Abstract

Speculative execution on superscalar processors demands substantially better branch prediction than what has been previously available. In this paper we present code replication techniques that improve the accurracy of semi-static branch prediction to a level comparable to dynamic branch prediction schemes. Our technique uses profiling to collect information about the correlation between different branches and about the correlation between the subsequent outcomes of a single branch. Using this information and code replication the outcome of branches is represented in the program state. Our experiments have shown that the misprediction rate can almost be halved while the code size is increased by one third. 1 Introduction Branch prediction forecasts the direction a conditional branch will take. It reduces the branch penalty in a processor and is a basis for the application of compiler optimization techniques. In this paper we are mainly interested in the latter use, since we will appl...

Citations

398	A study of branch prediction strategies - Smith - 1981
264	Y.N.Patt. Alternative Implementations of Two-level Adaptive Branch Prediction - Yeh - 1992
249	HANSEN: Profile Guided Code Positioning - PETTIS, C - 1990
234	A comparison of dynamic branch predictors that use two levels of branch history - Yeh, Patt - 1993
175	Improving the accuracy of dynamic branch prediction using branch correlation - Pan, So, et al. - 1992
145	Predicting conditional branch directions from previous runs of a program - Fisher, Freudenberger - 1992
144	Branch prediction for free - Ball, Larus - 1993
140	Predicting program behavior using real or estimated pro les - Wall - 1991
121	Reducing the cost of branches - McFarling, Hennessy - 1986
88	Ecient Program Tracing - Larus - 1993
31	Avoiding unconditional jumps by code replication - Mueller, Whalley - 1992