No abstract

Systems programs rely on fine-grain control of data representation and use of state to achieve performance, conformance to hardware specification, and temporal predictability. The robustness and checkability of these programs could be greatly improved if modern type systems and programming language ideas, such as polymorphism and type inference, could be applied to these programs.BitC is a higher-order programming language in the tradition of ML and Haskell, extended to incorporate both state and the expression of unboxed and low-level datatypes. State and unboxed value types interact in subtle ways with polymorphic type-inference. Unless handled with care in the language design, interactions of these features can lead to unsoundness or results that are counter-intuitive to the programmer. Because instances of value types may have mutable components, a decision must be made concerning their compatibility at copy boundaries: should structurally equivalent types that differ only in their mutability be considered compatible. The choice impacts both the amount of polymorphism that the language can preserve and the burden of type annotation imposed on the programmer.This paper presents some of these challenges and our design for how to address these issues.

Dynamic translation is a general purpose tool used for instrumenting programs at run time. Many current translators perform substantial rewriting during translation in an attempt to reduce execution time. When dynamic translation is used as a ubiquitous policy enforcement mechanism, the majority of program executions have no dominating inner loop that can be used to amortize the cost of translation. Even under more favorable usage assumptions, our measurements show that such optimizations offer no significant benefit in most cases. A simpler, more maintainable, adaptable, and smaller translator may be preferable to more complicated designs. In this paper, we present HDTrans, a light-weight IA-32 to IA-32 binary translation system that uses some simple and effective translation techniques in combination with established trace linearization and code caching optimizations. We also present an evaluation of translation overhead under non-ideal conditions, showing that conventional benchmarks do not provide a good prediction of translation overhead when used pervasively. A further contribution of this paper is an analysis of the effectiveness of post-compile static pre-translation techniques for overhead reduction. Our results indicate that static pre-translation is effective only when expensive instrumentation or optimization is performed, and that efficient reload of pre-translated code incurs a substantial execution-time penalty.

Abstract. This paper introduces a new type system designed for safe systems programming. The type system features a new mutability model that combines unboxed types with a consistent typing of mutability. The type system is provably sound, supports polymorphism, and eliminates the need for alias analysis to determine the immutability of a location. A sound and complete type inference algorithm for this system is presented.

Dynamic translation is a general purpose tool used for instrumenting programs at run time. Performance of translated execution relies on balancing the cost of translation against the benefits of any optimizations achieved, and many current translators perform substantial rewriting during translation in an attempt to reduce execution time. Our results show that these optimizations offer no significant benefit even when the translated program has a small, hot working set. When used in a broader range of applications, such as ubiquitous policy enforcement or penetration detection, translator performance cannot rely on the presence of a hot working set to amortize the cost of translation. A simpler, more maintainable, adaptable, and smaller translator appears preferable to more complicated designs in most cases.HDTrans is a light-weight dynamic instrumentation system for the IA-32 architecture that uses some simple and effective translation techniques in combination with established trace linearization and code caching optimizations. We present an evaluation of translation overhead under both benchmark and less idealized conditions, showing that conventional benchmarks do not provide a good prediction of translation overhead when used pervasively.A further contribution of this paper is an analysis of the effectiveness of post-link static pre-translation techniques for overhead reduction. Our results indicate that static pre-translation is effective only when expensive instrumentation or optimization is performed.