Semantics-preserving inlining for metaprogramming

Inlining is used in many different ways in programming languages: some languages use it as a compiler-directive solely for optimization, some use it as a metaprogramming feature, and others lay their design in-between. This paper presents inlining through the lens of metaprogramming and we describe a powerful set of metaprogramming constructs that help programmers to unfold domain-specific decisions at compile-time. In a multi-paradigm language like Scala, the concern for generality of inlining poses several interesting questions and the challenge we tackle is to offer inlining without changing the model seen by the programmer. In this paper, we explore these questions by explaining the rationale behind the design of Scala-3’s inlining capability and how it relates to its metaprogramming architecture.

Compile-Time Type-Driven Data Representation Transformations in Object-Oriented Languages

Vlad Ureche

High-level languages allow programmers to express data structures and algorithms that abstract over the type of data they handle. This improves code reuse and makes it possible to develop general-purpose libraries. Yet, data abstractions slow down program execution, as they require low-level indirection. In this thesis we explore three compile-time approaches that leverage type systems to reduce the cost of data abstractions, thus improving program performance. In the first part of the thesis we present miniboxing, a compile-time transformation that replaces generic classes by more efficient variants, optimized to handle primitive types. These variants use the miniboxed data encoding, producing speedups of up to 20x compared to generic classes. The miniboxing transformation is the main result of this thesis and motivates the other techniques. Generalizing miniboxing, we show the Late Data Layout (LDL) mechanism, which uses the type system to guide performance-oriented program rewritings. It can be instantiated to perform a host of transformations, such as miniboxing generics, inlining value classes and unboxing primitive types. The LDL mechanism has many desirable properties, such as provable correctness in handling different data representations, reduced number of conversions and built-in support for the object-oriented paradigm. Finally, we show Data-centric Metaprogramming, a technique that allows programmers to go beyond standard compiler optimizations by defining custom representations to be used for their data. These representations are then automatically introduced by the compiler when translating programs. This technique, similar in spirit to metaprogramming, opens new directions in programmer-driven optimizations and shows encouraging results, with speedups of up to 25x. Under the hood, Data-centric Metaprogramming uses the Late Data Layout mechanism.

EPFL2015

Unification of Compile-Time and Runtime Metaprogramming in Scala

Eugene Burmako

Metaprogramming is a technique that consists in writing programs that treat other programs as data. This paradigm of software development contributes to a multitude of approaches that improve programmer productivity, including code generation, program analysis and domain-specific languages. Many programming languages and runtime systems provide support for metaprogramming. Programming platforms often distinguish the notions of compile-time and runtime metaprogramming, depending on the phase of the program lifecycle when metaprograms execute. It is common for different lifecycle phases to be hosted in different environ- ments, so it is also common for different kinds of metaprogramming to provide different capabilities to metaprogrammers. In this dissertation, we present an exploration of the idea of unifying compile-time and runtime metaprogramming in Scala. We focus on the practical aspect of the exploration; most of the described designs are available as popular software products, and some of them have become part of the standard distribution of Scala. First, guided by the motivation to consolidate disparate metaprogramming techniques available in earlier versions of Scala, we introduce scala.reflect, a unified metaprogram- ming framework that uses a language model derived from the Scala compiler to run metaprograms both at compile time and at runtime. Secondly, armed by the newfound metaprogramming powers, we describe Scala macros, a language-integrated compile-time metaprogramming facility based on scala.reflect. Thanks to the comprehensive nature of scala.reflect, macros are able to work with both syntactic and semantic information about Scala programs, enabling a wide range of previously impractical or impossible use cases. Finally, based on our experience and user feedback, we identify key strengths and weaknesses of scala.reflect and macros. We propose scala.meta, a new unified metapro- gramming framework, and inline/meta, a new macro system based on scala.meta, that take the best from their predecessors and address the most important problems.

EPFL2017

Cross-Platform Language Design

Sébastien Jean R Doeraene

Programming languages are increasingly compiled to multiple runtimes, each featuring their own rich structures such as their object model. Furthermore, they need to interact with other languages targeting said runtimes. A language targeting only one runtime can be designed to tailor its semantics to those of that runtime, for easy interoperability with other languages. However, in a language targeting multiple runtimes with differing semantics, it is difficult to cater to each of them while retaining a common behavior across runtimes. We call \emph{cross-platform language} a language that aims at being both \emph{portable} across platforms and \emph{interoperable} with each target platform. Portability is the ability for a program or a library to cross-compile for multiple platforms, and behave the same way on all of them. Interoperability is the ability to communicate with other languages on the same platform. While many cross-compiling languages focus on one of these two properties---only adding support for the other one as an afterthought---, languages that are designed from the ground up to support both are rare. In this thesis, we present the design of Scala.js, the dialect of Scala targeting the JavaScript platform, which turned Scala into a cross-platform language. On the one hand, Scala programs can be cross-compiled for the JVM and JavaScript with portable semantics. On the other hand, whereas Scala/JVM interoperates with Java, Scala.js interoperates with JavaScript, allowing to use any JavaScript library. Along the dissertation, we give insights that can be transferred to the design of other cross-platform languages, although with a bias towards those targeting the JVM and JavaScript. The first and most obvious challenge is to reconcile the static nature of Scala's object model with JavaScript's dynamic one. Besides the ability to mutate a class hierarchy at run-time in JavaScript, there are fundamental differences between the two models, in particular the difference between compile-time overloading and run-time overloading. We discuss how such semantic mismatches can live in harmony within the language. The second challenge is to obtain good performance from a language where interoperability with a dynamic and unknown part of the program is pervasive. To that end, we design and specify an intermediate representation (IR) with first-class support for dynamically typed interoperability features in addition to statically typed JVM-style operations. Despite its tight integration with the open world of JavaScript, most of the IR can be considered as a closed world where advanced whole-program optimizations can be performed. The performance of the overall system is evaluated and shown to be competitive with hand-written JavaScript, and even with Scala/JVM in some cases.

EPFL2018