Sorry for the slow reply... yes, I would expect that if you want to model time sharing across N programs in SE mode, the easiest way would be effectively to build a machine with N CPUs that share L1 caches, then add some event-based infrastructure to make sure that only one is active at a time (modeling the scheduling algorithm).
The switchCpus stuff is when you want different CPU models to run the *same* program at different points in time, so it's not directly relevant here. Steve On Thu, Aug 18, 2011 at 8:57 AM, Stevenson Jian <[email protected]>wrote: > Hi, > Is there a way to do time sharing in M5 se mode? I want to alternate > between two different workloads on the same processor to study the effect on > the cache. I know that there is the m5.switchCpus(switch_cpu_list) command > to switch from one type of cpu to another, such as simpleAtomic and O3. It > works fine as long as the workload is the same. When I set the workload of > the cpu in the switch_cpu_list to something else, I am getting a > segmentation error. What would be the proper way of doing time sharing in > M5? Alternately, since I am only concerned with effect on the cache, I can > create 2 CPUs, each dedicated to one workload, but I want only one CPU to be > running at any given time and alternate between the 2 CPUS to mimic the > behavior of time sharing. How would I go about doing that? Could someone > point me in the right direction? > Thanks, > Steve > > _______________________________________________ > gem5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
