Back to search

University of Waterloo

Specializing Scala with Truffle

Abstract

dc:description.abstract

Scala is a generic object-oriented programming language with higher-order abstractions. Programming abstractions in Scala exemplify reusability and extensibility in the context of type safety. In particular, generic programming allows user-defined data structures to behave identically irrespective of the types of their values while remaining free of type errors. The implementation of reusability in Scala comes at a cost; the standard implementation of Scala compiles to Java bytecode, where type erasure significantly reduces Scala program type information to create compatible Java bytecode. Consequently, autoboxing, operations needed when using primitive values in a generic context, are introduced into the final program. The current state-of-the-art techniques for eliminating boxing and achieving optimal data representations at runtime, known as specialization, rely on static program analysis. Such techniques must mitigate the problem of code duplication; static optimizations cannot use runtime information to best select which data structures to specialize. This thesis proposes a new approach to the specialization of Scala programs. The approach integrates type information from a high-level source-like input language with the mechanisms of just-in-time compilation. We propose an ad-hoc specialization mechanism using a whole program approach; specializations of data structures are created based on concrete type arguments. In our approach, specialized objects are compatible with non-specialized code. The thesis uses Truffle, a framework that simplifies the implementation of just-in-time compilers, to implement an experimental research prototype. We demonstrate that our approach is viable and produces improvements in throughput for simplified implementations of real-world Scala programs. While these programs are simple, it is still challenging for state-of-the-art approaches to specialize optimally. We show that our approach can improve performance by an order of magnitude in the context of polymorphic data structures and methods that use bulk storage. We compare the results of our approach to our interpreter without specialization and compiled Scala on GraalVM, a state-of-art Java Virtual Machine.

Degree

thesis:*
Grantor dc:publisher
University of Waterloo
Year dc:date.issued
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • You, James
Advisor dc:contributor.advisor
  • Lhotak, Ondrej

Subjects

dc:subject × 8

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Handle dc:identifier.uri
http://hdl.handle.net/10012/18928

Chain of custody

source
Harvested from
University of Waterloo
Base URL
dspacemainprd01.lib.uwaterloo.ca/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

You, James. Specializing Scala with Truffle. University of Waterloo, 2022. http://hdl.handle.net/10012/18928