Hanhwi, We don't modify the linux kernel, so MARSS doesn't have any influence over its' policies.
MARSS is a combination of PTLSim, an x86 simulator, and QEMU, a virtualization environment. MARSS filled in some gaps that PTLSim had and switched to QEMU to provide user-space simulation as opposed to Xen, which required root privileges. Brendan On Mon, Feb 4, 2013 at 12:08 PM, jang hanhwi <[email protected]> wrote: > Thanks for your reply, Brendan. > > I know MARSS is a kind of full system simulator, but if the simulator can > gives exact same environment every simulation, the simulated system should > replay same instruction sequences, I think. Is it right? > > Then, does MARSS simulator have some random scheduling decision in its > design. > > Thanks, > Hanhwi > > 2013. 2. 5., 오전 1:19, Brendan Fitzgerald <[email protected]> 작성: > > Hi, > > Because MARSS is an x86 full system simulator the instructions will not > always be in the same order due to context switches, exceptions, > interrupts, etc. > What you should see is the same output after running the same benchmark > for the same number of instructions. > > Brendan > > On Mon, Feb 4, 2013 at 11:02 AM, hanhwi jang <[email protected]> wrote: > >> Hi Dear all, >> >> Every time I execute SPEC2006 on the marss simulator, I get different >> instruction traces that include kernel and user code. >> >> So, I want to know the marss is deterministic or not. >> If it is not deterministic, where is the nondeterminism from? >> >> Thanks, >> hanhwi >> >> _______________________________________________ >> http://www.marss86.org >> Marss86-Devel mailing list >> [email protected] >> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >> >> > >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
