Summary of Benjamin Titzer (CMU): Managing a Managed Language in a Managed Language (21/11/22)

00:00:00 - 00:55:00

In this video, Benjamin Titzer of Carnegie Mellon University discusses the design of his new system language, Virgil, and how it is designed to be garbage-collected, support threading and memory management, and be low-level and unsafe. He also discusses the challenges of designing a managed language, and how Virgil can help.

• 00:00:00 Benjamin Titzer describes his work on managed languages and systems languages, both in academia and in industry. He goes on to say that, if you want to learn about these topics, you should always be implementing things.
• 00:05:00 Benjamin Titzer is working on a research engine for webassembly, which is an intermediate language for compiling to machine code. He also wrote a paper about systems languages in 2013.
• 00:10:00 This video discusses Benjamin Titzer's work on a system language called Virgil, which is designed to be safe and efficient. The main features of Virgil are its static typing and memory safety, its compilation model, and its ability to generate source stack traces when a program crashes.
• 00:15:00 Benjamin Titzer from Carnegie Mellon University discusses the importance of a systems language supporting interaction with low-level software and hardware, as well as the benefits of self-hosting such a language. He argues that a good systems language should be able to do all of the tricks that a language implementation requires, and that a systems language should be expressive and low level enough to be able to do all the tricks that a language implementation requires.
• 00:20:00 In this video, Benjamin Titzer from Carnegie Mellon University discusses the design goals for his new system language, Virgil. He explains that Virgil is designed to be garbage-collected, support threading and memory management, and be low-level and unsafe. He also states that Virgil is a pain to debug, but it is ultimately a very powerful and efficient system language.
• 00:25:00 In this video, Benjamin Titzer from Carnegie Mellon University discusses the challenges of designing a managed language, and how Virgil can help. He also discusses the design of the wizard research engine, which is designed to be flexible and easy to use.
• 00:30:00 Benjamin Titzer describes the managed language Virgil, which has a tagged value system, unifying references and integers, and a vector type. The interpreter for Virgil is slow and inefficient, and Titzer plans to rewrite it in assembly to gain more freedom and speed.
• 00:35:00 Benjamin Titzer (CMU) discusses the managed language design choices made in The Interpreter, and how they enable fast code execution. He also discusses the systems language requirement that interactions with dynamically generated code be possible. Finally, he demonstrates how these design choices allow for code generation and dynamic access to objects in the runtime environment.
• 00:40:00 Benjamin Titzer from CMU describes their research on implementing a virtual machine for running managed languages on a virtual machine. They discuss how they managed to achieve a research result using virtual as a systems language, and some of the challenges they still face.
• 00:45:00 Benjamin Titzer discusses the requirements for a systems language, including the need to be able to generate machine code efficiently, build large systems succinctly, and interface with low-level code. He also touches on garbage collection and how flexible fuel systems are for implementing different collectors.
• 00:50:00 Benjamin Titzer discusses how to manage a managed language in a managed language, how to generate code for a managed language barrier, and how to avoid problems when targeting a managed language in a VM.
• 00:55:00 Benjamin Titzer from Carnegie Mellon University discusses the design and implementation of a managed language, which allows for assembly code to be directly written. He states that while it is not a requirement, the managed language implementation makes it easier to generate machine code in memory, which is essentially the same as ajit.