Virtual Machines for Dynamic Languages Ausgewählte Kapitel der Systemsoftwaretechnik, WS2021 Mark Deutel Friedrich-Alexander-University Erlangen-Nürnberg (FAU) Erlangen, Germany
[email protected] ABSTRACT threading model. It is common that once a dynamic language has A common approach to execute dynamic languages is inside a vir- grown popular enough the interpreter is extended by a just-in-time tual machine (VM), usually implemented in a low level system (JIT) compiler to more efficiently execute interpreted code. language like C. Such VMs have to support the often complex However this approach does not come without problems. The syntax and semantics of the dynamic languages they are hosting. effort required to write a complete virtual machine from scratch Additionally, some properties of dynamic language code can only is huge. It also forces the language implementer to deal with plat- be inferred at runtime making the optimization of code execution form or target architecture specific details. This makes implement- another challenging task. Therefore, it is hardly possible to provide ing a VM time consuming, error prone, and complicated to port general purpose implementations that support a large range of to other systems. Additionally, modern dynamic languages like different dynamic languages. Furthermore, many commonly used Python, Ruby or JavaScript evolve rapidly and have complex syn- dynamic languages such as Python or Javascript are still evolving tax and semantics. This not only makes it hard to represent the which means that their VMs have to adapt constantly. This implies language properly in low level code, but makes maintenance of that it is necessary to write and maintain a sufficiently elaborated existing VM implementations a complicated task as well.