Hi everyone,
Steve Reinhardt schrieb:
Here are some other things that I came across:
1. When enabling the caches, m5 takes quite a lot of time to start the
simulation. I have not done any profiling yet, though.
Yes, we've noticed this too... we tracked it down (though I forgot
what the exact cause was) and hopefully we'll get it fixed for the
final 2.0 release.
Well, I have also ran a profiler and it seems like sorting the
statistics takes up most of the time spent before simulation starts. The
function Stats::StatData::less(), which compares two StatDatas, is the
main time consumer. Especially the tokenize() function on each
StatsData.name takes time.
It looks like less() does unneccessary work: it splits up the
hierarchical name at the "." (using tokenize()) and then compares the
token lists component-wise.
IMHO, a simple string compare of both StatsData.name should yeild the
same result w/o taking the strings apart. Thus, one could simplify
Stats::StatData::less():
bool
StatData::less(StatData *stat1, StatData *stat2)
{
// simple string compare (instead of tokenizing the names and
comparing the tokens)
return stat1->name < stat2->name;
}
In my measurements, using this implementation results in the same orders
m5stats.txt and runs MUCH faster.
What do you think?
Regards,
Jonas.
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
