A framework for optimistic program optimization

摘要: The problem of program optimization is a non-trivial one. Compilers do fair job, but can't always deliver the best performance. expressibility general-purpose languages limited, not allowing programmers to describe expected run time behavior, for example, and some programs are thus more amenable than others, depending on what compiler expects see. We present generic framework that allows addressing this in two ways: through specifying verifiable source annotations guide analyses, optimistically using assumptions analysis results subset seen so far. Two novel applications presented, one each above approaches: dynamic optimistic interprocedural type algorithm, mechanism immutability assertions. Both result measurable speedups, demonstrating feasibility approach.