Standard ML

Abstract: We present a new approach to proving type soundness for Hindley/Milner-style polymorphic type systems. The keys to our approach are (1) an adaptation of subject reduction theorems from combinatory logic to programming languages, and (2) the use of rewriting techniques for the specification of the language semantics. The approach easily extends from polymorphic functional languages to imperative languages that provide references, exceptions, continuations, and similar features. We illustrate the technique with a type soundness theorem for the core of Standard ML, which includes the first type soundness proof for polymorphic exceptions and continuations.

Abstract: From the Publisher: In teaching the methods of functional programming--in particular, how to program in Standard ML, a functional language recently developed at Edinburgh University, the author shows how to use such concepts as lists, trees, higher-order functions and infinite data structures.

Abstract: Machine-learning (ML) hardware and software system demand is burgeoning. Driven by ML applications, the number of different ML inference systems has exploded. Over 100 organizations are building ML inference chips, and the systems that incorporate existing models span at least three orders of magnitude in power consumption and five orders of magnitude in performance; they range from embedded devices to data-center solutions. Fueling the hardware are a dozen or more software frameworks and libraries. The myriad combinations of ML hardware and ML software make assessing ML-system performance in an architecture-neutral, representative, and reproducible manner challenging. There is a clear need for industry-wide standard ML benchmarking and evaluation criteria. MLPerf Inference answers that call. In this paper, we present our benchmarking method for evaluating ML inference systems. Driven by more than 30 organizations as well as more than 200 ML engineers and practitioners, MLPerf prescribes a set of rules and best practices to ensure comparability across systems with wildly differing architectures. The first call for submissions garnered more than 600 reproducible inference-performance measurements from 14 organizations, representing over 30 systems that showcase a wide range of capabilities. The submissions attest to the benchmark’s flexibility and adaptability.

Abstract: Programming computers is a notoriously error-prone process. It is the job of the programming language designer to make this process more reliable. One approach to this is to impose some sort of typing discipline on the programs. In doing this, the programming language designer is immediately faced with a tradeoff: if the type system is too simple, it cannot accurately express important properties of the program; if it is too expressive, then mechanically checking or inferring the types becomes impractical. This thesis describes a type system called refinement types, which is an example of a new way to make this tradeoff, as well as a potentially useful system in itself. Refinement type inference requires programs to have types in two type systems: an expressive type inference system (intersection types with subtyping) and a relatively simple type system (basic polymorphic type inference). Refinement type inference inherits some properties from each of these: as in intersection types with subtyping, we can use the type system to do abstract interpretation; as in basic polymorphic type inference, refinement type inference is decidable (preliminary experiments suggest refinement type inference may be practical as well). We have implemented refinement type inference for a subset of Standard ML to test these ideas. We have added new syntax, called rectype declarations, to allow the programmer to specify relevant domains for the abstract interpretation. A prototype implementation of refinement type inference can do some interesting case analysis for Standard ML programs; for example, if the programmer uses a rectype declaration to declare interest in whether a boolean expression is in conjunctive normal form (CNF), refinement type inference can efficiently prove that a function for converting boolean expressions to CNF does indeed always return a boolean expression in CNF. Rectype declarations and refinement type inference seem flexible and efficient enough to practically enforce many other useful program properties as well.