Ali, I tried that. I get a python error: "AttributeError: 'module' object has no attribute 'stats'". I'm not entirely sure why and haven't been able to figure it out. I'm sure there's a fairly straightforward reason, but I haven't been able to figure out m5's python hierarchy enough to sort it out.
I *did* find a workaround, but it's a bit of a hack. Since SIGUSR1 and SIGUSR2 are mapped to signal handlers that dump stats and dump & reset stats, respectively, I can have python send itself a signal which will cause the stats to be dumped on the next cycle: os.kill(os.getpid(), signal.SIGUSR1) (I assume the signal handlers are there for FS mode.) This does seem to work, but as I said, it's not the prettiest thing in the world. To use this, I have to import os and signal at the top of the file. Thanks, -Vilas On 10/31/07, Ali Saidi <[EMAIL PROTECTED]> wrote: > > Hi Vilas, > > You should be able to call m5.internal.stats.reset() from within > python after you do the switch. This will reset *all* stats. If you > just want to reset some the easiest way is probably to wrap the > resetStats() method is swig and then you could call something like > test_sys.cpu.dcache.resetStats() in python after the switchover happens. > > Ali > > On Oct 31, 2007, at 12:52 PM, Vilas Sridharan wrote: > > > Hello all -- > > > > I looked at the list archives and webpage and couldn't figure this > > out. Is there any way to dump & reset the stats midway through > > simulation *without* specifying a cycle number up front? > > > > I'd like to look at some cache statistics. I'm using the > > TimingSimpleCPU to warm the caches, but then the DetailedCPU to > > simulate. I'd like to reset the statistics before starting > > simulation (so I only see the memory statistics for the simulation > > phase) but I don't know up front how long that is - I am warming up > > for a fixed number of instructions, not a fixed number of cycles. > > Basically, I'd like to dump & reset (or even just reset) the > > *memory* stats when I do the CPU switch, but I can't figure out how. > > > > Thanks, > > > > -Vilas > > _______________________________________________ > > m5-users mailing list > > m5-users@m5sim.org > > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > > _______________________________________________ > m5-users mailing list > m5-users@m5sim.org > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users >
_______________________________________________ m5-users mailing list m5-users@m5sim.org http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
