In this dissertation I argue that modal type systems provide an elegant and practical means for controlling local resources in spatially distributed computer programs. A distributed program is one that executes in multiple physical or logical places. It usually does so because those places have local resources that can only be used in those locations. Such resources can include processing power, proximity to data, hardware, or the physical presence of a user. Programmers that write distributed applications therefore need to be able to reason about the places in which their programs will execute. This work provides an elegant and practical way to think about such programs in the form of a type system derived from modal logic. Modal logic allows for reasoning about truth from multiple simultaneous perspectives. These perspectives, called "worlds," are identified with the locations in the distributed program. This enables the programming language to be simultaneously aware of the various hosts involved in a program, their

This dissertation is about effects and type theory. Functional programming languages such as Haskell demonstrate that monads can safely encapsulated side effects. If you want to reason about effectful code, however, it is not enough to only encapsulate effects: you must explain their meaning. In this dissertation I provide pure specifications of effects in type theory. After initial investigations using Haskell, I show how such specifications can be made total in a language with dependent types. Hoare Type Theory takes an alternative approach to incorporating effects in a dependently typed language. Instead of giving functional specifications, they follow Haskell’s lead and postulate the existence of primitive functions. This dissertation shows how some of these primitives may be implemented in a programming language with dependent types. I believe that functional specifications may provide the foundations on top of which richer logics may be constructed. The results presented in this dissertation may be used to write and verify

Abstract. Locality-aware algorithms over distributed arrays can be very difficult to write. Yet such algorithms are becoming more and more important as desktop machines boast more and more processors. We show how a dependently-typed programming language can help develop such algorithms by hosting a domain-specific embedded type system that ensures every well-typed program will only ever access local data. Such static guarantees can help catch programming errors early on in the development cycle and maximise the potential speedup that multicore machines offer. At the same time, the functional specification of effects we provide facilitates the testing of and reasoning about algorithms on distributed arrays. 1

As a means of transmitting not only data but also code encapsulated within functions, higher-order channels provide an advanced form of task parallelism in parallel computations. In the presence of mutable references, however, they pose a safety problem because references may be transmitted to remote threads where they are no longer valid. This paper presents an ML-like parallel language with typesafe higher-order channels. By type safety, we mean that no value written to a channel contains references, or equivalently, that no reference escapes via a channel from the thread where it is created. The type system uses a typing judgment that is capable of deciding whether the value to which a term evaluates contains references or not. The use of such a typing judgment also makes it easy to achieve another desirable feature of channels, channel locality, that associates every channel with a unique thread for serving all values addressed to it. Our type system permits mutable references in sequential computations and also ensures that mutable references never interfere with parallel computations. Thus it provides both flexibility in sequential programming and ease of implementing parallel computations.

K42 is an open-source scalable research operating system well suited to support systems research. The primary goals of K42’s design that support such research include flexibility to allow a multitude of policies and implementations to be supported simultaneously, extensibility to allow new policies and implementations to be readily added, and scalability to enable good performance for both small and large applications on both small and large multiprocessor systems. The goals are accomplished via key features including an objectoriented structure that allows specialized resource management implementations and policies on a per-resource, perapplication basis, implementation in user-level servers of much of the system functionality, and a sophisticated set of underlying services that provides a programming model for developing system software in a scalable and modular fashion. These characteristics make K42 an attractive framework for prototyping new operating system ideas. In addition, K42 has a sophisticated performance monitoring infrastructure allowing a thorough understanding of new ideas to be gained. The above framework combined with a consistent emphasis on scalability makes K42 well suited for high-end computing initiatives. In this paper, we describe the structure of K42 which contributes to the advantageous prototyping environment, and demonstrate how to utilize it by describing ongoing research efforts.

Domain-specific languages raise the level of abstraction in software development. While it is evident that programmers can more easily reason about very high-level programs, the same holds for compilers only if the compiler has an accurate model of the application domain and the underlying target platform. Since mapping high-level, general-purpose languages to modern, heterogeneous hardware is becoming increasingly difficult, DSLs are an attractive way to capitalize on improved hardware performance, precisely by making the compiler reason on a higher level. Implementing efficient DSL compilers is a daunting task however, and support for building performance-oriented DSLs is urgently needed. To this end, we present the Delite Framework, an extensible toolkit that drastically simplifies building embedded DSLs and compiling DSL programs for execution on heterogeneous hardware. We discuss several building blocks in some detail and present experimental results for the OptiML machine-learning DSL implemented on top of Delite. 1

This paper outlines a general picture of our ongoing work under EU Mobius and Sensoria projects on a type-based compilation and execution framework for a class of multicore CPUs. Our focus is to harness the power of concurrency and asynchrony in one of the major forms of multicore CPUs based on distributed, non-coherent memory, through the use of type-directed compilation. The key idea is to regard explicit asynchronous data transfer among local caches as typed communication among processes. By typing imperative processes with a variant of session types, we obtain both type-safe and efficient compilation into processes distributed over multiple cores with local memories.

One of the major productivity issues in parallel programming arises from the use of lock/unlock operations or atomic/critical sections to enforce mutual exclusion. Not only are these constructs complicated to understand and debug, but they are also often an impediment to achieving scalable parallelism. In this paper, we propose to give the programmer the convenience of critical sections combined with the scalability of fine-grained locks, by solving two technical problems related to optimized assignment of locks to critical sections. The first problem, Minimum Lock Assignment (MLA), addresses the problem of finding the minimum number of locks needed to enforce mutual exclusion among interfering critical sections without any loss of concurrency. The second problem, K-Lock Allocation (K-LA) addresses the problem of allocating a fixed number (K) of locks to critical sections so as to minimize the serialization overhead. i

The Standard Template Adaptive Parallel Library (STAPL) is a C++ parallel programming library that provides a collection of distributed data structures (pContainers) and parallel algorithms (pAlgorithms) and a generic methodology for extending them to provide customized functionality. STAPL algorithms are written in terms of views, which provide a generic access interface to pContainer data by abstracting common data structure concepts. Briefly, views allow the same pContainer to present multiple interfaces, e.g., enabling the same pMatrix to be ‘viewed ’ (or used) as a row-major or column-major matrix, or even as a vector. In this paper, we describe the stapl View concept and its properties. stapl Views generalize the iterator concept — a View corresponds to a collection of elements and provides an ADT for the data it represents. stapl Views enable parallelism by providing random access to the elements, and support for managing the tradeoff between the expressivity of the views and the performance of the parallel execution. Views trade additional parallelism enabling information for reduced genericity. We illustrate the expressivity enabled by Views for several examples and examine the performance overhead incurred when using Views. 1

Abstract not found