SE mode has its own translation to "physical" addresses that are used for caching (to enable multiprogrammed workloads). I don't think there's anything in the virtual-to-physical mapping that guarantees that you'll end up with the same virtual address in different processes mapping to the same set.
Steve On Thu, Aug 26, 2010 at 10:41 AM, Weixun Wang <[email protected]> wrote: > Thanks. I understand now. > > One more question. If I run two identical workloads on two CPUs in a CMP > configuration, their mirroring instructions and data should be mapped to the > same set in L2 cache, right? I manually checked several of them and it seems > true. How about 3 or more CPUs? > > Thanks. > > On 8/26/2010 1:11 PM, Gabe Black wrote: >> >> Yes. That's probably the standard library which is compiled into your >> program initializing itself. That code is part of your workload and not >> provided by or controlled by M5. >> >> Gabe >> >> Weixun Wang wrote: >> >>> >>> Thanks for your reply. >>> >>> I'm running the M5 under SE mode, hence there should no be any OS. >>> However, if I execute a simple helloworld program (only a printf ...) >>> under M5 (TimingSimpleCPU), it requires 12492 instructions. Are you >>> saying all these instructions belongs to the helloworld program? >>> >>> On 8/26/2010 1:00 PM, Gabe Black wrote: >>> >>>> >>>> Weixun Wang wrote: >>>> >>>> >>>>> >>>>> Hi List, >>>>> >>>>> I have several questions regarding the start-up code of M5. >>>>> >>>>> 1) What code does M5 execute before the actual workload assigned to >>>>> it? What does it exactly do? >>>>> >>>>> >>>> >>>> None. All setup work is done within the simulator without running any >>>> simulated code. If you're using FS the OS would have to boot before you >>>> could run your user level application benchmarks, but then the OS is the >>>> workload. >>>> >>>> >>>> >>>>> >>>>> 2) If I assigned the same workload to two CPUs in a CMP config, will >>>>> the start-up code do EXACTLY the same (instructions...)? >>>>> >>>>> >>>> >>>> Yes, they would both start with the workload. >>>> >>>> >>>> >>>>> >>>>> 3) Is there a way to exclude the start-up code from the simulation >>>>> (from stats, etc.)? >>>>> >>>>> >>>> >>>> It's not included. >>>> >>>> Gabe >>>> _______________________________________________ >>>> m5-users mailing list >>>> [email protected] >>>> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >>>> >>>> >>>> >>> >>> >> >> _______________________________________________ >> m5-users mailing list >> [email protected] >> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >> >> > > > -- > Best Regards, > > Wang, Weixun > > Department of Computer& Information Science& Engineering > Gator College of Engineering > University of Florida > Gainesville, FL 32611 > http://www.cise.ufl.edu/~wewang > > _______________________________________________ > m5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > _______________________________________________ m5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
