Initial Results for Glacial Variable Analysis

. Run-time code generation that uses specific values to generate specialized code is called value-specific optimization. Variables which provide values for value-specific optimization are called candidate variables. They are modified much less frequently than they are referenced. In current systems that use run-time code generation, candidate variables are identified by programmer directives. We describe a novel technique, staging analysis, for automatically identifying candidate variables. We refer to such variables as glacial variables. Glacial variables are excellent candidate variables. Glacial Variable Analysis is an interprocedural analysis. We perform several experiments with glacial variable analysis to characterize the programs in the PERFECT benchmark suite. We explain the imprecision of our results due to procedure boundaries. We examine the structure of the programs to determine how often value-specific optimization might be applied. We will explain how staging analysis r..