Hello, I am planning to extend the SimpleCPU with a branch predictor. Because I need statistics of the penalty induced by mispredicted branches quickly this will be my way to go. Does anybody has any suggestions for me?
Since SimpleCPU works with staticInsts and not DynamicInsts, the integration with the existing modules will not be very smoothly/easy I guess. Should I rebuild the branch predictor or use the existing ones? I am willing to ignore branch predictions of branches along a speculative path (i.e. a branch instruction on the path predicted by a mispredicted branch). These would be very hard to model, since you need detailed simulation for them. Because I want to model the branch penalty of an in-order core this will probably not affect the total number of inpredicted branches, nor the state of the branch predictor. Thanks! Max _______________________________________________ m5-dev mailing list [email protected] http://m5sim.org/mailman/listinfo/m5-dev
