Stack Renaming of the Java Virtual

This study proposes a scheme to map the operand stack of the Java Virtual Machine to hardware registers and evaluates the performance beneets of the proposed scheme. Using the technique of register renaming while mapping the stack to registers , we are able to exploit the inherent parallelism in the instruction stream. The simulation results conducted show an improvement of about 15%-26% for the case with register renaming over the case without renaming. In addition, we show that allocating registers to local variables and renaming loads/stores of these locals can lead to further speedup.