Programs often call the same function with the same arguments, yielding the same results. We call this phenomenon, “function reuse”. Previously, we have shown such a behavior for some of the SPEC2000 integer benchmarks using HP ATOM instrumentation tools. However, this required extensive analysis by hand, and assumptions regarding side-effects caused by functions. In this paper, we modified a well-known architecture simulator, SimpleScalar, to analyze multiple benchmarks to investigate the function reuse behavior. Key words. Function reuse, Speculative Execution, Value Prediction, Instruction Reuse, Basic Block Reuse, Instruction Level Parallelism, SimpleScalar
Peng Chen, Krishna M. Kavi, Robert Akl