As you can imagine, the second way is absurdly more performant. Another is to use a SIMD instruction that will compare a 64-byte chunk of text to 64 repeated ' characters and report which bytes are equal. One way is to increment an index until buffer = '''. Basically, instead of processing it character by character, it does it in chunks of up to 512 bits (64 characters) per CPU instruction! Imagine trying to find the end of a JSON string. It can process gigabytes of JSON per second thanks to clever use of SIMD instructions. If you’ve never heard of it, simdjson is (to my knowledge) the world’s fastest JSON parser. The first thing I tried was to use simdjson. Step 1: Parse JSON using SIMD instructions Improving V8 (Chrome, Node) or JSC (Safari, Bun) would have greater impact, but Hermes is relevant to my work. The JavaScript engine we’ll target is Hermes, used primarily by React Native.

By making it twice as fast, it could have real impact on users, as opposed to optimizing some obscure microbenchmark. There are real world use cases where you’d want to parse large JSONs, such as initial login to an offline-first app, or deserializing persisted state at launch. We’ll make a common operation, JSON.parse(), faster.

Speed is a feature, and when working in an interpreted language, squeezing every last bit of performance can be the difference between a great product and unusable garbage.Īnyway, how cool would it be to make JavaScript itself go faster? I’m not a C++ programmer, but that didn’t stop me before, so I thought I’d give it a try anyway! The objective Part of my job is to make JavaScript things go fast.

Json helper parse