Float16Array in JavaScript

Float16Array stores 16-bit floating point numbers in half the memory of a Float32Array. How it works, when to use it and which browsers support it.

Float16Array is a new type of array in JavaScript that stores 16-bit (2-byte) floating point numbers. It joins the existing family of TypedArrays like Int8Array, Uint8Array, Float32Array, and Float64Array

If you’ve never used TypedArrays before, think of them as specialized arrays that can only contain specific types of numbers.

Unlike regular JavaScript arrays that can mix strings, objects, and numbers, TypedArrays are fixed-size, memory-efficient arrays dedicated to numeric data.

Here’s how you’d create and use a Float16Array

JavaScript
// Create a Float16Array with 4 elements
const myArray = new Float16Array(4);
myArray[0] = 1.5;
myArray[1] = 42.0;
myArray[2] = 0.333;
myArray[3] = -27.5;
console.log(myArray); // Float16Array [1.5, 42, 0.3330078125, -27.5]

Notice something interesting with the third value? 0.333 became 0.3330078125. This happens because 16-bit floating point numbers have limited precision compared to JavaScript’s standard 64-bit numbers.

You might wonder: “We already have Float32Array and Float64Array, so why do we need a 16-bit version?”

The answer comes down to three things: memory, performance, and compatibility.

TypedArrayElement TypeBytes Per ElementDescriptionValue RangePrecision
Float16Array16-bit float216-bit IEEE floating point±65,504~3 decimal digits
Float32Array32-bit float432-bit IEEE floating point±3.4×10^38~7 decimal digits
Float64Array64-bit float864-bit IEEE floating point±1.8×10^308~15 decimal digits

A Float16Array uses half the memory of a Float32Array. If you’re working with large datasets, this can make a significant difference.

For applications processing millions of values, this memory saving can be crucial. In certain contexts, especially with GPU-accelerated computing, 16-bit floats can be processed more efficiently than 32-bit or 64-bit numbers. Newer GPUs often have specialized hardware for handling 16-bit floating point math, making operations potentially faster.

References


Trevor I. Lasn

Trevor I. Lasn

Building whiteheap.com and 0xinsider.com.



Related Articles

Check out these related articles that might be useful for you. They cover similar topics and provide additional insights.

JavaScript
3 min read

Copy and paste text, images, and files using the new navigator.clipboard API

Dec 7, 2024
Read article
JavaScript
4 min read

How the Suspense component manages async dependencies and improves loading states in Vue apps

Aug 23, 2024
Read article
JavaScript
5 min read

The Stage 2 pipeline operator passes a value through a chain of calls left to right, with % marking where it goes, instead of nesting h(g(f(x))).

Oct 29, 2024
Read article
JavaScript
6 min read

Closures are essential for creating functions that maintain state, without relying on global variables.

Sep 6, 2024
Read article
JavaScript
4 min read

Stop mixing try/catch with Promise chains - JavaScript's new Promise.try handles return values, Promises, and errors uniformly

Nov 10, 2024
Read article
JavaScript
8 min read

How to handle unique values and key-value pairs properly without type coercion and performance issues

Nov 17, 2024
Read article
JavaScript
6 min read

Peaks.js is a client-side JavaScript component to display and interact with audio waveforms in the browser

Oct 22, 2019
Read article
JavaScript
5 min read

scheduler.postTask() runs JavaScript at user-blocking, user-visible or background priority, giving finer control than setTimeout(0) over when work runs.

Nov 26, 2024
Read article
JavaScript
6 min read

The creator of JavaScript and Node.js are challenging Oracle's control over the JavaScript name

Nov 23, 2024
Read article

This article was originally published on https://www.trevorlasn.com/blog/float16array-javascript. It was written by a human and polished using grammar tools for clarity.