Hmmmm.... I think in detailed mode, the PC is used for branch prediction, > not the next PC. I glanced through the fetch_unit.cc file in o3, not sure > completely.
you need to look for an "advancePC" before the branch prediction. Check in "branch_predictor.cc" in InOrder....Check in fetch_impl.hh for the O3. > But as you suggested, I would mostly want to change it to prediction on the > same PC, not next one. > I'm trying to say that both are logically equivalent from the CPU' s point of view. > > And any suggestions about my first email in this chain regarding the likely > bug in the always taken variable updation? > You can make the change to the inst->predTaken variable if you wish. I'm hesitant to do any deep debugging into this issue, because I have a number of uncommitted InOrder patches that cleans up this code and likely fixes this issue. I'm hoping to get that in the tree at the end of this week, but keep tabs on the m5-dev list for that. -- - Korey
_______________________________________________ m5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
