I'm seeing branch _misprediction_ rates of between 40% and 98% for a set of spec benchmarks. The simulations have been performed over billions of instructions (1B and 10B) with the following configuration: choiceCtrBits=2 choicePredictorSize=8192 globalCtrBits=2 globalHistoryBits=13 globalPredictorSize=8192 localCtrBits=2 localHistoryBits=11 localHistoryTableSize=2048 localPredictorSize=2048 predType=tournament
These prediction rates do not mesh with the rates I was seeing on a much older version of m5 nor on the O3 cpu. Any ideas on what may be causing this issue? Anyone else see very high misprediction rates with the inorder core? Thanks! Henry
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
