation.

       I'm not sure whether this calculating is right or not. Would anyone give me some hints?

       Thanks a million!

--
Regards
Qi Zuo

School Of Computer Science and Technology
Beijng Institute of Technology, China, 100081
[← Prev in Thread] Current Thread [Next in Thread→]
  • [Gems-users] link power in orion 2.0, Qi Zuo <=