Skip to content


Version History
Date Description
Oct 16, 2021 Move compilers section from the summary file

The general ones:


Today (Oct 14, 201) I was reading Hacker News and came across this post A viable solution for Python concurrency. It was about removing the Global Interpreter Lock (GIL) in the cpython compilers. Quite interesting. The technique is to use Biased Atomic Reference Accounting. Basically, it uses non-atomic operation if it is single-thread so to avoid the cost of atomic instructions. But for multiple thread case, it will normal atomic instructions (which will be much better the original GIL implementation).

So I decide to take another look at the cpython source code, which I have cloned (repo)quite a while ago when I had a broken leg. Once I decided to read the code, I google some cpython internals and these links pop up quite nicely. There are A LOT good contents out there, I probably don’t have time reading that now.

I briefly read the code, a lot typedefs for sure. The PyStatus structure is interesting. And the way they organize the repo is also interesting. For a common python library, say csv, there will a python library file under Lib/, then optionally a C accelerated version in Modules/_csv.c. Essentially the whole thing is built like a Exokernel, the base is written in C for performance and portability among OSes. Then a more rich python wrapper on top of that, which will be the default built-in python libraries we use day-to-day.

For those common built-in functions, they are organized here

  1. Your Guide to the CPython Source Code
  2. Exploring CPython’s Internals
  3. Design of CPython’s Compiler
  4. Yet another guided tour of CPython


  • OpenJDK

    • JRE = JVM + Runtime Classes => JVM is the one parsing the bytecode, along with some extra classes/libraries, they form JRE.
    • JDK = JRE + Development Tools => JDK as in Development Kit therefore consists of some tools in addition to JRE.
    • JDK is a monster collection of resources in one place. The JVM here is called HotSpot, a reference JVM implementation written in C++, Since JDK also has so many runtime support, it has a lot Java code.
    • Personally I haven’t written Java since 2013 or so. Although I’m not using it anytime soon, I’m curious how it performs nowadays.
    • The repo is VERY WELL organized. see src/
  • HotSpot JVM

    • This one is included in the OpenJDK Repo, written in C++.
    • e.g., the GC code is under src/hotspot/share/gc.
  • Eclipse Openj9 JVM

    • A JVM for OpenJDK that’s optimized for small footprint, fast start-up, and high throughput
  • ASM

    • ASM is an all purpose Java bytecode manipulation and analysis framework.
    • It can be used to modify existing classes or to dynamically generate classes, directly in binary form.

All these OpenJDK components follow the Java Language Spec and JVM Spec.

An important note: Java is NOT the only language that can run on a JVM. A lot of other languages are using JVM as well! Such as Kotlin, Scala, Clojure etc. I think the reason is that JVM is production-ready and proven to be stable across platforms. If a new language compiles into JVM bytecode, then this new language can instantly run all architectures. Without it, the new language’s compiler needs to emit different ISA’s assembly, which is difficult and quite an effort.

JVM-based Languages

There is a List of JVM languages: This list of JVM Languages comprises notable computer programming languages that are used to produce computer software that runs on the Java virtual machine (JVM). Some of these languages are interpreted by a Java program, and some are compiled to Java bytecode and JIT-compiled during execution as regular Java programs to improve performance. The most popurlar ones are: 1) Java, 2) Groovy, 3) Scala, 4) Clojure, 5) Kotlin.

So follow up on the Java section, I want to spend some time on JVM-based languages, the rationale and benefits behind it, and how should one create a new language on JVM. Great explanation here.

JVM is a virtual MACHINE, with its own machine model and ISA. Hence it has assembly instructions and assmeblers (e.g., Jasmin) compiling annoted assembly into Java class file / bytecode (or binary for the JVM, in some sense). Some people seem to use Java ASM tool to generate bytecode as well.

Several blogs I found via google that try to build a new language on top of JVM. There must be more. 1), 2)

Last update: October 17, 2021


Back to top