i just wanted to know if there are instances where compilers assist in branch prediction like the one brought up in class. but i was thinking in terms of assisting the predictor in initialisations or warm up by means of profiling or compiler analysis. this was interesting because a significant portion of the remaining 4% above the 96% limit may be due to warm up misses. another general point of interest was a reconfigurable dynamic branch predictor which could say for example be used as gAp or pAp depending on the amount of correlation between the branches in the program.

Subramanian Rajagopalan