Hi Max, If you find time, can you rerun your deep pipeline experiments using this patch? It seems like the CPU Model was marking the address as "unused" too early, and in the case where there was a squashing of instructions it's possible that entries do not get removed from the list.
On Sun, Mar 14, 2010 at 8:50 AM, Maximilien Breughe < [email protected]> wrote: > > Op 13-mrt-10, om 21:08 heeft Korey Sewell het volgende geschreven: > > > > > Most of the addresses are indeed added to the list more than once > > (but get removed only once)! > > Some of the addresses are added to the list once, and never removed. > > (I built a small program to identify the adresses that get added but > > not removed, if you are interested I can send it to you). > > I think just adding an assertion there as well as inspecting the > > trace will reveal the source of the problem. Hopefully wont be too > > long to find the bug, but I'm looking into the fix. > I'll try that, it's a good idea. thanks! > > > > What do you mean with a time sync? > > Maybe I should've said "time sink"? Aarrgh, I'm just communicating > > badly! > > > Thats okay =) > > Basically I was just saying its a major source of time wasted in > > that function. > > > > > Max > > _______________________________________________ > m5-users mailing list > [email protected] > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users > -- - Korey
addrList.patch
Description: Binary data
_______________________________________________ m5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
