As heterogeneous parallel systems become dominant, application developers are being forced to turn to an incompatible mix of low level programming models (e.g. OpenMP, MPI, CUDA, OpenCL). However, these models do little to shield developers from the difficult problems of parallelization, data decomposition and machine-specific details. Most programmers are having a difficult time using these programming models effectively. To provide a programming model that addresses the productivity and performance requirements for the average programmer, we explore a domainspecific approach to heterogeneous parallel programming. We propose language virtualization as a new principle that enables the construction of highly efficient parallel domain specific languages that are embedded in a common host language. We define criteria for language virtualization and present techniques to achieve them. We present two concrete case studies of domain-specific languages that are implemented using our virtualization approach.

Self-representation – the ability to represent programs in their own language – has important applications in reflective languages and many other domains of programming language design. Although approaches to designing typed program representations for sublanguages of some base language have become quite popular recently, the question whether a fully metacircular typed selfrepresentation is possible is still open. This paper makes a big step towards this aim by defining the F ∗ ω calculus, an extension of the higher-order polymorphic lambda calculus Fω that allows typed self-representations. While the usability of these representations for metaprogramming is still limited, we believe that our approach makes a significant step towards a new generation of reflective languages that are both safe and efficient.

As the size of datasets continues to grow, machine learning applications are becoming increasingly limited by the amount of available computational power. Taking advantage of modern hardware requires using multiple parallel programming models targeted at different devices (e.g. CPUs and GPUs). However, programming these devices to run efficiently and correctly is difficult, error-prone, and results in software that is harder to read and maintain. We present OptiML, a domain-specific language (DSL) for machine learning. OptiML is an implicitly parallel, expressive and high performance alternative to MATLAB and C++. OptiML performs domain-specific analyses and optimizations and automatically generates CUDA code for GPUs. We show that OptiML outperforms explicitly parallelized MATLAB code in nearly all cases. 1.

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

Existing approaches to extend a programming language with syntactic sugar often leave a bitter taste, because they cannot be used with the same ease as the main extension mechanism of the programming language—libraries. Sugar libraries are a novel approach for syntactically extending a programming language within the language. A sugar library is like an ordinary library, but can, in addition, export syntactic sugar for using the library. Sugar libraries maintain the composability and scoping properties of ordinary libraries and are hence particularly well-suited for embedding a multitude of domain-specific languages into a host language. They also inherit self-applicability from libraries, which means that sugar libraries can provide syntactic extensions for the definition of other sugar libraries. To demonstrate the expressiveness and applicability of sugar libraries, we have developed SugarJ, a language on top of Java, SDF and Stratego, which supports syntactic extensibility. SugarJ employs a novel incremental parsing technique, which allows changing the syntax within a source file. We demonstrate SugarJ by five language extensions, including embeddings of XML and closures in Java, all available as sugar libraries. We illustrate the utility of selfapplicability by embedding XML Schema, a metalanguage to define XML languages.

Abstract. Developing rich web applications requires mastering different environments on the client and server sides. While there is considerable choice on the server-side, the client-side is tied to JavaScript, which poses substantial software engineering challenges, such as moving or sharing pieces of code between the environments. We embed JavaScript as a DSL in Scala, using Lightweight Modular Staging. DSL code can be compiled to JavaScript or executed as part of the server application. We use features of the host language to make client-side programming safer and more convenient. We use gradual typing to interface typed DSL programs with existing JavaScript APIs. We exploit a selective CPS transform already available in the host language to provide a compelling abstraction over asynchronous callback-driven programming in our DSL.