Varying the latency between simulations. As an example, if for a 4-way CMP, the average cache to cache (both LOAD and STORE) transfer latency is 50 cycles, I want to see the performance impact if the average latency goes down to 40 cycles.
> This would be better served on the mailing list [email protected]. > > > > The short answer is yes, M5 does do cache to cache transfers for a CMP. > > > > By varying the latency, what do you mean? Varying the latency during > simulation, or between simulations? > > > > > > > > I am new to m5. Could you please tell me what m5 does in case of LOAD for > shared cache line. Basically I need a CMP system with private L2 caches, > and > in case > of LOAD, the request should be satisfied by peer L2 caches (of course if > it > is available) instead of going to memory. I want to vary the latency for > cache to cache transfer. > Does the current coherence in m5 work this way?? Or, how hard will it be > to > change? > > thanks > > -- > Md. Kamruzzaman (Shawon) > Graduate Student > Computer Science and Engineering Department > University of California, San Diego (www.cse.ucsd.edu) > Member, Elite Problemsetters' Panel, VOJ, Spain > Homepage: http://www.cse.ucsd.edu/users/mkamruzz > > No virus found in this incoming message. > Checked by AVG - www.avg.com > Version: 8.5.339 / Virus Database: 270.12.53/2154 - Release Date: 06/04/09 > 05:53:00 > > _______________________________________________ > m5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users -- Md. Kamruzzaman (Shawon) Graduate Student CSE Dept., University of California, San Diego Cell# 858-205-9350 Homepage: http://www.cse.ucsd.edu/users/mkamruzz _______________________________________________ m5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
