Search code examples
javascripthtmlasm.js

How does assembly (asm.js) work in the browser?

Asm.js comes from a new category of JavaScript apps: C/C++ applications that’ve been compiled into JavaScript. It’s a subset of JavaScript that’s been spawned by Mozilla’s Emscripten project.

But how does it work, and why would I use it?

Solution

  • Why compile to JavaScript?

    JavaScript is the only language which works in all web browsers. Although only JavaScript will run in the browser, you can still write in other languages and still compile to JavaScript, thereby allowing it to also run in the browser. This is made possible by a technology known as emscripten.

    Emscripten is an LLVM based project that compiles C and C++ into highly performant JavaScript in the asm.js format. In short: near native speeds, using C and C++, inside of the browser. Even better, emscripten converts OpenGL, a desktop graphics API, into WebGL, which is the web variant of that API.

    How does asm.js fit into the picture? enter image description here

    Asm.js, short for Assembly JavaScript, is a subset of JavaScript. An asm.js program will behave identically whether it is run in an existing JavaScript engine or an ahead-of-time (AOT) compiling engine that recognizes and optimizes asm.js—except for speed, of course!

    In terms of speed, it’s difficult to offer a precise measurement of how it compares to native code, but preliminary benchmarks of C programs compiled to asm.js are usually within a factor of 2 slowdown over native compilation with clang, the compiler frontend for C, C++, and Obj-C programming languages. It’s important to note that this is a “best” case for single-threaded programs. More on this limitation of the JavaScript language below.

    On the backend, Clang uses LLVM, which is a library for constructing, optimizing and producing intermediate and/or binary machine code (those 0s and 1s again). LLVM can be used as a compiler framework, where you provide the “front end” (parser and lexer such as Clang) and the “back end” (code that converts LLVM representation to actual machine code)

    Further reading: Alon Zakai of Mozilla has a fantastic slide deck which goes into further detail about how this all works.

    So how cool is asm.js? Well it has its own Twitter account, @asmjs. While the asm site is a bit sparse, it does cover the W3C spec, in addition to having a thorough FAQ. Even better, Mozilla coordinated the Humble Mozilla Bundle in 2014, which allowed you to buy a bunch of gamest that took advantage of asm.js.

    Why not just turn your JavaScript code into asm.js?

    JavaScript can’t really be compiled to asm.js and offer much of a benefit, because of its dynamic nature. It’s the same problem as when trying to compile it to C or even to native code – a VM with it would be necessary to take care of those non-static aspects. You could write asm.js by hand, however.

    If one could already translate standard Javascript in a fully static manner, there would be no need for asm.js. Asm.js exists so for the promise that Javascript will get faster without any effort from the developer. It would be very difficult for the JIT to understand a dynamic language as well as a static compiler.

    To better understand this, it is important to comprehend why asm.js offers a performance benefit at all; or why statically-typed languages perform better than dynamically-typed ones. One reason is “run-time type checking takes time,” and a more thought out answer would include the enhanced feasibility of optimizing statically-typed code. A final perk of going from a statically typed language such as C is the fact that the compiler knows the type of each object when it is being compiled.

    Asm.js is a restricted subset of JS that can be easily translated to bytecode. The first step required would need to break down all the advanced features of JS to that subset for getting this advantage, which is a bit complicated. But JavaScript engines are optimized and designed to translate all those advanced features directly into bytecode – so an intermediate step like asm.js doesn’t offer much of an advantage.

    I go into greater detail and pictures in this post.